National Emission Standards for Hazardous Air
Pollutants from Petroleum Refineries
Background Information for Final Amendments
Summary of Public Comments and Responses
-------
This page intentionally left blank.
-------
National Emission Standards for Hazardous Air Pollutants
for Petroleum Refineries
Background Information for Final Rule
Summary of Public Comments and Responses
Contract No. EP-D-11-084
Work Assignment No. 3-08
U.S. Environmental Protection Agency
Office of Air Quality Planning and Standards
Sector Policies and Programs Division
Research Triangle Park, North Carolina 27711
September 2015
-------
Disclaimer
This report has been reviewed by the Sector Policies and Programs Division of the Office of Air
Quality Planning and Standards, EPA, and approved for publication. Mention of trade names or
commercial products is not intended to constitute endorsement or recommendation for use.
-------
TABLE OF CONTENTS
1.0 Introduction/Comment Period Extension 1
2.0 Risk Assessment 3
2.1 Emissions Data 3
2.1.1 Actual emissions and release characteristics 3
2.1.2 Allowable emissions 12
2.1.3 Revisions submitted for the emissions data set 17
2.2 EPA overestimated human health risks 18
2.2.1 Modeling assumptions overestimate risk 18
2.2.2 Risk assessment is too broad in scope 21
2.3 EPA underestimated human health risks 23
2.3.1 Modeling methodology understates risks 23
2.3.2 Cancer unit risk estimates, non-cancer reference doses and acute benchmarks used by
EPA understate risks 25
2.3.3 Additional pollutants and health effects should be considered 39
2.3.4 Additional emissions and source categories should be considered 41
2.3.5 Multipathway risk assessment 47
2.3.6 Recommendations for strengthening risk assessment 52
2.4 Demographic Analysis / Environmental justice 54
2.5 Ecological risk assessment 62
2.6 Rule changes are not needed because risks are acceptable 65
2.7 Stronger standards are needed because risks are unacceptable 67
3.0 Refinery Flares Control Device Provisions 74
3.1 Halogenated vent stream 77
3.1.1 Prohibition of halogenated vent streams 77
3.1.2 Definition of halogenated vent stream 78
3.2 Pilot flame requirement 78
3.2.1 Automated relight systems 80
3.2.2 Pilot monitoring requirements 81
3.3 Visible emissions and velocity requirements 82
3.3.1 Need for visible emission limit 82
3.3.2 Time allowed for visible emissions 84
3.3.3 Monitoring requirements for visible emissions 85
3.3.4 Need for flare tip velocity requirement 88
3.3.5 Velocity limit and calculation method 89
3.4 Flare SSM Issues 92
3.4.1 VE and velocity exclusions during SSM events 92
3.4.2 Other flare SSM issues 93
3.5 Combustion zone operating limits (general, non-H2-olefin) 94
3.5.1 Applicability (steam-assist only or all flares) 97
3.5.2 Selection of parameters & limits 98
l
-------
3.5.3 Allowance to use any parameter at any time 100
3.6 Dilution parameters for air-assisted flares (general, non-H2-olefin) 101
3.7H2-olefin interaction 103
3.7.1 Appropriateness of interaction criteria concept 103
3.7.2 Selection of limits when interaction is present 104
3.7.3 Applicability (steam-assist only or all flares) 106
3.8 Flow rate monitoring requirements 106
3.8.1 Flow monitors 107
3.8.2 Engineering calculations (temp/press monitors) 108
3.9 Flare vent gas composition monitoring requirements 109
3.9.1 Continuous monitoring systems Ill
3.9.2 Grab sampling systems 114
3.10 Averaging time and calculation of flare operating limits 115
3.11 QA/QC requirements for flare CPMS 118
3.12 Alternative means of emissions limitations (site-specific) 123
3.12.1 Site-specific test plan 125
3.12.2 Request/approval of site-specific operating limits 126
3.13 Definitions for flare control devices 126
3.13.1 Air assist terms 126
3.13.2 Steam assist terms 128
3.13.3 Flare gas terms 129
3.14 Cost and emission impacts for flares 130
3.15 Other/Ancillary Flare Comments 133
4.0 Miscellaneous Process Vent Provisions 136
4.1 Revisions to the definition of miscellaneous process vents 138
4.1.1 Revision of fuel gas system exclusion 139
4.1.2 Removal of in situ sampling exclusion 143
4.2 Revision of the definition of periodically discharged 144
4.3 Bypass line provisions 145
4.4 Monitoring requirements 150
4.5 Recordkeeping and reporting requirements 151
5.0 Storage Vessel Provisions 152
5.1 Technology review results 152
5.1.1 Other controls (degassing, geodesic domes, roof landings) 155
5.1.2 Revision of definition of Group 1 storage vessels 157
5.1.3 Revision of definition of reference control technology for storage vessels 158
5.2 Compliance schedule provisions 159
5.3 Impact estimates for storage vessel requirements 161
5.4 Monitoring and inspection requirements 161
5.5 Recordkeeping and reporting requirements 162
5.6 Other 163
6.0 Equipment Leak Standards 165
6.1 Technology review results 165
6.2 Optical gas imaging provisions 170
6.3 Clarification of seal for open-ended lines 172
6.4 Requirements for pressure relief devices 174
11
-------
6.4.1 Pressure release (remonitor) requirement 179
6.4.2 No atmospheric discharges 179
6.4.3 Monitoring system requirements 186
6.5 Compliance time for equipment leak/PRD requirements 187
6.6 Impact estimates for equipment leak/PRD requirements 187
6.6.1 Emissions Changes 187
6.6.2 Control costs 189
6.7 Recordkeeping and reporting requirements 190
7.0 Delayed Coking Units 193
7.1 Technology review/need for MACT standard 193
7.2 MACT floor determination 194
7.2.1 Existing source alternatives 196
7.2.2 New source & NSPS Ja 199
7.3 Feasibility of 2 psig requirement 199
7.4 New/revised definitions related to delayed coking units 200
7.4.1 Delayed coker vent 200
7.4.2 Decoking operations 201
7.5 Impact estimates for DCU requirements 202
7.5.1 Emission reductions 202
7.5.2 Control costs 204
7.5.3 Cost effectiveness and risk 205
7.6 Monitoring requirements 206
8.0 Fenceline Monitoring Provisions 209
8.1 Proposed siting procedures (including comments on Method 325A) 212
8.2 Proposed analysis procedures (including comments on Method 325B) 224
8.3 Need for alternative monitoring technologies for fenceline monitoring 234
8.4 Applicability of fenceline monitoring requirements 240
8.4.1 Require tiered levels of monitoring 241
8.4.2 Limit time or number of locations that need ongoing monitoring if concentrations are
low 242
8.5 Benzene as target analyte 243
8.6 Adjusting for background emissions 246
8.6.1 Delta C approach 248
8.6.2 Site-specific approach for near-field sources 251
8.7 Action level 253
8.8 Non-compliance and Corrective Action Plan 255
8.9 Recordkeeping and reporting requirements 263
8.10 Cost of Fenceline Monitoring 266
8.11 Other 269
9.0 Other Refinery MACT 1 Provisions 270
9.1 Applicability and Affected Sources 271
9.1.1 Applicability dates and Table 11 (other than for PRD) 271
9.1.2 Overlap provisions 272
9.2 Wastewater 274
9.3 Gasoline loading racks 278
9.4 Marine Vessel Loading Operation Provisions 279
in
-------
10.0 Refinery MACT 2 Amendments 281
10.1 Catalytic Cracking Units (fluid and other) 282
10.1.1 Inclusion ofNSPS Ja compliance option 283
10.1.2 Phase-out ofNSPS J compliance option 287
10.1.3 Startup shutdown provisions 290
10.1.4 Performance testing every 5 years 295
10.2 Catalytic Reforming Units 295
10.3 Sulfur Recovery Units 298
10.3.1 Inclusion ofNSPS Ja compliance option 298
10.3.2 Startup shutdown provisions 299
10.4 CPMS Requirements in Table 41 303
10.5 General provisions applicability (Table 44) 305
11.0 General Compliance Requirements for MACT standards 306
11.1 Removal of SSM plan/exemptions 309
11.2 Testing and monitoring requirements 323
11.3 General duty provisions 333
11.4 Electronic reporting requirements 334
12.0 Refinery NSPS Subpart Ja 338
12.1 Sulfur Recovery Plant Provisions 341
12.2 Performance Test Requirements (for flares for H2S) 341
12.3 Other Revisions 344
13.0 Economic Impacts 346
13.1 Economic Impact Analysis 346
13.2 ICR Burden Estimates 347
14.0 Statutory and Executive Orders 349
15.0 Other Comments 352
15.1 General Support 352
15.2 Editorial Corrections 352
15.3 Miscellaneous Other Comments 352
15.4 Out-of-Scope 365
iv
-------
LIST OF ABBREVIATIONS, ACRONYMS AND UNITS OF MEASURE
AB2588
Air Toxics "Hot Spots" Information and Assessment Act
Admin.
Administrative
AEGL
acute exposure guideline levels
AERMOD
air dispersion model used by the HEM-3 model
AFPM
American Fuel & Petrochemical Manufacturers
AMEL
alternative means of emission limitation
AMOS
ample margin of safety
AMP
alternative monitoring plan
APCD
air pollution control devices
API
American Petroleum Institute
AQMD
Air Quality Management District
ARB
Air Resources Board
Ass'n
association
ASTM
American Society for Testing and Materials
AT SDR
Agency for Toxic Substances and Disease Registry
BAAQMD
Bay Area Air Quality Management District
BLD
bag leak detectors
BP
British Petroleum
Btu
British thermal units
BWON
Benzene Waste Operations NESHAP
CA
corrective action
CAA
Clean Air Act
CAA
corrective action analysis
CalEPA
California EPA
CASRN
Chemical Abstract Services Registry Number
ecu
catalytic cracking units
CCV
Continuing Calibration Verification
Ccz
combustion zone combustibles concentration
CDC
Centers for Disease Control
Cdii
combustibles concentration dilution parameter
CE
combustion efficiency
CEMS
continuous emissions monitoring system
CERCLA
Comprehensive Environmental Response, Compensation, and Liability Act
cert.
certiorari (writ or order by which higher court reviews decision of lower court)
cfm
cubic feet per minute
CFR
Code of Federal Regulations
CGA
cylinder gas audit
ch4
methane
CHIEF
Clearinghouse for Inventories and Emissions Factors
CHPAC
Children's Health Protection Advisory Committee
Cir.
Circuit Court
CO
carbon monoxide
C02
carbon dioxide
v
-------
CChe carbon dioxide equivalents
COMS continuous opacity monitoring system
CPMS continuous parameter monitoring system
CRU catalytic reforming units
CVS closed-vent system
D.C. District of Columbia
DCS distributed control system
DCU delayed coking units
DE destruction efficiency
DIAL Differential Absorption Light Detection and Ranging
dL deciliters
DNA deoxyribonucleic acid
DOT Department of Transportation
dscf dry standard cubic feet
DTSC Department of Toxic Substances Control
EBU enhanced biological unit
ECHO Enforcement and Compliance History Online
e.g. exempli gratia (for example)
EIA Economic Impact Assessment
EJ environmental justice
EO Executive Order
EPA U.S. Environmental Protection Agency
EPCRA Emergency Planning and Community Right-to-Know Act
ERPG emergency response planning guidelines
ERRRNSPS Electronic Reporting and Recordkeeping Requirements for the New Source
Performance Standards Rulemaking
ERT Electronic Reporting Tool
ESP electrostatic precipitator
etc. et cetera (and so forth)
ETV Environmental Technology Verification
FCCU fluid catalytic cracking units
FQPA Food Quality Protection Act
FR Federal Register
ft feet
ft3 cubic feet
FTIR Fourier transform infrared spectroscopy
GC gas chromatograph
GEAE Generic Ecological Assessment Endpoint
GHG greenhouse gases
GHGRP Greenhouse Gas Reporting Program
GP General Provisions
g grams
gr grains
H2S hydrogen sulfide
HAP hazardous air pollutants
HC1 hydrogen chloride
vi
-------
HCN
hydrogen cyanide
HEM-3
Human Exposure Model, Version 1.1.0
HF
hydrogen fluoride
HFC
highest fenceline concentration
HIA
Health Impact Assessment
HON
Hazardous Organic NESHAP
HQ
hazard quotient
hr
hour
HRVOC
highly reactive volatile organic compounds
ICR
Information Collection Request
id.
idem (the same)
i.e.
id est (that is)
IQ
intelligence quotient
IRIS
Integrated Risk Information System
IS
Internal Standard
ISO
International Organization for Standardization
1ST
inherently safer technologies
kg
kilograms
km
kilometers
L
liters
L&E
Locating and Estimating Air Emissions from Sources of Benzene
LACEEN
Los Angeles Community Environmental Enforcement Network
lb
pounds
LDAR
leak detection and repair
LDEQ
Louisiana Department of Environmental Quality
LEL
lower explosive limit
LFL
lower flammability limit
LFLcz
combustion zone lower flammability limit
LFLdii
lower flammability limit dilution parameter
m3
cubic meters
M325A
Method 325A
M325B
Method 325B
MACT
maximum achievable control technology
MDL
method detection limit
MFC
measured fenceline concentration
mg
milligrams
min
minutes
MIR
maximum individual risk
ml
milliliters
MMbbl
million barrels
mmBtu
million British thermal units
MOA
mode of action
MOE
margin of exposure
MON
Miscellaneous Organic NESHAP
mph
miles per hour
MPV
miscellaneous process vent
vii
-------
MSS maintenance, startup and shutdown
MTBE methyl tert-butyl ether
NAAQS National Ambient Air Quality Standards
NAS National Academy of Sciences
NEI National Emissions Inventory
NESHAP National Emission Standards for Hazardous Air Pollutants
NFS near-field interfering source
ng nanograms
NHV net heating value
NHVcz combustion zone net heating value
NHVdii net heating value dilution parameter
NHVvg net heating value of the vent gas
Ni nickel
NIST National Institute of Standards and Technology
No. number
NOAA National Oceanic and Atmospheric Administration
NOAEL no-ob served-adverse-effect level
NOCS notification of compliance status
non-PBT not persistent, bioaccumulative and toxic
NOx nitrogen oxides
NPS National Park Service
NRC National Research Council
NRDC Natural Resources Defense Council
NSPS new source performance standards
NTP National Toxicology Program
02 oxygen
03 ozone
OAQPS Office of Air Quality Planning and Standards
ODW Office of Drinking Water
OEHHA Office of Environmental Health Hazard Assessment
OEL open-ended line
OGI optical gas imaging
OHEA Office of Health and Environmental Assessment
OMB Office of Management and Budget
OMMP Operating, Monitoring, and Maintenance Plan
OOC out of control
OSC off-site source contribution
OSHA Occupational Safety and Health Administration
OTM other test method
PAH polycyclic aromatic hydrocarbons
PB-HAP hazardous air pollutants known to be persistent and bioaccumulative in the
environment
PBM Perimeter Boundary Monitoring
PBT persistent, bioaccumulative, and toxic
PFTIR passive Fourier transform infrared spectroscopy
PM particulate matter
viii
-------
PM2.5
particulate matter 2.5 micrometers in diameter and smaller
PM10
particulate matter 10 micrometers in diameter and smaller
POM
polycyclic organic matter
ppb
parts per billion
ppbv
parts per billion by volume
ppm
parts per million
ppmv
parts per million by volume
ppmw
parts per million by weight
PRA
Paperwork Reduction Act
PRD
pressure relief devices
PRV
pressure relief valves
PS
Performance Specification
psi
pounds per square inch
psia
pounds per square inch absolute
psig
pounds per square inch gauge
PSM
Process Safety Management
PTE
potential to emit
P/V
pressure/vacuum
PVC
polyvinyl chloride
QA/QC
quality assurance/quality control
RATA
relative accuracy test audit
RC
root cause
RCA
root cause analysis
REL
reference exposure level
REM
Refinery Emissions Model
RfC
reference concentration
RfD
reference dose
RIA
Regulatory Impact Analysis
RMP
Risk Management Plan
RQ
reportable quantity
RTC
response to comments
RTR
residual risk and technology review
RV
relief valves
RWET
Refinery Wastewater Emission Tool
SAB
Science Advisory Board
SCAQMD
South Coast Air Quality Management District
scf
standard cubic feet
SCR
selective catalytic reduction
s
seconds
S02
sulfur dioxide
SOF
Solar Occultation Flux
SOx
sulfur oxides
SRP
sulfur recovery plant
SRU
sulfur recovery units
ss
startup and shutdown
SSM
startup, shutdown and malfunction
ix
-------
STERPP Storage Tank Emission Reduction Partnership Program
TAB total annual benzene
TAC Texas Administrative Code
tbl. table
TCEQ Texas Commission on Environmental Quality
Tex. Texas
TexasAQS Texas Air Quality Study
THC total hydrocarbons
TOSHI target organ-specific hazard index
tpy tons per year
TRI Toxics Release Inventory
TRS total reduced sulfur
UB uniform background
UF uncertainty factor
UF-H intraspecies (human) uncertainty factor
URE unit risk estimate
U.S. United States
U.S.C. United States Code
USW United Steel workers
USWS U.S. Weather Service
UV ultraviolet
UV-DOAS ultraviolet differential optical absorption spectroscopy
v. versus
VE visible emissions
VOC volatile organic compounds
VOHAP volatile organic hazardous air pollutant
vol. volume
WGS wet gas scrubber
AC concentration difference between the highest measured concentration and the
lowest measured concentration
°F degrees Fahrenheit
$ dollars
|ig micrograms
% percent
x
-------
1.0 Introduction/Comment Period Extension
The U.S. Environmental Protection Agency (EPA) promulgated national emissions
standards for hazardous air pollutants (NESHAP) pursuant to the Clean Air Act (CAA) section
112(d)(2) and (3) for petroleum refineries located at major sources in three separate rules. The
first of which was promulgated on August 18, 1995 in 40 CFR, subpart CC (also referred to as
Refinery MACT 1) and regulates miscellaneous process vents, storage vessels, wastewater,
equipment leaks, gasoline loading racks, marine tank vessel loading and heat exchange systems.
The second rule was promulgated on April 11, 2002 in 40 CFR subpart UUU (also referred to as
Refinery MACT 2) and regulates process vents on catalytic cracking units (CCU, including fluid
catalytic cracking unit (FCCU)), catalytic reforming units (CRU), and sulfur recovery units
(SRU). Finally, on October 28, 2009, the EPA promulgated maximum achievable control
technology (MACT) standards for heat exchange systems which were not originally addressed in
Refinery MACT 1. This same rulemaking included updating cross-references to the General
Provisions in 40 CFR part 63.
Section 112(f)(2) of the CAA requires the EPA to determine for each CAA section
112(d) source category if promulgation of additional standards is required "in order to provide an
ample margin of safety to protect the public health." The EPA may also impose a more stringent
emission standard to prevent adverse environmental effect if such action is justified in light of
costs, energy, safety, and other relevant factors. Section 112(d)(6) of the CAA requires the EPA
to review NESHAPs and to revise them "as necessary (taking into account developments in
practices, processes, and control technologies)" no less frequently than every 8 years. On
September 27, 2012, a lawsuit was filed which alleged that the EPA missed statutory deadlines
to perform review and revise, as necessary, the Refinery MACT 1 and 2. The EPA has
completed both a technology and risk review of Refinery MACT 1 and 2 to fulfill its obligations
under 112(f)(2) and 112(d)(6). The EPA proposed amendments as a result of these reviews on
June 30, 2014. In addition to the proposed amendments relative to the Petroleum Refinery
MACT 1 and 2, the EPA also proposed amendments to the refinery new source performance
standards (40 CFR part 60 subparts J and Ja) to address technical corrections and clarifications
raised in a 2008 industry petition for reconsideration. These were addressed in the proposed
rulemaking because they affect sources included within the Refinery MACT 1 and 2
amendments. The proposed amendments include new standards for storage vessels, delayed
coking units (DCU), flares, fenceline monitoring of fugitive emissions sources and FCCU.
The proposal provided a 60-day comment period ending on August 29, 2014. Upon
request, the EPA extended the comment period for an additional 60 days ending October 28,
2014. During the comment period, the EPA held two public hearings on July 16, 2014 in
Wilmington, CA and on August 5, 2014 in Galena Park, TX, to provide the public the
opportunity to present data, views, or arguments concerning the proposed amendments. The EPA
received 202,650 comments (including those from mass mail campaigns and public hearing) on
the proposed amendments from refiners, industry, trade associations, and consultants, state and
local environmental and health departments, environmental groups, private citizens and other
interested parties during the comment period. These comments were reviewed and it was
1
-------
determined that there were 201 unique and substantive comments. All of the comment letters,
including transcripts of the public hearings, have been placed in the docket for this rulemaking
(Docket No. EPA-HQ-2010-0682). Key comment summaries and EPA responses are included in
the preamble of this rulemaking package. This document includes a summary of the other
substantive comments received and written responses from the EPA.
2
-------
2.0 Risk Assessment
2.1 Emissions Data
2.1.1 Actual emissions and release characteristics
Comment 1: Several commenters expressed concerns that U.S. refineries are receiving greater
quantities of unconventional crudes including tar sands, oil shale and tight oil, and the impacts of
this shift in raw material has not been incorporated into the health risk assessment for this
rulemaking, thus making it unlawfully incomplete, and arbitrary and capricious.
The commenters claimed that the use of these unconventional crudes poses a number of
additional environmental, health and safety risks relative to standard conventional crude oil. The
risks that they outlined include:
Higher risk of refinery accidents: Tar sands and tight oil are more corrosive than
conventional crude oil, which increases the risk of refinery accidents and raises the
potential for significant additional air emissions during upsets and accidents.
Increases in air pollution & associated health impacts: The chemicals used to dilute and
blend tar sands contain highly volatile and sometimes toxic organic chemicals at much
higher concentrations than conventional crude oil. Tar sands also contain many toxic
constituents including heavy metals, such as lead, at much higher concentrations than in
conventional crude oil.1 The much heavier, denser tar sands crude requires greater use of
heaters, boilers, hydro-treating, coking, cracking and greater hydrogen use, all of which
creates greater emissions of smog- and soot-forming pollutants and toxic chemicals.2
Tight oil has an unusually high flashpoint and Reid Vapor Pressure leading to potentially
double the emissions of many light organic hazardous air pollutants (HAP) such as
benzene from storage and handling.
Irreversible Climate Damage: Carbon pollution from extracting and upgrading tar sands
can be 3 to 5 times greater than for conventional crude oil. Over the course of production
1 R.F. Meyer, E.D. Attanasi, and P.A. Freeman, Heavy Oil and Natural Bitumen Resources in Geological Basins of
the World, U.S. Geological Survey Open-File Report 2007-1084, 2007, p. 14, Table 1;
http://pubs.usgs.gov/of/2007/1084/OF20Q7-1084vl.pdf.
2 Excerpt from Phyllis Fox Report to NRDC, July 1, 2013.
3
-------
(well-to-tank), tar sands release 80% more global warming pollution than the U.S.
average refined crude.3'4
Other environmental & community impacts: Tar sands processing releases strong odors
due to the higher levels of sulfur compounds, particularly noxious mercaptans. Exposure
to mercaptans may cause irritation of the skin, eyes, and upper respiratory tract, and
mercaptans negatively affect the central nervous system. Refining tar sands also leads to
roughly 50% more petroleum coke, [Lattanzio CRS Report, 2014] which is a hazardous
by-product that in some cases is stored in open piles creating a serious health hazard.
The commenter concluded that the EPA neglected to consider the many serious impacts to
refinery emissions from unconventional crude oil including tar sands and Bakken crude, both of
which are relatively new types of crude oil and yet are already in significant and increasing use
at U.S. refineries. Unconventional crude oil refining is likely to significantly increase volatile
HAP as well as heavy metals, polycyclic aromatic hydrocarbons (PAH), and other HAP,
resulting in significant health impacts to refinery fence-line communities.
Response 1: First, the types of control systems used to limit the emissions from petroleum
refinery process units are essentially independent of the types of crude processed. The EPA finds
that the same MACT requirements would be needed to ensure sources using these
unconventional crude slates are achieving the MACT emissions limitations. For
example, delayed coking units are commonly used to process heavy crudes into lighter distillates
(gasoline or diesel fuel). The MACT standards that we are finalizing for delayed coking units
would apply to units processing unconventional heavy crude oils exactly the same as to units
processing conventional heavy crudes. Thus, the MACT standards will help to limit the
emissions from refineries regardless of the type of crude oil processed.
Second, the precise impact of different crude slates is very difficult to project, because it is
difficult to predict how much of each type of unconventional oil will be processed and which
conventional crude slates will be replaced. Heavy and sour crude oils are already processed in
the U.S. If crude oil produced from tar sands replaces these heavy sour crudes, there may be very
little difference in the amount of upgrading equipment used or the amount of emissions from this
shift in crude feedstocks. Even if significant quantities of unconventional crude oil are processed
in U.S. refineries, the fuel standards and the refinery MACT requirements will limit HAP (and
sulfur) releases from refineries and subsequent petroleum product use. For example,
hydrotreaters are commonly used to remove sulfur and heavy metals from petroleum
intermediate streams. This processing will help to ensure emissions from downstream refinery
processes are similar to those processing sweeter crudes. As such, there is significant uncertainty
3 Richard K. Lattanzio, Congressional Research Service report: Canadian Oil Sands: Life-Cycle Assessments of
Greenhouse Gas Emissions, March 10, 2014. This source also reports a 17% increase in GHGs from tar sands v.
conventional crude over the lifecycle, including the use of the fuel in vehicles.
4 Jessica P. Abella and Joule A. Bergerson, Model to Investigate Energy and Greenhouse Gas Emissions
Implications of Refining Petroleum: Impacts of Crude Quality and Refinery Configuration Environ. Sci. Technol.,
2012, 46 (24), pp 13037-13047 DOI: 10.1021/es3018682
4
-------
as to whether these unconventional crudes will really cause any appreciable increase in refinery
HAP emissions.
With respect to climate impacts, we consider these comments to be beyond the scope of this
rulemaking, particularly as many of the impacts cited relates to upstream production and
processing that occurs prior to the petroleum refinery.
At this time, we cannot scientifically defend revisions to refinery-specific emission inventories in
an effort to estimate health risk assessments for some future time with assumed changes in crude
slates. Furthermore, we consider that any changes in crude slate that actually do occur in the
future will have a relatively small impact on refinery HAP emissions because the MACT control
requirements we are finalizing will adequately control the refinery HAP emissions, and these
changes would not alter our conclusion that the risks from petroleum refineries are acceptable.
Comment 2: Two commenters stated that emissions are over-stated because of the use of the
section 114 information collection request (ICR) data collection effort in 2011 and the refinery
emissions protocol document. Both commenters provided specific objections related to the
EPA's requirements for preparing the emissions inventory for use in risk modeling, stating that
the inventory approach required to be used tended to over-estimate the refinery's emissions,
citing a critique of the EPA's risk analysis entitled "Refinery Air Emission Metrics" by Jess A.
McAngus and Erin Valley of Spirit Environmental LLC which compared ICR inventory
estimates to the Toxic Release Inventory (TRI) data. Commenters also state that there were
unsupported adjustments made to the ICR data including increasing the hydrogen cyanide (HCN)
emissions for FCCUs in the emissions inventory by a factor of 10 for those sites that did not
provide actual stack test data.
Several commenters stated that many significant emission reductions have occurred since 2010
and those changes should have been reflected in the emission estimates and risk modeling. Most
significant are the reductions anticipated due to new source performance standards (NSPS) Ja.
One commenter stated other changes that should also have been reflected in the emission basis
for the risk modeling include the following:
The EPA states that the floor for the proposed delayed coker work practice standard is
based on eight such units having to meet a 2 pounds per square inch gauge (psig) pressure
limit. The risk input should reflect that limit, not their 2010 emissions. Furthermore, the
South Coast Air Quality Management District (SCAQMD) has a rule in place that
requires all 8 DCUs in that District to meet a 2 psig limit by late 2017 and those units
should have been modelled on that basis as well.
Consent decrees specifically addressing flares, a focus of this proposal, have been put in
place since 2010 for six Marathon petroleum refineries, the Shell Deer Park Refinery, the
Flint Hills Port Arthur Refinery, the Countrymark Indiana Refinery and the British
Petroleum (BP) Whiting Refinery. The reductions in flaring and other emissions resulting
from these and other refinery consent decrees should be reflected in the baseline emission
and risk estimates.
5
-------
One commenter also added that emissions of volatile organic compounds (VOC) and toxic air
contaminants have been reduced through current and previous emission control efforts, such as
state regulations. The commenter provided California's Air Toxics "Hot Spots" Information and
Assessment Act (AB2588) process as an example, which reduced toxic emissions and risk such
that most refineries pose risks that are already less than 10-in-l million. The commenter
recommended that the EPA should recognize all efforts to reduce emissions into the agency's
analysis on a regional level. This will allow the EPA the opportunity to incorporate updated
emissions data and cost information to determine if implementation of the NESHAPs on a local
level is reasonable and feasible.
Response 2: The commenter's objections to the 2011 ICR data collection efforts and emissions
inventory protocol have been noted and we appreciate the commenter's submission of the
ICR data and emissions inventory in spite of their objections. We prepared the Refinery Protocol
to provide guidance to refinery owners or operators to use the best available, site-specific data
when developing their emissions inventory, to ensure that all emission sources are included in
the inventory, and that there is a consistent set of emission factors that all respondents use if no
site-specific emissions data were available. We disagree that the refinery emissions protocol
would lead to overly conservative emissions estimates for the inventory. If site-specific
emissions data were available, sites were to use these data preferentially over the default
factors. The default factors provided in the Protocol were developed from the best data available
at the time.
While TRI reporting is required, there is no uniform set of requirements for refinery owners or
operators to follow in developing those emissions inventories. The idea that the TRI emission
estimates are more accurate and complete than the inventories produced for the ICR effort is not
supported by the facts. The Refinery Protocol included methodologies and emission factors for
sources that have been historically omitted from emission inventory estimates and it is precisely
because of the Refinery Protocol document that we consider the ICR inventories to be more
thorough, consistent and complete than the TRI data.
We acknowledge that we increased the emission inventory estimated for hydrogen cyanide based
on the emissions test data results from the ICR testing. We determined that the HCN emissions
factor we provided in the Refinery Protocol, while likely be greater than what refineries used for
TRI reporting, which understated the actual HCN emissions. Therefore, we increased the HCN
emissions based on an updated HCN emissions factor derived from more recent ICR source test
data. An added advantage of requiring facilities to use the Refinery Protocol factors when no
site-specific emissions data are available is that, if we determine a specific emissions factor is
inaccurate, it is relatively simple and straight forward to adjust the ICR emissions inventory
appropriately. We strongly disagree with the commenter's assertion that this adjustment was
somehow "arbitrary" as the revised HCN emissions factor was based on a greater quantity and
quality of data than the original factor provided in the Refinery Protocol. We note that,
subsequent to the ICR effort, many refiners have increased their HCN emission estimates in TRI
reports.
We also disagree with commenters that our risk assessment was overly conservative in light of
emission reductions purported future emission reductions. While we appreciate the fact that
6
-------
some refineries have entered into consent decrees for their flares, we note that the refinery
Protocol recommended that flare emission control efficiencies of 98 percent should be used, so
the flare emissions inventory provided in response to the ICR already reflects the improvement
in flare performance expected as a result of these consent decrees (as well as the proposed
operating limits for flares).
We routinely assess the impacts of a rule based on the information available at the time and it is
impossible to continually revise the baseline emissions and emissions reductions based on on-
going changes in the industry. While the commenters only discuss cases they expect would
reduce future emissions, we also know that there have been some significant refinery expansions
that were completed since 2010, and these changes would increase baseline emissions. We
performed the analysis based on the 2010 data set because it was the most recent and complete
data set available by which to assess health risks resulting from refinery emissions.
Even if, for the sake of argument, the emissions inventory developed using the emissions
Protocol resulted in a conservative emissions inventory, this only strengthens the conclusion that
risks are acceptable since the risks determined using those "conservative" emissions
inventory are less than 100-in-l million.
Comment 3: Several commenters stated that the EPA underestimates exposure because
emissions are underreported and underestimated. The commenters noted that for the risk
assessment for the refineries rule, the EPA evaluated (1) the emissions reported to the agency
pursuant to the 2011 Petroleum Refinery ICR as sources "actual" emissions based on some tests
and some estimation, and (2) the emissions the EPA estimates that the existing standards
currently allow sources to emit, which it describes as "allowable" emissions, based on the REM.
According to the commenters, both data sets are incomplete and undercount emissions, causing
the EPA to significantly underestimate the resulting risk they cause in its risk analysis.
According to one commenter, the ICR emissions inventory significantly underestimates
emissions because the emission inventory is largely calculated from emission factors and
engineering judgment and it is well documented that emission factors underestimate emissions
for a variety of reasons including inherent bias in the factors themselves and the inability to
account for equipment malfunctions and environmental conditions. The commenter cited a
variety of Differential Absorption Light Detection and Ranging (DIAL) studies and other studies
to support the assertion that emission factors understate emissions, particularly for fugitive
sources. Furthermore, the commenter suggested that there is evidence that facilities did not use
the emission factors as directed or applied them incorrectly and the EPA has recognized some of
these problems in the technology review component of this rulemaking and made some
adjustments.
Another commenter similarly stated that a Texas Air Quality Study (TexasAQS) found large
volumes of unreported airborne organic chemicals (see:
http://www.utexas.edu/research/ceer/texaqsII/visitors.htm, and TexasAQS II Field Study at:
http://vvvvvv.tceci.state.tx.us/airciualitv/research/texacis/texacis data.html). One of the early phases
of the first TexasAQS found that VOC and air toxics concentrations in the Houston area above
the major oil refineries were on average 6X-12X higher than those that were being reported by
7
-------
the refineries in the annual emissions inventory (EI) submitted to the Texas Commission on
Environmental Quality (TCEQ). The TCEQ, scientists with the Texas AQS, and industrial
officials concluded that the large volumes of unreported VOCs and air toxics were being released
through four plant processes: flares, process vents, fugitive leaks, and cooling towers. The
commenter stated that the significant under-reporting of VOC and air toxics indicates that the
EPA needs stronger rules for more accurate VOC/HAP monitoring that that the EPA needs to
require more reductions in refinery HAP, particularly from elevated flares.
In conclusion, the commenters stated that the EPA's risk analysis underestimates exposure
because it relied on inaccurate data and assumptions to calculate and model the total amount of
HAP emissions released by petroleum refineries every year, both in terms of the so-called
"actual" and "allowable" emissions. For this reason alone, the EPA should find the level of
current health risk to be "unacceptable," and set standards under 112(f)(2) to assure the requisite
"ample margin of safety to protect public health" from additional emission points at refineries
(including wastewater treatment, equipment leaks, FCCUs and process vents), to reduce these
emissions.
Response 3: We used the best and most robust facility-specific HAP emissions inventory
available to us, which was the 2011 ICR, in performing the analysis for the proposed rule. We
conducted a thorough and exhaustive review of the data submitted through the ICR and we
followed up on source-specific information on a facility-by-facility basis, as documented in the
"Emissions Data Quality Memorandum and Development of the risk Model Input File" (see
Docket Item No. EPA-HQ-2010-0682-0076). In addition, we took steps ahead of issuing the
2011 ICR to make sure that facilities could, as accurately, as practicable as possible, estimate
their HAP emissions for purposes of responding to the inventory portion of that ICR. We
prepared a Refinery Protocol to provide guidance to refinery owners or operators to use the best
available, site-specific data when developing their emissions inventory, to ensure all emission
sources are included in the inventory, and that the agency has a consistent set of emission factors
that all respondents use if no site-specific emission data were available. If site-specific emissions
data were available sites were to use these data preferentially over the default factors. We
developed default factors provided in the protocol from best data available at the time.
The ICR-submitted information for allowable emissions did not include emission estimates for
all HAP and all emission sources. Consequently, we used the REM to estimate allowable
emissions. The REM relies on model plants that vary based on throughput capacity. Each model
plant contains process-specific default emission factors, adjusted for compliance with the
Refinery MACT 1 and 2 emission standards.
Regarding the findings of the TCEQ AQS, we promulgated new MACT standards in 2010
for heat exchange systems to reduce emissions from cooling towers. Based on TCEQ and other
studies, we proposed and are finalizing significant new operating and monitoring requirements
for flares to ensure flares are achieving the required control efficiency. We also note that we
proposed and are finalizing fenceline monitoring requirements in part to ensure the fugitive
emissions inventory that we used in the risk modeling file are an accurate representation of
actual refinery emissions.
8
-------
If the industry commenters are correct, then the emissions inventory used in the risk assessment
overstates the emissions and the risk conclusion is conservative. If the emissions inventory used
in the risk assessment understates the emissions, then the fenceline monitoring and corrective
action requirements will ensure refineries reduce their emissions to levels comparable to their
emissions inventory and the surrounding communities would be protected to acceptable
risk levels.
Comment 4: One commenter stated that a detailed review of the EPA's risk assessment by
ENVIRON Corporation was commissioned by American Petroleum Institute (API) and they
attached a copy of this report to their comment (Attachment A-l to 0583). This risk modeling
assessment was performed for the 33 refineries reported by the EPA that have the highest cancer
and chronic noncancer risk to assist API in verifying the EPA's risk estimates and to identify key
risk-drivers at those facilities for further review by API and member companies and correcting
data if warranted. The commenter summarized key findings as follows:
While modeled results are consistent with the EPA's judgment in its proposed rule that
refinery risks are "acceptable," certain potential data issues were identified that, once
corrected, will lower risk estimates for the highest-risk refineries and others.
For certain refineries, the EPA modeled emission-outlier sources that overestimated
actual emissions, sources that were mislocated or incorrectly extended offsite, or
calculated maximum risks at nonresidential locations, causing risks to be misstated.
For certain refineries, the EPA's risk modeling does not reflect corrected emission
inventory data, causing risks to be overstated.
Correction of data errors, if present, and re-modeling of risk by the EPA using corrected
data is warranted. For example, using corrected data from companies, the Human
Exposure Model, Version 1.1.0 (HEM-3) was used to re-calculate risk for the two
refineries with the highest EPA-reported facility-wide cancer risk, correcting modeled
cancer risk from 67-in-l million to 15-in-l million for one refinery, and from 67-in-l
million to 13-in-l million for the other.
A few commenters (Western Refining and Northern Tier Energy) confirmed the errors found by
ENVIRON and stated they would resubmit their corrected emissions data using the EPA's tool.
The commenters stated that there are serious technical issues in the EPA's risk modeling for
refineries that greatly overstate the risk. Consequently, the EPA's modeling does not provide a
credible basis to evaluate refinery risk or to substantiate any purported risk reduction benefits to
be derived from the proposed rulemaking.
Response 4: We note that most of the issues noted by the commenters are issues with the
emissions inventory files that were submitted to the EPA. We appreciate the commenters' review
of the emission sources that were driving high risk results and the emissions inventory
corrections submitted by various refineries. While a detailed modeling analysis of the revised
emissions received during the comment period is not possible, the EPA risk modelers and
engineers performed a screening analysis of these emissions revisions to determine their
potential impact on the source category risk estimates. We estimate that the source category
chronic cancer MIR based on actual emissions may be closer to 40-in-l million as compared to
9
-------
the 60-in-l million predicted by the initial risk modeling. Since allowable emissions are
primarily based on model plant emissions, the model predicted chronic cancer maximum
individual risk (MIR) of 100-in-l million does not change based on the submitted emissions
revisions. Other risk metrics changes (i.e., incidence, populations in risk bins, multipathway and
ecological analyses) were not quantified but would likewise be expected to show some minor
reductions from those presented in the draft risk assessment report.
Comment 5: A commenter stated that for decades, communities have faced substantial amounts
of uncontrolled "off-the-books" pollution during periods of startup, shutdown, and malfunction
(SSM). Another commenter stated that the EPA's emission estimates, developed to support its
assessment of risks, are incomplete because they do not adequately account for the fact that HAP
releases during malfunctions or maintenance (which are regular events at most refineries) are
often much higher than emissions during "normal" operations, and are either under-reported or
not reported at all. Commenters cited some specific examples from the ICR Component 1 when
thousands of pounds of HAP were released during upset or malfunction events such as piping
failure or leaks which occurred over a few minutes or many days, respectively, to support this
argument. The commenter stated that the EPA must revise its risk-analysis of allowable
emissions to account for these problems.
One of the commenters added that refineries release millions of pounds of emissions during SSM
periods each year. According to the commenter, data from Texas shows that SSM emissions are
a severe public health problem. Between 2009 and 2013 Texas refineries alone released nearly 1
million pounds of HAP. The commenter pointed out that while facilities reported more than
100,000 pounds of HAP emissions in Component 1 of the EPA's ICR, 96.77% of these
emissions were attributable to the 43 Texas and Louisiana refineries. It seems unlikely that the
other 103 refineries across the country accounted for less than 5% of startup, shutdown, and
malfunction pollution.
A commenter also added that a recent investigation of 18 Texas oil refineries between 2003 and
2008 found that "upset events" were frequent, with some single upset events producing more
toxic air pollution than what was reported to the federal Toxics Release Inventory database for
the entire year.
The commenters stated that it is not clear from the record that the EPA considered these
hazardous events in its risk review of whether to require stronger standards, including leak
detection and repair (LDAR) requirements, or the total and acute risks that communities near
refineries are exposed to. The commenters stated that the EPA needs to account for the fact that
upsets or repair events may not even be treated as a reportable malfunction or maintenance event.
The commenters requested that the EPA should further require facilities to install the most
responsive monitoring available on the release of emissions from pressure relief devices, control
devices, bypass lines and flares.
Response 5: We agree that SSM emissions can be significant and that these releases, particularly
when directed straight to the atmosphere rather than to a flare or other control device can quickly
exceed emissions from routine operations. While large release events can significantly impact a
facility's annual emissions, we disagree with the commenter that these large events are
10
-------
frequent. The commenters note that "upset events" are frequent, but the large "upset events" that
significantly alter a facilities annual emissions only make up a small portion of the total "upset
events." Based on our review of the TCEQ incident database, many of these large release events
had to do with sulfur dioxide (SO2) emissions and only a few had significant HAP emissions.
Thus, we do not characterize large releases that may occur once every 3 to 5 years to be
"frequent."
Episodic malfunction events cannot be predicted so it is impossible to estimate the location and
quantity of these emission events for direct inclusion into the risk assessment. However, since we
are now allowing a limited number of these events from pressure relief devices and flares, we
performed a screening level analysis for the risks from these non-routine emissions. To estimate
emissions for chronic risk calculations, emissions data was extracted from the EPA ICR in
Component 1 of the survey. We summed the HAP emissions from all individual events at each
facility, since chronic inhalation risk depends on the total quantity of emissions released over a
year and is independent of the release time. This screening level analysis indicated we can expect
these emissions to contribute up to about 2-in-l million to the chronic cancer MIR. Noncancer
risks resulting from these emissions are expected to remain well below 1. To estimate potential
risks for acute events, we examined both the quantity and duration of each event such that higher
mass emitted in a shorter time period has an increased potential for an acute impact. Acute risks
from these events estimated a Hazard Index based on the REL of up to 14 from emissions of
benzene. In all cases the acute exposure guideline levels (AEGL) levels are well below 1. It is
important to note that the estimated acute risk assume a catastrophic release such that all the
emissions from an event are emitted during a single hour time period (for more detail on this
analysis, see Appendix 13 of the Final Residual Risk Assessment for the Petroleum Refining
Source Sector in Docket No. EPA-HQ-OAR-2010-0682).
We also proposed revisions to the MACT requirements to eliminate "off-the-books" emissions
(by reducing these events and making refinery owners or operators report these emissions when
they do occur). In the final rule, we revised the monitoring requirements for bypass lines from
the proposal, but we are retaining the requirements to estimate and report the emissions released
during control system bypasses. We have also revised the requirements for pressure relief
devices (PRD) from what was proposed, but we are still requiring that facility owners or
operators determine the quantity of emissions released and report these emissions to the
EPA. We proposed significant monitoring requirements for flares and we are finalizing these
monitoring requirements with minor revisions from what was proposed. Finally, we proposed
fenceline monitoring requirements to more quickly identify leaks and other fugitive emissions
and to ensure the emission inventories provided to the EPA were reasonably accurate. As such,
we proposed significant enhancements in the monitoring and reporting requirements petroleum
refineries to limit and report emissions for these "upset events" and we are finalizing these
enhanced monitoring and reporting requirements (with some revisions from proposal) in the final
rule.
Comment 6: One commenter stated that emission totals for the petroleum refinery sector were
underestimated and requested the that EPA detail their plans for future Information Collection
Requests, new stack tests, and technology and risk as well as applicability review for petroleum
refineries that are minor sources. The commenter explained that some of these refineries are
11
-------
minor sources because of third party lease agreements and argued that these business agreements
allow sources to remain exempt from the Sector Rule proposal when if considered one entire
source would constitute a major source. The commenter also inquired whether co-located
recycling facilities processing motor oil or animal fats that are used to produce diesel fuel on a
refinery's campus, but not included on their permit would be addressed through future
rulemakings (e.g., ICR, stack test, residual risk and technology review (RTR)).
Response 6: At this time, we do not have any plans to conduct an additional, detailed ICR for
the petroleum refining source category. With respect to business agreements, 40 CFR part 63
includes a specific definition for a "facility." If contiguous plants are not under a common
ownership or control, then they are separate facilities, and if they are minor (area) sources for
HAP emissions, then they are not subject to the Refinery MACT 1 and 2. We do not have
specific plans to develop standards for area source petroleum refineries, but we may consider the
need for such standards in the future. For our refinery risk assessment, we conducted a facility-
wide risk assessment and did not necessarily limit the risk assessment to refinery MACT 1 and 2
emission sources, so co-located sources at a refinery facility were included in the risk
assessment.
2.1.2 Allowable emissions
Comment 1: The EPA received comments in support of and against the use of allowable
emissions in the risk assessment.
Those in support of the use of allowable emissions stated that facility emission estimates rely on
outdated emissions factors which undercount emissions. Another commenter concern was that
using actual emissions from a single point in time could underestimate risk from the source
category. Commenters also stated that the use of allowable emissions allows the EPA to consider
the potential health risks under the existing standards and added that using allowable emissions is
consistent with the basis of the major source HAP thresholds and the basis of issuing air quality
permits. Some of the commenters supporting the use of allowable emissions estimates, however,
did not agree with the use of the REM to develop the allowable emissions estimates asserting
that it is not representative of best available data.
On the other hand, those against the use of allowable emissions argued that the EPA's
consideration of "MACT allowable emissions" is arbitrary and not authorized under the law. The
commenter stated that section 112(f)(1)(A) requires EPA to report to Congress on "methods of
calculating the risk to public health remaining, or likely to remain, from sources subject to
regulation under this section after the application of standards under subsection (d) of this
section." Because a risk assessment based on allowable emissions would not reflect risk that
remains or is likely to remain, it is not permissible under the statute. Further, section 112(f)(1)(C)
requires the EPA also to report on "the actual health effects with respect to persons living in the
vicinity of' affected sources. These requirements clearly signal that Congress expected EPA to
focus on actual risk and not hypothetical risk in implementing the requirements of section 112(f).
Thus, it is not reasonable for EPA to construe section 112(f) as authorizing the Agency to
conduct risk assessments based on hypothetical "MACT allowable" emissions.
12
-------
The commenter added that EPA has failed to provide any reasoned explanation for why risk
assessments based on actual emissions estimates are inadequate, nor has the EPA provided data
or analyses indicating that its current methodology results in negative bias. Further, there is no
rational basis for EPA to conclude that allowable emissions are reasonably likely to occur,
making them irrelevant to its risk analysis. EPA has provided no data demonstrating that affected
sources actually do emit at levels significantly higher than the actual data show for any
significant period of time. Nor has EPA shown that emissions will systematically and
consistently increase across the industry such that allowable emissions would be a reasonable
basis for conducting a risk assessment. The commenter noted that there is already significant
conservatism in the EPA's risk assessment methodology, including the use of the REM, for
establishing health benchmarks and in the dispersion model.
Response 1: We evaluated both actual and allowable emissions and considered the risk
results using both of these emissions estimates to assess whether risks are acceptable. When
estimating the impacts of the proposed revisions, we generally assess those impacts based on the
actual emissions estimates, but the risk acceptability decisions consider both the actual
and allowable emissions.
We disagree with the commenter that suggests consideration of allowable emissions to be
unauthorized by the CAA. We interpret section 112(f)(1)(A) to require the EPA to assess
allowable emissions because these are the potential emissions and therefore the "risk to public
health remaining, or likely to remain, from sources subject to regulation under this section after
the application of standards under subsection (d) of this section." The record is replete with
evidence that emissions from refinery MACT sources may be understated. The fenceline
monitoring approach was proposed largely to address concerns that emissions, particularly from
fugitive sources (storage tanks, wastewater treatment, and equipment leaks), are difficult to
characterize and studies have shown measured emissions to be many times higher than inventory
reported values (see EPA Docket Item No. EPA-HQ-OAR-2010-0682-0210). As such, the
commenter is mistaken in their assertion that the EPA did not provide any data in the record to
suggest that the actual emissions may be understated and that an assessment of risk using
allowable emissions is reasonable.
Comment 2: Some commenters stated that the EPA's method to develop allowable emissions
estimates, including the use of the REM, is not representative of best available data and resulted
in understated emissions, and provided specific criticisms of the flare and wastewater allowable
emissions estimates.
Flares
The commenter stated that the EPA's estimate of allowable HAP emissions from flares
significantly underestimates exposure because the REM is based on the following incorrect
factually incorrect premises and thus the EPA must revise the REM.
The EPA incorrectly used permit limits exceeding the assumed 98% flaring efficiency for
facilities in Louisiana which biased the estimates for these facilities. Additionally,
commenters disagreed that 98% efficiency is appropriate for allowables as an EPA
13
-------
analysis of data provided by petroleum refiners concluded that the average flare achieves
approximately 93.9% destruction efficiency.
Some facilities calculated emissions based on the combustion of natural gas, not refinery
flare gas identifying the Murphy Oil permit as an example. The commenter further stated
that refinery flare gas has significantly higher HAP concentrations than natural gas (as
seen by the benzene concentration in the data the EPA collected from the passive Fourier
transform infrared spectroscopy (PFTIR) tests).
The REM used an emission factor of 0.14 pounds per million British thermal units
(mmBtu) of VOC for some facilities from AP-42. This emissions factor is four times
lower than EPA's proposed revision of the AP-42 emissions factor to 0.55 lb/mmBtu
VOCs. According to the commenter, the EPA must assume that benzene and organic
HAP releases were also under-estimated by the same amount and that the EPA's earlier
risk estimate must be increased by a factor of four.
The REM emission factor purports to account for all flaring emissions, but the underlying
permits and emission factors in the EPA's Emission Estimation Protocol for Petroleum
Refineries state that SSM events may not be incorporated. The EPA's attempt to account
for the uncertainty in the data by applying a multiplier of three to the modeled emissions
cannot correct for the structural errors in the EPA's model.
The commenter stated that the EPA's recent consent decrees recognize that flaring
emissions are proportional to the refinery's Nelson Complexity Index. The complexity of
the U.S. Refineries increased nearly 12% between 2002 and 2010, warranting a
corresponding 12% increase in flare emissions.
Wastewater
The commenter stated that the EPA's underlying assumptions about wastewater treatment
systems are factually incorrect and that the EPA has severely underestimated the allowable
emissions from petroleum refinery wastewater systems for the following reasons.
The EPA's 2011 ICR for the Petroleum Refining industry shows that the average
benzene concentration and wastewater production rates are much higher than the
assumptions used to estimate allowable emissions. The commenter noted that benzene
loading rates were derived from a 1998 document, Locating and Estimating Air
Emissions Sources of Benzene (L&E document), which used data from the EPA's ICR
for the original MACT 1 regulations for petroleum refineries issued in 1994.
The EPA uses two contradicting models to estimate the control efficiency for wastewater
treatment system pollution controls. To estimate allowable emissions, the EPA assumed
92% control efficiency, significantly higher than the average 86.5% control efficiency
assumed in the Technology Review for Industrial Wastewater Collections and Treatment
Operations at Petroleum Refineries. The commenter stated that EPA required five
facilities to conduct emissions testing to determine the control efficiency of their
biological treatment units and that EPA must use this data to inform its analysis of the
allowable emissions from wastewater treatment facilities.
The commenter stated that the EPA must address these problems, remodel the allowable
emissions from petroleum refineries, and then recalculate the risk posed by refineries.
14
-------
Response 2: We disagree with the commenter that the allowable emission estimates are
understated. We did note in the documentation of the allowable emissions estimates
that "allowable emissions" are very difficult to estimate from several of the refinery
emissions sources, including flares and wastewater.
Regarding our allowable emission estimates for flares, we based them on permit application
limits, which reflect the "allowable" emissions requested for these flares. The available
data supports the validity of the emission factors developed for the REM model. For example,
we have measurements of 1,3-butadiene in the flare plume from the PFTIR data. We estimated
1,3-butadiene emissions for four refinery flares at three refineries using the data set for which the
new VOC emissions factors were developed. The flare emissions factor for 1,3-butadiene
developed from these data yields an identical emissions factor as used in the REM model. This
suggests that the commenter's claim that HAP emissions should be 4 times higher based on the
new VOC factor is unfounded. It also suggests that concerns that refinery fuel gas contains more
HAP than natural gas may not be a significant issue either.
The commenter suggested that the PFTIR data shows more benzene in refinery fuel gas than in
natural gas. We note that there is typically very little benzene in refinery fuel gas (or other C6+
hydrocarbons, which are liquid at ambient temperatures). Only one facility measured benzene in
its refinery fuel gas. Assuming 98 percent combustion efficiency, the emission factor developed
for this unit would be an order of magnitude lower than the REM emission factor for benzene.
This again appears to support the relative accuracy and conservatism in the REM flare emissions
factors.
It is difficult to set a maximum allowable quantity of gas that can be flared, so this is the key
uncertainty in the analysis and this is why a factor of 3 was applied to the basic REM flare
estimates so as to account for these potentially higher volumes of gas that could be flared. We
note that the Greenhouse Gas Reporting Program (GHGRP) requires all refineries to report flare
emissions including emissions that occur from SSM events. Flare emissions have not increased
significantly over the 4 year reporting period, and the factor of 3 emissions is considered to be
sufficient to account for any increased flaring due to increases in refinery complexity.
We agree with commenters that studies have shown that many refinery flares are operating less
efficiently than 98 percent. Prior to proposing this rule, we conducted a flare ad hoc peer review
to advise our consideration of factors affecting flare performance (see discussion in the June 30,
2014, proposal at 70 FR 36905). However, we disagree with the commenters that the risk
analysis should consider this level of performance since the existing MACT standard does not
allow it. For purposes of the risk analysis, we evaluate whether it is necessary to tighten the
existing MACT standard in order to provide an ample margin of safety. Thus, in reviewing
whether the existing standards provide an ample margin of safety, we review the level of
emissions the MACT standards allow. In the present case, we considered the level of
performance assumed in establishing the MACT standard for purposes of determining whether
the MACT standard provides an ample margin of safety. However, we did recognize that
facilities were experiencing performance issues with flares and that many flares were not
meeting the performance level established at the time we promulgated the MACT standard.
Thus, we proposed, and are finalizing, revisions to the flare operating requirements to ensure that
15
-------
flares used as control devices meet the required performance level. These provisions are
consistent with the EPA's goals to improve the effectiveness of our rules.
Regarding the wastewater data in the ICR, in our initial review of these data, we determined that
there were discrepancies that made the data unusable. For example, some facilities reported
concentrations exceeding 1,000,000 parts per million by weight (ppmw), which is impossible.
We were unable to quality assure the data prior to the time we were obligated to issue our
proposed rule, and determined in light of these deficiencies it was more appropriate to utilize the
L&E document factors. The data collected from the ICR efforts has been reviewed in more detail
since proposal. As discussed in Chapter 9.2 of this response to comments (RTC), we performed
an analysis of the facilities that submitted a complete data set (meaning values for flow, benzene,
and total organic HAP), and calculated the average flow factor, flow-weighted average benzene
concentration, and flow-weighted average HAP concentration. We applied these factors to the
model plants used in the technology review and for 5 of the 6 model plants, the benzene loading
rates determined from the L&E document factors were higher than the loadings derived from the
2011 ICR factors. While there are still issues with the 2011 ICR data for wastewater, this
analysis indicates that the L&E document factors provided a reasonably conservative estimate
for the allowable emissions from petroleum refinery wastewater treatment systems. The analysis
of the ICR wastewater data has been added to the refinery docket EPA-HQ-OAR-2010-0682.
We also reviewed the ICR "source test" data for wastewater treatment systems. Based on these
data, the control efficiencies of the wastewater treatment systems that were evaluated exceeded
the 92 percent control efficiency used in the allowable emission estimate suggesting the
allowable emissions estimates were appropriately conservative. As described in further detail in
Section 9.2 of this document, we consistently used a control efficiency of approximately 92
percent for the enhanced biological treatment unit in both the allowables emissions estimate and
the technology reviews, which we note is different from the overall collection and treatment
system control efficiency. For some model plants in both the risk assessment and technology
review, wastewater collection system controls are not required and the emissions from the
collections system reduces the overall control efficiency of the system.
Comment 3: One commenter stated that EPA's inventory of hydrogen cyanide emissions
underestimates these emissions from FCCUs. The commenter noted that EPA used the same
factor to estimate "actual" and "allowable" emissions of HCN. The commenter attached a report
from Dr. Phyllis Fox that, according to the commenter, explains that both actual and allowable
emissions of HCN were underestimated for two main reasons:
The underlying FCCU stack test show that HCN emissions are highly variable between
process units. Based on only 9 stack tests, EPA found an HCN emission rate that ranged
from 114 to 22,100 lb/MMbbl. With such a small sample size and such a high level of
variability, EPA must use a conservative uncertainty factor to characterize "actual"
emissions from all 203 FCCUs in use at refineries.
The existing MACT standards do not regulate HCN emissions from the FCCUs, and the
existing carbon monoxide (CO) standard actually leads to higher HCN emissions as CO
is reduced.
16
-------
In light of these issues, the commenter stated that HCN emissions at many FCCUs could be
greater than the highest measured HCN emissions, from Hovensa, of 22,000 lb/MMbbl. The
resulting target organ-specific hazard index (TOSHI) from inhalation exposure to "MACT-
allowable" HCN emissions from FCCU's could be much greater than the TOSHI significance
threshold of 1.
To fulfill the agency's duty to assess health risks from all emitted pollutants, prevent all
unacceptable risk, and assure an "ample margin of safety to protect public health," as section
112(f)(2) requires, EPA must evaluate the HCN data and evidence provided in the Fox Report
and set standards that will assure limits on HCN for the first time. EPA must also do so, as
"developments" and ensure it is limiting all emitted HAPs as National Lime Association makes
clear that the CAA requires.
Response 3: We agree that the HCN emissions were highly variable and we adjusted reported
HCN emissions by a factor of 10 to account for the magnitude of emissions we saw in the stack
test data.5 Fortunately, we also had actual HCN stack test data for a number of FCCU and we
modeled these emissions directly. Based on our consideration of risks associated with HCN
emissions, we did not identify any facilities that had unacceptable HCN emissions.
We disagree that the existing CO limit leads to higher HCN emissions. As discussed in greater
detail in our response to comments regarding HCN emissions in the preamble to the final rule,
we consider the control strategy used by the best performing facilities is the use of complete
combustion as defined by reducing CO emissions to 500 parts per million by volume
(ppmv). Combustion theory suggests that partial combustion FCCU with no post-combustion
device would have true uncontrolled (high) HCN emissions. The only data we have available is
for facilities that have CO concentrations at or below 500 ppmv, so all of these emissions are for
controlled units. The poor correlation between CO and HCN emissions below CO levels of 500
ppmv merely confirms the idea that CO limits are needed. In any event, since all FCCU must
meet the 500 ppmv CO limit, the 9 source tests that are available should be representative of the
allowable emissions from controlled, complete combustion units. This, combined with our
augmentation of reported HCN emissions for risk modeling of FCCU HCN emissions, provides
us adequate confidence that the actual or allowable emissions for FCCU HCN do not result in
unacceptable risk.
2.1.3 Revisions submitted for the emissions data set
Comment 1: Commenters encouraged EPA to incorporate all corrections of the model input data
provided by facilities and to re-evaluate the source category risk. One commenter provided
suggested data revisions to the risk modeling file for 33 facilities. Seventeen facilities submitted
detailed revisions in the EPA data revisions spreadsheet tool.
Response 1: Due to the limited time and resources available for completing the final rulemaking,
the EPA was unable to remodel the entire source category risk. Additionally, information was
5 The Refinery Emission Protocol emission factor for HCN from FCCU was also revised by a factor of 10 in April,
2015.
17
-------
provided by a commenter for 33 facilities that was incomplete and would have required
additional follow-up with each of the facilities in order to revise the risk modeling file. However,
17 facilities provided revisions in the EPA data revisions tool. Of those 17 facilities, 13 of them
provided emission data revisions. EPA post processed the risk for four of these facilities because
they were the four highest MIR risk facilities. The fifth highest facility did not submit any
emission revisions. Since their MIR risk will remain the same and the top-four MIR values are
reduced below the fifth facility, the fifth facility will now be the new highest MIR facility. No
additional post-processing of the remaining nine facilities will change the overall risk picture. By
examining the specific pollutant contributions to the MIR we can adjust each pollutant risk by
the ratio of the revised emissions versus the ICR emissions. A revised facility risk is then
estimated by substituting the revised pollutant risks in place of the ICR pollutant modeled risks.
The approach assumes that the reductions are spread evenly across the sources emitting that
specific pollutant and that the facility maximum risks will still be located at the MIR location.
Other revisions submitted included deleting emission units that are no longer in service and
moving the coordinates of emission points to more accurately reflect their actual location in the
facility.
2.2 EPA overestimated human health risks
2.2.1 Modeling assumptions overestimate risk
Comment 1: Modeling assumptions are overly conservative and erroneous: One commenter
argued that EPA has overstated the risk associated with refinery emissions because EPA's
modeling assumptions are overly conservative. Regarding chronic risk estimates, the commenter
describes EPA's methodology including the MIR based on the unit risk estimate (URE) of each
HAP. The commenter points out that the URE is an upper bound estimate and may overestimate
both the individual risk levels and the total estimated number of cancer cases.
Regarding acute risk estimates, the commenter again described EPA's methodology and
underscores that the accuracy of the acute inhalation exposure assessment depends on the
simultaneous occurrence of independent factors that may vary greatly, such as hourly emission
rates, meteorology and human activity patterns. The commenter noted that EPA conservatively
assumes that individuals remain for 1 hour at the point of maximum ambient concentration. The
commenter further contends that EPA amplified its conservative approach, by claiming that the
maximum hourly emission rate data collected in the ICR was inadequate, and instead estimating
hourly emissions based on reported annual emissions, multiplied by "escalation factors" ranging
from 10 to 60 times the annual emissions. According to the commenter, this the escalation factor
methodology reflects EPA's arbitrary and excessively conservative attempt to assign an emission
variability factor to each source type, which deviates from past RTR rulemakings.
Response 1: Uncertainty and the potential for bias are inherent in all risk assessments, including
those performed for this rulemaking. Although uncertainty exists, our approach, which uses
conservative tools and assumptions, ensures that our decisions are appropriately health protective
and environmentally protective. A brief discussion of the uncertainties in the RTR emissions
datasets, dispersion modeling, inhalation exposure estimates and dose-response relationships can
18
-------
be found in the proposal for this rulemaking [79 FR 36879], Regarding the comment that we
amplified our conservative approach by using refinery-specific hourly emission rate factors to
estimate acute risks, we note that we derived process specific factors based on our knowledge of
refinery processes. We intended to refine the default factor of 10 that is applied to the average
hourly rate produced by dividing annual emissions by annual hours (8760 hours) that has been
used in many other source categories to produce a more realistic value, not to add to the
conservancy of the analysis. We did not, as the commenter asserts, use values from 10 to 60 to
predict hourly rates. In fact, for many processes, our refinery specific factor was much less than
10; for example, we applied hourly multipliers of 2 to estimate maximum hourly rates for
fugitives, FCCU, SRU, chemical processes and miscellaneous process vents. For storage and
wastewater emissions sources, we used a value of 4; it was only in the case of cyclic and semi-
regenerative catalytic reformers, and delayed cokers that we applied a factor greater than 10; in
these processes, emissions, as reported on an annual basis, only occur for very limited periods of
time. We used the number of hours in the venting cycle and the variability of emissions expected
over the cycle to develop a factor specific to each type of process. For example, semi-
regenerative reformers only vent a few weeks a year during regeneration. Our factor of 60 was
intended to account for the period of time that they vent, and for variability over the venting
period. Therefore, we do not agree that our approach was arbitrary or excessively conservative as
it considers the variability of various types of emissions sources in petroleum refining and it is
based on the actual venting characteristics of these sources. As we discussed in the proposal, we
elected not to simply model reported hourly emissions as these values were often left blank or
appeared to be reported in units other than those asked for in the ICR and we were not confident
in the quality of these data. A more thorough discussion of these uncertainties is included in the
Final Residual Risk Assessment for the Petroleum Refining Source Sector in support of the
September 2015 Risk and Technology Review Final Rule, in Docket No. EPA-HQ-OAR-2010-
0682 .
Comment 2: Unit risk estimates are too conservative: One commenter expressed concerns with
EPA's assumptions concerning benzene, naphthalene, and other polycyclic organic matter
(POM) unit risk values, and stated that the unit risk factors are extremely conservative adding to
the conservatism of the risk analysis. The commenter claimed that the EPA singled out benzene
and naphthalene by attributing to them 98% of the MIR resulting from equipment leaks and
storage tanks.
Regarding benzene, the commenter stated that the inhalation unit risk value developed under
EPA's IRIS program is a range of values (2.2 x 10"6 to 7.8 x 10"6) and that EPA has gone on
record stating that each estimate has equal scientific plausibility6. Several commenters asserted
that EPA's risk assessment utilizes the upper end of this range, biasing towards a greater cancer
risk for each micrograms (jug)/ cubic meters (m3) of exposure to this substance, rather than
utilizing the entire range. If the lower end of the range were used, commenters stated that the
resulting cancer risk estimates may have been 3-4 times lower than those shown in the report.
One of the commenters noted that EPA often defaults to a more conservative approach when
estimating risk, therefore utilizing the conservative upper end toxicity value; however, it would
be beneficial to see a balanced approach, demonstrating the range of potential cancer risk,
6 http://www.epa.gov/iris/subst/0276.htm
19
-------
especially when actions are being required by the refinery sector on the basis of the risk
calculations.
Regarding naphthalene, one commenter stated that for several years, industry and EPA have
been working together to resolve the question of whether naphthalene has been inappropriately
classified as a possible or reasonably anticipated human carcinogen. The commenter asserted
that EPA's analysis of the possible risk associated with exposure to naphthalene does not take
into account the most recent scientific information. According to the commenter, the naphthalene
URE used in EPA's risk analysis originates from California EPA's Office of Environmental
Health Hazard Assessment (OEHHA), which was established based upon National Toxicology
Program (NTP) studies. The commenter noted that EPA dropped its own URE value that same
year, citing a lack of evidence on naphthalene carcinogenicity in humans, and furthermore that
current research by the National Research Council (NRC) has called into question the NTP
studies (in which the animals were dosed at concentrations exceeding the Maximum Tolerated
Dose resulting in inflammation incidence near 100%, suggesting that cytotoxicity played a
significant role in the tumor responses [NRC comments submitted to EPA's IRIS Docket; August
20, 2014]). The commenter contended that the evidence used to support the URE adopted by
OEHHA is irrelevant to actual human risk, that the California EPA URE is outdated and that
therefore the validity of classifying naphthalene as a human carcinogen is highly questionable.
The commenter also points out that, as a result of the NRC work, EPA is performing another
IRIS program review of naphthalene carcinogenicity and derivation of cancer risk, which is
delayed, but which will consider all the experimental evidence on naphthalene published over the
past several years. Many commenters concluded that, in light of the latest experimental evidence
on naphthalene, use of the OEHHA URE is overly conservative and mischaracterizes the human
risk posed by naphthalene. Specifically, the conclusion that naphthalene carries more weight than
benzene (22% vs 21%) in driving the cancer risk should be reviewed, according to the
commenters. One commenter asserted that, at a minimum, recognition of this uncertainty and the
work underway to resolve this uncertainty should have been addressed in the preamble, in an
effort to provide a balanced representation of the potential risk attributed to naphthalene.
Response 2: EPA's chemical-specific toxicity values are derived using risk assessment
guidelines and approaches that are well established and vetted through the scientific community,
and follow rigorous peer review processes.7 The RTR program gives preference to EPA values
for use in risk assessments and uses other values, as appropriate, when those values are derived
with methods and peer review processes consistent with those followed by the EPA. The
approach for selecting appropriate toxicity values for use in the RTR program has been endorsed
by the Science Advisory Board.8
The commenter is correct that for benzene the high end of the reported cancer URE range (7.8E-
06 per |ig/m3) was used in our assessments to provide a conservative estimate of potential cancer
risks. Use of the high end of the range provides risk estimates that are approximately 3.5 times
7 Integrated Risk Information System (IRIS). IRIS Guidance document available at
http://www.epa.gov/iris/backgrd.html
8 http://yosemite.epa.gOv/sab/sabproduct.nsf/0/b03 lddf79cffded38525734f00649caf!OpenDocument&TableRow
=2.3#2
20
-------
higher than use of the equally plausible low end value (2.2E-06 per |ig/m3). If the estimated
benzene-associated risks exceed 1-in-1-million, we would also evaluate the impact of using the
low end of the URE range on our risk results. For the petroleum refineries source categories, the
maximum individual risk for all carcinogens was estimated to be up to 60-in-l-million with
benzene contributing 21% to the cancer incidence in exposed population. Estimating the cancer
risk using the low end URE value for benzene would change the maximum individual risk to be
up to 50-in-l-million for these source categories (i.e., driven by pollutants in addition to
benzene) and would reduce the incidence by less than 6%, which would not impact our risk-
based decisions. We have updated the risk assessment report for the final rulemaking to provide
this information.
The commenter is correct that the EPA used the unit risk factor for naphthalene derived by
OEHHA, which is considered by the EPA to be the best available value and most appropriate for
use in RTR assessments. The derivation of OEHHA's naphthalene unit risk is based on a well
conducted 2 year NTP bioassay showing clear evidence of carcinogenic activity in rats. We
disagree with the commenter that the EPA dropped its own URE citing a lack of evidence on
naphthalene carcinogenicity in humans. EPA did not drop the URE from the naphthalene IRIS
assessment. This assessment, which was last revised in 1998, does not include a URE. Based on
the 1996 Proposed Guidelines for Carcinogen Risk Assessment, the EPA determined that the
human carcinogenic potential of naphthalene via the oral or inhalation routes "cannot be
determined" at this time based on human and animal data; however, the naphthalene assessment
also states that there is suggestive evidence (observations of benign respiratory tumors and one
carcinoma in female mice only exposed to naphthalene by inhalation).9
Several commenters noted the ongoing assessment of naphthalene in the IRIS Program, which
will consider the more recent studies sponsored by the Naphthalene Research Council, and
suggested consideration of those studies alone as a reason to add greater uncertainty to the
current cancer risk conclusions regarding naphthalene. EPA will not prejudge or presuppose the
outcome of that ongoing assessment, or the relative merits of any particular set of studies; to do
so would be premature and counterproductive to the established IRIS Process. EPA will continue
to use the current cancer assessment for naphthalene.
2.2.2 Risk assessment is too broad in scope
Comment 1: Facility-wide emissions and risks should not be considered: One commenter stated
that EPA is not authorized to consider total facility emissions in conducting risk assessments for
particular source categories According to the commenter, the CAA requires EPA to conduct
residual risk determinations on a category-by-category basis, citing the substance and timing
required in section 112(f), in particular language in section 112(f)(2)(A) referring to "emissions
from a source in the category or subcategory. " The commenter asserted that nothing in the text
9 National Toxicology Program (NTP). (1992) Technical Report on the Toxicology and Carcinogenesis Studies of
Naphthalene (CAS No. 91-20-3) inB6C3Fl Mice. (Inhalation Studies). DHHS, PHS, NIH, Rockville, MD.
https://ntp.niehs.nih.gov/ntp/htdocs/lt rpts/tr410.pdf
21
-------
of the statute indicates that EPA may combine multiple source categories into a single risk
assessment and therefore EPA does not have the authority to consider emissions from any
sources other than those in the source category or subcategory under review at that time. The
commenter further argued that it is not reasonable to construe these provisions as authorizing
EPA to consider emissions from entire facilities in conducting risk assessments, because
Congress clearly envisioned that full implementation of the MACT program would take longer
than eight years (which, according to the commenter, would be impossible for facilities that
contain sources in a category where the eight-year deadline precedes the adoption of MACT
standards for other sources at these facilities). For these reasons, the commenter concluded there
is no practical way to use combined source categories for purposes of conducting the risk
assessment, and thus the statute cannot reasonably be construed as authorizing EPA to do so.
Response 1: We disagree that examining facility-wide risk in a risk assessment conducted under
section 112(f) exceeds the EPA's authority. The development of facility-wide risk estimates
provides additional information about the potential cumulative risks in the vicinity of the RTR
sources, as one means of informing the ample margin of safety step of our risk analysis for the
RTR source categories in question.
Section 112(f)(2) of the CAA expressly preserves our use of the two-step process for developing
standards to address residual risk and interpret "acceptable risk" and "ample margin of safety" as
developed in the Benzene NESHAP (54 FR 38044, September 14, 1989). In the Benzene
NESHAP, the EPA rejected approaches that would have mandated consideration of background
levels of pollution in assessing the acceptability of risk, concluding that"... comparison of
acceptable risk should not be associated with levels in polluted urban air. With respect to
considering other sources of risk from benzene exposure and determining the acceptable risk
level for all exposures to benzene, EPA considers this inappropriate because only the risk
associated with the emissions under consideration are relevant to the regulation being established
and, consequently, the decision being made." (54 FR 38044, 38061, September 14, 1989).
Although not appropriate for consideration in the determination of acceptable risk, we note that
background risks or contributions to risk from sources outside the source category under review
could be one of the relevant factors considered in the ample margin of safety determination,
along with cost and economic factors, technological feasibility, and other factors. Background
risks and contributions to risk from sources outside the facilities under review were not
considered in the ample margin of safety determination for these source categories, mainly
because of the significant uncertainties associated with emissions estimates for such sources. Our
approach here is consistent with the approach we took regarding this issue in the Hazardous
Organic NESHAP (HON) RTR (71 FR 76603, December 21, 2006), which the court upheld in
the face of claims that the EPA had not adequately considered background (NRDC v. EPA, 529
F.3d 1077 (D C. Cir. 2008)).
22
-------
2.3 EPA underestimated human health risks
2.3.1 Modeling methodology understates risks
Comment 1: Deposition methodology understates metal toxicity: One commenter stated that
particulates have an inversely proportionate bonding to particle size, with 99% of toxic metal
gases like lead and mercury in an effluent stack adhered to particles less than 2 microns in size.
According to the commenter, the deposition models attempt to show the general location where
something that "went up" actually "comes down" as a deposited particle. The commenter further
explained that, in these computer models, a "reflection coefficient" is a user-adjustable factor
that "floats" very small particles and with artificially high values, small particles that go up do
not come down. The commenter argued that the use of this reflection coefficient could enable
modeling demonstrations to hide the bulk of the metal toxicity and requested that the EPA
provide valid guidance for the use of these types of coefficients to ensure the models will to
accurately project the real health impacts on the public.
Response 1: EPA disagrees with the commenter that the modeling methodology employed for
the analysis underestimates risks. The commenter is concerned that the mass of the plume is not
being made available for exposure at receptor locations at the surface. In determining downwind
ambient concentrations the HEM-3 model treats each emitted pollutant (including metals) as a
gaseous pollutant and thus does not account for potential removal of mass in the particle phase
by contacting objects on the surface or settling. By modeling the pollutant only as a gas, the
plume is brought to the ground for potential exposure at receptor locations by simple dispersion
(movement of the plume by the wind and atmospheric stability). Although the emitted metals are
particulates, at such low ambient concentrations the plume will behave more like a gas and EPA
believes this approach will most realistically simulate the path of the plume as it travels
downwind from the facility. This approach, in some situations, may result in an over-
prediction of ambient concentrations (i.e., health protective).
Comment 2: Centroid location representing exposed population understates risk: Two
commenters objected to EPA's estimation of chronic exposure at the census block centroid
instead of at the facility property line or location of the maximum exposed individual.
Commenters noted that census blocks can be geographically large and the population distribution
within the blocks are not necessarily homogenous and thus this approach understates risk.
Furthermore, commenters stated that, except for lead, EPA made no effort to move receptor
points closer to the facility to assess chronic or cancer risk, even where local residents live nearer
to a facility than the census block centroid. According to the commenter, this conflicts with the
recommendation of the Science Advisory Board (SAB), which has urged EPA to consider
"specific locations of residences" [SAB May 2010], The commenter stated that taking
geographic variation out of the equation fails to properly account for exposure to the "individual
most exposed to emissions" as required by section 112(f)(2)(A), and fails to provide an accurate
estimate of risk. The commenter stated that EPA's failure to adjust receptor points for residents
living on the fenceline is particularly inexcusable given that the HEM-AERMOD system allows
for such an adjustment, and that such an adjustment was appropriately made for the estimation of
acute health risks [79 FR 36890], The commenter concluded that, having recognized that the
23
-------
maximum exposed individual for acute risks is likely present at the fence-line, EPA cannot
justify failing to analyze cancer and other chronic health effects in a similar manner.
Response 2: In a national-scale assessment of lifetime inhalation exposures and health risks
from facilities in a source category, it is appropriate to identify exposure locations where it may
be reasonably expected that an individual will spend a majority of his or her lifetime. In
determining chronic risks, it is appropriate to use census block information on where people
actually reside, rather than points at the property line, to locate the estimation of exposures and
risks to individuals living near such facilities. Census blocks are the finest resolution available as
part of the nationwide population data (as developed by the US Census Bureau); on average, a
census block is comprised of approximately 40 people and about 10 households. In the EPA risk
assessments, the geographic centroid of each census block containing at least one person is used
to represent the location where all the people in that census block live. The census block centroid
with the highest estimated exposure then becomes the location of maximum exposure, and the
entire population of that census block is assumed to experience the maximum individual risk. In
some cases, because actual residence locations may be closer to or farther from facility emission
points, this may result in an overestimate or underestimate of the actual annual concentrations
(although there is no systematic bias for average levels). Given the relatively small dimensions
of census block in densely-populated areas, there is little uncertainty introduced by using the
census block centroids in lieu of actual residence locations. There is the potential for more
uncertainty when the census block are larger, although there is still no systematic bias. The EPA
concludes that the most appropriate locations at which to estimate chronic exposures and risks
are the census block centroids because: 1) census blocks are the finest resolution available in the
national census data; 2) facility fence lines do not typically represent locations where chronic
exposures are likely (i.e., people do not typically live at the fence line of facilities); and 3) there
is no bias introduced into the estimate of MIR by using census block centroid locations. In its
peer review of the methodologies used to estimate risks as part of the RTR rulemaking efforts,
the EPA's SAB endorsed this approach.10
In addition to the approach described above, the EPA recognizes that where a census block
centroid is located on industrial property or is large and the centroid is less likely to be
representative of the block's residential locations, the block centroid may not be the appropriate
surrogate. For these source categories, as described in the Draft Residual Risk Assessment for
the Petroleum Refining Source Sector (May 2014) in Docket Item No. EPA-HQ-OAR-2010-
0682-0225, in cases where a census block centroid was within 300 meters of any emission source
(and therefore possibly on facility property), we viewed aerial images of the facility to determine
whether the block centroid was likely located on facility property. Likewise, we examined aerial
image of all large census blocks within one kilometer of any emission source. If the block
10 U.S. Environmental Protection Agency, Office of the Administrator Science Advisory Board.
Review of EPA's draft entitled, "Risk and Technology Review (RTR) Risk Assessment Methodologies: For
Review by the EPA's Science Advisory Board with Case Studies - MACTI Petroleum Refining Sources and
Portland Cement Manufacturing." EPA-SAB-10-007. May 7, 2010.
http://vosemite.epa. gov/sab/sabproduct.nsf/4AB3966E263D943A8525771F00668381/$File/EPA-SAB-10-007-
unsigned.pdf.
24
-------
centroid did not represent the residential locations within that block, we relocated it to better
represent them and/or we added receptors for residences nearer to the facility than the centroid.
For these source categories, we added several receptors for census blocks where the centroid
location was not representative of the residential locations. Appendix 7 of the risk assessment
document cited above provides additional information on these changes.
2.3.2 Cancer unit risk estimates, non-cancer reference doses and acute benchmarks
used by EPA understate risks
Comment 1: Cancer unit risk estimates and chronic non-cancer reference concentrations used
are not protective or based on the most up-to-date science: One commenter stated that EPA
failed to use the best available reference values for a number of key HAPs including benzene,
and thus underestimated risk from these pollutants.
Regarding cancer risk, the commenter recommended using the cancer potency values as
published by the California EPA (CalEPA). The commenter performed a comparison of the
CalEPA values and those used by the EPA and claimed for some HAP (including 1,3 butadiene,
benzene, Cadmium Compounds, and Chromium VI Compounds) the UREs are at least 57%
lower. The commenter also stated an emissions weighted analysis of the different potency values
used by EPA versus those recommended by CalEPA shows that cancer risk may have been two
times higher had EPA utilized up-to-date factors. Similarly, another commenter stated that
California's OEHHA has developed a more stringent cancer inhalation unit risk value for
benzene and if this number had been used with the proposed 9 |ig/m3 fenceline action level, the
lifetime risk would be close to 260-in-l million.
Regarding non-cancer chronic risk, the commenter also recommended that the EPA use the
chronic inhalation reference concentrations (RfC) published by California EPA. The commenter
claimed, that EPA's RfC for some HAP (including benzene, toluene, manganese, mercury
(elemental), and nickel) are less protective than those published by CalEPA by at least a factor of
3. Likewise, another commenter stated that California's OEHHA finalized new and revised
benzene and nickel reference exposure levels (REL) which are more stringent than those used in
the residual risk assessment. One commenter further explained that the new OEHHA acute REL
for benzene is 27 |ig/m3, is nearly equal to the value EPA used for the chronic RfC. The
commenter noted that the EPA proposed a fenceline action level of 9 |ig/m3 would exceed the
California chronic REL by a factor of three.
Regarding risks from lead exposure, one commenter stated that EPA should utilize the best
science on assessing children's health risk of exposure to lead, rather than consider only the 2008
Lead National Ambient Air Quality Standards (NAAQS). The commenter stated that EPA must
address and incorporate the best currently available information on children's exposure,
including the Children's Health Protection Advisory Committee (CHPAC) recommendation11 of
lowering the lead standards to 0.02 |ig/m3 from the current EPA NAAQS level of 0.15 |ig/m3.
11 Letter from Dr. Melanie A. Marty, Chair, Children's Health Protection Advisory Comm., to Administrator
Stephen L. Johnson, (June 16, 2008), http://www2.epa.gov/sites/production/files/2014-05/documents/61608.pdf
25
-------
According to the commenter, the Centers for Disease Control (CDC) has now recognized that
there is no safe level of exposure, and has replaced the now outdated 10 |ig/dL standard with a
recognition that action is required at the reference level of 5 |ig/deci liters (dL). 12The commenter
concluded that because EPA relies on the lead NAAQS in its proposed residual risk rule, EPA
has not met the legal standard of section 112(f)(2), noting that the residual risk standards are
designed to do more than just replicate other statutory protections, such as those provided by the
NAAQS. The commenter argued that, if Congress had intended EPA simply to replicate the
NAAQS or some other different CAA requirement in its section 112(f)(2) residual risk
rulemaking, the section 112(f)(2) requirement would become redundant for any hazardous air
pollutant that also has any relationship to any other regulated pollutant. Furthermore, according
to the commenter, for any ambiguity on this question, statutory construction requires a reading of
section 112(f)(2) that preserves its independent value and meaning.
Response 1: EPA's chemical-specific toxicity values are derived using risk assessment
guidelines and approaches that are well established and vetted through the scientific community,
and follow rigorous peer review processes.13 The RTR program gives preference to EPA values
for use in risk assessments and uses other values, as appropriate, when those values are derived
with methods and peer review processes consistent with those followed by the EPA. The
approach for selecting appropriate toxicity values for use in the RTR Program has been endorsed
by the SAB.14
The commenter recommended that EPA use California OEHHA's new toxicity values for several
chemicals and provided some references for the approaches used to derive those values. The
EPA scientists reviewed the information provided by the commenter regarding the California
values and concluded that further information is needed to evaluate the scientific basis and
rationale for the recent changes in California OEHHA risk assessment methods. The EPA will
work on gathering the necessary information to conduct an evaluation of the scientific merit and
the appropriateness of the use of California OEHHA's new toxicity values in the agency
decisions. Until the EPA has completed its evaluation, it is premature to determine what role
these values might play in the RTR process. Therefore, EPA did not use the new California
OEHHA toxicity values as part of this current rulemaking.
We disagree with the comment that the use of the lead NAAQS does not sufficiently protect
children. While recognizing that lead has been demonstrated to exert "a broad array of
deleterious effects on multiple organ systems," the lead NAAQS targets the effects associated
with relatively lower exposures and associated blood lead levels, specifically nervous system
effects in children including cognitive and neurobehavioral effects (73 FR 66976). EPA
establishes the NAAQS at a level to protect sensitive sub-populations, such as children and
12 CDC, What do Parents Need to Know to Protect Their Children?,
http://www.cdc.gov/nceh/lead/acclpp/blood_lead_levels.htm
13 Integrated Risk Information System (IRIS). IRIS Guidance documents available at
http://www.epa.gov/iris/backgrd.html
14http://yosemite.epa.gov/sab/sabproduct.nsf/0/b031ddf79cffded38525734f00649caf!OpenDocument&TableRow=2.
3 #2
26
-------
pregnant women. The 2008 decision on the lead NAAQS was informed by an evidence-based
framework for neurocognitive effects in young children. In applying the evidence-based
framework, we focused on a subpopulation of U.S. children, those living near air sources and
more likely to be exposed at the level of the standard; to the same effect see 73 FR 67000/3
"The framework in effect focuses on the sensitive subpopulation that is the group of children
living near sources and more likely to be exposed at the level of the standard. The evidence-
based framework estimates a mean air-related intelligence quotient (IQ) loss for this
subpopulation of children; it does not estimate a mean for all U.S. children"; 73 FR 67005/1
"the air-related IQ loss framework provides estimates for the mean air-related IQ loss of a subset
of the population of U.S. children, and there are uncertainties associated with those estimates. It
provides estimates for that subset of children likely to be exposed to the level of the standard,
which is generally expected to be the subpopulation of children living near sources who are
likely to be most highly exposed." In addition, in reviewing and sustaining the lead primary
NAAQS, we note that the D.C. Circuit specifically noted that the rule was targeted to protect
children living near lead sources: "EPA explained that the scientific evidence showing the
impact of lead exposure in young children in the United States led it 'to give greater prominence
to children as the sensitive subpopulation in this review' and to focus its revision of the lead
NAAQS on the 'sensitive subpopulation that is the group of children living near [lead emission]
sources and more likely to be exposed at the level of the standard.' Given the scientific evidence
on which it relied, the EPA's decision to base the revised lead NAAQS on protecting the subset
of children likely to be exposed to airborne lead at the level of the standard was not arbitrary or
capricious." Coalition of Battery Recyclers, 604 F. 3d at 618.
As noted in the risk assessment document, there is no RfD or other comparable chronic health
benchmark value for lead compounds. In 1988, the EPA's IRIS program reviewed the health
effects data regarding lead and its inorganic compounds and determined that it would be
inappropriate to develop an reference dose (RfD) for these compounds, stating:
A great deal of information on the health effects of lead has been obtained through
decades of medical observation and scientific research. This information has been
assessed in the development of air and water quality criteria by the Agency's Office of
Health and Environmental Assessment (OHEA) in support of regulatory decision-making
by the Office of Air Quality Planning and Standards (OAQPS) and by the Office of
Drinking Water (ODW). By comparison to most other environmental toxicants, the
degree of uncertainty about the health effects of lead is quite low. It appears that some of
these effects, particularly changes in the levels of certain blood enzymes and in aspects of
children's neurobehavioral development, may occur at blood lead levels so low as to be
essentially without a threshold. The Agency's RfD Work Group discussed inorganic lead
(and lead compounds) at two meetings (07/08/1985 and 07/22/1985) and considered it
inappropriate to develop an RfD for inorganic lead.
The EPA's IRIS assessment for lead and compounds (inorganic) (c 7439-92-1) can be found at
http://www.epa.gov/iris/subst/0277.htm.
With regard to the information identified by the commenters, much of this information was
similar to information available at the time of the 2008 NAAQS decision. For example, in 2005,
27
-------
the CDC recognized the evidence of adverse health effects in children with blood lead levels
below 10 |ig/dL, and that there is no safe level of blood lead in young children (CDC, 2005). The
commenter also cites a benchmark analysis by California OEHHA that was completed during the
time of the last lead NAAQS review (Carlisle and Dowling, 2007). The quantitative relationship
from this analysis is a correlation of 1 IQ point change with a 1.0 |ig/dL change in blood lead is
actually a substantially smaller change in IQ per |ig/dL blood lead than the slope of 1.75 IQ
points per |ig/dL blood lead used in the evidence-based framework that the Administrator relied
upon in his 2008 decision on a revised level for the lead NAAQS in 2008 (73 FR 66964).
Regarding the CHPAC, a recommendation on the level and averaging time for the revised
NAAQS referenced by the commenter was made to EPA in January of this year in the context of
the current NAAQS review and the same comment was made and considered in the 2008 review,
that concluded with the current lead NAAQS.
The commenter also makes the legal argument that the primary NAAQS provides an "adequate"
margin of safety, but that section 112 (f)(2) requires that a residual risk standard provide an
"ample" margin of safety, reasoning from this that a NAAQS cannot just substitute as the
measure for evaluating acceptability of risk and that some greater level of protection is required.
We do not accept the commenter's argument. The EPA is considering the primary NAAQS for
lead which incorporates an adequate margin of safety in determining whether risks (taken
together with cancer and other non-cancer health risks) from air-borne lead from petroleum
refinery facilities are acceptable or unacceptable. Thus, to the extent the commenter's argument
rests on the difference between 'adequate' and 'ample' margin of safety, the argument is
misplaced. Margin-of-safety determinations for this rule are conducted separately, in accord with
the two-step framework set forth in the Benzene NESHAP and the en banc opinion in Vinyl
Chloride. See not only Vinyl Chloride. 824 F. 2d at 1165, 1166 but NRDC v. EPA, 902 F. 2d
962, 973-74 (D.C. Cir. 1990) (distinguishing the NAAQS process, whereby the margin of safety
analysis is incorporated as part of the standard without a two-step analysis, from residual risk
determinations).15 Using that framework, with its consideration of costs, cost-effectiveness,
technological feasibility, and other factors set out in the Benzene NESHAP, at proposal we did
not identify any additional controls beyond those that would need to be implemented to ensure
an acceptable level of risk with an ample margin of safety. The EPA thus disagrees with the
commenter that section 112 (f)(2) standards must be more stringent than a primary NAAQS as a
matter of law.
Comment 2: Acute benchmark concentrations used are not protective or based on the most up-
to-date science: A commenter claimed that the acute reference value used to evaluate benzene
exposure is two orders of magnitude too high, leading to a very significant underestimation of
the acute non-cancer health hazards of benzene emissions from refineries. Two commenters
noted that the EPA used an acute REL of 1.3 mg/m3, while California EPA's OEHHA uses a
15 The court was referring to the predecessor provision to the current section 112 (f), but its
analysis is equally applicable to the revised provision.
28
-------
value of 0.027 milligrams (mg)/m3 based on the latest science.16 One commenter explained that
since developmental toxicity may occur in response to just one exposure during a specific
window of susceptibility, the acute REL is intended to be health-protective with respect to
infrequent one-hour exposures.
Notwithstanding EPA's decision not to use the California OEHHA benzene REL, one
commenter expressed gratification to see that EPA has increased its reliance on the California
RELs to address acute exposures in the residual risk assessments and urged EPA to continue to
use the RELs for these assessments.
Another commenter stated that they have expressed concerns in the past with EPA's use of
AEGL or Emergency Response Planning Guidelines (ERPG) values to address acute exposures
in the residual risk assessments. The commenter asserted that these limits were developed for
accident release emergency planning and are not appropriate for assessing daily human exposure
scenarios (and also cited EPA's December 2002 document, "A Review of the Reference Dose
and Reference Concentration Processes," in which EPA stated that the primary purpose of the
AEGL program is to develop guidelines for once-in-a-lifetime short-term exposures to airborne
concentrations of acutely toxic chemicals). The commenter argued therefore that the AEGL
values are not meant to evaluate the acute impacts from routine emissions that occur over the life
of a facility and, unlike the RfCs for chronic exposures, the AEGLs and ERPGs do not include
adequate safety and uncertainty factors that can be relied upon to protect the public from the
adverse effects of exposure to toxic air pollutants. The commenter concluded that the use of
AEGLs or ERPGs in residual risk assessments is not appropriate and does not ensure that public
health is adequately protected from the acute impacts of HAP exposure. Likewise, according to
one commenter, EPA recognizes that many pollutants creating acute risks are pollutants for
which it has no reference value [Draft Risk Assessment, tbl. 2.6-3 at 27-28.] and for that reason,
the EPA looks at inappropriate values that are not health-protective (i.e., the AEGLs and ERPGs)
designed only for emergency exposure response.17
16 CalEPA Table of Reference Exposure Levels, Guidelines for conducting health risk assessments under the Air
Toxics Hot Spots Program, May 2014, http://www.oehha.ca.gov/air/allrels.html; EPA Dose-Response Values for
Chronic Inhalation Exposure, as reported in the Draft Risk Assessment for the Petroleum Refining Source Sector
(Doc. ID -0225), Table 2.6-2.
17 The AEGL values (and Emergency Response Planning Guidelines (ERPG) values, which EPA also should not
use) were created for emergency exposure scenarios. Levels defined for "once-in-a-lifetime, short-term exposures"
and "emergency" chemical releases or accidents, 76 F R at 52,772, are not appropriate tools to measure long-term,
lifetime acute exposure risk. As the Science Advisory Board has explained: The incorporation of the available
California Reference Exposure Levels (RELs) for the assessment of acute effects is a conservative and acceptable
approach to characterize acute risks. . . . The Panel has some concern with the use of the Acute Exposure Guidelines
Limits (AEGLs) and Emergency Response Planning Guidelines (ERPGs).... AEGL-2 and ERPG-2 values should
never be used in residual risk assessments because they represent levels that if exceeded could cause serious or
irreversible health effects. Sci. Adv. Bd., Review of EPA's draft entitled, "Risk and Technology Review (RTR) Risk
Assessment Methodologies: For Review by the EPA's Science Advisory Board with Case Studies - MACTI
Petroleum Refining Sources and Portland Cement Manufacturing," EPA-SAB-10-007 at 6 (May 07, 2010)) ("SAB
May 2010"), EPA-HQ-OAR-2010-0682-0103 (emphasis added). The AEGL and ERPG numbers would be expected
to underestimate risk. Using these numbers is likely to discount or cloak the level of risk to the maximum exposed
29
-------
Response 2: We disagree with the comment that the EPA should adopt the California OEHHA
acute REL for benzene. The EPA has an approach for selecting appropriate health benchmark
values and in general, this approach places greater weight on the EPA derived health benchmarks
than those from other agencies. The approach favoring EPA benchmarks (when they exist) has
been endorsed by the SAB,18 and ensures values most consistent with well-established and
scientifically-based EPA policy. The EPA is currently evaluating the most appropriate use for the
California OEHHA derived reference doses. This evaluation on the appropriateness of these
values in the context of the RTR Program and EPA science policy is necessary prior to using
these references doses.
The EPA does not rely exclusively upon AEGL or ERPG values for assessment of acute
exposures. Rather, the EPA's approach is to consider various acute health effect reference values
(see 79 FR 37857), including the California REL, in assessing the potential for risks from acute
exposures. To better characterize the potential health risks associated with estimated acute
exposures to HAP, and in response to a key recommendation from the SAB's peer review of the
EPA's RTR risk assessment methodologies,19 we generally examine a wider range of available
acute health metrics (e.g., RELs, AEGLs) than we do for our chronic risk assessments. This is in
response to the SAB's acknowledgement that there are generally more data gaps and
inconsistencies in acute reference values than there are in chronic reference values. In some
cases, when Reference Value Arrays20 for HAP have been developed, we consider additional
acute values (i.e., occupational and international values) to provide a more complete risk
characterization. As discussed in the preamble to the proposed rule (79 FR 37857), the exposure
guidelines EPA considers depends on which exposure guidelines are available for the various
hazardous air pollutants emitted. The EPA uses AEGL and ERPG values (when available) in
conjunction with REL values (again, when available) to characterize potential acute health risks.
However, it is often the case that HAP do not have all of these acute reference benchmark
values. In these instances, the EPA describes the potential acute health risk in relation to the
acute health values that are available. Importantly, when interpreting the results, we are careful
to identify the benchmark being used and the health implications associated with any specific
benchmark being exceeded.
Comment 3: Non-cancer reference values should assume no safe level of exposure, similar to
cancer risk values, to avoid understating risks from non-carcinogens: One commenter asserted
that the EPA must recognize that chronic (non-cancer) risk-causing pollutants have no safe level
individual. These values are therefore not appropriate for rely on as health-protective in a section 112(f)(2) residual
risk analysis. They simply do not provide sufficient protection for health.
18 The SAB peer review of RTR Risk Assessment Methodologies is available at:
http://voscmitc.cpa.aov/sab/sabproduct. as 174 AB3966E263 D943 A8525771F()( 166838 l/$File/EPA-SAB-l(M)()7-
unsigned.pdf
19 Ibid.
20 U.S. EPA. (2009) Chapter 2.9 Chemical Specific Reference Values for Formaldehyde in Graphical Arrays of
Chemical-Specific Health Effect Reference Values for Inhalation Exposures (Final Report). U.S. Environmental
Protection Agency, Washington, DC, EPA/600/R-09/061, and available on-line at
httv://cfpub. epa. eov/ncea/cfin/recordisplav. cfm ?deid=211003.
30
-------
of exposure, noting that the National Academy of Sciences (NAS) recommends that cancer and
chronic non-cancer risk assessment use the same approach in order to address the fact that very
low levels of non-carcinogen exposures can pose health risks. The commenter noted that the use
of RfDs for dose-response risk assessments of chronic non-cancer health effects may
significantly underestimate risk, because the NAS defines the RfD or RfC as a dose "likely to be
without an appreciable risk of deleterious effects" over a lifetime of exposure.
The commenter identified what they see as the problems with traditional toxicology as the basis
for risk assessments, including the presumption that health effects are related to dose, and that a
dose can be found for virtually all chemicals where no effect is found. The commenter claimed
that the greatest public health threat of chemicals is for fetal exposure, meaning that the dose
may be less important than the timing, and furthermore that for some toxic chemicals, the
clinical effect can actually increase as the chemical concentration decreases, meaning that there
is no safe level of exposure. For example, the commenter noted that a 2009 statement by the
Endocrine Society stated that "[e]ven infinitesimally low levels of exposure, indeed, any level of
exposure at all, may cause endocrine or reproductive abnormalities, particularly if exposure
occurs during a critical developmental window. Surprisingly, low doses may even exert more
potent effects than higher doses."21 The commenter also noted that a recent panel of twelve
national endocrine disruptor specialists recently stated that "[for] every chemical that we looked
at that we could find a low-dose cutoff, if it had been studied at low doses it had an effect at low
doses."22 Finally, the commenter cited a report published in The New England Journal of
Medicine, regarding the toxicity of volatilized compounds from oil, which states that:
"Mutagenic effects theoretically can result from a single molecular DNA alteration. Regulatory
prudence has led to the use of "one-hit models" for mutagenic end points, particularly cancer, in
which every molecule of a carcinogen is presumed to pose a risk."23
Response 3: Chronic noncancer dose response values used in the RTR program, including those
derived by EPA and similar authoritative agencies (e.g., Agency for Toxic Substances and
Disease Registry (ATSDR) and CalEPA) represent chronic exposure levels that are intended to
be health-protective. Those values are derived using an approach that is intended not to
underestimate risk in the face of uncertainty and variability. When there are gaps in the available
information, uncertainty factors (UFs) are applied to derive reference values that are intended to
be protective against appreciable risk of deleterious effects. Uncertainty factors are commonly
default values24 (e.g., factors of 10 or 3) used in the absence of compound-specific data; where
21
Endocrine Society, Scientific Statements, https://www.endocrine.org/endocrine-press/scientific-statements.
22 Vandenberg L, et al. Hormones and endocrine-disrupting chemicals: low-dose effects and nonmonotonic dose
responses. Endocrine Rev; doi:10.1210/er.2011-1050 [online 14 Mar 2012],
23 Goldstein B, Osofsky H, LichtveldM. The Gulf Oil Spill N Engl J Med 2011; 364:1334-1348 April 7, 2011.
24 According to the NRC report Science and Judgment in Risk Assessment (NRC, 1994) "[Default] options are
generic approaches, based on general scientific knowledge and policy judgment, that are applied to various elements
of the risk-assessment process when the correct scientific model is unknown or uncertain." The 1983 NRC report
Risk Assessment in the Federal Government: Managing the Process defined default option as "the option chosen on
the basis of risk assessment policy that appears to be the best choice in the absence of data to the contrary" (NRC,
31
-------
data are available, data-derived extrapolation factors may also be developed using compound-
specific information. When data are limited, more assumptions are needed and more default
factors are used. Thus there may be a greater tendency to overestimate riskin the sense that
further study might support development of reference values that are higher (i.e., less potent)
because fewer default assumptions are needed. However, for some pollutants there is some slight
possibility that risks may be underestimated. With regard to consideration of a potential
vulnerability of a specific lifestage, including time period before conception, the EPA includes
this information in its derivation of cancer and noncancer toxicity assessments. For example,
prenatal developmental studies in rodents, when available, are regularly considered in IRIS
toxicity assessments and data are then extrapolated to predict effects in humans. As mentioned
above in this response, in some instances, the available literature is unavailable for a robust
characterization of risk during a specific lifestage and in that case the potential susceptibilities
are accounted for by applying the appropriate uncertainty factors.
The EPA agrees with the NAS that the recommendations on harmonization of cancer and
noncancer approaches are important issues in risk assessment and EPA incorporates NAS
recommendations as feasible. The NAS has agreed with the EPA, specifically on the derivation
methodology of RfCs and RfDs , that the available scientific information does not always allow
for assessment derivation issues to be fully considered and it has reviewed and supported the
approaches currently used in the derivation of the RfCs and RfDs. The NAS has also recognized
that many of the recommended changes for the IRIS Program will need to be incorporated over a
number of years and further recommend continuation of the development of assessments as the
recommendations are implemented (i.e., the process should not be halted until all
recommendations can be enacted). As such, improvements will be made over time and existing
assessments will need to be used in the interim. Further, EPA has a legal obligation to proceed
with regulatory action based on the best, currently available tools.
The commenter states that there are problems associated with traditional toxicology presumption
that health effects are related to dose, however the commenter does not provide any information
to consider an alternative paradigm to risk assessment that would not include an analysis of dose
response relationships in the risk assessment process.
The commenter provided a reference to support the statement that there is no safe level of
exposure and that for fetal exposure, the dose may be less important than the timing, and
furthermore that for some toxic chemicals, the clinical effect can actually increase as the
chemical concentration decreases. The review article on hormones and endocrine-disrupting
chemicals focuses on a broad category of chemicals that appear to act at low concentrations. We
disagree with the interpretation of the commenter on the referenced review. First, the authors of
1983a, p. 63). Therefore, default options are not rules that bind the agency; rather, the agency may depart from them
in evaluating the risks posed by a specific substance when it believes this to be appropriate. In keeping with EPA's
goal of protecting public health and the environment, default assumptions are used to ensure that risk to chemicals is
not underestimated (although defaults are not intended to overtly overestimate risk). See EPA 2004 An examination
of EPA Risk Assessment Principles and Practices, EP A/100/B-04/001 available at:
http://www.epa.gov/osa/pdfs/ratf-final.pdf.
32
-------
the review article do not conclude that there is no safe level of exposure for chemicals in general,
not even for endocrine disruptors. We disagree with the comment that clinical effect increases as
dose of the chemical decreases; rather, the authors of the review conclude that the effect of low
doses of these group of chemicals cannot be predicted by effects observed at high doses, and
they encourage investigators to make changes in chemical testing approaches to identify
potential endocrine disruptors. We concluded that the review article does not provide
information that is relevant to this regulatory action.
We agree with the comment that the regulatory community, including the EPA, should use
conservative approaches to evaluate cancer risks especially when considering cancer risks to
early life stages. When chemical-specific data is available on which age or life-stage specific risk
estimates or potencies can be determined, default age dependent adjustment factors can be
applied when assessing cancer risk for early-life exposures to chemicals which cause cancer
through a mutagenic mode of action (MOA). With regard to other carcinogenic pollutants for
which early-life susceptibility data are lacking, it is the Agency's long-standing science policy
position that use of the linear low-dose extrapolation approach (without further adjustment)
provides adequate public health conservatism in the absence of chemical-specific data indicating
differential early-life susceptibility or when the mode of action is not mutagenicity.25 The basis
for this methodology is provided in the 2005 Supplemental Guidance.26
Comment 4: The EPA inappropriately and unlawfully treated risk as zero for some pollutants:
One commenter stated that EPA underestimates health risks by not using the best available
information on pollutants and by treating various types of risk as zero even when the science
shows risk is present. The commenter asserted that just because EPA has not yet developed a risk
function for a pollutant, type of exposure, or type of risk, does not mean risk does not exist and
can be ignored. The commenter suggested that EPA develop default approaches to support the
evaluation of risk from chemicals which lack chemical-specific data. One of the approaches
detailed by the commenter is the inclusion of an uncertainty factor to account for the additional
risk that a HAP likely causes, until such time as EPA does have a reference value to use. The
commenter added that if a default approach is not developed, the EPA should at a minimum
engage in the interim in a qualitative assessment of the additional, missing risks, and account for
them in its analysis. The commenter stated that EPA must include the risk from all HAP to
satisfy their legal obligation under section 112(f)(2) to prevent unacceptable risk and ensure an
"ample margin of safety to protect public health."
The commenter also argued that some pollutants continue to have no reference values over 20
years after the CAA was amended and that the IRIS review process has been bogged down for
many pollutants.27 One commenter asserted that, for pollutants currently under IRIS assessment,
25 Id.
26 U.S. EPA. (2005), Supplemental Guidance for Assessing Susceptibility from Early-Life Exposure to Carcinogens.
EPA/630/R-03/003F. Washington, DC. Available online at: http://www.epa.gov/cancerguidelines/guidelines-
carcinogen-supplement. htm.
27 The commenter cites to U.S. Gov't Accountability Office, GAO-12-42, Chemical Assessments: Challenges
Remain with EPA's Integrated Risk Information System Program 17-18 (2011).
33
-------
the EPA must use the best available scientific information from the IRIS review during current
rulemakings.28 At minimum, commenters concluded that the EPA must account for the lack of
reference values or the lack of an up-to-date final IRIS assessment rather than not including any
consideration of health risks caused by such pollutants.
Finally, according to the commenter, the EPA violated the notice and comment requirements by
not providing public notice of all HAPs for which it did not evaluate cancer, chronic non-cancer,
acute, or multipathway risk. The commenter contended that this prevented commenters from
having a meaningful opportunity to present data to EPA that may be useful in EPA's evaluation
of the risk from pollutants for which EPA is currently treating as zero risk.
Response 4: This issue was addressed by the EPA's SAB in its May 7, 2010 response to the
EPA Administrator.29 In that response, the SAB panel recommended that, for HAP that do not
have dose-response values from the EPA's list, the EPA should consider and utilize, as
appropriate, additional sources for such values that have undergone adequate and rigorous
scientific peer review. The SAB panel further recommended that the inclusion of additional
sources of dose-response values into the EPA's list should be adequately documented in a
transparent manner in any residual risk assessment case study. We agree with this approach and
have considered other sources of dose-response data when conducting our risk determinations
under RTR. However, in some instances no sources of information beyond the EPA's list are
available. Compounds without health benchmarks are typically those without significant health
effects compared to compounds with health benchmarks, and in such cases we assume the
compounds will have a negligible contribution to the overall health risks from the source
category. For a tabular summary of HAPs that have dose response values for which an exposure
assessment was conducted, refer to Table 3.1-1 of the "Final Residual Risk Assessment for the
Petroleum Refining Source Sector", Docket ID No. EPA-HQ-OAR-2010-0682.
The EPA agrees that it is important to develop toxicity values for all HAP utilizing all credible
and relevant toxicity information. The need to update assessments with newly available data as
well as the need to complete toxicological assessments for all HAP lacking dose-response
assessments increases the importance of Agency activities to streamline and fully utilize the
EPA's already overloaded IRIS program. To that end, the EPA has always prioritized for IRIS
assessments those HAP without dose-response values but with the greatest potential for public
exposure. As a result of this prioritization, while not all HAP may have scientifically accepted
dose-response values that can be used in residual risk assessments, it is clear that the vast
majority of HAP which might carry the potential to significantly impact the results of residual
risk assessments do, in fact, have credible dose-response values. Thus, while we are not yet at the
point where all HAP have dose-response values, we are generally capable of deriving reasonable
risk estimates for those HAP which dominate the risks from any one source category. In the
course of each residual risk assessment, should we encounter HAP without dose-response values
which carry the potential to create significant risks, we shall clearly point those out as
28 Integrated Risk Information System (IRIS); Announcement of 2012 Program, 77 FR 26,751 (May 7, 2012).
29 The SAB peer review of RTR Risk Assessment Methodologies is available at:
http://vosemite.epa. gov/sab/sabproduct.nsf/4AB3966E263D943A8525771F00668381/$File/EPA-SAB-10-007-
unsigned.pdf.
34
-------
uncertainties and target them for future IRIS assessments. In general, we strive to strike a
balance in our assessments, meaning that while some factors likely lead to underestimates of
risk, others likely lead to underestimates of risk. We conclude that the risk assessment for this
source category is sufficient to support a decision on the acceptability of the risk and ample
margin of safety.
Comment 5: The EPA did not rely on the latest science and underestimated risk by not
adequately accounting for pre-natal and early life exposures: One commenter stated that EPA's
cancer risk assessment for refineries does not adequately account for early-life exposure or the
greater risk to and susceptibility of children. According to the commenter, EPA must account for
increased early-life susceptibility by applying age-dependent adjustment factors for all
carcinogens emitted by a source category. The commenter noted that EPA has restricted its
application of age-dependent adjustment factors to those HAPs included in EPA's list of
carcinogens that act by a mutagenic mode of action.30 The commenter pointed out that EPA's
2005 Guidelines recognized that updates would be needed if more data became available,31 and
claimed that such data are now available from the NAS and OEHHA, yet the EPA has not issued
such updates to implement age-dependent adjustment factors for all carcinogens.32
Regarding pre-natal cancer risk, the commenter noted that EPA's risk assessment does not take
into account increased susceptibility to carcinogens due to pre-natal exposures, even for known-
to-be mutagenic carcinogens, and that EPA must do so for these as well as for all carcinogens.33
The commenter argued that despite EPA's own recognition that exposures of concern include
pre-conception exposures of both parents through adolescence, it has not developed adjustment
factors for pre-natal exposures.34'35 This omission from EPA's 2005 Guidelines was noted by
30 Draft Risk Assessment (-0225) at 29-30; See EPA, "Guidelines for Carcinogen Risk Assessment," EPA/630/P-
03/00 IF, at 1-19 to 1-20 (Mar. 2005),
http://www.epa.gov/raf/publications/pdfs/CANCER_GUIDELINES_FINAL_3-25-05.PDF; EPA, "Supplemental
Guidance for Assessing Susceptibility from Early-Life Exposure to Carcinogens," EPA/630/R-03/003F (2005),
http://www.epa.gov/raf/publications/pdfs/childrens_supplement_final.pdf.
31 See EPA, 2005 Supp. Guidance at 21, 31 ("EPA expects to expand this Supplemental Guidance to specifically
address modes of action other than mutagenicity when sufficient data are available and analyzed.").
32 Cal. EPA, OEHHA, TSD for Cancer Potency Factors, supra note 313. EPA should also update the 2005
Guidelines to fully reflect current science as described in OEHHA's 2009 review of the scientific literature on
increased susceptibility to carcinogens from early life exposures
33 Draft Risk Assessment (-0225) at 29 (noting that EPA applied factors only to known mutagens to account for
"children aged 0-1" but not younger than that).
34 EPA 2005 Guidelines for Carcinogen Risk Assessment, EPA/630/P-03/00 IF, at 1-16.
35 EPA, 2005 Supp. Guidance at 4-5, 14 & tbl. la (A-l) (discussing research on human and animal cancer risks from
prenatal exposure).
35
-------
NAS.36 The commenter recommended that EPA use the OEHHA methods and adjustment
factors, as well as procedures to assess exposure during fetal development, to account for pre-
natal susceptibility and exposures.37'38'39 According to the commenter, OEHHA specifically
discusses the use of a 10X adjustment factor for cancer risk to account for pre-natal (third
trimester) to age 2 exposures, and the EPA should consider using this same factor.40
Regarding non-cancer risk, another commenter asserted that exposure to toxic agents in the
intrauterine stage of life has one of the most important, potentially irreversible impacts on life-
long health, but EPA's rules are insufficient to protect human health at the critical stage of
embryonic development. Most of EPA's IRIS toxicity threshold values (reference concentrations
and doses) used for chronic non-cancer risk assessment do not incorporate the latest science on
increased susceptibility of children and, according to the commenter, EPA must consult and
apply child-specific reference values, where available.41 The commenter argued that until the
IRIS values fully account for the increased risk caused by early-life exposure to an emitted
pollutant, the EPA should use the OEHHA child-specific reference doses or benchmarks
available to assess chronic non-cancer health risk from ingestion for certain pollutants. The
commenter asserted that EPA should also assess such risk from inhalation by using standard
methods to translate these values into child-specific reference concentrations to assess
inhalation-based risk.
Where child-specific reference values are unavailable, the commenter asserted that EPA must
consult science on early exposure impacts and use an additional default or uncertainty factor.
Until EPA has child-specific or child-based reference values available for a given pollutant, the
EPA should apply a default or uncertainty factor of at least 10, according to the commenter, to
account for increased risk from early-life exposures for non-cancer risk in this rulemaking and
36 NAS 2009, supra note 264, at 112-13; see also id. at 112, 196 (noting that it is a "missing" default that EPA
recognizes in utero carcinogenic activity, but fails to take account of it or calculate any risk for it as "EPA treats the
prenatal period as devoid of sensitivity to carcinogenicity").
37See Cal. EPA, OEHHA, "Technical Support," supra note 313, App. J: "In Utero and Early Life Susceptibility to
Carcinogens: The Derivation of Age-at-Exposure Sensitivity Measures" conducted by OEHHA's Reproductive and
Cancer Hazard Assessment Branch, .http://oehha.ca.gov/air/hot_spots/2009/AppendixJEarly.pdf.
38 Id. App. J at 7-8 & tbl. 1
39 See Cal. EPA, Air Toxics Hot Spots Program Risk Assessment Guidelines: Technical Support Document for
Exposure Assessment and Stochastic Analysis at 1-6 to 1-7 (Aug. 27, 2012) ("OEHHA 2012 Guidelines"),
http://www.oehha.ca.gov/air/hot_spots/tsd082712.html.
40 See id.; 2014 Air Toxics Hot Spots Program Guidance Manual, supra note 121, at 2.
41 OEHHA has explained why child-specific reference doses or values are needed and provided a list of chemicals.
See, e.g., Cal. EPA, OEHHA, "Prioritization of Toxic Air Contaminants - Children's Environmental Health
Protection Act" (Oct. 2001),
http://oehha.ca.gov/air/toxic_contaminants/pdf_zip/SB25%20TAC%20prioritization.pdf; Cal. EPA, OEHHA,
"Development of Health Criteria for School Site Risk Assessment Pursuant to Health and Safety Code 901(g):
Identification of Potential Chemical Contaminants of Concern at California School Sites, Final Report" (June 2002),
http://oehha.ca.gov/public_info/public/kids/pdf/ChildHealthreport60702.pdf.
36
-------
other risk assessments. The commenter noted that this would be consistent with the NAS
recommendation on the need for EPA to use default factors to account for greater risk42, with the
science developed and considered by OEHHA, and with the 10X factor enacted by Congress in
the Food Quality Protection Act. The Food Quality Protection Act (FQPA) enacted, a Ten-fold
Margin of Safety, or "10X factor." Specifically, the Act provided that "an additional tenfold
margin of safety for the pesticide chemical residue and other sources of exposure shall be applied
for infants and children to take into account potential pre- and post-natal toxicity and
completeness of the data with respect to exposure and toxicity to infants and children." 43
Congress's recognition of the need to use this default factor provides a model that EPA should
consider and incorporate into its residual risk assessment, according to the commenter. The
commenter stated that it would be appropriate and within EPA's authority under CAA section
112(f)(2) to determine that EPA must use a children's ten-fold margin of safety factor, to fulfill
the CAA's "margin of safety" requirement.
Response 5: We disagree with the comment that this risk assessment underestimates risk to
children and lacks consideration of early-life susceptibility. We acknowledge that population
subgroups, including children, may have a potential for risk that is greater than the general
population due to greater relative exposure and/or greater susceptibility to the toxicant. The
assessments we undertake to estimate risk account for this potential vulnerability. With respect to
exposure, the risk assessments we perform implicitly account for this greater potential for
exposure by assuming lifetime exposure, in which populations are conservatively presumed to be
exposed to airborne concentrations at their residence continuously, 24 hours per day for a full
lifetime, including childhood. With regard to children's potentially greater susceptibility to non-
cancer toxicants, the assessments rely on EPA (or comparable) hazard identification and dose-
response values which have been developed to be protective for all subgroups of the general
population, including children.
For example, a review44 of the chronic reference value process concluded that the EPA's RfC
derivation processes adequately considered potential susceptibility of different subgroups with
specific consideration of children, such that the resultant RfC values pertain to the full human
population "including sensitive subgroups," a phrase which is inclusive of childhood. With
respect to cancer, the EPA uses the age-dependent adjustment factor approach referred to by the
commenter, but limits the use of those factors only to carcinogenic pollutants that are known to
act via mutagenic mode of action, in contrast to the OEHHA approach, which uses them across
the board for all carcinogens regardless of MO A. In lieu of chemical-specific data on which age
or life-stage specific risk estimates or potencies can be determined, default age dependent
adjustment factors can be applied when assessing cancer risk for early-life exposures to
42 NAS 2009, supra note 264, at 190-93, 203.
43 21 U.S.C. 346a(b)(2)(C) (requiring that, in establishing, modifying, leaving in effect, or revoking a tolerance or
exemption for a pesticide chemical residue, "for purposes of clause (ii)(I) an additional tenfold margin of safety for
the pesticide chemical residue and other sources of exposure shall be applied" to protect infants and children).
44 U.S. EPA. A Review of the Reference Dose and Reference Concentration Processes. U.S. Environmental
Protection Agency, Risk Assessment Forum, Washington, DC, EPA/630/P-02/002F, 2002.
37
-------
chemicals which cause cancer through a mutagenic MOA. With regard to other carcinogenic
pollutants for which early-life susceptibility data are lacking, it is the Agency's long-standing
science policy position that use of the linear low-dose extrapolation approach (without further
adjustment) provides adequate public health conservatism in the absence of chemical-specific
data indicating differential early-life susceptibility or when the mode of action is not
mutagenicity.45 The basis for this methodology is provided in the 2005 Supplemental
Guidance.46
We also disagree with the comment that the risk assessment for these source categories did not
consider the groups that may be most at-risk (e.g., pregnant women and children). When the
EPA derives exposure reference concentrations and URE for hazardous air pollutants, it also
considers the most sensitive populations identified in the available literature, and importantly,
these are the values used in our risk assessments.47 With regard to consideration of a potential
vulnerability of a specific lifestage, including time period before conception, the EPA includes
this information in its derivation of cancer and noncancer toxicity assessments. For example, a
prenatal developmental studies in rodents, when available, are regularly considered in IRIS
toxicity assessments and data are then extrapolated to predict effects in humans. In some
instances, the available literature is unavailable for a robust characterization of risk during a
specific lifestage and in that case the potential susceptibilities are accounted for by applying the
appropriate uncertainty factors.
We disagree with the general comment that the EPA should adopt the Cal OEHHA child-
protective scientific approach on for deriving health benchmarks. The EPA has an approach for
selecting appropriate health benchmark values and in general, this approach places greater
weight on the EPA derived health benchmarks than those from other agencies. The approach of
favoring EPA benchmarks (when they exist) has been endorsed by the SAB,48 and ensures use of
values most consistent with well-established and scientifically-based EPA science policy. The
EPA is currently evaluating the most appropriate use for the Cal OEHHA child-specific
reference doses. We note that there are currently no such values for HAP inhalation, therefore
the current utility may be limited to persistent and bioaccumulative (PB)-HAP, which may be
associated with non-negligible ingestion exposures. This evaluation on appropriateness of these
values in the context of the RTR Program and EPA science policy is necessary prior to using
these child-specific RfD.
45 Id.
46 U.S. EPA. (2005), Supplemental Guidance for Assessing Susceptibility from Early-Life Exposure to Carcinogens.
EPA/630/R-03/003F. Washington, DC. Available online at: http://www.epa.gov/cancerguidelines/guidelines-
carcinogen-supplement. htm.
47 US EPA. (2002). A review of the reference dose and reference concentration processes. EPA/630/P-02/002F. Risk
Assessment Forum, Washington, DC. Available online at httv://www. eva. sov/raf/vublications/vdfs/rfd-final, vdf
48 The SAB peer review of RTR Risk Assessment Methodologies is available at:
http://vosemite.epa. gov/sab/sabproduct.nsf/4AB3966E263D943A8525771F00668381/$File/EPA-SAB-10-007-
unsigned.pdf
38
-------
The estimated risks must also be considered in the context of the full set of assumptions used for
this risk assessment. Our unit risk estimates for hazardous air pollutants are considered a
plausible upper-bound estimate with an appropriate age dependent adjustment; actual potency is
likely to be lower and could be as low as zero. Our chronic noncancer reference values have been
derived considering the potential susceptibility of different subgroups, with specific
consideration of children. In addition, an extra 10X uncertainty factor is not needed in the
RfC/RfD methodology because the currently available factors are considered sufficient to
account for uncertainties in the database from which the reference values are derived. After
considering these and other factors, we continue to consider the risks from emissions after
application of the revisions to the NESHAP for these source categories to provide an ample
margin of safety (consistent with the Benzene NESHAP framework).
The EPA disagrees with the commenter in that a children's default safety factor of 10 or more
should be added to EPA's reference values. In response to the 10X factor enacted by Congress in
the FQPA (1996)49 to the EPA non-cancer reference value derivation, the Agency evaluated their
methods for considering children's risk in the development of reference values. As part of their
response, the U.S. EPA (i.e. the Science Policy Council and Risk Assessment Forum) established
the RfD/RfC Technical Panel to develop a strategy for implementing the FQPA and examine the
issues relative to protecting children's health and application of the 10X safety factor. One of the
outcomes of the Technical Panel's efforts was an in depth review of a number of issues related to
the RfD/RfC process (U.S. EPA 2002). The most critical aspect in the derivation of a reference
value pertaining to the FQPA has to do with variation between individual humans and is
accounted for by a default uncertainty factor (UF-H) when no chemical-specific data are
available. EPA reviewed the default UF for inter-human variability and found the EPA's default
value of 10 adequate for all susceptible populations, including children and infants. The EPA
also recommended the use of chemical-specific data in preference to default uncertainty factors
when available (US EPA, 1994, 2011) and is developing Agency guidance to facilitate
consistency in the development and use of data-derived extrapolation factors for RfCs and RfDs
(U.S EPA, 2011). Additionally, the EPA also applies a database uncertainty factor (UF-D) which
is intended to account for the potential for deriving an under protective RfD/RfC as a result of an
incomplete characterization of the chemical's toxicity. In addition to the identification of toxicity
information that is lacking, review of existing data may also suggest that a lower reference value
might result if additional data were available.
In conclusion, an additional uncertainty factor is not needed in the RfC/RfD methodology
because the currently available factors are considered sufficient to account for uncertainties in
the database from which the reference values are derived (and does not exclude the possibility
that these uncertainty factors may be decreased or increased from the default value of 10).
2.3.3 Additional pollutants and health effects should be considered
Comment 1: Additional pollutants to those considered pose serious health risks: Commenters
described the health effects including asthma and other respiratory diseases, IQ loss and other
49 US Environmental Protection Agency, Pesticide: Regulating Pesticides. The Food Quality Protection Act (FQPA).
1996. Available at http://www.epa.gov/pesticides/regulating/laws/fqpa/backgrnd.htm.
39
-------
developmental impacts, cancer, heart disease, birth defects and other reproductive system
impacts, kidney, liver and other organ damage, and even premature death they claimed result
from toxic air pollutants emitted by refineries and asserted they must be reduced. The pollutants
referenced by the commenters include: benzene, toluene, carbonyl sulfide, 1,3-butadiene, ethyl
benzene, mixed xylenes, n-hexane, hydrogen sulfide (H2S), hydrogen chloride (HC1), HCN,
hydrogen fluoride (HF), PAHs (POM), mercury (including methylmercury), naphthalene, non-
mercury metals (arsenic, beryllium, cadmium, lead, hexavalent chromium, manganese, nickel,
and selenium), formaldehyde, acetaldehyde, SO2, nitrogen oxides (NOx), particulate matter 2.5
micrometers in diameter and smaller (PM2.5), VOCs as precursors to ozone (O3), and methane
(CH4). Another commenter supported the need for public health impacts research data on all
chemicals, including determination and annual reporting of all carcinogenic, neurological, non-
carcinogenic, bio-cumulative, and physical and developmental public health impacts. Another
commenter stated that HF is not addressed in the rule, but the risks from HF are catastrophic
risks if something goes wrong which the commenter noted has happened several times in Corpus
Christi.
The commenters provided information on risks from specific emissions and pollutants. One
commenter stated that there is clear evidence of harm from direct emissions of sulfur dioxide,
nitrogen oxide and VOCs, and further noted that these pollutants are precursors to O3 and PM2.5,
which also pose a significant threat to human health (including increased risk of cardiovascular,
respiratory, other acute and chronic systemic damage, and potentially cancer). Another
commenter stated that although EPA's rule targets section 112-listed HAPs, refineries emit vast
quantities of other pollutants, including criteria pollutants that can negatively interact with and
exacerbate the impacts of HAP exposure. Of gravest concern, according to the commenter, are
the numerous studies that have documented a wide range of adverse health impacts from
exposure to PM2.5.
Response 1: Section 112(f) of the CAA evaluates risks associated with emissions of hazardous
air pollutants listed under section 112 (b)(1) or those added to this list under section 112(b)(2).
The EPA does not regulate VOC, particulate matter, NOx or SO2 under section 112 except to the
extent that individual HAP are also VOC (e.g., benzene, toluene, xylene) or particulate matter
(PM) (e.g., nickel, mercury). The pollutants are regulated under Title 1 of the CAA as NAAQS
or precursors to the NAAQS.
The EPA conducted an assessment of the cumulative cancer risks from all emitted carcinogens
and the cumulative noncancer hazard indices from all emitted non-carcinogens that are HAP
affecting the same target organ system for both the source category emissions and the facility-
wide emissions. To address the effect of mixtures of carcinogens, the individual cancer risks for
the source categories were aggregated for all carcinogens. In assessing noncancer hazard from
chronic exposures for pollutants that have similar modes of action or that affect the same target
organ, we aggregated the hazard quotients (HQ) to provide a TOSHI. We further modeled
whole-facility risks for both chronic cancer and non-cancer impacts to understand the risk
contribution of the sources within the petroleum refineries source categories.
HF was included in both the inhalation and, along with six other environmental HAP, in the
environmental screening analysis for these source categories. The average modeled
40
-------
concentration around each facility (i.e., the average concentration of all off-site data points in the
modeling domain) did not exceed any human or ecological benchmarks. In addition, each
individual modeled concentration (i.e., each off-site data point in the modeling domain) was
below the human and ecological benchmarks for all facilities.
Comment 2: Hydrogen sulfide should be included in risk assessment: One commenter stated
that EPA should include H2S emissions in the technology and risk review as well as develop an
actionable level and a national ambient air quality standard for H2S. The commenter stated that
there are health risks associated with H2S including eye and respiratory irritations, nausea,
dizziness, confusion and headache, which may disproportionality affect children.
Another commenter stated that there is a pending petition to list H2S as a HAP, and EPA should
grant that petition without further delay and regulate it at refineries.50 The commenter added that
the EPA's addition of H2S reporting to the TRI, under the Emergency Planning and Community
Right-to-Know Act (EPCRA), demonstrates EPA's acknowledgment of existing risk and further
supports listing this compound as a HAP.51
Response 2: Section 112(f) of the CAA evaluates risks associated with emissions of hazardous
air pollutants listed under section 112 (b)(1) or those added to this list under section 112(b)(2).
H2S is neither on the 112(b)(1) list nor has it completed the listing process under section
112(b)(2) as of the time of this analysis. If, in the future, H2S, or any other HAP, is added to the
list under section 112(b) (2) they will be included in all future risk analyses.
2.3.4 Additional emissions and source categories should be considered
Comment 1: EPA understates risk by ignoring emissions from upsets and malfunctions: One
commenter stated that EPA must account for the acute, cancer, and chronic non-cancer health
risks from emissions during upsets and malfunctions, instead of ignoring these risks, particularly
since these emissions can be significantly higher than emissions at any other time of source
operation. Ignoring these emission spikes is equivalent to treating additional health risk caused
by exceedances as zero, according to the commenter.
Commenters added that EPA's own scientists stated that EPA may be underestimating actual
maximum short-term emissions, through the use of low, short-term emissions factors and "data
filtering" such that "accidental releases were dropped", thus also underestimating maximum
health risk for the most-exposed person.52 The commenter asserted that the dropping of all so-
called "[accidental releases" removes most of the maximum short-term emissions numbers that
50 Letter from Sierra Club et al. to U.S. EPA Administrator Jackson, Hydrogen Sulfide Needs Hazardous Air
Pollutant listing under CAA Title III, (Mar. 25, 2009), available at
http://www.texas.sierraclub.org/press/newsreleases/H2SLetterT0EPA.pdf
51 U.S. EPA, Lifting of Administrative Stay for Hydrogen Sulfide, 76 FR 64,022, 64,024 (Oct. 17, 2011), see also
http://www.epa.gov/tri/lawsandregs/hydrogensulfide/indexf.html (last accessed Sept. 23, 2011)
52 Risk Assessment App. 3, Ted Palma & Roy Smith, Analysis of data on short-term emission rates relative to long-
term emission rates, at 3
41
-------
the EPA must consider if it indeed wishes to fulfill the Act's requirement and its own
interpretation of its responsibility.
Regarding the issue of accidental releases, commenters objected to the use of the term
"accidental" since according to the commenters, many of these events can be prevented. The
commenters also asserted that emissions from SSM are still lawful and thus should be considered
to be part of the "allowable" emissions in the risk evaluation. In contrast to EPA policy, the
commenters suggested that emissions from accidental releases must be evaluated even if they
exceed the "allowable" emissions and are considered a violation of the standard. Commenters
also stated that EPA did not say that it removed only emissions that were accidental releases that
exceeded the level of emissions standards.
In support of their arguments, commenters noted that the SAB has also questioned how the risk
from these emissions are addressed and has even criticized the EPA's estimation of maximum
short-term emissions. Commenters expressed concern that EPA's method of calculating acute
risk using a "worst-case" scenario is not actually representative of the "worst case" because it is
ignores all malfunctions which exceed the standards. The commenter stated that EPA could
simply use a more accurate factor (based on statistical methods and probability factors) to
account for malfunctions for acute and other types of health risk, to close the gap and respond
appropriately to the SAB's criticism of its current method.
The commenter noted that, to create representative factors to assess the health risk from
malfunctions, EPA has information available or can collect information on major sources'
malfunction and violation histories.53 According to the commenter, EPA has already collected
significant information on upset incidences as part of the Information Collection Request for this
rulemaking [See ICR Component 1, at Part III] and the agency must evaluate and use them to
address emission spikes at least to the extent reported by these existing sources. The commenter
further noted that the Agency has already collected approximately two years' worth of this type
of information from the TCEQ.54 The commenters suggested that EPA should consider more of
these data from TCEQ and other states that have delegated air programs, as facilities are required
53 See, e.g., EPA, Enforcement and Compliance History Online (ECHO), www.epa.gov/echo; Kelly Haragan, Envtl.
Integrity Project, "Gaming the System: How Off-the-Books Industrial Upset Emissions Cheat the Public Out of
Clean Air" (Aug. 2004), 1-2, 5, http://www.environmentalintegrity.org/news_reports/Report_Gaming_System.php
(finding significant likelihood of an upset at refineries, chemical plants, gas plants and a carbon black plant, and
finding that the resulting emissions release is many times higher than the amount of otherwise-reported annual
emissions and that "releases from upsets actually dwarf a facility's routine emissions.").
54 Palma & Smith memo at 2-3 ("The Texas Commission on Environmental Quality (TCEQ) collects emissions data
using online reporting required of any facility releasing 100 pounds or more of a listed chemical (primarily ozone-
forming VOCs) during a non-routine event. ... The database we utilized in our analysis was a subset of the TCEQ
data covering emission events that occurred in an eight-county area in eastern Texas during a 756-day period
between January 31, 2003 and February 25, 2005."); see Tex. Comm'n on Envtl. Quality, Search the Air Emission
Event Report Database, http://wwwll.tceq.texas.gov/oce/eer/index.cfm.
42
-------
to report malfunction releases to the states under EPA's existing regulations.55 The commenter
offered additional data with these comments, including: (1) EIP has created two major reports on
upset or malfunction incident data that refineries reported to Texas and more recent follow-up
letters summarizing data since 2012;56 (2) Louisiana Bucket Brigade has compiled similar data
from Louisiana reports by refineries to the state.57
Response 1: While we appreciate the additional information provided by the commenters about
specific emissions events in violation of the standards, we disagree with the commenter that such
emissions, whether or not they are caused by malfunction events, should be considered as part of
the risk analysis. The purpose of the risk review is to evaluate whether the emission limits - the
"standards promulgated pursuant to subsection (d)" not the non-compliance with those standards
[section 112(f)(2)(A)] should be made more stringent to reduce the risk posed after compliance
with the underlying MACT standard. To the extent that a source is violating the underlying
MACT standard, no tightening of the emission standard under the residual risk rule will avoid or
mitigate against such violations. In other words, a source that is violating the MACT emissions
standard promulgated under section 112(d) would not be any more likely to be able to avoid such
violations and comply with a different, presumably more stringent, standard promulgated under
section 112(f). Such events are violations and subject to enforcement by the EPA, the States or
citizens, and an action for injunctive relief is the most effective means to address such violations,
whether or not they are caused by malfunctions if an emissions event poses a significant health
or environmental risk.
While we agree with the commenter that the original standards did provide some relief for
malfunctions, we proposed to remove those provisions from the MACT standard consistent with
the Court's decisions in Sierra Club. Thus, at proposal, we evaluated risk based on the MACT
standards as they would be modified to incorporate this revision. We continue to follow that
approach for the final rule. Because we are finalizing standards that would apply to certain
releases from flaring events and PRDs, we are evaluating emissions that would be allowed (i.e.,
that would not be a violation) during these events as part of our risk review.
We performed a screening level analysis to estimate the risks from these non-routine emissions
(flaring events and PRDs). To estimate emissions for chronic risk calculations, emissions data
was extracted from the EPA ICR in Component 1 of the survey. We summed the HAP emissions
from all individual events at each facility, since chronic inhalation risk depends on the total
quantity of emissions released over a year and is independent of the release time. This screening
55 See, e.g., 40 CFR 63.10(b) & (d); 40 CFR 63.655(g)(6)(i)-(ii) (pre-proposed rule) (requiring reporting of periods
of "excess emission"); see also 40 CFR part 60 subpart J and Ja, 40 CFR part 63 subparts CC, UUU (other similar
requirements)..
56 Accident Prone, supra note 222; Letter from Eric Schaeffer, Environmental Integrity Project, Executive Director,
to Inspector General Arthur Elkins, Environmental Protection Agency, Clean Air Act Enforcement of Excess
Emissions and the Affirmative Defense (Apr. 23, 2012) [hereinafter EIP SSM Enforcement Letter] available at
http://www.environmentalintegrity.org/news_reports/documents/EmissionEventLetter2009-2012FINAL001 .pdf
57 Louisiana Bucket Brigade, Common Ground Reports I-IV.
43
-------
level analysis indicated we can expect these emissions to contribute up to about 2-in-l million to
the chronic cancer MIR. Noncancer risks resulting from these emissions are expected to remain
well below 1. To estimate potential risks for acute events, we examined both the quantity and
duration of each event such that higher mass emitted in a shorter time period has an increased
potential for an acute impact. Acute risks from these events estimated a Hazard Index based on
the REL of up to 14 from emissions of benzene. While the analysis shows that there is a potential
at these facilities for HQs exceeding 1 (benzene), because of the many uncertainties and the
screening level nature of this analysis, we believe the potential for achieving a HQ=14 is low. In
all cases the AEGL levels are well below 1. Further, it is important to note that the estimated
acute risk assume a catastrophic release such that all the emissions from an event are emitted
during a single hour time period (for more detail on this analysis, see Appendix 13 of the Final
Residual Risk Assessment for the Petroleum Refining Source Sector in Docket No. EPA-HA-
OAR-2010-0682).
Comment 2: EPA understates risk by not accounting for cumulative risk from facility-wide
emissions and multiple nearby sources impacting an area: One commenter asserted that in
addition to performing a cumulative assessment from refineries alone, the EPA must perform a
cumulative analysis which aggregates or adds the emissions for the most-exposed communities
coming from: (1) the source category (including all individual sources within it); (2) facility-
wide risk from collocated sources outside of this category; and (3) all other sources of toxic air
pollution in the area.58 The commenter stated that this is particularly important for the many
communities containing refineries and other nearby sources of toxic air emissions such as:
Wilmington, California, Port Arthur, Texas, and Delaware City, Delaware. The commenter noted
that EPA has recognized this need in its recent risk report59, yet has failed to propose any
changes to the emission standards based on the combined exposure with any other sources. In
support of their argument, the commenter also cited recommendations from the SAB and NAS
which called for the incorporation of cumulative health risk into its residual risk analysis.
The commenter also noted that EPA has not used its calculated "facility-wide" risk for co-
located sources to set standards, and it has ignored different sources across the street or in close
proximity in its Draft Risk Assessment. In addition, the commenter claimed that EPA has
provided no information on how it reached the "facility-wide" risk number.
The commenter recommended that EPA use the risk assessment results available for those source
categories for which it has already performed a risk assessment review - such as those covered
58 We support EPA's recognition of the need to assess whether the maximum exposed individual is exposed to
emissions from more than one source within each source category. We also appreciate that EPA has considered
facility-wide risk in some way in this rulemaking. However, those assessments offer only part of the picture. And,
even on both of these issues, EPA has provided very little information about what it included in such assessments, as
discussed elsewhere in these comments. EPA just states numbers found for facility-wide risk, without explaining
where those numbers came from, how they were calculated, or what emission sources they cover.
59 U.S. EPA, "Concepts, Methods and Data Sources," supra, at xxxii (defining a cumulative risk assessment as
including "aggregate exposures by multiple pathways, media and routes over time, plus combined exposures to
multiple contaminants from multiple sources").
44
-------
by the hazardous organic NESHAP and other chemical plant sources that are frequently located
near refineries in its cumulative risk assessment. According to the commenter, the combined risk
results for these sources near one another (not just co-located) would likely have higher overall
risk than has been estimated for the most-exposed person near a refinery, and therefore stronger
standards are needed under section 112(f)(2) for multiple emission points at refineries (not just
storage vessels). Finally, the commenter suggested that EPA consider the existing research
regarding health risk from toxic air pollution in urban communities nationwide60 as well as the
OEHHA cumulative assessment approach and apply a similar science-based approach in this
residual risk assessment.61
In addition, one commenter noted that the NAS has recommended that EPA evaluate
"background exposures and vulnerability factors," as well as use "epidemiologic and toxicologic
evidence" in its risk assessments.62 Rather than separating an environmental justice analysis and
considerations of inequality from the risk assessment, considering these factors as part of the
cumulative risk assessment - because of the increased vulnerability created - would be a more
effective, meaningful, and scientific approach, according to the commenter. The commenter
asserted that, in assessing a source category's emission contributions in affected communities
and considering whether these contributions cause the most-exposed people to experience an
unacceptable level of public health risk when combined with the existing baseline from past
emissions, other HAP emissions, and the community's health status, EPA can describe and
manage uncertainties, similar to many other analyses.63
The commenter concluded that the EPA must develop a data-driven approach to
comprehensively model cumulative risk or impacts from multiple sources, EPA must incorporate
60 See, e.g., Rachel Morello-Frosch & Bill M. Jesdale, Envtl. Health Perspectives, Separate and Unequal: Residential
Segregation and Estimated Cancer Risks Associated with Ambient Air Toxics in U.S. Metropolitan Areas, 114(3)
Envtl. Health Perspectives 386 (2006) (assessing toxic air pollution cancer risk for 309 metropolitan areas
encompassing 45,710 tracts); "National Air Toxics Program: The Integrated Urban Strategy," 64 FR 38,706, 38,738
(July 19, 1999).
61 See, e.g., Cal. EPA, "Cumulative Impacts," supra.
62 NAS 2009, supra note 264, at 221-23 (discussing Menzie et al. 2007 model); id. at 230 (discussing the role of
epidemiology and surveillance data).
63 See, e.g., 42 U.S.C. 7475(a)(3), 7503(a)(1) (requiring a localized, cumulative assessment of whether or not a new
or modified source's additional emissions will cause an attainment area to deteriorate, or will make it difficult for a
nonattainment area to make progress toward achieving the national ambient air quality standards); New York v. EPA,
443 F.3d 880, 883 n.l (D.C. Cir. 2006) (citingNew Yorkv. EPA, 413 F.3d 3, 11-14 (D.C. Cir. 2005)); see also 40
CFR. 1508.27(b)(7) (requiring a consideration of "[w]hether the action is related to other actions with individually
insignificant but cumulatively significant impacts. Significance exists if it is reasonable to anticipate a cumulatively
significant impact on the environment. Significance cannot be avoided by terming an action temporary or by
breaking it down into small component parts"); see also 40 CFR. 1508.7; Nat'l Wildlife Fed'n v. Nat'lMarine
Fisheries Serv., 524 F.3d 917, 930 (9th Cir. 2008) (applying 16 U.S.C. 1536(a)(2) to enforce the Endangered
Species Act duty to ensure against jeopardy which includes the requirement to assess a newly proposed action in the
context of all other impacts, and determine whether or not the specific action will "tip a species from a state of
precarious survival into a state of likely extinction," or, where baseline conditions already jeopardize a species,
whether it will "deepen[] the jeopardy by causing additional harm").
45
-------
a default or uncertainty factor to adjust the degree to which each individual source category is
contributing to the total risk experienced by the most-exposed individuals. For example,
according to the commenter, for a source category in an area with up to 10 other HAP-emitting
facilities, this default or uncertainty factor should equal at least 10, consistent with the common
scientific use of this factor for other kinds of vulnerability.64 This commenter suggested that
wherever there is a history of other exposures or other source categories, the "unacceptable"
level of cancer, non-cancer chronic, and acute risk from the source category must be adjusted
downward (such that no single source category could constitute all acceptable risk, when the
most-exposed person is exposed to many other source categories).
Response 2: We disagree with the claim that additional quantitative assessment of risks from
sources outside the source category is required under the statute. Section 112(f) requires the EPA
to perform a review of whether there is risk remaining from emissions from a source category
after promulgation of the technology based MACT standards for that source category. To this
end, the EPA conducts a comprehensive assessment of the risks associated with exposure to the
HAPs emitted by the source category and supplements that with additional information that is
available about other possible concurrent and relevant risks. While the incorporation of
additional background concentrations from the environment in our risk assessments (including
those from mobile sources and other industrial and area sources) could be technically
challenging, they are neither mandated nor barred from our analysis. In developing the decision
framework in the Benzene NESHAP currently used for making residual risk decisions, the EPA
rejected approaches that would have mandated consideration of background levels of pollution in
assessing the acceptability of risk, concluding that comparison of acceptable risk should not be
associated with levels in polluted urban air (54 FR 38044, 38061, September 14, 1989).
Although EPA rejected such approaches for considering the acceptable level of risk, EPA
recognized in the Benzene NESHAP that background levels (including natural background)
could be considered as part of EPA's ample margin of safety (AMOS) analysis, as appropriate
and as available, along with other factors, such as cost and technical feasibility.
For the petroleum refinery source categories, the EPA conducted an assessment of the
cumulative cancer risks from emitted carcinogens and the cumulative noncancer hazard indices
from all emitted non-carcinogens affecting the same target organ system for both the source
category emissions and the facility-wide emissions. While the emissions for the source category
and whole facility emissions have been collected as part of the information data request and
reviewed by project engineers and scientists, emissions data for sources outside of the refinery
facility are not readily available at the level of detail and quality that is required for a refined
risks analysis. Thus, because of uncertainties in this data we do not include the risks from
stationary and mobile sources outside of the refinery in our facility-wide risk analysis.
The risk assessment modeling for the Refineries MACT accounted for the effects of multiple
facilities within the source category that may be in close proximity when estimating
concentration and risk impacts at each block centroid. When evaluating the risks associated with
64 For areas with more facilities, which cause an even greater level of health risk combined, the UF should be
adjusted accordingly, i.e., 11-20 facilities would result in an UF of 20, and more than 20 would result in an UF of
100, so the source category's contribution is no higher than 1/100 of the threshold.
46
-------
a particular source category we combined the impacts of all facilities within the same source
category, and assessed chronic exposure and risk for all census blocks with at least one resident
(i.e., locations where people may reasonably be assumed to reside rather than receptor points at
the fenceline of a facility). The maximum individual risk considers the combined impacts of all
sources in the category that may be in close proximity. We do not agree with the idea that we
should apply an uncertainty factor related to the number of non-refinery facilities near a refinery.
First and foremost, MIR is highly location-specific and neighboring facilities seldom have any
significant contributions to the MIR posed from an individual facility. Second, given the
geographic size of refineries and their emissions, it is non-scientific to suggest neighboring
facilities can have such an additive impact on the MIR. When EPA considers the use of an
uncertainty factor in a risk analysis, it only includes such factors that have been developed based
on scientific data that support their use.
2.3.5 Multipathway risk assessment
Comment 1: Multipathway risk assessment is not sufficient and underestimates risks: One
commenter supported performing a multipathway (i.e., non-inhalation-based) risk assessment but
stated that EPA's multipathway analysis is deficient and could be improved by incorporating the
following suggestions:
Perform a multipathway (or non-inhalation) risk assessment assessing the "allowable"
emissions.
Assess the non-inhalation-based risk created by refineries' emissions of all hazardous air
pollutants known to be persistent and bioaccumulative in the environment (PB-HAP)
instead of restricting its multipathway risk screening assessment to the 14 contaminants
identified in the 2004 Risk Assessment Guidance as PB-HAP.65 The commenter
specifically recommended the following compounds be included in the assessment:
arsenic, hexavalent chromium, nickel, diethylhexylphthalate, beryllium, selenium,
manganese and naphthalene. The commenter asserted that these pollutants have been
shown to have a significant potential for deposition and retention within the environment
and present a risk to nearby communities. The commenter cited the California OEHHA
2012 Guidelines for Exposure Assessment as the rationale for including these HAP in the
assessment and recommended that EPA review and adopt the methods in these guidance
documents. Specifically for naphthalene, the commenter stated that this compound has
been demonstrated to be persistent and bioaccumulative and is a PAH, and as such must
be considered in the POM category which is already listed as a PB-HAP.
EPA must perform a proper multipathway assessment for lead in lieu of a reference value
for multipathway risk rather than comparing the emissions to the National Ambient Air
Quality Standards for lead (0.15 micrograms/m3).
EPA's table summarizing emissions and dose-response values suggests that the Agency
did not evaluate the health risks at all from 2,2,4-trimethylpentane, phosphorus,
65 EPA, PB-HAP Compounds, Risk Assessment and Modeling - Air Toxics Risk Assessment Reference Library,
Vol. I Tech. Resource Manual, Ch. 4 Air Toxics: Chemicals, Sources, and Emissions Inventories, at 4-10, Exhibit 4-
2 (2004), http://www2.epa.gov/sites/production/files/2013-08/documents/volume_l_reflibrary.pdf.
47
-------
dibutylphthalate, 3,3'-dimethylbenzidine, 3,3'-dimethoxybenzidine, N,Ndimethylanline,
p-Phenylenadiamine. The risk from these compounds must be fully assessed.
Consistent with the 2004 guidance document and OEHHA 2012 Guidelines, EPA must
recognize the deposition of persistent HAPs as a source of soil contamination presenting
a potentially significant route of exposure, particularly for children. The commenter
argued that the EPA has been relying on outdated estimates of incidental soil ingestion
exposures and EPA must update these values to ensure that it considers the urban child
scenario in its multipathway risk assessment.66 The risk assessment of exposure to soil
contaminants should evaluate both direct exposure, hand-to-mouth, and indirect, object-
to-mouth exposure as a study has found a statistically significant positive correlation
between the frequency of object or food in mouth activity and blood lead levels,
according to the commenter.67 The commenter added that the 2011 update to EPA's
Exposure Factors Handbook includes more recent studies and estimates of hand-to-mouth
behavior, which must be used to assess risks from exposures to contaminated soils.68
EPA's analysis shows that the highest level of multipathway risk, to which the most-
exposed individual is exposed, may well be above the risk estimated from EPA's more
refined multipathway case study of the Marathon Petroleum refinery, near Garyville in
St. John the Baptist Parish, Louisiana.
Furthermore, the commenter stated that EPA must account for the aggregate impact of inhalation
and multipathway cancer and chronic non-cancer risk by adding each type of similar risk
together for all pollutants. The commenter stated that the purpose of the multipathway
assessment is to allow EPA to look at a person's exposure overall - not just inhalation or other
exposure pathways, in isolation. According to the commenter, failing to add up each type of risk
in order to come up with a total cancer risk number and a total non-cancer number, and then a
cumulative burden metric, makes EPA's overall risk assessment incomplete. The commenter
asserted that because of these deficiencies, EPA's analysis underestimates health risks from
refineries and thus EPA's proposed decision not to set residual risk standards for any part of this
source category other than storage vessels is flawed.
66 As an additional problem, California's lead in soil standard is more stringent than EPA's due to more recent
science on the harm of lead exposure. EPA has recognized that its standard is based on out-dated information about
lead, that previously assumed children's blood-lead levels below 10.0 ng/dL was safe. EPA now admits that number
is not protective, but has not updated its soil standard. See, e.g., "EPA fails to revise key lead-poisoning hazard
standards," USA Today (Mar. 10, 2013), http://www.usatodav.eom/storv/news/nation/2013/03/10/epa-hasnot-
revised-lead-hazard-standards-for-dust-and-soil/1971209 ("The EPA has not revised key hazard standards that
protect children from lead poisoning since 2001, despite science showing harms at far lower levels of exposure than
previously believed."); Children's Health Advisory Protection Comm., Letter to Administrator Jackson Regarding
Childhood Lead Poisoning (Mar. 29, 2012),
http://vosemite.epa.gov/ochp/ochpweb.nsf/content/chpac childhood lead poison lctter.htm.
67 Ko, S., Schaefer et al., Relationships of Video Assessments of Touching and Mouthing Behaviors During Outdoor
Play in Urban Residential Yards to Parental Perceptions of Child Behaviors and Blood Lead Levels, 17 J. of
Exposure Science and Environ. Epidemiology 47 (2007).
68 EPA, Exposure Factors Handbook, 2011 Edition (http://cfjpub.epa.gov/ncea/risk/recordisplay.cfm?deid=236252).
48
-------
Response 1: Regarding the commenter's assertion that we did not base the multipathway risk
assessment on allowable emissions, we believe it is reasonable for the multipathway risk
assessment to be based on actual emissions for this source category, and not the allowable level
of emissions that facilities are permitted to emit. The uncertainties associated with the
multipathway screen along with uncertainties in the allowable emissions estimates would make a
multipathway risk assessment based on allowable emissions highly uncertain. The ICR-
submitted information for allowable emissions did not include emission estimates for all HAP
and sources of interest. Consequently, we used our REM model to estimate allowable emissions.
The REM model relies on model plants that vary based on throughput capacity. Each model
plant contains process-specific default emission factors, adjusted for compliance with the
Refinery MACT 1 and 2 emission standards. Basing the multipathway assessment on these
modeled emissions along with the uncertainties inherent in the multipathway screening analysis
would result in a risk estimate that is too uncertain to support a regulatory decision.
Naphthalene is not treated as a POM chemical in the multipathway analyses. Naphthalene is
listed individually as a HAP under section 112(b) of the CAA. POM also is listed as a HAP
under section 112(b) and is defined as organic compounds with more than one benzene ring and
a boiling point greater than or equal to 100 degrees Celsius. Under this definition, naphthalene
potentially could be considered as part of the POM listing. However, naphthalene is short-lived
in environmental media due to its tendency to volatilize and biodegrade and, consequently, will
not build up in environmental media over time.69 It has a moderate affinity for lipids and will
undergo short-term bioaccumulation is tissues; however, biochemical processes lead to its
biodegradation and elimination. Because it is neither persistent nor bioaccumulative, we do not
consider it a PB-HAP, and its inclusion as POM is inappropriate and would result in less
accurate and less meaningful estimates of media concentrations and multipathway risk.
While we acknowledge we do not have screening values for some of the PB-HAP, we do not
agree that this results in an inadequate multipathway assessment. In the Air Toxics Risk
Assessment Reference Library,70 we developed the current PB-HAP list considering all of the
available information on persistence and bioaccumulation (see http://www2.epa.gov/fera/air-
toxics-risk-assessment-reference-library-volumes-1-3), specifically Volume 1 Appendix D). This
list considered HAP identified as PB-HAP by other EPA Program Offices (e.g., the Great Waters
Program), as well as information from the persistent, bioaccumulative and toxic (PBT) profiler
(see http://www.pbtprofiler.net/). This list was peer-reviewed by the SAB and found to be
reasonable for use in the RTR program. Based on these sources and the available information on
the persistence and bioaccumulation of other HAP, we do not believe that the potential for
multipathway risk from other HAP rises to the level of the PB-HAP currently on the list.
69 US Agency for Toxic Substances and Disease Registry, 2005. Toxicological Profile for Naphthalene, 1-
Methylnaphthalene, and 2-Methylnaphthalene.
http://www.atsdr.cdc.gov/ToxProfiles/tp.asp?id=240&tid=43#bookmark08
70 The Air Toxics Risk Assessment Reference Library provides information on the fundamental principles of risk-
based assessment for air toxics and how to apply those principles in different settings (e.g., facility-specific) as well
as strategies for reducing risk at the local level.
49
-------
We have combined risk assessment results to the extent that it is appropriate to do so. We
consider the effect of mixtures of carcinogens and use a TOSHI approach for our chronic
noncancer assessments. We do not use a TOSHI approach for our acute analyses nor do we
combine the results of our inhalation and multipathway assessments. The multipathway risk
assessment conducted for proposal was a screening-level assessment followed by a more refined
site-specific assessment. The screening assessment used highly conservative assumptions
designed to ensure that facilities with results below the screening threshold values did not have
the potential for multipathway impacts of concern. The screening scenario is a hypothetical
scenario, and due to the theoretical construct of the screening model, exceedances of the
thresholds are not directly translatable into estimates of risk or hazard quotients for these
facilities. In situations where we have previously summed the inhalation and multipathway
assessment results (e.g., secondary lead smelters), it was for two refined assessments, and still
serves as a conservative overestimate of potential risks given that it is highly unlikely that the
same receptor has the maximum results in both assessments. For the petroleum refinery source
categories, the refined multipathway analysis predicts a potential lifetime cancer risk of 4-in-l
million to the maximum most exposed individual (MIR). The non-cancer HQ is predicted to be
well below 1 for all target organs. The chronic inhalation cancer risk assessment estimated
inhalation cancer risk around this same facility to be approximately 10-in-l million, due in large
part to emissions of naphthalene and 2-methylnaphthalene (both are not persistent,
bioaccumulative, and toxic (non-PBT) HAP). Thus, although highly unlikely, if around this
facility the person with the highest chronic inhalation cancer risk is also the same person with the
highest individual multipathway cancer risk, then the combined, worst-case MIR for that facility
could theoretically still be 10-in-l million (risk estimates are expressed as 1 significant figure).
While this refined assessment was performed on only a single facility, the results of this single
refined analysis indicate that if refined analyses were performed for other sites, the risk estimates
would consistently be lower than those estimated by the Tier II analysis. In addition, the risks
predicted by the multipathway analyses at most facilities are considerably lower than the risk
estimates predicted by the inhalation assessment, indicating that the inhalation risk results are in
all likelihood the primary factor in our residual risk determination for this source category.
We disagree with the commenter that a separate multipathway risk assessment should have been
performed for lead. As noted in previous responses, the NAAQS for lead was used as a health
based standard in this review. It was not simply adopted mechanically, but rather is justified
under the independent decision framework in section 112(f)(2). We note that the NAAQS for
lead was set to protect, with an adequate margin of safety, the health of the most susceptible
children and other potentially at-risk populations against an array of adverse health effects, most
notably including neurological effects, particularly neurobehavioral and neurocognitive effects
(which are the most effects to which children are most sensitive) (73 FR at 67007). We further
note that in developing the NAAQS for lead, air-related multipathway effects were already taken
into account. That is, as noted at 73 FR at 66971: "As was true in the setting of the current
standard, multimedia distribution of and multipathway exposure to Pb that has been emitted into
the ambient air play a key role in the Agency's consideration of the Pb NAAQS."
In addition, the EPA's analysis addresses the cumulative, long-term impacts to individuals in the
local affected communities from prior lead emissions. First, as part of the Risk and Exposure
Assessment supporting the lead primary NAAQS, the EPA assessed the IQ loss associated with
50
-------
the cumulative environmental impact of long-term emissions in communities living near
stationary sources of lead emissions, specifically lead smelters. See Lead Human Exposure and
Health Risk Assessments for Selected Case Studies; Volume I Human Exposure and Health Risk
Assessments 00 Full Scale; and Volume II: Appendices ((USEPA, 2007) (Office of Air Quality
Planning and Standards, Research Triangle Park, NC EPA-432/R-07-014a and EPA-452/R-07-
014b; see also 73 FR 66979-66983. The Risk and Exposure Assessment assessed exposure to
lead from the inhalation route as well as the multiple ingestion pathways. Although the EPA
relied primarily upon the evidence-based framework (i.e. epidemiological studies and related
information) in developing the lead primary NAAQS, the EPA nonetheless found the air-related
lead risk estimates from the Risk and Exposure Assessment to be generally supportive of that
scientific evidence-based framework. (See 73 FR at 67003-04.)
The evidence-based framework used in developing the lead NAAQS is focused on children with
total lead exposure closer to the current U.S. average (than the more highly lead exposed upper
percentiles of the population), since these children will have greater response to each increment
of air-related lead exposure than would children with higher overall blood lead levels. Other
populations, including children with higher blood lead levels due to past exposures, will
necessarily be accorded protection from air-related lead as well. (See EPA-HQ-OAR-2006-0735-
5894 at page 13).
As noted in the risk assessment document, there is no RfD or other comparable chronic health
benchmark value for lead compounds. In 1988, the EPA's IRIS program reviewed the health
effects data regarding lead and its inorganic compounds and determined that it would be
inappropriate to develop an RfD for these compounds, saying, "A great deal of information on
the health effects of lead has been obtained through decades of medical observation and
scientific research. This information has been assessed in the development of air and water
quality criteria by the agency's Office of Health and Environmental Assessment (OHEA) in
support of regulatory decision-making by the Office of Air Quality Planning and Standards
(OAQPS) and by the Office of Drinking Water (ODW). By comparison to most other
environmental toxicants, the degree of uncertainty about the health effects of lead is quite low. It
appears that some of these effects, particularly changes in the levels of certain blood enzymes
and in aspects of children's neurobehavioral development, may occur at blood lead levels so low
as to be essentially without a threshold. The agency's RfD Work Group discussed inorganic lead
(and lead compounds) at two meetings (07/08/1985 and 07/22/1985) and considered it
inappropriate to develop an RfD for inorganic lead." The EPA's IRIS assessment for Lead and
compounds (inorganic) (CASRN 7439-92-1), http://www.epa.gov/iris/subst/0277.htm.
Regarding incidental soil contamination, the EPA has estimated risks from both direct and
indirect pathways. Direct routes of exposure include direct ingestion of soil and exposure
through dermal contact. However, because it has been demonstrated in past analyses that
exposure levels associated with dermal contact are but a small fraction of exposure levels
associated with ingestion and inhalation pathways, dermal exposures were not assessed for this
source category.
51
-------
2.3.6 Recommendations for strengthening risk assessment
Comment 1: Suggestions for remedying claimed deficiencies in risk assessment: Several
commenters claimed there are shortcomings in EPA's risk assessment, as detailed in above
sections, and suggested improvements. These suggestions which have been included elsewhere
in this chapter of the RTC are summarized in the following list:
Assess and account for receptors with increased vulnerability such as children, seniors,
pregnant woman, prenatal fetuses, persons with pre-existing health conditions, disabled
and Environmental Justice Communities through the use of specific risk adjustment
factors and reference values.
Fully integrate the environmental justice analysis into the risk assessment and rulemaking
set more protective standards.
Assess and aggregate exposure from multiple pathways including by adding inhalation
and non-inhalation-based cancer risks.
Use a margin of exposure (MOE) framework for non-cancer impacts and adjust the target
MOE according to known vulnerability factors.
Account for exposure to multiple sources of multiple pollutants via multiple pathways to
assess the total cumulative risk, and use the estimated impacts of these cumulative risks
to shift the level of risk which triggers policy action.
Reduce the acceptable lifetime cancer risk.
Base risk assessments on the best available scientific research including incorporating
recommendations from NAS and SAB.
Response 1: This compilation of suggestions for improving the perceived deficiencies in our
residual risk assessment have been address in other responses in Sections 2.2 through 2.7 of this
document. Further, it is important to note that the EPA's approach to performing risks
assessments in support of residual risk program has been reviewed and supported by the SAB
several times over the past 17 years. First, in 1998 they examined our analytical and policy
approach for assessing residual risk from hazardous air pollutants emitted from stationary
sources, followed by a second review in 2000 to verify that our application to a specific source
category was consistent with the approved approach. A third SAB consultation in 2006 focused
on development of emissions inventories for source categories and updated methods for
characterizing human exposure and risks. Again in 2009, the SAB reviewed and supported our
updated and expanded air toxics risk assessment methods, including our multipathway
assessment, refinement of acute risk screening, and the methods of assessing potential
environmental risk. We will continue to seek SAB consultation as our risk assessment methods
develop and revise our approach as appropriate.
Comment 2: EPA must undertake a cumulative risk assessment: Many commenters stated that
EPA needs to perform an up-to-date health risk and impact assessment of the cumulative effects
refinery chemicals on communities and children. Commenters stated that this assessment should
evaluate the combined impact of each type of risk from multiple pollutants and assess the total
cumulative risk burden from all pollutants to make an ample margin of safety determination. The
commenter noted that in the refineries risk assessment, EPA only assesses the combined impact
of cancer risk and chronic non-cancer risk that operates on the same target organ, while the
52
-------
commenter argued that the EPA should assess the total and synergistic cancer risk and total
chronic noncancer risk for different pollutants. In addition, the commenter noted that EPA
should apply these same principles to create a mechanism for assessing the total acute risk to
chemical mixtures, such as the TOSHI for chronic risk that aggregates the acute impacts on the
same organ systems for all pollutants.
The commenter asserted that consistent with NAS recommendations, the EPA must create a
metric to assess the total and cumulative risk burden, rather than only looking at each type of risk
separately. After first assessing the total cancer, chronic non-cancer, and acute risks, for both
inhalation and multipathway exposure, the commenter argued that EPA also must create a metric
to assess the total risk. EPA must aggregate health risk for each pollutant, and each type of health
risk, according to the commenter, to create a cumulative risk determination for the individual
"most exposed" to emissions as section 112(f)(2) of the Act requires. The commenter asserted
that without performing this cumulative assessment, EPA has failed to gather the information
needed to assess whether the risk to public health is acceptable under section 112(f)(2).
Response 2: We have combined risk assessment results to the extent that it is appropriate to do
so. We consider the effect of mixtures of carcinogens and use a TOSHI approach for our chronic
noncancer assessments. We do not use a TOSHI approach for our acute analyses nor do we
combine the results of our inhalation and multipathway assessments.
For the petroleum refinery source categories, the EPA conducted an assessment of the
cumulative cancer risks from emitted carcinogens and the cumulative noncancer hazard indices
from all emitted non-carcinogens affecting the same target organ system for both the source
category emissions and the facility-wide emissions.
Concerning comments that we should consider aggregate risks from multiple pollutants and
sources, we note that we have done this to the extent it is appropriate to do so. We modeled
whole-facility risks for both chronic cancer and non-cancer impacts to understand the risk
contribution of the sources within the Petroleum Refinery source categories. The individual
cancer risks for the source categories were aggregated for all carcinogens. In assessing noncancer
hazard from chronic exposures for pollutants that have similar modes of action or (where this
information is absent) that affect the same target organ, we aggregated the HQ. This process
creates, for each target organ, a TOSHI, defined as the sum of hazard quotients for individual
HAP that affect the same organ or organ system. Whole facility risks were estimated based on
the 2011 ICR data obtained from facilities, which included emissions from all sources at the
refinery, not just Refinery MACT 1 and 2 emission sources (e.g., emissions were included for
combustion units and units subject to the Hazardous Organic NESHAP, if present at the
refinery).
As described in the Draft Residual Risk Assessment for the Petroleum Refining Source Sector
(Docket Item No. EPA-HQ-OAR-2010-0682-0225), we do not sum results of the acute
noncancer inhalation assessment to create a combined acute risk number that would represent the
total acute risk for all pollutants that act in a similar way on the same organ system or systems
(similar to the chronic TOSHI). The worst-case acute screen is a conservative scenario. That is,
the acute screening scenario assumes worst-case meteorology, peak emissions for all emission
53
-------
points occurring concurrently and an individual being located at the site of maximum
concentration for an hour. Thus, as noted in Section 2.4 of the Draft Residual Risk Assessment
for the Petroleum Refining Source Sector (Docket Item No. EPA-HQ-OAR-2010-0682-0225),
"because of the conservative nature of the acute inhalation screening and the variable nature of
emissions and potential exposures, acute impacts were screened on an individual pollutant basis,
not using the TOSHI approach."
We did not combine the inhalation and multipathway risk assessment results for each pollutant
because it is inappropriate to do so. The multipathway risk assessment conducted for proposal
was a screening-level assessment followed by a more refined site-specific assessment. The
screening assessment used highly conservative assumptions designed to ensure that facilities
with results below the screening threshold values did not have the potential for multipathway
impacts of concern. The screening scenario is a hypothetical scenario, and due to the theoretical
construct of the screening model, exceedances of the thresholds are not directly translatable into
estimates of risk or hazard quotients for these facilities. For the refined multipathway analysis, it
is unlikely that the person with the highest chronic inhalation cancer risk is also the same person
with the highest individual multipathway cancer risk. Also, in our experience, the risk estimates
predicted by the multipathway analysis at most facilities are considerably lower than the risk
estimated predicted by the inhalation assessment, indicating that the inhalation risk results are in
all likelihood the primary factor in our residual risk determination.
2.4 Demographic Analysis / Environmental justice
Comment 1: Lower income, minority, less educated. non-English speaking and younger
communities are disproportionately impacted by refinery emissions: One commenter stated that
the emissions from the 149 petroleum refineries in 32 states pose serious harm to millions of
Americans who live nearby or downwind of these facilities. The commenter noted that among
the approximately 5.2 million people exposed to levels of carcinogens from refineries, fully half
are from minority groups, one quarter of those exposed are under the age 17 and one in five had
incomes below the poverty line. Furthermore, the commenter cited that one in four adults lacked
a high school diploma and that one in ten had no one in the household over age 14 who spoke
English. The commenter highlighted that a greater percentage of African-Americans, Hispanics
and people living below the poverty line face an increased risk of cancer from exposure to
emissions from refineries than do whites or those with higher incomes. Therefore, according to
the commenter, as supported by EPA's own analysis the harm that communities are exposed to is
not equal and a serious environmental and social injustice exists.71 In particular the commenter
noted, half of the people who currently face a cancer threat from refineries' pollution are racial
minorities even though the U.S. population is only 28% minority. The commenter stated that of
the approximate 4 million people who will still face an increased cancer threat from refineries
after the proposed rule, more than 50% will be racial minorities, 31% will be African Americans,
24%) will be Hispanic or Latino, and 22% will be people living below the poverty level. The
71 79 FR 36937; Analysis of Socio-Economic Factors for Populations Living Near Petroleum Refineries, EPA-HQ-
OAR-2010-0682-0226
54
-------
commenter stated that the disparity increases under the EPA's proposed rule.72 The commenter
concluded that the EPA must take stronger action than it has proposed to resolve the need for
robust national air toxics standards and to reduce the socioeconomic disparity in environmental
health impacts caused by U.S. refineries.
Another commenter stated that considerations of the cost/benefits to industry have been taken
into account, while the costs to those living near the fenceline of refineries have not, noting that
these populations often have a lower quality of health and do not have access to health insurance.
Response 1: With regards to the comment concerning unacceptability due to socioeconomic
disparity, under Executive Order (EO) 12898, the EPA is directed to the greatest extent
practicable and permitted by law, to make environmental justice (EJ) part of its mission by
identifying and addressing, as appropriate, disproportionately high and adverse human health or
environmental effects of its programs, policies and activities on minority populations and low-
income populations in the United States. Consistent with EO 12898 and the Presidential
Memorandum that accompanies it, the EPA's environmental justice policies promote justice by
focusing attention and EPA efforts on addressing the types of environmental harms and risks that
are prevalent among minority, low-income and indigenous populations. EO 12898 and the EPA's
environmental justice policies do not mandate particular outcomes from an action, but they
demand that decisions involving the action be informed by a consideration of EJ issues. With
respect to this rule, the EPA found the overall level of risk from the source categories to be
acceptable and to provide an ample margin of safety for all populations in close proximity to
these sources, including minority and low-income populations.
Because minority groups make up a large portion of the population living near refineries, as
compared with their representation nationwide, those groups would see a greater benefit from the
implementation of the controls required by this rule. For example, we estimate that after
implementation of the controls {i.e., post-controls), about 1,000,000 fewer people will be
exposed to cancer risks greater than 1-in-l million {i.e., 4,000,000 people). Further, we estimate
that approximately half of those who risk would be reduced to below 1-in-l million, or about
500,000 people, would be in a minority demographic group.
Comment 2: Specific examples of communities cited by commenters as having environmental
justice concerns: Commenters expressed environmental justice concerns for several communities
where people of low income, less education, and/or communities of color live within close
proximity to a refinery fenceline, particularly for those living in Manchester, Pasadena, Galena
Park, Baytown, Deer Park, Port Arthur, Beaumont, and Texas City which are located near the
largest petrochemical complex in the nation. The commenter stated that these communities suffer
the burdens of industry including health effects (e.g., higher than normal incidence of cancer,
72 Draft Residual Risk Assessment for the Petroleum Refining Source Sector, EPA-HQ-OAR-2010-0682-0225;
Analysis of Socio-Economic Factors for Populations Living Near Petroleum Refineries, EPA-HQ-OAR-2010-0682-
0226; Analysis of Socio-Economic Factors for Populations Living Near Petroleum Refineries Post Control Scenario
at 8, EPA-HQ-OAR-2010-0682-0227.
55
-------
asthma, chronic bronchitis, and other negative respiratory health outcomes) without enjoying the
benefits (e.g., highly skilled and technical jobs).
Another commenter provided a report on the risks refineries pose to Pasadena, a city east of
Houston,that has dozens of chemical plants and a refinery. The commenter stated community
members from this area have issues with asthma and other health problems, but have limited
access to health care outside of a single public clinic that serves a huge swath of this portion of
the Houston Ship Channel. The commenter stated that it is especially important to strengthen
rules considering the vulnerable demographics of the area combined with poor air quality in the
region.
Additional commenters referenced a 2006 study by the University Of Texas School Of Public
Health concluding that children living within 2 miles of the Houston Ship Channel, which is
home to five oil refineries, have a 56 percent greater risk of leukemia than children living 10
miles away. One of these commenters also referenced a 2008 study published in the
journal Environmental Science and Technology that found for residents of the eastern portion of
the Houston Ship Channel, the greatest contributor to an increased risk of cancer is point-
source emissions from petrochemical refineries. Another commenter stated that there are 86
schools in the Houston Independent School District which are in the top 5 percent of the most
toxic schools in the nation according to reports by the Center for Health and Environmental
Justice.
Response 2: We are not able to determine the baseline health status of individuals or
communities in a national rulemaking. Individual privacy issues as they relate to health records
and the costs that would be associated with such an analysis make the analysis infeasible.
Through the EPA's interim guidance on Environmental Justice and the Action Development
Process, the agency is encouraging rule writers and policy makers to look at the whole range of
factors that impact communities and population groups when crafting rules. The EPA is
continuing to discuss and pilot approaches that are consistent with the agency's responsibilities
regarding environmental justice as outlined in EO 12898. In determining the need for tighter
residual risk standards, the EPA strives to limit to no higher than 100-in-l million the estimated
cancer risk for persons living near a plant if exposed to the maximum pollutant concentration for
70 years and to protect the greatest number of persons to an individual lifetime risk of no higher
than 1-in-l million. Considerations are made for all people regardless of racial or socioeconomic
status.
To examine the potential for any environmental justice issues that might be associated with the
source categories, we performed a demographic analysis of the population close to the facilities.
In this analysis, we evaluated the distribution of HAP-related cancer and non-cancer risks from
petroleum refineries across different social, demographic, and economic groups within the
populations living near facilities identified as having the highest risks. The methodology and the
results of the demographic analyses are included in a technical report, Risk and Technology
ReviewAnalysis of Socio-Economic Factors for Populations Living Near Petroleum Refineries,
available in the docket for this action (Docket Item Number EPA-HQ-OAR-2010-0682-0226).
These results, for various demographic groups, are based on the estimated risks from actual
emissions levels for the population living within 50 kilometers (km) of the facilities.
56
-------
The results of the demographic analysis indicate that emissions from petroleum refineries expose
approximately 5,000,000 people to a cancer risk at or above 1-in-l million. Implementation of
the provisions is expected to reduce the number of people estimated to have a cancer risk greater
than 1-in-l million due to HAP emissions from these sources from 5,000,000 people to about
4,000,000. Our analysis of the demographics of the population within 50 km of the facilities
indicates potential disparities in certain demographic groups, including the African American,
Other and Multiracial, Hispanic, Below the Poverty Level, and Over 25 without a High School
Diploma. The population living within 50 km of the 142 petroleum refineries has a higher
percentage of minority, lower income and lower education persons when compared to the
nationwide percentages of those groups. For example, 50 percent are in one or more minority
demographic group, compared to 28 percent nationwide. As noted above, approximately
5,000,000 people currently living within 50 km of a petroleum refinery have a cancer risk greater
than 1-in-l million. We would expect that half of those people are in one or more minority
demographic groups.
Because minority groups make up a large portion of the population living near refineries, as
compared with their representation nationwide, those groups would similarly see a greater
benefit from the implementation of the controls. For example, we estimate that after
implementation of the controls proposed in this action {i.e., post-controls), about 1,000,000 fewer
people will be exposed to cancer risks greater than 1-in-l million {i.e., 4,000,000 people).
Further, we estimate that approximately 500,000 people no longer exposed to a cancer risk
greater than 1-in-l million would be in a minority demographic group.
Although the EPA's fenceline monitoring requirement is intended to ensure that owners and
operators monitor, manage and, if necessary, reduce fugitive emissions of HAP, we also expect
the collected fenceline data to help the EPA understand and identify emissions of benzene and
other fugitive emissions that are impacting communities in close proximity to the facility. While
currently-available emissions and monitoring data do not indicate that risks to nearby
populations are unacceptable, we recognize that the collection of additional data through routine
fenceline monitoring can provide important information to communities concerned with potential
risks associated with emissions from fugitive sources. We note that the data we will collect on a
quarterly basis may include exceedances of the fenceline action level that a facility could have
addressed or could still be actively addressing at the time of the report. Requiring the electronic
reporting of fenceline monitoring data on a quarterly basis will ensure that communities have
access to data on benzene levels near the facility, which is directly relevant to the potential health
risks posed by the facility. The requirements for fenceline monitoring and corrective action when
fugitive emissions from a facility exceed the specified corrective action level will serve as an
important backstop to protect the health of the populations surrounding the facility, including
minority and low-income populations.
Comment 3: Recommendations for improvements to the environmental justice demographic
analysis: Commenters made several recommendations for improvements to the environmental
justice demographic analysis including:
Performing an environmental justice analysis more similar to the recent work done on
the definition of solid waste.
57
-------
Considering other factors in terms of the overall disproportionate impact on communities
(e.g., a complete risk profile) including: susceptibility of the community, the number of
children, vulnerable populations, higher disease rates, the ability of the community
to participate in decision-making or receiving information, lack of information,
language barriers, lack of social capital, the ability of the community to prepare for
impacts (such as evacuation issues in emergencies), the ability of the community to
recover from environmental insults such as the lack of health care and the lack of
financial resources, the potential impacts of land use on the community, the availability
of recreation and other enrichment opportunities, cumulative impacts (which is critically
important and may include all of the above), the distribution of environmental burdens,;
and the location of other potential sites of pollution (such as Superfund sites and
hazardous waste facilities within the same community, and their compliance rates,
particularly comparing facilities subjected to other regulations).
Develop more conservative resulting emission standards because, according to the
commenter, a California EPA screening tool which identifies California communities that
are disproportionately burdened by multiple sources of pollution as determined by
evaluating multiple pollutants and stressors in these communities shows that those
identified as being threatened by the burden of serious pollution are located in the same
locations as the majority of California's refineries.
Consistent with NAS recommendations, the EPA should consider the use of
socioeconomic factors as part of EPA's consideration of both vulnerability and
variability, as core elements of the risk assessment.73 According to the commenter, data
describing these factors are available from the CDC's Environmental Public Health
Tracking Program, the U.S. ATSDR, state and local health agencies, and academic
researchers.
Better account for other types of human variability such as genetics and baseline health
status, as recommended by NAS and current science. According to the commenter,
socioeconomic status has been shown to act as a proxy for other types of human
variability to chemical risk that EPA has not adequately addressed in its draft risk
assessment for the refineries rule.
In addition to looking at the demographic census data on race, ethnicity, poverty level,
and similar factors, EPA must assess the starting point or baseline overall health status of
the affected individuals and communities using the best available data at a local and
national level, including the baseline cancer levels, respiratory problems, and health
problems associated with the toxic chemicals emitted by a source category. The
commenter asserted that doing so would be consistent with the 1999 Residual Risk
Report74 and would also follow EPA's own statements (in the 2014 Second Integrated
73 NAS 2009, supra note 264, at 109-10 & tbl. 4-1 (describing the need to consider increased susceptibility due to
prior and concurrent exposures; and to 'social and economic factors'); id. at 220-21 (describing ways to assess
cumulative risk including by consideration of "epidemiologic concepts" and information, and by considering "what
the burden of disease is in the context of simultaneous exposure to a number of stressors"); id. at 230 (discussing the
role of epidemiology and surveillance data).
74 U.S. EPA, "Residual Risk Report to Congress" at 42, 67 (Mar. 1999), EPA-453/R-99-00 (discussing factor of
"overall health" and recognizing the need to consider sensitive subpopulations that "consist of a specific set of
58
-------
Urban Air Toxics Report) that more work is needed to reduce excess cancer risks in
urban areas that continue to face elevated risks [Second Integrated Urban Air Toxics
Report, at xiv]. As an alternative, EPA could use a default factor to account for
socioeconomic and other community-based stressors.
Better communicate and document the findings of the analysis, as one commenter stated,
the EPA has presented environmental justice numbers in the abstract, which makes it
difficult for communities to understand the risk assessment and the EPA's findings.
Conduct a demographic analysis on individuals projected to experience a risk greater than
1-in-1-million for cancer or an HQ above one and on individuals living within five
kilometers of the facility, regardless of projected risk, consistent with the approach used
for the Hard and Decorative Chromium Electroplating and Chromium Anodizing Tanks
source category.
EPA rule writers should consult with the EPA Office of Environmental Justice to develop
criteria and specific guidance on how to interpret and apply the outcome of these types of
analyses in the rulemaking process.
Response 3: The EPA continues to evaluate and refine environmental justice analysis, such as
the proximity analysis performed for the Definition of Solid Waste. The draft EPA document
Technical Guidance for Assessing Environmental Justice in Regulatory Analysis is currently
undergoing SAB review. The Office of Air and Radiation is contributor to this draft document
and cites numerous examples of EJ analysis performed in agency rulemakings. Environment
justice considerations are a part of the rulemaking process; however, various factors influence
the scope and complexity of an assessment. These factors may include, but are not limited to
statutory mandates, data availability, resources and/or timeframe limitations.
For this rulemaking, the EPA conducted both pre- and post- control risk-based assessment with
analysis of various socio-economic factors for populations living near petroleum refineries.
While partially proximity-based, this assessment also uses air quality modeling HEM-3 to show
a decrease in potential cancer risks for all populations. These reports can be found at the
consolidated petroleum refinery rulemaking repository at:
http://www.epa.giv/airtoxics/petref.html. In addition, the EPA's meaningful involvement
activities included webinars, community calls and all day trainings held in San Francisco,
California, New Orleans, Louisiana and Research Triangle Park, North Carolina. Additionally,
we have held a number of meetings with community organizations on this rulemaking.
The commenter is correct that we performed the demographic analyses for the petroleum
refinery source categories differently than we did for the October 2010 proposed RTR for the
chrome source categories. We performed the demographic analyses for the chrome source
categories using two approaches as examples of how such analyses might be developed, and
invited public comment on the approaches used and the interpretations made from the results. In
the first approach, we focused the analysis on the total populations residing within 5 km of each
facility, regardless of their estimated risks, and examined the distributions across various
demographic groups within those 5 km circles. That analysis was a "proximity" analysis in that it
individuals who are particularly susceptible to adverse health effects because of physiological (e.g., age, gender, pre-
existing conditions), socioeconomic (e.g., nutrition), or demographic variables, or significantly greater levels of
exposure," based on various demographic factors).
59
-------
considered only the distance from the emission sources to surrounding populations, and not the
estimated risks to those populations.
In the second approach, we focused the analysis on the populations within 5 km of any facility
estimated to have exposures to HAP which result in cancer risks of 1-in-l million or greater or
non-cancer hazard indices of 1 or greater. We examined the distributions of those risks across
various demographic groups. In each approach, we compared the percentages of particular
demographic groups to the total number of people in those demographic groups nationwide. We
stated in the proposed rule for the chrome source categories that in future rules we planned to
extend the analyses to cover the entire modeled domain for a facility (50 km radius) to capture
all individuals with risks above 1-in-l million or greater or non-cancer hazard indices of 1 or
greater. We also stated that generally we have found that using a 5 km radius in the analysis will
capture more than 90 percent of the individuals with cancer risks above 1-in-l million.
In the February 2012 supplemental proposal for the RTR for the chrome source categories, our
demographic analyses included populations within 50 km of each source (including those very
near the sources) with risks of 1-in-l million or greater or non-cancer hazard indices of 1 or
greater (77 FR 6628, Feb. 8, 2012). We did not include analyses using a 5 km radius in that
supplemental proposal nor in the RTR proposal for the petroleum refinery source categories.
Where a risk assessment has been performed, it is more informative to consider the
demographics of all populations (including those beyond 5 km) with elevated risks than to limit
the demographics analysis to populations located within 5 km of a facility. Where a risk
assessment has been performed, these populations are identified, and the source parameters are
taken into account. As discussed above, we have found that most exposure locations with the
highest estimated risks are within 5 km of a facility, so extending the radius to 50 km has little
impact on an analysis based on risks, but makes more sense because 50 km corresponds to the
risk modeling radius and includes all populations with elevated risk estimates. We also note that
we are working with the Office of Environmental Justice in an ongoing effort to develop new
tools for considering environmental justice in rulemakings.
In the Urban Air Toxics Strategy Report to Congress we acknowledge that national rules and
standards can address part of the risk to communities, but because the assessments did not
include background risks or contributions to risk from sources outside the facilities more needs
to be done at the community level with other tools available within the CAA and within state,
local, and other federal programs. EPA is committed to our efforts to make a difference in
communities of concern and developing an integrated strategy focusing work in communities
with the most need for EPA's assistance. We have been, and will continue to work in thousands
of communities across the country. For example, through community-scale grants, we provided
funding and technical support to the City of Philadelphia over a 2 year period in which we
deployed numerous technologies in the South Philadelphia communities bordering a refinery to
investigate how passive samplers and sensor based, stand-alone air measures can help improve
information on air pollutant concentrations in areas that have many potential sources. The study
has found that low cost passive samplers and sensors are effective in determining the origins of
pollutants in communities that have a number of potential sources. Furthermore, these low cost
passive sensors are able to determine emissions of multiple pollutants at low concentrations.
Some of the information that resulted from this study is also discussed in Chapter 8. Over the
60
-------
next two years we will look for opportunities to enhance our partnership with communities to
strengthen and improve their health - both environmental and economic. This effort to enhance
coordination across our EPA programs and with other federal agencies will improve how we
support community needs. We will focus on those communities where we think we have
opportunities to leverage resources and actions to make a real difference. As we learn lessons on
coordinating and focusing our efforts, we will use these lessons to help more communities in the
future.
Comment 4: Consideration of environmental justice issues are outside the scope of EPA
authority: One commenter stated that EPA's consideration of environmental justice issues are
outside the scope of its authority under section 112(f). The commenter noted that section
112(f)(2)(A) expressly instructs EPA to impose additional emissions controls if needed to
provide an ample margin of safety "to protect public health." Although the term "public health"
is not defined in section 112 or in EPA's part 63 regulations, the commenter argued such a
definition would not include non-health related demographic factors and EPA's analysis is,
therefore, inapplicable to its statutory mandate to protect public health. The commenter provided
an example of how this term is used in the context of EPA's NAAQS program. The commenter
stated in this context, the term "public health" should be dictated by the meaning of the word
"public" which, according to the commenter means "of, relating to, or affecting all of the people
or the whole area of a nation or state" and "of or relating to people in general." These definitions
emphasize, according to the commenter, that the word "public" should be construed expansively
as describing the people as a whole, and not particular demographic segments.
The commenter further stated that EPA's established approach to assessing potential impacts on
public health under the NAAQS program is consistent with this meaning. The EPA reasonably
interprets the term "public health" to include consideration not only of potential impacts to the
population as a whole, but also to sensitive subpopulations, according to the commenter,
recognizing that the objective is to protect the group rather than any particular individual in the
group. See, e.g., 71 Fed. Reg. 61144, 61145 fn. 2 (Oct. 17, 2006). The commenter asserts that
sensitive subpopulations are identified according to their particular health-based sensitivities
(e.g., asthmatics), rather than demographic classifications unrelated to particular health-based
sensitivities. According to the commenter, with this backdrop, it would not be reasonable to
construe the term "public health" as used in section 112(f) as allowing consideration of
demographic classifications that bear no relationship to the potential health effects presented by
the HAPs at issue for the given source category or subcategory.
Response 4: As noted above, the EPA is continuing to discuss and pilot approaches that are
consistent with the agency's responsibilities regarding environmental justice as outlined in EO
12898. The EPA defines "environmental justice" to mean fair treatment and meaningful
involvement of all people, and this definition represents a commitment to ensuring that EPA
works to improve conditions affecting the public health of all Americans so that everyone has
access to clean water, clean air and healthy communities.
As stated in the Benzene NESHAP ((54 FR 38044, 38061, September 14, 1989), in determining
the need for residual risk standards, we strive to limit to no higher than approximately l-in-10
thousand (100-in-l million) the estimated cancer risk that a person living near a plant would
61
-------
have if he or she were exposed to the maximum pollutant concentrations for 70 years and, in the
ample margin of safety decision, to protect the greatest number of persons possible to an
individual lifetime risk level of no higher than approximately 1 -in-1 million. These
considerations are made for all people, regardless of racial or socioeconomic status. However, in
determining whether to require additional standards under section 112(f), these levels are not
considered rigid lines, and we weigh the cancer risk values with a series of other health measures
and factors in both the decision regarding risk acceptability and in the ample margin of safety
determination. We also consider cost of controls in the ample margin of safety determination.
While we analyzed the demographic makeup of the at-risk populations surrounding the facilities
within the source categories addressed by this rulemaking, we ultimately concluded that the risks
were acceptable for all population groups and protected public health with an ample margin of
safety. Thus, while the results of our demographic analyses served to provide information about
the demographic makeup of the populations exposed to the source category, the EPA did not
base its decision solely on these analyses.
2.5 Ecological risk assessment
Comment 1: Additional pollutants should be considered in ecological risk assessment: One
commenter noted that EPA acknowledged in the proposal that there are HAPs beyond the seven
the agency evaluated in the environmental risk screening assessment that "may have the potential
to cause adverse environmental effects" [79 FR 36898], stating that additional HAPs may be
evaluated in the future "as modeling science and resources allow." The commenter strongly
urged EPA to evaluate additional pollutants that are emitted by this source category, including
arsenic and nickel, and ensure that measures are undertaken to reduce the public's exposure to
them.
Response 1: The environmental screen focuses on the following seven environmental HAP: five
PB-HAP - cadmium, dioxins/furans POM, mercury (both inorganic mercury and methyl
mercury) and lead; and two acid gases - HC1, and HF. HAP that persist and bioaccumulate are of
particular environmental concern because they accumulate in the soil, sediment and water. The
five PB-HAP we evaluate as part of our screening analysis account for 99.8 percent of national
PB-HAP emissions from all stationary sources (on a mass basis from the 2005 NEI). The acid
gases, HC1 and HF, were included due to their well-documented potential to cause direct damage
to terrestrial plants. According to the 2005 NEI, HC1 and HF account for about 99 percent (on a
mass basis) of national acid gas emissions from stationary sources. The commenter did not
provide any documentation to support the claim that arsenic and nickel air emissions from the
petroleum refineries category cause adverse environmental effects. Therefore, we have not
revised our environmental risk screen to include arsenic and nickel.
Comment 2: Refinements to the ecological risk assessment are needed to address region-specific
impacts: The commenter stated that EPA has not adequately examined environmental, wildlife,
and other ecological risks such as region-specific impacts to wildlife, including federally listed
species under the Endangered Species Act, and aquatic resources in rivers and estuaries. For
example, the commenter noted that EPA is legally required to assess impacts to endangered and
threatened species, and yet EPA's assessment includes no discussion of potentially affected
species located near refineries, much less any evaluation of the risks they face. The commenter
62
-------
stated that the EPA also says nothing about ESA consultation which, according to the
commenter, provides further evidence that it has not considered or addressed its duty to prevent
adverse environmental effects. The commenter asserted that EPA must assess potential
endangered and threatened species near the existing sources in this source category.
The commenter provided specific details of the area of Anacortes, Washington and Cherry Point
and cited numerous scientific studies showing environmental impacts in that area. This area has
four large refineries near where the Nooksack and Skagit River deltas provide fertile salmon
habitat. The commenter also note that shellfish and orcas - the latter listed as endangered under
the Endangered Species Act - are of particular concern in the area. According to the commenter,
recent research suggests that PAHs in streams are creating big problems for salmon.
Furthermore, current research by the National Oceanic and Atmospheric Administration
(NOAA) concerning storm water runoff and PAHs indicates regional species' sensitivity to PAH
emissions from any source.75 The commenter stated that EPA's assessment appears to look only
at lakes, which is not giving EPA the full picture for impacts to salmon and orcas.
The commenter stated that it is also important to consider that air emissions from the Anacortes
area refineries (and from the aluminum smelter that is also located there) significantly affect
Olympic National Park. The commenter noted that emissions from the Anacortes sources (the
Park Service was concentrated on the Tesoro Refinery) are the most significant source of haze
impairment in the Olympics and presumably air toxics are deposited in Olympic lakes, rivers,
and estuaries as well. It is unclear whether EPA has done an assessment of those impacts,
according to the commenter.
The commenter further asserted that it is unclear if EPA's risk assessment documents analyzed
particular existing problems and sensitivity to airborne emissions over time. According to the
commenter, but it is difficult to tell from EPA's documentation what water bodies actually were
modeled. The commenter asserted that the latest studies about mercury in Olympic National Park
are available and EPA must consider these and other similar scientific research in this
rulemaking in order to fulfill its legal duty to evaluate the adverse environmental effect of
refineries' pollution under section 112(f)(2).
Response 2: The environmental risk screen is designed to be a conservatively protective screen
that identifies potential adverse environmental effects from refinery air emissions to the most
sensitive organisms, including threatened and endangered species. We evaluate four exposure
media in the environmental risk screen: terrestrial soils, surface water bodies, fish consumed by
wildlife, and air. Within these four exposure media, we evaluate nine Generic Ecological
Assessment Endpoints (GEAEs). The GEAEs reflect the overall "health" of aquatic and
terrestrial ecosystems and any important organisms that could be exposed in those ecosystems,
including threatened and endangered species. Therefore, we did not conduct an ecological
75 See NOAA Northwest Fisheries Science Center, Polycyclic aromatic hydrocarbons,
http://www.nwfsc.noaa.gov/research/divisions/efs/ecotox/pah.cfm and NOAA Northwest Fisheries Science Center,
Stormwater science: ecological impacts, http://www.nwfsc.noaa.gov/research/divisions/efs/ecotox/ecoimpacts.cfm
and publications cited there.
63
-------
assessment that focused on specific species of organisms found in the Anacortes region as
suggested by the commenter.
Specifically, the environmental risk screen includes the following conservative assumptions to
ensure that adverse environmental effects to the most sensitive species are identified:
Selection of species for the population-level assessments for PB-HAPs was based on
those organisms that are likely to be the most sensitive and/or highly exposed due to
bioaccumulation of the PB-HAP through aquatic and terrestrial food chains.
For hydrogen fluoride, we evaluate the ecological endpoint for damage to plants, rather
than fluorosis in livestock, because the HF concentrations at which fluorosis in livestock
occur are higher than those at which plant damage begins.
Where possible we used ecological benchmarks for each ecological assessment endpoint
at no-observed-adverse-effect levels (NOAEL).
In the Tier 1 screen for PB-HAP we use worst case meteorology and lake distance.
As the commenter indicates, for PB-HAP we screen for adverse environmental effects
primarily for lakes. In general, lentic bodies of water (lakes and ponds) accumulate
higher levels of contaminants in both sediments and biota than lotic systems (rivers,
streams). Therefore, if no potential adverse environmental effects are identified in lakes
we would not expect there to be an adverse environmental effect on
rivers/stream s/estuari es.
We did not include background levels of pollutants, such as pollutants emitted from other non-
refinery sources in the environmental risk screen as suggested by the commenter. Similar to the
AMOS analysis for human health, background levels (including natural background) are not
barred from the EPA's assessment of adverse environmental effects, and the EPA may consider
them, as appropriate and as available, along with other factors, such as cost, economic impacts
and technical feasibility. However, this assessment excludes background contributions because
the available data are of insufficient quality upon which to base a meaningful analysis. We
reviewed the information the commenter cited in their comments and found the following:
The mercury study titled "Mercury in Fishes from 21 National Parks in the Western
United StatesInter- and Intra-Park Variation in Concentrations and Ecological
Risk" published by the U.S. Geological Survey does not draw any connection
between mercury measurements and petroleum refinery facilities. In fact, the study
cites the global nature of mercury pollution with sources as far away as China.
The National Park Service (NPS) Western Airborne Contaminants Assessment
Project Report also makes no mention of HAP emissions from petroleum refineries.
The report cites regional agriculture and global emissions as the primary sources of
airborne contaminants.
The NOAA studies cited by the commenter focus on storm water runoff from roads,
parking lots, and other impervious surfaces, not air emissions from petroleum
refineries.
The comments submitted by the National Park Service to EPA, Region 10 on
February 15, 2013 are in regards to the Regional Haze program. The comments
64
-------
related to Tesoro Refinery focus on NOx and SOx controls for improving visibility,
not on HAP, which are the focus of this rule.
Based on our conservative environmental screening analysis, we do not expect adverse
environmental effects, as defined in CAA section 112 (a)(7), as a result of HAP emissions from
any sources addressed in this rulemaking. The EPA does not have information, nor did the
commenter provide information to support the position that HAP emissions from facilities
addressed in this rulemaking would result in environmental effects. Because our environmental
risk screen does not indicate an environmental effect from this source category, we disagree with
the commenter's statement that the EPA must engage in consultations (e.g., with the U.S. Fish &
Wildlife Service and the National Marine Fisheries Service) under 16 U.S.C. 1536(a)(2).
2.6 Rule changes are not needed because risks are acceptable
Comment 1: Several commenters agreed that the EPA has correctly concluded that the proposed
rule requirements protect the public with an ample margin of safety from refinery emissions.
Other commenters noted that EPA found residual risks remaining after implementation of the
MACT standards to be acceptable, and in light of the acceptability determination argued that the
proposed changes to the rule are not justified. The commenters noted that the EPA's detailed
emissions inventory assessment and risk modeling results demonstrated that, at every U.S.
refinery, category-specific risks are below the EPA's presumptive limit of acceptable risk (i.e.,
cancer risk of less than 100-in-l million).
The commenter noted that the estimated annual cancer incidence based on the maximum risk
results is less than 1 case per year (0.3 and 0.6 cases per year based on actual and allowable
emissions, respectively), and argued that this incidence is insignificant compared to the over
1,665,000 cases predicted to occur in 2014.76 The commenter added that the proposed
requirements (e.g., storage vessels, delayed coking units) would not affect maximum individual
risk and would only result in an estimated 2 to 15 percent reduction in the conservatively
estimated cancer incidence for the source category. The commenter claimed that such modest
available reductions (2 to 15 percent) in theoretical upper-bound risk demonstrate that the
existing standards are already providing an ample margin of safety.
The commenter further asserted that because risk was found to be acceptable, the agency must
follow the requirements of section 112(d)(6) of the CAA when proposing actions to further
increase the margin of public safety and thus must consider costs when issuing new controls of
emissions. Commenters argued that proposed rule costs are estimated to be $20-40 billion. Due
to these costs burdens, commenters recommended that EPA withdraw the proposed rule.
Commenters also stated in response to EPA's request for comment on whether to find the risks
unacceptable, that there are flaws in the risk analysis based on actual and allowable emissions
which the commenter believes if corrected would decrease the overall estimated risk.
76 http://seer.cancer.gov/statfacts/html/all.html
65
-------
For these reasons, the commenter asserted that to be consistent with the finding of sufficient
protection, EPA should retain the existing standards, as allowed by the FCAA.
Response 1: Regarding the comment that in light of the acceptability determination the proposed
changes to the rule are not justified, we note that we also are required to ensure that the standards
provide an ample margin of safety to protect public health. That analysis is separate from the
acceptability analysis, and the determination of acceptability does not automatically lead us to
conclude that the standards provide an ample margin of safety to protect public health.
In response to the comment on cost burden, as note elsewhere in this RTC, we are not finalizing
the atmospheric PRD prohibition. While some facilities will need to implement additional
prevention measures on some atmospheric PRD, we estimate that the costs of the final work
practice standards to be reasonable.
We address various comments regarding commenter identified potential flaws in the estimation
of actual and allowable in the risk assessment in Sections 2.1.1 (comment/response 2 and 4),
2.1.2 (comment/response 1), and 2.1.3.
Comment 2: EPA's proposed changes are not demonstrably beneficial or cost effective:
Commenters stated that despite significant emission reductions and financial investments made
by refiners over the last two decades coupled with the fact that both the 2008 and current RTR
risk met the acceptability criteria, the EPA has proposed regulations which impose significant
costs. According to the commenter, EPA has done so without demonstrating that such
regulations are cost effective or justified by the factors specified in section 112(d)(6) of the
CAA. The commenter added that EPA took no further action in its 2008 rulemaking after
determining risk was acceptable. Another commenter also stated that EPA's analysis shows that
the public health benefits from this rule, under best case scenarios, would be quite small even
when based on the conservative assumptions, extrapolations and conclusions which are
unsupported by science, making the projected benefits of these rules illusory.
One commenter stated that the balance of this proposal is not based on risk reduction and thus
must be justified on other grounds and must be achievable and cost effective. Even where a
change is claimed to be driven by judicial decisions, the commenter asserted that EPA must
follow the provisions of the CAA in implementing the Court decision, which has not been done
in many cases. The commenter provided the example that extending emission limitations derived
for normal operation to maintenance, startup, and shutdown (MSS) periods may be the easiest
regulatory path, but that approach is not required and is sometimes not cost effective or even
feasible; alternative standards must be developed and implemented in such cases.
The commenter stated that although EPA's analysis has overstated the risk associated with
refinery emissions, it does not affect the agency's conclusion that risk levels are acceptable and
that only additional controls on small storage tanks are justified as a means to increase the ample
margin of safety. The commenter noted that EPA evaluated the reduction in risk associated with
a number of additional potential controls on storage vessels, equipment leaks, gasoline and
marine loading racks, cooling towers and heat exchangers, wastewater collection and treatment
systems, FCCUs, flares, and other refinery emissions sources. With the exception of changing
66
-------
the applicability threshold to include controls on small storage vessels, no controls were
identified that cost-effectively reduced the levels of risk and - even in the case of storage vessels
- the cost-effectiveness was largely driven by EPA's questionable assumption of credits
associated with VOC recovery, according to the commenter.
Response 2: We maintain that the proposed rule is targeting significant, scientifically
identified deficiencies in the current MACT standards that must be corrected to ensure that
health risks are acceptable and to provide an ample margin of safety. Since 2008, we have new
data documenting poor flare performance and the magnitude of emissions from decoking
operations which have led to significant revisions that we are now including that were not in the
2008 rule. Additionally, we are required to make several other revisions to ensure the refinery
MACT standards are consistent with the Sierra Club v. EPA decisions; another development
since the 2008 rulemaking.
We note that this rulemaking includes both a technology review and a risk review. We only
included the storage vessel requirements as part of our risk review (similar to the conclusions in
the 2008 rulemaking). Given the advances in technology and practices since the 2008
rulemaking, the additional revisions we are making in the refinery MACT standards are
appropriate and consistent with the CAA requirements in section 112. We disagree that the
proposed amendments are not justified. We made several revisions to address SSM
exemptions. These revisions are not only justified, but required by the CAA as determined in the
Sierra Club v. EPA decision. We also revised existing MACT requirements or set new MACT,
under the authority of 112 (c)(2) and (3) as we determined that the original MACT were deficient
in achieving expected emission reductions or missing. For example, the DCU decoking
operations had significant HAP emissions and contributed significantly to the overall cancer
incidence from refineries was effectively exempted from MACT requirements. Therefore, we
developed MACT requirements for DCU decoking operations. Similarly, minimal flare
monitoring requirements were included in Refinery MACT 1 and 2, leading to the operation of
flares with poor control efficiencies. Therefore, we determined that significant revisions were
needed in these requirements. Our technology review also identified numerous issues with the
ongoing performance and monitoring requirements for some control systems. These
requirements were based in part on the technology review conducted under the New Source
Performance Standards which found that the existing monitoring requirements were insufficient
to ensure compliance at all times. For these reasons, we enhanced the monitoring requirement for
FCCU and in addition to the current 30 percent opacity compliance option for subpart J, we are
adding a 20% opacity 3-hour average operating limit. Consequently, even though we are not
promulgating these specific requirements under the authority of section 112(f)(2), we maintain
that significant revisions to Refinery MACT 1 and 2 are "necessary" to ensure refinery owners
and operators are complying with the MACT requirements at all times.
2.7 Stronger standards are needed because risks are unacceptable
Comment 1: EPA must find the current health risk unacceptable and set protective standards
with an adequate margin of safety: Many commenters argued that EPA underestimated risks and
that the true level of risk is unacceptable. One of these commenters argued that the residual risk
assessment should have found the risk from refinery pollution "unacceptable" and that EPA's
67
-------
rule will not protect communities with the "adequate margin of safety" required by the CAA.
The commenter stated that outdated and inadequate science cannot estimate true risk. EPA's risk
analysis has failed to adequately account for early life exposure, multipathway exposure, and
cumulative impacts from multiple source exposure, according to the commenter. The commenter
further noted that EPA has also understated risk by ignoring emissions during unplanned SSM
events and by ignoring hazardous air pollutants for which no reference value is established. As a
result, the commenter asserted that the 100-in-l million MIR is an underestimation.
The commenter asserted that 0.6 excess cancer cases per year - or at least 1 extra cancer case
every other year - is one additional cancer case is too many. It is plainly unacceptable that EPA
proposes to allow at least 1 new cancer case every other year, and 5 new, additional cancer cases
every decade, in communities that have refineries, because Congress enacted the 1990 CAA
Amendments to prevent cancer from toxic air pollution, and to do so especially in communities
overburdened by such pollution.77 According to the commenter, when EPA recognizes that at
least 7 million people are exposed to extra cancer risk from refineries, it should plainly find this
inhalation-based cancer risk, alone, to be unacceptable under section 112(f)(2). Another
commenter noted that EPA has determined that the risks from petroleum refinery emissions are
"acceptable", however the commenter expressed serious concerns with MACT-allowable
emission cancer risks are up to 100-in-l million. Another commenter (who supports steps to
reduce the use of flaring, including banning flaring and monitoring and reporting of flaring
emissions to the public) added that even with the proposed improvements, only 18 percent more
people will be protected, leaving nearly four million people breathing air that gives them an
unacceptable risk of cancer, which is far from the only health risk they will continue to face.
According to the commenter, in addition to the reasons EPA identified as why risk could be
considered unacceptable (79 FR 36940), the EPA should also find the current health risks to be
unacceptable for the following reasons, which have been included in more detail elsewhere in
this chapter:
Underestimation of the cancer risk from inhalation particularly from early exposure.
Underestimation of the cancer risk from multipathway exposure.
Need to recognize that the combination of cancer, high chronic non-cancer and acute
risks, together, create unacceptable risk.
Need to recognize that the cumulative impacts and multiple source exposure from various
sources, including refineries.
EPA did not evaluate facility-wide risk based on "allowable" emissions, which would
likely cause the presumptive acceptability benchmark for cancer to be exceeded. In
addition, the chronic non-cancer risk EPA found is 4 (4 times EPA's TOSHI threshold of
1). [79 FR 36937]
There is a high population exposed to refineries' risk (at least 7 million), and resulting
incidence of cancer risk (at least one case every 1.5 years). There are 83 million people
living within 50 km (approx. 31 miles) of refineries, with an additional risk of
catastrophic exposure.
77 S. Rep. No. 101-228, at 128-29, 1990 U.S.C.C.A.N. at 3513-14.
68
-------
It is unjust and inconsistent with the Act's health protection purpose to allow the high
health risks caused by refineries to fall disproportionately on communities of color and
lower income communities who are least equipped to deal with the resulting health
effects.
The rule likely does not address all emissions, particularly fugitive emissions.
Response 1: We believe we have adequately estimated risk from these source categories and that
the risk is acceptable. As discussed in previous responses, there are many aspects of the risk
assessment that commenters disagreed with, and we are providing references to the responses to
these comments here:
Early life exposure: Chapter 2.3.2, 2.3.6
Multipathway exposure: Chapter 2.3.5
Cumulative risks and impacts from multiple source exposure: Chapter 2.3.4, 2.3.6
Use of allowable emissions: Chapter 2.1.2, 2.3.5
Non-cancer risk: 2.3.2, 2.3.5, 2.3.6
Environmental justice concerns: 2.4
Concerns over the inclusion of emissions from SSM events: 2.3.4
Comment 2: Risk is unacceptable even when less than 100-in-l million: Many commenters
disputed the acceptability criteria of 100-in-l million for risk and stated that it should be lower,
particularly when there is uncertainty built into the EPA's risk assessment, the EPA lacks
necessary information on certain pollutants, and these rules affect people of color and poor
people disproportionately. Since 1990, however, EPA has used this criteria. The commenter
asserted that the EPA based the criteria on an unusual study of people's perceptions of their own
risk from 1988, known as the Survey of Societal Risk (July 1988), rather than science.78 EPA
looked at an odd collection of risks, according to the commenter, such as dangers from driving a
car, and found that "the presumptive level established for MIR [maximum individual risk of
cancer] of approximately l-in-10,000 is within the range for individual risk in the survey, and
provides health protection at a level lower than many other risks common "in the world in which
we live." [54 FR 38044, 38046 (Sept. 14, 1989).]
The commenter noted that EPA has failed to revisit or update this number for the decades, even
though scientists have made breakthroughs on early-life exposure and children's vulnerability;
biomonitoring and other data on adult body burdens of chemicals; the vulnerability of
overburdened communities, including socioeconomic disparities; and on ways to analyze and
control the impacts of pollutants on human health. The commenter listed numerous landmark
policies concerning health risks since 1990 to substantiate the importance of the issue.
The commenter asserted that it is time for EPA scientists and policymakers to revisit the
outdated assumption EPA makes regarding what level of cancer risk triggers policy
78 Benzene Rule Docket No. OAQPS 79-3, Part I, Docket Item X-B-l, EPA Air Docket (cited at Nat'l Emission
Standards for Hazardous Air Pollutants; Benzene Emissions from Maleic Anhydride Plants, Ethylbenzene/Styrene
Plants, Benzene Storage Vessels, Benzene Equipment Leaks, and Coke By-Product Recovery Plants, 53 FR 28,496,
at 28,512-13 (July 28, 1988)).
69
-------
interventions. EPA's own policy regarding carcinogens recognizes that they have no safe
threshold of exposure, according to the commenter. EPA has appropriately recognized that
cancer risks add up to increase lifetime risk. Importantly, the commenter stated that EPA's
presumption regarding cancer risk ignores the experience of communities exposed to multiple
sources and types of sources of pollution. The commenter noted that even if some level of risk
might otherwise be acceptable, that cannot be assumed to be true for communities exposed to
more than one source that is causing that level of health risk. EPA has a responsibility to address
the science on cumulative impacts and risk, according to the commenter, and update its
assumptions accordingly, to acknowledge that cancer risks below 100-in-l million cannot be
presumed safe.
The commenter also argued that EPA should also reform how it evaluates chronic and acute
hazard indices, in which a risk number below 1 does not result in policy changes or standards.
The commenter suggests that EPA should instead factor in uncertainties and vulnerability factors
that adjust the "acceptable level of risk." The commenter notes that this is currently done under
the FQPA when EPA uses factors to determine a Target MOE and risks below this level warrant
increased scrutiny and changes to allowable exposures.79
Response 2: The approximately 100-in-l million benchmark was established in the Benzene
NESHAP (54 FR 38044, September 14, 1989), which Congress specifically referenced in CAA
section 112(f)(2)(B) . While this presumptive level provides a benchmark forjudging the
acceptability of MIR, it is important to recognize that it does not constitute a rigid line for
making that determination. The EPA considers the specific uncertainties of the emissions, health
effects and risk information for the source category in question when deciding whether the risk
posed by that source category is acceptable. In addition, the source category-specific decision of
what constitutes an acceptable level of risk is a holistic one; that is, the EPA considers all
potential health impacts chronic and acute, cancer and non-cancer, and multipathway along
with their uncertainties, when determining whether the source category presents an unacceptable
risk.
We also note in this response, that EPA has addressed comments specifically related to the use of
OEHHA toxicity values, early life exposure (including prenatal), cumulative impacts and risks,
and the use of uncertainty and vulnerability factors in chapters 2.3.2 and 2.3.4 of this RTC
document.
Comment 3: Other rule decisions support claim that refinery risks are unacceptable: One
commenter noted that in other rules, EPA has found or is proposing to find similar or lower
levels of risk to be unacceptable. For example, the commenter noted that in Secondary Lead
Smelting, EPA found that health risks were "unacceptable" based partly on high chronic non-
cancer risk (due to fugitive lead emissions) and partly on: "[t]he fact that maximum individual
cancer risks due to actual emissions are above 1-in-l million also contributes to our
79 See, e.g., EPA, Sulfuryl Fluoride; Proposed Order Granting Objections to Tolerances and Denying Request for a
Stay, Proposed Rule, 76 FR 3422, 3427 (Jan. 19, 2011) (explaining use of MOE).
70
-------
determination of unacceptability."80 The commenter asserted that for petroleum refineries the
maximum individual cancer risk due to actual emissions is well above 1-in-l million, is
underestimated (as explained in these comments), and the risk number based on allowable
emissions is 100-in-l million. Thus, the commenter argued that it would be inconsistent with its
risk acceptability determination in Secondary Lead Smelting not to find risk in this case similarly
unacceptable.
The commenter stated that EPA has long recognized that a combination of cancer and non-
cancer risks can show unacceptable health risks, as stated as its policy in the current
rulemaking.81 The commenter notes that, in its proposed rule for Ferroalloys, EPA found that
allowable cancer risk is 100-in-l million and that cancer risk based on reported "actual"
emissions is lower; EPA also found significant chronic non-cancer risks from manganese and
mercury. EPA stated, according to the commenter, that the fact that "risks from allowable
emissions are at the upper end of the range of acceptability ... combined with" high non-cancer
risk, lead EPA to conclude that the current risk from Ferroalloys sources is "unacceptable."82
EPA similarly found significant non-cancer risks here, as well as cancer risk at the "upper end"
of the range EPA considers acceptable - and thus, the commenter asserts, should similarly find
the health risks from refineries to be "unacceptable." Specifically, the commenter noted that EPA
has found a high acute risk (HQ of 5 based on "actual" emissions) and high chronic non-cancer
risk (TOSHI of 1), as well as multipathway risks that did not screen out, in a refined analysis for
a single refinery (which EPA recognizes is not representative of the highest multipathway risk
for any exposed individual).83.
As another example, the commenter stated that EPA found in regard to the Wool Fiberglass
source category that, although EPA's assessment of "actual" emissions did not create risk above
100-in-l-million (as is true with refineries), under the potential cancer risk EPA evaluated:
"8,100 people would be exposed to risks greater than 100-in-l-million, 460,000 people would be
exposed to risks of greater than 10-in-l-million, and over 7 million people would be exposed to
cancer risks of greater than 1-in-l-million."84 The commenter noted that regarding petroleum
refineries, there are millions of people exposed to cancer risks above 1-in-l million, about
100,000 exposed above 10-in-l million, and EPA has found that based on allowable emissions,
there is exposure at 100-in-l million, based on inhalation risk alone. By contrast with Wool
Fiberglass, where EPA found no other relevant health risks above its threshold, with petroleum
80 Secondary Lead Smelting, Final Rule, 77 FR 556, 563 (Jan. 5, 2012) (EPA has since granted the reconsideration
petition of environmental petitioners on the issue of how EPA underestimated health risks in the secondary lead
smelting risk assessment, and reconsideration remains pending).
81 79 FR at 36,899 ("the level of the MIR [maximum individual lifetime cancer risk] is only one factor to be
weighed in determining acceptability of risks ... the Agency may find, in a particular case, that a risk that includes
MIR less than the presumptively acceptable level is unacceptable in the light of other health risk factors.") (quoting
Benzene Rule).
82 Ferroalloys, Supplemental Proposed Rule, 79 FR 60,238, 60,269-70 (Oct. 6, 2014).
83 79 FR at 36,934-38. Risk Assessment (-0225).
84 Wool Fiberglass, Proposed Rule, 76 FR 72,770, 72,801 (Nov. 25, 2011).
71
-------
refineries EPA has found significant acute, chronic non-cancer, and multipathway risks, and
there are many more people exposed to refinery risks, on top of the cancer risks, which all
provide additional reasons for EPA to find petroleum refinery risk "unacceptable."
In general, the commenter argued that it would be out of step with EPA's interpretation of
section 112(f)(2), and with the policy reflected in other EPA risk determinations, and thus
"arbitrary" and "capricious", for EPA not to find that the refineries risk is similarly unacceptable.
According to the commenter, EPA must provide equal protection for people near air toxics
sources. Because people living near refineries are disproportionately people of color and lower
income people, the commenter notes that disparate treatment by EPA would be especially
problematic and raise discrimination concerns under all applicable requirements for equal
treatment and equal protection. The commenter also noted that EPA should not ignore the reality
that communities near refineries face daily, and instead should find that their health risks are
currently "unacceptable," and set residual risk standards for to reduce those risks.
Response 3: We generally draw no bright lines of acceptability regarding cancer or noncancer
risks from source category HAP emissions, and acknowledge that it is always important to
consider the specific uncertainties of the emissions and health effects information regarding the
source category in question when deciding exactly what level of cancer and noncancer risk
should be considered acceptable. In addition, the source category-specific decision of what
constitutes an acceptable level of risk should be a holistic one; that is, it should simultaneously
consider all potential health impacts chronic and acute, cancer and noncancer, and
multipathway along with their uncertainties, when determining the acceptable level of source
category risk. The Benzene NESHAP decision framework of 1989 acknowledged this; in today's
world, such flexibility is even more imperative, because new information relevant to the question
of risk acceptability is being developed all the time, and the accuracy and uncertainty of each
piece of information must be considered in a weight-of- evidence approach for each decision.
This relevant body of information is growing fast (and will continue to grow even faster),
necessitating a flexible weight-of-evidence approach that acknowledges both complexity and
uncertainty in the simplest and most transparent way possible. While this challenge is
formidable, it is nonetheless the goal of the EPA's RTR decision-making, and it is the goal of the
risk assessment to provide the information to support the decision-making process.
Comment 4: Technology developments require stronger standards: One commenter stated that
for both equipment leaks and wastewater, EPA has found that "developments" have occurred in
practices, processes, and control technologies within the meaning of CFR 112(d)(6), but has not
proposed to strengthen the existing standards for these emission points at all. [79 FR 36915
(leaks), 36919 (wastewater), 36920 (wastewater and leaks)] The commenter stated that it would
be "unlawful, arbitrary, and capricious" for EPA not to set stronger standards for these emissions
under section 112(d)(2)-(3).
In addition, the commenter noted that EPA has recognized that these sources drive the health risk
EPA found under section 112(f)(2). [79 FR 36934] Therefore, under section 112(f)(2), EPA must
reduce all unacceptable risk (which it has underestimated and should find to be unacceptable
here), according to the commenter, and set stronger standards to assure an "ample margin of
safety to protect public health" from these sources. EPA's purely cost-based reasons for not
72
-------
updating the equipment leak and wastewater standards are both inconsistent with the text and
purpose of the Act, according to the commenter, and are irrational based on the record showing
that methods to reduce emissions from these sources are available, are in use, and are not cost-
prohibitive. Further, the commenter argued that EPA should not use a cost-per-ton test to assess
the value of HAP reductions because, as the reference exposure values show, the pollutants
emitted by these sources are harmful at levels well below 1 ton, and even at the level of
micrograms. The commenter asserted that EPA has no valid basis not to acknowledge that it is
indeed valuable to public health to reduce less than 1 ton of benzene and the other air toxics
emitted by these sources.
Response 4: For the Petroleum Refineries NESHAP, we are revising emissions controls for
coking units, and storage and loading processes concurrently under both the CAA section
112(d)(6) technology review and the CAA section 112(f)(2) risk review. As noted in the
preamble to the proposed rule (79 FR 36879), in keeping with the Benzene NESHAP two-step
process for developing residual risk standards, in the risk review the EPA first examines risk
acceptability and then considers whether the existing standards provide for an "ample margin of
safety." The two-step process does not conclude with a decision of acceptability. In the second
step, the EPA determines whether the emissions standards provide an ample margin of safety,
considering all of the health risks and other health information considered in the acceptability
determination, as well as other relevant factors. While we proposed that the risks from the
petroleum refineries source category are acceptable, we also found that the cancer risk estimates
for 5,000,000 individuals in the exposed population were above 1-in-l million, with an MIR of
up to 60-in-l million, based on actual emissions. We then proceeded to the ample margin of
safety analysis. For this source category, we estimated what the risks would be if all of the
petroleum refinery facilities adopted control measures to limit emissions from coking units, and
storage and loading processes. We estimated that after implementation of the proposed controls,
about 1,000,000 fewer people will be exposed to cancer risks greater than 1-in-l million. Finding
that the revised requirements would reduce cancer risks and considering the associated costs,
economic impacts, and technological feasibility among other information, we proposed to revise
the standards for these emissions sources under CAA section 112(f). The commenter has
identified no flaw in our reasons or analysis under either section 112(d)(6) or 112(f)(2) for
adopting the revised requirements.
73
-------
3.0 Refinery Flares Control Device Provisions
Comment 1: Some commenters expressed support for the proposed flare operating requirements
but other commenters suggested that the EPA needs to do more. Suggestions included:
establishing direct emission limitations (release quantities, number of events, or other "hard
caps"); requiring specific technologies such as flare gas recovery systems; requiring back-up
power generation; and requiring root-cause analysis and reporting for flaring events. One
commenter suggested that the EPA needs to ban emissions from routine flaring and minimize
their use in other circumstances to protect people from these emissions.
Response 1: Flares are being used as air pollution control devices (APCD) to meet the Refinery
MACT 1 control requirement for affected emissions sources such as miscellaneous process vents
(MPVs), gasoline loading operations, marine vessel loading operations and storage vessels. The
flare itself is not an affected emissions source within Refinery MACT 1. As such, we disagree
with the commenter that we should ban emissions from routine flaring in the MACT rule. The
commenters do not recommend similar actions to minimize or eliminate the use of thermal
oxidizers, carbon absorbers or other control devices that may be employed to control HAP
emissions from the affected emission sources at the petroleum refinery. Eliminating the routine
use of flares as an acceptable APCD would only increase the use of these other types of APCD
(at potentially significant cost) without any net emissions reductions from the refinery (provided
that the flare is meeting the required control efficiency). Therefore, the flare operating
requirements we are finalizing in this action are focused on those requirements necessary to
ensure that refineries that use flares as APCD meet the MACT standards at all times when
controlling HAP emissions. In other words, the flare operating requirements are being finalized
to ensure that flares operate with appropriately high destruction efficiencies at all times when
controlling HAP emissions.
Comment 2: Commenters stated that there are other non-conventional types of flares, such as
pressure-assisted flares, ground flares, and enclosed flares, and recommended that the EPA set
separate (or case-by-case) limits for these "other" flare types.
Response 2: We note that the types of flares referred to by the commenter are in most instances
covered under the proposed definition of flare that we are finalizing in this action: "Flare means
a combustion device lacking an enclosed combustion chamber that uses an uncontrolled volume
of ambient air to burn gases. For the purposes of this rule, the definition of flare includes, but is
not necessarily limited to, air-assisted flares, steam-assisted flares and non-assisted flares." We
also note that if the combustion device has an enclosed combustion chamber that uses a
controlled volume of ambient air, then the device is a thermal oxidizer or incinerator and you
must verify the control device achieves 98% destruction efficiency or 20 ppmv following the
requirements in 40 CFR 63.645.
With respect to setting separate (or case-by-case) limits, commenters provided either insufficient
information (i.e., pressure-assisted flares) or no additional information (i.e., ground flares and
enclosed flares) for the agency to consider for use in developing separate limits, and therefore,
74
-------
we are not establishing separate limits for these types of non-conventional flares at this time. We
note however that to the extent a flare owner or operator considers the final rule requirements too
restrictive, the final rule includes provisions in 40 CFR 63.670(r) for owners or operators to test
their flares and develop alternative operating limits on a case-by-case basis.
Comment 3: Several commenters recommended that emergency or standby flares should have
separate standards.
Response 3: The commenter does not provide, and we are not aware of, any basis for revising
the flares standards to provide separate, different limits for emergency or standby flares. If a gas
stream containing HAP is required to be controlled under the Refinery MACT standards, then
that control device must meet the applicable control efficiency requirements, regardless of
whether that control device is used every day or once per year. Therefore, we are not establishing
separate standards specifically for emergency or standby flares.
Comment 4: Several commenters recommended that temporary flares (flares on-site for less
than 12 months) should not be subject to the flare requirements in subpart CC, but should
comply only with the General Provisions (GP) requirements in 40 CFR 63.11.
Response 4: The EPA disagrees with the commenters. Subpart CC applies to the identified
emission sources located at refineries and thus these emissions sources have been covered by the
MACT since it was first promulgated. As noted in the previous response, if a gas stream
containing HAP is required to be controlled under the Refinery MACT standards, then that
control device must meet the applicable control efficiency requirements. We have determined
that the GP requirements in 40 CFR 63.11 are not sufficient to ensure that refinery flares achieve
the applicable HAP emissions control requirements. The commenters provided no reason, and
we see no basis, to exclude temporary flares from the flaring requirements of subpart CC in this
action.
Comment 5: Three commenters suggested that 98 percent flare gas recovery should be
considered in compliance with the flare provisions.
Response 5: While we encourage the use of flare gas recovery systems, we disagree that this
requirement would be consistent with the flare requirements in the MACT and we do not believe
we should revise the MACT pursuant to section 112(d)(6) or (f)(2) to provide for this option.
First, 98 percent flare gas recovery is not equivalent to the 98 percent destruction efficiency that
forms the basis for the MACT flaring provisions. This is because 98 percent of the gas that is
diverted away from the flare to a recovery system is not 100 percent controlled. The recovered
flare gas is used in process heaters and/or boilers which should achieve a 98 percent destruction
efficiency.
We acknowledge that emissions from process heaters and boilers can be measured, so it is
possible to "prove" that these combustion devices are achieving greater than 98 percent
destruction efficiency. If, for example, 99 percent recovery of flare gas is achieved and 99
percent control efficiency of the recovered flare gas is achieved, then the overall control
efficiency of the recovery and control system would be 98 percent. However, we also have
75
-------
concerns about how a refiner could demonstrate compliance with this provision, if it were
provided. Continuous emissions monitoring system (CEMS) would be needed on all individual
process heaters and boilers receiving the recovered gas. Additionally, one would need to
continuously determine the percentage of flare gas that is being recovered. Process gases are
routinely recovered at certain process units as the standard practice. Therefore, there would be a
need to identify only that fraction of the fuel gas that is actually recovered at the flare (or from
the flare header system) that is in addition to the traditional process gases that are recovered.
Finally, if a big flaring event occurs that overwhelms the flare gas recovery system, it is quite
possible, for that day, month, or year (which ever averaging time was provided) that the flare gas
system would not meet the targeted 98 overall system control efficiency. Therefore, the refinery
owner or operator would likely need to install the appropriate flare monitoring systems even if a
flare gas recovery option were provided in order to demonstrate compliance during large
malfunction events. As we could identify no reasonable way a facility could demonstrate
continuous compliance with the flare provisions using flare gas recovery efficiencies, we have
not provided the suggested compliance alternative in the final rule.
Comment 6: Several commenters requested that the EPA require continuous PFTIR monitoring
of flares to ensure they are achieving the targeted 98 percent destruction efficiency.
Response 6: Based on our understanding of the PFTIR testing method and our review of the
PFTIR test data, we determined that it is technically infeasible to apply this technology on a
continuous basis. First, the monitor must be appropriately positioned so that it receives infrared
light signals from the flare exhaust plume. Shifts in wind directions would require manual
repositioning of the monitoring system and it would be impossible to differentiate short shifts in
wind speed or direction that cause the monitor to lose key carbon dioxide (CO2) signals (used to
calculate combustion efficiency) from poor combustion performance. The instrument would also
have significant interferences from rain, snow, and fog that would generally yield no useable
data during these events. In addition to technical infeasibility, we also note that there is an
inadequate supply of PFTIR monitoring systems as well as highly trained and specialized
personnel to operate these systems for every flare in the refining industry and that the cost of a
continuous PFTIR monitoring requirement would be cost prohibitive considering the equipment
necessary to process the data and the personnel needed to interpret the results. Lastly, no
standardized EPA or industrial method for operating a continuous PFTIR monitoring system for
flares currently exists. For all these reasons, we find that the suggested continuous PFTIR
monitoring requirement is infeasible at this time. Further, we find that the operating and
monitoring requirements that we are finalizing for flares are adequate to ensure that flares
achieve the targeted 98 percent control efficiency at all times so that additional monitoring
requirements are not needed to ensure compliance with the final standards.
Comment 7: Some commenters agreed with and supported the EPA's conclusion that wind does
not affect flare performance while other commenters disagreed with this assertion.
Response 7: We have extremely limited data to suggest that wind adversely impacts the
combustion efficiency of flares, let alone the combustion efficiency of industrial-sized refinery
flares. Commenters submitted no new data to otherwise support the assertion that wind does
indeed affect flare performance and as such, we are unconvinced that changing our position from
76
-------
the proposal that no flare operating parameter(s) are needed to minimize wind effects on flare
performance is warranted.
Comment 8: Several commenters noted that 10 years must be provided to comply with the
proposed requirements for flares and PRDs, stating that the time is needed to install new flares so
that PRDs will not be vented to the atmosphere and also to comply with the visible emissions
and velocity limits during SSM events.
Response 8: The statute provides a maximum compliance period of 2 years for standards
promulgated pursuant to CAA section 112(f)(2) and 3 years for standards promulgated pursuant
to CAA section 112(d). We note that the final rule does not include a prohibition on venting PRD
to the atmosphere, but instead establishes work practice standards for PRD releases that are not
vented to a control device, and it includes separate emissions limitations in the form of work
practice standards that apply to flaring events that exceed the smokeless capacity of the flare.
Comment 9: One commenter expressed concern that the proposed rules do not require refineries
to install refinery flare gas recovery systems. They stated that the technology can be
implemented with relatively small capital investments, that it has short pay back periods, and that
it appears to them that most refineries would profit by implementing a strategy of utilizing flare
gas recovery.
Response 9: The cost-effectiveness of a flare gas recovery system is a very site-specific
determination based on the current amount of gas being flared and the ability to offset natural gas
purchases with the recovered flare gas. Many facilities have installed flare gas recovery systems
because of the favorable economics for their particular site-specific conditions. However, it is
incorrect to assume that flare gas recovery systems would always be cost-effective in every
application because it was cost-effective for certain applications. In response to this and other
comments, however, we are requiring flares to minimize flaring during SSM events and we
specifically identify installation of a flare gas recovery system as a means to reduce flaring.
3.1 Halogenated vent stream
3.1.1 Prohibition of halogenated vent streams
Comment 1: Several commenters suggested that while the EPA proposed a new requirement to
prohibit the flaring of halogenated vent streams greater than 0.45 kilograms (kg)/hour (hr) (1
lb/hr), that no justification or analysis was provided for this newly proposed requirement in the
preamble or elsewhere, that no explanation of the legal authority was given, and that no costs or
emissions impacts were presented to determine if such standards are even necessary and that this
violates the EPA's obligations under section 307(d)(3) of the CAA as well as the Administrative
Procedure Act. In addition, commenters suggested these newly proposed requirements raise
significant safety concerns because they would apply under all operating circumstances,
including start-up, shutdown, process upset, and malfunction and recommended based on all of
these facts that the EPA remove this requirement from the rule. Commenters suggested that if the
EPA really believes that the flaring of halogenated vent streams is an issue that the EPA address
it in the future either through a new proposal or during the next technology review. A few
77
-------
commenters suggested that the ban on flaring a halogenated vent stream containing more than 1
lb/hr of halogen will significantly impact HF alkylation units, the catalytic reforming units, and
the isomerization units at refineries because these three types of process units contain and/or use
process streams with halogenated compounds. One commenter further stated that refineries
already have incentives to minimize halogen concentrations due to metallurgy and corrosion
concerns and that there is already a high level of care refiners take when processing these
streams (as is evident by their company's use of scrubbing systems to remove halogenated
species). They also stated that due to the solubility of halogens in water, they believe
halogenated species will drop out with the liquid in the flare knock-out drum and not be released
as an emission to the flare.
Response 1: First, we note that no refinery included dioxin emissions in their inventories for any
of the refinery flares, which would indicate that halogens are not being combusted in the flare,
and only a few reported HC1 emissions from a flare. We are not, however, finalizing the ban on
halogenated vent streams at this time because we did not include sufficient justification or
include cost estimates for this proposed provision and we did not include any monitoring
requirements to ensure compliance with this ban on halogenated vent streams. Also, we find that
more emissions data, control technology information and cost information are needed to
establish whether additional standards for halogenated vents are needed. Therefore, we are not
finalizing the proposed halogenated vent gas prohibition for flares in this final rule.
3.1.2 Definition of halogenated vent stream
Comment 1: One commenter recommended clarifications to the definition of "Halogenated vent
stream" should the EPA decide to promulgate the requirements to not flare halogenated vent
streams in 40 CFR 63.670(a).
Response 1: As noted above, we are not finalizing the requirements proposed for flaring
halogenated vent gas streams.
Comment 2: One commenter suggested three different alternative compliance options to the
requirements to not flare halogenated vent streams in 40 CFR 63.670(a) that the Agency could
make should they decide to finalize the aforementioned requirements.
Response 2: As noted above, we are not finalizing the requirements proposed for flaring
halogenated vent gas streams.
3.2 Pilot flame requirement
Comment 1: A few commenters suggested that although most flares within the industry have
multiple pilots, that the EPA revise the flare flame monitoring language in both Refinery MACT
1 and Refinery MACT 2 to make clear that the proposed standards only require one flare pilot
flame be lit since this is consistent with the wording in the current General Provisions
requirements at 40 CFR 63.11(b)(5).
78
-------
Response 1: We note that the requirement in 40 CFR 63.670(b) is to . .operate each flare with
a pilot flame present at all times." That is, the requirement is to operate with "a" pilot flame,
which we interpret to mean at least one pilot flame must be present. The language proposed to be
included in 40 CFR 63.670(g) includes the optional plural form in the event that a refinery owner
or operator elects to monitor for a flame at multiple pilot locations. This may be done to help
demonstrate that "a" pilot flame is present at all times. If one pilot goes out and the other does
not, monitoring both pilots allows that refinery to demonstrate continuous compliance. If the
owner or operator elects to monitor only one of the pilots and that pilot goes out, the refinery
owner or operator may not be able to demonstrate compliance with the standard. Therefore, we
are finalizing 40 CFR 63.670(b) as proposed, with the exception of no longer requiring an
automated device to the relight the pilot (see Section 3.2.1 for more details).
Comment 2: A few commenters suggested that the EPA should allow alternatives to continuous
pilot flame presence and monitoring for different flare sizes and type of flare operations that may
not be continuous like those controlling emissions from batch operations or truck and railcar
loading operations. The commenters suggested that the EPA should allow these intermittent type
of operations to operate with no pilots until the need is triggered to operate the flare, and only
require pilot monitoring when flares are actually receiving streams requiring control. On the
other hand, one commenter suggested a variety of additional requirements for the EPA to
consider that would strengthen the requirements for flare pilot flames. These suggested
recommendations included pointing to the self-imposed industry requirements per API-537
which specify multiple requirements for stable pilot flame operation and re-ignition, removing
the regulatory language "when regulated material is routed to the flare" to ensure that the pilot
must be continuously lit, and adding a requirement that flare pilots must have a reliable fuel
source separate from the flare vent gas as some pilot designs have the potential to use a portion
of the flare vent gas as a slip stream fuel source.
Response 2: The first set of commenters appear to misunderstand the proposed rule. The
proposed pilot flame requirements "...apply at all times when regulated material is routed to the
flare," [40 CFR 63.670(b)], In response to other comments, we have added the following
definition of regulated material: Regulated material means any stream associated with emission
sources listed in §63.640(c) required to meet control requirements under this subpart as well as
any stream for which this subpart or a cross-referencing subpart specifies that the requirements
for flare control devices in §63.670 must be met. If no "regulated material" is being discharged
through the flare, there is no requirement to operate a pilot flame. Thus, the proposed rule would
allow flares to operate with no pilots if no regulated material is routed to a flare, which appears
in part to be what the first set of commenters are requesting, however we note that refiners still
have a general duty to minimize emissions as required in 40 CFR 63.642(n). Also, we understand
that some flares that are used on a discrete basis continually vent purge gas to the flare to prevent
oxygen ingress into the flares. If refinery fuel gas is used as the flare's purge gas, then it is likely
that the flare will always be in regulated material service and the rule would require that these
flares operate continually with a pilot flame present. If the flare is purged using natural gas or
nitrogen, these gas streams are not regulated HAP streams and the requirements in 40 CFR
63.670(b) would not apply during those times that only these streams without "regulated
material" are used in the flare. We are not removing this phrase "when regulated material is
routed to the flare" from this requirement as requested by one commenter because the flare is not
79
-------
a separate affected emissions source within the Refinery MACT rules - it is an APCD. Section
112 of the CAA only authorizes us to set standards for the control of HAP; natural gas or
nitrogen do not contain HAP and we do not have the authority under section 112 to regulate
these vent gas streams when they are not mixed with a HAP-containing vent gas. We consider
the requirement to have a pilot flame present at all times is sufficient to ensure flare vent gas is
ignited as it exits the flare and we do not consider it necessary to provide additional
specifications to refinery owners or operators on how to operate their pilot flame system. We are
finalizing this requirement as proposed.
3.2.1 Automated relight systems
Comment 1: Several commenters suggested that the EPA remove the proposed requirement that
flares be equipped with automatic pilot re-ignition systems. The commenters state that the EPA's
assumption that nearly all refinery flares are already equipped with an automated device to re-
ignite the pilot flame is incorrect and rather, most refinery flares are equipped with manual re-
ignition systems. The commenters point to the fact that it is very uncommon for a flare to lose all
of its pilots and/or all pilot monitors and if that happens, there is already adequate redundancy
for operators to be quickly alerted so they can manually re-ignite the pilot. Commenters also
stated that past experience has shown auto re-ignition systems to be unreliable, and that there is a
potential significant cost burden to the industry of $1.5-2.0 billion ($3-4 million per flare) to
require a system that would provide little to no improvement in compliance assurance compared
to manual re-ignition systems. One commenter added that the EPA has not identified what CAA
provision authorizes this change or included any cost or associated emission reduction in the rule
record or the Paperwork Reduction Act (PRA) Information Request Supporting Statement to
support the proposed revisions.
One commenter suggested that if the goal of the proposed automatic pilot re-ignition system was
to remove the possible delay of manual interaction to relight the pilots that the following re-
wording should be considered: "The pilot system must be equipped with an automated device to
relight the pilot(s) without human interaction if extinguished."
Another commenter suggested that the fact that the EPA proposed a requirement for a flare pilot
automatic re-ignition system demonstrates that a pilot flame system will not be able to operate
continuously. The commenter suggested that the rule needs to allow for a time gap to for
automatic or manual relighting of the pilot flame if it were to go out.
Response 1: Our intent for an automated re-light system was to ensure that the pilots were re-lit
as soon as practical. We proposed this requirement under CAA section 112(d)(2) and (d)(3)
along with the other flare requirements. We expected most refineries had automated re-light
systems so we did not attribute any costs to the proposed requirement. As noted in the previous
response, we consider the requirement to have a pilot flame present at all times is sufficient to
ensure flare vent gas is ignited as it exits the flare and we do not consider it necessary to provide
additional specifications to refinery owners or operators on how to operate their pilot flame
system. Therefore, we are not requiring in the final rule that all flares be equipped with automatic
re-ignition systems. We agree with the commenter that suggested that even with automatic re-
ignition there will be at least some small gap when the pilot is not lit once it goes out and the
80
-------
commenter recommended that the rule should address this. We are clarifying in 40 CFR
63.670(b) that one minute in any 15 minute block where a pilot flame is not present (when
regulated material is routed to the flare) is a deviation of the standard and that deviations in
different 15 minute blocks from the same event are considered separate deviations (such that
failure to relight the pilot flame prior to the next 15-minute block will be a separate deviation).
3.2.2 Pilot monitoring requirements
Comment 1: One commenter stated that while the proposed regulation language at section
63.670(b) requires pilot flame monitoring, it has no requirements to monitor the presence of a
flare flame or other such requirements that would indicate flare combustion.
Response 1: There are numerous other provisions in 40 CFR 63.670 that place requirements on
the flare vent gas to ensure that the gas discharged by the flare is flammable. These provisions in
conjunction with the pilot flame requirement will ensure proper performance of the flare.
Comment 2: One commenter suggested that reporting each instance when a pilot flame has gone
out is unnecessary because many flares are designed with more than one pilot and can maintain a
flame and safely operate even during periods when one of the pilots is not lit. The commenter
further suggested the absence of a pilot flame does not indicate the absence of a flare flame or
whether destruction of gases being routed to the flare is occurring.
Two commenters suggested alternatives to the options proposed for pilot flame monitoring. The
commenters claimed that the ultraviolet beam sensor and infrared sensor options in the current
proposal are not capable of distinguishing between the main flare flame and the pilot flame. The
commenters identified an ionization detector, an ultrasonic (sound) detector, or possibly a flame
rod as options that could make such a distinction.
Response 2: The commenter is misreading the proposed regulatory requirement. The
requirement is to have "a" (i.e., at least one) flare pilot flame present at all times when regulated
material is sent to the flare and to report "each period when regulated material is routed to a flare
and a pilot flame is not present." Thus, this reporting requirement would not require reporting of
each instance when a single pilot went out if multiple pilots are used and monitored, but only
those instances that all of the pilots were not lit.
We agree that most pilot flame monitors would likely read compliant if a flare flame is present
even if the flare pilot goes out. We do not consider this to be a very realistic condition
considering flare and flare pilot designs. However, even if it were to occur, we note the objective
of these requirements is to ensure that an ignition source is always present to ignite and
adequately combust the flare vent gases discharged to the flare. Thus, we do not consider it
necessary to finalize a requirement for use of a monitoring system that can always delineate
between the flare flame and the pilot flame at this time.
Comment 3: One commenter suggested that the proposed requirements in 40 CFR 63.671(a)
should clearly limit Table 13 applicability to only the instrument types specified in that table and
not apply those requirements to remote sensing monitors such as infrared and ultraviolet (UV)
81
-------
pilot monitors as well as visible emissions video monitors. One commenter similarly suggested
that the EPA keep the flare flame monitoring language in the existing regulations at 40 CFR
63.644(a)(2) and 40 CFR 63.11(b)(5) and not convert the flare pilot monitors to continuous
parameter monitoring system (CPMS) due to challenges associated with flare and process
shutdown, accessing an elevated flare tip safely, and lost production time due to additional
requirements with respect to performing CPMS quality assurance/quality control (QA/QC).
Response 3: Table 13 only applies to those CPMS that are listed in Table 13. We agree with the
commenters that flare pilot monitoring systems need not be subject to the performance and
quality control/quality assurance requirements in Table 13. For example, it is not necessary to
determine a precise measurement of temperature (when using a thermocouple) to detect the
presence of a flame, so it is not critical to have a temperature monitor that is accurate to ±1
percent as would be required if Table 13 applied. We have revised 40 CFR 63.671(a)(1) to add
the clause "Except for CPMS installed for pilot flame monitoring..and to include the word
"applicable" (i.e., "Except for CPMS installed for pilot flame monitoring, all monitoring
equipment must meet the applicable minimum accuracy..." ) to help clarify that the
requirements in Table 13 are applicable only to those systems included in Table 13 and that they
are not applicable to flare pilot flame monitoring systems. We are also excluding flare pilot
monitoring systems from the out-of-control period requirements in 40 CFR 63.671(c) as these
are not applicable for CPMS that are not required to meet the accuracy requirements in Table 13.
3.3 Visible emissions and velocity requirements
3.3.1 Need for visible emission limit
Comment 1: One commenter suggested that when the EPA promulgated the visible emissions
requirements in the General Provisions in part 63, it provided no explanation for the foundation
of this requirement in its proposed or final preamble and that this error was repeated when the
EPA incorporated those requirements in the Refinery MACT 1 standards. The commenter
claimed that now that the EPA is re-codifying and changing portions of these requirements they
must support the legal and technical foundation for the requirement. Specifically, the commenter
suggested that the EPA must provide an explanation for why it believes that daily opacity
monitoring will better ensure VOC/HAP compliance and why elevated opacity would signify
that the flare is not being properly maintained and operated for VOC/HAP control efficiency. In
addition, with regard to the newly proposed daily observation requirements, the commenter
stated that EPA must explain how it is supported under either sections 112(d)(6) or section
112(d)(2) and (3) in light of the fact that a daily Method 22 observation of flares is not currently
in use as an industry practice.
Several commenters suggested that a smoking flare does not imply reduced combustion
efficiency, that there is a potential conflict for demonstrating compliance between operating a
flare close to the incipient smoke point and operating with no visible emissions, and that PM2.5 is
not a regulated pollutant under section 112 of the CAA and regulating these emissions in
Refinery MACT 1 is inconsistent with the General Provisions requirements. To support their
claims, the commenters point to the 1994 RTC for the General Provisions rulemaking, in which
the EPA promulgated the flare requirements in 40 CFR 63.11. The commenters noted that the
82
-------
RTC discusses the frequency of the Method 22 observations. The commenter also noted that the
TCEQ flare study, which analyzed flare soot, found it to be elemental carbon and not PM
containing HAP. The commenter also referenced the TCEQ operator training discussing proper
flare operation and the visible emissions observation point as being outside the flame envelope,
which is something the EPA should clarify in the rule.
One commenter suggested that flares with flare gas recovery systems, flares that only receive
purge gas or sweep gas (with a daily average molecular weight of less than 20 grams (g)/mole),
and flares that only receive high hydrogen streams have a low likelihood of smoking and should
not be required to perform the daily Method 22 observation and monitoring for visible emissions.
Two commenters suggested that performing daily visible emissions observations using EPA
Method 22 is overly burdensome and an unnecessary for refiners. To support this argument, the
commenters point to the fact that flare smoking events are rare and unscheduled, that many flares
now are equipped with flare gas recovery systems that minimize flaring and make routine flaring
of regulated material more intermittent in nature, and that additional personnel and costs will be
incurred unnecessarily when video cameras are already available and used by operators.
Commenters further stated that flare test data show that visible emissions do not suggest poor
destruction efficiency by the flare.
One commenter suggested that flares are not designed to adequately combust benzene and other
aromatic hydrocarbons, especially when smoking and that evidence of smoking is extensive and
is a big problem.
Response 1: We disagree with the commenter that we did not provide an explanation of the
visible emissions standard when it was first established for flares and we find no record of this
requirement being challenged on this ground in the 1994 RTC. We included that visible
emissions requirement in the General Provisions as a means to ensure that flares were operating
with destruction efficiencies of 98 percent of higher. In addition, for the present rulemaking we
also provided the basis for this requirement in Section IV.A.3.b of the preamble to the proposed
rule. As discussed in the proposal preamble, smoking flares indicate reduced combustion
efficiency. Furthermore, as discussed in other comment responses, this requirement is part of a
suite of flare requirements that ensure that the flare is well-operated as necessary to achieve the
MACT requirements for affected emissions sources using flares as a control device.
We also think commenters clearly misunderstood how the TCEQ flare study PM measurements
were taken and analyzed. The TCEQ made a concerted effort to obtain flare test data that were in
compliance with the current General Provisions flaring requirements. Specifically, they state on
page 49 of the final report that "A requirement of this study was that all data be obtained at flare
operating points that comply with 40 CFR 60.18. A requirement of 40 CFR 60.18 is that flares be
operated with no visible smoke emissions, except for periods not to exceed 5 minutes during any
two consecutive hours." This same requirement is also a requirement for refinery flares
complying with Refinery MACT 1 and 2 standards at 40 CFR 63.11 and as such, the data
collected from this study in no way supports the argument that the EPA should not limit smoking
from flares because it was collected under this purview. However, even if the data in figure 12 of
the "Appendix-Aerodyne Research Mobile Laboratory Particulate Measurements TCEQ 2010
83
-------
Flare Study" were collected from flare flame smoke, it clearly shows that not just elemental
carbon, but organics, which were not speciated and could contain HAP, are also emitted.
Therefore, we conclude and continue to believe that it is appropriate to limit smoking of flares
under the Refinery MACT standards.
In our review of the existing MACT requirements for flares, we found that there are very little or
no on-going compliance requirements for the visible emissions limit. We agree that the 1994
RTC specifically stated that "This [no visible emissions] requirement was intended to provide a
practical method for occasional observation. While this paragraph does not state the frequency
that Method 22 must be applied, the EPA certainly did not intend for facilities to continuously, or
even daily, monitor the flare to comply with the no visible emissions requirement." However, we
now find that the lack of on-going monitoring requirements in the General Provisions to be a
major stumbling block in our efforts to ensure refinery owners or operators are complying with
the emission limitation "at all times." We disagree that performing a daily 5-minute EPA Method
22 observation is overly burdensome and we are finalizing these requirements as proposed.
However, we also agree that daily EPA Method 22 monitoring for only 5 minutes during the day
would not be as effective at ensuring continuous compliance with the visible emissions standard
as "continuous" or "on-going periodic review" of video camera displays. Therefore, we are
including in the final rule the use of video camera monitoring as an alternative to the proposed
daily EPA Method 22 monitoring requirement.
3.3.2 Time allowed for visible emissions
Comment 1: Several commenters suggested that the proposed changes will force many flares to
operate near the incipient smoke point and that in order to have enough time to fine-tune flare
operations in the event the incipient smoke point is reached, the EPA should extend the not to
exceed visible emissions requirements in 40 CFR 63.670(c) from a period of 5 minutes to 10
minutes during any 2 consecutive hours.
Several commenters suggested that allowing 5-minutes of smoking or visible emissions in a two-
hour period from flares is unacceptable and that the EPA should prohibit visible emissions from
flares altogether as flare smoking can release large volumes of HAP, toxic soot, and fine
particulate matter emissions.
Response 1: Neither set of commenters provided data to support their position that a longer
allowance period for visible emissions or that no allowance period should be provided. In the
preamble to the proposed rule, we determined that a short allowance period was appropriate. It is
unrealistic to suggest that flares can be effectively operated at high combustion efficiencies while
never having a short period of visible smoke. Flare vent gas flow rates can change abruptly, and
even with advanced steam flow controls, an abrupt increase in flare vent gas flow can cause a
short period of smoking while steam system flow rates are adjusted. These adjustments should
take no more than the 5-minutes, so we considered the historical 5-minute allowance to be
reasonable and consistent with the best-performing flares.
We considered but rejected a visible emissions limit allowance of 10 minutes. Given the
improved vent gas flow monitoring and steam system controls projected in our flare impact cost
84
-------
estimates, we determined that refinery owners and operators would be able to meet the flare
operating limits without increased incidence of smoking.
Comment 2: Two commenters suggested that the EPA remove the requirement in 40 CFR
63.670(h) to extend the Method 22 observation period to 2 hours when visible emissions
observations are made and occur at any point during the 5-minute period. The commenters
suggested that there is no value added by this requirement, especially when other monitoring
(e.g., video and/or operator observation, daily visible emissions (VE) check under state rules like
TCEQ rule 111) occurs routinely, that some situations will arise where high flare venting occurs
only for a short period of time, and that any instances of smoke emissions (greater than 5 minutes
in 2 hours) are already reported as part of the semi-annual Title V deviation reports.
Response 2: We disagree with the commenter that there is no value to require an extended
monitoring period of 2 hours in circumstances where visible emissions are observed and persist
for at least one minute. The requirement is to operate with no visible emissions except for
periods not to exceed 5 minutes during any 2 consecutive hours. Where visible emissions are
observed for a period of at least one minute, it is important that the owner or operator tracks and
documents whether there are a series of short 1 to 2 minute smoking events that occur
sporadically over a period of 2 hours. The only possible means to demonstrate whether the
source is in compliance with the no visible emissions requirement is to observe the flare flame
over that 2 hour period. We note that in the final rule we have provided for the use of video
camera surveillance monitoring as an alternative to EPA Method 22 monitoring.
3.3.3 Monitoring requirements for visible emissions
Comment 1: Several commenters suggested that the visible emissions monitoring requirements
proposed at 40 CFR 63.670(h) not be finalized. Alternatively, they suggested that if these
requirements are maintained, they should be revised and clarified, or that that the EPA allow a
work practice type approach where use of a video camera or visual observation to monitor and
manage visible emissions from flares in place of the EPA Method 22 monitoring requirements
that were proposed. Commenters suggested that if the EPA does decide to maintain the visible
emissions monitoring requirements that the EPA:
Remove the initial visible emissions determination or clarify its intent of applying as just
the first time that a flare becomes subject to the requirements at 40 CFR 63.670(h);
Include only a simple requirement to maintain a record of the time and duration of any
visible flare emissions and any greater than 5 minute occurrence in 2 hours be reported in
the next periodic report;
Clarify that two observers are not required given that rest periods are needed;
Allow sources to stipulate that the 5 minute period has been exceeded in lieu of two or
more hours of monitoring;
Clarify that if a visible emission occurrence has been performed before the daily
monitoring has been carried out that this counts for the daily observation;
Allow a second tier of monitoring via Method 9 using a trained observer as follow-up to
Method 22 observations where some level of visible emissions are noted;
85
-------
Adopt a similar approach to visible emissions as those in the TCEQ rule 30 TAC Chapter
111, Subchapter A- Visible Emissions and Particulate Matter, Division 1: Visible
Emissions Rule section 111.111; and
Allow some relief when weather inhibits observation.
Response 1: We are retaining the proposed EPA Method 22 observations but we have included
an alternative to use video surveillance cameras provided that the cameras are installed at an
appropriate vantage point where the camera would reasonably be expected to continuously
record (at least one frame every 15 seconds) as well as output to a control room or other
continuously manned location the video images of the flare and smoke generated from the flare.
We confirm that the "initial visible emissions demonstration" is only for newly-affected flares
and that the "subsequent visible emissions observations" pertains to all subsequent daily
evaluations. We agree with the commenter that, if the 5-minute observation is extended to 2
hours and the observer has already observed 5 minutes of visible emissions, the observation
period is complete and we note that this is already specified in Section 11.4.2 of Method 22. We
also are clarifying that a single observer may be used to perform the 2 hour observation and that
the observer is to observe the rest breaks required in Section 11.4.3 of Method 22 and must not
include those rest breaks in the 2 hour observation period (as specified in Section 11.4.1 of
Method 22).
The commenters appear to suggest that clarification is needed that, if a required 5-minute
observation is needed prior to the "daily 5 minute observation" then the daily 5-minute
observation is not needed. The basic requirement is that the visible emissions test must be
conducted at least once per day using an observation period of 5 minutes. Any 5-minute
observation period satisfies this requirement. We do not distinguish this as a "daily observation"
separate from the requirement to monitor for 5 minutes if visible emissions are observed except
to note that, even if you have satisfied your "daily observation" requirement, additional 5-minute
observations are required when visible emissions are observed. Refinery owners and/or operators
are required to record all periods of observed visible emissions and must report all 2-hour
periods where the visible emissions were observed for more than 5 minutes in a given 2-hour
period. Therefore, while any 5-minute observation of visible emissions satisfies the "daily
observation" requirements, any new instance of visible emissions will require a new observation
period, even if there was an exceedance period earlier in that day.
It is unclear why the commenter requests to perform Method 9 monitoring "as follow up to
Method 22." Observer certification is more stringent for Method 9, and quantification of the
opacity is not required. We consider the use of video surveillance as a more reasonable and more
applicable alternative to direct Method 22 monitoring, and we have provided this as an
alternative in the final rule, which should also address weather-related observation issues. We
find that the proposed requirements are similar to those specified in TCEQ Title 30 Rule 111.111
except that they allow Method 9 and they do not require readings during process upsets. We
consider that we have adopted an approach similar to that used in the TCEQ rule but we disagree
with the provisions to exclude periods of upset from the visible emissions requirement.
Comment 2: A few commenters suggested that in order to encourage flare operation as close to
the incipient smoke point a possible that the EPA should make changes and clarifications to the
86
-------
definition of visible emissions as it relates to the Method 22 procedure. Specifically, the
commenters suggested that the language be revised to read: "Smoke emissions means visible
emissions persisting beyond one flame length from the visible flame tip. Smoke occurring within
the visible flame is not considered a smoke emission."
Response 2: We agree that black smoke within the flame envelope should not be considered
visible emissions, but we note that EPA Method 22 already clarifies that "Smoke emissions
means a pollutant generated by combustion in a flare and occurring immediately downstream of
the flame. Smoke occurring within the flame, but not downstream of the flame, is not considered
a smoke emission." Flame lengths can vary widely depending on the velocity of the release, and
one flame length can be too far from the flame for the visible emissions observation. While we
recognize that identifying the exact location where the flame ends is somewhat subjective, the
commenters provided no data to support that EPA Method 22 needs to be amended. Therefore,
we are not amending EPA Method 22 directly or as it applies to flares subject to the Refinery
MACT 1 standards.
Comment 3: Several commenters suggested that the additional burden to require refiners to have
a visible emissions evaluator available for 2-hour daily checks and during any smoking incident
is not reflected in the record or the Information Collection Supporting Statement. Commenters
also stated that plant personnel's primary task should be to focus on alleviating the issues
causing visible emissions at the flare rather than just performing a visual observation.
Additionally, commenters suggested that the EPA clarify that personnel that are conducting
Method 22 observations need not be certified as some situations could arise where flaring ceases
before a certified individual is available to perform the visual observation.
Response 3: The daily check only requires 5-minute observations and these observations were
included in the ICR burden estimate. EPA Method 22 does not require the same certification
procedures as EPA Method 9, but rather, only requires that an observer be knowledgeable with
procedures for determining the presence of visible emissions. The commenter did not provide
any reason why it would be appropriate to waive this requirement and we believe it is important
that an observer understand how to perform this type of monitoring. We clarify in this response
that all that is required for personnel that will be expected to conduct the visible emissions
observations is that they have read Method 22 and understand the basic requirements for visible
"smoke emissions" monitoring for a flare. As noted elsewhere in this document, we are also
including an option for refiners to use a video surveillance camera to conduct the visible
emissions monitoring, which will eliminate the need to have a separate Method 22 observer.
Observation via the video camera feed can be conducted readily throughout the day and will
allow the operators of the flare to watch for visible emissions at the same time they are adjusting
the flare operations.
Comment 4: One commenter supported the proposed visible emissions observation
requirements while another commenter wanted the Agency to provide a webcam feature for
citizens living near refineries to use and actually monitor refinery flares in real time. Another
commenter wanted more information about flaring and air permits for the ExxonMobil
Chalmette refinery, wanted the EPA to provide additional clarity on what was meant by "visible
87
-------
emissions" since it can be subjective, and wanted the EPA to provide state agencies with training
regarding the no visible emissions requirement.
Response 4: While we are allowing an option for refinery owners and operators to use video
surveillance cameras for visible emissions monitoring, we are not requiring them to provide
public access to the video feed. Refiners must keep records of either their Method 22
observations or the continuous surveillance video records, but in general we would not require
additional reporting for one option over what would be required for refiners opting to comply
with the Method 22 observations. Regarding one commenter's request for more information
about flaring and air permits for a specific refinery, we note that request is beyond the scope of
this rulemaking. Information relating to permitting for individual facilities should be directed to
the delegated air agency, e.g. the Louisiana Department of Environmental Quality in the case of
ExxonMobil's Chalmette refinery.
We disagree that additional clarity is needed for the term "visible emissions" and that the EPA
needs to provide state agencies with training regarding this requirement as no revisions to EPA
Method 22 are being made in this action. The term "visible emissions" is the same as smoke
emissions as defined in EPA Method 22: "Smoke emissions means a pollutant generated by
combustion in a flare and occurring immediately downstream of the flame. Smoke occurring
within the flame, but not downstream of the flame, is not considered a smoke emission." The
EPA Method 22 is available online at: http://www.epa.gov/ttn/emc/promgate/m-22.pdf. Persons
interested in understanding the method should read the method and the references cited in the
method if more information is needed. Questions on EPA methods may also be directed to the
Emission Measurement Center. A contact list is available online at:
http ://www. epa. gov/ttn/emc/ staff dir. html.
3.3.4 Need for flare tip velocity requirement
Comment 1: Several commenters suggested that the EPA remove the flare tip velocity
requirements from the rule altogether. To support this request, commenters suggested that flares
operating at low velocities will achieve a high level of combustion efficiency due to the new
flare combustion zone requirements and that at high exit velocities any lit flame which has
sufficient heat content to support combustion will achieve the same result. Also, they suggested
that if the maximum velocity was limited to only 400 feet (ft)/second (s) that new flare capacity
would be needed which would increase flare purge gas rates as well as greenhouse gas
emissions. One commenter suggested that it may be difficult for the EPA to remove the
numerical requirements that have become conventionalized over time and as an alternative
recommended that a new equation representing required heat content as a function of velocity be
developed on a combustion zone heating value basis that has maximum of 600 British thermal
units (Btu)/ standard cubic feet (SCF) at the high flare tip velocity end of 400 ft/s or greater.
Response 1: We disagree that we should abandon the flare tip velocity requirements. At high
velocities, flare flame lift-off can occur leading to reduced combustion efficiencies and
potentially complete loss of flame. Additionally, the flare tip velocity limits ensure that refiners
operate their flares within a stable operating regime. The commenters provide no data to suggest
that high flare combustion efficiencies are achievable for traditional, elevated flare tips at
88
-------
velocities over 400 ft/s as well as provide no data to suggest what the stable flare operating
envelope would look like for elevated refinery flares operating above 400 ft/s. We do not expect
that flare velocities will be over 400 ft/s under normal flaring events so we disagree that this
requirement will cause refineries to have to build new flare capacity. However, as detailed later
in this response to comment document, we are also implementing a work practice standard for
these significant events to minimize their occurrence while also preventing the need to build new
flares in efforts to prevent exceedances of the 400 ft/s flare velocity limits.
Comment 2: A few commenters suggested that the EPA revise the proposed standards for flare
tip velocity to address pressure-assisted (or multi-stage ground) flares because they are designed
to operate above the 400 ft/s regulatory threshold and that other new requirements related to
visible emissions and SSM proposed in this rule will potentially drive facilities to consider this
technology going forward. Other commenters suggested that the EPA remove the flare tip
velocity equations for certain flare designs and operations altogether, particularly pressure-
assisted flares and non-assisted flares. As an alternative, one commenter suggested that the EPA
provide the ability for facilities to set site-specific flare tip velocity evaluations when the
equations proposed in 40 CFR 63.670(d) are not appropriate.
Response 2: We do recognize that pressure-assisted flares have specially-designed flare tips
capable of achieving high combustion efficiencies at high flare tip velocities and are in the
process of reviewing alternative operating parameters for such installations on a site specific
basis. We also note that we already proposed methods to allow refinery owners or operators to
develop alternative operating parameters or limits, including, but not limited to, flare tip velocity
limits.
3.3.5 Velocity limit and calculation method
Comment 1: Several commenters suggested that the EPA should reassess the flare tip velocity
requirements for flares combusting low net heating value gases, particularly hydrogen, because
the current requirements in the General Provisions provide a different flare tip velocity equation
for flares that have a diameter of 3 inches or greater, are non-assisted, and have a hydrogen
content of 8.0 percent (by volume) at 40 CFR 63.1 l(b)(6)(i)(A) and that flares combusting lots
of hydrogen will not be able to achieve a sufficient net heating value to comply with any flare tip
velocity limit other than 60 ft/s even though hydrogen flame stability at high tip velocities is well
established.
Response 1: As described in Section 3.5.2 of this document, we are providing an allowance to
use an effective net heating value of hydrogen of 1,212 Btu/scf. With this provision, we consider
that the flare velocity equations will work effectively for flares that have high hydrogen content.
Comment 2: Several commenters suggested that the EPA revise the averaging time for
demonstrating compliance with the flare tip exit velocity provisions from a 15-minute block
average period to either a 1-hour or 3-hr block average and that the EPA revise the volumetric
flow rate calculation requirements to be at standard conditions (pressure of 14.696 pounds per
square inch absolute (psia) and temperature of 68 degrees Fahrenheit (°F)) and not actual
conditions. To support these suggestions, commenters pointed out that the Texas highly reactive
89
-------
volatile organic compounds (HRVOC) Vent Gas Control rule under 30 TAC 115, Subchapter H,
Division 1 [section 115.726(d)(4), 115.725(d)(6)], which is considered to be among the most
stringent set of monitoring provisions for flares in the US, utilizes a 1-hr block average
monitoring requirement, that facilities should not incur a violation when flare tip velocity
requirements are exceeded because flares can handle highly variable and intermittent waste gas
streams and use of flares in these situations is a proper emission control technique, and that the
current exit velocity flare requirements in the General Provisions use standard conditions and not
actual conditions.
Response 2: We disagree with the commenters' suggestion that the flare tip velocity should be
evaluated on a 1-hour or 3-hour average. The primary purpose of the maximum velocity
requirements is to ensure that flares are operated within a stable operating regime that prevents
"flame lift off or complete extinction of the flame if flare vent gas is released too fast. Even a
small period of time of flame lift-off or extinction can make it impossible for a flare to achieve a
98 percent destruction efficiency over a 3-hour period (the typical testing period for other control
devices used to control Refinery MACT organic HAP emissions). As the commenters noted,
flares often receive highly variable and intermittent waste gas streams and using a 1-hour or 3-
hour average can easily mask short term events when large quantities of gas may be expelled
from a flare. For example, if a flare that receives a vent gas stream with a net heating value of
600 Btu/scf the Vmax calculation would yield a maximum permitted velocity of 135 ft/s. If the
event lasted approximately 30 minutes releasing vent gas at a velocities of 200 ft/s during two
15-minute periods but the flow rate was under 20 ft/s during the previous and subsequent two 15-
minute periods, the average velocity would be 110 ft/s across the time period. However, during
periods of the highest flow, it is likely that the combustion efficiency was reduced because the
flame would be expected to have operated in an unstable regime and could potentially be
completely extinguished. Due to potential flame stability issues at high flows (and the sharp
reduction in combustion efficiencies that result from unstable flare flames), we find it critical
that the flare tip velocity (as well as the other operating limits) are determined and adjusted as
quickly as possible, which we determined to be based on the 15-minute block periods (due to
time needed for compositional analysis determinations when using a gas chromatograph).
With respect to the comments stating that the flare tip velocity should be determined based on
standard conditions, in our initial review of the historical 1985 flare efficiency report, we noted
that the exit velocities were reported as "Actual Exit Velocity (ft/s)." There was a footnote
indicating that the calculation was based on the open areas of the flare, but no other indication of
how the flow volume was determined. Based on the table headings, we interpreted this to mean
the reported velocities were determined based on the actual exit velocities (i.e., actual volumetric
flow divided by the open area). We do note that 40 CFR 63.1 l(b)(7)(i) includes the following
instructions: "The actual exit velocity of a flare shall be determined by dividing by the
volumetric flow rate of gas being combusted (in units of emission standard temperature and
pressure), as determined by Test Method 2, 2A, 2C, or 2D in appendix A to 40 CFR part 60 of
this chapter, as appropriate, by the unobstructed (free) cross-sectional area of the flare tip"
(emphasis added), lending credence to the commenter's suggestion. We reviewed the 1985 report
again and found it very difficult to determine the actual basis of the "Actual Exit Velocity"
calculation. Appendix D provides some detail regarding the orifice meter flow rate
measurements. After our review of this equation, and understanding that orifice meters generally
90
-------
correlate to a mass flow rate, we agree that the volumetric flow rate was determined using a fixed
molar volume correction factor (i.e., "standard conditions") although it is impossible to tell from
the report what conditions were used as "standard conditions." In any event, based on our further
examination, we agree with the commenter that for flare tip velocity should be determined based
on the volumetric flow at standard conditions, consistent the procedures in 40 CFR
63.1 l(b)(7)(i), and we have revised the rule accordingly.
Comment 3: Several commenters suggested that flares that would be demonstrating compliance
with the combustion efficiency parameter of combustion zone combustibles concentration (Ccz)
using a total hydrocarbon (THC) analyzer would be constrained to a maximum flare tip velocity
of 60 ft/s and that this would effectively eliminate the use of using a THC analyzer as a viable
monitoring option for demonstrating compliance with the flare tip velocity requirements since
you need to know the net heating value of the vent gas (NHVvg). One commenter suggested that
the EPA should add flexibility to the rule to allow for the use of process knowledge or
engineering calculations to determine the NHVvg for flare tip velocity compliance
demonstrations when there is no continuous on-line monitor or monitor capable of determining
NHVvg.
Response 3: First, we are not finalizing operating limits related to Ccz or combustion zone lower
flammability limit (LFLCZ). These operating limits were included primarily to provide more
appropriate limit for flares with high hydrogen content. We have determined that these
alternatives are no longer necessary given the adjusted net heating value for hydrogen of 1,212
Btu/scf included in the final rule. We agree with the commenter that facilities that elected to use
the proposed THC monitoring option would have had to comply with the 60 ft/s flare tip velocity
limit or they would have had to install a calorimeter to determine the net heating value of the
vent gas. As such, the proposed THC monitoring option had limited utility and we have removed
this monitoring option from the final rule.
Comment 4: One commenter suggested that the EPA clarify the language at 40 CFR
63.670(k)(l) by adding the word "exit" after cross sectional area of the flare tip.
Response 4: We are unclear how adding the word "exit" to the flare tip discussion further
clarifies the discussion. Based on our review of the 1985 flare study, we understand that the
proposed approach (to reduce the flow area based on obstructions in the flare tip) is consistent
with the approach used in determining the "actual exit velocity" reported in the 1985 flare study.
We also note that we have received comments from refinery owners or operators suggesting that,
if these obstructions end before the flare tip "exit" (even if only by Vi inch), than the flare tip
"exit" should not be reduced by these obstructions. We disagree because this would not be
consistent with the treatment of the data used to determine Vmax. We specifically include in 40
CFR 63.670(k)(l) the obstructions in the flare tip that must be considered in determining the
unobstructed cross-sectional area. We find that the description provided clearly reflects our
intent and that including the word "exit" will actually lead to inappropriate determinations of the
unobstructed cross-sectional area.
91
-------
3.4 Flare SSM Issues
3.4.1 VE and velocity exclusions during SSM events
Comment 1: A few commenters suggested that the flare tip velocity and visible emissions
limitations proposed by the EPA have not been achieved and are not achievable during periods
of malfunction or emergencies, even by the best performing 12 percent of sources and that the
EPA is significantly misapplying the D.C. Circuit Court ruling in Sierra Club v. EPA,
551F.3dl019 (D.C. Cir. 2008), cert, denied, 130 S. Ct. 1735 (2010). Because of this, the
commenters suggested that the EPA is required under section 112 to demonstrate that
compliance with proposed flare standards is compliant with the statute for all operating
conditions and that the EPA must either establish a numeric emission standard in accordance
with section 112(d)(2)-(3) or, if that is not feasible, establish a work practice standard under
section 112(h) that is achievable under these operating conditions.
Response 1: For the reasons provided in the preamble to the final rule, we are establishing work
practice standards that would apply when a flare is operating above its smokeless capacity.
Comment 2: A few commenters suggested that flare tip velocities in excess of 400 ft/sec and/or
the presence of smoke (visible emissions) do not indicate a reduced combustion efficiency from
flares. The commenters suggested that available data demonstrate that HAP and VOC
destruction efficiencies are maintained at emergency release velocities as long as a flame is
present and the new combustion zone requirements are met; that the data on high velocity flaring
is consistent with combustion theory, which shows that high velocity flames result in better air
entrainment and mixing and result in higher combustion efficiency; that both literature searches
and computational evaluations noted that smoke does not necessarily indicate a significantly
inefficient combustion; that soot formation typically comprises less than 0.5 percent of unburned
hydrocarbons; and that smoking flares can have a combustion efficiency of greater than 99
percent.
Another commenter suggested that when flares are releasing black smoke, it is an indicator of
incomplete combustion, and based on their conversation with Houston city staff and at least one
flare supplier, operating a flare in this smoking regime could lead to destruction efficiencies as
low as 50 to 70 percent. The commenter also suggested that flares operating in this regime burn
heavier aromatics less efficiently, which poses a greater potential for adverse health impacts on
surrounding communities.
Response 2: The commenters do not provide any verifiable data to support their conclusion that
gases discharged from a flare in excess of 400 ft/s (270 miles per hour (mph)) will remain in the
combustion zone for sufficient time to ensure high combustion efficiencies or that these
velocities will not lead to flame instability for conventional elevated flares. Even if the high
velocity does not cause flame instability, there are no data to support claims for traditional
elevated flare tips that "if a flame is present, combustion efficiencies are high." While we agree
that it may be possible to achieve high combustion efficiency when small quantities of smoke are
produced for short time periods (hence the allowance to have 5 minutes of visible emissions in a
two hour period), we agree with the commenter that suggests destruction efficiencies can become
92
-------
greatly reduced during periods of high velocities when large quantities of black soot are
produced, as seen during "hydraulic load" events. We do not have direct emissions test data from
large industrial flares during such an event, but based on the flame stability studies performed in
the 1980s, we conclude that flame stability issues that produce significantly lower flare
combustion efficiencies are likely to occur at flare tip velocities exceeding the limits provided in
the rule and/or when significant black smoke is produced. Therefore, we are retaining the
proposed limits on flare tip velocities and visible emissions. However, as noted previously, we
are providing a work practice standard that apply specifically when the volumetric flow rate
exceeds that smokeless capacity of the flare. Below the smokeless capacity of the flare,
exceedances of the visible emission limits or flare tip velocity limits are a deviation of the
standard. Above the smokeless capacity of the flare, exceedances of the visible emission limits or
flare tip velocity limits triggers the work practice standard (i.e., root cause analyses and
corrective actions with hard limits on these events that can occur in a 3-year period) as a means
to reduce the frequency and magnitude of these events.
3.4.2 Other flare SSM issues
Comment 1: One commenter suggested that the loss of exemptions for periods of startup,
shutdown, and upset conditions should drive refiners to make improvements in equipment
integrity and operability, lead to better maintenance programs, and reduce breakdown conditions
to diminish upset flaring.
Response 1: We agree that emissions limits that apply at all times encourage sources to maintain
equipment and improve procedures to ensure proper operation.
Comment 2: One commenter suggested that the use of flares is a backdoor malfunction
exemption that EPA must prohibit. The commenter claimed that the CAA requires that emissions
standards apply at all times and prohibits exemptions for malfunction. The commenter claimed
that the EPA's proposal to allow facilities to route refinery fuel gas and miscellaneous process
vent gas to flares caused by malfunctions would violate these requirements; that some gas routed
to flares from process upsets comes from the FCCU and since the FCCU is subject to an
independent limit on organic HAP that this illegally circumvents the applicable requirements;
and that when refinery fuel gas that is normally burned in heaters and boilers is routed to a flare
that the control efficiency decreases from a 99.9 percent control to 98 percent control efficiency.
Response 2: We disagree that the use of flares is a "backdoor" malfunction exemption. As the
commenter recognizes, flares are a means of controlling emissions and, at the time the EPA
promulgated Refinery MACT 1, the EPA determined that use of a flare meeting certain
requirements in the General Provisions at 40 CFR 63.11 was an equivalent option to achieving a
98 percent control efficiency, the MACT floor level of control. We also note that the use of
flares is specifically provided as an acceptable control device to comply with the organic HAP
limits for FCCU in Refinery MACT 2. It is irrelevant whether the flare is receiving these organic
HAP streams as a primary (routine) control device or as a back-up control system to other
controls. We consider that flares operated in accordance with the requirements in the final rule
will achieve the necessary MACT floor level of control and are acceptable control devices.
93
-------
Therefore, we are not circumventing the MACT control requirements by allowing gases
produced as a result of an upset or malfunction to be discharged to the flare.
With respect to heaters and boilers, we do not agree that all heaters and boilers achieve 99.9
percent control efficiency at all times or that the long-term average control efficiency for flares is
only 98 percent. Combustion efficiencies in heaters and boilers are dependent on a variety of
factors (fuel quality, air feed rates, burner design, etc.) and we have historically determined that
these combustion controls achieve 98 percent control (or 20 ppmv total organic carbon). As with
flares, the combustion efficiency may vary with time and the 98 percent control level was
established as the minimum acceptable control efficiency over the performance test period (3-
hour average). The flare requirements we are finalizing are designed to ensure flares meet this
minimum combustion efficiency for any 3-hour average period, consistent with the MACT
control requirements. It is clear from the flare performance test data that flares can achieve 99.9+
percent controls at certain times and we estimated that the long-term nationwide average control
efficiency achieved by flares meeting the final flare requirements will be well-over 99 percent,
which is equivalent to the combustion performance of process heaters and boilers.
3.5 Combustion zone operating limits (general, non-H2-olefin)
Comment 1: One commenter supported the EPA's proposal on flares but suggested that
additional improvements are required under sections 112(d) and (f) of the CAA. Specifically, the
commenter suggested that the EPA: (1) require flares to achieve 98 percent control efficiency,
(2) limit routine flaring (pursuant to sections 112(d)(2),(3) and (6)), and (3) require continuous
monitoring of the flare gas in order to assure continuous compliance with all applicable limits.
Response 1: As discussed in the preamble and elsewhere in this section, we established the flare
operating requirements and monitoring provisions in order to ensure flares are achieving the
targeted 98 percent or higher destruction efficiency.
Regarding the request that the EPA limit "routine flaring," we note that where Refinery MACT 1
and 2 allow the use of flares as a control option, they don't specify this control device must be
used to demonstrate compliance with the applicable MACT standards. In fact, refinery owners or
operators can elect to use non-flare control devices (e.g., thermal oxidizers or other control
systems) to demonstrate compliance with these same MACT standards and that these other non-
flare control devices also have no "hard caps" or limits on their routine use. Therefore, we do not
consider it appropriate under the MACT requirements to limit routine flaring to control Refinery
MACT regulated gas streams.
Finally, as to the comment regarding continuous monitoring, the final rule generally requires
continuous monitoring of flare gas in order to ensure continuous compliance with the applicable
limits. We note that, as proposed, flare owners or operators may elect to use once every 8-hours
grab samples for compositional analysis. We expect this alternative will only be used by
"emergency only" flares that very seldom receive flare waste gas and owners or operators of
these flares must still install continuous monitoring systems for determining flare vent gas flow
rates. We also included a provision for flare owners or operators that have a consistent
composition of flare vent gas (such as for gasoline loading racks) to determine, through an initial
94
-------
sampling program, a fixed net heating value of the flare vent gas to be used in the calculations.
See the preamble to the final rule and Section 3.0 of this response to comment document for
further discussion on this alternative. We determined that these alternatives are appropriate for
these situations and will provide a means by which refinery owners or operators can demonstrate
continuous compliance with the final operating limits for each 15-minute block average period.
Comment 2: One commenter suggested that the EPA should provide flexibility in monitoring of
flow rates and composition for infrequently used flares and instead require that the flares be
operated by trained and certified personnel when they need to be used.
Response 2: The operating limits apply to all flares that manage Refinery MACT regulated gas
streams. We proposed alternative monitoring methods for flares, some of which were specifically
targeted for flares that were not routinely used, such as the use of pressure monitoring systems
and engineering calculations to determine flow rates and the use of grab samples to determine
gas composition. We have finalized these monitoring alternatives and have also added an
alternative for flares with a consistent composition of flare vent gas (such as for gasoline loading
racks) to determine, through an initial sampling program, a fixed net heating value of the flare
vent gas to be used in the calculations. We have determined that these alternatives are
appropriate for these situations and will provide a means by which refinery owners or operators
can demonstrate continuous compliance with the final operating limits for each 15-minute block
average period, while helping to reduce the burden related to installing and operating a
continuous monitoring system for a flare that receives waste gas only a few times a year or only
once every few years. However, we disagree that a simple requirement to use "trained flare
operators" is sufficient to ensure that these flares are achieving 98 percent control efficiencies at
all times. We note that refineries currently use "trained flare operators" and it is not unusual to
find flares being over-assisted. Without the necessary information on flare vent gas properties, as
would be collected via the monitoring requirements in the final rule, a "trained flare operator"
simply does not have information required to properly operate the flare.
Comment 3: A few commenters suggested that the rule language should be flexible enough to
permit the use of direct combustion efficiency measurement as a potential future compliance
option and that the EPA should allow use of an Alternative Monitoring Plan (AMP) for flare
control and monitoring.
Response 3: We note that the General Provisions at 40 CFR 63.8(f) provide a mechanism by
which refinery owners or operators can apply for an AMP. Therefore, if a technology becomes
available by which refinery owners or operators can directly monitor flare combustion
efficiency, refinery owners or operators can request to use this alternative monitoring method.
We also note that we included provisions in 40 CFR 63.670(r) to provide a mechanism by which
refinery owners or operators can establish different operating limits, which may also include
different monitoring requirements.
Comment 4: Several commenters suggested that poor destruction efficiency from flares is not a
widespread problem and that overly stringent combustion zone metrics could lead to unnecessary
emissions of criteria air pollutants and greenhouse gas (GHG) from natural gas supplementation
beyond what is necessary. The commenters also stated that the EPA needs to account for
95
-------
situations of poorly mixed steam in the combustion zone and that steam flow rate measurements
could actually be lower at the flare tip than where they are being measured due to heat losses.
Response 4: We disagree that poor destruction efficiency from flares is not a widespread issue.
Many companies that use flares to control emissions have become more cognizant of the fact that
flares can be over-assisted and not achieve the intended destruction efficiencies. However, not all
companies are required to continuously monitor the combustion zone characteristics of the flare
and excessive quantities of steam or air can still be introduced in an effort to minimize smoking
events or limit the visible flame from the flare. Additionally, and as discussed further in the
preamble, we project that the flare operational and monitoring requirements being finalized have
the potential to actually reduce excess emissions of both criteria air pollutants (i.e., VOC) and
GHG.
Regarding the concerns about the introduction and monitoring of steam into the combustion
zone, refinery owners or operators should place steam flow monitors at a location that is
representative and also ensure they are efficiently operating flares in a manner where situations
of poorly mixed steam in the combustion zone are minimized. For properly installed steam
monitoring systems, we disagree that the steam flow rate would be lower at the flare tip due to
heat losses because of the relatively high heat capacity and heat of vaporization of steam. Given
the thermodynamic properties of steam, any heat losses between the representative measurement
location and the flare tip will have a negligible effect on the quantity of steam actually
introduced at the flare tip.
Comment 5: One commenter suggested that the EPA should encourage operation near the
incipient smoke point and provide an option for operating at or just above that point regardless of
the net heating value in the combustion zone.
Response 5: We disagree with the commenter that we should provide an option to operate at or
near the incipient smoke point. First, without the detailed monitoring systems that we proposed
and that we are requiring in this final rule, flare owners or operators will not have the
information needed to determine if they are at or near the incipient smoke point. In other words,
if no smoke is present, there would be no way to know if the system is being operated "near" the
incipient smoke point. Alternatively, a flare owner or operator would have to operate at all times
with some smoke present in the flame but not in the gas beyond the flame. This would be
impossible given typical variations in flare gas flow rates, so incidence of smoking would be
prevalent and likely in violation of the no visible emissions limitation.
Comment 6: Several commenters generally supported the proposed suite of flares provisions,
use of combustion zone metrics instead of vent gas metrics, and the 15-minute sampling period
proposed by the Agency.
Response 6: We appreciate the support for the proposal. We have retained several of aspects of
the proposal noted by the commenter, such as use of combustion zone metrics and a 15-minute
block average, but have modified the flare operating requirements as described in the preamble
to the final rule.
96
-------
3.5.1 Applicability (steam-assist only or all flares)
Comment 1: Several commenters suggested that the EPA should only apply the proposed flaring
requirements at 40 CFR 63.670 and 63.671 to steam-assisted flares and that there is no
justification for imposing the revised requirements on air-assisted, pressure assisted, and
unassisted flares and that these flares should remain subject to the requirements at 40 CFR 63.11.
In support, commenters suggested that the EPA has not collected adequate performance data on
which to establish or justify limits for flares that are not steam-assisted; that the EPA has not
properly considered the differences in design, waste gas composition, and use between these
flare types and steam-assisted flares; and that the EPA has not properly evaluated the costs and
benefits associated with the flare types that are not steam-assisted of meeting the limits derived
from the steam-assisted flare dataset.
Response 1: First, while we acknowledge that there are limited new data for other flare types,
we do not see any difference in the combustion efficiency curves between steam and unassisted
flares when we consider the combustion zone properties, which is consistent with combustion
zone theory. Furthermore, recent data clearly indicates that combustion efficiencies begin to
deteriorate at combustion net heating values above 200 Btu/scf and that an operating limit of 200
Btu/scf in the flare vent gas (which is the combustion zone gas for unassisted flares), as currently
provided in the General Provisions for unassisted flares, does not ensure that these flares will
achieve an average destruction efficiency of 98 percent. Finally, the data that we do have on air-
assisted flares clearly indicates that air-assisted flares can be over-assisted leading to poor
combustion efficiencies. Therefore, we consider the available pool of data, which includes data
for air, steam and unassisted flares, supports our conclusion that the combustion zone net heating
value target for all flares must be significantly higher than 200 Btu/scf in order to ensure flares
are achieving 98 percent control and that special provisions are also needed to ensure air-assisted
flares are not over assisted.
Comment 2: A few commenters suggested that the EPA should clarify that refinery flares are
only subject to the 40 CFR 63.670 and 63.671 requirements when they are receiving Group 1
Refinery MACT 1 or regulated Refinery MACT 2 vent streams. Specific examples of refinery
flares that commenters listed that never receive Refinery MACT 1 or 2 regulated vent streams
include: flares that receive fuel gas where there are no Group 1 emission points routed to that
fuel gas system, dedicated acid gas flares, dedicated hydrogen plant flares, flares that are
dedicated to non-HAP pressure storage (e.g. propane/butane spheres), and flares equipped with
flare gas recovery systems where no flaring of "regulated material" is taking place.
Response 2: We consider that the rule is clear and we are confirming in this response that flares
that do not receive any "regulated material" would not be subject to the new flare provisions in
40 CFR 63.670 and 63.671. We are adding a definition of "regulated material" to help clarify
this phrase. Nonetheless, we are hesitant to provide definitive answers to specific examples
provided by the commenter because there are potentials for "exceptions" to the general rule. For
example, the new flare requirements would not generally apply if only Group 2 streams are
controlled by the flare. However, if a refinery is using emissions averaging and is controlling a
Group 2 stream using a flare, we consider that the flare must be in compliance with the new flare
requirements in order to take emissions averaging credits. Similarly we expect that
97
-------
propane/butane spheres would not have sufficient HAP content to trigger Group 1 storage vessel
control requirements, so we generally expect that flares dedicated to serving these vessels would
not be subject to the new flare requirements. However, if the purity of these products in these
tanks allows 2 to 4% HAP materials (depending on the applicable Group 1 storage vessel
definitions), then this flare could be subject to the new flare requirements. Refinery owners or
operators should carefully review the rule requirements and definitions to assess the applicability
of the new provisions and are encouraged to submit an applicability determination for any flare
for which they are uncertain whether the new flare requirements apply in a specific application.
We do note that we expect that most fuel gas systems would receive a Group 1 stream since any
fuel gas recovery compressor that receives gas from a process unit in HAP service would be a
Group 1 miscellaneous process vent had these gases not been recovered in the fuel gas system.
Again, refinery owners or operators can submit an applicability determination for their specific
fuel gas system to determine if there is any question as to whether the fuel gas system receives a
Group 1 stream.
Comment 3: One commenter stated that available data strongly suggests that the 8% (vol.) rule
for hydrogen rich flaring for unassisted flares works well for the chemical industry practices but
that it may not work for hydrogen rich flaring in refineries.
Response 3: We are providing provisions to use 1,212 Btu/scf for the net heating value of
hydrogen, which accounts for the flammability of hydrogen. While this provision does not
directly allow gas streams with 8 percent hydrogen to comply with the rule regardless of the
other constituents in the gas stream (e.g., 8 percent hydrogen in an otherwise inert gas stream
would not have sufficient heat content to ensure high combustion efficiencies), it does greatly
improve the correlation between combustion efficiency and combustion zone net heating value.
Therefore, we are finalizing a provision to use 1,212 Btu/scf for the net heating value of
hydrogen. The final rule is applicable to flares at petroleum refineries; we are not revising the
requirements for flares in the chemical manufacturing industry or other industry sectors at this
time.
Comment 4: One commenter raised a concern that the proposed flare requirements would
become a template for future NSPS rule revisions and that these regulations may be misapplied
to other source categories with flares or to flares controlling non-VOC vent streams.
Response 4: We are only applying these requirements to petroleum refinery flares at this time.
If, in the future, we propose similar requirements for NSPS, the commenters can raise their
concerns in the context of that rulemaking.
3.5.2 Selection of parameters & limits
Comment 1: A few commenters suggested that the EPA based the proposed combustion zone
limits on an invalid data analysis (i.e., invalid use of CO2 PFTIR signal wavelengths and invalid
data acceptance criteria), that the one minute PFTIR data should not be used to establish
combustion efficiency correlations because of temporary shifts in the flare plume due to changes
98
-------
in wind speed or direction, and that the emission limits should be set so as to provide an equal
chance of false positives and negatives.
Response 1: We re-analyzed the data using the "approved" CO2 signal wavelength when it was
available. While the arguments provided by the commenters regarding the use of 1 minute
average data are not compelling, we note that the flare gas composition data used to determine
the NHVcz is generally based on gas chromatography (GC) analysis, with cycle times of 10 to 15
minutes. Because the compositional analysis data are not available each minute, we compiled the
data using approximate 15-minute run averages, as suggested by the commenters, and we then
evaluated the impacts of different operating limits using the resulting run-average data set.
Details of this analysis are provided in the memorandum entitled "Monte Carlo Analysis of Flare
Performance Data" included in Docket ID No. EPA-HQ-OAR-2010-0682. This analysis
essentially confirmed that the proposed operating limits were reasonable although we finalizing
some revisions to the flare operating limits as described in the preamble to the final rule. We
disagree that we should develop limits that provide an equal chance of false positives as false
negatives because the operating limits are intended to ensure that all flares are achieving the
targeted 98 percent destruction efficiency at all times rather than half of the flares meeting that
target half of the time.
Comment 2: A few commenters suggested that the EPA should assign hydrogen a heating value
of 1,212 Btu/scf to more accurately reflect its flammability in a net heating value (NHV) basis
and that doing so is consistent with some recent flare consent decrees and will help reduce
natural gas supplementation for facilities complying only with the NHVcz metric.
Response 2: We evaluated the 15-minute average run data using the normal net heating value for
hydrogen of 274 Btu/scf and the suggested value of 1,212 Btu/scf, which is based on a
comparison of the lower flammability limit and the net heating value of hydrogen compared to
light organic compounds. Based on our analysis, using the 1,212 Btu/scf value for hydrogen
greatly improves the correlation between combustion efficiency and combustion zone net heating
value. Therefore, we are finalizing a provision to use 1,212 Btu/scf for the net heating value of
hydrogen. For more detail on the data analysis, see the memorandum entitled "Monte Carlo
Analysis of Flare Performance Data" included in the Docket ID No. EPA-HQ-OAR-2010-0682.
Comment 3: One commenter supported the EPA's decision to not include perimeter assist air in
the calculation of combustion zone metrics (i.e., NHVcz, LFLCZ, and Ccz).
Response 3: We appreciate the support.
Comment 4: One commenter generally supported and urged the EPA to stand by the updated
flare regulations.
Response 4: We appreciate the support. We are finalizing the requirements with revisions as
described in the preamble to the final rule.
99
-------
Comment 5: One commenter requested that the EPA provide actual refinery flare operating data
that confirms that affected flares can achieve the proposed standards under the range of process
conditions and control scenarios found in the refining industry.
Response 5: Many of the flare performance studies, such as those conducted at Marathon and
Flint Hills Resources,85'86"87 were conducted on actual industrial flares at petroleum refineries,
which indicate affected flares can achieve the proposed requirements.
Comment 6: One commenter requested that the EPA formulate and provide the equation for
calculating the combustibles concentration, Cvg, in proposed section 63.670(1)(6) to be used
when a total hydrocarbon analyzer is used for total volumetric hydrocarbon concentration.
Response 6: At the time of proposal, we expected that a THC analyzer calibrated using propane
would provide a direct measure of Cvg. However, we are not finalizing the use of Cvg operating
limit because we are providing an allowance to use 1,212 Btu/scf as the net heating value for
hydrogen. The Cvg parameter (as well as the lower flammability limit (LFL) parameter) was
proposed largely to account for instances where the net heating value did not correlate well with
combustion efficiency. With the hydrogen adjustment, the net heating value alone provided
excellent correlation with combustion efficiency and the use of Cvg is no longer needed.
3.5.3 Allowance to use any parameter at any time
Comment 1: One commenter suggested that no flare monitoring system currently exists that is
capable of both monitoring and simultaneous adjusting flare operations to ensure compliance
with at least one of the three parameters in the combustion zone proposed by the EPA (i.e.,
NHVcz, LFLcz, and Ccz) for each measurement taken. The commenter further stated that while
such an advanced control system may be possible to implement, it estimates a cost of such a
system to be $850,000 and that the EPA has not justified such a cost, especially when recent
consent decrees use values in the vent gas rather than the combustion zone gas.
Response 1: In the proposal, we reasoned that if a GC was used to determine gas composition,
calculation of all three parameters would be straight forward and one could easily determine the
parameter that would allow the most steam addition (the largest compliance margin). We are not
finalizing the proposed approach but rather requiring refinery owners or operators to comply
85 Clean Air Engineering, Inc. 2010a. Performance Test of a Steam-Assisted Flare with Passive FTIR. Prepared for
Marathon Petroleum Co., LLC, Texas Refining Division, Texas City, Texas. May. Available at:
www.tceq.texas.gov/assets/public/implementation/air/rules/Flare/2010flarestudy/mpc-txc.pdf.
86 Clean Air Engineering, Inc. 2010b. Performance Test of a Steam-Assisted Elevated Flare with Passive FTIR -
Detroit. Prepared for Marathon Petroleum Co., LLC. Detroit Refinery, Detraoit, Michigan. November. Available at:
www.tceq.texas.gov/assets/public/implementation/air/rules/Flare/2010flarestudy/mpc-detroit.pdf.
87 Clean Air Engineering, Inc. 2011. PFTIR Test of a Steam-Assisted Elevated Flares - Port Arthur. Prepared for
Flint Hills Resources Port Arthur. LLC. Port Arthur Chemicals, Port Arthur, Texas. June 17. Available at
http://www.regulations.gov/, Docket Item No. EPA-HQ-OAR-2010-0682-0167.
100
-------
with the NHVcz parameter using an adjusted net heating value of 1,212 Btu/scf for hydrogen.
Under this requirement, we expect that most refinery owners or operators will incur monitoring
costs for a calorimeter (costs for which were included at proposal) and a hydrogen analyzer
(costs for which were not included at proposal) and have revised our cost projections for the final
rule accordingly. As noted in previous responses, the LFLCZ, and Ccz parameters were proposed
largely to account for instances where the net heating value did not correlate well with
combustion efficiency. With the hydrogen adjustment, the combustion zone net heating value
alone provided excellent correlation with combustion efficiency and the use of these other
parameters is no longer needed.
Finally, we note that we are requiring monitoring of the flare vent gas with calculation of the
combustion zone properties. This is consistent with recent consent decrees and it is thus far the
only means by which the proper performance of the flares can be assured.
Comment 2: One commenter suggested that the EPA should give further consideration on how
to apply the target combustion zone gas limits. To support this claim, the commenter referenced
conditions of certain combustion zone gas mixtures that may meet some of the target combustion
zone gas properties but not others and referenced the fact that certain combustion zone gas
mixtures may meet the proposed criteria but that there now exists a potential conflict with the
EPA flaring requirements in the General Provisions, specifically for NHV and streams with high
hydrogen content.
Response 2: We proposed that compliance could be demonstrated with any of the operating limit
formats (i.e., Btu, LFL, combustibles) at any time because we were aware that some operating
parameters worked better under certain conditions while others worked better under other
conditions. As provided in the preamble to the final rule, we have determined that using a net
heating value of 1,212 Btu/scf for hydrogen, the combustion zone net heating value appears to
work well for nearly all cases. We do not agree that the proposed requirements (many of which
we are not finalizing) were or the final requirements we are promulgating are in conflict with the
GP although they are more stringent than those in the GP. Importantly, refinery flares are not
required to comply with both sets of provisions; they need to comply with the new monitoring
requirements in 40 CFR 63.670 and 63.671 and are not subject to the requirements in the GP and
we have added specific overlap provisions to clarify this point.
3.6 Dilution parameters for air-assisted flares (general, non-H2-olefin)
Comment 1: A few commenters suggested that the EPA did not evaluate or provide costs,
burden evaluations, or demonstrations of cost effective control methodologies for the proposed
air-assisted flare combustion control requirements that would require facilities to convert their
air-assisted blower motors to continuous control systems and upgrade their electrical and digital
control logic systems.
Response 1: First, we disagree that we did not project costs for control systems for air-assisted
flares. See the memorandum entitled "Petroleum Refinery Sector Rule: Flare Impact Estimates"
included as Docket Item No. EPA-HQ-OAR-2010-0682-0209. Second, while some retrofits may
require new, adjustable blower motors, the adjustable blower motors, in addition to providing
101
-------
functionality needed to prevent over-assisting the flare, will reduce operating costs of the flare
during lower flow events. We did not attempt to quantify or take credit for the cost savings
associated with the air flow controllers. While individual applications may vary significantly
from the average cost that we included for air-assisted flares, we consider that the nationwide
estimate of the impacts for air-assisted flares is reasonable and we did not revise the cost
estimates specific to air control systems for air-assisted flares.
Comment 2: A few commenters suggested that the EPA did not provide enough data analyses to
demonstrate that air-assisted flares can be "over assisted." The commenters also stated that the
proposed dilution factor approach, which considers both flare tip diameter and perimeter assist
air flow rates to be critical parameters in flare combustion efficiency (CE) is not better correlated
with CE than the stoichiometric air ratio approach, and that the EPA did not demonstrate that
these sorts of occurrences are common or even occur in practice. Another commenter suggested
that compliance with the proposed dilution parameters using the current variable frequency drive
technology will require the minimum gas flow to the flare (and subsequently CO2 emissions) to
be increased as the air blowers must remain on at all times vent gas is being routed to the flare
for purposes of equipment integrity and that this problem could be exacerbated with the proposed
exit velocity and no visible emissions requirements.
One commenter suggested that the air-assisted flare monitoring parameters do not appear to have
been extensively tested in practice because blowers are often not capable of performing fine
adjustments in air assist rates and that the EPA should investigate alternatives to these proposed
requirements to ensure air-assisted flares in routine service will obtain the combustion efficiency
improvements being proposed in this rule.
Response 2: The performance data for air-assisted flares clearly demonstrate that air-assisted
flares can be over-assisted (see Figure 5 in the memorandum entitled "Petroleum Refinery Sector
Rule: Operating Limits for Flares" included as Docket Item No. EPA-HQ-OAR-2010-0682-
0206). The ability of the dilution factor to align the data from flare tips of different diameters
clearly suggests that the proposed parameters are better than the stoichiometric air ratio approach
(see Figures 6 and 7 in the memorandum entitled "Petroleum Refinery Sector Rule: Operating
Limits for Flares" included as Docket Item No. EPA-HQ-OAR-2010-0682-0206). The fact that
the industry practice is to use a single speed blower fan as suggested by some commenters
ensures poor combustion efficiencies at low flare vent gas flow rates. We do not understand how
using a variable speed fan would increase the amount of vent gas that needs to be sent to the flare
when the exact opposite is true. In fact, using a variable speed fan will lower electricity use by
the facility during low vent gas flows and it would be the use of a single air blower rate that
would require more supplemental gas flow to prevent over-dilution of the vent gas. We are
finalizing the net heating value dilution parameter (NHVdii) parameter requirements for air-
assisted flares with limited revision from those proposed.
Comment 3: One commenter suggested that the EPA significantly change the final rule with
respect to the proposed target dilution parameters for flares with perimeter assist air, particularly
for the target dilution parameter limits proposed for air-assisted flares that meet Hydrogen-Olefin
interaction criteria because there is no data currently available that support them. The commenter
also suggested that the EPA allow flexibility in the rule for facilities to demonstrate compliance
102
-------
by using a stoichiometric ratio of 10:1 or less or allow facilities to operate air-assisted flares right
at or above the incipient smoke point.
Response 3: We are not finalizing the hydrogen-olefin target parameters at this time. We are not
allowing use of a set stoichiometric ratio of 10:1 because the flare performance data we have
clearly suggests that a 10:1 stoichiometric ratio would not ensure combustion efficiencies
anywhere near 98 percent for some flares (see Figure 6 in the memorandum entitled "Petroleum
Refinery Sector Rule: Operating Limits for Flares" included as Docket Item No. EPA-HQ-OAR-
2010-0682-0206). We have previously described issues with operating at or above the incipient
smoke point in this response to comment document. We are finalizing the NHVdii parameter
requirements for air-assisted flares as proposed, except that we are not finalizing separate
operating limits that were proposed to address concerns of reduced combustion efficiency when
both hydrogen and olefins are present in the vent gas.
Comment 4: One commenter suggested that the EPA make some clarifications to the regulatory
language at 40 CFR 63.670(n). Specifically, the commenter pointed out that the EPA should use
an "or" rather than an "and" in the first paragraph since only one of the dilution parameters needs
to be calculated for compliance in paragraph (f) of that section. The commenter also points out
that while the units and equations used for NHVdii and lower flammability limit dilution
parameter (LFLdii) at 40 CFR 63.670(n)(l) and n(2) are different than the NHVCZ and LFLCZ
parameters presented at 40 CFR 630.670(m)(l) and m(2), that the definitions are the same and
need revision. Lastly, the commenter also points out that combustibles concentration dilution
parameter (Cdii) should be defined at 40 CFR 63.670(n)(3) and not Ccz.
Response 4: First, we are not finalizing the options to comply with the LFLCZ , Ccz, LFLdii, or Cdii
operating limits, so these revisions are no longer needed. We agree with the parameter definition
revisions suggested by the commenter to fix the editorial error in the equation term descriptions.
Therefore, we are revising the parameter description in 63.670(n)(l) to describe NHVdii as the
"Net heating value dilution parameter" rather than "Net heating value of the combustion zone
gas."
3.7 H2-olefin interaction
3.7.1 Appropriateness of interaction criteria concept
Comment 1: Several commenters suggested that neither scientific literature nor the available
flare test data support the EPA's claim of an adverse Hydrogen-Olefin Interaction on combustion
efficiency and that the EPA should not finalize the more restrictive combustion zone operating
limits for all flare types. To support their claims, commenters suggested that the EPA did not
provide any evidence that the assumed hydrogen/olefin effect actually exists, that statistical
analysis demonstrates that the EPA developed their limit based on random differences in data,
that the PFTIR data analysis method of using the individual minute by minute data instead of the
test average data is flawed and leads to invalid conclusions, and that proper analysis of the data
demonstrates that the more stringent operating limits for hydrogen/olefin conditions cannot be
supported.
103
-------
Several commenters suggested that there is some evidence to support more stringent flare
combustion zone limits for a narrowly defined high concentration propylene-only condition as
outlined in some of the recent flare consent decrees and that the flare test data do not support
more stringent operating limits for the proposed hydrogen/olefins criteria by the EPA.
Additionally, one of the commenters suggested that if the EPA decides to proceed with the more
restrictive target combustion zone limits for the hydrogen/olefins interaction cases that the final
rule should not expand beyond an interaction between hydrogen and propylene.
Response 1: We disagree that the hydrogen-olefin interaction effect had not been demonstrated.
The data presented in Figure 2 in the memorandum entitled "Petroleum Refinery Sector Rule:
Operating Limits for Flares" (included as Docket Item No. EPA-HQ-OAR-2010-0682-0206)
clearly indicated an early drop in combustion efficiency at relatively high combustion zone net
heating values relative to the other data points. When using the minute-by-minute data, there
were numerous data points meeting the hydrogen-olefin interaction criteria and the early
degradation of combustion efficiencies was quite pronounced. When we analyze the run average
data, we still see low combustion efficiencies above a combustion zone net heating value of 270
Btu/scf, and all of the lowest combustion efficiencies in this range are from runs that meet the
hydrogen-olefin interaction criteria that we proposed. The reduced combustion efficiencies were
clearly present at two different Marathon refineries; hydrogen olefin criteria were also present
for one Flint Hills Resources flare, but this flare generally operated above the target combustion
zone net heating value proposed for gas streams meeting the hydrogen-olefin interaction criteria.
Nonetheless, we acknowledge that the precise mechanism of the combustion efficiency
degradation is not well known (e.g., is it limited to propylene or applicable to all light olefins)
and we do not have additional data available to evaluate the ability of the proposed interaction
criteria to predict poor combustion efficiency for other refinery flares outside of the data from
which the criteria were developed. Therefore, at this time we have decided to not finalize the
hydrogen-olefin interaction criteria. Based on our Monte Carlo assessment (see the memorandum
entitled "Monte Carlo Analysis of Flare Performance Data" included in Docket ID No. EPA-HQ-
OAR-2010-0682), we determined that the NHVCZ limit of 270 Btu/scf evaluated on a 15-minute
average provided adequate assurance that all flares would meet an average destruction efficiency
of 98 percent over a 3-hour period (the time period for which other conventional control systems
must demonstrate compliance).
3.7.2 Selection of limits when interaction is present
Comment 1: One commenter suggested that the EPA misinterpreted Marathon Petroleum
Company's PFTIR flare testing resulting in errors in the proposed Olefins-Hydrogen operating
limits. The commenter suggested these errors were that the EPA based the limits on all olefins
rather than solely on propylene, that the limits are calculated in the combustion zone rather than
the vent gas, that the limits are unbounded on the upper end, and that the limits do not take the
relative amounts of hydrogen and olefins into account.
Response 1: We selected our hydrogen-olefin criteria so as to include the worst performing data
when combustion zone net heating values were at or over 300 Btu/scf. Since combustion occurs
in the combustion zone, we considered it most appropriate to base the criteria on the
104
-------
concentrations of these compounds in the combustion zone. Because we have a minimum
concentration for each compound and a minimum combined concentration, we conclude that we
did consider "relative amounts" of the hydrogen and olefins. At high concentrations of these
compounds in the combustion zone, the net heating value of the combustion zone gas will be
well above the proposed target (particularly now that we are providing 1,212 Btu/scf for the net
heating value of hydrogen), so there is no need to establish an upper concentration limit.
However, as noted previously, we are not finalizing separate standards for cases meeting the
proposed hydrogen-olefin interaction criteria.
Comment 2: One commenter suggested that TCEQ investigations that form the bulk of the
dataset on which the regulations are based show that 98% destruction efficiency is achieved by
focusing on operation at the incipient smoke point and that while the EPA implies these flare
regulations will cause operation on the lean side of the incipient smoke point, that that notion is
not proven and that it is possible the numerical regulatory limits proposed by the EPA will
mandate flare operation in the smoking region, especially for flares burning hydrogen and
olefins, thereby resulting in greater particulate emissions and visible emissions above existing
allowable limits.
Response 2: The TCEQ is one of seven flare studies evaluated, so it is not "the bulk of the
dataset." While visible emissions records were not always provided with the combustion
efficiency data, the available data indicate that refinery flares were able to operate with no visible
emissions at combustion zone net heating values well above 400 Btu/scf.
Comment 3: One commenter wanted clarification that in the special provisions for olefins and
hydrogen in the combustion zone that diolefins are included and that unsaturated hydrocarbons
such as aromatic compounds are not.
Response 3: As proposed, diolefins (e.g., 1,3-butadiene) were included in the olefins
concentrations and aromatic compounds were not included. As noted previously, however, we
are not finalizing these provisions.
Comment 4: One commenter suggested that while some of the data in the EPA flare dataset was
collected at flares serving ethylene production units, that it is not representative of ethylene
industry flaring or other olefin production flaring and that the proposed combustion zone
parameters are not appropriate for flares outside of the petroleum refining source category. To
further support this suggestion, the commenter notes that chemical industry flares combust a
wide variety of hydrocarbon compounds, some of which have not been included in the flare
dataset, and that flare tip sizes can range from 2" to 54" and larger in the chemical industry,
whereas the refinery flare testing likely focused only on larger flare tips.
Response 4: As noted previously, we have not definitively concluded whether the poor
combustion efficiencies identified in the performance test data are due to a general olefin-
hydrogen interaction or a propylene-hydrogen interaction. More data are needed to determine if
flares with high ethylene and hydrogen may have similar performance issues as the refinery
flares tested, regardless of their size. In any event, we did not propose, nor are we finalizing,
hydrogen-olefin requirements for the chemical industry as part of the Refinery Sector Rule.
105
-------
3.7.3 Applicability (steam-assist only or all flares)
Comment 1: A few commenters suggested that the EPA should remove the requirements for air-
assisted flares meeting the hydrogen/olefins interaction criteria to comply with the more stringent
combustion zone and dilution parameter operating limits until actual data are obtained for air-
assisted flares operating in this regime because these limits could be over-specified and result in
operation of a smoking flare.
Response 1: We are not finalizing separate operating requirements for hydrogen-olefin
interaction operating limits for steam-assisted, unassisted or air-assisted flares.
3.8 Flow rate monitoring requirements
Comment 1: One commenter suggested that the EPA should specifically allow all required
CPMS outages for flare gas flow monitors due to activities required to comply with NSPS Ja as
well as flow out-of-control periods due to required temperature and pressure correction
instrument outages.
Response 1: The commenters did not identify and we are unaware of activities required to
comply with subpart Ja that would cause a flare flow CPMS outage. We do not agree that we
should allow flow CPMS outages when temperature and pressure monitoring systems are not
functioning. We have revised the QA/QC requirements for pressure tap inspections and we are
unaware of other issues that would cause monitoring system outages. We note that flare flow
rates must be determined under "standard conditions" so the temperature and pressure
monitoring systems are considered an integral part of the flare flow monitor for this application.
Refinery owners or operators may employ redundant temperature and pressure monitors, if
desired, to limit outages of these inputs needed to correct measured flow rates to standard
conditions.
Comment 2: A few commenters suggested that if air-assisted flares are included in the final rule
that the EPA should specifically allow for use of air blower motor speed and design curve air
flow data as the basis for estimating assist-air flow rates as an alternative to having to install air-
assist flow rate monitoring systems.
Response 2: Generally, we agree that air blower motor speeds and design air flow data can be
used provided these have the required accuracy and the accuracy of the design curves are
verified like other flow CPMS.
Comment 3: One commenter suggested that utilization of flare vent gas composition data from
time periods before the 15-minute block periods with measured flow data as prescribed in the
rule at section 63.670(1)(2) appears misleading and problematic and that an approach to use a
continuous calorimeter in concert with a GC should be encouraged to avoid such problems.
Response 3: Based on the data we have reviewed, flow rates vary more significantly than
compositions. Given the short time period for the operating limit, we considered that it would be
more important to use a feed forward compliance approach to provide a means for owners or
106
-------
operators to be able to know the allowable steam to vent gas flow allowed within a given 15-
minute time period, particularly when using a GC. The GC allows speciation of chemicals which
may assist the refinery owner or operator in identifying and correcting discharges to the flare
from, for example, PRD that are not properly seated that might go undetected when using a
calorimeter. As such, we do not wish to discourage the use of GC, although we recognize the lag
time of the GC analysis presents challenges with the control logic. In the final rule, we allow
refinery owners or operators to comply with either the feed forward calculation that was
proposed or a direct calculation procedures using the flow and composition results available for a
given 15-minute block to calculate the vent gas and combustion zone gas properties during that
15-minute block period. While we expect that the direct method will be used primarily for flares
equipped with calorimeters, we allow the owner or operator to elect either the feed-forward or
the direct calculation methodology. The flare owner or operator must select which calculation
method they will use and must use that method at all times.
Comment 4: One commenter suggested that if flow monitors are employed in concert with other
monitors that will let refiners evaluate flare combustion zone gas characteristics, that the rule
should allow for manual operation and not dictate automated responses, particularly when the
relevant data is not synchronized based on delayed response times.
Response 4: The rule does not prohibit manual operation of the flare and refinery owners of
operators can elect how to best control steam additions to prevent smoking while maintaining
adequate net heating value in the combustion zone.
Comment 5: One commenter suggested that the proposed rules do not recognize infrequently
used flares like NSPS Ja and that provisions should be made to reduce or eliminate the
requirements for infrequently used flares.
Response 5: We disagree with the commenter. While we did not specifically define "emergency
only" flares as we did in NSPS Ja nor limit the alternative grab sampling approach specifically to
"emergency only" flares, these provisions were provided specifically to reduce the requirements
for infrequently used flares.
3.8.1 Flow monitors
Comment 1: Several commenters suggested that the EPA flow monitor accuracy specifications
are inconsistent with those in the South Coast Flare Rule and many Refinery Consent Decrees
and that as indicated in their NSPS Ja Reconsideration Petition, it is unclear what "measurement
sensitivity" means. Thus, in order to resolve the NSPS Ja reconsideration issue as well as to be
consistent with recent consent decrees, the commenters recommended revising both the flare
flow meter sensitivity specification and accuracy specification in Refinery MACT 1 Table 13
and in 40 CFR 60.106a(a)(6)(i)(B) of NSPS Ja to be consistent with the accuracy specification
from the Shell Deer Park Consent Decree, Appendix 1.10, which specifies the required flare flow
meter accuracy as "±20% of reading over the velocity range of 0.1-1 ft/s and ± 5% of reading
over the velocity range of 1-250 ft/s." One commenter similarly suggested that the accuracy
requirements in Table 13 for flows less than 0.1 ft/s be set within 20% to be consistent with
Petroleum Refinery consent decrees and the monitoring technology available today.
107
-------
Response 1: We agree that the term "measurement sensitivity" is not a term that is commonly
used in flow monitoring system's technical specification sheets. To be consistent with the
terminology used by instrument vendors and used in Refinery MACT 1 and 2, we are revising
Refinery NSPS Ja to replace the term "measurement sensitivity" with "accuracy." We agree that
flares can normally operate at very high turndown rates. We also recognize that flares can have
fairly large diameters. These two issues combined can cause some difficulties meeting the 10
cubic feet per minute (cfm) lower flow accuracy requirement. Therefore, we are revising the
flow rate accuracy provisions specific for flares to provide an accuracy requirement of ±20%
over the velocity range of 0.1-1 ft/s and ±5% for velocities exceeding 1 ft/s in 40 CFR
60.107a(f)(l)(ii) and in Table 13 of subpart CC. Note, we believe that the commenter cited an
incorrect paragraph in Refinery NSPS Ja. 40 CFR 60.106a(a)(6)(i)(B) pertains to gas flow
measurements related to the sulfur recovery plant. We do not consider it appropriate to revise
that accuracy requirements for these units. The accuracy requirement applies to normal flow
ranges and these units do not normally operate at high turn down rates. Therefore, if a gas stream
generally operates at low flow, the diameter of gas line, or at least the flow monitoring insert,
should be sized appropriately to accurately determine these low flow volumes. Thus, we are only
providing the dual range accuracy provisions for flares; all other flow monitoring systems must
meet the lower flow limit accuracy of 10 cfm.
Comment 2: One commenter suggested that the EPA remove the annual calibration and QA/QC
requirements for ultrasonic and optical flare gas flow monitors required in Refinery MACT 1
Table 13 and Refinery MACT 2 Table 41 because they can only be calibrated at the factory and
not in the field. In lieu of such requirements, the commenter suggested that the EPA allow these
flow monitors to use a monitoring plan that specifies following the manufacturer's
recommendations instead.
Response 2:, While we recognize that ultrasonic and optical flare gas flow monitors are
calibrated at the factory, there are techniques that manufacturers recommend to evaluate the
accuracy of the instruments in the field on an ongoing basis and these calibration evaluations
should be conducted periodically. To reduce the burden of these evaluations, we are changing
calibration evaluation requirements to be biennially (every two years) rather than annually (or
semi-annually as proposed in Table 41 of subpart UUU).
3.8.2 Engineering calculations (temp/press monitors)
Comment 1: Two commenters were supportive of the proposed language at 40 CFR 63.670(i)(4)
that would allow use of engineering calculations in lieu of flow monitoring in some cases,
especially where the molecular weight of the gas is known. One of those commenters raised the
question of whether both pressure and temperature monitoring are required for each use of
engineering calculations. This commenter also gave some specific examples where the final rule
should allow engineering calculations where either pressure or temperature information is
known, including but not limited to, calculations to determine the flow rate of steam though an
orifice plate or a control valve, calculations of vent gas flow rate through a process control valve,
and calculations of the air-assist flow rate using blower curves and actual operating conditions.
108
-------
Response 1: All of the flow rate measurements or engineering estimates will need to be
converted to "standard conditions" (68 degrees F and 1 atmosphere pressure). Depending on the
information available for the engineering calculation, knowledge of system temperature and
pressure may be needed.
3.9 Flare vent gas composition monitoring requirements
Comment 1: Several commenters suggested that 40 CFR 63.671 (e)(3)(ii) assumes that there are
heavy compounds present in flare gas which is an invalid assumption, that these species are
treated as C5+ in the calculations and that because of this, this paragraph should deleted. If it is
not deleted, commenters suggested that the EPA try and be consistent with state regulations like
the TCEQ HRVOC which requires analysis of C1-C4 compounds and uses a C5+ grouping or
that the EPA specify a concentration level at which the calibration range no longer needs to be
extended.
Response 1: We agree with the commenter that very few if any C5+ compounds exist routinely
in the flare vent gas. We are revising these calibration requirements to allow the use of n-pentane
as a surrogate for all C5+ organics in the flare vent gas.
Comment 2: Some commenters suggested that the EPA should not require continuous GCs for
all refinery flares as other more cost effective monitoring approaches are available that will
achieve the same reductions in emissions, that both identifying leaks into the flare system and
incremental steam savings and natural gas use savings from being able to fine tune flare
operations isn't as practical as the EPA claims. The commenters stated that requiring
sophisticated instrumentation like online GC on all flares has a disproportionate effect on small
refineries. Other commenters generally supported the use of GCs on all refinery flares,
suggesting it would increase operational awareness of real-time flaring events and give refiners a
better understanding of the true emission potential of the gas being flared.
Response 2: We are not requiring all flares to install a GC. At proposal, we considered that
many refineries would elect to use a GC in order to evaluate all compliance options. Because we
are finalizing only the NHVCZ operating limit (with the provision allowing use of 1,212 Btu/scf
for the net heating value of hydrogen for determining compliance), we consider it more likely
now that refinery owners or operators will elect to install a calorimeter along with a hydrogen
analyzer than we did at proposal. The advantage of this monitoring technique is quicker response
times, which helps to improve process control of the flare. While the full GC monitor would
provide additional data for both compliance and troubleshooting purposes, it is no longer needed
to allow a means to demonstrate compliance with any of the three proposed flare operating
parameters and the longer response time, which generally provides only one reading per 15
minute period, causes lag in process control of the flare. The final rule allows either the use of a
GC or a calorimeter and, if a calorimeter is used, provides the refinery owner or operator the
option to also install a hydrogen analyzer in order to take advantage of the provision to use 1,212
Btu/scf for the net heating value of hydrogen for determining compliance.
Comment 3: A few commenters suggested that the EPA add a definition for "Pipeline Natural
Gas" and that in lieu of using the default composition laid out in 40 CFR 63.670(j)(5) that the
109
-------
EPA allow periodic sampling results (e.g., weekly, every six months, or annually) of pipeline
natural gas to be used as an alternative.
Response 3: The term we use in the regulations is "pipeline quality natural gas." We do not
consider it necessary to add a definition for this term as it is a common term used to refer to
natural gas that meets certain specifications for interstate distribution. When this term is used in
the regulations, it generally refers to purchased natural gas (i.e., natural gas purchased from local
natural gas suppliers), and we are clarifying this intent in the final rule. We agree that periodic
sampling is a reasonable alternative to using the default composition provided in the proposal,
and we are including this allowance in the final rule. Since we are finalizing only the NHVCZ
operating limit, we are only providing a default net heating value for natural gas rather than
providing a default composition. Considering all of these revisions, we are finalizing 40 CFR
63.670(j)(5) as follows "Direct compositional or net heating value monitoring is not required
for purchased ("pipeline quality") natural gas streams. The composition of purchased In lieu
of monitoring the composition of a pipeline quality natural gas streams may be determined
using annual or more frequent grab sampling at any one representative location.
Alternatively, the net heating value of any stream, the following composition can bo used for
purchased pipeline quality natural gas stream can be assumed to be 920 Btu/scf."
Comment 4: One commenter suggested that the EPA should allow the use of any standard
reference text or Internet site to obtain the LFL for any component that is not listed on Table 12.
Response 4: We are not finalizing the option to comply with a LFLCZ operating parameter.
Therefore, it is not necessary for refiners to determine the LFL for purposes of the flare
requirement revisions to Refinery MACT 1.
Comment 5: One commenter suggested that the proposed calculation for Cvg does not reflect
combustibility because normalizing the calculation to propane penalizes CI and C2 compounds
and that the EPA should explain the basis for reporting results as volume percent as propane v.
volume percent as methane.
Response 5: At this time we are not finalizing requirements that include this calculation.
Comment 6: Several commenters suggested that the calibration requirements for on-line GC in
Table 13 should be modified to allow the approach used in the Texas HRVOC regulations,
specifically in 30 TAC 115.725(d)(2)(A)(i). Specifically, commenters suggested that in order to
compensate for the significant GC downtime associated with calibrations over the wide range of
compounds specified in the proposal, the EPA should permit variations from Appendix B, PS9
by allowing a multi-point calibration quarterly rather than monthly and by allowing a mid-level
calibration once per week instead of daily.
Response 6: We have already permitted variations from Performance Specification 9 by
allowing for calibration using surrogate compounds instead of every compound expected to be in
the stream. While we do not fully agree with the Texas HRVOC regulatory approach, we do
agree that if a mid-level calibration check is performed daily, it provides enough assurance of
110
-------
proper operation to allow for quarterly multi-point calibration checks in this application. We
have updated Table 13 accordingly.
Comment 7: One commenter wanted the EPA to clarify in what instances 40 CFR 63.670(d) and
(f) are not "applicable" from the following regulatory text at 40 CFR 63.670(j): "The owner or
operator shall determine the concentration of individual components in the flare vent gas using
either the methods provided in paragraphs (j)(l) or (j)(2) of this section, to assess compliance
with the operating limits in paragraph (e) of this section and, if applicable, paragraphs (d) and (f)
of this section." Also, the commenter pointed out that "continuously" is not defined in
paragraphs 40 CFR 63.670(j)(3) or (4) and that while it is defined in (j)(l) as occurring at least
once every 15 minutes it may not need to be defined for Btu analyzers or THC monitors as they
can probably measure continuously.
Response 7: The "as applicable" reference in Paragraph 63.670(j) is intended to refer to the fact
that it is not necessary to determine NHVvg for flare tip velocity if the owner or operator elects to
comply with the 60 ft/s velocity limit in (d)(1). Similarly, since paragraph 63.670(f) is only
applicable for air-assisted flares with perimeter assist air that provision would not be applicable
within the meaning of paragraph (j) for flares that do not have perimeter assist air. With respect
to continuous, 40 CFR 63.671(a)(1) states "All CPMS must complete a minimum of one cycle of
operation (sampling, analyzing and data recording) for each successive 15-minute period." We
do expect that most CPMS will cycle much more quickly. For example, we expect flow monitors
used for flares would provide output by the minute to afford control of steam flows to maintain
appropriate combustion zone concentrations. We specifically included this requirement in 40
CFR 63.670(j)(l) because the flare vent gas composition monitors (i.e., GC systems) were
expected to have the longest cycle times of any of CPMS.
3.9.1 Continuous monitoring systems
Comment 1: Commenters suggested that the proposed GC calibration requirements at 40 CFR
63.671(e)(2) are infeasible, unnecessary and that they must be revised. Specifically, commenters
raised issues with the requirements that all compounds in the flare gas stream be included in the
calibration gas because there could be hundreds of compounds, suggested that the speciation
requirements should focus on species that matter (i.e., hydrogen, olefins, hydrocarbons through
C3 present above 1%, 1,3-Butadiene, n-butane, and n-pentane), and that 1,2-Butadiene is
difficult to separate from 1,3-Butadiene and that there is negligible impact in ignoring it.
Commenters suggested that 40 CFR 63671 (e)(2)(ii) be revised to only require calibration using
normal hydrocarbons through n-pentane.
A few commenters suggested that the EPA should limit the calibration gas requirements in 40
CFR 63.671(e)(2)(i) to compounds that are present in higher concentrations. One commenter
suggested a concentration threshold of greater than or equal to 5 mole % on an annual average
basis while another commenter suggested a concentration of 10 mole % or greater.
One commenter suggested that in order to limit the composition requirements to what may be
accomplished with a single GC within the required 15 minute cycle time that the following
component list be adopted in 40 CFR 63.670(j)(l): 1. Hydrogen 2. Oxygen 3. Nitrogen 4. Carbon
111
-------
Dioxide 5. Carbon Monoxide 6. Methane 7. Ethane 8. Ethene (aka: Ethylene) 9. Acetylene 10.
Propane 11. Propene (aka: Propylene) 12. 2-Methylpropane (aka: iso-Butane) 13. Butane (aka: n-
Butane) 14. Butenes (including butene-1, isobutene, c & t-butene-2) 15. 1,3 butadiene 16.
Pentane plus (aka: C5 plus) (i.e., all HCs with five Cs or more). The commenter also suggested
that the EPA clarify that additional species such as hydrogen sulfide may be measured as long as
the cycle time limit (15 minutes) is not exceeded, that the Items 2-4 be allowed to be grouped
and reported as a single value, and that the EPA try and be consistent with the Texas HRVOC by
allowing Propadiene and 1,2-Butadiene to be unresolved since they are difficult to resolve from
methyl acetylene and 1,3-Butadiene and found only in very small quantities. Lastly, this
commenter suggested that Table 12 be revised as appropriate to correspond with whatever
changes are made.
Response 1: For most flares, flare gas composition will be variable, so it is difficult to know
when high concentrations of some compounds may occur, but we find that it is generally
necessary to include compounds at concentration of 1 percent or less to accurately characterize
the flare vent gas stream. We are revising 40 CFR 63.671(e)(2)(i) to specify the target analytes as
follows: hydrogen, methane, ethane, ethylene, propane, propylene, n-butane, iso-butane, butene
(general), 1,3-butadiene, and pentane (as a surrogate for C5+ hydrocarbons). We note that
speciation of specific butenes (cis-, iso- and trans-butene) is not necessary, but properties for
trans-butene would be used for non-speciated butane. We also note that speciation of 1,2-
butadiene is not required, but properties for 1,3-butadiene should be used for co-eluting
butadienes. Additional compounds, such as acetylene, carbon monoxide, propadiene, and
hydrogen sulfide can be included if the refinery owner or operator has sufficient quantities of
these pollutants such they want to specifically include them in the calculation analysis. We are
also revising the requirements in 40 CFR 63.671 (e)(2)(ii) to limit the quantification of hydrogen
and CI through C5 normal alkanes. After incorporating these revisions, we consider the
calibration requirements to be reasonable. We are not requiring calibration for oxygen, nitrogen,
and carbon dioxide because these components do not need to be included in the calculations
(NHV = 0). We are adding hydrogen sulfide properties to Table 12 in the event a refinery owner
or operator elects to monitor FhS (or has a separate monitoring system for FhS for Refinery
NSPS Ja).
Comment 2: One commenter suggested that the requirement in proposed Table 13 that
Performance Specification (PS) 9 of part 60 Appendix B applies to GC used to measure flare gas
net heating value needs modification. Specifically, the commenter suggested that the EPA waive
the temperature specification for flare gas GC samples in PS 9 since they typically have
temperatures much lower than 120 degrees C and that condensation of heavy hydrocarbons is not
an issue. The commenters cited the requirements in the Shell Deer Park consent decree of
maintenance of temperature of greater than or equal to 135 degrees F for the sample transport
line and greater than or equal to 125 degrees F for the sample conditioning system as reliable and
cost effective measures the EPA should require because electric heaters can be used.
Response 2: We agree that the sampling line for this source can be maintained at a lower
temperature and we note this exception when referencing PS 9 in Table 13.
112
-------
Comment 3: One commenter suggested that if the EPA does not delete the flare flow monitoring
QA/QC requirements in favor of using a monitoring plan that the EPA should defer back to the
flare flow monitoring requirements in 40 CFR 60 subpart Ja.
Response 3: We are revising Table 13 of Refinery MACT CC, Table 41 of Refinery MACT
UUU and Refinery NSPS Ja to have consistent specifications for gaseous flow monitoring
systems for flares.
Comment 4: A few commenters suggested that 40 CFR 63.671(a)(4) allow for at least a 5%
downtime limit for continuous monitoring data outside of maintenance periods, instrument
adjustments and calibration checks, similar to the requirements in Texas Sampling Rule protocol
found at 30 TAC 115.725(d)(3).
Response 4: As proposed, 40 CFR 63.671(a)(4) requires operation of the CPMS at all times
except during "maintenance periods, instrument adjustments or checks to maintain precision and
accuracy, calibration checks, and zero and span adjustments." We have more clearly written this
requirement in the final rule to require operation of the CPMS at all times when regulated
emissions are routed to the flare except for periods of "monitoring system malfunctions, repairs
associated with monitoring system malfunctions and required monitoring system quality
assurance and quality control activities." We believe that the original wording created some
confusion, as a calibration check and checks to maintain precision and accuracy can be
considered the same thing.
We note that the cited Texas rule states that the time required for normal calibration checks are
not considered downtime; this exclusion does not apply to monitoring system malfunctions or
repair periods. Therefore, 40 CFR 63.671(a)(4) allows more periods to be excluded from what is
considered downtime than the Texas rule does, and as such the additional 5% downtime limit
from the Texas rule is not warranted. We are not including any specific minimum data
availability requirement in 40 CFR 63.671(a)(4). We do not believe that numerical missing data
allowances provide incentive to conduct monitoring in a manner consistent with good air
pollution control practices. We are excepting periods of monitoring system malfunctions, repairs
associated with monitoring system malfunctions, and required monitoring system quality
assurance or quality control activities from the provisions to collect data continuously. Because
the data recorded during these periods are not generally considered valid data and are not
generally allowed to be used in calculations used to report emissions or operating levels, it is not
necessary for CPMS to provide data during these periods.
Also, we note that a monitoring system malfunction is a sudden, infrequent, not reasonably
preventable failure of the monitoring system to provide valid data. Monitoring system failures
that are caused in part by poor maintenance or careless operation are not malfunctions. Owners
and operators are expected to complete monitoring system repairs in response to monitoring
system malfunctions and to return the monitoring system to operation as expeditiously as
practicable.
Comment 5: A few commenters showed general support for proper flare monitoring and
operator training and suggested that a three year or longer delay before improving monitoring of
113
-------
flares is unacceptable and that facilities must be held accountable for proper operation of their
flares now with existing equipment, regardless of whether it is automated or manual.
Additionally, one of these commenters suggested that the EPA should consider amending its
proposal to require refiners to install emissions sensors on flare stacks to alert refinery personnel
in real time when flares malfunction and suggested these sensors could be used by refiners as a
way to maintain a log of those malfunctions.
Response 5: While we appreciate the general support for the proposal to improve monitoring of
flares, we disagree with the suggestion that these improved monitoring methods can be
implemented immediately upon promulgation of the final rule. Without the proper
instrumentation and control systems, refinery owner or operators cannot demonstrate
compliance. For example, we assumed all flares would need to upgrade their steam control
systems to provide both fine and coarse steam adjustment capabilities. Without such control
systems, the refinery owner or operator cannot easily adjust steam addition rates to comply with
the combustion zone gas operating limits. Based on the GHGRP data, three-quarters of all flares
do not currently have a compositional or heat content monitors, and these monitoring systems are
necessary in order for refinery owners or operators to be able to determine NHVCZ and assess
compliance with the flare operating limits. While we understand and support the desire to
implement these requirements as quickly as possible, we have determined, based on the flare
system modifications needed to comply with the operating limits as well as the monitoring
systems that must be installed to evaluate these operating limits, that 3 years will be needed.
3.9.2 Grab sampling systems
Comment 1: Several commenters supported the grab sampling option in the rule and suggested
that the EPA change the sample frequency from once every 8 hours to once daily or once
weekly; that the EPA should consider a provision to allow reduced grab sampling frequency for
infrequently used flares, flares with a small total annual flaring volume, or flares that consistently
exceed the minimum combustion zone requirements; and that the EPA should include Method
ASTM D-1946 as another option for the compositional analysis.
A few commenters suggested that the grab sampling approach should ideally be used for refinery
flares that only receive flare gas during emergencies, major refinery startups or shutdowns and/or
unusual flare gas recovery outages and requested that the EPA clarify where the grab samples
should be taken, whether grab samples are required for flares burning purge gas only (including
those with FGR), and if an event occurs between routine samples whether the EPA would
assume that the composition of gases flared during that event is the same as the composition of
the routine samples.
Several commenters generally did not support the EPA's proposed use of grab samples as a
compliance option for monitoring the flare vent gas. One of the commenters suggested that the
EPA must demonstrate that monitoring requirements are sufficient to assure compliance and that
the option to allow grab sampling once every eight hours in lieu of continuous monitoring does
not meet this standard.
114
-------
Response 1: Although not specifically limited to "emergency only" flares or flares with well-
designed flare gas recovery systems, these are the types of flares that we anticipated would use
this monitoring option. We consider this to be the "reduced monitoring frequency" option for
seldom used flares, and we are not changing the frequency of the grab sampling so that it can
occur less frequently. We provided this option because installing and operating a CPMS can be
costly and labor intensive and we determined this was not warranted for a flare that receives
regulated material for only a few hours or a few days a year.
For flares that receive regulated material on a consistent basis, this compliance option is not
practical. First, it is labor intensive to manually collect the 8 hour grab samples on a regular basis
and thus the cost of this approach can exceed the costs of a CPMS for routinely operated flares.
More importantly, there is significant risk of non-compliance under this option. This is because it
takes time to get the grab samples analyzed and thus there may be a significant period where the
flare operator will be adjusting steam rates based on experience and process knowledge without
knowing the actual vent gas composition. As a result, if the results of the grab sample analysis
indicates that the vent gas has lower heat content than expected, it is likely that the operating
limits would not be met and that there could be a significant number of 15-minute block periods
for which the operating limit was exceeded and the refiner could be subject to enforcement
action.
In response to request for clarifications of when the grab samples must be taken, we note that the
grab sampling is required whenever regulated material is sent to the flare. In general, a flare that
does not have a water seal and uses sweep gas for the entire flare header system would be
expected to monitor the flare vent gas continually because it would be extremely difficult if not
impossible to ensure no regulated material is entering the flare gas header (from leaking relief
valves or other issues), regardless of the sweep gas used (natural gas or refinery fuel gas). For
flares that normally operate with a water seal (i.e., emergency only flares or flares with well-
designed flare gas recovery systems), the purge gas is not subject to the grab sampling
requirements or the operating limits provided the purge gas is not regulated material. For
example, if natural gas is used as the purge gas, the flare is not subject to the requirement in 40
CFR 63.670 and 671 during time periods when only purge gas is provided to the flare (i.e., when
the water seal is intact). Generally, we expect the first grab sample collected for a given event
would be taken within the first 15-minutes of the start of the event.
3.10 Averaging time and calculation of flare operating limits
Comment 1: One commenter suggested that the claim that flares must meet a 98 percent
Destruction Efficiency (DE) at all times (i.e., every 15 minutes) is invalid and is a significant
change in the Refinery MACT 1 112(d)(2) and (d)(3) MACT Floor. The commenter claimed that
the EPA has not supported the extension of the historical "typical operation" expectation to an
every 15 minute minimum requirement. Further, the commenter asserted, the 98 percent DE
assumed at the time the MACT was promulgated was for a daily average and not for 98 percent
at all times. Finally, the commenter disagreed that these amendments should impose the MPV
floor (98 percent DE) for emission types with a 95 percent DE floor. In addition, the commenter
suggested that changing the averaging time established in 1995 for the MPV floor is a significant
change in the stringency of the MPV limitation established pursuant to sections 112(d)(2)and (3)
115
-------
and cannot be done without demonstrating that it is achieved in practice and meets all of the
CAA section 112(d)(6) criteria, including cost effectiveness.
Response 1: First, we disagree with the suggestion that the MACT floor control efficiency is
based on a daily averaging time. The 98 percent or 20 ppmv control requirements in the original
MACT were based on the results of performance tests, generally consisting of three 1-hour test
runs. Additionally, the initial compliance demonstration in Refinery MACT 1 is based on a
performance test consisting of three test runs of no less than 1 hour. Both of these facts indicate
that the MACT floor performance requirement is based on a 3-hour average performance. While
we require evaluation of the NHVCZ operating limit on a 15-minute block average, we are not
requiring and do not expect that all flares would achieve 98 percent control efficiency for each
15-minute block period. In our revised analysis, we specifically evaluated how well various
operating limit values and averaging times ensured that flares achieve 98 percent control
efficiency over a 3-hour period. We found that the NHVCZ operating limit needed to be
maintained at or above 270 Btu/scf for each 15-minute period in order to ensure flares achieve 98
percent control efficiency over a 3-hour average period due to the variability inherent in flare
operations and the drastic drop in flare control efficiencies that can occur as a result of this
variability (see memorandum entitled Flare Control Option Impacts for Final Refinery Sector
Rule in Docket ID No. EPA-HQ-OAR-2010-0682).
We note that to demonstrate on-going compliance with the MACT for MPV controlled using
process heaters, boilers, or thermal oxidizers, which generally have stable operations, we elected
to provide a 24-hour average firebox temperature based on the expected stability of these
systems in the original MACT standards. We considered whether it was necessary to revise the
requirements in Table 10 to record the 3-hour rolling average temperatures and report all 3-hour
rolling average periods when temperatures are outside the range established in the Notification of
Compliance Status Report (NOCS) or operating permit. While these amendments would be
consistent with the proposed amendments to revise the operating limit averaging times in
Refinery MACT 2, we did not propose to revise these MPV operating limits. Therefore, we
cannot finalize revisions to the averaging time of the MPV operating limits at this time.
However, we appreciate the commenter for pointing out the discrepancy between the
performance requirements and reporting requirements for MPV and we will consider revising the
MPV operating limits to 3-hour averages (consistent with the performance test) in the future.
However, this does not alter our determination of the MACT floor emissions limit is determined
by a performance test (i.e., a 3-hour average basis) and it does not alter the need for a 15-minute
averaging period for the NHVCZ operating limit for flares in order to ensure that flares achieve 98
percent control efficiency over a 3-hour average. Thus, we maintain that revisions to operating
limits used to ensure continuous compliance with the emissions limits do not constitute a change
in the underlying MACT standard, MACT floor or the performance demonstration requirements.
Comment 2: A few commenters suggested that the rule must clarify that it is not a deviation if
flare gas flow is too short to allow the control system to obtain a measurement and respond.
Response 2: We consider that the flare standards apply for any 15-minute block that the flare
receives regulated material. We recognize that there may be short periods of flaring that last only
a few minutes (such as when cycling-up a new compressor in a staged compressor system). We
116
-------
are clarifying that grab samples of these short events are not required. While a flare is on
standby, we expect that assist rates will be set at their minimum value, so that, when gases are
vented to the flare, there will be minimum steam or air supplied to the flare, so the flare will not
be over-assisted during these brief events. Due to the logistics of collecting a sample of the flare
vent gas during very brief events, we are clarifying that, for events that last less than 15-minutes,
calculation of certain operating limits (i.e., the flare tip velocity limit, NHVCZ and NHVdii) is not
required.
Comment 3: One commenter suggested that the rule needs to address how to handle time
periods when there is no flare gas flow and periods when there is no regulated material flow in
the calculations. The commenter recommended that no average be calculated when there is no
flow to a flare for the entire averaging period and those averages be recorded and reported as "no
flow" rather than zero and similarly, "no regulated material flow" be recorded where there is no
regulated material flow to the flare within an averaging period. Where there is some flow or
some regulated material flow to a flare during an averaging period, the commenter suggested that
the average should be calculated based on the cumulative flow and the combustion zone
properties associated with that flow (i.e., no flow periods should not be included in the average,
since any assumed combustion properties would be incorrect and would invalidate the calculated
average). For instance if flow to a particular flare only occurred for 30 minutes in an hour, the
three hour average would be the average value for those 30 minutes. Otherwise the remaining
150 minutes of the three hours would have to be taken as zero and would yield a false indication
of a failure to meet the minimum combustion zone properties.
Response 3: We agree with the commenter regarding the calculation procedures but not the
suggested averaging period. The operating limits only apply when the flare receives regulated
material. Within a 15-minute block, refinery owners or operators would only need to include the
cumulative volume of flare vent gas and, if applicable, volume of assist air or steam during those
periods when flow of regulated material is sent to the flare. It may be easier for some refinery
owners or operators not to distinguish within a 15-minute block average, specific minutes when
there is no flow of regulated material, so we are allowing the cumulative flow rates within a 15-
minute block to be determined by excluding flows during periods which regulated materials are
not sent to the flare. To clarify this, we are making several revisions to 63.670(k) and renaming
this section "Calculation methods for cumulative flow rates and determining compliance with
Vtip operating limits." In this section, we specify that, if desired, the cumulative flow rates for a
15-minute block period only needs to include flow during those periods when regulated material
is sent to the flare, but owners or operators may elect to calculate the cumulative flow rates
across the entire 15-minute block period for any 15-minute block period there is regulated
material flow to the flare.
Comment 4: One commenter suggested that a 15 minute frequency for taking flare samples is
appropriate given the time required to collect, inject and analyze a sample, especially when using
a GC. Additionally, the commenter suggested that there is no apparent significant benefit to more
frequent grab sampling unless the system is fully automated to take advantage of the extra data.
Response 4: Thank you for the comment. We are finalizing this requirement as proposed.
117
-------
3.11 QA/QC requirements for flare CPMS
Comment 1: One commenter suggested the EPA make the following revisions to 40 CFR
63.671: Paragraph (b)(3) References paragraph (c)(3)(vii) which does not exist, Paragraph
(b)(3)(v) - There is a reference to (b)(10), which is not included in section (b), and Paragraph
(b)(3)(vii) - There is no reference to a parametric signal analyzer.
Response 1: Thank you for the comment. Our responses to the specific comments are provided
below.
40 CFR 63.671(b)(3): This paragraph should reference ".. .the information specified in (b)(3)(i)
through (vii)" rather than referencing paragraphs (c)(3)(i) through (viii). We note that a similar
error was made in 63.671(b)(5), where the reference should be "... procedures listed in
paragraphs (b)(5)(i) through (vi)" rather than (c)(5)(i) through (vi).
40 CFR 63.671(b)(3)(v): The provisions anticipated in paragraph (b)(10) were not proposed, we
are deleting the phrase ".. .and meet the requirements in paragraph (b)(10) of this section."
40 CFR 63.671(b)(3)(vii): The parametric signal analyzer is that part of the CPMS that converts
the measured signal into the actual output of the CPMS when the parameter measured is different
from the parameter monitored. This is common for flow monitoring systems that may measure
pressure (from an orifice plate) or velocity and need to convert these to volumetric flow rates
(sometimes corrected to standard conditions). Sometimes these calculations require additional
inputs to perform the calculations (e.g., density or molecular weight of the gas; temperature and
pressure of the system). This section requires identification of these parameters and the
algorithms used to convert the parameters actually measured by the monitor to the final output
used in the operating limit. We are finalizing this requirement as proposed.
Comment 2: A few commenters suggested the EPA consider the following for the revisions to
the flare instrumentation portions of Refinery MACT 1 and 2: (1) Per NSPS Ja flare
minimization requirements, excess flaring should not be required in order to perform marginal or
unnecessary instrument testing, (2) The special situations associated with flare instruments must
be addressed (e.g., inaccessible pilot monitors, sonic and optical flow meters, optical
composition monitors), (3) Flare instrumentation requirements must be consistent with the
requirements in NSPS Ja, State Rules (e.g., Texas HRVOC monitoring, SCAQMD and
BAAQMD monitoring), and applicable refining consent decrees, so that wasteful duplication or
replacement of existing instrumentation due to minor differences is avoided, (4) Flare instrument
requirements must be technically feasible using, as much as reasonable, industry standard
equipment, (5) Flare instrument requirements should impose no more instrumentation outage
than is absolutely necessary and periods of instrument outage and maintenance must be excused
from compliance demonstration requirements.
Response 2: We appreciate the comment, and we have revised several of the monitoring
requirements to improve the consistency between Refinery MACT rules (subpart CC, UUU) and
Refinery NSPS Ja. With respect to item 1, we are unaware of any requirements in the proposed
of final rules that would require excess flaring.
118
-------
With respect to item 2, we expect the inaccessible pilots to reference the inspection requirements.
As noted in Section 3.2.2, we are not requiring flare pilot monitoring systems to be subject to the
performance and quality control/quality assurance requirements in Table 13. For other
monitoring systems, we recommend flare owners or operators employ redundant monitoring
systems (thereby eliminating the need for visible inspections) for inaccessible monitors. We have
revised the accuracy requirements for flare flow meters. We are unaware of any "optical
composition monitoring" systems.
With respect to item 3, we have revised accuracy requirements for flare flow meters to improve
the consistency between Refinery MACT rules (subpart CC, UUU) and Refinery NSPS Ja. We
also revised the some of the requirements for Performance Specification 9 to address some
requirements in State rules and consent decree requirements; however, we are not required to
incorporate provisions in these rules that we do not consider adequate to ensure continual
compliance with the MACT standards.
With respect to item 4, we are unclear as to what provisions the commenter is referring to. We
have revised the calibration requirements (using n-pentane for C5+ rather that require
calibrations up to C7 hydrocarbons). We maintain that the requirements included in the final rule
are technically feasible using available industrial equipment.
With respect to item 5, we have minimized, to the extent practical, the calibration requirements
for GC to minimize instrument outages. The proposed rule specifically allows for flare operation
and compliance during these periods. We have added a specification consisted with the Texas
HRVOC rule that these periods will not exceed 5 percent of operations.
Comment 3: One commenter suggested that a safety over-ride must be provided for 40 CFR
63.671 and Table 13 requirements where flare access is required because requiring instrument
QA/QC for instruments that are not safe to access puts workers at risk. Also commenters
suggested that temperature CPMS are unworkable and unnecessary for flare pilot monitors
because the purpose of monitoring is to indicate the presence of a flame and an accurate reading
of the flame temperature is not required to just show the presence of a pilot flame.
Response 3: As discussed elsewhere in this chapter, we are removing the proposed requirement
of making flare pilot monitoring systems CPMS in the final rule. For other monitoring systems,
flare owners or operators can use redundant monitoring systems if the monitors are unsafe to
inspect.
Comment 4: One commenter suggested that all required CPMS outages be specifically allowed.
The commenter specifically suggested that while 40 CFR 63.671(a)(4) lists activities during
which a CPMS may be off-line, that not all activities required by 40 CFR 63.671 and Table 13
are covered. For instance, outages associated with the required Table 13 check of pressure CPMS
taps for plugging and the 63.671 (b)(5)(iii) daily response checks would not be allowed by
proposed 63.671(a)(4). Therefore the commenter recommended this paragraph be generalized by
adding a phrase to cover any outage required to comply with this rule or the procedures specified
in the CPMS monitoring plan (e.g., manufacturer's recommendations or activities required by
applicable Performance Standards).
119
-------
Response 4: We reviewed these sections and conclude that 63.671(a)(4) specifically allows for
all "required CPMS outages." As proposed, checks for plugged pressure taps would be
considered "checks to maintain precision or accuracy." Obviously, a plugged pressure tap would
not be accurate. As noted elsewhere in this document, we have revised the requirements for
identifying plugged pressure taps to be daily review of pressure reading to ensure the CPMS is
responding/fluctuating as normal and these checks would not require a CPMS outage. As
discussed previously in this document, we also revised 63.671(a)(4) to place a maximum time
the monitor can be offline due to "maintenance periods" and other instrument checks or
adjustments.
Comment 5: One commenter suggested revisions be made to the proposed CPMS monitoring
plan requirements at 40 CFR 63.671(b). Specifically, the commenter suggested that this section
require the CPMS monitoring plan to list the manufacturer and model number for all monitoring
equipment components and that the plan should be limited to the monitoring instrument itself by
deleting the word "all", that the requirement of the monitoring plan to identify the parameter
detected by the parametric signal analyzer and the algorithms used to convert these values into
the operating parameter monitored to demonstrate compliance be deleted or only required where
it is available and not confidential business information of a third party, and that the EPA remove
the requirement to include a spare parts inventory as this requirement is ambiguous, wasteful,
burdensome, and unnecessary.
Response 5: We note that in paragraph 40 CFR 63.671(a) we clearly delineate that the section
applies specifically to CPMS installed to comply with §63.670. Additionally, we then reference
the CPMS monitoring plan from 40 CFR 63.671(a)(5), which also supports that this plan only
needs to cover CPMS installed to comply with §63.670. For additional clarity, in the final rule,
we repeat this applicability phrase as follows:
In 63.671(b): We are modifying the first sentence to clarify that".. .a CPMS monitoring
plan that covers each flare subject to the provisions in §63.670 and each CPMS
installed to comply with applicable provisions in §63.670."
In 63.671 (b)(3)(i): We are revising to read: "Manufacturer and model number for all
monitoring equipment components installed to comply with applicable provisions in
§63.670"
In 63.671(c): We are modifying the first sentence to clarify that "For each CPMS
installed to comply with applicable provisions in §63.670, ..."
In 63.671(d): We are modifying the first sentence to clarify that "The owner or operator
shall reduce data from a CPMS installed to comply with applicable provisions in
§63.670 as specified in paragraphs (d)(1) through (3) of this section."
We believe that it is imperative that a facility understand how its monitoring equipment work.
When an algorithm is needed to convert the signal measured by the equipment to the monitored
output (e.g., in the case of a flowmeter measure pressure differential), we believe that it is
reasonable to expect the facility to be aware of how this calculation is performed. We also do not
believe that it is burdensome to provide a spare parts inventory. Our intent is not to require the
facility to maintain a list of what spare parts are available at any given moment in the CPMS
120
-------
plan. The intent of this requirement is to provide a list of what spare parts are planned to be kept
on site in order to quickly remedy any issues with the CPMS.
Comment 6: One commenter had specific comments and suggestions related to Table 13 in
Refinery MACT 1 including: (1) Proposed Table 13 does not define how temperature
performance evaluation is to be conducted, (2) Proposed Table 13 requires "visual inspections
and checks of CPMS operation every 3 months, unless a CPMS has a redundant temperature
sensor." How are such inspections to be conducted and recorded? (3) How is flow evaluation to
be conducted under proposed Table 13? (4) Proposed Table 13 requires daily checks for
obstructions, such as pressure tap pluggage. This requirement is excessive and unnecessary and
should be removed, (5) Proposed Table 13 requires pressure performance evaluations, but does
not specify how such evaluations are to be conducted.
Response 6: In general, we are not specifying how performance evaluations are to be performed,
just the timeframes and types of inspections. Due to the many variations in monitoring
equipment, we believe that the manufacturer of the equipment is the best source for determining
the proper technique for performing most performance evaluations. The CPMS monitoring plan
must include information on routine quality assurance and quality control procedures. The plan
should include not only a schedule for performing the performance evaluations, but also a
description on how the evaluations will be performed. Unless specifically specified, we are
providing the facility with discretion to determine the best method to perform these evaluations
for these site-specific monitoring systems.
As noted elsewhere in this document, we have revised the requirements for identifying plugged
pressure taps to be daily review of pressure reading to ensure the CPMS is responding
(fluctuating) as normal.
Comment 7: One commenter suggested that most CPMS do not measure concentrations, so the
requirement for span of the analyzers at 40 CFR 63.671(b)(3)(v) must be limited to those
analyzers that do or generalized to cover other parameters and that the reference at (b)(10) needs
to be corrected. If it is generalized, the commenter suggested that it is made clear that the Table
13 accuracy requirements only need to be met for the normal range (as specified in Table 13) and
not all expected values (as specified here). The commenter points out that generally, the Table 13
accuracy specifications cannot be met at the extremes of values that instruments can measure,
which is particularly true for flare flow meters, which cannot meet + or - 5% accuracy below
about 1 foot per second flow, but can measure with less accuracy down to about 0.1 ft./sec.
The commenter also suggested that 40 CFR 63.671(b)(3)(vi) should be clarified to specify just
which analyzers it includes or it should be generalized by substituting "CPMS" for "analyzer."
Response 7: The "span of the analyzer" is applicable to other types of CPMS, not just those that
measure concentration. We also note that a CPMS typically includes multiple components, such
as the sensor (measuring device) and analyzer (that converts the sensor signal to a final output).
For example, a thermocouple temperature sensor provides a voltage output. This voltage output
is then converted to a temperature output reading by a meter (analyzer). We are revising 40 CFR
63. 671(b)(3)(v) to clarify that the span requirement applies to both the sensor and the analyzer
121
-------
as follows: "Span of the CPMS. The span of the CPMS sensor and analyzer must encompass
the full range of expected values." We are also revising 40 CFR 63. 671(b)(3)(vi) to refer to the
"span of the CPMS" rather than the "analyzer's span."
Comment 8: One commenter suggested that the paragraph at 40 CFR 63.671(b)(5)(ii) be revised
to allow having an alarm as an alternate to an operator check.
Response 8: We believe the commenter is referring to 40 CFR 63.671 (b)(5)(iii) rather than (ii).
We consider the commenter's suggestion to be reasonable. We are revising the second sentence
of 40 CFR 63.671 (b)(5)(iii) to reads as follows: "If the CPMS system includes an internal system
check, the owner or operator may use the results to verify the system is responding, as long as
the system provides an alarm to the owner or operator or the owner or operator checks the
internal system results daily for proper operation and the results are recorded."
Comment 9: One commenter requested that the EPA include specific language in the rule
providing that lack of compliance data for entire averaging periods is not a deviation from the
combustion or velocity emission limitations. To support this suggestion, commenters pointed out
that QA/QC and maintenance activities for the monitors take more than 15 minutes to perform
and that this will result in several averaging periods every day for which no data is available,
especially since daily calibrations of the GC, heat content and/or hydrocarbon analyzers are
required and daily checks of pressure taps for pluggage are required.
Response 9: We consider that the current language is sufficiently clear on how to manage the
availability of compositional data. For example, 63.670(l)(2)(ii) states that "For all other 15-
minute block periods, use the results that are available from the most recent sample prior to the
15-minute block period for that 15-minute block period." We clarify in this response that this
language does not require that the "most recent sample prior to the 15-minute block period" must
be the results for the 15-minute block immediately preceding the 15-minute block for which the
calculations are being made. We recognize that a 15-minute block compositional analysis will
typically be missed during the daily calibration check and we allow for this "miss." If the most
recent analysis available prior to the 12:45 a.m. to 1:00 a.m. became available at 12:20 am
(because calibration check was performed after the 12:20 a.m. result analysis so no flare vent gas
composition result is available prior to 12:45 a.m.), then the rule allows (and even requires) the
use of the 12:20 a.m. result for both the 12:30 a.m. to 12:45 a.m. 15-minute block period and the
12:45 a.m. to 1:00 a.m. 15-minute block period. Given the cycle times of composition monitors,
we consider this allowance necessary, particularly for GC compositional analyses. Calorimeters
and hydrogen analyzers cycle times are much shorter, so it is less likely that data will be
"missed" in any given 15-minute block period, but we have provided this allowance for each of
these monitoring systems. We have not revised these requirements as a result of this comment.
As noted previously in this document, we have revised the pressure pluggage check
requirements, which eliminates the need for a CPMS outage for this check.
Comment 10: One commenter generally did not support parameter monitoring systems because
they suggested that they do not tell you what air pollutants in what concentrations are emitted
and suggested that they cannot tell you whether an actual air pollution violation has occurred.
122
-------
Response 10: We appreciate the comment and understand the concern; however, continuous
speciated emissions monitoring systems for flare gas exhaust emissions are not commercially
available at this time. We consider the flare combustion efficiency requirements we are finalizing
to be the best approach available to ensure refinery flares are operated with high destruction
efficiencies.
3.12 Alternative means of emissions limitations (site-specific)
Comment 1: A few commenters supported the EPA's inclusion of the provisions for obtaining
an Alternative Means of Emission Limitation (AMEL) in the rule and suggested that facilities
seeking an AMEL should be able to rely on the results of a previously completed emissions tests
if the flare and vent gas characteristics are similar to the flare seeking the AMEL request. The
commenters also suggested that additional testing should not be required for every flare seeking
an AMEL.
One commenter suggested that the rule should also allow for flare manufacturers to request
approval for equipment specific operating limits.
Response 1: In general, we disagree that where a source performs tests for an AMEL for one
flare that it can rely on that test for an AMEL for other flares. For the special case that identical
flares are used in series (as in a cascaded flare system), then it seems reasonable that the AMEL
developed for the primary flare in the cascaded system be relied on for an AMEL for identical
secondary flares. But in most other situations, however, there is too much variation in the vent
gas sources and compositions, flare diameters, flare tip design, etc., to be able to rely on the same
test for an AMEL for different flares. In the final rule, we require in 40 CFR 63.670(r)(l) that an
owner or operator must submit a flare test plan for approval prior to the source test. As such, we
do not consider it appropriate to allow facilities to rely on an already conducted a flare
performance evaluation in seeking an AMEL without a review and evaluation of the tests
conducted. A flare owner or operator may submit the test plan for a previously conducted source
test for approval, but if the test plan is deficient, a new performance evaluation would be
required. If the test plan is approved, the owner or operator can continue the process specified in
40 CFR 63.670(r)(2) and (3) to request an AMEL. A complete test report must be submitted
containing all of the pertinent data, including test methods used, calibrations, other quality
control results, raw field data sheets, and laboratory data, if applicable. In any request for an
AMEL, the EPA reserves the right to reject the request based on a determination that the test was
insufficient to adequately characterize flare combustion efficiency; that the results have not been
adequately documented (e.g., instrument not calibrated correctly, evaluation did not test
for/quantify enough organic compounds likely to be in the flare plume, or no raw data); or that
the reduction in emissions is not at least equivalent to the flare combustion efficiency
performance requirements in the final rule.
Comment 2: One commenter suggested that the rule should be flexible enough to permit direct
combustion efficiency measurement as a potential future compliance option since this
technology is rapidly evolving and may one day prove to be a viable cost-effective alternative to
the monitoring being proposed in this rulemaking.
123
-------
Response 2: Although this technology may not be sufficiently advanced to be able to provide the
requested provisions at this time, we can reconsider the need to revise these standards if a "flare
CEMS" technology becomes available. If this technology becomes available prior to the next
technology review, sources have the option to submit an alternative monitoring request for these
systems.
Comment 3: One commenter recommended that the rule allow refineries the option of preparing
alternative compliance approaches. The commenter states that flexibility could provide a viable
method for resolving inconsistencies among flare combustion efficiency requirements in recent
consent decrees, the NSPS Ja rule requirements, and the requirements of this proposed rule. It
would also be an appropriate mechanism for refineries that have established their own
combustion efficiency protocols separate and apart from any requirement to do so.
One commenter suggested that the majority of the flares that currently are subject to consent
decrees with flare specific limits comply with a dynamic operating limit which requires a gas
chromatograph to obtain speciated vent gas compositions. They go on to state that these flares
also have the ability to dynamically control assist steam and supplemental gas flow rates. The
commenter suggests that the EPA allow sources that want to establish and use a flare specific
operating limit be allowed to do so and be able to comply with a dynamic operating limit without
having to go through the proposed procedures at 40 CFR 63.670(r).
The commenters noted that the EPA's existing flare combustion efficiency consent decrees state
that as long as an entity subject to a consent decree is meeting its requirements, it is deemed to be
meeting the required combustion efficiency. The commenter believes that refineries that invest in
compliance with applicable consent decrees should not lose the benefit of these expenses by
being required to comply with the EPA's proposed rules.
Additionally, the commenter urged the EPA to allow an AMP that would allow monitoring and
control similar to those included in an existing flare combustion efficiency consent decree,
though the refinery may not itself be subject to a consent decree. Examples where such flexibility
would be appropriate include a situation where a refinery purchases equipment to comply with a
consent decree that is under negotiation, but the refinery is never subject to the consent decree
(sold before the decree is finalized; or where a refinery invests in equipment based on consent
decree negotiations before being subject to a consent decree itself). The commenter believes
that refineries in these cases should have an opportunity to utilize the monitoring and emission
control equipment purchased in good faith under these scenarios.
Response 3: We have provided mechanisms to develop site-specific operating limits through the
AMEL provisions in 40 CFR 63.670(r). We do not consider it appropriate to bypass these
provisions for flares with consent decrees. As noted in the previous response in this section, we
will allow owners or operators to submit test plans for a previously conducted performance
evaluation for approval, but we maintain the provisions in 40 CFR 63.670(r) are necessary to
ensure that the alternative limits are at least equivalent to performance achieved via the operating
limits included in the final rule.
124
-------
3.12.1 Site-specific test plan
Comment 1: One commenter suggested that the EPA make clear in the final rule preamble that
there is no initial or general flare performance testing requirements in the rule and that
monitoring for and determining the combustion zone operational limits satisfies these
requirements.
Response 1: There are several operating limits that apply to flares: a visible emissions limit,
flare pilot requirements, flare tip velocity limits and combustion zone net heating value limits.
We note that there is a requirement to conduct an initial 2-hour visual emissions observation, so
there is an initial performance test for the visible emissions limit, but there are no other
performance testing requirements for the other flare operating limits (unless the owner or
operator elects to request an AMEL) to demonstrate that the flare is achieving 98 percent control
efficiency. As suggested by the commenter, compliance with the MACT flare requirements
(other than the visible emissions limits) is determined based solely on the operating and
monitoring requirements in the final rule.
Comment 2: One commenter suggested that the EPA not refer to the testing required for
development of an Alternative Emission Limitation as proposed in 40 CFR 63.670(r) as a
"performance test" but rather call it a "flare specific test" because there are many requirements
imposed for performance tests that are not for the test required for an AMEL, such as electronic
reporting tool (ERT) reporting and testing at maximum representative conditions and that all of
the requirements needed to assure a good flare performance test are already included in 40 CFR
63.670(r).
Response 2: Given that 40 CFR 63.670(r) specifies the conditions of the tests, which override
GP requirements with respect to the maximum representative conditions and that ERT reporting
is not applicable to these test methods, we see little substance in the commenter's suggestion.
However, we do note that the "performance test" is really a "performance evaluation test" in that
it specifically looks to identify conditions at which performance of the flare begins to deteriorate.
Therefore, we have revised 40 CFR 63.670(r) to describe the test as a "performance evaluation"
rather than just a "performance test" and we note here that we consider the "performance
evaluation" to be different and separate from a conventional "performance test."
Comment 3: One commenter raised questions about the AMEL process proposed at 40 CFR
63.670(r). Specifically, the commenter wanted to know (1) the time-frame for the EPA approval
when a deviation request and test procedure are submitted to prove that different operating
parameters yield good destruction efficiency; (2) the time for the EPA to review the procedure
submittals and re-submittals; (3) whether, if prior testing has been conducted which adequately
shows good combustion, that data be re-submitted; and (4) what qualifies as a "sufficient
number" of test runs to identify the point at which the flare efficiency deteriorates and what
qualifies as deterioration?
Response 3: Generally, we try to review and respond to these types of AMEL requests within 60
days. However, if deficiencies are noted, the approval time can be significantly longer. Refinery
owners or operators should begin their evaluation early in the process and determine if they wish
125
-------
to seek an AMEL 12 to 18 months before the applicability date of the new flare provisions in
order to have time to get approval for and conduct their site-specific flare performance
evaluations and then submit and get approval for site-specific operating limits. We are allowing
refinery owners or operators to submit test plans for a previously conducted test and continue
through the process to request an AMEL if that test plan is approved. The number of test runs
will be outlined in the site-specific evaluation plan. The phrase "a sufficient number of test runs"
is included in the event the performance of the flare is different from expected in order to require
additional test runs to identify the point at which the flare efficiency deteriorates. We consider
test runs with combustion efficiency less than 90 percent to satisfy this requirement. If the
combustion efficiency for all test runs under given fuel type are less than 96.5 percent, we also
interpret the testing provision to require additional runs to identify the point where 96.5 percent
combustion efficiency is achieved. A "sufficient number" would be two or three runs in a given
combustion efficiency range.
3.12.2 Request/approval of site-specific operating limits
Comment 1: One commenter suggested that the EPA include an alternative to site-specific
limits with a five year time frame in which facilities be allowed to gather site specific data upon
which compliance limits would be based. The commenter suggests that this will encourage
proper flare operation and allow the regulated community to deal with any site or equipment
specific issues prior to incurring any potential violations.
Response 1: We are providing 3 years to comply with the new flare provisions in the final rule
consistent with the maximum time allowed to implement standards under CAA section 112(d)(2)
and (d)(3). We encourage refinery owners and operators to act swiftly in the event they need to
install new equipment or if they are considering submitting a request for an AMEL. We consider
5 years is more time than is needed and, moreover, it is more time than allowed under the CAA.
3.13 Definitions for flare control devices
Comment 1: One commenter suggested that the EPA clarify what constitutes a flare "event" for
purposes of 40 CFR 63.670(l)(2)(i) and (3)(i) as the duration of an "event" could be so short that
no vent gas sample can be collected.
Response 1: For the purposes of these requirements, an event is any time regulated material is
sent to the flare. We have added provisions to the final rule such that certain operating limits are
only applicable for an "event" that is at least 15 minutes in duration.
3.13.1 Air assist terms
Comment 1: A few commenters suggested that the EPA make revisions to the definition of
"Assist air" or propose a definition for "air-assisted flare" to distinguish these flares from steam-
assisted flares and make clear that the assist air addition is for the purpose of aiding combustion
and that incidental air addition due to entrained air in the steam, from process sources, or leakage
of instrument air does not qualify a flare necessarily as using "assist air." One commenter
suggested that the EPA make the following change to the second sentence in the definition of
126
-------
Assist Air: Assist Air includes premix assist air, whether injected or induced, and perimeter
assist air.
Response 1: The definition of "assist air" already includes the notion that assist air is . .air that
intentionally is introduced..so we disagree that further clarification is needed. The second
clarification appears to be more appropriate for the definition of premix air and we are clarifying
the definition of premix assist air as follows: "Premix assist air means the portion of assist air
that is introduced to the flare vent gas, whether injected or induced, prior to the flare tip.
Premix assist air also includes any air intentionally entrained in center steam."
Comment 2: Several commenters suggested that the EPA make revisions or clarifications to the
definition of "Perimeter assist air" because most air-assisted flares send air assist along the axis
of the flare and not necessarily to the perimeter and, they believe, the definition was intended
cover air conveyed through the upper or lower steam lines but not to cover the typical 3-10
pounds of unmeasurable ambient air entrainment per pound steam from upper and lower steam.
Additionally, one commenter suggested the following definition change for "Perimeter Assist
Air": means the portion of assist air introduced at the perimeter of the flare tip or above the flare
tip. Porimotor assist air includes air intentionally entrained in the lower and upper stream.
Perimeter assist air includes all assist air except premix assist air. Perimeter assist air does not
include the surrounding ambient air.
Response 2: We think it is important to consider all air that is intentionally introduced in or
around the combustion zone. As we understand it, lower and upper steam is not mixed with the
flare gas until the flare tip, so we consider this to be similar to "traditional" perimeter assist air
and we are retaining that portion of the definition. Therefore we are finalizing the definition of
perimeter assist air as proposed.
Comment 3: Several commenters suggested that the EPA make revisions or clarifications to the
definition of "Premix assist air" as they stated that assist air is typically supplied at the flare tip
and not prior to the flare tip to avoid creating a potentially explosive mixture inside of the flare,
that it is not clear whether "prior to the flare tip" means prior to the flare tip inlet or flare tip exit,
and that the definition needs to differentiate between air conveyed to the flare tip and ambient air
entrainment.
A few commenters suggested specific regulatory definition language changes. One commenter
suggested we delete the second sentence altogether and replace it with the following: "Premix
assist air does not include the surrounding ambient air." Another commenter suggested we add
the following to the definition of Premix assist air: means the portion of assist air that is
introduced to the flare vent gas prior to the flare tip exit. Premix assist air also includes any air
intentionally entrained in center steam or assist gas injected into the flare vent gas prior to the
flare tip exit.
Response 3: We are revising the definition of premix assist air as noted in Response 1 of this
section. We consider it important to emphasize that assist air is air intentionally added to the
flare vent gas or tip. We see no need to clarify that premix assist air does not include surrounding
ambient air since that does not seem pertinent to premix assist air (air added prior to that flare
127
-------
tip) and we have already made that clarification in the definition of assist air (the general term,
which applies to both perimeter and premix assist air).
3.13.2 Steam assist terms
Comment 1: One commenter suggested the EPA revise the definition of "Assist Steam" to
exclude "Center Steam" because no air is included with this steam, it is a diluent and used to
reduce burn back, it increases exit velocity, and it creates confusion for users and is in error. The
commenter provided the following regulatory definition language change to the second sentence
of the definition of Assist Steam: Assist steam includes, but is not necessarily limited to, center
steam, lower or tube steam and upper steam.
Response 1: We do not follow the logic that we should not include center steam in the definition
of assist steam because no air is included in this stream. It is still steam added prior to the
combustion zone and must be include in the assist steam flow determination. We understand that
it is unusual to use "premix assist air" but flare vendors discussed this method of intentionally
inducing air in with assist steam and the "premix assist air" definition was developed to address
cases where are is introduced into the vent gas prior to the flare tip, whether directly or entrained
in center steam. This is the same reason we clarified that air intentionally entrained in lower or
upper steam must be included with perimeter assist air. We are not revising the definition of
assist steam based on this comment.
Comment 2: Several commenters suggested the EPA revise the definition of "Center Steam" to
clarify that this steam does not entrain ambient air and that it is injected in the flare tip to reduce
burn back as well as into burning in the stack of a flare. One commenter suggested the following
regulatory definition change for Center Steam: means the portion of assist intentionally added
steam introduced into the stack of a flare to reduce burnback. There is no ambient air induced
with this steam to assist the burning of the flare gas for smokeless operation. Another
commenter suggested revision to the definition of Center Steam as follows: means the portion
of assist steam introduced into the stack of a flare body of the flare tip to reduce burnback.
internal burning and provide dilution for small flows offlare gas that will generate visible
emissions on a continuous basis and would be adversely effected by other assist steam
injection. Center steam does not induce air into the flare tip.
Response 2: As the definition of "assist steam" already includes the notion that assist steam is
"... steam that intentionally is introduced..." so we deem no further clarification is needed on this
issue because the term assist steam is used. As noted previously, we understand that it is unusual
to use "premix assist air" but flare vendors discussed this method of intentionally inducing air in
with assist steam. If no air is intentionally induced in the center steam, then the "premix air
assist" flow will be zero, but we do not consider it necessary to outlaw or ignore this practice (if
it does occur). We are not revising the definition of center steam based on these comments.
Comment 3: A few commenters suggested the EPA revise the definition of "Lower Steam" to
include the fact that this steam induces ambient air before flowing through the tubes and
minimizes smoke formation. One commenter suggested the following regulatory definition
change for Lower Steam: means the portion of assist steam piped to an exterior annular ring
128
-------
near the lower part of a flare tip, which induces ambient air, and then the mixture flows
through the tubes to the flare tip outlet, and ultimately mixes with the flare gas exiting the flare
tip. Another commenter suggested revision to the definition of Lower Steam as follows: means
the portion of assist steam piped to an exterior annular ring near the lower part of a flare tip,
which then induces air and flows in through tubes passing through t© the flare tip body, and
ultimately exits the tubes at the flare tip exit with no premixing with the flare vent gas stream.
Response 3: We are unaware that all lower steam would "intentionally" induce air and we are
not necessarily concerned about this in the definition of lower steam, nor are we concerned
whether mixing of the steam occurs in/at the flare tip or as exiting the flare tip. The key point is
that any steam, whether mixed with the flare vent gas before the flare tip (center steam), through
tubes at the flare tip (lower steam), or through nozzles at the perimeter of the flare tip (upper
steam), must be considered in the assist steam flow term when calculating combustion zone gas
properties. Assist air intentionally entrained in the lower steam is included in the definition of
perimeter assist air and must be considered when determining the dilution parameters, but it is
not really relevant to the lower steam definition. We are not revising the definition of lower
steam based on these comments.
3.13.3 Flare gas terms
Comment 1: One commenter suggested that for clarity, the EPA revise the term "vent gas" in 40
CFR 63.670(k)(2) to "flare vent gas" since that is defined.
Response 1: We agree with the commenter and have revised 40 CFR 63.670(k)(l), (k)(2), and
(1) [lower case L] to use this term.
Comment 2: One commenter suggested that phrase "minimum amount of gas necessary" in the
definition of flare sweep gas is ambiguous and not enforceable without clarification and that the
EPA should either delete this phrase or add clarifying language to the definition to provide the
regulated community with the concise criteria that the agency will use to determine if the flare
sweep gas rates are "minimized."
Response 2: There is no requirement to minimize the quantity of sweep gas. The intent here is to
recognize that some quantity of sweep gas is necessary for the safe operation of the flare header
system and that even the minimum flow of sweep gas needed for flare operation is to be included
as part of the flare vent gas. However, for the purpose of the flare requirements, there is no
specific need to refer to this as the "minimum amount of gas." We are replacing the phrase "the
minimum amount of gas necessary" with the phrase "the gas intentionally introduced into the
flare header system" in two places within the definition of "flare sweep gas."
Comment 3: One commenter suggested that the EPA should include pilot gas as part of the flare
vent gas because, for some flares, the heat release of the pilots can be a significant portion of the
total heat release from the flare. Alternatively, one commenter supported EPA's assessment and
agreed with the EPA that it is unnecessary to consider pilot gas in calculating the combustion
zone parameters.
129
-------
Response 3: We appreciate the comments. The data available to us suggests that heat release
from the flare pilots are generally negligible when regulated materials are sent to the flare and
exclusion of the flare pilot gas simplifies the NHVCZ calculation. Even when only purge gas is
used, the flare pilots typically only provided about 10 percent of the total heat input to the flare
and typically well less than 1 percent in the recent PFTIR flare tests when potential regulated
material is routed to the flare (this is dependent on the size of the flare, number of pilots, and
flare tip design, which impacts minimum purge flows). We are finalizing the definition of flare
vent gas as proposed, which excludes pilot gas.
3.14 Cost and emission impacts for flares
Comment 1: Several commenters suggested that the EPA underestimated the cost burden that
would be imposed on industry by the proposed new flaring requirements. To support their
claims, the commenters suggested that, among other things, the EPA failed to account for the
500+ new flares and needed to comply with the flare tip velocity and visible emissions
requirements during emergencies (and 200+ new flares needed to control atmospheric relief
valves in organic HAP service), costs for new automatic pilot ignition systems, costs for
personnel needed to perform daily Method 22 observations, and that the EPA should have
evaluated the cost burden considering at a minimum a 10% compliance margin. Additionally,
commenters went through their own cost analysis, making assumptions that a much larger
percentage of flares will need continuous controls (and in most cases it will need to be redundant
controls of both a GC and Btu analyzer) as well as larger excesses of natural gas because of the
proposed 15-minute compliance time, proposed dual combustion zone limits, and additional
operational 10% compliance margin and suggested that the new flare requirements would lead to
a capital investment of at least $343 million and annualized costs of $118 million versus the
EPA's estimates of $147 million capital and $36.3 million annualized. One of these commenters
also noted that the GC specifications in the proposal cannot be met by the GCs typically in place
to meet State, consent decree, and permit requirements, as the EPA assumed. This commenter
stated that their existing GC installations may have to be supplemented, adding an additional
$500,000-$600,000 cost. Lastly, several commenters suggested that a combustion zone net
heating value of 200 Btu/scf as a 3-hr average (and equivalent lower flammability limit and total
combustibles combustion zone metrics) is all that is justified under a cost-effective basis to
provide reasonable assurance of 98% destruction efficiency for steam-assisted refinery flares.
Response 1: First, it was not our intent or expectation that refineries would have to install
hundreds of new flares in attempts to meet the velocity and visible emissions limitations at all
times. We understand that, under high "hydraulic" load events, that the flare tip velocity and
visible emissions limits possibly cannot be met, but we also understand that these events are
infrequent and typically caused be power outages or other activities out of the refinery owners or
operators control and that discharging these gases to the flare is better for both plant safety and
environmental control, even when these emission limitations are not met. Having reviewed all of
the comments, we have implemented a work practice standard to manage these large but
infrequent events. This work practice standard is expected to reduce the frequency of these
instances while obviating the need for hundreds of new flares. As such, we deem the extreme
costs estimates for the installation of new flares to be unnecessary and inapplicable to the
requirements in the final rule.
130
-------
Next, and as discussed previously in Section 3.5 of this document, we have included an
allowance to use 1,212 Btu/scf for the net heating value of hydrogen and eliminated the proposed
dual combustion zone NHVCZ operating limits. The provision for hydrogen allows the NHVCZ
limit to be an excellent indicator of good flare performance without "Type 2" errors previously
expected from this compliance option so that refinery owners or operators may be less compelled
to install full GC systems. With the elimination of the LFL and combustibles concentration
operating parameters, we consider it unnecessary for flare owners or operators to install
monitoring systems of both a GC and BTU analyzer as suggested by the commenters. We did
revise our cost estimates for refinery flares to include costs to install hydrogen analyzers needed
to determine hydrogen concentration in the flare vent gas (to allow use of the adjusted heat
content of 1,212 Btu/scf for hydrogen). We also made minor changes to the supplemental natural
gas and steam savings that results in complying with the "hydrogen adjusted" NHVCZ operating
limit. Thus, as a result of these actions, we project the costs for refinery flare combustion
efficiency requirements to increase slightly from $147 MM capital cost and $36.3 MM
annualized cost at proposal to $156 MM capital cost and $45.6 MM annualized cost in this final
action (see Table 8 in memorandum entitled Flare Control Option Impacts for Final Refinery
Sector Rule in Docket ID No. EPA-HQ-OAR-2010-0682).
Lastly, the data available clearly refute the unsubstantiated assertion that a 200 Btu/scf 3-hour
average NHVCZ provides "a reasonable assurance of 98 percent destruction efficiency for steam
assisted flares." Our Monte Carlo analysis suggests that 90 percent of flares would not be
achieving a 98 percent destruction efficiency if the NHVCZ operating limit was established at 200
Btu/scf on a 3-hour average basis.
Comment 2: A few commenters suggested that the EPA overestimated the emissions reductions
that the newly proposed flaring requirements would achieve while another commenter suggested
that the EPA underestimated these benefits.
One of the commenters that suggested that the EPA overestimated the emissions reductions
points out that recent refinery flare consent decrees and the September 2012 finalized NSPS Ja
amendments have driven flaring baseline emissions down and that they would project a 2015
baseline of 22,300 tons per year (tpy) VOC and 2,560 tpy HAP compared to the EPA's estimate
of 33,100 tpy VOC and 3,810 tpy HAP. This commenter also did another analysis with differing
assumptions from those of the EPA's (e.g., a different CE v. DE relationship, a higher baseline
destruction efficiency, a 10% compliance margin for complying with the combustion zone limits
and operating at 300/400 BTU/scf (proposed limits of 270/380 BTU/scf)), and this analysis
determined that the proposed new flare requirements would result in smaller reductions of
11,150 tpy VOC, 1,280 tpy HAP, and actually cause a net increase of 111,700 metric tpy C02e
(compared to the EPA's estimated 327,000 metric tpy carbon dioxide equivalents (CO2Q)
reduction).
Alternatively, one commenter suggested that the EPA underestimated the benefits of the
proposed new flaring requirements (0568) supported this assertion by stating that the average of
the PFTIR test data shows that flares only reduce emissions by 92% (and not 93.9% as was
assumed in the EPA's analysis) and that evidence from specific refineries shows that flare DE
could be 12-25 times higher than the presumed 98% destruction efficiency and that it is arbitrary
131
-------
and capricious for the EPA to rely on the industry's data when evidence suggests that it is
inaccurate. This commenter also suggested that the EPA has determined that the AP-42 emission
factors for VOC emissions from flares operating at 98% destruction efficiency are approximately
four times more on a heat input basis and that if this same ratio holds at lower destruction
efficiencies that emissions reductions should be significantly higher.
Response 2: We appreciate these comments. Given that much of the information used to support
this rulemaking came from an industry-wide 2011 ICR (with 2010 data), we still continue to feel
that use of data from 2010 and 2011 from the EPA's GHGRP in concert with flare specific
industry submitted data used to assess additional potential emissions reduction impacts from the
same time period is consistent with our overall approach of using best available data as well as a
consistent timeframe in estimating the impacts of this rulemaking. We reviewed the more recent
GHGRP data for refinery flares and see no evidence of reduced use of flares, so our baseline
emissions estimates appear to accurately portray current industry flaring practices. Therefore, we
consider the approach we used to be a reasonable projection of the current emissions and
projected emission reductions. We have revised our analysis slightly, considering the allowance
to use 1,212 Btu/scf as the net heating value for hydrogen and the use of a single NHVCZ
operating limit. Based on these revisions, we estimate 3,670 tpy emissions reduction of HAP,
31,900 tpy emissions reduction of VOC, and 396,000 metric tonnes per year emissions reduction
of CChe as a result of implementing the refinery flare NHVCZ operational requirements as
provided in the final rule.
Comment 3: Two commenters stated that cost-effectiveness of the NESHAP proposal must be
adjusted to reflect any current local programs that address the same pollutant but perhaps in
different ways. For example, consider the Flare Efficiency Requirements. Most refineries in
California, including those in Wilmington and the Bay Area, already are under regulations that
minimize or monitor flare utilization. As a result in California, flare use has already been
minimized by regulation since the early 2000's and the reduction in flare use and flare emissions
have already been documented.
These commenters recommended that the EPA's cost-effectiveness analysis be adjusted to
reflect the capital and operating costs of emissions control technology. Rather than looking at
overall costs for the nation, the EPA should document cost-effectiveness of the NESHAP
controls by region because it seems likely that the cost-effectiveness calculations are likely too
low. One of these commenters added that this is probably also true for toxics emissions as well
as for VOC emissions and even for GHG emissions such as CO2.
One of the commenters also stated that the EPA should be aware that the AB2588 process in
California, for example, reduced toxic emissions and risk, such that most refineries posted risks
that are already less than 10, certainly less than 25-in-l million, which appears to be
meeting EPA's target for residual risk, after implementation of all these controls. The commenter
asked how refineries should address the EPA proposal if they have already achieved target
levels.
Response 3: Our emission reduction estimates considered consent decrees as well as state and
local requirements where these were known. Therefore, we disagree with the commenters that
132
-------
we overstated the emissions reductions. With respect to California provisions for flares, we
acknowledge that these requirements helped to reduce flare emissions by reducing the amount of
gas flared, but these requirements do not necessarily ensure that the flare is achieving the
desired control efficiency. The flare impact estimates relied on nationwide data reported under
the GHGRP (40 CFR part 98) based on 2010 and 2011 data, which would have accounted for
reduced flare use from early 2000s regulations.
As this is a federal standard, we generally present the costs and environmental impacts on a
nationwide basis. We also develop facility-specific cost estimates for use in the economic impact
analysis. We acknowledge that the impacts and cost-effectiveness of the new standards varies for
different facilities, but unless there are specific adverse impacts on a class of refineries (such as
small businesses), we generally use the nationwide impact estimates to support our decisions.
With respect to AB2588, we note that the new flaring provisions are being promulgated under
CAA section 112(d)(2) and (d)(3) and are not being promulgated under section 112(f). All major
source refineries will be required to comply with all of the flaring requirements in the final rule,
regardless of what their contribution to risk is projected to be. We also expect that the risks
projected by the commenter assumed that all flares were achieving 98 percent destruction
efficiency. The final Refinery MACT requirements for flares will ensure that these flares achieve
that control efficiency.
3.15 Other/Ancillary Flare Comments
Comment 1: One commenter suggested that the implementation of low sulfur gasoline standards
and clean fuels projects created dirty neighborhoods because the EPA failed to simultaneously
require upgraded pollution control technology at the refineries and that fence line communities
near refineries may be more impacted even though emissions from vehicles will go down.
Because of this, they argue that fence line monitoring is crucial for these communities and
wanted the EPA to clarify how many and which refineries are using flares as a pollution control
device as well as to clarify how these proposed requirements will prevent more pollution in the
event that more clean fuels projects are implemented.
Response 1: The commenters do not explain how they EPA should address their concerns
through the residual risk and technology review rule at issue here. We assume their concerns are
with the SRU, which many refineries had to install (or increase the use of existing units) to
manage and recover the additional sulfur removed from the fuels as a result of the clean fuel
standards. However, the commenters do not explain why they believe our risk and technology
review for SRU was insufficient.
Second, we note that we have included requirements for fenceline monitoring. The fenceline
monitoring program is for the purpose of addressing fugitive emissions and thus would not
address emissions from the SRU.
With respect to the request that the EPA clarify how many and which refineries are using flares
as a pollution control device, we note that the 2011 Refinery ICR included information regarding
the number of flares and the general use of the flares (routine use classified by hours of operation
133
-------
as well as emergency only) and this information was included in the docket for the proposed rule
(see Docket Item Nos. EPA-HQ-OAR-2010-0682-0061 through 0069). Any flares for which the
General Provisions of 40 CFR part 60 or subpart 63 apply are assumed to be used as control
devices. With respect to the request that the EPA clarify how the proposed requirements will
prevent more pollution in the event that more clean fuels projects are implemented, we maintain
that the final rule will ensure any new SRU or any flare receiving gases from clean fuels related
project will be controlled to the MACT performance level.
Comment 2: One commenter suggested that stronger regulations for flaring at night are needed
and that stronger community monitoring systems are needed to track and chart events and alert
nearby residents when the events occur, through both web-based and phone-based systems.
Response 2: The final rule will ensure high flare combustion efficiencies, whether they occur
during the day or night. As noted previously, we consider the existing emergency notification to
be reasonable and appropriate and that the community will be notified in the event of a release or
flaring event that creates a risk that would require evacuation of nearby residents. It is
unreasonable to notify the public of every instance in which regulated materials are sent to a
flare, therefore, we did not revise the rule requirements in response to this comment.
Comment 3: One commenter suggested that the flare monitoring requirements for presence of a
pilot flame and no visible emissions as well as continuous monitoring of flare tip velocity
accounting for the assist media to determine the combustion zone gas properties will give
refiners a more accurate picture of the actual combustion efficiency of flares and that owners
now have options and flexibility on how to demonstrate compliance.
Response 3: We agree that the monitoring requirements will provide refinery owners or
operators with significantly more information than most refinery owners or operators currently
have and that this information, along with fine air or steam assist controls that we included in or
cost estimates, will enable them to operate their flares in a more efficient manner.
Comment 4: One commenter suggested that the requirement to replace flare tips is being
implemented on an impractical timeline and that doing so is minimally beneficial at best.
Response 4: We did not propose, nor are we finalizing, requirements to replace flare tips.
Comment 5: One commenter suggested that the EPA consider providing exceptions or reduced
requirements for refineries already complying with more stringent local rules pertaining to
flaring and fugitive emissions.
Response 5: We do not agree that such exceptions are warranted because we have found no local
requirements that are equivalent in performance to those in the final rule.
Comment 6: One commenter suggested that Refinery MACT 1 and 2 must specifically
authorize the additional flaring required for compliance. Other commenters suggested this is
needed because NSPS Ja requires minimizing flare emissions and Refinery MACT 1 and 2 will
require new flare instrumentation and some of the QA/QC requirements for this instrumentation
134
-------
could cause an increase in the base emissions for the purpose of establishing the basis for the
NSPS Ja 500,000 scf/d flare excess flow trigger. Additionally, commenters suggested these flare
flows should not be counted in determining whether a flare meets the "emergency flare" criterion
in NSPS Ja of 4 releases per year or less.
Response 6: We disagree. Generally, flare refinery owners will comply with the flare
requirements by reducing steam flow, so we consider it unlikely that the final Refinery Sector
Rule will significantly impact baseline flow characteristics. We also do not agree that the
monitoring systems or QA/QC requirements will ever have an appreciable impact on baseline
flare flows. For those few flares where supplemental fuel is needed to meet the operating limits
in the final rule, refinery owners or operators may always request a revision to their NSPS
baseline flow rates. Finally, we see no reason the requirements in this final rule would ever cause
additional emergency flare releases. In fact, we are implementing work practice standards as part
of this final rule that are expected to reduce the frequency of emergency release events.
135
-------
4.0 Miscellaneous Process Vent Provisions
Comment 1: Many commenters stated that where a CPMS flow meter will now be required,
three years is needed to engineer, procure, and install that system and thus the effective date
should be three years after promulgation. The commenters added that eighteen months is
required to bring all other instruments up to the new requirements, change procedures, train
personnel, and revise permits, obtain and/or revise AMPs, and revise Consent Decrees and/or
comply with an alternate work practice if adopted.
Response 1: The primary requirement for a CPMS flow meter in the MPV provisions is for
bypass lines. Previously these bypass lines only required a "flow indicator that determines
whether a vent stream flow is present at least once per hour." We agree that, for bypass lines, the
new CPMS flow requirements are significantly more stringent than previously required. As
described in more detail in subsection 4.4 of this section, we are not finalizing this requirement
for bypass lines, so no additional time is needed to comply with the bypass line monitoring. It is
possible that flow meters may be used for alternate monitoring provisions in 40 CFR 63.644(b),
but it is expected that such flow meters would be used for the purpose of quantifying flow rates,
not simply as an indicator of flow. The 5 percent accuracy requirement for flow rate monitoring
systems is a reasonable accuracy requirement for industrial flow meters and we do not expect 18
months is needed to meet these requirements.
The other primary CPMS for MPV are temperature monitoring devices used for thermal
destruction devices. We do not consider the requirements in Table 13 to be onerous for these
systems. As the normal operating range is likely 1,300 to 1,700 degrees F, the ± 1 percent
accuracy requirement requires measurement to ± 13 or 17 degrees F. For refineries using
alternate monitoring systems as provided in 40 CFR 63.644(b), we do not expect that any
reasonable monitoring system to have difficulty meeting the accuracy requirements in Table
13. Therefore, we disagree with commenters that additional time is needed to upgrade
monitoring systems used for MPV.
Comment 2: One commenter explained that in proposing revisions to the Refinery MACT 1
MPV provisions under the authority of section 112(d)(2) and (3), the EPA may be subject to
claims that it reopened the entirety of its original MACT standard for comment and challenge,
because the data used to support the EPA's proposal calls into question the foundation of that
standard. When the EPA originally set the MPV standard, it based the standard on a finding that
"combustion control" qualified as the MACT floor. Lacking specific data on the control
efficiency that could be achieved by "combustion control," the EPA relied on information from
its experience under the NSPS and HON source category regulations to assume an achieved level
of control at 98% VOC emissions reduction. The commenter argued that now, the EPA offers
information suggesting flares are not achieving this level of emissions control.
The EPA's flares impacts analysis claims that combustion control under the current MACT
standard achieves a nationwide average control efficiency of only 93.9%. In this regard, the
EPA's rationale for the proposed rule could appear diametrically opposed to its original standard.
For example arguably, the EPA should not simultaneously claim that a 98% combustion control
efficiency is not assured without the proposed rule enhancements, and yet continue to claim that
136
-------
the MACT floor was 98% emissions reduction when the best performing 12% of sources were
not implementing these enhancements when the EPA established the MACT floor.
The commenter asserted that they do not believe that the EPA is necessarily compelled to re-
examine the basis for its MACT floor decision to make improvements to the monitoring,
recordkeeping and reporting requirements of a standard, in all cases. Here, however, because the
EPA arguably presents data to support its proposed rule that calls into question the EPA's
original basis for the control level, and also relies on section 112(d)(2) and (3) as the legal
authority for the proposed changes, the EPA may be vulnerable to claims that the EPA must
address the impact that this new information has upon its original decision-making.
"Rulemakings that significantly change the context for a regulatory provision can re-open it for
comment, even if an agency does not change the provision itself." Sierra Club v. EPA, 551 F.3d
1019, 1024 (D C. Cir. 2008).
The commenter stated that they believe that the EPA established 98% as the appropriate control
efficiency for the original section 112(d)(2) and (3) standard. Even if the EPA should have
properly established the MACT floor at a lower control efficiency, such as 93.9%, the
commenter remains confident that the EPA could have justified a 98% control efficiency as a
control measure beyond the MACT floor that was cost effective and reasonable considering the
non-air quality health and environmental benefits, and energy requirements. Nonetheless, to
avoid reopening the original standard, and then defending the original basis for the standard, or
re-affirming the original standard as cost-effective, the commenter urged the EPA to consider
relying on section 112(d)(6) rather than section 112(d)(2) and (3) as the source of authority for
its final regulations, because this will not have the effect of potentially reopening the original
basis for the standard.
Response 2: We disagree that our proposal regarding flares calls into question our MACT floor
analysis for MPV. We determined in the 1995 Refinery MACT 1 (subpart CC) that the MACT
floor level of control for MPV is 98 percent VOC control (or reduction to 20 ppmv on a dry
basis, corrected to 3 percent oxygen) based on the number of MPV controlled and the types of
controls used on those MPV. At the time we set the MACT standard for MPV, we found that
well over 12 percent of MPV used "combustion controls" that achieved 98 percent control
efficiency. We note that not all "combustion controls" used for MPVs were flares (some
facilities used thermal oxidizer/incinerators and some facilities used boilers or process heaters to
control MPV). Additionally, we believe that flares prior to the MACT standard generally
achieved 98 percent control efficiency and that it was only after the implementation of the
Refinery MACT 1 standards, which imposed visible emissions limits through the General
Provisions requirements for flares, and the increased use of flare gas recovery systems that
refineries began to over-steam their flares had have reduced combustion efficiency. Even if some
flares in 1995 were not meeting the 98 percent control efficiency at the time that MACT standard
was set, we still expect that most flares were meeting this control efficiency. Even today, with
average performance of all flares estimated to be 93.9 percent based our analysis of the recent
API flare data, that same data set indicates that 20 of the 38 flares achieved 98 percent
destruction efficiency in all operating modes. Thus, we are confident that the best performing
flares (along with thermal oxidizers, boilers and process heaters) were achieving 98 percent
137
-------
destruction efficiency, so there is no question that the original MACT requirements for MPV
were correctly established at 98 percent control efficiency.
While the best performing MPV sources were (and still are) meeting 98 percent control
efficiency, the API data together with recent flare performance studies indicate that the on-going
monitoring requirements established in the MACT rule (i.e., reliance only on the General
Provision requirements) were not adequate to ensure that all flares were performing at the level
equivalent to the top 12 percent of sources. As such, we are finalizing the monitoring standards
for flares to ensure these control systems achieve the MACT performance level established in
1995 under CAA section 112(d)(2) and (d)(3).
Comment 3: One commenter stated that carbon adsorption controls are widely used to control
remotely located, small and/or dilute or low pressure emission sources, such as waste
management unit vents and certain tankage. The commenter explained that carbon adsorption
generally cannot be used during shutdown of this equipment if steam is present (e.g., during
hydrocarbon-freeing equipment) because the water will damage the carbon and/or prevent
adsorption of hydrocarbons and/or cause desorption of already adsorbed hydrocarbons. In such
cases, good air pollution practice is to bypass the carbon beds to prevent damage to the carbon
and desorption of previously adsorbed hydrocarbons. The commenter recommended language be
added to Refinery MACT 1 clarifying that carbon adsorption controls may be bypassed when the
equipment they control is being steamed for hydrocarbon freeing purposes.
Response 3: This request to bypass the control device appears to be a request to allow
unregulated emissions from MPV for a period of time. We do not believe that this is consistent
with the requirement that emission standards must apply at all times. Best practices would be to
have a knock-out drum and/or cooler to prevent steam or very hot gases from entering the carbon
adsorber (and stripping VOC/HAP from the carbon in the adsorber). In addition, water spray
cooler with a knock-out drum is a reasonably efficient and inexpensive means to lower the vent
stream temperature prior to the carbon adsorber. Other MPV control systems, such as
combustion devices, would allow sources to meet the applicable emission standard during these
steaming events.
4.1 Revisions to the definition of miscellaneous process vents
Comment 1: One commenter stated that the proposed rule would preserve the exemption for
emissions associated with "vents from storage vessels", however this fails to comport with the
definitional provision of EPA's broader statements regarding SSM provisions. The commenter
specifically stated that it is unclear whether emissions associated with vessel degassing would be
subject to MACT requirements under the Refinery MACT as SSM related, or governed by the
exemption from the definition of MPVs, and therefore not subject to any standards under
Refinery MACT. The commenter suggested that the proposed rule be revised to clarify the
inapplicability of MACT standards during these operating scenarios.
Response 1: Vents from storage vessels are not MPV as provided in item 12 of the MPV
definition. We address requirements applicable to storage vessels separate from the requirements
for MPV. We understand some storage vessels will have PRD which may release as a result of a
138
-------
process upset or malfunction. Storage vessel PRD in organic HAP service are subject to the
equipment leak monitoring provisions and, if the set pressure is greater than 2.5 psig, the
pressure release management provisions in 40 CFR 63.648(j).
Comment 2: One commenter interprets EPA's proposal to remove high point bleeds, open-
ended valves or lines, and pressure relief valves from the list of exceptions to the requirements of
40 CFR 63.644(c), to mean that these types of equipment are subject to the 40 CFR 63.644(c)
monitoring requirements unless they meet the exception of being subject to 40 CFR 63.648.
Many commenters objected to the inclusion of high points bleeds as MPVs and suggested this
change not be finalized. Additionally, one commenter explained that the examples of MPVs
included in the proposed definition create confusion, and that it is the material vented, not the
piping component through which the venting occurs that is considered in deciding whether a
stream sent to a control device or to the atmosphere is a MPV. Many commenters argued
that these changes are unclear and should be clarified or the existing language left unchanged.
Response 2: The provisions in 40 CFR 63.644(c) are specific to Group 1 MPV bypasses. High
point bleeds are expected to be used primarily on liquid transport lines to collect and remove
gases that might enter the system. In this application, we agree that the high point bleed would
not be a bypass of a Group 1 MPV, rather the high point bleed would be an MPV itself, thus
engineering calculations would be used to determine if this vent is a Group 1 MPV requiring
control or a Group 2 MPV. In rare instances, the owner or operator may classify a release point
on a gaseous vent system associated with a Group 1 MPV as a "high point bleed." In this case,
the high point bleed when open acts as a bypass line (allowing direct atmospheric release) of a
Group 1 MPV stream). These examples demonstrate that depending on the circumstance, a high
point bleed could be construed as an MPV or a bypass line. Thus, we see no reason to
categorically allow use of high point bleeds to bypass controls required for a Group 1 MPV
streams. Additionally, in the case of liquid transport lines, we expect that this exemption has
been used to neglect these releases from evaluation as a Group 1 or 2 MPV. We therefore retain
the revisions to not include high point bleeds in the list of equipment that are not considered to
be a bypass line.
We consider open-ended valves or lines and pressure relief devices to be equipment subject to 40
CFR 63.648 so it is redundant to list these items separately in 40 CFR 63.644(c). The first
sentence remains the primary definition of MPV, i.e., "Miscellaneous process vent means a gas
stream containing greater than 20 parts per million by volume organic HAP that is continuously
or periodically discharged from a petroleum refining process unit meeting the criteria specified
in section 63.640(a)." Consistent with the definition, we agree that the first criterion to consider
in determining whether a stream is an MPV is whether the gas stream contains greater than 20
ppmv organic HAP discharged from the petroleum refining process at a major source of HAP
emissions. We are retaining the list of potential release points that could be an MPV because we
consider them to be useful and add clarity to the definition.
4.1.1 Revision of fuel gas system exclusion
Comment 1: Two commenters raised a concern that the MPV provisions would apply to flares
that are connected to a fuel gas system and that typically operate only during SSM because of
139
-------
EPA's proposal to remove the fuel gas exemption for MPV, which according to one commenter
directly conflicts with the EPA's Refinery MACT 1 finding, "that it is not necessary to revise
Refinery MACT 1 requirements for miscellaneous process vents..One of these commenters
added that refinery fuel gas systems are processes that generate fuel gas for internal and
sometimes external use, and thus fuel gas is a refinery product just as gasoline is a refinery
product. This commenter further noted that the proposed rule could trigger immediate
applicability of Refinery MACT 1 or 2 evaluation and control requirements including permitting,
closed vent system and control device requirements for gas streams routed to that fuel gas
system. The commenter stated that the requirement is unlawful because it would arbitrarily and
without justification convert countless internal process streams into MPVs and estimated that
hundreds or thousands of streams could be impacted.
Another commenter recommended two changes to the proposed regulatory text if the EPA
eliminates the current exemption for vent streams routed to fuel gas systems from the
requirements for miscellaneous process vents. First, the commenter stated the definition of
"miscellaneous process vent" should include an exemption for vent gas streams routed to
another refinery processing unit for further processing, internal use or sale. Second, the
commenter requested that the regulations include an exemption for vent gas streams routed to the
fuel gas system provided that any fuel gas going to flare is normally recovered by a flare gas
recovery system which is designed with excess capacity under normal operations and operates at
least 95 percent of the time.
Response 1: First, we note that the full quotation from Section IV.B. 1 .a of the proposal
preamble is "that it is not necessary to revise Refinery MACT I requirements for miscellaneous
process vents pursuant to CAA section 112(d)(6)." This conclusion follows our discussion of the
inadequacy of the General Provisions at 40 CFR 63.11 to ensure flares used as control devices
achieve a 98 percent destruction efficiency (in Section IV. A 3 of the proposal preamble) and
need to ensure flares used with fuel gas systems currently exempted from the definition of MPV
achieve the desired control efficiency (in Section IV.A.4.d. of the proposal preamble), revisions
that were taken pursuant to CAA sections 112(d)(2) and (d)(3). Therefore, the technology review
conclusion pursuant to CAA sections 112(d)(6), which focused on the primary control
requirements for MPV (i.e., 98 percent reduction or 20 ppmv), does not conflict with conclusions
made pursuant to CAA sections 112(d)(2) and (d)(3) to ensure that flares achieve the 98 percent
destruction efficiency that was the determined to be the MACT floor.
Second, we did not remove the exemption for fuel gas systems, rather the exemption has been
revised to specify that flares used with the fuel gas systems are operated with appropriately high
combustion efficiencies. In our impact analysis, we assumed that all refiners would elect to
comply with the new flare monitoring and operating requirements rather than consider all fuel
gas streams to be a Group 1 MPV. While it is possible that a refinery could elect to treat their
fuel gas as an MPV (if, for example, all of the fuel gas is used in process heaters or boilers with
heat input capacity of 44 megawatts or greater, so no other monitoring is required), such an
action would not exclude the flares that might receive that fuel gas from meeting the new flare
monitoring and operating requirements. We expect that facilities will elect to upgrade their flare
monitoring and operations and retain the fuel gas exemption provided in Item 1 of the definition
of miscellaneous process vent. Since we are not eliminating the exemption for vent gas streams
140
-------
routed to the fuel gas system, we do not see a need to add additional exclusions suggested by the
commenter. However, we do note that gases recycled internally to a process or that are purified
and sold as a product and that are not vented would not meet the definition of MPV so these
streams would be excluded in any case.
Finally, with respect to the suggestion that we allow an exemption for fuel gas systems with
flares with flare gas recovery, we find it extremely difficult to develop requirements for a flare
gas recovery system that could be considered equivalent to the 98 percent MPV control
requirements. It is important to note that the recovered flare gas is generally being used in a
process heater or boiler. While we agree these units achieve 98 percent combustion efficiency or
higher without additional monitoring or operating requirements (and thus we are retaining the
primary exemption for gases routed to the fuel gas system), they do not necessarily achieve 100
percent HAP emission reduction. Thus, a flare gas recovery system recovering 98 percent of the
gas that would otherwise be sent to a flare is not necessarily equivalent to 98 percent emissions
reduction of all HAP in the Group 1 MPV unless the 2 percent that is not recovered is also
efficiently controlled. While we would like to encourage the use of flare gas recovery systems,
we find that the exemption requested by the commenter is not equivalent with the MACT control
requirements for MPV, and we have not provided such an exemption from the flare monitoring
and operating requirements or MPV provisions for flares utilizing flare gas recovery
systems. We do note that facilities have the right to request alternative monitoring options to
demonstrate that their overall system achieves 98 percent HAP reduction at all times, regardless
of the flare's level of performance.
Comment 2: One commenter stated the proposal to eliminate the route-to-fuel gas exemption
imposes costs and burdens for no benefit and should not be finalized. The commenter noted that
the EPA has failed to provide any explanation of the legal authority on which it is relying for this
new requirement. In the preamble, the discussion of this new requirement is located in a section
entitled "What actions are we taking pursuant to CAA sections 112(d)(2) and 112(d)(3)?", but
according to the commenter, the EPA has not provided the explanation and analysis needed to
understand how this new provision satisfies the requirements of sections 112(d)(2)/(3). The
commenter stated the EPA did not identify the best performing source or sources, did not
calculate a MACT floor, and did not set a standard based on consideration of the mandatory
statutory factors (such as consideration of the cost of any "above the floor" alternatives). In
short, one commenter argued that the EPA has effectively proposed a flare performance standard
for flares in which fuel gas is combusted, but has not asserted any factual or legal justification for
this new standard and, therefore, has violated EPA's obligation under section 307(d)(3) to
provide an explanation in the proposed rule of the basis and purpose for this requirement.
Response 2: First, we note that we are not eliminating the route-to-fuel gas exemption; rather,
we are simply conditioning it to those systems that may discharge to a flare that does not meet
the Group 1 MPV control efficiency requirements. As explained in Section IV.A.3, the EPA
determined that the current requirements for flares are not adequate to ensure compliance with
the MACT. The previous exemption for fuel gas systems was predicated on an assumption that
all fuel in a fuel gas system would be introduced in the flame zone of a combustion device,
effectively ensuring efficient combustion with no on-going monitoring requirements, or sent to a
flare. As noted elsewhere, the MACT floor level of control has not changed; however, we are
141
-------
implementing requirements, consistent with our determination that improperly operated flares do
not meet the required control efficiency, to ensure Group 1 MPV streams that are discharged to
the flare from the fuel gas system are properly operated and achieving the MACT floor control
level.
We note that the limitation of the fuel gas exemption does not become effective for 3 years,
providing time for refinery owners or operators to upgrade their flare systems. In our impact
estimates, we assumed all refinery flares would install the enhanced monitoring system,
including those serving a refinery fuel gas system.
Comment 3: One commenter stated that the section 63.640(d)(5) affected facility exception
applies to emission points that are routed to fuel gas and that the current rule repeats the "route to
fuel gas exception" in several places indicating its importance and the potentially large impact of
these proposed revisions. The commenter suggested that only refinery flares should be subject to
Refinery MACT 1 or 2 flare requirements and specific exclusions should be given for flares not
owned or operated by the refinery (e.g., it is a flare in an associated chemical facility). For flares
which are not refinery flares, a commenter suggested an hours per year criterion (based on
various flare APCDs) be used to avoid imposing the costly refinery flare requirements and the
burdens associated with this change on flares that only receive refinery fuel gas occasionally or
due to unusual circumstances. The commenter asserted that there is no legal, logical, or
environmental basis for imposing large costs and operating burdens on non-refinery flares
because of the potential for a small amount of regulated HAP to reach that flare via a fuel gas
system. Thus, the commenter asserted the proposed revision should not be finalized and the
routing of fuel gas to flares should be addressed directly by defining fuel gas vented from a fuel
gas system that might contain what would otherwise be Group 1 vents, as a Group 1 MPV. The
commenter provided specific suggested changes to the regulatory text to address their concerns.
Response 3: Contrary to the concerns of the commenter, Refinery MACT, including the
revisions being promulgated in this final action, do not apply to flares that do not receive refinery
fuel gas or that are not located at the refinery affected facility (e.g., associated with an off-site
chemical facility). In most applications, it is readily apparent which flares could receive gas from
a fuel gas stream that receives gases that would otherwise be a Group 1 MPV gas stream.
However, if a facility is uncertain regarding the applicability of the rule in a specific instance,
then that facility can request an applicability determination. In our impact analysis, we assumed
all flares at the refinery would be subject to the new flare monitoring requirements primarily as a
result of the consideration of flares used as back-up control systems to a refinery's fuel gas
systems. We recognize that some flares may only occasionally act as a Refinery MACT 1 (or 2)
control device. We explicitly included language in the flare monitoring and operating limit
requirements to specify that these flares only need to meet the Refinery MACT requirements
when the flare is acting as a Refinery MACT control device, but we see no reason to disregard
the Refinery MACT requirements simply because the flare is not always acting as a Refinery
MACT control device. We would hope that, given the improved information on flare gas flow
and composition afforded by the required monitoring systems, refinery owners or operators
would elect to operate the flare efficiently at all times whenever the flare is receiving waste gas.
However, we recognize that these standards are specific to flares used as Refinery MACT
control devices, so the operating limits apply only during times the flare is acting as a Refinery
142
-------
MACT control device. Nonetheless, when being operated as a Refinery MACT control device
the control requirements must apply at all times, so flares that are receiving "regulated material",
i.e., a Group 1 MPV stream, must comply with the flare monitoring and operating limits at all
times. For these same reasons, we are not changing the proposed regulatory text as the
commenter suggested.
Comment 4: Two commenters stated that if the proposed approach is finalized, it is critical that
the wording of proposed 40 CFR 63.640(d)(5), 63.1562(f)(5) and exception 1 to the MPV
definition be revised to remove the language requiring "compliance" with 40 CFR 63.670. One
commenter explained this language would result in the affected facility and MPV exceptions
being lost for any 15 minute period where compliance is not being achieved while fuel gas is
routed to that flare. Loss of those exceptions for even 15 minutes would trigger the massive
burdens of creating new MPVs and revising the affected facility. The commenter argued that
failure to meet a flare requirement should only result in a deviation for that flare. Thus, the
commenters stated that the fuel gas exemption should use the phrase "subject to 40 CFR 63.670"
rather than "in compliance with 40 CFR 63.670."
Response 4: We did not intend that a deviation of the flare requirements would cause all other
combustion devices associated with the fuel gas system to become Group 1 MPV and subject to
any applicable MPV monitoring requirements (either ongoing or only during the deviation).
While we do not consider it likely that the current language would be interpreted in this manner,
we agree that the phrase "subject to" rather than "in compliance with" would more accurately
express our intent in this case, so the final regulatory text uses the phrase "subject to."
Comment 5: One commenter suggested that fuel gas used as sweep gas, purge gas, assist gas,
and pilot gas for flares be excluded from the Group 1 requirements, as the their amount of uses is
minimized per NSPS Ja requirements.
Response 5: We disagree that an exemption should be given for sweep, purge, assist and pilot
gases because if the fuel gas comes from a fuel gas system that receives a Group 1 MPV stream,
then the combustion of the fuel gas must meet the minimum efficiency requirements for a Group
1 MPV stream because the standard applies at all times. We also disagree that the subpart Ja
requirements minimize the use of sweep, purge, assist and pilot gases as the requirements were
primarily focused on reducing waste gas that is flared. As we understand, a minimum quantity of
gas is needed as purge/sweep gas to ensure air (oxygen) does not enter the flare tip and facilities
already have an economic incentive not to use more gas than needed for this purpose. Similarly,
pilot gas is used to maintain a constant pilot flame and is not varied with flare use or reduced as a
result of the NSPS Ja flare provisions.
4.1.2 Removal of in situ sampling exclusion
Comment 1: One commenter stated that the definition of MPV should not be revised with
respect to the "in situ sampling (on-stream analyzer)" exclusion. If the definition is revised in the
final rule, the commenter contended that the proposed revisions are a confusing approach to
address analyzer vents and that a clearer regulatory approach would be to define "in situ
sampling systems" and maintain an exclusion for these systems in the definition of MPV. The
143
-------
commenter also recommended separately defining on-stream analyzer vents which would be
covered as a MPV.
Response 1: The definition of MPV historically contained an exclusion for "in situ sampling
system (on-stream analyzers)." We did not change the terminology; we merely proposed to phase
out this exclusion. The term "in situ sampling system" is not defined and given the previous
wording, it appears that in situ sampling systems and on-stream analyzers are considered to be
interchangeable terms. Given the historic phrasing, we have concluded that the most clear and
direct approach to resolve concerns with large sampling system or analyzer vent releases is to
eliminate this exclusion in its entirety. As the rule amendments were only to phase out this
exclusion (eliminating it within 3 years of publication of the final rule), we do not consider it
necessary to define this term at this time, which will be irrelevant after the removal of the
exemption. We did specifically request comment on issues as to why Group 1 sized analyzer
vents would not be amendable to control and we received no comments suggesting such control
was not feasible. Therefore, we are finalizing the elimination of this exclusion as proposed.
4.2 Revision of the definition of periodically discharged
Comment 1: Several commenters stated EPA's proposed definition of "periodically discharged"
is broad and would result in all gas stream containing greater than 20 ppmv organic HAP that are
continuously or periodically discharged from a petroleum refining process unit to be considered
MPV. These commenters argued that this is a significant expansion of the universe of MPVs.
The commenters stated that in order to ensure this change does not apply retroactively or
immediately on promulgation of the final rule, the EPA must expressly provide the specific
effective date. The commenters specifically identified MPV exceptions 1, 2, 4, and 11 as
proposed for modification without an applicability date. The commenters recommended adding
applicability dates for each addition to the exception language and maintaining the old language
and adding a deactivation date for that language in each entry.
Response 1: Owners or operators of affected facilities must meet the requirement that exist at
the time the venting event occurs. Exception 1 of the MPV definition was proposed with an
effective date 3 years after the date of publication to allow refineries to upgrade their flare
systems, as needed. Exception 2 of the MPV definition is in the existing regulation; EPA's
proposed revision was not intended to change the substance but rather to provide additional
clarity regarding why these discharges are not MPV (i.e., because they are regulated under 40
CFR 63.648). Since we did not substantively change Exception 2, we intended (and are
finalizing) that it will be effective as of the effective date. Exception 4 of the MPV definition
(episodic or non-routine releases) was proposed to be removed as of the effective date. The
rationale for this was that these periodic events could be planned and, if the planned event would
trigger the Group 1 MPV thresholds, then the plans could include the control of these larger vent
streams. Based on other comments received, we have included special provisions for vents
associated with process unit startup or shutdown to reflect best industry practices for these vents.
Those provisions provide that the requirements apply as of the effective date for the reasons
provided in the preamble to the final rule. Exception 11 of the MPV definition was proposed to
fully exclude vents associated with decoking operations and is broader than the previous
144
-------
exception, although equivalent in practice. This provision will be effective upon the effective
date of the final rule.
4.3 Bypass line provisions
Comment 1: One commenter stated that most flows through vent system bypass lines are not
violations and there is no basis for arbitrarily declaring that they are by prohibiting all bypass
flows as proposed 40 CFR 63.644(c) and 63.660(i)(2). The commenter stated that this would
make refineries inoperable.
The commenter explained that in most cases such bypasses are done purposely (e.g., as part of a
startup or shutdown, control device maintenance, or for some other process purpose) and the
bypassed material is routed to a process, a fuel gas system or an alternative control device. The
commenter claimed that while these flows will be detected by the current flow indicators and the
proposed flow monitoring, removal of a car seal or opening of a lock, no atmospheric release
occurs in such cases and flow through these connections is not an indicator of an uncontrolled
release to the atmosphere and certainly is not a violation. Thus, the commenter concluded that
the proposal to declare any flow through a bypass a violation is arbitrary and unsupportable and
that proposed language must not be finalized.
The commenter also stated even when there is flow to the atmosphere through a bypass, there is
not necessarily a violation. The commenter provided an example of when a stream containing
mostly nitrogen would be bypassed around a combustion control device to avoid upsetting its
operation. Alternative controls in these circumstances may not be feasible or justifiable for the
minimal HAP release potential. The commenter provided another example when Group 1 vents
may not be going to the vent system at the time other gases in the vent header are bypassed or the
flow through the bypass is a start-up or shutdown stream that does not occur during normal
operation and is not a Group 1 stream. This situation can readily occur for intermittent Group 1
vents and during maintenance, startup, and shutdown situations.
Further, the commenter added that operators retain records as required in 63.655(g)(6)(iii) of
potential bypasses to the atmosphere, and these must be included in the Periodic Report. The
commenter argued that the EPA has the necessary information to determine if a violation has
occurred rather than making them a violation by rule.
The commenter suggested the wording of 40 CFR 63.644(c) and 63.660(i)(2) be revised to apply
to bypasses that route Group 1 streams to the atmosphere or to a control that does not meet the
applicable control requirements of Refinery MACT 1 and are not exempted under equipment
leak work practice to clarify applicability. The commenter also stated that bypasses to another
process or to fuel gas should be specifically excluded from the requirements in 40 CFR 63.644(c)
and 63.660(i)(2).
Response 1: We did not intend to include lines that divert gases from one control system to
another (such as diverting flow from an incinerator to a flare) to be a bypass line. Thus, we are
revising 40 CFR 63.644(c) to clarify that a bypass line is one that diverts a vent stream away
145
-------
from the control device used to comply with paragraph (a) of this section and vents the stream
either to the atmosphere or to a control device that does not comply with 40 CFR 63.643(a).
In several of their comments, it appears that the commenters are suggesting that there may be
times that there is flow in the bypass line that is "associated with a Group 1 MPV," but that a
specific flow event may in that line may not be flow of a Group 1 MPV stream. We note that the
proposed language at 40 CFR 63.644(c) that the "[u]se of the bypass line at any time to divert a
Group 1 miscellaneous process stream is an emissions standards violation" [emphasis added].
The proposed language did not say that the use of the bypass line is always an emissions
standards violation, but only when that use specifically diverts a Group 1 MPV stream from the
required control. We have clarified the recordkeeping requirements to include records to
demonstrate whether flow detected in the bypass line includes flow from the Group 1 MPV. We
have also clarified the reporting and recordkeeping requirements regarding the quantity of HAP
released is specific only to those periods when the bypass line is used to divert a Group 1 MPV
stream to the atmosphere or to a control device that does not comply with the requirements in
§63.643(a). If the commenters are suggesting that Group 1 MPV streams may be directed to the
atmosphere without a violation of the standard, we disagree. We maintain that any diversion of a
vent stream meeting the definition of a Group 1 MPV stream (i.e., the thresholds for HAP
concentration and mass emissions) either to the atmosphere or to a control device that does not
comply with 63.643(a) is a violation of the MPV emissions standards.
We also note that we have established special provisions for maintenance MPV associated with
startup and shutdown events, which are expected to address many of the issues raised by the
commenter. Please see the preamble to the final rule for further details of the maintenance MPV
provisions.
Comment 2: One commenter asserted that prohibiting bypasses and, thereby, declaring all
bypasses to be violations complicates the management of any unavoidable bypass and required
bypasses associated with maintenance, startup, and shutdown. The commenter argued that
control is not always feasible, because of the properties of the stream or the control device and if
the EPA believes there are feasible controls available for bypass streams, a CAA justification for
imposing them must be provided and cost estimates and emission impact estimates to install such
control must be provided for comment.
Response 2: As described in more detail in the preamble, we are finalizing requirements for
maintenance vents (vents only used as a result of startup, shutdown, maintenance, or inspection),
which we expect will typically address the concerns expressed by the commenter. We note that
maintenance MPV may appear to be a bypass line if opened during normal operations; however,
as long as the vent line is opened only for startup, shutdown, maintenance, or inspection, these
vents can be classified as a maintenance MPV. This does not eliminate the need to monitor the
vent line for flow if the vent line "could" act as a bypass line for a Group 1 MPV stream;
however, special provisions apply to these vents when they are only used for startup, shutdown
or maintenance. If the bypass line may be used for other purposes, including routine venting or
emergency venting due to equipment or control device malfunctions, then the maintenance
provisions are not applicable, and the bypass provisions as described in the previous comment
response applies and it is a violation of the emissions standards to divert a Group 1 MPV stream
146
-------
to the atmosphere or a control system that does not comply with the requirements in §63.643(a).
Based on the final rule requirement, we consider the MPV and bypass provisions can be
achieved at all times.
Comment 3: One commenter stated that there are significant legal, technical, and cost issues
associated with installing flow CPMS in bypass lines. The commenter added that flow
monitoring was available in the initial Refinery MACT 1 rulemaking and does not represent new
technology and thus is not authorized under section 112(d)(6). There is no evidence presented in
the record to demonstrate this change reduces emissions or increases compliance assurance nor is
there any indication that the costs and burdens of replacing the current compliance monitoring
has been tabulated, published for comment, and justified. The commenter also argued that flow
monitors offer no advantage to a flow indicator in alerting the operator of a release. The
commenter concluded by stating no such change is proposed for other potential bypasses and
there is no legal basis for this proposed change.
The commenter asserted that bypasses to the atmosphere are not a significant problem and
reporting under federal and state rules and facility permits is already required. Any bypass to the
atmosphere of a Group 1 stream, as indicated by the flow indicator required under the existing
Refinery MACT 1 for any potential closed-vent system (CVS) bypass, already has to be recorded
and reported (including an emissions estimate) under Refinery MACT 1 requirements, under
permit reporting requirements, and under a variety of federal and state release reporting
requirements. The commenter argued that the proposed requirement adds no new information
while significantly increasing the instrumentation costs and burdens. The commenter added that
the EPA provided no data that indicate unplanned bypasses to the atmosphere are frequent or
common.
The commenter stated that there are feasibility and cost issues with requiring flow monitors
instead of flow indicators. The commenter specifically stated that one criterion is to meet the
specified accuracy requirement at normal flow which they argued is zero, but the bypass flow
may be high. If the meter is set for no or low flow, it may serve as a flow indicator, and if it is set
for the bypass flow, it will miss low flows.
The commenter also discussed the QA/QC requirements in Table 13 and stated that monthly
and quarterly frequency is nonsensical if there is no flow. Additionally, the commenter stated
that an annual performance check is required, but could not be completed without having to
bypass material which would result in a violation. The commenter also noted that meeting the
swirling flow or abnormal velocity distribution requirements will require revisions of bypass
lines to install long straight meter runs. Such piping revisions come at significant cost in the
typically congested areas where these bypasses are typically located.
Conversely, another commenter supported the requirement for flow monitoring through bypass
lines, noting that such monitoring will discourage facilities from manually bypassing flares to
disguise potential activation of pressure relieving devices and give the appearance of having less
HAP released than is actually leaving the process.
147
-------
Response 3: The issue of whether the use of bypass is frequent or common is not pertinent to a
determination that when such an event occurs it is a violation of the emissions standards. Our
proposal was for the purpose of ensuring, through monitoring, that the applicable emissions
standards are continuous and are not circumvented by a bypass of the control device.
We considered all of the comments that we received on this issue and we re-evaluated the need
to use a quantitative CPMS for flow in bypass lines. Based on that re-analysis, we have
determined that the use of a flow indicator along with engineering estimates and process
knowledge are sufficient for the purpose of estimating the magnitude of the release without
having to install new quantitative flow monitoring systems. Therefore, we are not finalizing the
proposed requirements to install flow CPMS, and are, instead, retaining the existing
requirements to have flow indicator monitoring systems. We are finalizing, similar to the
proposal, recordkeeping and reporting requirements for bypass lines to require an estimate of
emissions (previously only the duration of the flow was required to be reported, not an estimate
of the mass emissions as suggested by the commenter). We are also finalizing clarifications that
these estimates can be made using instrument readings, engineering calculations, and/or process
knowledge.
Comment 4: One commenter stated that defining a bypass in 40 CFR 63.644(c) as including
occurrences of air intrusion into a control device is new, unjustified, technically flawed,
ambiguous, and should be deleted. The commenter explained that this has never been considered
a bypass in the past and does not involve emissions to the atmosphere. The commenter added
that air is added to many combustion control devices to allow combustion to occur (i.e., air-
assisted flares) and some regulated vent streams include air (i.e., some tank vapor spaces and
waste treatment unit vapors), and these would presumably now be a disallowed with this
proposed prohibition.
The commenter also stated that the term "bypass" is ambiguous since it would mean that a
stream with any amount of air intrusion would be considered a bypass. The commenter provided
an example of minute, unpreventable leakage into a vacuum system through fittings as a stream
which could be construed as a bypass under the proposed definition. The commenter concluded
by asserting that a bypass under Refinery MACT 1 should be limited to releases of Group 1
streams to the atmosphere without their passing through a compliant control device, being routed
to a process (including fuel gas process), or being allowed under special circumstances, such as
during identified maintenance, startup, and shutdown scenarios, and all proposed language
associated with air intrusion in this proposal must be deleted.
Response 4: We did not intend to define a bypass as including occurrences of air intrusion and
the language in 40 CFR 63.644(c) does not include any language that suggests that a bypass
would include occurrences of air intrusion. However, we did inadvertently include
a recordkeeping requirement for bypass lines regarding air intrusion and we are not
including that reference to air intrusion in 40 CFR 63.655(i)(4) in the final rule.
Comment 5: One commenter stated that the prohibition on routing MPV and storage vessel
bypasses to the atmosphere will require installation of controls, where they are feasible. The
commenter explained that the installation of monitoring equipment will require new flow meter
148
-------
installations as well as piping revisions wherever flow indicators are currently used for MPV
bypass monitoring to meet the Table 13 requirements for these new monitors. The commenter
stated that at least 3 years will be required to implement controls and monitoring equipment
required by the proposal. The commenter also suggested that applicability dates should be
provided in 40 CFR 63.644(c) to indicate when the flow monitor requirements take effect and
when the flow indicator requirements are no longer applicable.
Response 5: We have included specific startup and shutdown provisions for maintenance MPV
that we anticipate will eliminate the need to install new control devices. Also, we are not
finalizing the requirements to install flow CPMS for bypass lines, so no additional time is
needed. Based on the provision included in the final rule as described in the preamble and in the
responses to the other comments in this section, we have determined that no changes in
compliance dates are necessary.
Comment 6: With regard to the proposal to remove the exception from 40 CFR 63.644(c)
monitoring requirements for PRD on MPV vapor collection systems, two commenters expressed
concerns over closing or blocking (even by a flow meter) relief paths which they stated could
cause potentially catastrophic events. One of these commenters added that the EPA has proposed
40 CFR 63.648(j) to address PRD and thus the exception for equipment subject to 40 CFR
63.648 might be adequate for excluding PRD from 40 CFR 63.644(c). However, 3 years is
provided for the PRD monitoring requirements in 40 CFR 63.648(j) to be met and there may be a
question as to whether the 40 CFR 63.648 exception applies prior to that PRD compliance date.
Furthermore, controlled PRD (e.g., PRD routed to a flare) are excepted from 40 CFR
63.648(j)(l)-(3) and thus there may be a question as to whether the 40 CFR 63.648 exception
applies to controlled PRD. Since the impact of misunderstanding the applicability of 40 CFR
63.644(c) to this critical safety equipment is dire and the current rule calls them out separately
from 40 CFR 63.648 regulated equipment, the commenter strongly suggested that the separate 40
CFR 63.644(c) exception for PRD be maintained.
Response 6: We agree with the commenter that the phrase "regulated under §63.648" provides
some ambiguity regarding whether certain PRD are MPV or exempt because they are regulated
as equipment components in 40 CFR 63.648. In our proposal, we intended that PRD would be
subject to the requirements in 40 CFR 63.648 and we consider PRD vented to a flare or other
control device to be subject to the requirements in 40 CFR 63.648(j). In the final rule, we are
proposing additional work practice requirements for PRD under 40 CFR 63.648 which will likely
further confuse the issue (with respect to timing of certain requirements). Therefore, we are
maintaining the general exclusion for PRD; we are simply rephrasing item 2 in the definition of
MPV to exclude "Pressure relief device discharges" rather than "Relief valve discharges" to
more clearly note that this exclusion applies to PRD (including rupture discs) and is not limited
to "valves"; this revision also provides more consistency with the terms used in final rule at 40
CFR 63.648(j).
Comment 7: One commenter requested that the EPA add an option for a manual block valve
equipped with a valve position indicator, to be used in lieu of a flow indicator, under 40 CFR
63.644(c). Associated with the use of a manual block valve, the EPA could require continuous
149
-------
monitoring of the indicator and reporting of any diversions to the atmosphere. The arrangement
would be similar to a lock and key or car-sealed valve (i .e. manually operated block valve).
Response 7: We agree that a manual bock valve with a valve position indicator is a type of flow
indicator and have provided language in §63.644(c)(1) to allow this as an option.
Comment 8: One commenter stated the proposed 40 CFR 63.644(c)(l)(i) and Table 10 revision
requires that the flow monitor "record the volume of the gas stream that bypassed the control
device". However, flow monitors measure flow rates, not volumes, and, thus, this wording is
imprecise. The commenter suggested that the wording should be that the monitor "record the
flow rate of the gas stream" and that the estimated volume be calculated from the flow rate data.
Response 8: We are not finalizing the flow CPMS provisions, and, thus, we are not revising the
provisions in Table 10 concerning bypass lines. However, we are requiring refinery owners or
operators to estimate the volume of gas and mass of HAP diverted to the atmosphere via the
bypass lines.
4.4 Monitoring requirements
Comment 1: One commenter stated that the EPA failed to explain or justify the proposed
amendment to the second sentence in section 63.644(a). The commenter noted there is no change
in instrumentation proposed for monitoring combustion controls other than flares (in sections
63.670 and 63.671) and no demonstration that there are compliance assurance problems that
justify the large additional burdens imposed by Table 13 or that justify the change of the existing
monitoring instrumentation to meet the new Table 13 accuracy requirements. The commenter
continued by stating there has been no change in monitoring instrumentation that might authorize
these revisions under section 112(d)(6), and this change is not reflected in the cost and burden
analyses for this rulemaking.
Response 1: We disagree with the commenter's assertion that the Table 13 requirements are
arbitrary. The existing rule requires monitoring equipment be installed, calibrated, maintained,
and operated according to manufacturer's specifications or other written procedures to ensure
that the equipment will monitor accurately. In order to make this requirement clearer, ensure the
accuracy of monitoring equipment, and provide framework for enforceability, we have proposed
and are finalizing reasonable and achievable calibration and quality control requirements in
Table 13.
Comment 2: One commenter stated that flare tips cannot be safely accessed when the flare is in
service. The commenter argued that the proposed revisions to 40 CFR 63.644(a)(2) put people at
great risk, because they require accessing the flare tip frequently, and the commenter urged EPA
to retain the current language.
Response 2: We revised 40 CFR 63.644(a)(2) to transition the requirements for flares used for
MPV to meet the new flare monitoring and operating requirements in 40 CFR 63.670. We
consider these revisions necessary because we found that the current monitoring requirements
(which only includes pilot flame monitoring) is insufficient to ensure that flares are achieving 98
150
-------
percent control efficiency. For these reasons, we disagree that revisions to 40 CFR 63.644(a)(2)
are not necessary.
We expect that the commenter is concerned with the application of Table 13 to monitoring
systems to detect the presence of a flare or to pilot flame monitoring requirements in 40 CFR
63.670, similar to the comments summarized in Section 3.2.2 of this document. As noted in the
responses in that section, we do not consider monitors used to detect the presence of a flame to
be a temperature CPMS or any other type of CPMS listed in Table 13, and we have revised 40
CFR 63.671(a)(1) to specifically exclude pilot flame monitoring systems from the QA/QC
requirements in Table 13. The existing language used in 40 CFR 63.644(a)(2) for flare pilot
monitoring is to "monitor the presence of the pilot flame(s) using a device (including, but not
limited to, a thermocouple, ultraviolet beam sensor, or infrared sensor) capable of detecting that
the pilot flame(s) is present." While we allow the use of a thermocouple, we do not refer to it as a
temperature monitor, and we clearly do not limit the devices to thermocouples or other direct
temperature monitoring systems. As noted in previous responses in Section 3.2.2, we do not
consider it necessary that monitors used to detect the presence of a flame require the same
accuracy as a monitor used to determine a temperature used to determine compliance with a
specific temperature operating limit. While we maintain that there are no direct CPMS
monitoring requirements in Table 13 for CPMS used to "detect the presence of a pilot flame," we
are revising 40 CFR 63.644(a) to include the phrase "except for CPMS installed for pilot flame
monitoring" with regards to requiring monitors to meet the requirements in Table 13, similar to
the revisions we included in revised 40 CFR 63.671(a)(1).
4.5 Recordkeeping and reporting requirements
Comment 1: One commenter stated that the proposed 40 CFR 63.655(g)(6)(iii), 63.655(i)(4)(ii),
and Table 10 should all be revised to remove references to air intrusion events. Also, the
commenter added that most flows through bypasses do not involve regulated materials or
releases to the atmosphere and there is no reason to require records or reports unless regulated
materials are present and are not routed to a process, fuel gas system or compliant control device.
The commenter noted that the proposed 40 CFR 63.655(i)(4) contains the recordkeeping
provisions associated with the proposed bypass monitoring requirements. The introductory
paragraph appears to apply the recordkeeping requirements to all closed vent system bypasses,
not just those regulated by 40 CFR 63.644(c) and to include drains and vents, which are
specifically excluded from monitoring by those paragraphs. Thus, the commenter recommended
section 63.655(i)(4) be revised as follows.
(4) For each closed vent system that contains bypass lines that are required to be monitored
by section 63.644(c)could divert a vont stream away from the control device and to the
atmosphere, or cause air intrusion into the control device, the owner or operator shall keep a
record of the information specified in either paragraph (i)(4)(i) or (ii) of this section when
Groupl streams are present as applicable.
Response 1: We agree that the references to air intrusion in these sections and Table 10 should
be removed. The emission limitations in Refinery MACT 1 are not susceptible to air dilution so
these provisions are not needed. We have revised the Group 1 MPV requirements, and there may
151
-------
be flow in a line that may serve as a bypass line during normal operations, but may have
permissible flow during certain periods, such as prior to equipment openings. Therefore, we are
revising the reporting requirements to limit the reporting for bypass lines to instances when flow
of "regulated material" (i.e., Group 1 MPV stream requiring control) is discharged to the
atmosphere. Records must still be kept of instances of flow in the bypass line as well as records
needed to document that the flow was not "regulated material."
5.0 Storage Vessel Provisions
5.1 Technology review results
Comment 1: One commenter asserted that the EPA's proposed enhanced tank requirements
should be based only on the EPA's section 112(d)(6) authority, not on section 112(f). The
commenter explained that in the proposal, the EPA conducted a technology review under CAA
section 112(d)(6) to "identify the latest developments in practices, processes, and control
technologies for storage vessels." 79 Fed. Reg. at 36914. The EPA evaluated three options for
requiring additional tank controls. Based on its cost-effectiveness analysis, the EPA proposed
adopting option two, which included the controls in option one. Id. at 36915. The EPA also
evaluated these three options in its section 112(f) risk assessment. The EPA determined that the
controls in option two also "are necessary to provide an ample margin of safety to protect public
health." 79 Fed. Reg. at 36940
The commenter argued that the EPA's approach in this roposal is contrary to the statute in that
sections 112(d)(6) and 112(f)(2) address fundamentally different issues using fundamentally
different factors. The commenter asserted that the agency's obligation under section 112(d)(6) is
separate from the EPA's duty to evaluate the health risks of hazardous air pollutant ("HAP")
emissions from a source category on a one-time basis through section 112(f)(2). The commenter
suggests that if Congress had intended the two review processes (which address wholly different
concerns) to be dependent, it could have required the EPA to conduct them concurrently or
otherwise relate them. In recent years, the EPA has chosen to conduct the two reviews
simultaneously as a matter of convenience, but that approach is not required by the statute. The
EPA could conduct them in separate proceedings at separate times. The commenter argued that
the EPA's analysis has improperly conflated these separate provisions and argues that addressing
112(d)(6) and 112(f)(2) in the same rulemaking would deprive section 112(d)(6) of any meaning.
Another commenter argued that EPA's approach in this rulemaking is inconsistent with and
contravenes Congressional intent.
Response 1: These amendments were appropriately implemented under the authorities of both
sections 112(d) and 112(f) of the CAA. There is nothing in the CAA that precludes us from
performing the duties under section 112(d) and under section 112(f) within the same rulemaking.
We performed a technology review and risk review, separately, and found that additional
controls were appropriate for storage vessels under the authority of 112 (f) (2) because the
additional tank controls were found to be cost effective and would reduce cancer incidence and
the number of people exposed to risks of greater than 1-in-l million, and therefore we are
finalizing our decision that these controls are necessary to provide an ample margin of safety.
Because we also found these controls to be cost effective, we proposed and are finalizing our
152
-------
decision to require tank controls under 112 (d)(6) as well. We note that for other emission
sources, we proposed and are finalizing additional controls under the authority of 112 (d)(6) only
as they were not found to reduce risk but were found to be cost effective. Therefore we disagree
that section 112 (d)(6) has no meaning independently of section 112 (f)(2).
Comment 2: One commenter generally supported the EPA's proposal to require better controls
for external and internal floating roof tanks and expand the applicability of the standards to
smaller storage vessels and/or those with lower vapor pressure due to pollution control
developments that have occurred and to assure an ample margin of safety to protect public
health. However, the commenter asserted that the EPA also has no lawful basis to not require the
following improvements based on costs: degassing controls; retrofit external floating roof tanks;
additional monitoring of tanks using the EPA Method 21 or optical gas imaging; and warning
monitors for liquid overfill and roof landings. The commenter asserted that the statute contains
no authorization to place cost above the statutory objectives of section 112(d) or 112(f). The
commenter further supported its assertions by citing examples of existing SCAQMD
requirements for external floating roof any storage vessel at a refinery having a capacity larger
than 75 cubic meters for materials having a vapor pressure greater than 3.0 pounds per square
inch (psi) and existing Texas and the SCAQMD requirements for refineries to control emissions
from tanks when they are emptied and degassed for any reason.
The commenter contended that the EPA should revise its cost analysis of the rejected control
options on the most recent data from the petroleum ICR. The commenter asserted that the EPA
failed to provide any reasoned analysis for how it estimated the resulting VOC and HAP
reductions from augmented monitoring and warning control. The commenter requested
clarification on how EPA derived potential emission reduction in the conclusory estimate of
VOC and HAP possible reductions for implementing these controls in the memorandum to the
docket "Impacts for Control Options for Storage Vessels at Petroleum Refineries" (EPA-HQ-
OAR-2010-0682-0199). The commenter also contended that there is significant evidence that
leaking storage vessels are significant sources of fugitive HAP emissions and that emissions
from tanks are significantly underestimated, citing the report to the docket "Critical Review of
DIAL Emission Test Data for BP Petroleum Refinery in Texas City, Texas" (EPA-HQ-OAR-
2010-0682-0070). The commenter requested that the EPA include these potential emissions
reduction in its analysis of requiring augmented monitoring requirements.
The commenter also requested that the EPA reevaluate the benefits of requiring degassing
controls and the conversion of external floating roof tanks into internal floating roof tanks based
on the actual vapor pressures of the stored material as reported in the ICR, citing that earlier
analysis of control options assumed that that the vapor pressure of the material stored by external
floating roof tanks is significantly lower than what was actually reported to the ICR.
Response 2: The EPA appreciates the general support to require better controls for storage
vessels at refineries. Regarding the commenters assertion that the EPA has no lawful basis to
consider costs of controls, section 112(d)(2) of the CAA states that we shall take cost into
consideration in establishing emission standards under that section of the Act. If the application
of cost effective controls under section 112(d) results in unacceptable residual risk, then any
standards promulgated to mitigate such unacceptable risk under section 112(f) would not take
153
-------
costs into consideration. Because the proposed standards for storage tanks based on technology
review left no unacceptable risk, we had no rationale for requiring controls are that not cost
effective for storage vessels at refineries. Regarding the emission reduction estimates used,
"Impacts for Control Options for Storage Vessels at Petroleum Refineries" states that it was
assumed that the additional monitoring requirements would reduce emissions by 10 percent. We
have not received any comments providing evidence of an alternative estimate. Any estimate
would involve gross assumptions about the number of avoided floating roof landings and the
occurrences of tank control equipment failures or improper operations that would be detected by
Method 21 or optical sensors.
Regarding fugitive emissions from leaking storage vessels, to the extent that benzene emissions
are present in such, these types of emissions are expected to be detected in the fenceline
monitoring requirements promulgated herein. Facilities will have an incentive to find the cause
of higher than normal concentrations (including leaking storage vessels) and fix the cause of the
higher than normal readings before an action level is triggered.
Regarding degassing controls and retrofitting fixed roofs on tanks that are currently equipped
with external floating roofs, we considered these controls in developing the proposal, but were
not included in the proposal because we considered them not to be cost effective. Because there
was no unacceptable residual risk after application of cost effective controls, these controls were
not needed to mitigate unacceptable risk. The rationale for this consideration is expressed in the
document "Impacts for Control Options for Storage Vessels at Petroleum Refineries." Regarding
the assertion that the ICR reported vapor pressures are higher than the vapor pressures that we
used to assess the cost effectiveness of degassing and retrofitting fixed roofs onto existing
external floating roofs, we note that we used annual average vapor pressure values, which we
consider appropriate when developing nationwide annual emissions estimates. The vapor
pressures reported in the ICR are maximum true vapor pressures and not annual averages
(footnote: Table 4-1 of
https://refmervicr.rti.Org/Portals/0/Petroleum Refinery ICR Component l.pdf). so the ICR
values are expected to be higher than those used in emissions impact analysis for various stored
liquids.
Comment 3: One commenter stated that the existing Refinery MACT 1 storage vessel
provisions in section 63.119 through 63.121, where the requirements for use of a closed vent
system and control device are specified in 40 CFR 63.119(e). Paragraph 40 CFR 63.119(e)
includes 1) a provision to grandfather control devices that were installed on or before December
31, 1992 and achieve at least 90% effectiveness, and 2) a provision that addresses planned
routine maintenance. The commenter explained that the proposed change in 40 CFR 63.660 to
invoke the requirements of subpart SS does not include these provisions from 40 CFR
63.119(e)88, and no rationale or analysis is presented to justify removal of these provisions and
no compliance time is provided if a grandfathered control device must be upgraded. If the
grandfathering provision specified in 40 CFR 63.119(e)(2) is not preserved in the revisions to the
88 Subpart SS contains the recordkeeping and reporting requirements associated with the planned routine
maintenance provisions, but relies on the referencing subpart to contain the planned routine maintenance provision
itself.
154
-------
rule, then a control device that has been in compliance with this provision will need to be either
upgraded or replaced. Consistent with CAA section 112(i), 3 years should be allowed for
upgrade or replacement of an existing control device. The provisions for planned routine
maintenance in 40 CFR 63.119(e)(3) and (e)(4) are needed to allow for the routine maintenance
necessary in order to maintain control devices (and in some cases closed vent systems) in good
working order. These paragraphs from 40 CFR 63.119(e) should be carried over into the new
requirements of 40 CFR 63.660(i).
The commenter suggested specific revisions to proposed 63.660(c) to maintain grandfather and
the planned routine maintenance provision provisions in their comment on pages 204-205.
Response 3: EPA disagrees that a grandfathering provision for closed vent control devices that
were installed on a storage vessel on or before December 31, 1992 should be added to §63.660.
By the promulgation date of this RTR rulemaking, storage vessels with controls installed prior to
December 31, 1992 will have had 23 years to upgrade controls, and thus we maintain that ample
time has been provided. EPA did not, however, intend to eliminate the provisions for planned
routine maintenance. The routine maintenance provision was originally established in the HON
(see 40 CFR 63.119(e)(3)-(4); 57 FR 62710, December 31, 1992 (proposed); 59 FR 19402, April
22, 1994 (final)) for facilities that elected to use a closed vent system and control device to
comply with the emission limitation requirements for tanks. Consistent with other recent
rulemaking packages, including the NESHAP for Off-site Waste and Recovery Operations
(OSWRO) codified in subpart DD of part 63, we intended to include the routine maintenance
provision in the HON for tanks routing emissions to control devices. We are including these
provisions in the final rule because the estimated HAP emissions to degas the tank would be
greater than the emissions that would result if the tank emitted directly to the atmosphere for a
short period of time during routine maintenance of the control device.
5.1.1 Other controls (degassing, geodesic domes, roof landings)
Comment 1: Many commenters generally supported the rule provisions for storage tanks. The
commenters did provide additional suggestions or clarifications for the proposed rule regarding
emissions detection technology, stormwater run-off, primary seals, regulating more categories of
tanks, and the grandfathering exemption.
One of these commenters added that the EPA should consider the use of newer emissions
detection technology including Fourier Transform Infrared Spectroscopy (FTIR), Solar
Occultation Flux (SOF) Time Correlation Tracer, Vertical Radial Plume Mapping, and DIAL.
Another commenter added that the EPA should also require monitoring at the bottom of the tank
to capture all evaporate leakage.
One commenter also asked how evaporative emissions from rain accumulation and run-off from
the roofs of the storage tanks would be addressed in this rulemaking.
This commenter also asked the EPA to clarify which primary seal (liquid mounted foam or
mechanical shoe) is more protective of human health.
155
-------
Another commenter added that the EPA should upgrade roof deck fitting controls by adding
requirements to cover roof openings, sleeves, and wipers for guide poles as well as expand the
regulation to tanks that are in excess of 20,000 gallon capacity and exceed 1.9 psi or are in
excess of 40,000 gallon capacity and exceed 0.75 psi. Two commenters recommended that all
tanks be required to be covered by domes to prevent leakage.
One commenter added that the BAAQMD and SCAQMD regulations in California governing
organic liquid storage tanks are at least as stringent as those proposed in this rulemaking and, in
some cases, require emission controls at even lower product vapor pressures. The
commenter supported the requirements in California and stated that the EPA should adopt the
SCAQMD requirement for all refinery tanks to have internal floating roofs.
Two commenters stated that the grandfathering exemption for tanks in subpart WW of part 63
should be removed from the proposed rule.
Response 1: The EPA appreciates the support for enhanced control requirements for storage
vessels and the additional suggestions.
Regarding the suggested detection methodologies, the EPA did consider these before proposing
these amendments. The fenceline monitoring provisions in this final rule specify a technology
that is highly effective at detecting emissions of benzene from storage vessels at concentrations
expected to be present at refinery property boundaries. In the fenceline monitoring section of this
response to comment document, please see responses to comments about the possible use of
alternative detection technologies in the fenceline monitoring provisions of this rule.
Regarding emissions from rain accumulation and run-off from the roofs, any product that
becomes entrained in rain water drains is required to be addressed in the wastewater provisions
of this rule.
Regarding what type of primary seal is most effective, theoretically, according to equations and
rim-seal loss factors in AP-42, all else being equal, a storage tank with a liquid mounted primary
seal would have less rim seal loss emissions than one with a mechanical shoe primary seal. This
is the case for either a primary only seal or if a rim mounted secondary seal is also present.
However, another important factor in the emissions is how well the primary seal is maintained.
Proper maintenance of rim seals are at least as important as the type of primary seal on a storage
vessel. There is anecdotal evidence that the service history of mechanical shoe seals are better
than liquid mounted seals.
Regarding the recommendations to upgrade desk fitting controls on certain tanks and to require
internal floating roofs on all storage vessels at refineries, the EPA did examine enhanced controls
when it developed the proposal. Based on the technology review the EPA performed, additional
storage tank controls beyond what are provided in the final review would not be cost effective.
156
-------
5.1.2 Revision of definition of Group 1 storage vessels
Comment 1: One commenter stated that small storage vessels should be included in the storage
vessel definition if a pressure exception is not added to the proposed atmospheric PRD
requirements. The commenter explained that the definition of a storage vessel in Refinery
MACT 1 contains an exclusion for "[vjessels with capacities smaller than 40 cubic meters." This
exclusion has not been an issue in the past. Under this proposal, however, the commenter argued
that the pressure/vacuum (P/V) vent on these small tanks would be miscellaneous process vents
or, if the tank is controlled with a closed system, relief valves and would be subject to evaluation
under those sections and potentially control. Thus, these smaller vessels, which are permitted as
tanks and were considered tanks in the development of Refinery MACT 1, should be clearly
identified as storage vessels in Refinery MACT 1.
Response 1: The final rule now provides exemptions for PRDs in §63.648(j)(5), including relief
devices with a designed set relief pressure of less than 2.5 psig, from complying with the
pressure release management requirements in §63.648(j)(3). We expect most P/V vents on
storage vessels will meet this criteria and, therefore, will not be subject to the additional
monitoring, prevention measures, and RC/CA requirements finalized for atmospheric PRDs.
Comment 2: One commenter stated that the bypass monitoring requirement in 63.660(i)(2) need
to be clarified to match those applicable to potential miscellaneous process vent bypasses.
Proposed 63.660(i)(2) includes language that makes any flow through a Group 1 storage vessel
vent system potential bypass a violation. The commenter referred the EPA to the comments
made on the MPV bypass provisions as they are similar in nature.
Response 2: We maintain that bypassing the emissions controls of a CVS for a storage vessel to
the atmosphere is a violation of the standards of this final rule. The final rule requires the
installation, maintenance and operation of a flow indicator for any bypass line that could divert
the Group 1 storage vessel vent stream to the atmosphere or to a control device that does not
comply with the requirements in subpart SS. We made minor revisions to the proposed language
to clarify that a bypass line that diverts the vent gas to an alternate control system meeting the
requirements in subpart SS is an allowable bypass and would not be a violation of the final
standards. Please see our responses to comments of the MPV bypasses in this document which
are consistent with this response.
Comment 3: One commenter stated that emissions associated with vessel degassing could not be
practically controlled to any specific MACT-level standard under numerous operating scenarios,
including the degassing of storage tanks with significant solids in the bottom; continuous control,
which would otherwise be required to satisfy such standard, could not be applied during solids
removal from the tank. Further, the commenter argued extending the nitrogen purging would
adversely affect flare gas recovery operation, likely resulting in flare gas recovery bypass. The
commenter asserted that the rule should expressly exclude from all MACT standards emissions
associated with tank venting and degassing, regardless of any contention that any such emissions
would be related to SSM activity.
157
-------
Response 3: For storage vessels, the requirements that apply during normal operations also
apply during startup and shutdown. We evaluated degassing controls as a control alternative
for Group 1 storage vessels that use floating roofs (i.e., that do not already route emissions to
a control device) and do not consider the installation and use of controls during the degassing
of such storage vessels to be cost effective (see memorandum Survey of Control Technology
for Storage Vessels and Analysis of Impacts for Storage Vessel Control Options, Docket Item
Number EPA-HQ-OAR-2010-0871-0027). Based on this review, neither did we propose, nor
are we finalizing any additional standards for floating roof storage vessels during startup or
shutdown. However, if a control device is used to comply with this final rule during normal
operations, then such a control device must be used at all times, including during degassing of
the storage vessel. Any bypassing of emissions from being routed to a control device to being
routed to the atmosphere would be considered a violation of the standard.
Comment 4: One commenter is concerned that the vapor pressure of 0.75 psia or greater in the
proposed revision of the definition of Group 1 storage vessels at an existing source for storage
tanks greater than 40,000 gallons may not be low enough to achieve the gains sought. The
commenter based their discussion off of results from the DIAL study conducted by the Houston
Department of Health and Human Services through its Bureau of Pollution Control and
Prevention funded by an EPA grant. The commenter stated that the DIAL report noted that
measured emissions from process areas and storage tanks exceeded the emission factor estimates
for benzene and VOC. Some of the surveyed tanks had true vapor pressures ranging from 0.00 to
0.56 psia, and thus would not be subject to the Group 1 control requirements under the proposed
regulation because their true vapor pressure is less than 0.75 psia.
Response 4: In the study cited by the commenter, direct measurement of the composition or
vapor pressure of the materials stored in these tanks was not performed. As noted in "EPA
Review of Available Documents and Rationale in Support of Final Emissions Factors and
Negative Determinations for Flares, Tanks, and Wastewater Treatment Systems," cutting fluids
are often used for heavy liquid materials that generally have low vapor pressures and the effect
of these cutting fluids are not always considered when determining the true vapor pressure of the
tank contents when doing emission inventory calculations. However, the rule requires that the
true vapor pressure be determined based on all materials in the storage vessel. There are also
potentially significant differences in short-term emissions that may occur during, for example,
tank filling in the heat of summer, than the annual average emission rate. Therefore, the short-
term DIAL measurements are often not representative of the long-term annual average emissions
rate. Based on our technology review, the 0.75 psia vapor pressure threshold was determined to
be the appropriate threshold for cost-effective additional storage vessel control requirements.
5.1.3 Revision of definition of reference control technology for storage vessels
Comment 1: One commenter suggested revisions to 63.660(b) to clarify requirements for
floating roof deck fittings by adding requirements for ladders that have at least one slotted leg
that are identical to the proposed requirements for slotted guidepoles. The commenter added that
these provisions were specified in the proposed Uniform Standards, but were left out of the
proposed revisions. The proposed options included a ladder sleeve, the effectiveness of which
was demonstrated to the EPA prior to proposal of the Uniform Standards, and no rationale is
158
-------
given for the omission of this control option in this proposal. Thus, the commenter has requested
that the control options from the proposed Uniform Standards be included here.
Response 1: EPA agrees with and has amended this final rule in response to this
comment. These revisions would require the same controls for slotted ladder poles as those
proposed for slotted guidepoles. These two types of slotted poles are similar deck fitting
structures having similar emissions characteristics, even though they provide different structural
functions. These amendments prevent an unintended loophole that would have allowed more
emissions of HAPs to the atmosphere.
Comment 2: One commenter stated that the requirements for floating roof deck fittings are in
need of editing for clarity and completeness, as well as for correcting minor typographical errors.
The flexible enclosure control option is allowed under the STERPP for internal floating-roof
tanks as well as for external floating-roof tanks. The proposed language to revise Refinery
MACT 1 limits this option to external floating-roof tanks, but there is no rationale given for this
limitation. The commenter assumed this to simply be an oversight, and requested that the flexible
enclosure control option be allowed for both internal floating-roof tanks and external floating-
roof tanks. The commenter also stated that if this is a purposeful change, the cost impact for
internal floating roof tanks using this control to make this change and the justification should be
provided and published for comment. This issue pertains to 63.640(n)(8)(vii) and (n)(10)(vii),
with respect to the overlap provisions, and to 63.660(b), with respect to citations to subpart WW.
Response 2: EPA agrees the flexible enclosure control option is applicable to both internal and
external floating-roof tanks and has amended the final rule requirements to extend the use of this
control option for both types of tanks.
5.2 Compliance schedule provisions
Comment 1: One commenter stated that the format for introducing the new storage vessel
requirements must be improved. The commenter added that applicability dates are needed for the
new requirements. While the commenter endorsed the approach of adding a new section (i.e.,
63.660) for the new standards and adding a statement at the beginning of the prior standards (i.e.,
63.646) explaining that this section will no longer apply after demonstration of compliance with
the new section, they recommended further clarification in the applicability section of the rule.
The commenter explained it would be very helpful if the compliance schedule provisions in
63.640(h) alerted the reader to the separation of the prior and new standards for storage vessels
in sections 63.646 and 63.660, respectively.
Response 1: EPA maintains that the applicability dates for storage vessels are adequately
described in the introductory paragraphs to §§63.646 and 63.660, as well as in Table 11. The
provision in §63.640(h) specifically direct the reader to Table 11 where the compliance dates are
listed separately for §§63.646 and 63.660. It should be noted that compliance with §63.646 for
new and existing sources will only be permitted for up to 90 days after publication of the final
rule. Therefore, we consider it unnecessary to revise the referenced provisions, and they will be
finalized as proposed.
159
-------
Comment 2: One commenter stated that editorial corrections are needed in 63.660(d) to clarify
the compliance timing for storage vessels that become Group 1 as a result of these amendments.
The commenter explained that 63.660(d) pertains to equipping a fixed-roof storage vessel with
controls, if it has not been subject to control prior to these amendments. The specified control
could be achieved by routing the vapors to a control device or by installing a floating roof.
However, the citation to "the requirements of section 63.1062" could be construed as being
applicable only to the installation of a floating roof. It would be apparent that the control
requirements in question could be either routing to a control device or installing a floating roof if
this phrase were changed to read "the requirements of this section," in that the introductory text
of this section (63.660) addresses complying with either subpart WW (for installing a floating
roof) or subpart SS (for routing vapors to a control device). In addition, the term "storage vessel
fixed roof' in 63.660(d) should read "fixed roof storage vessel."
Response 2: EPA agrees that corrections are needed to the provision contained in §63.660(d), as
it was not our intent to limit compliance options to the installation of a floating roof. The
provision has been amended in the final rule to allow compliance for an uncontrolled fixed roof
storage vessel that commenced construction on or before June 30, 2014 and that meets the
definition of Group 1 storage vessel in item 2 in §63.641 but not the definition of Group 1
storage vessel in item 1 in §63.641, to comply by installing a floating roof or by routing vapors
to a control device.
Comment 3: One commenter stated that the compliance date is unclear for storage tanks
complying with the Kb option under 40 CFR 63.640(n)(8). The commenter explained that Table
11 (4)(iii) references section 63.660, but that section should not be applicable since section
63.660 references tanks complying with 40 CFR part 63 subparts WW or SS. The commenter
inquired if the citation of section 63.640 was intended to provide the compliance date for the
subpart Kb option found in Table 11 (4)(ii). The commenter added if so, the EPA should provide
a 10-year upgrade option as found in subpart WW (63.1063(a)(2)(ix)). Finally, the commenter
stated that the table contains two references to subsection "(iii)" with respect to requirements
applicable to units with a date of construction on or before July 14, 1994. The latter reference
should be changed to "(iv)" to avoid confusion.
Response 3: EPA expects storage vessels which are in compliance with subpart Kb of part 60 to
already be using guidepole controls as outlined in the notice of the STERPP (described at 65 FR
19891 on April 13, 2000)). The STERPP notice specifically stated, "that uncontrolled slotted
guidepoles do not comply with the "no visible gap" requirement in NSPS subparts Ka and Kb.
See 65 FR 2336 (January 14, 2000)." Appendix I in the STERPP notice went on to identify
acceptable controls for slotted guidepoles. Facilities have now had over 15 years to implement
these controls. Therefore, the EPA maintains that sufficient time has passed to implement
controls for storage vessels complying with subpart CC of part 63 through the overlap
provisions, and no 10-year upgrade option will be given. We are clarifying in this response that
the citation of 63.660 is intended to provide compliance for the Kb option found in Table 11. We
note that the overlap provisions provide facilities with a way to comply with the applicable
requirements of subpart CC for storage vessels (i.e., §63.660), and thus the applicable
compliance dates in subpart CC apply to tanks utilized the overlap provisions. In the final rule,
160
-------
we are correcting the typographical error in Table 11 (4) in which two rows in the proposed table
were given the same "(iii)" label.
5.3 Impact estimates for storage vessel requirements
Comment 1: One commenter asserted that the EPA's cost effectiveness determinations in the
RTR proposal may not apply to them. Because of the distance of St. Croix from the continental
U.S., costs for a given project are 1.3 to 2.3 times higher at the commenter's refinery than a
continental refinery. Also, the commenter is almost entirely dependent on marine movement of
products and raw materials and has many more storage tanks (due to limit supply interruptions)
than continental refineries. These storage tanks will be subject in many cases to added control
costs by the RTR Rule. The commenter recommended that the EPA consider these added
impacts to non-mainland facilities in finalizing the RTR Proposal provisions at section 63.600
and the potential for enhanced tank repair requirements under the fenceline monitoring
provisions.
Response 1: The EPA develops impacts as an average of data collected nationwide, and thus we
expect some facilities to be above and below the cost effectiveness determinations, and thus the
cost to comply will vary. Furthermore, we believe the refinery identified in this comment is no
longer in operation, and thus the requirements no longer apply.
5.4 Monitoring and inspection requirements
Comment 1: In response to the EPA request for comment in the preamble, one commenter
stated that consistent with the allowance in subpart WW, the EPA should allow in-service
inspections of internal floating roof tanks. The commenter asserted that performing inspections
in this way avoids significant hazardous waste generation, shutdown emissions, the operating
risk associated with having tanks out of service, and is much less costly than performing an out-
of-service inspection. The commenter argued that to require a facility to incur the otherwise
unnecessary emissions associated with tank cleaning, when the inspection can be performed
adequately with the storage vessel in service, is unjustifiable and less favorable for the
environment. Thus, the commenter recommended the EPA not take any action that would
interfere with this proven subpart WW provision.
Response 1: After reviewing the solicited comments regarding the internal inspection provisions
for in-service tanks in subpart WW, we agree with the commenter and have finalized these
referenced provisions as proposed.
Comment 2: One commenter contended that the requirement for tank inspections is too
infrequent, and 10 years is too long to go without an inspection.
Response 2: Although there are types of inspections with a frequency of every 10 years, storage
vessels are subject to different types of inspection with varying frequency. An internal floating
roofs floating roof deck, deck fittings, and rim seal are required to be visually inspected through
openings in the fixed roof annually. The inspection with a frequency of 10 years requires owners
and operators to inspect these same items from within the tank unless all deck components are
161
-------
visually accessible from the top side. Therefore, the allowance to perform this 10-year inspection
without taking the tank out of service does not reduce the number or frequency of all required
tank inspections, but rather provides facilities with a less costly and labor intensive way to
perform the inspection (i.e., not requiring it to be taken out of service).
Comment 3: One commenter stated that the bypass monitoring requirement in 63.660(i)(2)
needs to be clarified to match those applicable to potential MPV bypasses. The commenter
explained that 63.644(c), which addresses MPV bypass monitoring, contains exceptions to the
bypass monitoring requirements including: Equipment such as low leg drains, high point bleeds,
analyzer vents, open-ended valves or lines, PRVs needed for safety reasons, and equipment
subject to 63.648 are not subject to this paragraph.
The commenter asserted that while revisions to these exceptions have been proposed and the
commenter provided comments on those revisions, it is important that 63.660(i)(2) contain the
same exceptions as are finalized for 63.644(c). The commenter argued that there is no logical
reason for treating these two situations differently and equipment that is subject to other Refinery
MACT 1 requirements or is required for safety should not also be considered potential bypasses
and subject to these sometimes conflicting requirements as well.
Response 3: We agree with the commenter and have finalized a consistent list of exceptions for
bypass for MPV and storage vessels.
Comment 4: One commenter stated that the citation in 63.660(i)(2) to 63.985(a)(3)(i) or (ii)
should be to 63.983(a)(3)(i) or (ii).
Response 4: We have corrected these references and will finalize these changes.
5.5 Recordkeeping and reporting requirements
Comment 1: Commenters requested several editorial changes and clarifications to the reporting
requirements for storage vessels including:
Clarifying the scenarios under which information related to gaskets and slotted
membranes are required to be reported, specifically with respect to provisions in
§§63.655(g)(2)(i)(B)(l) and 63.655(g)(3)(i)(C)(l).
Adding a reporting requirement to report whether compliance is being demonstrated for
§§63.646 or 63.660.
Adding a note from §63.119 to §63.655(g)(2)(i)(A)(l) which allows the floating roof to
rest on the leg supports.
Modifying the term "slotted membrane" to read "slotted membrane fabric cover" specific
to roof drains in external floating roofs which empty into the stored liquid consistent with
§63.119.
Modifying the term "slotted membrane" specific to the sample wells in internal floating
roofs to be referred to as "slit fabric cover" consistent with §63.119.
Adding a reference to the reporting requirements for facilities utilizing the subpart Kb of
part 60 overlap provisions as outlined in §63.648(n)(8) to §63.655(g).
162
-------
Response 1: We disagree with the commenter that these changes are necessary. We did not
materially change the provisions identified above in the proposed rule. Further, these provisions
are for storage vessels complying with the requirements in §63.646, and they will only be
available as a method for demonstrating compliance for up to 90 days after publication of the
final rule. Therefore, the suggestions will not be incorporated into the final rule.
Comment 2: One commenter requested that the EPA clarify how the requirements in section
63.655(g)(3)(i)(C) is invoked. The commenter explained that section 63.655(g)(3)(i) only
references (g)(3)(i)(A) and (g)(3)(i)(B).
Response 2: To correct the referential error identified by the commenter, we have amended
paragraph 63.655(g)(3)(i) in the final rule to reference subparagraphs (A) through (C) and not
just (A) and (B).
Comment 3: Commenters recommended that external floating roof tanks converted to internal
floating roof tanks should be subject to the same requirements as internal floating roof tanks.
One commenter added that the reference in 63.655(g)(2)(i)(A)(2) to (g)(2)(i)(C) should refer to
(g)(2)(i)(A)(3). The commenter also stated that the reference to (g)(2) through (5) in
63.655(g)(1) should be to (g)(2) and (3). The commenter also added that paragraph 63.655(g)(4)
is unnecessary if paragraphs 63.655(g)(2) & (3) are complete and self-contained. The commenter
provided draft regulatory text in their comment package for the recommended revisions.
Response 3: Section 63.655(g)(2)(ii) references 40 CFR part 63 subpart WW, which makes no
differentiation between an internal floating roof tank. In fact, the definition of internal floating
roof in §63.1061 states, "[f]or the purposes of this subpart, an external floating roof located in a
storage vessel to which a fixed roof has been added is considered to be an internal floating roof."
We would like to clarify in this response that we did not intend to differentiate between internal
floating roofs and external floating roofs which have been converted to internal floating roofs in
§63.660. We have updated (i) and removed (iii) from the definition of Reference control
technology for storage vessels consistent with this clarification. We agree with the commenter
that the reference in 63.655(g)(2)(i)(A)(2) should be to (g)(2)(i)(A)(3) and have finalized this
change. We disagree that the reference in 63.655(g)(1) should be changed, as it appropriately
references citations with recordkeeping and reporting requirements for storage vessels and are
finalizing this provision as proposed. We also agree with the commenter that the paragraph in
63.655(g)(4) is redundant and unnecessary and will not be finalized.
5.6 Other
Comment 1: Two commenters identified a typographical error at 63.640(n)(l), specifically that
the citation in the phrase "... is required to comply only with the requirements of 40 CFR part
60, subpart Y, ..." should be "40 CFR part 61, subpart Y".
Response 1: EPA agrees with this comment and has corrected the citation in the final rule.
Comment 2: Two commenters noted typographical errors at 63.640(n)(8), specifically that the
citation in the phrase "... in paragraphs (n)(8)(i) through (n)(8)(vi) of this section ..." should be
163
-------
"(n)(8)(i) through (n)(8)(viii)" and that the citation in the phrase "... in paragraphs (n)(8)(i)
through (n)(8)(vii) of this section." should be "(n)(8)(i) through (n)(8)(viii)"
Response 2: EPA disagrees with the commenters regarding the need to amend §63.640(n)(8)
from "... in paragraphs (n)(8)(i) through (n)(8)(vi) of this section ..." to "(n)(8)(i) through
(n)(8)(viii)". This part of the provision applies to Group 2 storage vessels in §63.640(n)(l) which
do not require add-on controls and thus the requirements in §63.640(n)(8)(vii) and (viii) are not
expected to apply. EPA agrees with the commenters regarding the need to amend the citation in
the phrase "... in paragraphs (n)(8)(i) through (n)(8)(vii) of this section." to "(n)(8)(i) through
(n)(8)(viii)", as this provision applies to Group 1 storage tanks which are expected to have add-
on controls and has finalized this change.
Comment 3: Several commenters discovered a typographical error at 63.640(n)(10). The
citations to paragraphs in the phrase "...paragraphs (n)(10)(i) through (n)(8)(vi) of this section"
should be "(n)(10)(i) through (n)(10)(viii) of this section"
Response 3: Similar to the response above, EPA disagrees with the commenters regarding the
need to amend the citations to paragraphs in the phrase "...paragraphs (n)(10)(i) through
(n)(8)(vi) of this section" to "(n)(10)(i) through (n)(10)(viii) of this section". This part of the
provision applies to Group 2 storage vessels in §63.640(n)(l) which do not require add-on
controls and thus the requirements in §63.640(n)(10)(vii) and (viii) are not expected to apply.
Comment 4: One commenter stated that the requirements for floating roof deck fittings are in
need of editing for clarity and completeness by correcting typographical errors. Specifically, at
§63.660(b), the citations to subpart WW should be to 63.1063(a)(2)(viii) for slotted guidepoles,
rather than to §63.1063(a)(2)(vii).
Response 4: EPA agrees with the commenter and has corrected the typographical errors at
§63.660(b) in the final rule.
164
-------
6.0 Equipment Leak Standards
6.1 Technology review results
Comment 1: One commenter stated as a legal matter, a leak is a type of malfunction. The
commenter argued that authorizing facilities to leak below a given threshold, and to do so
without repairing or ending the leak, means that the EPA's standards do not apply continuously,
as the CAA requires. The commenter asserted that the EPA must finalize a rule that sets a leak
prohibition and establishes up-to-date, stronger leak detection and repair requirements.
The commenter also identified the provisions in the current rules that allow sources to delay and
defer leak repair indefinitely as unlawful malfunction exemptions. The commenter argued in
particular that the EPA proposes to continue to rely on 40 CFR part 60, subpart VV, and 40 CFR
part 63, subpart HH, which according to the commenter contain a number of deferred monitoring
and repair provisions that are unlawful exemptions that the EPA must remove. The commenter
asserted that the EPA must set a firm, enforceable deadline for the repair of all leaks found and
must not exempt any valves, connectors, or other equipment from the leak detection and repair
standards and requirements.
Response 1: We disagree with the commenter's claim that the types of equipment leaks
addressed in Refinery MACT 1 are "malfunctions." Equipment leaks typically occur from
equipment such as pumps, compressors, agitators, sampling collection systems, open-ended
valves or lines, valves, and connectors.89 At the time we developed the MACT standards for this
source category, we recognized that these emission points even at the best performing facilities
regularly emit small quantities of HAP, and we promulgated standards regulating equipment
leaks from these components based on what the performance of the best performing sources at
40 CFR 63.648. These provisions require petroleum refiners to monitor for leaks and to repair
any detected leaks. While any specific equipment leak is not predictable, the types of equipment
leaks addressed by the regulations at 40 CFR 63.648 are fairly routine emissions from sources
are not the type of unpredictable or infrequent event for which we cannot anticipate when, where
or how they may occur, but that we generally consider to be malfunctions.
The delay of repair provisions were included to prevent the undesirable impact of creating more
emissions from shutting down and evacuating major process equipment than are emitted from
the leaking equipment component. In such cases, the environment is better served by allowing a
small leak to persist until the next scheduled shutdown than to shut the unit down to replace the
leaking component, and the owner or operator must make this demonstration to avail themselves
of these provisions90. Contrary to the suggestion of the commenter, the delay of repair
89 40 CFR 63.648; see also 60 FR 43260 (August 18, 1995)
90 40 CFR 60.482-9 (c) (1) Delay of repair for valves will be allowed if(l) The owner or operator demonstrates
that emissions ofpurged material resulting from immediate repair are greater than the fugitive emissions likely to
result from delay of repair,...
165
-------
requirements do establish firm timelines by which a leak must be repaired. Further, difficult to
monitor equipment are still subject to audible, olfactory and visible emissions inspections
(although from a distance), so these equipment components are not exempt from the LDAR
provisions.
Comment 2: One commenter noted that the EPA concluded in the proposal that updating the
standards for equipment leaks was not "necessary" based in part on the annualized cost and "cost
effectiveness" (dollars per ton of HAP reduced) estimated for available methods to reduce
emissions. However, the commenter argued that section 112(d)(6) does not authorize the EPA to
refuse to update standards based on cost. Where "developments" have occurred, the EPA must
"account" for those. The D.C. Circuit has held that, when setting revised air toxics standards, the
EPA may consider cost, but it has not held that the EPA may not update a standard, at all, as it
proposes to do for leaks and wastewater, based on cost. That Court has recognized that
"developments" are the core requirement of this provision.91 Thus, the commenter asserted that
if they are present, the EPA must set revised standards. The commenter further opined that the
D.C. Circuit's holding on cost in Association of Battery Recyclers v. EPA was wrongly
decided.92 Notwithstanding that decision, the EPA should decide to follow the plain text of
section 112(d)(2)-(3) and applicable precedent requiring explicit authorization to consider cost.
The commenter argued that the EPA's decision to make cost-per-ton the standard-setting
criterion and to choose a number it deems unreasonable, without a rational explanation, is
arbitrary and capricious.
Commenters added that certain facilities have complied with stronger LDAR provisions in the
state of California or under consent decree demonstrating that such requirements are
technologically and economically feasible and the commenters asserted that because the EPA's
cost analysis does not consider this, it is thus arbitrary and incomplete. The commenters also
stated that the EPA has ignored positive economic impacts that the rule improvements can have,
such as job creation.
The commenter also stated that the cost-per-ton does not consider the benefits associated with
reducing health risk by reducing the amount of HAP emitted by these sources. The commenter
stated that the EPA's cost-focused analysis ignored the statutory objective of assuring the
"maximum" achievable degree of emission reduction provided in section 112(d)(2) and
implemented through the review required by section 112(d)(6). The commenter added that it also
ignored the statutory goal of protecting public health, which is the core purpose behind this
provision and the stated purpose of section 112(f)(2). Thus, the commenter asserted that the
EPA's proposed inaction is unlawful, arbitrary, and capricious because it ignored key statutory
purposes that are required factors to consider.
Moreover, the commenter detailed that the cost the EPA found here is lower than that in other
rules or for other source categories within this proposed rule, for which the EPA has determined
the cost-per-ton to be appropriate. For example, the EPA recognized that an annualized cost of
91 NRDCv. EPA, 529 F.3d 1077, 1084 (D.C. Cir. 2008).
92ABR v. EPA, 716 F.3d 673 (D.C. Cir. 20132003); NASF v. EPA Opening and Reply Briefs of Environmental
Petitioners.
166
-------
$36.3 million/year is appropriate for the flare provisions. That is orders of magnitude higher than
the annualized costs of the leak reduction provisions the EPA proposed to reject.
The commenter added that the EPA has recognized as appropriate or cost-effective much higher
cost-per-ton ratios. For example, in the NESHAP from Secondary Lead Smelting (77 FR 556,
576; Jan. 5, 2012), the EPA determined that the following cost-per-ton measures were
appropriate: $0.33 million/ton ($ 170/lb) (for stack lead emission limit); $1 million/ton ($500/lb)
(for enclosure requirements); $1.5 million/ton ($550/lb) (for fugitive control work practices).
Response 2: We do not agree that in reviewing a standard under CAA section 112(d)(6), the
CAA mandates that the EPA must promulgate revised standards if "developments" have been
identified. CAA section 112(d)(6) requires that EPA review and revise standards "as necessary."
Under CAA section 112(d)(6), the EPA retains very significant discretion in balancing relevant
factors in determining whether it is "necessary" to revise the existing technology-based MACT
standards. See, e.g., Sierra Club v. EPA, 325 F. 3d 374, 378 (D.C. Cir. 2003) (under CAA
section 202(1)(2), EPA is to consider factors beyond pure technological capability, and the statute
does not direct how EPA should weigh such factors). In reviewing section 112(d)(2) standards,
and determining whether to revise them is "necessary" under section 112(d)(6), the EPA may
take into consideration cost and feasibility when evaluating developments in practices, processes
and control technologies.
With regard to the comment referring to more stringent state requirements and consent decree
requirements, we did consider these, and in fact, as noted in the proposal preamble at 79 FR
36916, we acknowledged that some owners and operators are required to repair leaking valves as
low as 100 ppm and pumps as low as 500 ppm. However, these lower leak definitions would
have been more costly than the leak definitions of 500 ppm for valves and 2,000 ppm for pumps
that we already rejected as not cost effective.
We also reject the commenter's argument that in rejecting more stringent LDAR requirements,
we ignored the statutory goal of protecting public health, which is the core purpose of section
112(f)(2) and that rejecting more stringent LDAR requirements based on cost-per-ton failed to
consider the ensuing health benefits associated with reducing the amount of HAP emitted by
these sources. In the proposed rule, as part of the CAA section (f)(2) risk assessment, we
evaluated the impacts on risks of all the equipment leak advancements in technologies and
practices that we identified. We concluded that the available control options for equipment leaks
for petroleum refineries do not provide quantifiable risk reduction and therefore, we proposed
that further controls were not necessary to provide an ample margin of safety, regardless of costs
or cost effectiveness (see 79 FR 36941, June 30, 2014).
The commenter's comparison of dollar per ton cost values against other rules and other
requirements within this final rule are also misplaced. First, the commenter draws a comparison
to an analysis for metal HAP in the Secondary Lead NESHAP RTR, where those costs per ton
were determined to be within the range of metal HAP values for other section 112 rules (see 77
FR 576, January 5, 2012). However, organic HAP are the issue of concern for equipment leaks,
and the EPA has historically used a different cost effectiveness scale for organic HAP versus
metal HAP (i.e., tons versus pounds) due to their relative toxicity. As for the flare operating
167
-------
requirements in this final rule, the annualized cost of $36 million is not related to the RTR, but
rather, is an estimate of the cost of proposed revisions to the flare monitoring requirements to
ensure that flares used as APCD meet the MACT standards at all times when controlling HAP
emissions under the authority of CAA section 112(d)(2) and (d)(3). A more appropriate
comparison would be how LDAR controls were considered under technology reviews in recent
rulemakings. For example, we conducted technology reviews for LDAR requirements in the final
RTR rules for the Polymers and Resins IV, Pesticide Active Ingredient and Polyether Polyols
Production MACT standards (see 79 FR 17340, March 27, 2014), the Acrylic and Modacrylic
Fibers Production, Amino/Phenolic Resins Production and Polycarbonate Production MACT
standards (see 79 FR 60898, October 8, 2014) and the Offsite Waste and Recovery Operations
MACT standards (see 80 FR 14248, March 18, 2015), and our decisions for LDAR under the
technology review for this rule are consistent with these recent rulemakings. Further, the
commenter fails to acknowledge that we proposed and are finalizing a fenceline standard in this
rule as part of our technology review to ensure that petroleum refinery owners and operators
properly monitor and manage fugitive HAP emissions, which includes equipment leaks. As
stated in the proposal preamble (see 79 FR 36919, June 30, 2014), it is impractical to directly
measure emissions from fugitive sources at refineries, thus requiring the use of emission
estimation models, which in turn introduces uncertainty into the inventory for emission sources,
such as equipment leaks. We evaluated the developments in processes, practices and control
technologies for measuring and controlling fugitive emissions from the petroleum refinery as a
whole and identified fenceline monitoring as a development. We evaluated several fenceline
monitoring techniques and proposed to require a system of passive monitors. Fenceline
monitoring will identify increases in HAP emissions in a timely manner, which will allow
corrective action measures to occur more rapidly than would happen if a source relied solely on
the traditional infrequent monitoring and inspection methods, such as those associated with
periodic Method 21 LDAR requirements.
It should also be noted that positive impacts to the economy as a result of the rulemaking were
considered as documented in 4.3.2.2 Environmental Protection Sector and 4.3.2.3 Labor Supply
Impacts of the Economic Impact Assessment (EIA) which can be found in the docket for this
rulemaking (EPA-HQ-OAR-2010-0682).
Comment 3: One commenter argued that the EPA must strengthen the leak detection and repair
standards so they are consistent with evidence showing leaks can be prevented and minimized.
The commenter stated that the EPA must set standards to limit leak emissions based on currently
available "zero emissions technologies" and practices, at a minimum, as demonstrated by the
best performing sources in order to follow the CAA's requirements, under section 112(d)(2)-(3).
For example, the commenter stated that the EPA proposed to retain the leak definition of 10,000
parts per million (ppm) for valves and pumps at existing sources and 1,000 ppm for valves at
new sources, which is inconsistent with lower leak definitions in other EPA rulemakings (77 FR
49490, August 16, 2012) and the leak definitions in the BAAQMD and SCAQMD rules. The
commenter asserted that it would be arbitrary and capricious for EPA not to finalize LDAR
requirements that are at least as stringent as these rules.
168
-------
The commenter also noted that the EPA identified additional developments in its memo on the
uniform standards and certain chemical plant sources that have occurred that the EPA has not
acknowledged in this rulemaking, including the use of low-leak technologies; lowering the leak
definitions to even tighter levels, and requiring tighter timelines for minimization of leaks to
within 24 hours of identification and repairs within seven days.93 In addition, the commenter
explained that leak-free pumps, leakless valves, and improvements in practices that reduce the
number of leaks by using a greater percentage of other kinds of leakless devices reflect
developments within the meaning of section 112(d)(6) that the EPA is required to consider.94
These types of provisions are included in the consent decrees for Hovensa, BP Whiting, Murphy
Oil, Shell Chemical, and Dow Chemical, as well as numerous chemical plants. The commenter
stated that an increase in enforcement actions surrounding equipment leaks demonstrates the
need for stricter standards.
Response 3: As the commenter stated, we did identify developments in practices, processes, and
control technologies, documented in Analysis of Emissions Reduction Techniques for Equipment
Leaks (Docket Item No. EPA-HQ-OAR-2010-0869-0029); this memorandum was referenced
from Impacts for Equipment Leaks at Petroleum Refineries (Docket Item No. EPA-HQ-OAR-
2010-0682-0207), which documents the costs and environmental impacts for the regulatory
options considered as part of the technology review for equipment leaks at petroleum refineries.
While the memorandum Analysis of Emissions Reduction Techniques for Equipment Leaks
mentions some of the developments identified by the commenter, such as sealless pumps and
"low leak" packings for valves, the memorandum does not include an evaluation of the cost and
emission reduction impacts of these developments, as the necessary data are not available.
However, these requirements are more stringent and more labor and equipment intensive than
other approaches and thus almost certainly more costly. Since we rejected those other approaches
as cost effective, we also would have rejected other more costly approaches as not necessary
under section 112(d)(6).
In addition, we note that the requirements in consent decrees are negotiated settlements and are
not based on an analysis of the nationwide impacts, including costs, conducted as part of a
technology review. The analysis conducted for this final rule package supports the requirements
in this national rulemaking and is based on the nationwide costs and emissions reductions
achieved.
93 Analysis of Emissions Reduction Techniques for Equipment Leaks, EPA-HQ-OAR-2010-0869-0029 (uniform
standards docket); Supplemental Technology Review for Equipment Leaks in Group IV Polymers and Resins,
Pesticide Active Ingredient Production, and Polyether Polyols Production Source Categories (Jan. 31, 2014), EPA-
HQ-OAR-2011-0435-0082.
94 In addition to the documents cited herein, we have also attached and incorporate by reference the discussion of
developments and leak detection and repair methods discussed by the Bruce Buckheit Report on oil and gas, as all or
many of the same problems are issues that EPA must address in this rule for Refineries. See Buckheit Report (Nov.
2011).
169
-------
6.2 Optical gas imaging provisions
Comment 1: Two commenters supported the use of optical gas imaging as an alternative means
for leak detection on Refinery MACT 1 equipment leaks as proposed {i.e., in place of Method 21
instrument monitoring). Two commenters requested that the EPA finalize the protocol for optical
gas imaging as Appendix K to 40 CFR part 60. Another commenter stated that optical gas
imaging should also be provided as an alternative for equipment in VOC service, regulated in 40
CFR part 60, to provide consistent approaches for the facility's overall LDAR program. Another
commenter also requested that the EPA establish effective dates for using the optical gas imaging
techniques after the rules and protocol are finalized to ensure facilities have time to develop the
systems necessary to comply with the requirements and are able to properly detect leaks.
One commenter added that ultrasound or optical scanning or imaging programs should be a part
of an overall LDAR program. The commenter noted that they can be used to conduct daily or
weekly scans to identify areas to target during LDAR inspections. The commenter also added
that remote scanning devices should be used to monitor equipment that is not currently required
to be monitored using Method 21, and EPA should require facilities to report the results of any
optical gas imaging scans conducted.
To the contrary, two commenters opposed the use of optical gas imaging because the instrument
cannot provide instantaneous measurement of actual VOC concentrations in ppm.
Response 1: We do not plan to propose the optical gas imaging protocol in Appendix K to 40
CFR part 60 until we develop procedures and specifications that we determine will result in
effective detection of leaks. As we noted in the preamble to the proposed rule, we plan to
propose Appendix K and request comments on that appendix and how those requirements would
apply to equipment subject to Refinery MACT 1. We will not take final action adopting use of
Appendix K to 40 CFR part 60 for optical gas imaging for refineries subject to Refinery MACT
1 until such a time as we have considered any comments on that protocol as it would apply to
refineries.
In response to the comment that optical gas imaging should be provided as an alternative for
equipment in VOC service, we note that Refinery MACT 1 regulates HAP emissions, and any
change in requirements for equipment in VOC service is beyond the scope of this rulemaking.
However, while we are first considering how Appendix K to 40 CFR part 60 would apply to
equipment in HAP service at petroleum refineries, we anticipate that it may be an alternative to
Method 21 for other monitored process units for which the owner or operator can verify that the
instrument selected can image prevalent chemicals. Therefore, owners and operators of those
process units with equipment in VOC service would be able to request to use Appendix K to 40
CFR part 60 as an alternative to Method 21, just as they would request to use any other
alternative test method.
As the use of optical gas imaging would be a voluntary alternative to the standards in Refinery
MACT 1, we do not at this time anticipate needing to establish effective dates for compliance.
Depending on the content and format of the final Appendix K, we may decide to establish dates
within Refinery MACT 1 before which optical gas imaging using Appendix K cannot be used,
170
-------
although we do not anticipate this will be necessary at this point. We think most owners and
operators will take the time they need to ensure they can comply with Appendix K and that it
will not be necessary to prescribe a beginning date.
As part of the development of Appendix K to 40 CFR part 60 and how it will apply to equipment
subject to Refinery MACT 1, we will consider the appropriate monitoring and repair
requirements along with the recordkeeping and reporting requirements. At that time, we will
evaluate the suggestions raised by the commenter for integrating optical gas imaging programs
into existing LDAR programs, as well as similar requirements in existing standards such as the
Alternative Work Practice To Detect Leaks From Equipment (40 CFR 63.11(c), (d) and (e)).
While current optical gas imaging technology does not provide instantaneous readings of VOC
concentration, we do not believe that this is a deterrent to the use of this technology once an
appropriate protocol for its use has been developed. Because anything that is imaged by a camera
is considered to be a leak, there is no real need for the instrument to provide a numerical reading,
such as is necessary when complying with a numerical leak definition. We also note that this
technology is quickly evolving and believe that these cameras may be able to quantify leaks in
the near future, should quantification be necessary. Sources may also choose to quantify leaks in
other ways (e.g., use of a Method 21 instrument, bag sampling) if this information is needed for
other purposes.
Comment 2: One commenter stated that the requirement in 40 CFR 63.661(a)(2) requiring that
the owner or operator be in compliance with the fenceline monitoring provisions of 40 CFR
63.658 in order to make use of the optical gas imaging (OGI) alternative (if the other conditions
specified in 40 CFR 63.661(a) are met) should not be finalized because the fenceline and leak
detection requirements are work practice based standards and compliance for these requirements
are not tied to an emissions limit.
The commenter added that compliance with the fenceline requirements is based on a two week
sampling period for which the refinery may not know the status for up to 30 days, while OGI is
performed periodically at various locations within the refinery and thus the compliance status
may be unknown. The commenter also requested guidance for a scenario in which non-
compliance with the fenceline program occurred in between OGI runs for a particular portion of
the refinery, but the fenceline monitoring program was in compliance for the two week period
during which the OGI for that portion of the refinery was performed. The commenter also stated
that using Method 21 when the OGI alternative is not being utilized presents further challenges
because of the amount of time it can take to perform Method 21 for a particular unit or the entire
refinery.
Finally, the commenter added that from a practical standpoint, is it unclear how one could move
in and out of the OGI program every 2 weeks. Presumably, when OGI is not being used one must
do Method 21 monitoring, but the whole point of OGI is to save the effort associated with
Method 21 monitoring. Furthermore, Method 21 monitoring typically takes more than a 2-week
period to complete for any particular unit and takes at least a quarter to complete for the entire
refinery. Therefore, the commenter stated, it would be impossible to switch back and forth for
171
-------
unplanned 2-week periods, making the use of OGI impractical and forcing sources to simply
remain with Method 21 monitoring rather than using the OGI alternative.
Response 2: As noted in the response to Comment 1 of this section, we do not plan to propose
Appendix K to 40 CFR part 60 until we have developed procedures and specifications that we
determine will result in detection of leaks at least as effectively as Method 21. We had
anticipated that Appendix K would have been proposed prior to finalizing the Refinery MACT 1
requirements. Given the current status of Appendix K, we are not finalizing the proposed
amendments at 40 CFR 63.661.
Comment 3: One commenter stated that it is unclear just which provisions in 40 CFR part 60,
subpart GGGa (and through it, subpart VVVa) may be replaced by OGI and which still apply.
The commenter requested that the individual sections of 40 CFR part 60, subparts GGGa be
identified in the second sentence of 40 CFR 63.640(p)(2). Additionally, the commenter stated
that the last part of the second sentence in 40 CFR 63.640(p)(2) should be revised to match the
wording in 40 CFR 63.661(a)(1) and indicate that facilities still have to comply with the other
provisions of subpart GGGa, but not that a facility must be "in compliance" with those other
provisions. If this language is not changed, the commenter noted that, for example, a failure to
perform a single weekly pump check somewhere in the refinery would mean the facility would
not be permitted to comply with OGI.
The commenter also requested that the reference to paragraph (a)(2) be changed to paragraph
(a)(3) in proposed 40 CFR 63.661(b)(3).
Response 3: We had anticipated that Appendix K would have been proposed prior to finalizing
the Refinery MACT 1 requirements. Given the current status of Appendix K, we are not
finalizing the proposed amendments at 40 CFR 63.661.
6.3 Clarification of seal for open-ended lines
Comment 1: Several commenters objected to finalizing the proposed clarification of "seal" as
proposed because it would cause owners and operators to incur a violation if the open-ended line
(OEL) is equipped with a proper plug, cap, or second valve but the "seal" is found leaking above
the leak threshold of 500 ppm. Two commenters added if OELs are subject to a leak standard,
they should not also be required to have a cap or plug; the requirement should be either install a
cap/plug or comply with a 500 ppm leak definition. Two commenters also stated that it is
arbitrary and capricious to establish an emission limitation in the case of OELs when other
similar fugitive emission components (e.g., connectors) are regulated through an LDAR work
practice standard
Another commenter requested the proposed rule language be revised to specify that if a missing
cap or plug is found, sources will have 24 hours to monitor the opening and determine if the
leakage exceeds 500 ppm. The commenter suggested that if the instrument reading is above 500
ppm, a deviation from the cap or plug requirement would be identified, and if not, no deviation
would be identified and the source would have to install a cap or plug.
172
-------
Response 1: For reasons provided in the preamble to the final rule, we are not finalizing any
changes to the rule language related to the requirements for OELs subject to Refinery MACT 1.
Therefore, the clarification requested by the commenters is not necessary.
Comment 2: One commenter stated that the proposed 500 ppm leak detection limit for OELs is
not necessary for components in heavy-liquid service but noted the provisions did not clarify
this. The commenter requested clarification that the requirement does not apply to heavy-liquid
components.
Response 2: The OEL standards apply to all open-ended valves and lines subject to Refinery
MACT 1 except those specifically described in 40 CFR 60.482-6(d) and (e). We did not propose
to change those provisions, and we are not finalizing any changes to those provisions.
Comment 3: Several commenters stated that the proposed approach established a 500 ppm
emissions limitation for controlled OELs in addition to the existing cap and plug standard, rather
than establishing it as a LDAR program action level. These commenters suggested that if the
EPA determines that OELs should become subject to a 500 ppm leak definition, the EPA should
treat these OELs like other pieces of equipment and provide a period of time to repair the leak,
rather than considering the leak to be an immediate non-compliance. Several commenters agreed
and stated that to the extent that the EPA should propose a monitoring schedule and a traditional
LDAR work practice standard of 5 days for a first attempt and 15 days for a final repair.
Response 3: We did not propose periodic instrument monitoring of OELs at petroleum
refineries, and we are not finalizing such a requirement in this rulemaking. If we determine that
periodic instrument monitoring of OELs at petroleum refineries is appropriate to add to Refinery
MACT 1 at a later date (e.g., as part of a future technology review), we will propose those
requirements and document our analyses to ensure that there is an opportunity for public
comment.
Comment 4: Two commenters stated that the proposed requirement to "seal" OEL is a
significant change and would result in an estimated 20 percent increase in LDAR costs
associated with the proposal. The commenters suggested adopting the Texas regulations, which
require OELs to be equipped with a second valve, cap, plug, or blind flange. This would make
Refinery MACT 1 consistent with requirements already in place, avoiding additional
unnecessary compliance costs.
Response 4: The provisions of Refinery MACT 1 currently require that OELs be equipped with
a cap, blind flange, plug, or a second valve. Therefore, no change was made to the requirements
in Refinery MACT 1 as a result of this comment.
Comment 5: Several commenters stated that if the clarification of "seal" is finalized as
proposed, at least two years should be provided for compliance, allowing time to identify and
properly tag all OELs and to equally distribute the monitoring throughout the year. One
commenter agreed with the need for additional compliance time, but requested 3 years be
provided.
173
-------
Response 5: We are not finalizing any changes to the rule language related to the requirements
for OELs subject to Refinery MACT 1. Therefore, the additional compliance time requested by
the commenters is not necessary.
6.4 Requirements for pressure relief devices
Comment 1: Commenters stated that the definition of relief device includes a wide universe of
relief devices, not all of which should be subject to standards, particularly those with little
potential for loss to the atmosphere (e.g., chemicals with very low vapor pressures). Commenters
provided specific recommendations for exemptions from the work practice standard for PRDs
including:
Relief devices for ethylene glycol, polymerizing materials, heavy liquids, and those with
less than 5% HAP.
Relief devices in liquid service, as commenters argued providing this exemption would
be consistent with previous rulemakings relative to handling under the LDAR provisions,
noting that existing LDAR monitoring requirements and reporting requirements under the
Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA)
and EPCRA regulations are sufficient to identify, notify, calculate, and report
unauthorized releases from these relief valves. Commenters also noted that the EPA has a
long history of considering the physical properties of HAPs when making control
requirement decisions, and provided storage tank and wastewater rules as examples (e.g.,
HON, MON). The commenters suggested that the EPA compare Table 1 of subpart CC to
Table 9 of subpart G volatile organic hazardous air pollutant (VOHAPs) and exclude
those non-VOHAPs from the pressure release management requirements. Alternatively,
the commenters suggested that the EPA could just refer to Table 9, as well as exclude
equipment in heavy liquid service from the requirements.
Thermal PRDs which are liquid-only relief valves used to manage thermal expansion of
liquid and are often located in difficult to access and/or remote areas of the refinery and
are typically associated with equipment that is not continuously monitored. The
commenters also asserted that the expected emissions from such devices are small.
High/low point bleeds which, the commenter stated, are used to clear lines for important
maintenance purposes and result in no material HAP emissions.
Another commenter requested that if the proposed 40 CFR 63.648(j)(3) applies to liquid RVs,
those routed to process drains be excluded just as gaseous RVs routed to control are excluded
consistent with the BAAQMD rule.
Several commenters added that the EPA must demonstrate that the change to apply these
requirements to liquid PRD is justified and demonstrated in practice under the provisions of
CAA section 112(d)(6) and no such demonstrations have been made. One commenter stated that
liquid RVs are regulated by 112(d)(2) and no basis is provided for overriding these section
112(d)(2) sensory leak check requirements and converting the current liquid RV work practice
requirement into a prohibition and violation. The commenter argued that if the EPA proceeds
with imposing atmospheric RV monitoring requirements such monitoring must be limited to RVs
in gas-vapor service. The commenter added that that the EPA must, at a minimum limit the
174
-------
pressure relief management to those devices with a reasonable expectation that a possibility
exists for significant organic HAP emissions.
One commenter stated that liquid RVs present much less environmental risk relative to the
increased safety risk resulting from requiring control of those not already controlled, and several
commenters stated that they are much more difficult and costly to safely monitor, if such
monitoring is feasible at all. One commenter concluded that their inclusion is not warranted and
certainly has not been justified. The work practice requirements associated with the Process
Safety Management (PSM) and Risk Management Plan (RMP) programs and the audio, visual
and odor monitoring already required by the applicable equipment leak rules are more than
adequate to assure these RVs are properly handled.
Response 1: As a practical matter, we note that the PRD standards only apply to PRD in organic
HAP service (as defined in §63.641), so the provision is already limited to PRD associated with
process streams with a HAP content of 5 percent or more. The majority of the other comments
summarized here relate to the applicability of the PRD provisions to PRD in light liquid or heavy
liquid service. First, we are not revising the applicability of the operating and release
requirements contained in §63.648(j)(l) and (2) to ensure the PRD is properly sealed. For
reasons provided in the preamble of the final rule, we are promulgating a work practice standard
for PRD releases in place of the proposed prohibition on atmospheric releases and are providing
exemptions for PRD in heavy liquid service and for liquid thermal expansion valves (see Section
IV.C.4 for the preamble to the final rule for additional details). We generally do not consider low
point or high point bleeds to be PRD. If a PRD is used to release material (i.e., used as a low
point bleed), then the PRD release would be subject to the work practice standard unless the
liquid release is specifically exempted (e.g., heavy liquid release or the "light liquid" hard-piped
to a controlled drain system). As discussed in Section 4.1 of this document, we consider most
high point bleeds to be an MPV. If a PRD is used as the release point of a high point bleed, the
high point bleed would be subject to the work practice standards (and would not be MPV, based
on the "exclusion 2" in the definition of MPV.
Comment 2: One commenter stated that the requirements for PRDs should include de minimis
thresholds to exclude systems with little potential for loss to the atmosphere. The commenter
suggested that EPA set a de minimis reporting value equal to the reportable quantity (RQ) values
in EPCRA and/or CERCLA. The RQ values vary from chemical to chemical, and the RQ value
is typically lower for more hazardous and/or volatile chemicals. The commenter stated that the
EPA should re-propose a work practice standard that applies only to devices with the potential
for significant gas/vapor emissions of the Table 1 HAP. Significant would mean a release of a
Table 1 HAP above an established RQ value.
Response 2: We considered the commenter's suggestion to provide a de minimis release quantity
for the PRV requirements. We disagree that RQ values from EPCRA and/or CERCLA should be
used as there is some dispute on how these apply to petroleum derivatives in the refinery. We
also are concerned with a post-event exclusion evaluation that generally requires engineering
calculations and assumptions. However, we understand there may be very small PRD on certain
lines, such as a PRD on a sample analyzer line to protect the analyzer that could not, considering
the size of the valve opening and design release pressure, release more than 72 pounds per day
175
-------
VOC. We are using the size threshold for MPVs since these PRD releases, would be considered
Group 2 MPV and would not require additional control. Consequently, we consider it reasonable
to exclude these very small PRD from the pressure release management requirements, consistent
with the control requirements for miscellaneous process vents.
Comment 3: Three commenters recommended that small mobile tanks should not be subject to
the PRV limits on HAP emissions, as the openings on these tanks are too small to meet the
technical specifications (regarding calculation of inlet losses) adopted by API for routing to a
flare. One of the commenters requested that atmospheric relief valves on portable and mobile
tankage including trucks, railcars, and marine vessels should be excluded from the relief valve
control and/or monitoring requirements. The commenter stated that RVs on this mobile
equipment are regulated by the Department of Transportation (DOT) or the Coast Guard (for
marine vessels) and prohibiting their release to the atmosphere and modifying them to allow
monitoring requires at least consultation with those Agencies and, likely, revision of their
regulations, which the commenter asserted has not been performed. Furthermore, the commenter
added, the openings on these tanks would typically not be sized large enough to route to a flare
under the industry inlet loss standard calculation for a RV.
The commenter stated that loading and unloading of this equipment is regulated under the
various air regulations that deal with transfer operations and/or receiving tankage and that the
risk of a release at other times is very low since these tanks are not an active part of a process.
Furthermore, the commenter state, these RVs are sized for release to the atmosphere since that is
the only option during transport and thus routing to another disposition would reduce their ability
to relieve in the case of a fire or other emergency. The commenter asserted that the increased
safety risk, their temporary nature, their need to be easily moved, and their construction make it
difficult to route these RVs to control or to continuously monitor them.
Response 3: Except for equipment needed to control loading and unloading activities, we
consider that the construction and safety design for mobile equipment regulated by the DOT or
the Coast Guard to be adequate to ensure the safety and minimize releases. We also note that,
due to their size and typical pressures, this equipment typically has a limited potential to emit
(PTE).
Comment 4: One commenter suggested that atmospheric relief valves on temporary
equipment be excluded from the relief valve control and/or monitoring requirements. Various
pieces of equipment are used by refineries for short periods of time to allow for equipment
maintenance (e.g., temporary storage or temporary pump), as temporary controls (e.g., portable
thermal oxidizer or diesel engine for degassing a tank), or to temporarily offset equipment issues,
such as fouling or internal damage (e.g., temporary exchanger to provide additional cooling). All
of this equipment will have RVs, designed for atmospheric release. This equipment, by its
nature, will not be in a location that is already equipped with CVS connections or monitoring
instrument connections. In some cases, the need for this temporary equipment will not have been
anticipated. Even where small distances are involved such connections likely could not be
designed and installed in the time the equipment is in use, nor could the RVs be resized for relief
to a closed system quickly, and the expense for these changes certainly cannot be justified for the
low risk of a release in that short time. The commenter requested that RVs on temporary
176
-------
equipment (i.e., equipment that is in use for less than 12 months) be excluded from the new
atmospheric RV pressure relief requirements.
Response 4: As described in greater detail in the preamble to the final rule, we are establishing
work practice standards based on the implementation of prevention measures to limit the
occurrence of releases and the magnitude of releases when they occur. Consequently, we are less
concerned with the ability to pipe this equipment into a CVS (although for longer "temporary"
service periods, this can still be accomplished), and more concerned with how to limit the
likelihood that a release will occur. Unlike mobile equipment that are separately regulated by
DOT and the Coast Guard and that are not directly tied to the process units, this temporary
equipment is not otherwise subject to regulations and often becomes an integral part of the
refinery process unit. The work practice standards we are finalizing in this rule requires
prevention measures be implemented for this equipment to reduce the chance of a release and to
minimize the size of the release if it occurs similar to "permanent" process equipment.
Comment 5: One commenter requested that low pressure RVs and RVs handling oxygen-
containing vapors be excluded from the RV control and/or monitoring requirements. The
commenter stated that P/V vents on storage vessels that are controlled with CVSs routed to
control (whether for compliance with regulation of not) often serve as RVs for those vessels,
should the CVS system be blocked. Typically these type vents are excluded from RV
requirements by excluding RVs with a design relief pressure of greater than (we assume the
commenter meant "less than") 2.5 psig from the definition. While the definition of relief valve in
Refinery MACT 1 would exclude storage tank P/V vents in their normal service (diurnal
pressure relief), it would not exclude them on tanks controlled with a CVS and control
device. The commenter requested a specific exclusion for RVs with a design release pressure
below 2.5 psig needs to be added to the RV definition.
The commenter added that there are technical challenges in applying the 500 ppm LDAR
standard definition to the P/V vent on a fixed-roof tank. These weighted-pallet type P/Vs do not
close to that standard of leak tightness, while higher set point RVs do not provide adequate
protection against tank failure. Similarly, some RVs protect equipment in services where oxygen
is present, such as in most wastewater management units. These RVs cannot be routed to
combustion controls without potentially explosive results and generally have such low levels of
condensables present that routing to other types of controls is ineffective. The commenter
requested that the RV definition be modified to exclude RVs where unsafe levels of oxygen are
present.
Response 5: We are not finalizing the prohibition on atmospheric releases, which is the
provision in the proposal that spurred these comments but are instead finalizing work practice
standards for these PRD releases. For the reasons discussed in the preamble, we determined it
practical to exclude PRD with design release pressure of less than 2.5 psig from the pressure
release management requirements as requested by the commenter.
Comment 6: Two commenters recommended that 40 CFR 63.648(j)(4) exclude RVs that are
routed to a process or to fuel gas. The commenters stated that most leakage from RVs is routed
to process dispositions including to fuel gas systems. As done throughout Refinery MACT 1 and
177
-------
in the referenced equipment leak rules (part 60 subpart VV and part 63 subpart H), these
alternatives need to be included in 40 CFR 63.648(j)(4). In these situations, as with RVs routed
to control devices, leakage and releases will not reach the atmosphere and, thus, leakage and
release requirements are not appropriate. Furthermore, recovery of this leakage is
environmentally beneficial relative to destruction in a control device and should be, and has
historically been, encouraged.
The commenters added that eliminating the route to process and route to fuel gas alternatives
will create substantial problem, because it will essentially eliminate the exception for most
currently controlled RVs, since most controlled RVs are routed to fuel gas. The piping and,
maybe, the disposition would have to be modified for most existing installations to allow the
monitoring required by this proposal if these RVs were no longer excluded. Compliance time,
costs, and burdens for such changes are not addressed in this rulemaking and are not justified, in
any case, and thus, there is no legal basis for not allowing the same control options as allowed by
the current Refinery MACT 1.
Response 6: We agree with the commenter, and we did not intend to treat PRD that are routed to
the fuel gas system any different from those routed to a CVS. We generally expected that PRD
would not typically vent to the fuel gas system as the fuel gas system operates at much higher
pressures than the flare gas system or other closed vent system. However, we agree with
commenters that, if the PRD can be routed to the fuel gas system, then the exclusions 40 CFR
63.648(j)(4) should apply and we are revising this paragraph to specifically indicate this.
Comment 7: One commenter stated that there are many small atmospheric PRDs in refineries
that pose little release risk and would be costly, difficult or infeasible to monitor or to outfit with
equipment based preventative measures. The commenter suggested excluding PRDs which have
a diameter of less than or equal to 2 inches from the equipment related portions work practice,
but not the RC/CA portion. The commenter asserted that this exclusion will focus the costs for
additional hardware on those PRDs that would have higher emissions in the event of a release
while still addressing the smaller valves in the event of a release to the atmosphere from one of
them.
The commenter explained a large number of these size PRDs are thermal relief valves (discussed
in a comment above) and that another large population of PRDs with a diameter less than 2
inches are on instruments such as process analyzers. These are typically very small
(e.g., 0.25 inches) and protect against blockage within the instrument. The commenter asserted
that monitoring and equipment based preventive measures are infeasible or inordinately
expensive relative to the potential release risk.
Response 7: A 2-inch valve can release a significant amount of material in a relatively short
period of time, so we disagree with the commenters that suggest these types of releases are
"small." As discussed previously, we understand there may be very small PRD on certain lines,
such as a PRD on a sample analyzer line to protect the analyzer, that could not, based on valve
size and maximum pressure, release more than 72 pounds per day VOC, and we have excluded
these PRD from the pressure release management requirements, consistent with the control
requirements for miscellaneous process vents.
178
-------
6.4.1 Pressure release (remonitor) requirement
Comment 1: One commenter asserted that there is no valid reason to add a Method 21
monitoring requirement to proposed 40 CF 63.648(j)(2)(ii) or to remove the current delay-of-
repair provision. The commenter stated that proposed 40 CFR 63.648(j)(2)(ii) reflects the
requirements currently imposed by Refinery MACT 1 for rupture disk PRDs and PEDs that have
a rupture disk associated with a valve to assure no leakage when the relief valve is not relieving.
The proposed language, however, imposed a new monitoring requirement not in the existing rule
and does not include language referencing the delay of repair provisions. The commenter stated
that neither of these changes is explained in the preamble and neither is justified; therefore the
commenter requested these changes not be finalized.
One commenter also added in general, rupture disks are not used in refineries as PRDs, but are
used upstream of a valve to assure that there is no leakage when the PRD is not relieving.
Therefore, monitoring serves no purpose and the current equipment leak rules, including
Refinery MACT 1, do not require such monitoring. This proposal would impose a 5 day Method
21 monitoring requirement in addition to the 5 day rupture disk replacement requirement. The
commenter argued that there is no emission or compliance basis for this change, since leakage is
not possible.
The commenter also stated under Refinery MACT 1, delay-of-repair of a PRD system that
includes a rupture disk is allowed if the conditions in 40 CFR 60.482-9 or 40 CFR 63.171, as
applicable, are met. This provision is critical for rupture disk PRD systems, because it may take a
process shutdown to change the rupture disk. In such cases, there is little environmental reason
not to allow delay because the rupture disk is a voluntary option and the facility could have just
the PRD, as long as the less than 500 ppm limit was met. The commenter requested that 40 CFR
63.648(j)(2)(ii) include the delay-of-repair language from NSPS VV and HON, or provide that
after a pressure release the owner/operator must within 5 days either replace the rupture disk or
demonstrate a leak rate of less than 500 ppm. This would be equivalent to including the current
delay-of-repair requirements for this particular situation.
Response 1: We proposed monitoring of PRD after a release to ensure the PRD is properly
seated following the relief event. We agree that, if a rupture disk is used in conjunction with a
direct atmospheric PRD and the atmospheric PRD is properly sealed, then it is unnecessary to
replace the rupture disk immediately. In the final rule, we are revising 40 CFR 63.648(j)(2)(ii) to
allow the owner/operator to either replace the rupture disk or demonstrate a leak rate of less than
500 ppm must no later than 5 calendar days after the PRD returns to organic HAP service.
6.4.2 No atmospheric discharges
Comment 1: A number of commenters objected to the prohibition of emissions from PRDs to
the atmosphere and further stated that EPA has no authority under section 112 to set a zero
emissions standard for these devices. Several commenters stated that the EPA has improperly
misconstrued the holding of Sierra Club v. EPA, 551 F.3d 1019 (D.C. Cir. 2008) by prohibiting
emissions to the atmosphere from PRDs. Commenters asserted that the Sierra Club ruling does
not prohibit SSM emissions and does not require that a single standard apply at all times.
179
-------
According to commenters, the ruling stated that the emission limits under section 112 of the
CAA must be continuous, which can be numerical, or where the statutory criteria are met, a work
practice can be established when the source is in operation. Commenters concluded that although
EPA claimed in the proposal that it cannot set an emissions standard for these malfunction
emissions, the proposed prohibition on emissions from atmospherically vented PRDs is in fact an
emissions standard. Therefore, the commenters argued stated that the EPA's legal justification is
inherently contradictory and, therefore, fatally arbitrary.
Commenters stated that PRDs are currently regulated through the equipment leak provisions
which apply during normal operations as well as to leaks which may occur if a valve does not
reseat after discharge. Commenters added that the EPA may revise the existing MACT standards
only under section 112(d)(6) authority and only "as necessary (taking into account developments
in practices, processes, and control technologies)." This commenter argued that the EPA has not
made a showing in the proposed rule that a revision of the existing standards is "necessary" upon
consideration of the specified factors. It is well established, according to commenters, that a
section 112(d)(6) standard must be cost effective and technically feasible. These commenters
argued that a zero-emission standard is plainly not cost effective or always technically feasible.
Therefore, the commenters concluded that no change to the existing standards is warranted.
Other commenters noted that the EPA's assertion that a zero-emission standard is grounded in
section 112(d)(2)/(d)(3) fails because the "MACT floor" for atmospherically vented PRDs is not
zero, and a prohibition on emissions cannot be justified as an "above the floor" standard.
Several Commenters suggested that a work practice standard would be a more legally
supportable and reasonable approach. One commenter explained that because the timing, nature,
and extent of PRD emissions are inherently unpredictable, there is ample justification for the
EPA to conclude that it is not technically or economically practicable to measure PRD
emissions. The commenter asserted that because these releases are not malfunctions developing a
work practice standard rather than a numerical emission limitation, as provided in section 112(h),
is more appropriate.
On the other hand, another commenter expressed support for the proposed elimination of venting
refinery gas to atmosphere during a malfunction, stating that as the EPA recognized, the
limits established under 112(d) of the CAA apply at all times. The commenter stated the EPA
must set limits on all HAP sources as a result of National Lime Association v. the EPA (233 F.3d
at 642). The commenter asserted that the proposal will close a regulatory gap that currently
allows certain PRDs to vent emissions directly to the atmosphere during a malfunction.
Response 1: As proposed, the requirement was to control atmospheric PRD. In other words,
PRD releases could not be vented directly to the atmosphere, but must either be eliminated or
vented to a control device, such as a flare. Thus, the proposed requirement did not purport to
limit PRD to zero emissions, but rather that such emissions must be routed to a control device,
and any control device used, such as flares, would not achieve 100 percent control efficiency.
Direct atmospheric PRD releases were previously covered in the facility's SSM plan, which we
consider to have been considered illegal under the Sierra Club v. EPA, 551 F.3d 1019 (D.C. Cir.
180
-------
2008) ailing. We are not revising the MACT provisions for PRD as a result of our technology
review, we merely contended that, in the absence of a legal SSM plan, direct PRD releases
needed to be specifically addressed in the rule because they are sources of HAP emissions at
petroleum refineries. As such, we maintain that the pressure relief management requirements are
appropriately being developed under CAA 112(d)(2) and (d)(3), rather than (d)(6).
We agree with commenters that suggested that the CAA 112(d)(2) and (d)(3) standards must be
based on a MACT analysis. Therefore, we conducted a complete MACT analysis for PRDs. As
described in further detail in the preamble to the final rule, we concluded that a work practice
standard for regulating PRD emissions is justified under CAA section 112(h) and we are
finalizing a work practice standard based on the best performing PRD sources. See the preamble
to the final rule for more information regarding the requirements and our rationale for the PRD
work practice standards we are finalizing.
Comment 2: Several Commenters asserted that the venting of a PRD is not a malfunction as
defined in section 63.2 of the CAA and therefore the EPA does not have legal authority to
prohibit emissions to the atmosphere from PRDs. Commenter argued that the EPA cannot justify
its rule by stating that PRDs are malfunctions without any explanation as to how PRDs meet all
elements of the malfunction definition.
First, commenters asserted that PRDs are not malfunctions because emissions from these devices
are not "excess emissions" but "normal" operation of these devices because PRDs are designed
precisely to allow for the immediate venting of gas from process equipment in order to avoid
safety hazards or equipment damage. Commenters noted that while the release may be due to a
process upset, the regulatory definition of malfunction requires excess emissions.
Second, commenters argued it is unclear how the "not reasonably preventable" criteria is met
since the entire design premise of PRDs is to assure that unpreventable pressure buildups,
particularly those associated with highly unlikely occurrences, are relieved and do not cause
injury, death or property damage. The commenter concluded that if an emission were reasonably
preventable it would not be released through a PRD.
Third, commenters stated that the EPA has provided no evidence that emissions from PRDs
generally will be the result of sudden and infrequent failures.
One commenter stated that if EPA treats atmospheric PRD emissions as malfunction emissions
and proceeds with setting standards under section 112(d), then arguably, atmospheric PRDs are
not part of the regulated stationary source until a specific valve actually experiences a release
and the EPA would lack authority to regulate PRDs under section 112(d). This is because a
stationary source only includes pollutant-emitting activities that emit or have the "potential to
emit" a HAP. "Potential to emit" (PTE) does not include malfunction emissions, and is generally
computed based on normal source operations. As such, if the PRDs release only malfunction
emissions, these relief valves have no potential to emit HAP.
On the other hand, commenters asserted that PRD releases occur during upset conditions, but
that the actual release of emissions during the event is "normal operation" for a PRD. By the
181
-------
EPA's definition, a PRD is designed as a valve used only to release an unplanned, nonroutine
discharge. 40 CFR 63.641. Although relieving during an unplanned event is "normal operation"
for the PRD, it is not "normal operation" for the process equipment that the PRD protects and,
therefore, is not included in the PTE determination of that process equipment. According to the
commenter, establishing PRD releases as "normal operations" might provide the EPA with
authority to regulate these emissions under section 112(d). Nonetheless, under this interpretation,
the commenter asserted that the EPA must conduct a MACT floor analysis and provide this
analysis to the public for review and comment before finalizing a section 112(d) regulation.
Commenters suggested that PRD releases continue to be treated as equipment leaks and an
alternative work practice standard be established in conjunction with the general duty
requirements in the existing MACTs. According to commenters, this would be legally justified
and would avoid interfering with PRD operations during emergencies, large investments and
incurring increased emissions to address emissions that occur infrequently.
Response 2: We believe that the arguments about whether or not emissions from PRD are
malfunctions is somewhat of a red herring as it pre-supposes that EPA can only regulate PRD
releases if they are malfunctions. And, we disagree with such an assumption. EPA certainly has
authority under section 112 to regulate emissions of HAP, and, while we do not believe that we
are required to regulate malfunctions emissions under section 112, we do not believe that we are
prohibited from regulating such emissions. We also disagree with the commenter's assertion that
PRDs have no PTE and therefore are not an affected source subject to the MACT emissions
limitations. Atmospheric PRD at petroleum refineries can release HAP both when they do not
seat properly (equipment leak) and when they open "intentionally," and we considered that, for
this source category, it is necessary to regulate these HAP emissions sources. As we have
explained in more detail in the preamble to the final rule, we are not finalizing the prohibition on
atmospheric releases from PRD, which appears to be the fundamental concern of the
commenters, but are instead establishing work practice standards. See the preamble to the final
rule for more information regarding the requirements and our rationale for the PRD pressure
relief management standards we are finalizing.
Comment 3: One commenter stated that if the EPA proceeds to regulate PRD emissions under
section 112(d), and sets a no discharge standard for atmospheric PRDs, then the EPA should
explain how it considered these adverse impacts in setting such a proposed standard. The EPA's
proposal does not provide a rationale for distinguishing between classes, types and sizes of PRDs
in setting the standard, but according to the commenter, there are two different standards: PRDs
to flares, and atmospheric PRDs. PRDs routed to a flare must meet only 98% control efficiency
or contain less than 20 ppmv HAP, while those venting to atmosphere must meet 100% control
efficiency. The commenter stated that it is appropriate for the Administrator to distinguish
between classes, types and sizes of PRDs for purposes of setting a section 112(d) standard, but
not in the manner in which the EPA proposed. The commenter suggested that instead, the EPA
should look at the technical feasibility and ability to control the PRDs in distinguishing
appropriate classes, types or sizes for determining the MACT floors. The commenter noted that
this has been the EPA's standard practice on all MACT standards including the original Refinery
MACT under 40 CFR subpart CC.
182
-------
Rather than relying on section 112(d) for its legal authority, which would entail an extensive
analysis of MACT floor and other considerations, the commenter recommended that EPA rely on
either section 112(h) or section 112(r)(7) to establish work practice standards
Response 3: As provided in more detail in the preamble to the final rule, we are not finalizing
the prohibition on atmospheric releases from PRD but instead are establishing work practice
standards. We disagree with the commenter that we can rely on section 112(h) to set a work
practice standard in place of relying on section 112(d). We may establish work practice standards
as allowed under section 112(h), but these work practice standards must meet the definition of
MACT as provided in section 112(d)(2) and (3). As noted previously, we identified the
BAAQMD and SCAQMD rules to represent the "best performers" within the meaning of section
112(d)(3). We also disagree that we established two different standards, one for controlled PRD
and one for atmospheric PRD. In the proposal, we established a single standard that required that
PRD releases be vented to a control device and we clarified that, if the emissions were not vented
to a control device, it was a violation of the standard.
Comment 4: Several commenters expressed safety, feasibility, and cost concerns if all PRDs in
organic HAP service are routed to existing and new flares.
Commenters detailed a list of concerns for routing PRDs to existing flares including:
a. The existing flare system would have to have sufficient additional hydraulic capacity for both
the individual relief loads and common mode relief loads, without raising the backpressure on
other relieving PRDs. This could require additional flares to increase pressure relief capacity,
which will increase net annual emissions because of the continuous pilot, purge and sweep gas
combustion required.
b. The flare sterile area would need to be adequate for an increase in individual or common mode
flare loads.95
c. The flare would need to be able to meet the other applicable regulatory requirements (e.g.,
smokeless operation, velocity, combustion zone heating value) at the increased load, which is of
particular concern if the underlying circumstance giving rise to flaring is an emergency event.
d. Compatibility of fluids (freezing issues from fluids that can auto-refrigerate with water
containing streams for example, hot and cold stream).
95 Due to the thermal radiation reaching the ground during a major flaring event, which can happen at any time and
without warning, an area around the base of an elevated flare is treated as a no entry zone. This area is often referred
to as the sterile area. Providing this sterile area is the reason flares require significant plot space, which often is
unavailable near the equipment the flare serves and which will be a major concern if new flares are required as a
result of this rulemaking.
183
-------
e. Relief valves may have to be elevated so that they are free draining and are sloped into the
flare header. Increasing the height will increase the inlet pressure loss which may violate the 3%
rule and also lead to RV instability (e.g., chatter with the potential for loss of containment).
f. Over-pressure scenario could occur in which the refinery vacuum tower could experience loss
of cooling water at the vacuum ejector discharge condensers as a result of routing all RVs in
organic HAP service to existing and/or new flares. The commenter stated that this scenario
would result in very significant amounts of steam being discharged to the flare and have negative
effects on the flare system.
g. "Stepped" RVs will be banned by this proposal. "Stepped" RVs are when a primary RV is
routed to control and additional RVs with somewhat higher set pressures routed to the
atmosphere.
h. These controls require very large investments (millions of dollars per RV) in new flares and
flare header systems.
i. Space may not be available for siting new flares.
j. The surrounding community objects to the addition of flares.
Two commenters also explained that for non-thermal relief, liquid RVs, the flare system would
also need to meet the following requirements:
k. The column foundation and structure would need to be strong enough to support a liquid full
column;
1. The flare knockout drums would have to have sufficient capacity for the identified liquid
overfill relief rate;
m. The flare piping would need to be strong enough to handle the liquid weight and flow regimes
associated with the liquid relief and any coincident vapor flows to the flare;
n. The flare system would have to have adequate capacity to vaporize and/or recover the liquid
and is designed to handle any associated cold safety issues.
The commenter suggested that when the EPA considers these types of technical and safety
concerns, the EPA should appropriately identify classes, types or sizes of PRDs and would likely
determine that certain PRDs, which currently release to the atmosphere, are distinguishable from
other PRDs for the purpose of setting standards.
Response 4: We recognize many of the concerns raised by the commenters about routing all
PRDs to flares. For example, low pressure PRD on storage vessels cannot relieve to a flare
because the flare header may often be at a higher pressure than the storage vessel PRD. To
address these concerns (as well as concerns expressed by other comments in this section), we are
not finalizing the prohibition on the atmospheric releases of HAP from PRD, which is the basis
184
-------
for these concerns. Rather, we considered what the best-performing facilities do with respect to
atmospheric PRD and developed a work practice standard for PRD releases (as described in
more detail in the preamble). This work practice requires the use of 3 prevention measures as
well as root cause analysis and corrective action for releases. In the development of the MACT
standard, we did consider PRD type, size, design pressure, and type of service (vapor versus
liquid service). We expect that the final requirements for PRD resolve many of the potential
issues identified by these commenters. See the preamble of the final rule for more details
regarding the final pressure release management requirements.
Comment 5: A few commenters stated that the prohibition of any pressure release from a relief
device to the atmosphere directly violates the general duty provision in proposed 63.642(n) of
Refinery MACT 1. A few commenters added that while this general duty applies to owners and
operators, the EPA must allow for safe operation of refineries by not prohibiting such operations
to ensure compliance with the general duty provisions.
Some commenters noted that because of the safety implications of this change, this proposal
requires coordination with the Occupational Safety and Health Administration (OSHA) and the
RMP Group at EPA to assure there are no inconsistencies between this proposal and the OSHA
PSM Requirements and CAA section 112(r) RMP requirements and current update effort in the
context of Executive Order 16350 that the impacts of this proposal are appropriately coordinated
with safety related requirements. Other commenters stated that EPA should use its authority
under section 112(r) to ensure that the rule is requiring inherently safer technologies (1ST).
Other commenters stated that prohibiting atmospheric safety valve releases will increase the
chance of a catastrophic equipment failure, particularly during the period between promulgation
of these amendments and installation of the new flare systems and additional/larger safety valves
that would be required to control existing atmospheric safety valves. Several commenters
expressed similar concern that the prohibition of PRDs releasing directly to atmosphere will
increase safety risks at refineries from what many of the commenters deemed to be infrequent
and low impact emission sources. A few commenters stated that this prohibition requires
operators to deal with conflicting safety and environmental regulations. One commenter stated
that the need to control a PRD is normally addressed during the complex process of process
hazards analysis (PHA). PHAs take into account the specifics of the situations in order to
properly apply solutions in order of importance for the explicit purpose of protecting life and
property. The commenter asserted that the EPA should not mandate the imposition of a control
system that safety experts may deem to be unnecessary.
Response 5: As noted previously, we are not finalizing the prohibition on the atmospheric
release of HAP from PRD. The work practice standards that we are finalizing address the
concerns expressed by the commenters. See the preamble of the final rule for more details
regarding the final pressure release management requirements.
With respect to coordination with section 112(r) requirements and the need for this rule to
require 1ST, we note that the EPA's regulations on catastrophic releases appear in 40 CFR part
68. We currently have a petition for rulemaking to address 1ST under section 112(r) and part
68. Also, pursuant to Executive Order 13650, EPA has issued a "Request for Information"
185
-------
soliciting public views on the appropriateness of 1ST regulations and is considering pursuing
rulemaking as part of the regulation, standard and guidance modernization effort called for by
the Executive Order. Historically, EPA's authority to address catastrophic releases under the
NESHAP program was viewed as limited under the pre 1990 ('AA. Congress added section
112(r) to address this gap. In light of the extensive history and efforts of the agency on 1ST
specifically and catastrophic accidents generally under the section 112(r) program, and in light of
the statutory structure of section 112, we view the request to enact 1ST provisions in this rule to
be outside the scope of section 1112(d)(2), section 112(f)(2) and section 112(d)(6). Therefore,
the comment suggesting that this rule requi re 1ST is outside the scope of the current rulemaking
6.4.3 Monitoring system requirements
Comment 1: One commenter stated that, to ensure compliance with the proposed
requirements, they will need to purchase and install monitoring equipment on each
affected atmospheric PRD, hardwire the monitoring equipment to the distributed control system
(DCS), and program appropriate alerts and/or alarms. The commenter stated that the cost to
implement this program is not justified given that atmospheric PRDs rarely open. The
commenter argued that the EPA's expectation that atmospheric PRDs be alarmed so that
operators may minimize the venting to atmosphere as quickly as practicable is not sensible
because atmospheric PRDs are most likely to open during an emergency when dozens of alarms
may simultaneously be going off, and an operator's primary concern will be to safely regain
control or shut down operations. Under such circumstances, atmospheric PRDs vent to
atmosphere precisely so that refineries can safely contain the emergency and mitigate risks to
personnel and the public, and to prevent equipment and catastrophic loss of containment. The
commenter added that to expect refineries to minimize venting to atmosphere during extreme
emergencies is counter to an atmospheric PRD's intended purpose and could increase the danger
to personnel, and the public. The commenter stated that the EPA's desire to regulate a critical
safety device as an air pollution control device will increase the health and safety risk from
refineries.
Other commenters similarly noted that the parametric monitoring requirements could prove to be
both challenging and expensive, particularly for liquid PRDs in piping associated with tank
farms or other areas that are not co-located with a process unit. One commenter explained that it
is unclear that flow monitoring is even feasible for all liquid PRDs. The commenters concluded
that, given the high cost of continuously monitoring liquid/thermal PRDs, their low likelihood of
ever relieving and their minimal release potential, the existing work practice emission limitation
provisions of 60.482-8 and 63.169 of part 63 subpart H are more than adequate and appropriate
for this class of PRDs. One commenter asserted that the record for this rulemaking does not
show that continuous instrumental monitoring is required, authorized or justified for liquid PRDs
and that portion of the proposal, therefore, cannot be finalized.
Response 1: After reviewing these comments in light of the requirements in the BAAQMD and
SCAQMD rules, we are not requiring PRD monitoring requirements for: liquid PRD routed to
drain or process piping; PRD with set pressure less than 2.5 psig; PRD in heavy liquid service;
PRD that have a maximum release rate less than 72 lbs/d VOC; thermal expansion relief valves;
186
-------
PRD on mobile equipment. These sources have small emissions potentials, so we consider it
reasonable to reduce the monitoring burden for these PRD
6.5 Compliance time for equipment leak/PRD requirements
Comment 1: One commenter stated that if the PRD prohibition is finalized, a compliance time
of 10 years should be granted to install and configure the hundreds of additional flares
anticipated for compliance. This time is needed for permitting, greenhouse gas reviews, and
construction. Further the commenter stated that in order to modify existing flares to meet the new
requirements in this rulemaking as well as install new flares to address the RV prohibition, the
additional compliance time is needed to minimize significant outages and disruption of fuel
production to ensure flare capacity is maintained for safety purposes.
Another commenter suggested that a minimum of 5 years for compliance be given depending on
the refinery's turnaround schedule as well as an expedited process for obtaining permit and
construction authorizations for the new flare under parts 60, 61, and 63, and providing provisions
to request additional time, if needed.
Finally, one commenter stated that if their proposed work practice approach is adopted, a
compliance period of three years should be granted for installing the requirement monitoring and
instrumentation and the preventative measures. Some commenters requested that 18 months be
provided to develop procedures and systems, train personnel, and institute the suggested root
cause corrective action (RC/CA) approach in the proposed work practice.
Response 1: We are not finalizing the prohibition for atmospheric PRD releases. Instead, we are
establishing work practice standards requiring the use of prevention measures with the goal of
reducing the frequency of PRD releases and reducing the magnitude of releases if a release does
occur. We are providing a 3 year compliance time to implement the prevention measures
required in the PRD work practice standards based on the time needed to perform the hazard
analysis and purchase and install any equipment necessary for the prevention measures.
6.6 Impact estimates for equipment leak/PRD requirements
6.6.1 Emissions Changes
Comment 1: One commenter argued that while claiming there is significant HAP release
potential, the EPA makes no claims that there are, in fact, significant amounts of HAP released
from PRDs or that there is any risk impact from these typically short, very infrequent
occurrences that would justify the billions of dollars in investments and operating costs this
proposed prohibition imposes.
Response 1: We are promulgating the standards for PRD pursuant to CAA sections 112(d)(2) &
(3) and these standards are technology based, not risk based. Neither the amount of HAP
released or the risk potential of the emissions is considered in developing these standards.
However, we note that we do have specific information regarding emissions from PRD from
Component 1 of the 2011 Refinery ICR survey and we evaluated the potential health risks
187
-------
associated with these PRD emissions as detailed in Appendix 13 of the Final Residual Risk
Assessment for the Petroleum Refining Source Sector in Docket No. EPA-HQ-OAR-2010-
0682).
Comment 2: One commenter stated that the proposed regulation would result in more emissions
because of the need to add additional flares. In addition to the natural gas purge, the new flares
would require pilot gas to ensure flared gas is combusted; supplemental fuel gas to meet the Btu
content requirements for efficient combustion; and, steam injection to meet tip cooling
requirements and smokeless combustion requirements. Combusting additional purge gas, pilot
gas, supplemental gas, and fuel to provide the steam for steam injection would result in
additional emissions of regulated air pollutants and GHGs on a continuous basis, e.g. 8760 hours
a year in order to avoid an accidental release that may only occur 5 minutes every 10 years, if
ever. The commenter argued that this results in unnecessary adverse environmental and health
impacts related to increased natural gas production.
Response 2: We agree that the installation of flares to address infrequent emissions from PRDs
could well result in greater harm to the environment than would be provided by controlling the
PRD releases. We are not finalizing the atmospheric PRD prohibition but are instead establishing
a work practice standard that should not result in negative secondary impacts.
Comment 3: One commenter stated that there is no justification to change the equipment leak
control device emissions limitation from 95% to 98%. The commenter explained that the
proposed 63.648(j)(4) requires that the CVS and control device systems used for RVs that are
exempted from the section 63.648(j)(l) through (3) requirements meet the requirements in
63.644. 63.644, which is titled "Monitoring Provisions for Miscellaneous Process Vents" and
contains requirements that reflect those determined to be appropriate for Refinery MACT 1
Group 1 MPVs (i.e., 98% control). However, the commenter argued that those requirements are
more stringent than those determined to be applicable to equipment leaks in the Refinery MACT
1 rulemaking (i.e., 95% control) and those are the requirements that 63.648(j)(4) must reference
unless the change has been justified under the provisions of section 112(d)(6). RVs are clearly
identified as being an equipment leak source in Refinery MACT 1 and they are specifically
excluded from the MPV definition. Even in the proposal, RVs complying with 63.648(j)(4) are
excluded from the MPV definition. The commenter stated that no analysis or justification is
presented for changing the equipment leak control device emission limitation from 95% to 98%
and thus concluded 63.648(j)(4) cannot reference 63.644, but must reference the same control
provisions as are referenced today (i.e., 60.482-10 and 63.172).
Response 3: Historically, PRD have been included under equipment leaks to ensure the valves
are properly seated, and there are no emissions during normal operations. In this mode, the PRD
is an equipment component that may leak and is subject to the equipment leak requirements.
However, when the PRD opens, we consider this mode of operation to be analogous to an MPV.
Consistent with this interpretation, we are finalizing the control device requirement as proposed.
188
-------
6.6.2 Control costs
Comment 1: Two commenters stated that the EPA has provided incomplete costs associated
with complying with the proposed ban on atmospheric PRDs. One commenter stated that as
many as 300 new flares and new header systems may be required to control the EPA estimated
12,000 atmospheric RVs as a result of the proposal.
The commenter stated that in 2005, the BAAQMD evaluated requiring control of the 324
atmospheric safety valves at the 5 refineries in the District and concluded that approach would
require addition of seven flares and have an investment cost of $192 Million ($17.9 Million
annualized over 20 years), with a cost effectiveness of $890,000-1,500,000 per ton of VOC
reduced.96 The commenter asserted that scaling the BAAQMD data to the 12,000 atmospheric
RVs the EPA estimated, would result in a national investment cost of $7.1 Billion 2005 dollars
($663 Million annualized over 20 years) and require construction of 209 new flares. According
to commenters, these costs and the number of new flares would be increased significantly due to
the imposition of the 400 ft./sec velocity limit as proposed in this rulemaking, and the increase in
refinery construction costs which were not applied to the BAAQMD flares. The commenter
stated a current investment estimate would approach $12 Billion as a result of this prohibition.
Another commenter provided a company-specific estimate of greater than $25 million dollars
capitol cost for the addition of at least two flares to control 266 atmospheric relief valves and 1.1
million dollars in annualized operating costs. The commenter also stated that considering the
estimated emissions reductions, the cost effectiveness would be greater than $200,000/ton VOC.
Another commenter provided a company specific estimate of $200 million dollars to connect all
of its atmospheric relief valves to flare systems. The commenter estimated that it would need to
install 25 new flares for a $1.5 billion investment.
Another commenter also stated that additional large costs will accrue from the production losses
associated with process shutdowns to install the new flares and header systems and to add
nozzles and RVs to equipment because of the requirement for additional relief area to
compensate for the added backpressure of relieving into a flare system rather than the
atmosphere.
The commenter also expressed concern about whether land will be available for these additional
flares or if they can be permitted, in light of community concerns over flares. Another
commenter also highlighted concerns associated with obtaining and complying with
preconstruction NSPS requirements and the feasibility for obtaining offsets for facilities located
in nonattainment areas. The CAA requires that the Agency consider these costs and the increased
emissions associated with adding flares, among other issues, in establishing an emission
limitation. The rulemaking record needs to clearly identify that this proposal requires installation
96 Douglas, Victor and Crockett, Alexander, BAAQMD, Proposed Amendments to Regulation 8, Rule 28: Episodic
Releases from Pressure Relief Devices at Petroleum Refineries, Staff Report, November 2005, Pages 35-38. This
cost effectiveness assumes no existing spare flare system capacity, a situation that is generally true and will be
totally true under the proposed extension of the 400 fps velocity limit to emergencies. Even if 100% spare flare
capacity was assumed the BAAQMD estimated costs of $10-80 million for new piping.
189
-------
of additional flares and it must inform the public of that impact. No such analyses have been
performed in support of this proposal to prohibit atmospheric pressure releases.
Response 1: We are not finalizing the atmospheric PRD prohibition but are instead establishing
a work practice standard. Some facilities will need to implement additional prevention measures
on atmospheric PRD and we provide an estimate of the costs associated with the work practice in
the memorandum titled "Pressure Relief Device Control Option Impacts for Final Refinery
Sector Rule" in Docket ID No. EPA-HQ-OAR-2010-0682.
Comment 2: Regarding the EPA estimate of $9.54M (roughly $7,500 per installation), with an
annualized cost of $1.36M per year for the monitoring requirements, several commenters
provided the results of an industry trade group (API and AFPM) survey. The survey indicated
that the cost per wireless PRD monitoring system would range from $12,000 to $16,700 with an
average cost of $14,350 which is approximately double what was included in the rulemaking.
The commenters indicated that wireless systems are not usable at some locations including areas
associated with liquid valves and would require hard-wired systems. One commenter reported
costs of $75,000 - $100,000 for hard-wired systems.
Response 2: First, we are not finalizing a prohibition on atmospheric PRD. Secondly, we are
providing exemptions from the pressure release management procedures (which includes the
PRD monitoring system requirements in §63.648) for devices in heavy liquid service as defined
in §63.641, and those that only release material that is liquid at standard conditions (1
atmosphere and 68 degrees Fahrenheit) that are hard-piped to a controlled drain system or piped
back to the process or pipeline, among others. As a result, we do not expect nearly as many new
monitoring systems to have to be installed as suggested by the commenter and maintain the rule
requirements are cost effective.
6.7 Recordkeeping and reporting requirements
Comment 1: One commenter stated that the EPA cannot demonstrate that new and additional
reporting requirements for releases from relief devices are justified. The commenter argued that
the regulated community is well aware of relief device events that result in emission releases,
and that industry takes action both to report and correct such events. Some commenters cited
specific regulatory programs requiring the reporting of unpermitted releases of HAP including
polyvinyl chloride (PVC) NESHAP, the HON, Group IV Polymers and Resins NESHAP,
Polyether Polyol NESHAP, MON, Polymers and Resins I NESHAP, 40 CFR 63, subpart UU,
Title V Reporting, EPCRA, and CERCLA and asserted that the EPA neither collected nor
analyzed any of this information as part of this rulemaking.
Response 1: We disagree with the claim that regulated communities are aware of these releases
and comments from community groups support the fact that they are not aware. Regarding other
rules that require reporting of unpermitted releases, we note that most of these regulations do not
apply to emissions from sources regulated pursuant to Refinery MACT 1 and Refinery MACT 2.
Thus, it is unclear why the commenter is suggesting that EPA should have collected and
considered that information. We note that the reporting requirements we are establishing in the
190
-------
final rule are consistent with the reporting requirements in the BAAQMD and SCAQMD rules,
which represent what is required of the best performing sources.
Comment 2: One commenter stated that proposed 40 CFR 63.655(g)(10)(iii) requires reporting
information on any pressure release to the atmosphere that occurred during the reporting period.
It calls for reporting "estimate of quantity of substances released." The commenter stated
that "substances" is an ambiguous word and would include materials that are not pollutants and
are not HAPs. The commenter suggested that since section 112 deals with HAPs, this reporting
requirement should be limited to estimates of releases of organic HAPs regulated by Refinery
MACT 1. Other CAA rules and rules authorized under other statutes are already in place to
obtain information on releases of other substances.
Response 2: We agree with the commenter that the reporting requirements 40 CFR
63.655(g)(10)(iii) should be specific to quantities of organic HAP released, and we have made
this change in the finalized rule requirements. Note that this is consistent with the proposed
requirement 40 CFR 63.648(j)(3)(iii) to calculate the quantity of organic HAP released during
the event.
Comment 3: One commenter stated that EPA proposed to add a new paragraph (vii) to 40 CFR
63.655(f)(1) to address NOCS reporting requirements relative to PRD monitoring. However, this
paragraph appears to require information for PRDs that are not required to be monitored (i.e.,
PRDs excluded from monitoring by 40 CFR 63.648(j)(4)). The commenter suggested that the
proposed 40 CFR 63.655(f)(l)(vii) needs to be revised to only address PRDs in organic HAP
service that are subject to the 40 CFR 63.648(j)(3) monitoring requirements. There is no
justification or need for listing those PRDs that are routed to a process, a fuel gas system, or
control.
The commenter also stated that proposed paragraph (10) to 40 CFR 63.655(g) to address periodic
reporting requirements for PRDs appears to apply to all PRDs and not just those subject to
monitoring. The commenter suggested that 40 CFR 63.655(g)(10) introductory paragraph needs
to be revised to only apply the requirements in this section to PRDs that are not excluded from
the requirements of 40 CFR 63.648(j)(l) - (3) by 63.648(j)(4).
The commenter also suggested that 40 CFR 63.655(g)(10)(ii) be modified to read as italicized
below, because, the commenter claimed, the current wording suggests that monitoring of all
PRDs to show compliance is required in each semiannual reporting period. No such monitoring
frequency is specified in 40 CFR 63.648(j).
(ii) For relief valves in organic HAP gas or vapor service subject to 63.648(j)(2), report
confirmation that any monitoring required to be done during the reporting period to show
compliance was conducted.
Response 3: We agree with the clarifications suggested by the commenter. We did not intend to
require reporting of controlled PRD or reporting of monitoring information from PRD not
subject to the monitoring system requirements. Specifically, the following revisions are included
in the final rule,
191
-------
63.655(f)(l)(vii): "For pressure relief devices in organic HAP service subject to the
requirements in §63.648(j)(3)(i) and (ii), this report shall include the information specified
in paragraphs (f)(l)(vii)(A) and (f)(l)(vii)(B) of this section: ..."
63.655(g)( 10)(i): "For pressure relief devices in organic HAP gas or vapor service, pursuant to
§63.648(j)(l), report any instrument reading of 500 ppm or greater, more than 5 days after the
pressure relief device returns to service after a pressure release."
63.655(g)( 10)(ii): "For pressure relief devices in organic HAP gas or vapor service subject to
§63.648(j)(2), report confirmation that any monitoring required to be done during the
reporting period to show compliance was conducted."
63.655(g)(10)(iii): "For pressure relief devices in organic HAP service subject to §63.648(j)(3),
report each pressure release to the atmosphere, including duration of the pressure release and
estimate of the mass quantity of each organic HAP released, and the results of any root cause
analysis and corrective action analysis completed during the reporting period, including
the corrective actions implemented during the reporting period and, if applicable, the
implementation schedule for planned corrective actions to be implemented subsequent to
the reporting period."
192
-------
7.0 Delayed Coking Units
Comment 1: One commenter asked how the refinery rule affects coker units still included in
permits, but shut down due to business decisions. The commenter also asked about coker units
not in the petroleum refinery sector. According to the commenter, there are three Rain CII
calcining plants in Louisiana with their own terminals and coker units. One of these plants is
adjacent to the ExxonMobil Chalmette refinery, yet it will not be subject to the EPA proposed
refinery rules, because it is not in the petroleum refinery sector. According to the commenter,
this plant emits the "lion's share" (over 7,000 tons per year) of the sulfur dioxide emissions said
to designate St Bernard Parish non-attainment for the 2010 NAAQS one-hour health standard for
SO2. Two commenters recommended that the EPA should consider improving pollution controls
on all coker units, including those at coke calcining facilities.
Response 1: Delayed coking units at facilities that are currently not operating for business
reasons will need to meet the emission limitations of the final rule based on the compliance
schedule for an existing source. While delayed coking units may be used to supply coke
calciners, we are not aware of any independent (non-refinery) coke calcining facility that
operates a DCU. That is, we expect that the facility the commenter is referring to likely receives
coke from the adjacent refinery and the coker at that refinery will be subject to the final
requirements for DCU. Coke calciners are very different emission sources than DCU and the
emission controls applicable to coke calciners would be very different as well. Coke calciners are
not a source category regulated under section 111 or 112 of the CAA and the development of
NSPS or other regulations for the coke calcining industry is beyond the scope of this rulemaking
effort.
7.1 Technology review/need for MACT standard
Comment 1: Three commenters supported the EPA's determination that the DCU atmospheric
vent was previously unregulated and it is appropriate to establish a MACT standard for this
source. One commenter suggested that decoking units are archaic and obsolete and that the EPA
should encourage the refinery industry to conduct research to develop a continuous decoking
operation where emissions can be more easily reduced or captured. Other commenters argued
that delayed coking unit vents are already regulated as a MPV. Two commenters claimed that
DCU are clearly sources regulated under Refinery MACT 1 as they are "process units for
thermal cracking." These commenters suggested that the exclusion stating that "coking unit vents
associated with coke drum depressuring at or below a coke drum outlet pressure of 15 pounds
per square inch gauge [psig], deheading, draining, or decoking (coke cutting) or pressure testing
after decoking" is simply an exclusion for when emissions were considered de minimis.
According to these two commenters, this exclusion is analogous to the 20 ppmv concentration
threshold for Group 1 MPV. If these vents are discharged to the atmosphere at a pressure above
15 psig, they would need to be evaluated as MPVs and controlled if the Group 1 criterion were
met. One commenter noted that the background document for the 1995 Refinery MACT rule
suggests that EPA understood its 15 psig threshold not to constitute a determination to leave
delayed coker vents unregulated, but to set the threshold "to encourage vapor recovery" at
facilities using DCUs.
193
-------
Response 1: We disagree that the DCU is an obsolete technology and should be replaced with
continuous coking operations. The refining industry has already developed continuous coking
processes, but these processes generally have significantly higher emissions than DCU and most
refineries have elected to use DCU rather than these continuous coking processes.
We agree with commenters that the DCU are sources that are part of the affected source under
Refinery MACT 1. However, as explained in the preamble to the proposed rule, we believe that
there are two distinct types of vents associated with DCU and these two types of vents are treated
very differently within the existing Refinery MACT rule. The "delayed coker vent" (i.e., the vent
for the DCU blowdown system) is specifically defined and cited as an MPV, which requires the
blowdown system to be controlled rather than open to the atmosphere. The atmospheric vents
associated with decoking operations (the coker "steam vent" as well as releases from deheading,
draining and coke cutting) are completely different venting locations and these vents are
specifically excluded from the definition of an MPV (at pressures less than 15 psig), which
effectively excluded these vents from any regulatory requirements. The EPA did not attempt to
determine the emissions limitation achieved by the top performing decoking operation vents in
establishing the exclusion regardless of whether it might be "to encourage vapor recovery." We
also note that there are no monitoring requirements in the existing Refinery MACT requirements
to ensure compliance with this 15 psig threshold, which further undermines the assertion that this
limit was intended as a MACT standard. It is evident that refineries considered this a full
exclusion for decoking operation vents. See, e.g., ICR responses. In Appendix C of commenter
EPA-HQ-OAR-2010-0682-0583, the commenter indicated that one facility reported that they
typically vent to the atmosphere at 17 psig and did not control this vent as an MPV.
We agree that at the time the MACT standard was developed in the mid-1990s, the EPA did not
make a determination that the emissions from the decoking operation vents were de minimis.
More recently, the EPA has obtained data to truly characterize these emissions. Available data
indicate that the HAP concentration in the atmospheric steam vent is often greater than 20 ppmv
on a wet basis and much greater than 20 ppmv when determined on a dry basis. Available data
also suggests that these vents release greater than 33 kg/day VOC (especially when venting at
higher coke drum pressures). Therefore, we disagree that the 15 psig is analogous to the de
minimis emissions levels used for MPV (i.e., the 20 ppmv threshold).
Because the EPA has concluded that decoking operation vents were excluded from MACT
requirements at the time the standards were initially promulgated, EPA has authority at this time
to determine the appropriate MACT limit for these operations under Refinery MACT 1.
7.2 MACT floor determination
Comment 1: One commenter agreed with the EPA that the DCU standard must be a work
practice standard under section 112(d)(6) because "it is not feasible to prescribe or enforce an
emission standard for the DCU steam vent because the application of a measurement
methodology for this source is not practicable due to technological and economic limitations."
The commenter contended that the EPA is arbitrarily and unlawfully using the MACT standard
(MACT floor) setting approach for a standard where the Agency determined that the MACT
standard setting approach cannot be used, i.e., in establishing a work practice standard. The
194
-------
commenter stated that the EPA is obligated to provide an explanation of how section 112(d)(6)
can or should be construed to require or allow the use of the section 112(d)(2)/(3) MACT
standard setting approach in setting a work practice under section 112(d)(6). According to the
commenter, the failure to provide an explanation of the legal authority for this standard setting
approach violates the Agency's obligation under section 307(d)(3) to provide an explanation of
the basis and purpose for the proposed rule.
Response 1: The premises of the commenter's arguments are incorrect. As an initial matter, the
EPA is developing standards for the DCU decoking operation vents under section
112(d)(2)/(d)(3) not 112(d)(6) because we determined that these vents were inappropriately
excluded from the MACT requirements in subpart CC. Under section 112(d)(6), we concluded
that the current MPV requirements for the "delayed coker vent" were appropriate and that there
were no developments in practices, processes, or control technologies for MPV. Section
(112)(d)(2) clearly allows that the MACT standards can be a work practice standards and section
112(d)(3) states that the minimum (MACT floor) requirements "shall not be less stringent...than
the average emission limitation achieved by the best performing 12 percent of the existing
sources..." While strict application of the MACT floor analysis might create issues in terms of
practical application in the context of setting certain work practice standards, that is not the case
in this instance and use of the MACT floor was appropriate.
Comment 2: One commenter suggested that the EPA must supplement the 2 psig limit with an
additional long-term standard on the annual average maximum pressure of the coking unit prior
to release to the atmosphere. While the commenter noted that the EPA's short-term, never-to-be
exceeded limit is important, the commenter argued that the available data show that the top
performing 12% of coking units actually achieve much higher levels of control on average. For
new units, the law requires EPA to set a standard that is "not less stringent than the emission
control that is achieved in practice by the best controlled similar source." At existing units, EPA
must set a standard based on "the average emissions limitation achieved by the best performing
12 percent of the existing sources." According to the commenter, the ICR data shows that the
typical pressure of the coke drum prior to release to the atmosphere for the best performing
facility and the best performing 12 percent of facilities is 1 psig.
Response 2: We disagree with the commenter that a long-term, annual average standard is
required. Based on our analysis of permit requirements, no DCU is required to meet a short- or
long-term standard. The only existing requirements for DCU are based on pressure limits for a
single cycle. While we agree that a facility achieving a 2 psig emission limitation for each cycle
will likely have an average venting pressure well below 2 psig, that fact alone does not mandate
that we create for existing units a long term limit where one does not currently apply to any of
the top performing sources.
Regarding the comment that we must establish a 1 psig limit for new sources because it
represents the best performing facility, we disagree. The reported "typical venting pressure"
reported by ICR respondents is insufficient evidence to conclude that any facility meets an
annual average of 1 psig. For example, assume a facility operates a DCU vent at about 1 psig 60
percent of the time, 2 psig 35 percent of the time, and 3 psig 5 percent of the time. Such a facility
195
-------
would likely report that they "typically" vent at 1 psig, but this facility would not be able to meet
a 1 psig annual average emission limit.
7.2.1 Existing source alternatives
Comment 1: Two commenters requested that an alternate work practice standard should be
provided for "water overflow" DCUs. Water overflow DCUs are a type of DCU design that
employs a unique process operation in which the coke drum is filled with water and the overflow
from the coke drum is passed through the overhead vapor line. According to the commenters,
this capability must be designed into the DCU and the pressure in the overhead line is not
necessarily a reliable indicator of the extent of bed cooling in this type of operation. For these
DCUs, the temperature of the water in the overhead line would be representative of the
temperature in the top of the coke drum. The commenters requested that water overflow DCUs
should be allowed to comply with the 2 psig depressurization work practice standard if equipped
to make that demonstration or, as an alternate, demonstrate compliance with work practice
standard precluding atmospheric venting prior to water overflow and an overhead temperature of
220 °F (based on the EPA's analysis of bed temperature at 2 psig). One commenter also
requested that, it should be clearly identified that water overflow drums, where water is routed to
an atmospheric tank, while vapors are recovered to the closed blowdown system, are permissible.
Response 1: We agree with the commenter that a 220 °F overhead temperature is at least
equivalent to the 2 psig limit since this is the saturated steam temperature of a system at
equilibrium (i.e., a closed system) at 2 psig. Since the coke drum is actively venting, it is easier
for a traditional DCU to meet the 2 psig pressure limit than 220 °F and that a traditional DCU
achieving an overhead temperature of 220 °F would have emissions equivalent to or less than a
traditional DCU achieving 2 psig pressure limit. Similarly, we find that an overhead temperature
of 218 °F is at least equivalent to a 2.0 psig limit. During our site visits, we asked about water
overflow as a means to help cool the coke bed by allowing more cooling water flow through the
bed; however, the sites visited did not have a system that was designed to do this and it
was unclear if such a system was possible. We are supportive of DCU with water overflow
design, but we are concerned with superheated overflow water being exposed directly to the
atmosphere. We expect that the overflow line would be hard piped to the overflow water storage
tank. Given the height of the DCU overhead pipe, no pump would be needed to allow flow to the
storage tank through a submerged fill pipe. Submerged fill will prevent atmospheric exposure of
superheated water and allow cooling of the overflow water as it mixes with the tank
contents. Therefore, we have revised the provisions of the rule to expressly allow water overflow
DCU to discharge water to an atmospheric tank provided that submerged fill pipe (pipe outlet
below the existing water level in the tank) is used.
Comment 2: Two commenters requested that an average atmospheric vent time of less than 20
minutes should be established as an alternative standard and considered to have satisfied the
depressurization work practice. Short depressurization time is indicative of a well-cooled bed
and low vent emissions, even if the initial release pressure is somewhat above a specified do-not-
exceed limit. Such an alternate would allow for reduced investment in those cases where excess
emission risk is minimal, but back pressure fluctuations are sometimes encountered. The 2010
ICR acknowledged the de minimis emissions emitted by DCUs with such short vent times and
196
-------
provided that these units need not perform stack tests. Short duration vented cokes (defined as
those with average vent duration times of less than 20 minutes on a monthly basis) should be
considered to be compliant. One commenter requested that the EPA should allow a facility, at its
choice, to test and demonstrate that its VOC emissions (as a surrogate for the organic HAPs) are
not appreciably different at venting pressures between 5 psig and 2 psig. According to the
commenter, if the VOC emissions measured from a DCU at 5 psig (or other pressure) does not
result in more than one (1) ton per year of increased VOC emissions compared to emissions at 2
psig, then that facility should be allowed to depressure to 5 psig (or other pressure limit tested).
Response 2: Through site visits, review of source tests, and conversations with delayed coker
operators, we have greatly increased our understanding of DCU coke drum cooling and the
emissions from decoking operations. For the ICR, facilities that initiated draining within 20
minutes of opening the atmospheric steam vent were considered to have venting times of less
than 20 minutes and were not required to test. We now know that venting can occur for some
time after the initiation of draining, so many of these DCU that were not required to test were
likely to have steam vent emissions for more than 20 minutes. Moreover, had we fully
understood the emission sources, including the potential for emissions from the draining area
when draining the very hot and often super-heated water, we would have placed requirements
that the DCU could not be drained until the vent velocity was no longer measurable or 90-
minutes, whichever is shorter. Had we implemented this testing constraint at the time of the ICR,
we do not believe any DCU would have vent times of less than 20 minutes.
CAA section 112(h)(3) provides that "If after notice and opportunity for comment, the owner or
operator of any source establishes to the satisfaction of the Administrator that an alternative
means of emission limitation will achieve a reduction in emissions of any air pollutant at least
equivalent to the reduction in emissions of such pollutant achieved under the requirements of
paragraph (1), the Administrator shall permit the use of such alternative by the source for
purposes of compliance with this section with respect to such pollutant." We have codified
regulations to implement this provision at 40 CFR 63.6(g). Any facility may use that regulatory
provision to demonstrate to the Administrator that an alternative limit is equivalent to the final
work practice standard. The key point is that the alternative emissions limitation must be at least
equivalent to the emissions limitation in the final rule (not "appreciably similar to" or that "result
in more than one (1) ton per year of increased VOC emissions compared to emissions at 2 psig").
Given that our DCU testing requirements would prohibit or restrict draining during the source
test, at this time, we consider it unlikely that a facility will be able to demonstrate that a higher
venting pressure or that a set venting time is equivalent to the 2 psig limit. However, owners or
operators can submit a demonstration pursuant to 40 CFR 63.6(g), if they believe they can make
an acceptable equivalency demonstration.
Comment 3: One commenter requested that compliance with the SCAQMD Rule 1114 should
be deemed as compliance with proposed requirements in 40 CFR 63.657. According to the
commenter, the SCAQMD Rule 1114 is at least equivalent to this proposed
rule; therefore, compliance with that regulation be established as a compliance alternative to the
requirements in 40 CFR 63.657.
197
-------
Response 3: We disagree that the compliance requirements are equivalent because South Coast
refineries with multiple DCU can comply with a 5 psig limit for some time period prior to
compliance with the 2 psig limit. Additionally, the monitoring requirements to ensure
compliance with the 2 psig limit are not identical. For these reasons, we are not providing the
equivalency requested by the commenter.
Comment 4: Two commenters opposed the requirement prohibiting draining of a coke drum
until the 2 psig vent pressure limit is reached. According to the commenters, the reverse flow of
water through a coke bed that occurs by draining the bed allows the water to contact additional
hot coke, reducing cracked vapor emissions and reducing depressurization time. According to the
commenters, limiting pre-vent draining would therefore increase cycle time and reduce coke and
gasoline production. Three commenters stated that the proposed ban also imposes a safety risk
by prohibiting "double quenching of hot drums." In some cases, a coke drum does not cool
properly during the initial quench and overhead temperature cannot be reduced to the allowable
release limit. In such a case, a significant quantity of quench water must be drained to allow
fresh, cold quench water to be added to provide a second "double" quench. According to the
commenters, without the ability to double-quench, the drum would sit idle for indefinite periods
of time, while waiting for it to eventually cool down and once it reached the pressure limit it
would be opened and the coke bed could still contain hot spots that could cause safety concerns
as the coke is removed from the drum. Therefore, the commenters urged EPA to develop a work
practice standard that allows draining prior to achieving the overhead pressure requirement in
order to reduce emissions and allow "double quenching" of hot drums.
Response 4: While we agree that early draining of the coke drum does limit emissions from the
steam vent, as noted in the preamble to the proposal, early draining of the superheated water
increases emissions to the atmosphere in the drainage area. We do understand that there are
times when the normal cooling cycle is not effective and additional water may be needed to
adequately cool the bed. With improved evacuation/blowdown systems, steam loss from the bed
will reduce water levels over time and allow additional water to be added. However, we
understand that if there is significant channeling in the coke bed, adequate cooling of the coke
bed may not be possible using normal cooling algorithms without significant delays in the
overall cycle time. Therefore, we are allowing draining of the coke drum for the purpose of
double-quenching a coke drum that did not cool adequately using the normal cooling process.
DCU owners or operators may drain under certain circumstances, for safety-purposes, provided
that they maintain and operate a temperature CPMS at the bottom of the drum or drain line that
can measure the temperature of the drain water and they maintain the temperature of the water
being drained below 210 °F. This would allow draining and adding of some water (double-
quenching) to improve coke bed cooling without draining super-heated water. Owners and
operators would be required to record the date, time and duration of each pre-vent draining
event, the pressure of the drum for the 15-minute period prior to the pre-vent draining, and the
drain water temperature at 1-minute intervals from the start of pre-vent draining to the complete
closure of the drain valve. Owners and operators must include in their semi-annual report the
number of pre-vent draining events that occurred and, for each instance that the drain water
temperature reached 210 °F or higher, the maximum drain water temperature during the draining
event. We considered setting a limit on the number of pre-vent draining occurrences that could
be performed per year, but we do not consider that appropriate because, as long as high drain
198
-------
water temperatures are avoided, we consider double-quenching to be an effective means to
improve coke bed cooling thereby reducing the average coke bed temperature and thereby
reducing the overall emissions from decoking operations.
7.2.2 New source & NSPS Ja
Comment 1: One commenter stated that the EPA should incorporate the new Refinery MACT 1
DCU standards into NSPS subpart Ja. The commenter stated that it would be unlawful and
arbitrary to delete the Refinery subpart Ja standards because they regulate different pollutants.
Four commenters stated that, if the EPA does regulate DCU decoking operations in Refinery
MACT 1, the EPA should define the standard at a depressurization level of 5 psig, consistent
with requirements identified in 40 CFR part 60, subpart Ja. Three commenters noted that the
EPA previously determined that 2 psig was not cost-effective for new sources and it would be
less cost-effective for existing sources to meet a 2 psig standard. One commenter stated that the 5
psig limit would more accurately reflect a degree of reduction justifiably characterized as
"achieved in practice" and more closely reflect cost-effective, additional HAP emission
reductions for these sources. One commenter noted that lowering the MACT limit undermines a
company's investment made to comply with the 5 psig NSPS subpart Ja limit.
Response 1: As mentioned previously, our understanding of DCU operations has changed
during the process of developing these MACT standards and the methodology used to estimate
these emissions has changed significantly. We have also revised our emissions and cost estimates
based on the comments received. Even after these revisions, we still expect greater VOC and
HAP emissions compared to our earlier NSPS Ja estimates and thus a lower cost-effectiveness
than assumed at the time NSPS Ja was promulgated in 2008. More significantly, however, as
described previously, we have determined that we must establish a MACT floor performance
level for DCU decoking operations. The MACT standard can be no less stringent that the
average emissions limitation achieved by the top performing 12 percent of sources. The EPA
does not consider costs in determining the MACT floor, but does consider cost in evaluating
whether to promulgate more stringent, "beyond the floor" standards. As described in more detail
in the memorandum entitled "Reanalysis of MACT for Delayed Coking Unit Decoking
Operations" included in Docket ID No. EPA-HQ-OAR-2010-0682, we agree that it is not cost-
effective to further lower the DCU decoking operations venting pressure beyond the 2 psig limit,
which we have determined is the MACT floor for existing sources, and thus we are promulgating
a limit of 2 psig averaged over 60 events as the MACT standard for existing sources. We also
note that, based on the comments and information received during the public comment period,
DCU designed to meet a 5 psig limit on a per venting event basis will generally vent at an
average pressure of approximately 2 psig and will be able to comply with the 2 psig limit
averaged over 60 events we are establishing in the final rule.
7.3 Feasibility of 2 psig requirement
Comment 1: One commenter noted that California refineries have already implemented several
emission control measures consistent with U.S. EPA's proposed 2 psig requirements through
compliance with local air district rules and another commenter noted that several refineries in the
Houston area already depressure down to 2 to 5 psig, so the 2 psig limit is achievable. Two
199
-------
commenters supported the 2 psig limit for decoking operations, but recommended that EPA
establish a 0.2 psig limit. Two commenters supported the 2 psig limit and the inclusion of all
decoking operations (venting, draining, deheading, and coke cutting). One commenter stated that
the 2 psig limit may not be achievable for facilities with flare gas recovery systems due to
variability in the pressure of the flare header system. This commenter suggested that a 5 psig
limit would be more appropriate.
Response 1: We appreciate the support for the 2 psig limit for decoking operation vents. We
disagree with commenters that recommended a 0.2 psig limit should be established. As noted in
the preamble to the proposed rule, the low pressure point in the blowdown system is generally
operated at approximately 0.5 psig to prevent air infiltration into the vapor lines. Therefore, it is
not technically feasible to achieve a 0.2 psig pressure in the coke drum while venting to the
blowdown system. With respect to the comment that the 2 psig limit is not achievable for
facilities with flare gas recovery systems, we disagree. We recognize that different vapor
recovery or flare gas recovery systems operate at different inlet pressures and, for remote flare
gas recovery systems, there may be significant back pressure at the DCU. However, there are
flare gas recovery systems in-place that allow compliance with the 2 psig limit. Therefore, we
conclude that it is technically feasible to design and operate a flare gas recovery system to
control emissions from the DCU blowdown system that allows depressurization of the DCU
coke drums to 2 psig prior to atmospheric venting.
7.4 New/revised definitions related to delayed coking units
7.4.1 Delayed coker vent
Comment 1: One commenter noted that the coke drum overhead is not typically "routed to the
atmosphere through the delayed coking unit's blowdown system." Instead, the coke drum
overhead is routed to the closed blowdown system for recovery to the fractionator, fuel gas, or
for routing to a control device (i.e., a flare) until such time as a particular drum overhead
pressure is achieved, at which time the coke drum overhead vapor is routed directly to
atmosphere. The proposed definition seems to imply that sometimes the stream goes to
atmosphere via the blowdown system. One commenter noted that it is unclear from the
definitions, whether a delayed coker vent to the atmosphere is considered a bypass under the
MPV provisions of proposed 40 CFR 63.644(c). The commenter requested that the EPA either
confirm that section 63.644(c) applies to delayed coker vents vented to the atmosphere or amend
and clarify the proposed regulation such that these vents are subject to section 63.644(c). A third
commenter suggested that the definition of delayed coker vent should clarify that the vent
associated with the blowdown system is the miscellaneous process vent and the final steam
release vent is not a miscellaneous process vent.
Response 1: There are several potential release points or vents from delayed coking units and we
are clarifying in the regulation the difference between a "delayed coker vent", which is a
miscellaneous process vent, and decoking operation vents, which were excluded from the
definition of MPV. Prior to Refinery MACT 1, it was our understanding that some units used
open blowdown systems for their delayed cokers and the MPV provisions of Refinery MACT 1
has effectively required all blowdown systems to be controlled. Therefore, we consider any
200
-------
direct atmospheric releases from the DCU blow down system would be considered a
MPV. However, if a control system is in-place on the delayed coker vent and these controls are
by-passed, the atmospheric releases would be considered a bypass of the MPV provisions in 40
CFR 63.644(c). Under the amendments promulgated in the final rule, direct discharges to the
atmosphere prior to or that by-pass the DCU's blowdown system would be subject to
the decoking operation requirements in 40 CFR 63.657. We are clarifying the definition of
delayed coker vent to specifically refer to the vent as a miscellaneous process vent. We are also
clarifying that the delayed coker vent contains the uncondensed vapors from the DCU blow
down system rather than referring to the blowdown system as open or closed. The revised
definition reads:
Delayed coker vent means a miscellaneous process vent that contains uncondensed vapors from
the delayed coking unit's blowdown system. Venting from the delayed coker vent is typically
intermittent in nature, and occurs primarily during the cooling cycle of a delayed coking unit
coke drum when vapor from the coke drums cannot be sent to the fractionator column for
product recovery. The emissions from the decoking operations, which include direct atmospheric
venting, deheading, draining, or decoking (coke cutting), are not considered to be delayed coker
vents.
7.4.2 Decoking operations
Comment 1: One commenter stated that the second sentence of the proposed definition of
decoking operations refers to "steam released from the coke drum is no longer discharged via the
delayed coker vent to the unit's blowdown system," but the definition of delayed coker vent is
limited to vapors that go to the blowdown system, so there is an internal inconsistency in the
statement. The commenter recommended that the term delayed coker vent be deleted from the
definition of decoking operations.
Response 1: In response to comments on the revised definition of "delayed coker vent" we have
clarified that the delayed coker vent is the vent containing uncondensed vapors from the delayed
coking unit's blowdown system. Given this revision in the definition of delayed coker vent, we
agree that there is an inconsistency with the inclusion of the phrase "...via the delayed coker
vent..." Therefore, we have removed this phrase from the definition of decoking operations, as
recommended by the commenter. The final definition of decoking operations reads:
Decoking operations means the sequence of steps conducted at the end of the delayed coking
unit's cooling cycle to open the coke drum to the atmosphere in order to remove coke from the
coke drum. Decoking operations begin at the end of the cooling cycle when steam released from
the coke drum is no longer discharged via the unit's blowdown system but instead is vented
directly to the atmosphere. Decoking operations include atmospheric depressuring (venting),
deheading, draining, and decoking (coke cutting).
201
-------
7.5 Impact estimates for DCU requirements
7.5.1 Emission reductions
Comment 1: One commenter noted that although reducing the venting pressure by 12 psig does
not sound like much, these are very large vessels that release twice a day, so the rule would result
in appreciable emission reductions. Two commenters noted that the available test data are highly
variable and insufficient to assess the impacts of the proposed requirements and recommended
additional test data be collected in order to determine a clear measurable effect of the proposed
control requirements on DCU emissions. One of these commenters stated that the proposed limit
for DCU is premature because the initial steam vent testing was not coupled with any emission
testing of the draining, deheading, or coke cutting, which will include cutting through hot spots,
and because the testing did not lend itself to emission rate differentiation with respect to the size
of the drum or the release pressure. The other commenter stated that the EPA relied excessively
on the results of one source test (Hovensa), which is not representative of industry emissions,
and therefore significantly overestimated HAP emissions.
Response 1: We recognize that there are difficulties in comparing the available emissions data
due to differences in draining times relative to initiation of venting so that the "steam
vent" tested does not necessarily include all of the emissions from the unit. However, the data are
of sufficient quality and scope to indicate that the DCU is a significant source of HAP emissions
and that some of the HAP emitted from the DCU decoking operations (particularly naphthalene
and 2-methylnaphthalene) can contribute significantly to the total cancer incidence from
petroleum refineries. Additionally, the available emissions data indicate that DCU decoking
operations emissions are proportional to the quantity of steam released, which is, based on heat
balance considerations, related to the initial venting temperature or pressure. Therefore, we
maintain that there is sufficient evidence to support the conclusion that lowering the venting
pressure will reduce HAP emissions from these sources.
Comment 2: One commenter evaluated the SCAQMD model and provided detailed comments
on some of the model assumptions. Key issues noted by the commenter include:
1. There is a uniform temperature throughout the entire coke bed and the quench water at
the time the vent is started and that 100% of the water in the coke drum at the time of
venting is at its bubble point (i.e. all the heat evolved goes toward affecting evaporation
and none of it is used in heating the water to the boiling point).
2. There is a 10 percent convective loss term; commenters suggested a convective loss value
of 80 percent be used to account for heat needed to raise water to bubble point (see item
1).
3. The model uses a void fraction of 0.5 for the coke bed, resulting in overestimating the
amount of quench water that could be in the bed and underestimating the mass of coke. In
addition, the model assumption that all the void space is filled with water is not supported
by available information. If the model void fraction is decreased from 0.5 to 0.35, the
average bulk density of the coke bed would be more in line with available data for green
coke (55 lb/cubic foot (ft3)).
202
-------
4. The model overestimates the base case emissions because it assumed the coke and
quench water for the average coke drum in the base case was at a much higher
temperature than indicated by correlation to the overhead pressure.
The commenter also expressed concerns about the typical coke drum dimensions, moisture
content of the vent gas, and dry vapor pollutant concentrations used in the SCAQMD
model. According to the commenter, the assumptions used in the model estimates result in a
significant overestimation of DCU baseline emissions and an overestimate of the emission
reductions achieved by the proposed 2 psig emissions limitation.
Response 2: When developing the emission impacts for the DCU emission control options, we
used a heat balance model, which we shared with SCAQMD. Several of the comments specific
to the assumptions used in this model are pertinent to our impact analysis and we have
considered these comments and have revised our emission impacts to address some of the key
concerns raised by the commenter.
1. We recognize that there is a temperature gradient in the coke drum with the hottest
temperature at the top of the bed and cooler temperatures at the bottom of the
bed. Therefore, we recognize that using the overhead temperature for the average bed
temperature over-estimated the baseline emissions (and thereby the emission
reductions). In recently proposed updates of the Emissions Protocol, we recommended
using the average of 212 °F and the overhead temperature as the average bed
temperature. While there may be a small portion of the bottom of the bed that may be
lower than 212 °F, we also recognize that there may remain hot spots in the coke bed that
are much higher than the wall temperature (where measurements are generally taken), so
we consider this approach to provide a good estimate of average bed temperature.
2. Even accounting for the heat needed to raise some of the water temperature to the bubble
point, we do not believe that this effect can be used to estimate such a high convective
heat loss since the enthalpy of the phase change is 960 Btu/lb compared to the heat
capacity of water, which is 1 Btu/lb/deg F. Thus, if every pound of water that turned to
steam had to increase in temperature by 20 °F, the heat used to increase the temperature
of the water would only be 2 percent of the heat loss from vaporization. If the entire
amount of water in the drum increased 20 °F and 10 percent of that water volatilized,
only 20 percent of the total heat loss would be due to raising the water
temperature. Therefore, the 80 percent convective loss term is considered
unreasonable. Furthermore, using an average bed temperature (rather than an overhead
temperature) minimizes, to some extent, the amount of heat that may be used to raise the
water temperature. While we consider 10 percent convective heat loss to be a reasonable
assumption, we also evaluated the emission impacts using a 20 percent convective heat
loss assumption.
3. We have seen very limited data on the bulk density of coke as it exists in a coke
drum. Data for green coke (that has been removed from the coke drum) is expected to
have less void space than exists within the coke bed. Based on reported coke production
capacity, drum dimensions and drum outages, a coke drum porosity of 0.5 appears to be
much more appropriate than the suggested value of 0.35. Therefore, we have not revised
this value.
203
-------
4. The pressure-temperature correlation is only more accurate if one has a closed system
(and that closed system is allowed to come to equilibrium). As vapors are continuously
being vented from the drum, the overhead drum pressure is lower than the equilibrium
pressure of the system, which is commonly seen with pressure increases as the system is
sealed to close the vapor line valves to the blowdown system prior to venting to the
atmosphere. As noted in our response to item 1, we do agree that the overhead
temperature alone would overestimate emissions, but we consider using the average bed
temperature, as described in item 1, to provide at least as accurate an estimate of the
average coke bed temperature as the overhead venting pressure during active venting of
the vessel to the blowdown system.
We have revised our analysis to use average pollutant emission factors from the source tests so
comments on the moisture content of the emissions and dry pollutant concentrations are not
directly pertinent to the revised emission impact method. Comments concerning drum
dimensions are also not pertinent. Unlike the SCAQMD impacts, we used facility-specific drum
dimensions, outages, and water heights when calculating the emissions rather default average
values. Based on these model revisions, we have lowered our estimated HAP emissions but we
still estimate our final rule will achieve approximately 400 tons per HAP and 2000 tpy VOC. See
the memorandum entitled "Reanalysis of MACT for Delayed Coking Unit Decoking Operations"
in Docket ID No. EPA-HQ-OAR-2010-0682 for additional detail.
7.5.2 Control costs
Comment 1: Six commenters stated that the EPA underestimated the costs associated with
complying with the 2 psig DCU emissions limitation. According to the commenters, the EPA
relied on a single cost estimate for a steam ejector system of $l-mllion and many DCU would
need to do significantly more expensive upgrades of their system in order to meet the 2 psig
emissions limitation. According to the commenters, unit-specific upgrade costs range for $1- to
$20-million (with most in the $1- to $10-million range). According to the commenters, jet ejector
systems may not work if the blowdown system cannot handle the additional steam. Thus, some
facilities would also need blowdown system upgrades. Alternatively, some facilities would elect
to install a new compressor for the DCU blowdown system. Two commenters noted that the
EPA assumed a compressor would be needed to achieve 2 psig emission limitation in its impacts
for Refinery NSPS Ja and a cost of $12- to $18-million per unit. According to the commenters,
the total capital costs projected for the subpart CC impact analysis are about a factor of 10 too
low.
Response 1: We based our cost on a steam ejector system that was actually installed to achieve a
2 psig pressure limit. While a significant range of costs were provided by the commenters, a
majority of DCU were projected to have costs under $5-million and many had no costs at all, so
we disagree with commenters suggesting the capital costs are a factor of 10 too low. The costs
provided by the commenters were early stage engineering estimates, while the costs we used
were based on a system that had already been installed and used to comply with the venting
pressure limit. While we do expect some variability in the final project costs, many of the costs
suggested by commenters are caused more by capacity creep (trying to process more gas oil
and/or use shorter cycle times than the equipment was originally designed). In order to account
204
-------
for additional costs associated with additional emissions reduction measures, we assumed that
the steam ejectors would be sufficient to achieve a two-thirds (67 percent) emission reduction
(i.e., if the current venting pressure is about 6 psig or less). If a DCU would need more than 67
percent emissions reduction, then the capital costs were projected to be twice our estimate for a
single steam ejector system, or $2-million (again assuming the use of steam ejector systems
along with additional modifications to improve the blow down system capacity). The annual
operating costs for the $2-million system is expected to be the same as the simple steam ejector
system with the addition of 4 percent of capital costs (for taxes and insurance). On a nationwide
basis, the final capital cost estimates increased by 60 percent from proposal to $81-million and
the total annualized costs (considering VOC recovery credits) increased by a factor of 3 to $11.7-
million, primarily due to lower VOC recovery credits due to changes in the emission modelling
assumptions described previously in Section 7.5.2 of this document. See the memorandum
entitled "Reanalysis of MACT for Delayed Coking Unit Decoking Operations" in Docket ID No.
EPA-HQ-OAR-2010-0682 for additional detail.
7.5.3 Cost effectiveness and risk
Comment 1: Several commenters stated that the cost-effectiveness of the 2 psig standard is very
high (commenter estimates ranged from $76,000 to $636,000/ton HAP); therefore, the
commenters concluded that a 2 psig work practice standard is not cost-effective and cannot be
justified under CAA section 112(d)(6). One commenter noted that this conclusion is consistent
with the conclusion the EPA reached in promulgating a 5 psig depressurization work practice
standard under NSPS subpart Ja. Three commenters noted that, based on their cost estimates, a 5
psig pressure limit would also not be cost-effective.
Response 1: While our revised cost analysis is generally in line with the lower cost-effectiveness
values of the range provided by the commenters, we conclude that a MACT floor limit was never
established for this emissions source and that we must now establish proper MACT requirements
for this source under CAA section 112(d)(2) and (d)(3). Based on our cost estimates, we agree
that additional reductions beyond the MACT floor are not cost-effective, therefore MACT is
being established as the MACT floor level of control.
Comment 2: Several commenters stated that the EPA has overstated the risk associated with
DCU emissions. Two commenters suggested that the Protocol overstated PAH concentrations, a
risk driver for cokers, by approximately an order of magnitude because the Protocol predated the
required stack tests from the ICR, which, according to the commenters, demonstrated that the
data set used to develop the Protocol was an outlier. These commenters also alleged that there
is significant conservatism in the employed toxicological values for naphthalene and benzene
because, for example, the cancer URE which the EPA used for naphthalene did not take into
account the most recent scientific information. According to the commenters, results of the
research funded by the Naphthalene Research Council suggest that the URE used by the EPA in
this document is overly conservative and that development of cancer in humans due to
naphthalene exposure may not even be a relevant endpoint. The commenters stated that, even if
one accepts the conservative selection of toxicological values and the base case emissions as
described by the EPA, the proposed revision to the DCU standard of 2 psig is not cost effective
and proposes to address a very small risk of 0.05 incidences per year nationwide according to the
205
-------
EPA estimates. The commenters also asserted that, even considering the EPA's modeling
conservatism, the cancer incidences reductions are still overstated by approximately 65%
because approximately 64% of the cancer incidences reductions were attributed to units located
in the SCAQMD, units that already are or will be subject to a federally enforceable not-to-exceed
limit of 2 psig. On the other hand, one commenter urged the EPA to set section 112(f)(2)
standards for these emission points because the health risks from refineries are unacceptable. The
commenter asserted that the fact that many sources have achieved much lower levels of toxic air
emissions from DCUs shows that the EPA must set standards for DCUs under section 112(f)(2)
in order to assure the requisite "ample margin of safety to protect public health."
Response 2: We are not finalizing the DCU requirements under CAA section 112(f)(2) because
we maintain that these emission points were not previously regulated and that these standards are
being finalized under CAA section 112(d)(2) and (d)(3). After the application of the final MACT
standards for decoking operations, we have not identified any other technically feasible methods
to further reduce the emissions from decoking operations. We also conclude that the risks are
acceptable. Together, these findings lead us to conclude that the final MACT standards achieve
an ample margin of safety.
7.6 Monitoring requirements
Comment 1: With respect to proposed section 63.657(a), one commenter requested that the EPA
clarify that the coke drum vessel pressure referenced in this section is the pressure as measured at
the top of the drum.
Response 1: We agree that the requested clarification is reasonable and necessary, particularly
as we understand that most refineries generally measure the pressure in the overhead line from
the coke drum rather than the direct drum pressure at some point within the coke
drum. Therefore, we have clarified 63.657(a) as follows:
"Each owner or operator of a delayed coking unit shall depressure each coke drum to a closed
blowdown system until the coke drum vessel pressure, measured at the top of the coke drum or
in the overhead line of the coke drum as near as practical to the coke drum, is 2 pounds per
square inch gauge (psig) or less prior to venting to the atmosphere, draining or deheading the
coke drum at the end of the cooling cycle."
Comment 2: One commenter stated that reasonable pressure instrument specifications and
QA/QC requirements must be included in the rule and consistency with the instrumentation
requirements specified in the Greenhouse Gas Protocol achieved. To avoid costly addition of all
new instrumentation, pressure instrument spans should cover the entire typical range of pressures
encountered at coke drum overheads (typically up to 100 psig and 1000 °F) and specified
accuracy must coincide with standard instrumentation and the Greenhouse Gas Protocol.
Inadequate span requirements cause unnecessary outages, since, under EPA's proposed QA/QC
requirements, pressure instruments have to be replaced or at least recalibrated, if over ranged.
Furthermore, unreasonable accuracy specifications, as have occurred in the NSPS Ja rulemaking,
can make it infeasible to comply or require installation of all new instruments in addition to the
existing process instruments.
206
-------
Response 2: The rule requires an instrument that is capable of reading a 2 psig pressure to within
0.5 psig accuracy. We recognize that system operating pressures are much higher than the final
depressurization, but allowing a pressure monitor that is accurate to 5 percent of full-scale for a
monitor with a full-scale of 100 psig would suggest that the monitor would only have to be
accurate to 5 psig. Obviously, enforcing a 2 psig limit using an instrument that is only accurate to
+/-5 psig is unworkable. Therefore, regardless of what the full-scale operation pressures are, the
instrument must be able to distinguish between a compliant 2 psig pressure and a non-compliant
3 psig pressure. Thus, if the pressure monitor has a full-range of 100 psig, then the instrument
must have an accuracy of 0.5% of full scale. We are unsure of any specific accuracy or
calibration requirements for DCU venting pressures in the Greenhouse Gas Protocol and the
"standards" provided appear to be general guidance. With respect to the requirements to
recalibrate instruments that have been over-ranged, we maintain that it is reasonable to expect
that pressure system excursions can negatively impact the accuracy of the pressure monitors,
particularly when the occur for extended periods (more than 24 continuous hours). Therefore, we
have retained the requirement to re-evaluate the accuracy of the monitors when these events
occur.
Comment 3: One commenter explained that the term CPMS is used in discussing the coker vent
pressure monitor in 63.657(b) of Refinery MACT 1, but Table 13 requirements are not applied to
that monitor by that paragraph. Since Table 13 is clearly referenced for flares and Group 1 MPV
combustion controls and in the one place a bypass continuous flow monitor is proposed, the
commenter has taken the lack of a reference to Table 13 in 63.657(b) as meaning that the Table
13 requirements do not apply to coker vent pressure monitors. The commenter added that if
Table 13 is intended to be applicable to the coker vent pressure monitor, there are serious safety
and operability issues with applying Table 13 requirements to that vent that need to be addressed.
The commenter's concerns relative to coker vent pressure taps include the following:
Coke drum overhead pressure taps are kept free of coke using blowback steam. Turning
that steam on and off to perform the required daily check for plugging would ultimately
lead to plugging of the pressure tap, steam leaks, and increased risk of burns.
Coke drum overhead pressure taps are not active when the coke drum is open for coke
cutting and there is no reason to check the pressure tap when the drum is open. Thus,
daily checks are not always possible and that requirement should only apply when the
coke is being deposited in the drum through the time the drum vent has been opened to
the atmosphere.
Response 3: As the commenter noted, 40 CFR 63.657(b) does refer to the pressure monitor as a
CPMS but it does not specifically reference that Table 13 requirements must be met. Instead, we
specify the accuracy requirements of the monitoring system in 40 CFR 63.657(b) because we
determined that the accuracy requirements in Table 13 were inadequate for the 2 psig monitoring
requirement as DCU operate at much higher pressures (80 to 100 psig) while in the operating
cycle. After considering the issues identified by the commenter, we agree that the use of the term
CPMS makes it unclear whether Table 13 applies and have altered the phrasing of the
requirements in 40 CFR 63.657(b) to avoid this terminology. While we agree that the daily check
of the pressure taps is unnecessary as the monitor's expected fluctuation between operating and
cooling cycles provides sufficient evidence that the pressure taps are not plugged, we do want to
207
-------
clarify that some review and maintenance of the pressure monitor is performed. Therefore,
instead of the daily pressure tap inspection requirements, we are requiring a daily inspection of
the pressure trends across the coking and cooling cycles to ensure the pressure taps are not
clogged. We are also specifying the calibration and placement criteria in 40 CFR 63.657(b) so
that all DCU pressure monitoring requirements are clearly outlined in this section. Specifically,
40 CFR 63.657(b) is being finalized to read as follows:
(b) Each owner or operator of a delayed coking unit shall install, operate, calibrate, and maintain
a monitoring system, as specified in paragraphs (b)(1) through (5) of this section, to determine
the coke drum vessel pressure.
(1) The pressure monitoring system must be in a representative location (at the top of the coke
drum or in the overhead line as near as practical to the coke drum) that minimizes or eliminates
pulsating pressure, vibration, and, to the extent practical, internal and external corrosion.
(2) The pressure monitoring system must be capable of measuring a pressure of 2 psig within
±0.5 psig.
(3) The pressure monitoring system must be verified annually or at the frequency recommended
by the instrument manufacturer. The pressure monitoring system must be verified following any
period of more than 24 hours throughout which the pressure exceeded the maximum rated
pressure of the sensor, or the data recorder was off scale.
(4) All components of the pressure monitoring system must be visually inspected for integrity,
oxidation and galvanic corrosion every 3 months, unless the system has a redundant pressure
sensor.
(5) The output of the pressure monitoring system must be reviewed daily to ensure that the
pressure readings fluctuate as expected between operating and cooling/decoking cycles to verify
the pressure taps are not plugged. Plugged pressure taps must be unplugged or otherwise repaired
prior to the next operating cycle.
208
-------
8.0 Fenceline Monitoring Provisions
Comment 1: Commenters claimed that the proposed fenceline monitoring standard represents a
section 112(d)(6) review of the proposed outcome of the EPA's concurrent section 112(d)(6)
review of existing tank, LDAR, and wastewater operation emission standards. Commenters
argued such a concurrent review is unlawful because the EPA is not reviewing "an emission
standard promulgated under this section."
Response 1: Section 112(d)(6) states that "The Administrator shall review, and revise as
necessary (taking into account developments in practices, processes, and control technologies),
emission standards promulgated under this section no less often than every 8 years." Thus,
section 112(d)(6), while allowing review of each standard individually, does not prohibit review
of standards collectively. The fenceline monitoring standard is innovative and applies to the
collection of fugitive emissions sources within a petroleum refinery, which were individually
regulated in the original Refinery MACT 1 rule. In the preamble to the proposed rule, we
evaluated specific emission standards individually, but explained that fenceline monitoring is a
development in processes, practices and control technologies for measuring and controlling
collective fugitive emissions from petroleum refineries. This development provided distinct
advantages by addressing the infrequency of the monitoring and inspection requirements and
providing necessary assurance that the emission control levels projected for these sources are
achieved. Therefore, we deemed it necessary, under section 112(d)(6), to include a single
overarching requirement to fenceline monitoring.
Comment 2: One commenter noted that the EPA did not impose the fenceline monitoring
standard on other source categories, with essentially the same fugitive emission requirements. In
fact, in the Generic and Amino/Phenolic Resins NESHAP amendments finalized on October 8,
2014, where very similar requirements to those proposed here are imposed, the EPA confirmed
that existing fugitive emission compliance requirements are adequate to demonstrate compliance
Response 2: We relied on source-category specific information to propose and finalize
amendments to the Refinery MACT. Similarly, decisions made for other source categories are
based on information specific to those source categories and should not be construed as setting a
precedent for other source categories. As we have discussed earlier in this document, the record
is replete with evidence that emissions from refinery MACT sources may be understated,
particularly from fugitive sources (storage tanks, wastewater treatment, and equipment leaks),
which are difficult to characterize and are significant. The work practice standard is tailored to
detect and correct fugitive emissions using benzene as a surrogate for organic HAP in refinery
process streams. Additionally, the development and evaluation of Methods 325A and 325B were
ongoing until the finalization of the Refinery MACT amendments.
Comment 3: One commenter disagreed that the primary goal for a fenceline monitoring system,
should be to ensure that owners and operators properly monitor and manage fugitive HAP
emissions. Instead, the commenter stated, the primary goal should be to determine the residual
risk to the downwind community as required by law.
209
-------
Response 3: Although we believe that the information collected by refineries as part of this
fenceline monitoring work practice standard will be valuable to downwind communities in
understanding the location of areas that are impacted by refinery fugitive emissions, it will not
and cannot provide affected communities all the information necessary to understand the risk
they face as downwind communities.
As we stated in the preamble to the proposed rule, we are concerned about the potential for high
emissions from fugitive sources and due to the difficulty of measuring these emissions, there is
significant uncertainty associated with emission inventories. Because the requirements and
decisions that were proposed are based on the results of our risk modeling of emission
inventories (i.e., the expected control effectiveness of the existing MACT requirements), our
goal was to ensure that fugitive emissions were not, in fact, significantly higher than what was
reported in the inventories and that fugitive emissions would be managed to the levels expected
to be achieved by the best performing sources upon which the MACT standards were originally
based.
Comment 4: Commenters stated that there is no correlation between the fenceline maximum
benzene value and equipment leak or fugitives emissions and thus this requirement is not a
compliance assurance method for those emission types. The commenters added that this
measurement is a duplication of the source specific monitoring already specified for equipment
leaks in Refinery MACT 1, state and local rules and permits. Additionally, the commenters
contended that the fenceline benzene concentration is a function of the location of sources of all
types within the facility, not just fugitive sources, contributions from mobile and non-refinery
sources, release heights and velocities, and wind direction and speed.
Response 4: We disagree with the commenters. The EPA performed pilot studies and actual and
allowable emissions/dispersion modeling to support the development of the fenceline monitoring
program and ensure that the monitoring would represent emissions from refinery fugitive
emission points. We further note that an API study, in which a fenceline monitoring program
was implemented at 12 facilities, was submitted during the comment period, and also supports
that the fenceline monitoring program will achieve the stated goals of measuring benzene
emissions from refinery emission points (EPA-HQ-OAR-2010-0682-0583). Appendix D of the
API report presents plots showing facility fenceline concentrations at the sampler locations
prescribed in the proposal's siting requirements as well as extra samplers where a known
emission source is located within 50 meters (162 feet) of the fenceline. These plots also show
predominant wind direction and speed over each 14-day sampling period. A review of these plots
indicates that qualitatively there is a correlation between potential sources of fugitive emissions
and higher benzene concentrations. Generally, higher concentrations were detected either at the
extra sampler location or sampling location next to the extra sampler.
Further, while low level miscellaneous process vents may contribute to the fenceline benzene
concentration, most large stack emissions, including flares, have buoyancy effects (due to
elevated temperatures) and velocities that result in minimal contribution to the fenceline
concentrations. Even so, we used the entire emissions inventory (both fugitive and "stack"
emissions) in our determination of the action level. Thus, while the fenceline monitoring
standard is targeted specifically to ensure proper control of fugitive sources, we did not establish
210
-------
the action level based only on fugitive source emissions. As such, we consider the action level to
include contributions from stack emissions and is achievable for refineries regardless of the
height and location of their stack emissions.
While we agree that mobile sources and non-refinery sources would also contribute to fenceline
concentrations, in theory, they would generally not likely be the most contributing factor for
elevated concentrations of benzene. Our ICR data indicates that most benzene at refineries is
emitted from fugitive sources (80%); and refinery sources (93%); further, the final rule provides
a means to identify non-refinery MACT sources (on-site interfering sources) and subtract out the
contribution from these sources. To understand the contribution of mobile sources to benzene
fenceline contributions, we looked at background levels readings of benzene in urban and rural
settings from the API study and noted that there was no appreciable difference, and readings
remained relatively constant (from 0.1 to 0.3 parts per billion (ppb)). Meanwhile, the highest
fenceline concentrations of benzene at these same locations varied from 0.4 to 6.4 ppb,
indicating significant contributions from refinery sources. It is likely that most mobile emissions
sources will be external to the refinery and their emissions will increase the background
concentration and correction concentration (that may overstate the true background and cause an
underestimation of the refinery's impact on the net concentration). However, we do recognize
that there are cases where a significantly travelled roadway may bisect the refinery and
contribute to the highest concentration point but not the lowest concentration point. While we do
not consider that these situations will significantly impact the highest concentration measured,
the final rule provides a means to identify these sources, conduct additional monitoring, and use
a uniform background and near-field source correction for determining the refinery contribution
to the fenceline concentration.
Finally, we disagree with the commenter's suggestion that the fenceline monitoring standard is
"simply ... a duplication of the source specific monitoring already specified." Fugitive
monitoring and inspection requirements are required on a relatively infrequent basis. We
determined that improvements to individual fugitive emissions source requirements were not
cost effective and not necessary pursuant to section 112(d)(6). However, we concluded that the
fenceline monitoring standard was a cost-effective development that would augment the source-
specific requirements.
Comment 5: Commenters suggested that rather than being prescribed as an on-going MACT
requirement, the EPA should collect fenceline data using these methods for a period of two years
through CAA section 114 authority. This would allow the EPA to determine the value of the
program and identify and correct deficiencies in the methods should a longer period of
monitoring be demonstrated as necessary.
Response 5: We disagree that we should establish the program as a temporary requirement
pursuant to CAA section 114. We identified the fenceline monitoring program as a cost-effective
development within the meaning of CAA section 112(d)(6) and thus that it is necessary under
that provision.
Comment 6: One commenter stated that the proposed fenceline monitoring program is not an
appropriate substitute for an EPA-funded national ambient air toxic monitoring network.
211
-------
Another commenter stated that because benzene concentrations vary depending on the
geographical area and the emission source potential, the EPA should defer to the States to
implement fenceline monitoring. The commenter noted that TCEQ, for example, has identified
areas of elevated benzene concentrations not related to refinery emissions and in 2004,
established Air Pollution Watch List Area 1203 for benzene in an area of Texas City, Texas. The
commenter contended that according to TCEQ's analysis, benzene emissions were due to
pipeline operations in the area and subsequent actions to reduce emissions resulted in the area
being removed from the Watch List in 2007.
Response 6: We agree that the fenceline monitoring work practice standard is not a substitute for
ambient air toxic monitoring networks and never intended for it to be. The commenter noted
some examples of where ambient monitoring was used to identify previously unknown or under
reported emissions and to correct the excess emissions source. This is precisely the idea behind
the fenceline monitoring standard. However, this standard was developed specifically to address
refinery sources. If a local area has other significant HAP emission sources that are not
petroleum refineries, those sources are beyond the scope of this rulemaking. Nothing in this
rulemaking prevents local agencies from developing additional strategies to address these non-
refinery emissions sources. If these other emission sources are subject to other MACT standards,
the EPA is required to conduct a technology review of the MACT standards every 8 years to
determine if it is necessary to revise source category-specific standards, considering
developments in processes, practices and control technologies.
8.1 Proposed siting procedures (including comments on Method 325A)
Comment 1: Some commenters stressed the importance of monitor placement and maximum
coverage to ensure as much emissions data is being collected as possible to achieve the goal of
protecting human health. One commenter specifically disagreed with allowing 2,000 feet
between monitoring sites stating that this could allow large fugitive sources and clouds of VOC
or HAP to remain unmonitored and added that while the EPA proposal requires additional
samplers be deployed when an emissions source is within 50 feet of the fenceline, it is
insufficient to plug a gap that is nearly seven football fields long. One commenter added that
open-path systems address placement concerns because the energy source and detector pair can
be placed 100-500 meters apart ensuring 100% coverage between those distances.
Response 1: We agree that monitor placement and coverage is important, which is why we
require monitoring around the entire plant. However, we disagree with the comment suggesting
that the monitoring siting procedures will allow large clouds or VOC or HAP to remain
unmonitored. First, the 2,000 feet (610 meters) distance between monitors can only be used by
relatively large facilities (perimeter of 24,000 feet or greater, which is about 800 acres or
greater); smaller facilities are still required to have 12 monitoring locations, which would be
evenly spaced and less than 2,000 feet apart. Moreover, fugitive emissions are expected to
expand and disperse along the wind vector and are highly unlikely to pass between monitoring
stations undetected, particularly as shifts in wind direction are likely to cause the plume to pass
across the monitoring location for at least some portion of the 2 week monitoring interval. That,
combined with the low detection limit capabilities of the passive diffusive tube monitoring
212
-------
method allows the proposed monitoring placement method to detect unusual, large fugitive
releases.
We disagree with the comment that open-path systems provide better coverage than the passive
diffusive tube monitoring. This is because there are significant detection limit issues with current
open-path systems. While longer path lengths for open-path systems theoretically improve the
detection limit, the detection limit for an open path system is an average concentration over the
entire path length, so it is still possible for small, reasonably concentrated plumes to remain
undetected while the passive monitoring system could detect an increase in fenceline
concentrations even though the plume only occasionally passes across the fixed diffusive tube
monitoring location. Additionally, increasing the path length of an open-path monitoring system
also increases interferences with other ambient compounds and pollutants. To date, there are no
commercially-available, real-time open-path monitors capable of detecting benzene at the levels
necessary to demonstrate compliance with the fenceline benzene action level in this final rule.
For these reasons, we find that the proposed diffusive tube monitoring systems actually provides
better coverage and better capabilities for detecting fugitive emission plumes from a refinery
than current open-path systems.
Finally, we note that the proposed additional monitoring requirement is for sources within 50
meters (which is 162 feet) from the monitoring perimeter, rather than just 50 feet. This may have
been an editorial error, but we find that the additional monitor requirement for sources within 50
meters of the monitoring perimeter is more likely to require additional monitor placement and/or
it will provide a greater distance between the source and the monitoring location, which will
allow the plume more time to disperse prior to monitoring perimeter, than had the additional
placement requirement been set at 50 feet.
Comment 2: One commenter explained that passive diffusive tube monitoring method is subject
to four potential common errors affecting the Fick's law relationship, all of which can result in
low benzene measurements in the test results: 1. High face velocity; 2. Low face velocity; 3.
Desorption of analyte due to excessive sampler heating; and 4. Sample overloading. High wind
speeds (generally greater than 10 mph or 15 m/s) may reduce the effective benzene collection
rate because it moves past the sampler before it can be trapped. No wind or very low wind (0 to
greater than 1 mph) may also reduce the collection rate because the media scavenges the local
benzene faster than the air can replenish it to maintain the ambient benzene concentration. The
commenter argued that all of these are issues when using passive dosimeters for OSHA
compliance monitoring that should be considered when using similar monitors for the EPA
compliance monitoring. The commenter continued that overexposure to heat may cause the
sampling trap to re-emit some previously collected benzene; therefore, the protective covers
referenced in the sampling method should be improved. The use of a reflective coating to reflect
sunlight and placing an insulator between the housing and the sampler to prevent heat transfer
and potential desorption of analyte should be required. And lastly, sampler adsorbent capacity is
important because each sampling tube has a finite capacity to adsorb analytes and as that
capacity is approached, the collection rate for any collectable analyte will drop. Once the
adsorption capacity is exceeded, analyte collection will either stop or preferential replacement
may occur (analytes which are weakly bound are replaced by those more strongly attracted to the
adsorbent). The commenter asserted that it is critical when using passive methods to know this
213
-------
limit before overloading occurs. Active samplers use multiple sections to identify overloading
conditions during analysis by looking for analyte breakthrough, but that is not often an option
with a passive sampler. Measuring total hydrocarbons collected, in addition to benzene, allows
the user to identify such cases of overloading and flag the data, provided that the manufacturer
has provided an estimate of sampler capacity. Data collected from an overloaded sampler should
not be used without flagging because the true value may be much higher.
The commenter suggested that the proposal include more detail on areas that should and should
not be used for monitor placement to ensure that the plants have some consistency and limit the
potential for abuse of the system to reduce apparent emissions solely through the selection of
monitoring points (see earlier references to factors affection the benzene collection rate).
Response 2: The commenter's concerns regarding wind speed effects are unfounded. High wind
speeds will actually increase sampling rates. According to the ISO method at Section A.4.4.2
"Tube-type samplers are typically unaffected by low air velocities but those without a draught
shield may be affected by high speeds." High temperatures can affect sorbent capacity and may
reduce uptake rates, but a temperature as high as 140 °F only lowers the uptake rate by about 8%
(ISO method A.4.1). Issues regarding sorbent capacity are also unfounded. Given the typical
fenceline concentrations at the refinery, it would be extremely unlikely that the tubes would
become saturated even if the sampling time was extended to a month or longer. Furthermore, the
sorbent capacity of a sampling tube is known, so one could tell if a tube is saturated based on the
monitors reported concentration. While this highly elevated concentration may be understated,
AC will clearly exceed the 9 [j,g/m3 action level for that sampling period, if not the entire year
due to that single sample.
Comment 3: One commenter stated that the EPA included details in the rule regarding areas that
can and cannot be used to site monitors in order to ensure consistency among facilities and
reduce the possibility of abuse of the system though poor monitor placement. The commenter
added that EPA could rely on the same monitoring siting criteria used for the National Ambient
Air Quality Standards monitoring, in order to establish valid data collection.
Another commenter explained that section 63.658(c) refers owners and operators to the general
guidance for siting passive monitors in EPA-454/R-98-004, Quality Assurance Handbook for Air
Pollution Measurement Systems, Volume II: Part 1: Ambient Air Quality Monitoring Program
Quality System Development, August 1998 (incorporated by reference, see section 63.14). The
commenter also added that Section 4.1 of Method 325A includes reference to the general
guidance for siting in EPA-454/B-13-003, Quality Assurance Handbook for Air Pollution
Measurement Systems, Volume II: Ambient Air Quality Monitoring Program, May 2013
(incorporated by reference, see 63.14). The commenter argued that since Section 8.2 of Draft
Method 325A specifies that the samplers be placed at specific radial points along the property
boundary or, alternately, at specific distances along the boundary, it is unclear how EPA-454/R-
98-004 or Section 4.1 of Method 325A can apply and citing them only adds confusion. Thus, the
commenter recommended the second sentence of section 63.658(c) be deleted or, if the EPA's
intent is to allow a siting approach other than the two in Draft Method 325 A, it should
specifically identify that alternative in this paragraph as being allowed in addition to the two
alternatives in Draft Method 325A. If there are other items in the cited manual that sources are
214
-------
expected to follow those should be incorporated into Method 325A. A general citation, such as
this, to a large online manual, particularly one meant to apply to ambient monitors rather than
fenceline monitors, results in confusion and compliance uncertainty.
Response 3: We agree with the commenter and amended Section 4.1 of Method 325A to remove
the reference to the Handbook.
Comment 4: Proposed 8.2.2.1.5 and 8.2.3.5 of Draft Method 325A call for extra samplers to be
placed near known sources of VOCs that are within 50 meters of the property boundary.
One commenter stated that the proposed test method language is vague and flawed in the
proposed requirement to install additional monitoring stations where there are emissions sources
within 50 meters of the fenceline. The commenter argued that this is because fugitive emissions
will expand and disperse along the wind vector and are highly unlikely to pass between stations
undetected. The commenter recommended an approach in which the EPA would require
additional monitors when there are emission sources of benzene exceeding 10 tpy within 50
meters of the fenceline.
Another commenter added that specific clarifications are needed for these provisions including
changing "known emission source" in 63.658(c)(1) to "known sources of VOCs" and "potential
emission source" consistent with the terminology used in the method to avoid misunderstandings
and claims of violations.
The commenter added that there are wastewater treatment units and Group 1 storage vessels that
do not contain benzene (except perhaps at very low impurity levels). Two commenters
recommended that 63.658(c)(1) be limited to areas near sources of benzene such as wastewater
treatment units that contain at least 10 ppm benzene and Group 1 storage vessels that contain at
least 500 ppm benzene in the stored liquid. The commenter continued explaining that because
some Group 1 tanks are quite small and individual portions of wastewater treatment systems may
be relatively small, it is also possible that after adding an extra sampler to comply with this
requirement, there may be a source between the new sampler and one of the existing samplers.
The Method seems to call for the additional sampler to be placed halfway between the otherwise
required samplers, even if the "known emission source" is not exactly halfway between the other
samplers, but is closer to one than another. This seems appropriate to avoid samplers being too
close together and because the nearest point at the property boundary may not be in the
predominant downwind position. Several commenters stated that it would be helpful, however, if
it could be clarified that siting the additional sampler halfway between the existing samplers is
the EPA's intent.
Response 4: We agree that fugitive emissions will generally expand and disperse, which is why
the monitors do not all have to be closer together. However, if the source is near the fenceline, it
is much more likely that an emissions plume from this source could remain undetected by the
perimeter monitors if the 50 meter (162 feet) extra monitor provisions are not included. We
therefore finalized these requirements as proposed. Since this placement consideration applies to
all monitoring placement options, we moved the extra monitor placement requirements to
Section 8.2.1.
215
-------
We agree that the terminology in the rule text regarding "known emission source" should use the
correct terminology consistent with the terms in Method 325A. We disagree with the
commenters' suggestion that we should limit the placement only to sources with benzene
concentrations above a certain threshold. We are using benzene as a surrogate for organic HAP
and even sources with low levels of benzene must be included as these sources can contribute to
the fenceline benzene concentration. Therefore, because we are using benzene as a surrogate for
organic HAP, we are not placing restriction on the benzene concentrations of the "known
emission sources."
Comment 5: Commenters recommended that language be added to the proposed rule requiring
and detailing the placement of fenceline monitors at petrochemical complexes that include
refineries. Commenters also recommended including provisions in the proposed rule that address
the placement of monitors within or around these large petrochemical complexes to reduce the
likelihood that refinery emissions will be attributed to associated chemical plants.
Response 5: We provided guidance in this response to comment document and clarified in
Method 325A that monitors should be placed at or inside the facility boundary so as to
encompass all sources at the facility. The term "facility" is not defined in the general provisions,
but we have consistently considered a facility to mean the collection of equipment, activities, or
both within a single contiguous area and under common control. If an owner or operator is
unsure of what is considered to be the "facility" in their specific case, he/she should submit an
applicability determination request to the EPA (or designated authority). We are aware that some
petroleum refineries will also have petrochemical processes within the boundaries of the facility
and the fenceline monitors should encompass all sources at the facility, including these
petrochemical processes. As a practical matter, a facility can comply with the 9 [j.g/m3
concentration difference for the entire facility or perform additional monitoring to correct for the
contribution of these sources that are specifically excluded from Refinery MACT 1. Specifically,
the final rule allows under 63.658(i) a site-specific monitoring plan where the owner/operator
can determine the impact of these "near-field" sources for each monitoring location and subtract
this contribution from the measured concentrations for those sources exempt from Refinery
MACT 1 under 63.640(g).
Comment 6: One commenter stated that based on the definition for property line, the marine
vessel docks would be included for fenceline monitoring. The commenter explained that it would
require placement of monitors over 1/2 a mile from shore on a dock and in several locations
placing a significant burden on the refinery. Placing monitors on the dock does not fit with the
EPA's purpose as the facility is not required to have vapor emissions collection and control and
will not accomplish the goal of evaluating the controls on tanks, wastewater, and process
fugitives. The commenter recommended that the EPA exclude docks from the fenceline
definition. Another commenter agreed and added that many times tank farms or marine terminals
are separate facilities only connected to the refinery by a pipeline. These locations are typically
low emitters of HAP, but under the proposed rule, provisions would require additional monitors
and additional costs to the refinery. Many of these monitors will also be along the waterfront.
Response 6: Marine vessel loading and tank farms are considered part of the refinery source
category and were modeled as part of the assessment, so the maximum modeled off-site
216
-------
concentrations should include these emissions. Therefore, we see no reason to provide an
exemption for these operations. In fact, we are revising "known sources of VOCs" (formerly
referred to as "known emissions sources") to include other refinery fugitive emissions sources,
such as marine vessel loading operations. In the final rule, we provided guidance on monitor
placement for segregated/remote sources. With that said, we do not consider it necessary to place
monitors along or at the end of a dock. When we modeled facilities we generally considered the
shoreline as the facility boundary. Therefore, we are providing special provisions for monitoring
at the shoreline for marine vessel loading operations.
Comment 7: One commenter explained that complicating matters and compounding costs is the
determination of monitoring locations for small refineries with non-contiguous dock facilities,
remote tank farms or multiple ownership structures. Small business refiners in this situation face
disproportionate cost burdens because they will essentially have to install the same minimum
amount of monitors around the perimeter of non-contiguous portions of their refineries as well as
around the main part of the refinery or possibly even around facilities owned and operated by
others that are located within the boundaries of a refinery. This can multiply the compliance
costs necessary to comply with this mandate. Another commenter argued that this could result in
small refineries incurring greater monitoring cost than some relatively larger facilities that could
have far greater potential emissions and potential impact on the community. Another commenter
stated that for facilities with highly irregular property boundary, contain public roadways, or
have third parties co-located within their boundaries, the number of samples will be much higher
than the EPA estimated.
In addition, refineries with numerous sub-areas and/or multiple nearby sites that are each treated
separately for monitoring purposes will incur benzene monitoring costs much higher than a
regularly shaped refinery of similar size. The increase in monitoring costs will be further
exacerbated at those irregularly shaped refineries having many sources within 50 meters of the
facility property boundary, as this will necessitate the placement of more monitors.
Response 7: For reasons discussed in the preamble to the final rule, we clarified a number of
monitor placement provisions that address many of the issues noted by the commenter and
minimized the number of extra sample locations that may be needed due to irregularly shaped
(including segregated) facilities. As seen by the monitoring results presented in the API study,
the additional monitoring locations due to near-boundary sources frequently exhibited higher
concentrations. Therefore, we find that it is critical that additional samplers be used and finalized
the requirements as proposed. We do agree with the commenter that the cost impacts were
estimated considering only regularly shaped, contiguous facilities. Therefore, we revised our cost
impacts based on the number of monitoring locations used in the API study for different-sized
facilities. Small refineries had, on average, 6 additional monitoring locations and medium and
large refineries had, on average, 8 additional monitoring locations. Based on this analysis of
available data, we disagree with commenters that the fenceline monitoring program will have a
disproportionate impact on small refineries.
Comment 8: One commenter stated that Sections 8.2.2.1.4 and 8.2.3.4 call for the samplers to be
placed "securely on a pole or supporting structure at 1.5 to 3 meters above ground level." The
commenter added that while this is an improvement over early, more prescriptive statement, the
217
-------
wording should be slightly more specific by specifying that it is the bottom of the diffusive
sample cap that is to be 1.5 to 3 meters above ground level.
Response 8: We agree with this clarification. It does apply to all monitors regardless of the
placement options. This language was also in Section 8.5.5. In the final Method, this requirement
was removed from Section 8.2, but remains in Section 8.5.5 and states, "Secure passive samplers
so the bottom of the diffusive sample cap is 1.5 to 3 meters (4.9 to 9.8 feet) above ground using a
pole or other secure structure at each sampling location."
Comment 9: One commenter is concerned about the amount of human error that is introduced
by using a passive monitoring system including placing, handling, removing, packaging, and
transporting of the samples. The commenter provided an example in which a field operator
would be instructed not to refuel their vehicles while handling the samples so as not to introduce
gasoline vapor during the deployment of uncapped samples.
Response 9: We find little merit to the commenter's concern. The sampling tubes are often made
of metal and are very durable. Sampling tubes are to be capped at all times except when
installing or removing the diffusive cap. For example, Method 325A at 8.6.1 specifies that
sampling tubes should be "immediately" resealed after removing the diffusive cap, so there
should be no contamination during transport. The only potential source of contamination is
during the few seconds while the caps are removed or replaced. For this reason, it is good
practice to have clean hands when deploying/collecting the diffusive samplers. Thus, we find
that the instruction to field operators not to refuel their vehicle (or to wash their hands thoroughly
if they do refuel their vehicle) to prevent sample contamination during sample deployment or
collection is reasonable.
Comment 10: One commenter discussed the specifics of wind direction and surrounding
contributions to air emissions with respect to a nearby facility. The commenter argued that it will
be important for the EPA to require monitoring that is up wind and down-wind of the refinery
sources, as well as requiring monitoring that is ground-level and stack-level.
Response 10: The finalized approach requires sampling of the entire perimeter, which allows for
approximate upwind and downwind concentrations. Further, both theory and actual gradient data
have shown that concentrations due to the refinery fugitive emissions drop exponentially as you
move away from the source, practically regardless of topography and weather.
Comment 11: One commenter stated that the proposal does not appear to consider the effect on
relatively uniform wind directions on the trigger level and its measurement. The commenter
stated that at their refinery the wind blows regularly and steadily in a relatively small vector
range. The effect is that (1) the background concentration will almost always be zero and (2) the
fenceline monitors will be monitoring the effect of only certain sources that happen to be near
the fenceline and downwind of the prevailing wind direction. The commenter recommended that
the EPA exclude facilities in trade wind areas from fenceline monitoring at least until a better
study can be conducted to identify an effective means of monitoring.
218
-------
Response 11: We do not understand the commenter's concern that the monitoring is not
effective where uniform meteorology conditions persist. On the contrary, we note that this
situation, where the wind blows regularly and steadily in the same general direction is the
simplest scenario where a subtraction of upwind and downwind concentrations would be the
clearest indication of the refinery's contribution. In the case described by the commenter, if the
upwind concentrations (or background) are always near zero, then the concentration at the
downwind point is entirely attributable to the refinery. We do understand that there may be
facilities where there are no prevailing winds such that all monitoring locations surrounding the
refinery are impacted by the facility's emissions and the AC approach may understate the
refinery's impact on the concentration, but this does not provide a reasonable rationale to
preclude the AC approach where it is more accurate. Finally, in our modeling efforts we
considered the local meteorological data when estimating fenceline concentrations. That is, when
we developed the action level, we accounted for cases where there are consistent prevailing
winds. Therefore, we specifically considered site-specific conditions in establishing the action
level and concluded that all refineries could comply with the action level of 9 [j,g/m3.
Comment 12: Several commenters are concerned about the effectiveness of the proposed
passive diffusion tube sampling method during cold weather conditions. Stationary sources
located on Alaska's North Slope and in the Interior routinely experience winter temperatures as
low as -40 °C. Strong atmospheric temperature inversions can result in such extreme conditions
lasting for extended periods of days or even weeks. One commenter added that collecting
samples in extreme weather to meet an arbitrary 14-day period can put employees at risk or make
strict adherence to sampling protocols impossible. Several commenters expressed similar
concerns for parts of the northern continental U.S. such as Wyoming and added that the rule
must allow for refineries to use proper media for the operating climate and encourage refineries
to work with their laboratories to ensure media and sampling rates are designed for their
operating climate.
One commenter requested that as part of finalizing the rule, the EPA provide data and analysis
that show the proposed monitoring requirements can be met across the range of temperature
conditions that will be experienced by stationary sources across the country including Alaska and
other cold region states. The commenter further requested that the EPA explain whether any
comparative testing data performed at low temperatures demonstrated the effectiveness in terms
of accuracy and precision for this sampling protocol before finalizing the rule as proposed. The
commenter questioned if after evaluating such data, did the EPA find that there are limitations to
the sampling method due to low temperatures? Does the EPA have data on alternative, similarly
low cost, sampling methods which would be available for use in cold weather regions?
Response 12: Ambient temperatures will generally have little impact on the diffusive tube
samplers. At colder temperatures, the sorbent should work effectively. Molecular diffusion is
dependent on absolute temperature. The mass of a pollutant collected is expected to be
proportional to the square root of the absolute temperature (ISO Method, Section A), with lower
diffusion rates (lower sampling rates) with lower temperatures. However, typical shifts in
ambient temperatures do not significantly alter the absolute temperature or the sampling rate. A
temperature shift of 100 °F is needed in order to change the sampling rate by 10%. Consequently,
we do not see a significant issue with sampling in cold climates. Lab data demonstrated the
219
-------
passive tube recovery of benzene challenged at a known concentration was acceptable across a
range of temperatures (6 °F - 120 °F). The primary issue with the cold weather would be if ice
forms on the sampler and covers the diffusion cap holes. The weather protector is designed to
limit this, but in a high swirling snow storm, some issues may exist, but these issues will
generally lead to lower than expected sampling rates and would not cause facilities to exceed the
action level. Since volatile HAP emissions are temperature dependent as well, emissions during
these extreme cold conditions are generally low and we do not expect significant issues with the
samplers or the methods for facilities in cold climates.
Comment 13: Several commenters stated that the fenceline monitoring rule requires every
refinery to install an on-site meteorological station without any requirement for how data from
that station is to be used. One commenter stated that if the station is used to collect wind speed
and direction over a two week period, it will likely result in data indicating wind from all
directors and add little to no value in identifying sources of benzene. Several commenters
understand that weather data may be useful in determining source locations, but argued it should
not be a requirement to have a met station. The commenters requested that EPA drop this
requirement or allow the use of United States Weather Service (USWS) data if a station is within
25 miles of the refinery.
A number of commenters also added that the EPA should broaden site-specific monitoring plans
for the met station to allow for any off-site sources (not just "upwind") and other on-site sources,
and allow these plans to be used from the date of submission.
Several commenters argued that specifying that any weather station used for complying with
63.658(d) be "dedicated" could cause refineries to have to have multiple weather stations, since
some refineries have weather stations for compliance with other requirements or for their own
use. These commenters concluded by stating that the requirement for a met station should be
changed to an optional requirement but if retained, the word "dedicated" should be removed
from proposed section 63.658(d) language. One commenter provided sample rule text for
revising the requirement to be optional in Section 1.4.3.2 of their comment package.
Response 13: Meteorological data are important in understanding the fenceline concentration
data and to identify the source of emissions if high concentrations are measured. However, we
agree that these data are only critical if a site-specific monitoring plan is needed and near-field
source corrections are to be made. Therefore, in the final rule we are allowing facilities that have
met data available from USWS data within 25 miles of the facility to use it in place of an on-site
meteorological station. However, we are retaining the requirement for an on-site weather station
if near-field source correction method is used. We also agree that the requirement that the met
station be "dedicated" is unnecessary and we are deleting this provision in the final rule.
Comment 14: One commenter explained that proposed 63.658(e) and Sections 1.4, 3.8 and 8.5.9
of Method 325A specify that the length of the sampling episode must be fourteen days.
However, the commenter argued that all samplers cannot be changed out simultaneously and
there is some delay in shutting down a sampler and starting up a replacement, particularly in bad
weather, or when sample support, sample holder or security lighting or fencing maintenance is
required. The commenter added that by the nature of the methodology (two week samples used
220
-------
to develop an annual average) exactly fourteen day samples are unnecessary. Annual average
values will not be sensitive to relatively small variations in the sampling time. Nor, are the types
of emissions the EPA claims they are addressing with this program subject to short-term
aberrations that might be missed by short sampling outages. Thus, the commenter recommended
that rather than 14 days these paragraphs specify a sample duration of "approximately" fourteen
days. One commenter proposed that the requirement be modified to allow the sampling episode
length to be 12 -16 days, at the discretion of the refinery, if needed.
Response 14: The commenter is correct that the samplers will work well whether the sampling
duration is a few days longer or a few days shorter. The method requires recording the time/date
of deployment and the time/date of collection and the sampling time is used in the calculation of
the concentration for the sampling period. Therefore, the sampling duration is used specifically
in calculating the concentration for the sampling period. However, we currently do not require
the use of the sampling duration when calculating the annual average. By requiring consistent
sampling intervals, there is no need to weight each sample by the number of days or hours
represented by that sample. If the sampling durations are not consistent, then a direct average of
the 26 samples collected over the year may not represent the true average. Therefore, the
guidance remains that the sampling period should be 14 days. We are softening the language in
Section 8.4.2 from "sampling tubes must be changed at approximately the same time of day at
each of the monitoring sites" to "to the extent practical, sampling tubes should be changed at
approximately the same time of day at each of the monitoring sites." We find that best practice
would be to maintain a consistent day and time for sampler deployment and collection. However,
if there is a significant thunder storm, tornado warning or similar event, there is no issue with
altering the sampling time to avoid dangerous conditions. If the sampler deployment and
collection were delayed a day, that would not invalidate the sample. Therefore, we are clarifying
in 40 CFR 63.658(e) that "For the purpose of this subpart, a 14-day sampling period may be no
shorter than 13 calendar days and no longer than 15 calendar days." This provision provides
some of the flexibility requested by the commenter while minimizing potential issues associated
with directly averaging the sampling period values. Nonetheless, we believe every effort should
be made to deploy and collect the samples using a consistent 14-day sampling interval.
Comment 15: One commenter stated that proposed 63.658(c)(3) and Section 9.3.1 and 9.3.2 of
proposed Draft Method 325A call for one co-located duplicate sample for every 10 field samples
per sampling episode and at least two field blanks per sampling episode. However, the note in
Section 8.5.5 only calls for one duplicate sample for each monitoring episode and Section 8.5.10
requires one field blank for every 10 samplers, rather than two per sampling episode. The
commenter requested that these conflicts need to be resolved and made consistent.
Response 15: The co-located duplicate sample requirement is 1 for every 10 field samples per
sampling period and the field blank requirement is at least 2 field blanks per sampling period.
The note for Section 8.5.5 was revised to say "Duplicate sampling assemblies must be deployed
in at least one monitoring location for every 10 monitoring locations during each field
monitoring period." Section 8.5.10 actually says 1 field blank for every 10 samplers in one
sentence and then says that no less than 2 blanks are required per monitoring study in the next
sentence. In the final rule, we deleted the first sentence regarding the 1 field blank for every 10
samplers and retained the second sentence, consistent with the requirements in 9.3.2.
221
-------
Comment 16: One commenter recommended a number of editorial corrections/suggestions to
the EPA Method 325A including:
1.) Section 1.2 to the start of Section 1.3: The first sentence should read "..to determine the
average concentration of the select VOCs using the corresponding uptake rates listed..."
2.) Section 1.2 to the start of Section 1.3: Incorporate all relevant validated uptake rates from
cited standards into a new table numbered 12.2 and all qualified uptake rates from relevant peer-
reviewed literature into a new table numbered 12.3 and revise the regulatory text to reference
these tables, "listed in Method 325B, Tables, 12.1, 12.2 and 12.3. Additional compounds or
alternative sorbents must be evaluated as described in Addendum A of Method 325B, unless
already validated in later editions of one of the following national/international standard
methods: ISO 16017-2, ASTM D6196-03 or EN 14662-4, or reported in the peer-reviewed open
literature. 1.3 Methods 325A and 325B" It should also be noted that Section 1.3 states three
different sorbents Carbograph 1TD, Carbopack B, and Carbopack X or equivalent can be used.
Inclusion of the new tables would support this.
3.) Section 1.4, line 6: for consistency with Sections 2.2.2, 8.4 and 12.3.4, amend to: "The
duration of each sampling period is normally 14 days - See Sections 2.2.2, 8.4 and 12.3.4."
4.) Section 3.7: This definition of Retention Volume is very unusual and the commenter is not
sure its correct. The term is also defined differently, and more conventionally, in Method 325B.
To avoid confusion and minimize debate the commenter suggested deleting this definition and
refer readers to Method 325B.
5.) Section 3.9: Suggested correcting the typo in the definition of Sorbent tube, "Sorbent tube
(also referred to as "adsorbent tube")." Also, recommended omitting the words 'stainless steel
tubes' from the second sentence as it is redundant to the first sentence.
6.) Section 4.3, line 6: Suggested amending "excludes" to "exclude"
7.) Section 4.4, line 4: Suggested amending "samplers" to "sampler"
8.) Section 6.1, line 4: Suggested amending to: "and preparation requirements described in"
9.) Section 6.1, Note, lines 5/6: Suggested amending to: "Other sorbents may be used"
10.) Section 6.3, lines 4/5: Suggested amending to: "(see Section 4.4) For example this could
comprise an inverted cone/funnel constructed"
11.) Section 6.4 and elsewhere: Recommended the header font be changed to italic
12.) Section 8.2.1, last line: Suggested amending to "to assess the accuracy of PS results."
13.) Section 8.2.3.5, line 10: The commenter stated they think it's only two additional monitors
in Figure 8.4
222
-------
14.) Section 8.4.1: The commenter questioned if some of the cited uptake rates are for 24 hours
(e.g. Table 12.1) hand stated that it would make sense to extend the range to 24 hrs to 14 days.
15.) Section 8.5.5 note, given even the smallest refineries require a minimum of 12 sampling
locations the commenter suggested amending this note to say: "Duplicate sampling assemblies
must be deployed at least two monitoring locations during each field monitoring exercise" - for
consistency with Section 9.3.1 of Method 325A and Section 8.3.2 of Method 325B.
16.) Section 8.5.9: for consistency with Sections 2.2.2, 8.4 and 12.3.4, the commenter suggested
amending to: "Expose the sampling tubes for the required sampling period - normally 14 days -
See Sections 2.2.2, 8.4 and 12.3.4."
17.) Section 12.3.4; For clarity regarding "shorter periods" the commenter suggested referencing
Section 8.4 - "Additional monitoring for shorter periods (See Section 8.4) may be necessary ..."
Another commenter noted additional editorial corrections/suggestions to the EPA Method 325A
including:
1.) It appears 63.658 references are based on earlier versions of Draft Methods 325A and 325B
and that as a result references are incorrect. We have pointed out the specific reference errors we
have noticed, but recommend a careful review of all references to those methods prior to
publication of the final rule and methods, particularly if further changes to the methods are made
in response to comments.
2.) The reference in the last sentence of paragraph 63.658(c) to placing monitors at 2 kilometers
intervals, should be corrected. Section 8.2.3 of Draft Method 325A specifies spacing intervals
that depend on the overall length of the refinery fenceline, but in no case would that spacing be 2
km.
3.) The reference in the last sentence of paragraph 63.658(c) to Section 8.2.2.5 in Method 325A
should be to Section 8.2.3.5. Also, the same provision would apply to the radial method in the
first sentence and for consistency it should be included and reference Section 8.2.2.1.5 of
Method 325A.
4.) Section 8.4.1 of Method 325A still states that "sample collection may be performed for
periods from 48 hours up to 14 days." It is assumed that EPA intended to revise Section 8.4.1 to
be consistent with the other Sections that were revised to specify 14 days.
5.) 63.658(i)(l)(3) 'Diffusion tub monitoring' should be 'tube.' 8.) 63.658(c)(2) 'Site-specific
monitoring plant' should be 'plan.'
Response 16: We reviewed the list of edits and made changes as appropriate.
223
-------
8.2 Proposed analysis procedures (including comments on Method 325B)
Comment 1: One commenter stated that the proposed passive monitoring does not have any
EPA test method associated with it, unlike ultraviolet differential optical absorption spectroscopy
(UV-DOAS), which is specifically listed in EPA's Other Test Method 10 (OTM-10). The
commenter does not believe it is possible to develop a scientifically sound method for two week
average of passive samplers for benzene.
Response 1: As part of this proposal, we proposed both a siting method (Method 325A -
Sampler Deployment and Collection) and an analytical method (Method 325B - Sampler
Preparation and Analysis) and are finalizing these methods based on comments for Methods
325A and 325B. Methods 325A and 325B were written based on extensive research and
validation performed by EPA's Office of Research and Development and Office of Air Quality
Planning and Standards. Implementation studies using Methods 325A and 325B were
subsequently performed to demonstrate and validate these methods for measuring refinery
fenceline benzene concentrations. These methods are proposed in their own right as reference
methods fit for the purpose of this rule.
EPA's OTM-10 is a method for identification of "hot spots" and measurement of emission fluxes
through a vertical plane using path-integrated optical remote sensing techniques such as UV-
DOAS and other open-path optical spectroscopic approaches. OTM-10 does not describe the
specifics of long-term facility fenceline monitoring with open-path equipment including but not
limited to: definition of and determination of method detection and quantitation limits,
calibration and quality assurance procedures, limitations of operations, handling of data and
spectral interferences, etc. To our knowledge, standardized and validated reference methods for
use of open path UV-DOAS for long term facility fenceline monitoring have yet to be developed.
Even if OTM-10 included specific information on UV-DOAS for continuous fenceline
monitoring applications, methods posted on the EPA's Emission measurement center website as
OTMs have not undergone the Federal rulemaking procedure as have Methods 325A and 325B.
The posting of OTMs with their technical support documentation is done to provide the
information to the measurement community and support their continued development and
evaluation.
Comment 2: One commenter suggested that Draft Method 325B should provide complete
Quality Assurance/Quality Control Procedures to verify results prior to electronic data
submittals. Specifically, the commenter requested that the Draft Method 325B provide a more
robust procedure for identifying and excluding outliers, provide data substitution rules to address
lost, damaged, and delayed samples, and specify a minimum number of data points to calculate a
representative facility wide average. Some commenters also asked that in addition to procedures
for identifying invalid data, provisions for such exclusions should be added to 63.658(f)(1). A
record of excluded data and the reason for the exclusion should be required. Such an allowance
is alluded to in the reporting requirements in proposed 63.655(h)(8)(iii), where reports of such
exclusions are required, but the fact that such exclusions are allowed needs to also be included in
63.658(f).
224
-------
One commenter also stated that it is suspected that there is wide variability between different
laboratories conducting the analysis prescribed by the proposed test method and asked what the
results from the EPA's round robin study of three laboratories and the Agency's laboratory were
as they have not been publicly provided.
Response 2: We acknowledge that quality review of data is important to reporting data suitable
for the purpose of this rule. However, we have not included additional outlier procedures or
minimum data availability requirements for fenceline monitoring data nor do we provide a
specific tool for data substitution beyond what is currently in the method(s). We note that
Method 325A as proposed provides qualitative procedures for identifying and reporting outliers
in Section 9.2 of Method 325A. Procedures to identify outliers are typically case-by-case
specific. We determined that proposed procedures are adequate and performance based. Further,
we have clarified requirements for sampler placement and minimum number of samples in
Section 8.2 of Method 325A and 63.658. On the issue of applying a data substitution procedure,
we understand that issues may arise during review of data that are not easily understood without
further examination of test metadata. We would consider it very unusual that all data for a given
monitoring period would be lost. Furthermore, we do not want to lose important data through the
use of a prescriptive approach that may not be responsive to all situations. Generally, the AC
should be calculated based on the measurements completed during a sampling period. We would
find it suspicious if the monitors with the highest concentrations are "missing" more than once
per year and should the refinery or the enforcement authority be concerned about the
representativeness of data during periods of missing data, either may consider collecting
information through other means (e.g., supplemental sampling) to fill data gaps not only because
such gaps are deviations from the rule but such gaps can lead to uncertainty about compliance
status. We further believe that the final rule provides sufficient means to ensure ongoing
compliance without specifying an arbitrary numerical minimum data availability or data
substitution requirement. We consider that failure to collect required or otherwise excepted data
is a deviation from the rule and that this will provide the necessary incentive to collect data
sufficient to demonstrate compliance with the final rule.
The "Evaluation of VOC Passive Sorbent Tube Technology Performance" report is in the
Docket. The report states, ".. .variability among multiple laboratory analysis was equivalent to
the variability when one laboratory was performing the analysis. This also indicates that within
the error associated with different laboratories performing analysis, passive sorbent tube results
are comparable across the laboratories and EPA Draft Method 325 A/B provides reproducible
procedures."
Comment 3: One commenter stated that the 30 day time period from completion of each
sampling episode to determine if the action level has been met provided in section 63.658(f)
should be extended to 60 days because outside laboratory analysis may require more time to
receive and process samples, especially if there is delay in data transmission or if QA/QC follow
up is needed to evaluate the problem.
Response 3: We generally find that most situations can be resolved in 30 days, but we realize
that some events may cause additional delays. However, we find that an additional two weeks
should be sufficient to complete processing of the data even if there are QA/QC follow-up issues.
225
-------
Therefore, in the final rule, we provided a 45 day time period for completing all analysis needed
to assess AC for a given sampling period.
Comment 4: One commenter recommended use of half the detection limit for samples below the
detection limit where benzene is suspected to be present, rather than the proposed requirements
in 63.658 (f)(1) to use zero as the value for below detection limit values for the lowest sample
result (and for background and near-field source samples) and use method detection limit as the
highest value if all samples are below the detection limit. The commenter also requested that we
provide a definition for method detection limit.
Response 4: The passive diffusive tube monitoring method using 14-day monitoring periods was
specifically selected because of its ability to accurately assess benzene concentrations at very
low concentrations (detection limits as low as 0.05 ppb or 0.2 (J,g/m3). As seen from our pilot
study and API's pilot study, the question posed is largely a theoretical concern as no samples
were below detection limits. Further, while the proposed handling of non-detects provides a
"conservative" estimate of AC, it is not expected to cause issues related to compliance. A AC
based on a monitoring period of all non-detects (however improbable that is) should be on the
order of 2% of the 9 [j.g/m3 action level, so the facility would easily demonstrate compliance with
the AC action level. If only the upwind samples are below detection limits (more likely than all
non-detects, but still rare), using one half the detection limit versus zero would only cause about
a 1% difference in the AC for facilities near the AC action level so it is extremely improbable
that this assumption will make the difference between compliant AC and noncompliant AC
values. As such, we find the proposed treatment of non-detect values to be reasonable and we are
finalizing these requirements as proposed.
Comment 5: One commenter requested that the EPA study uptake rates for alternative sorbents,
such as Carbograph 5 TD, which they believe is a viable sorbent for benzene monitoring. The
commenter also asked if Carbograph 1 TD and Carbopak B would be suitable, even though the
known breakthrough volume is 2.3 L not 20 L. Two other commenters also requested that
Carbograph 5 TD be identified as equivalent to Carbopack X sorbent. Additionally, one of the
commenters stated that it is inappropriate to mention a vendor name in the sorbent descriptions.
This commenter indicated that Carbopack X is the only sorbent that meets the qualified
breakthrough volume requirements and uptake rates and is solely manufactured by Sigma-
Aldrich and is not readily available to all consumers. With only one company supplying sorbent
by default, per the current Methods 325A and 325B, the commenter indicated that there could be
concerns about supply shortages, monopoly, or price gouging. Further, this commenter was
concerned about not being able to enter the market for sorbent tubes.
Response 5: We recognize the need to allow equivalent sorbents in passive sampling tubes.
Method 325 is based on EPA research using Carbopack X. We could not determine viability and
equivalency of Carbograph 5 TD to Carbopack X during extensive internet search. Therefore, it
is not specifically included in the Method. We revised Method 325B to be consistent with
Method 325A which included more than Carbopack X and listed other options that are
commercially viable and not sole-source. In addition, the Method allows for alternative sorbents
if the uptake rate is published by a consensus body or the sorbent is evaluated using Addendum
A should a facility, tester, or vendor deem it needed and necessary to use an alternative sorbent.
226
-------
Comment 6: One commenter asked whether PVC is considered a "non-outgassing material" and
whether PE is considered a "non-emitting polymer" referring to the requirements of Method
325B. The commenter also asked what sorbent material should be used inside of the focusing
trap/secondary trap for Method 325B. The commenter also asked what material should be used
for sampling tubes, inert coated stainless steel or non-coated stainless steel. Another commenter
recommended in Section 6.1 line 5, for consistency with Section 3.9 of Method 325A amending
to state 'The tubes are made of stainless steel or inert-coated stainless steel with...' The
commenter also asked if there are certain requirements certifications or qualifications a lab or
refinery must obtain in order to perform the laboratory analysis for Methods 325A and 325B.
Response 6: PVC is an acceptable waterproof hood material based on research and subsequent
field tests conducted by EPA. High density linear polyethylene may be acceptable for shipping
samples as described in Method 325B. However, it is incumbent for sample management during
analysis to evaluate blank samples to confirm handling procedures do not contaminate samples.
The selection of absorbent material is a performance-based variable that allows manufacturers
and laboratories to optimize equipment performance now and in the future to meet quality
control method requirements. Therefore, we will not specify absorbent materials.
We removed stainless steel as the tube construction and clarified throughout Method 325 inert
coated stainless steel tubes must be used.
We do not require laboratory certification for parts 60 and 63 testing. The laboratory is free to
pursue their own certification. The performance criteria in Method 325 are adequate to generate
confidence on a sample-by-sample basis.
Comment 7: Two commenters noted that the analytical blank procedure in Section 9.1 states
that the system blank requires desorption of an empty tube or desorption of the focusing trap
alone. However the criteria for the internal standard response should be removed as it is not
possible to add an internal standard to an empty tube. One commenter recommended changing
"System blank analysis" to "Laboratory Blank" in Table 17-1. Further, commenters stated that
requiring an analytical system blank and application of data flags based on its failure is
unnecessary. If the system blank shows unacceptable artifacts, then the lab blank or Method
blank will also show unacceptable artifacts. One commenter also stated that the Method blank
will also allow more effective evaluation of the level of contamination given the fact that internal
standards can be used to calculate concentrations. The commenter further suggested that if
unacceptable background is observed, that it be remedied prior to sample analysis. Most data
processing software is unable to utilize an internal standard area count from a different data file
to calculate a concentration. Table 17.1, the commenter stated System blank Analysis should be
replaced with laboratory blank analysis. The system blank is an empty thermal desorption tube,
so cannot have Internal Standard (IS) parameters associated with it. One commenter also
indicated that the process to calculate the system blank concentrations using the factors of the
continuing calibration verification (CCV) is impractical as it requires a manual off-line
calculation for each target compound.
227
-------
Response 7: We agree that the quality checks are adequate and revised Method 325B to remove
system blank analysis and separated the information on laboratory blanks and field blanks into
two sections within Section 9.0. The EPA disagrees with the commenter that the CCV is
impractical, particularly as there is only one required target compound. The EPA determined that
laboratories have two viable alternatives to perform quantitative calculations using CCV.
Laboratories may modify their Laboratory Information Management System to generate
quantitative data based on CCV or convert data based on initial calibration to CCV concentration
using a stand-alone spreadsheet. Furthermore, based on data provided in public comments, EPA
determined that using this modification provides better consistency between facilities for
compliance purposes. The EPA provided further clarification on CCV in Sections 9.13.2 and
11.3.2.4.
Comment 8: Several commenters recommended a number of editorial changes to the language
of Method 325B to revise, expand, clarify, or modify requirements as follows. [For clarity, we
are providing responses to each bulleted item (or group of bulleted items) within this comment.]
Table 12-1, the commenter stated that based on the literature reference, these rates are
based on a 24-hour duration; however, there is no footnote stating this limitation.
Since the rule covers 14-day collection period, the commenter suggested that uptake
tables should list the appropriate duration as well as sorbent.
Response: We added note on duration to Table 12.1. However, there is not a
significant difference in uptake rates between 24-hour and 14-day duration periods.
Section 12.2.4 and Equation 12.6, the commenter recommended clarifying that Untp
is the published rate in table 12-1, and U is the sampling site uptake rate. It is U that
is plugged into equation 12.5 to report the corrected target concentrations. (See
reference ASTM D6196-03(2009) 14.2.2.1 Note 7. Additionally, the commenter
suggested that this "adjustment" is unnecessary as the small variations in the uptake
rate as a function of temperature and pressure is insignificant relative to the cited
uncertainty of the published rate and uncertainty of the analytical measurement.
Another commenter suggested removing Section 12.2.4, along with equation 12.6
because they felt it was confusing to add this in. The commenter also stated Untp
should be the published uptake rate and if this is the case then this equation is
incorrectly labelled. The commenter continued, given the normal uncertainty of
uptake rates, making minor adjustments to correct for ambient temperature and
pressure variation is not justified.
Response: We disagree that Untp is the published rate in Table 12.1. The term Untp
is the site-specific corrected value for the average temperature and pressure. In
addition, we determined it is necessary to adjust for temperature and pressure as
outlined in ISO 16017-2, EN 14662-4, and ASTM D6196-03. We agree that the
proposal was not clear in specifying the correction procedure and added clarifying
information on temperature, pressure, and uptake rates corrected for local conditions.
We also clarified the intent of correcting U for local conditions and reversed the order
of Equation 12.5 and 12.6.
228
-------
Section 1.3, the commenter suggested amending the text of this section from the 3rd
sentence to read: "Compounds with known uptake rates and associated sorbents are
listed in Tables 12.1, 12.2 and 12.3. This method provides performance criteria to
demonstrate acceptable performance of the method (or modifications of the method)
for monitoring one or more of the compounds listed in Tables 12.1 to 12.3. If
standard passive sampling tubes are packed with other sorbents or used for other
analytes than those listed in Tables 12.1, 12.2 and 12.3, then method performance and
relevant uptake rates should be verified according to Addendum A to this method
unless already validated in later editions of one of the following national/international
standard methods: ISO 16017-2, ASTM D6196-03 or EN 14662-4, or reported in the
peer-reviewed open literature."
Section 7.1.2 line 6, the commenter recommended amending to "...listed in Tables
12.1 to 12.3 and used..."
Section 7.1.3, from line 5 to 15, the commenter recommended amending to
"...according to Addendum A to this method unless already validated in later editions
of one of the following national/international standard methods: ISO 16017-2, ASTM
D6196-03 or EN 14662-4, or reported in the peer-reviewed open literature. A
summary table..."
Section 12.1.8, the commenter suggested amending the text to say "Diffusive uptake
rates for common VOCs using sorbent tubes of the dimensions specified in Section
6.1 are presented in Tables 12.1, 12.2 and 12.3. Adjust analytical conditions." The
commenter also proposed for clarity and consistency including uptake rates from
other validated standards and peer reviewed publications as Tables 12.2 and 12.3.
Section A. 1.1 line 4, the commenter recommended amending to "...in Tables 12.1 to
12.3 must be evaluated"
Section A. 1.1 lines 8 to the end, the commended suggested amending the current text
to "...this Addendum unless the compound or sorbent has already been validated in
later editions of one of the following national/international standard methods: ISO
16017-2, ASTM D6196-03 or EN 14662-4, or reported in the peer-reviewed open
literature."
Response: We clarified and expanded information about acceptable absorbents for
use in Method 325 in response to these comments. We did not add additional Tables
or "later editions" to national/international standard methods, but added additional
absorbent uptake rates for Carbograph 1TD and Carbopack B to Table 12.1.
Section 2.3 lines 6 and 7, the commenter recommended replacing "desorping" with
"desorbing".
Response: We agree and revised by replacing "desorping" with "desorbing".
Section 4.1.1 line 7, for clarity, the commenter suggested amending "polymer" to
"polymeric sorbent".
Response: We agree and revised by replacing "polymer" with "polymeric sorbent".
229
-------
Section 6.2.1 line 7, the commenter recommended deleting repeat of 'only'.
Response: We agree and revised to read: "Note that the analytical TD system should
be used for tube conditioning only if it supports a dedicated tube conditioning
mode...".
Section 9.3.4 lines 5-8, for consistency with Section 9.3.2 of Method 325A the
commenter requested amending the lines to read "...throughout the monitoring
exercise. The field blanks must be installed under a protective hood/cover at the
sampling location, but the long-term storage caps must remain in place throughout the
monitoring period (see Method 325A)."
Response: We agree and revised to read: "throughout the monitoring exercise. The
field blanks must be installed under a protective hood/cover at the sampling location,
but the long-term storage caps must remain in place throughout the monitoring period
(see Method 325A)."
Section 9.4 line 2, the commenter recommended correcting the term to be 'co-
located'
Response: We agree and corrected the grammar by replacing "collocated" with "co-
located".
Section 9.6 line 13, the commenter recommended amending 'measure' to 'detect'
because this refers to LOD.
Response: We agree and corrected the grammar by replacing "measure" with
"detect".
Section 9.10.1, for clarity the commenter requested the sentence be amended to "must
be demonstrated by quantitative re-collection and repeat analysis of a tube standard
(See Section 13.4.2 of ASTM D6196-03(2009)."
Response: This section was intentionally left general since different analysis
equipment provide different options for this quality check. We disagree with specific
language recommended by the commenter as what was proposed provides the
minimum requirements to demonstrate compound recovery in system and tubes used
for analysis. Requiring what the commenter requested may not be possible on all
equipment. Therefore, quantitative recollection and analysis is not necessary to
generate data sufficient to meet requirements.
Section 9.10.2 line 2, the commenter requested the line be edited to read "...system
can also be demonstrated by..."
Response: We agree and revised to read: "...system can also be demonstrated by...".
230
-------
Section 12.1, the commenter recommended the typo in sub-section title be corrected.
Response: It is unclear what typo is, so we did not make any revisions.
Section A.2.1 first sentence, the commenter suggested reducing the first sentence so it
just says: "Known concentrations of VOC are metered into an exposure chamber (see
Figure A.l for an example exposure chamber)."
Response: We agree and revised to read: "(see Figure A. 1 and A.2 for an example of
the exposure chamber and sorbent tube retaining rack)" instead of reducing sentence
as suggested by commenter.
Section A.4.3, the commenter recommended for extra clarity amending the text to
say: "As passive sampling devices are placed into the exposure chamber, they should
only be handled by personnel wearing clean, white..."
Response: We disagree and did not amend text as recommended. Performance based
requirements of the Methods will identify contamination issues and care should be
taken at all times when handling passive sampling devices.
Section A.6.1, the commenter stated for consistency with A.8.4 the text be amended
to say: "sufficient size to simultaneously expose a minimum of eight sorbent tubes."
Response: We agree and revised to read: ".. .expose a minimum of eight sorbent
tubes".
Section A.7.1, the commenter stated the percent relative humidity requirements of
A.7.1 must be consistent with the requirements of A.8.5.
Response: We agree and revised to read: ".. .between 35 percent and 75 percent
relative humidity.
Section A.7.1, the commenter stated the VOC concentration requirements of A.7.1
must be consistent with the requirements of A.8.4.1 and A.8.4.2.
Response: We agree and revised to read: ".. .evaluation at two to five times..." to
make consistent with Section A.7.1.
Section A.7.3, last bullet the commenter suggested correcting the typo amend to
"Purified dilution air containing less than"
Response: We agree and revised by adding "containing" between "air" and "less".
Section A.8.1 from line 9, the commenter stated given the note which follows this
text and allows users to control humidity by controlling the fraction of humidified and
dry pure air entering the chamber, should the text be amended to say: "You must
231
-------
control the relative humidity in the test gas throughout the period of passive sampler
exposure." NB RH monitoring frequency is covered in A.8.7.4.
Response: We agree and revised to read: "You must control the relative humidity in
the test gas throughout the period of passive sampler exposure."
Section A.8.7.4, the commenter stated they feel the EPA does not think you need to
record the time on an hourly basis. Plus, given this test must continue for at least 24
hours (A. 8.7), the last bit of the sentence is not needed.
Response: We agree and revised by removing "at hourly intervals or" and
"whichever is greater".
Section A.9.2, the commenter suggested amending "house air" to "humidified air"
Response: We agree and revised by replacing "house" with "humidified".
Section A. 12, the commenter suggested correcting the typo in title.
Response: It is unclear what typo the commenter is referring to, so we did not make
any revisions.
Comment 9: Two commenters stated that demonstrating tube performance each year following
Addendum A is not practical for most laboratories as it requires an environmental chamber. One
commenter suggested applying the criterion of replacing or demonstrating performance every 50
uses or every 2 years. Given that the tubes will be deployed in the field for 2 weeks, it is unlikely
that a tube will be used for sample collections more than 13 times in a year which is more
conservative than the alternative of 50 uses.
Response 9: We agree and revised Section 7.1.6 to read: "at least every 2 years or every 50
uses...".
Comment 10: One commenter suggested that the requirement to use lab blank tubes from the
same conditioning batch as the field tubes should be removed. It is not practical to manage this
requirement if the tubes from multiple projects are being conditioned in a given batch.
Furthermore, allocating a minimum of 2 laboratory blanks per monitoring episode becomes
unwieldy when conditioning batches are typically 20 tubes, and tubes for multiple projects may
be conditioned as a batch.
Response 10: We agree and revised information in Section 9.0 to eliminate the requirement that
the laboratory blank must be from same conditioning batch as field samples. In addition, we
removed the requirement for two laboratory blanks per monitoring episode and replaced it with
one laboratory blank per analytical sequence.
Comment 11: One commenter believes it is unclear how the lab is to flag data in Sections 9.3.2
and 9.3.5 if only one of two blanks fails to meet the criterion (in Section 9.3.2) or if one of two
232
-------
field blanks fails to meet the criterion (in Section 9.3.5). Finally, the commenter recommended
the application of flags to a data set as a result of field precision non-compliance be applied by
the data user rather than the laboratory. Field duplicates may be submitted to the laboratory
without this designation on the chain of custody form, and it may not be clear to the laboratory
from the submitted documentation as to which samples are field duplicates. The commenter
stated that guidance is needed in the rule as to how to handle these biased results when
calculating averages. Since a root cause analysis of an average exceedance that uses such data is
likely to find the exceedance is due to this biased data, the commenter recommended avoiding
that unnecessary burden and excluding biased data from the average calculation.
Response 11: We eliminated the need for data to be flagged based on unacceptable laboratory
blank by revising laboratory blank requirements in Section 9.0 to require corrective action and
successful testing of laboratory blank before analyzing samples. We clarified and added the
following phrase, "If either field blank fails, flag all data..Per Method 325A, field blanks and
duplicates are required to be documented on chain of custody form.
Comment 12: One commenter recommended expanding the range of volume aliquots of the
standard to the sampling end of conditioned tubes allowed from 0.5 to 500 ml to provide a wider
calibration range.
Response 12: We agree and revised to expand the range of volume for standards preparation.
Comment 13: One commenter suggested including EPA TO-17 as an alternative confirmation
technique to active SUMMA canister sampling. The commenter stated that, as highlighted in
A.4.4, canister sampling is not appropriate for all vapor-phase organics in air for example the
main components of middle-distillate fuels which are likely to be present in ambient air near
refineries (references available if required). Given Method 325 relates to sorbent tubes that are
passively sampled, a more appropriate independent method for checking chamber
concentrations, that would have a very similar analytical scope (analyte volatility and polarity
range), would be pumped sorbent tubes, i.e., US EPA Method TO-17. The commenter suggested
a number of amendments to allow the use of US EPA Method TO-17 to confirm chamber
concentrations.
Response 13: We disagree that EPA TO-17 should be included as an alternative to measure
chamber exposure concentration(s). We want an independent method to verify sorbent exposure
concentration(s) and do not want a sorbent method to verify another sorbent method. However,
we recognize that other methods besides TO-15 and Method 18 may be used to verify chamber
concentrations. We revised Addendum A to provide more flexibility by including whole gas
sample collection and analysis or direct interface volatile organic compound measurement
methods.
Comment 14: One commenter stated that one chamber air change per minute is not critical if the
face velocity and the depletion requirement (A.8.4) are both met. While the prescribed exchange
rate may be necessary to meet the face velocity and depletion requirement in cases where very
small volume chambers are utilized (such as the one shown in Figure A-l), chambers described
in ASTM D5116 are on the order of 119 L and can easily meet the face velocity and depletion
233
-------
requirements under conditions of less than one chamber air change per minute. The commenter
recommended removing the air exchange requirement.
Response 14: We agree and determined that other quality control procedures were adequate to
ensure constant target compound in test chambers. We removed the air exchange requirement
from the method.
Comment 15: One commenter noted that several chemicals in Table 12-1 and in the reference
method ISO 16017-2 Table 2 have a published uptake rate of less than 0.5 milliliters (ml)/minute
(min.) Method 325B states that sorbent tube performance is acceptable if the relative accuracy of
the passive sorbent sampler agrees with the active measurements by ± 10 percent at the 95
percent confidence limit and the uptake rate is greater than 0.5 ml/min. It is unclear how to apply
acceptance criteria for these chemicals. According to the commenter, given the range of analytes
of interest, 0.3 ml/min would be a more useful minimum. The commenter suggested amending
the text to "95 percent confidence limit and the uptake rate is greater than 0.3 ml/min (or >1.0
nanogram (ng)/ppm/min.)"
Response 15: We disagree with the suggestion to lower the limit because the lower uptake rates
are subject to greater uncertainty due to temperature and humidity variations (based on EPA's
experience with 1,3-butadiene). We revised to include "equal to" or greater than 0.5 mL/min. To
clarify this point, the upper 95% confidence limit of uptake rates in Table 12.1 are all equal to or
greater than 0.5 mL/min.
8.3 Need for alternative monitoring technologies for fenceline monitoring
Comment 1: Several commenters recommended that the EPA provide sufficient flexibility in its
regulations to allow state and local jurisdictions to develop alternative monitoring programs that
can be demonstrated to provide equal or greater access to monitoring information or the amount
of information. Other commenters stated refineries with open-path monitors along the fenceline
or ambient downwind monitoring stations should not be required to install fenceline monitors.
The commenters indicated that some petroleum refineries have, or may have in the future,
federally-enforceable obligations to install, maintain, and operate population oriented ambient air
monitoring systems. The commenters claimed that these programs focus on the prevalent wind
situation and, because of cost or other obstacles, such as adjoining water bodies, do not require
monitoring around an entire refinery perimeter but require monitoring in the prevalent downwind
direction.
Response 1: We currently find that available real-time, open-path monitoring systems are not
capable of measuring benzene concentrations within the range we expect to see at refinery
fencelines (on average). For that reason, we are not providing a direct allowance to use of an
open-path monitoring system in lieu of the required passive diffusive tube monitoring locations.
However, we believe that the state of technology is advancing and that the capabilities of these
systems will continue to improve. Under the provisions of 40 CFR 63.7(f), owners and operators
can request the use of an alternative test method and we have provided specific provisions at 40
CFR 63.658(k) of the final rule to outline the requirements related to the fenceline monitoring
program. In order to obtain approval for an alternative test method, the owner or operator must
234
-------
demonstrate that the system proposed meets certain minimum requirements. Because alternative
approaches must ensure the same level of protection as the standard that they replace, a request
for an alternative method must demonstrate how the alternative approach achieves that goal. At a
minimum, an alternative approach to fenceline monitoring using Methods 325A and 325B must
have the same level of spatial coverage as the work practice standard. Additionally, the
alternative approach must be able to achieve a nominal method detection limit that is appropriate
for the monitoring objective outlined in the work practice standard. In order to quantify the
emissions detected using an alternative method, we believe any alternative method must be able
to achieve method detection limits that are an order of magnitude below the action level, or 0.28
parts per million by volume (ppbv) benzene. For time-resolved measurement approaches,
detection limits can vary over time based on the specific operational principals of the
measurement instruments/sensors employed, operational conditions and on the maintenance state
of the systems. As an example, for open-path devices, detection limits can depend on optical
signal levels (alignment), atmospheric conditions, etc.; therefore, the alternative method must
include operational procedures to track detection limits over time to ensure data quality
objectives can be met on a continuous basis. Additionally, an alternative monitoring approach
could encompass several different types of instruments and measurement techniques (e.g.
sorbent tubes, open-path monitoring), as long as the entire system provides the same spatial
coverage that is provided by the work practice standard and each monitoring technique achieves
the appropriate method detection limit.
When submitting a request for an alternative monitoring approach, the following should be
included in the request at a minimum: a description of the monitoring instrument and
measurement technique (measurement principle) employed by the instrument; the method
detection limit of the instrument; a map showing the spatial coverage of the monitoring approach
(e.g., locations of paths, locations of point monitors, etc.); initial installation certification
procedures, including criteria for linearity, drift, and accuracy; ongoing quality assurance/quality
control checks and timeframes for such checks, including tracking the detection limit over time
through field verification, if appropriate; means for in-field verification of calibration and other
performance metrics using National Institute of Standards and Technology (NIST) traceable gas
standard challenges; definition for when the system is out-of-control and corrective action for
such periods; and frequency of measurement (e.g., once every 15 minutes). The detection limit
must be based on field verification. For measurements made on a path length (as opposed to a
point), the detection limit should be expressed as a path average concentration.
Comment 2: One commenter does not believe that fenceline monitoring should be required or is
necessary. The commenter stated that the TCEQ has an extensive air monitoring network that is
already in place and functioning to identify and address emissions issues. The commenter noted
that this network gives TCEQ the ability to monitor and manage benzene concentrations in the
communities located near refineries and other industrial sources and to seek and obtain
reductions in emissions, when necessary, from multiple facilities. The commenter went on to
describe the ambient air monitoring data for benzene and other toxics available to the public
through TCEQ's website.
The commenter noted that TCEQ has a program to address persistent high levels of pollutants
and that some areas near petroleum refineries have never had benzene emissions that warranted
235
-------
being placed in that program. Further, the commenter stated, areas near petroleum refineries that
have been placed in the program have successfully lowered emissions to a point that TCEQ has
removed the area from the program for benzene.
Another commenter suggested that the rule should also make provisions for fenceline monitoring
data to be used to improve ambient air standards, in addition to bringing about emissions
reductions. For fenceline monitoring to contribute most effectively to the EPA's goals of
improving health in communities next to refineries, the EPA should develop a plan for using
information generated by fenceline monitoring at refineries across the country to evaluate the
adequacy of action levels set in the rule so as to ensure that those levels are protective enough.
Response 2: We applaud TCEQ's proactive approach for protecting communities near industrial
complexes. As an initial matter, the TCEQ program applies only in Texas and thus does not
address emissions from the refineries located in 35 other states. Second, and significantly, the
purpose and goal of the TCEQ monitoring program is different than the purpose of the fenceline
monitoring program we are establishing in this final rule. Monitors placed in communities some
distance from the refinery, as with the TCEQ program, are not capable of providing information
regarding the emissions from any specific facility. These types of ambient monitors provide
valuable information in that they identify the level of pollutants within a community attributed to
a variety of sources; but they do not provide the type of information that would allow a specific
facility to identify whether it is managing its fugitive emissions consistent with the federal
requirements or to take any corrective action.
Comment 3: One commenter supported incorporating alternative approaches to fenceline
monitoring and is concerned that the proposed monitoring will not be sufficient in most cases to
meet the goals the EPA has set forth. The commenter stated that in light of EPA's justification
that fenceline monitoring is necessary because of uncertainties in the fugitive emissions
estimates and the potential for higher emissions, EPA should allow facilities the use of tools
which identify potential issues as close to their source as possible. Specifically, the commenter
recommended that sources should be allowed an alternative to fenceline monitoring that
increases the Method 21 and/or optical gas imaging obligations to those in subpart H of part 63.
The commenter also recommended that EPA allow Method 21 and/or optical gas imaging for
wastewater drains and floating roof tanks if cost-effective. Another commenter urged the EPA to
speed up development of the OGI camera protocol in 40 CFR part 60 Appendix K and allow it to
be used as an alternate to fenceline monitoring. One commenter suggested that more frequent
tank inspection would be a far less onerous solution than the proposed fenceline monitoring.
Response 3: Although we did propose to allow OGI as an alternative to Method 21, we cannot
finalize this requirement at this time because Appendix K has not been proposed. We continue to
work on Appendix K and we intend to propose that method as soon as it is fully developed. As a
practical matter, the refinery owner or operator may elect to use OGI to help identify a specific
source of emissions when a high fenceline monitoring concentration is measured, and we
specifically noted that OGI could be used as part of a required root cause and corrective action
analysis in 40 CFR 63.658(g)(2). Based on the comment to use a more frequent, direct source
monitoring alternative, we expect the refinery owner or operator would need to submit a request
for an alternative means of emission limitation pursuant to the General Provisions requirements
236
-------
at 40 CFR 63.6(g). We consider it will be very difficult to determine the alternative monitoring
requirements (frequency, detection limits, etc.) that would achieve an equivalent emissions
limitation as the fenceline monitoring program short of performing inspections of tanks,
monitoring of all equipment components, and monitoring for detectable emissions around all
wastewater drains and tanks and other sources of emissions (e.g., gasoline loading racks, marine
vessel loading operations) every two weeks. As noted previously, we included provisions at 40
CFR 63.658(k) of the final rule to allow refinery owners or operators to request an alternative
test method because we found that use of an alternative method would be relatively feasible and
common. Based on the anticipated costs of more frequent monitoring of all potential fugitive
emission sources, we do not consider it likely that facilities will seek an alternative emissions
limitation for the fenceline monitoring work practice standards and, therefore, we did not include
specific procedures in Refinery MACT 1 for requesting an alternative emissions limitation for
the fenceline monitoring work practice standards. However, refinery owners/operators may still
submit such a request to the Administrator pursuant to 40 CFR 63.6(g).
Comment 4: One commenter suggested that the EPA require use of active monitors at large
refineries and those located in close proximity to population centers and claimed active monitors
are preferable for carcinogens, such as benzene, and other substances of high toxicity. The
commenter stated that the EPA has a number of complex responsibilities in relation to U.S
refineries, including but not limited to reducing overall emissions as proposed in this rule,
preventing catastrophic accidental releases under section 112(r), and facilitating a community's
emergency response capabilities in the event of a significant release. The commenter noted that
while the EPA is in the process of finalizing this rule, it is also soliciting comments on possible
revisions to its current Risk Management Plan, 40 CFR part 68 (RMP). The commenter
contended that the EPA may in the near future require active monitoring systems in order to
achieve objectives under the RMP. The commenter suggested that the EPA instead consider
requiring the use of active fenceline monitoring systems under this rule and that such a system
could be expanded, if needed, to fulfill the purposes of the RMP.
Response 4: Based on the current state of technology, we are skeptical that one type of
monitoring system can serve both fugitive management purposes and emergency management
purposes. This is because the fenceline work practice standard appropriate for managing
fugitives is aimed at detecting benzene at low concentration on average over time. Also, as
discussed in the preamble to this final rule, passive sampling is capable of these low
concentration measurements but active real-time open path systems are not. In comparison,
monitoring for the purpose of emergency management requires real-time quality assured
measurement capability and may not need to detect low concentrations since emergency events
will presumably occur at elevated concentrations and for shorter periods of time.
Comment 5: A commenter supported the proposed passive sampling sorbent tubes fenceline
monitoring requirement, stating that it is reliable, economical, and scientifically sound.
Additionally, the commenter pointed out that it uses no electricity during sampling and is easy to
handle and ship. The commenters noted that these merits have been recognized and utilized by
other countries, especially countries in European Union and in Asia.
Response 5: We appreciate the support.
237
-------
Comment 6: One commenter suggested that robust community participation in planning,
implementing, and overseeing fenceline monitoring is necessary to ensure that monitoring
systems remain state-of-the art and are trusted by the public. The commenter asserted that
monitoring and information technology are constantly advancing, and residents of refinery
communities can be a source of ideas about how new social networking platforms, for example,
could make monitoring information more accessible and ensure that it is used more fully.
The commenter further claimed that active community involvement is important because,
without it, information generated by monitoring is unlikely to be credible to the public. The
commenter pointed out that community members often dismiss as biased information generated
by refinery-funded programs even when overseen by regulatory agencies. The commenter stated
that community involvement is necessary to counteract the perception that monitoring programs
are designed to show only what companies are willing to divulge by ensuring that companies do
not overlook issues of concern to the community. The commenter concluded that the rule should
require companies to arrive at plans for fenceline monitoring in consultation with neighboring
communities and to provide on-going opportunities for community involvement.
Response 6: The final rule requires facilities to monitor concentrations around the fenceline and
to make the fenceline monitoring data available to the public. The final rule also contains
substantive QA/QC requirements for the monitoring methods that ensure the data collected are of
high quality. While we agree that it is important that neighboring communities understand and
believe the data collected, we disagree that the rule should require that the facility owners or
operators engage each neighboring community in planning, implementing and overseeing
monitoring at a given facility. Thus, while we encourage refineries to keep open lines of
communication with local communities to let them know about the monitoring programs they
put in place, respond to concerns raised by communities and answer questions about the
monitoring program and its results, we find that the monitoring and reporting requirements
provided in the final rule are sufficient to ensure quality data are collected and that the
monitoring results are readily available to local communities in a timely and transparent manner.
Comment 7: One commenter asked why the EPA did not take the approach recommended by a
panel of national air monitoring experts who recently published a report on current air
monitoring capabilities near Bay Area refineries and potential air monitoring technologies,
methodologies and tools to provide air quality information for communities near refineries and
gather data to evaluate health impacts associated with air quality near refineries and track air
quality changes and trends over time near refineries. The Expert Panel generally agreed that an
approach that utilized a combination of fenceline, community, and mobile monitoring would be
required to adequately define exposures during normal operations and when upsets and incidents
occur. The fenceline monitoring would be leveraged primarily to identify non-routine emissions
during normal operation, while the community monitoring would be utilized to develop spatial
gradients of chronic exposures. Mobile monitoring would be used to supplement on-going
monitoring during major upsets and incidents and to help develop information on spatial
variability.
Response 7: As an initial matter, under both sections 112(f)(2) and (d)(6), the EPA was
evaluating if it was necessary to require further reductions from the refineries source category.
238
-------
We do not agree that community monitors can isolate emissions from specific sources and thus
they are not an effective tool, at least at this time, in regulating emissions as required under either
of these CAA provisions. Thus, we did not consider community monitors in this rulemaking. We
note, however, as indicated by a number of other commenters, that many communities near
refineries already have community monitors that provide data on pollutant levels in the
community that are attributable to a broad spectrum of sources.
Regarding mobile monitors to address upsets, we do not agree that CAA sections 112(d)(6) and
(f)(2) are the correct tools for determining how to identify or address violations of standards.
Rather, these provisions are focused at determining whether the existing standards should be
tightened either because of developments in processes, practices and control technologies or
because the existing standards do not protect public health with an ample margin of safety. We
note that separate from this rulemaking, EPA is involved in a leading research group in the
mobile measurement area and including the use of mobile measurements for fugitive emission
detection (Refer to OTM 33A and other). As part of EPA's research it has facilitated
development of and demonstrated the use of the first commercially available prototype mobile
benzene measurement system (based on cell-based UV optical spectroscopy) for detection of
benzene emissions near refineries and other facilities.
In identifying the available technologies, the EPA talked with experts, including vendors of
monitoring systems and facilities that operate them. We then evaluated a variety of monitoring
methods, as discussed elsewhere in the preambles to the proposed and final rules and this
response to comment document, considering technical capabilities, such as monitor detection
limits and coverage, and costs. We concluded that, for the purposes of improving fugitive
emissions management, fenceline passive diffusive tube monitoring was the best monitoring
alternative.
Comment 8: One commenter indicated that he does not think that the present proposal requires
monitoring stations to be elevated to a level that would reliably monitor fugitive emissions from
towers and other elevated equipment. The commenter noted that there are "continuous
emissions" that are found every time downwind measurements are taken, and they are not
detected by ground-level monitors. The commenter stated that SOF and DIAL have the
advantage of being able to measure concentrations from the ground to the sky and that they also
use wind speeds to quantify their pollutant leak rates.
Response 8: We identified fenceline monitoring as a development in practices, processes and
control technologies for improved management of fugitive emissions at refineries. "Elevated
equipment" typically means equipment that is vented to a stack or a control device and does not
fall within the ambit of fugitive emissions. While some equipment, like distillation columns are
tall and can have elevated equipment components that may leak, these tall sources are generally
located near the center of the refinery so that the emissions released are dispersed and
measureable at ground levels by the time they reach the property boundary. For the purposes of
improving fugitive emissions management, a system that can measure levels at ground level is
the most effective tool.
239
-------
8.4 Applicability of fenceline monitoring requirements
Comment 1: One commenter objected to the requirement that operators must be in compliance
with the fenceline monitoring provisions of section 63.658 in order to use optical gas imaging
(provided in 40 CFR part 60, Appendix K) as an alternative to using EPA Method 21. The
commenter claimed EPA provides no rationale for this requirement. The commenter asserted that
since Appendix K is equivalent to, and arguably better than Method 21 monitoring, there is no
reason for EPA to impose any additional conditions for the use of optical gas imaging
technology.
Response 1: Use of Appendix K would not alleviate the need to comply with the fenceline
monitoring requirements. Petroleum refinery owners and operators would be required to meet
both the equipment leak survey requirements and the fenceline monitoring requirements. We
were seeking to develop Appendix K as an alternative equivalent method to EPA Method 21, so
that refinery owners or operators could elect to use Appendix K as an alternative to Method 21.
However, Appendix K has not been proposed, and will not be available as an alternative to
Method 21 until it is finalized. Therefore, the final rule does not reference the use of the optical
imaging camera as an alternative to Method 21.
Comment 2: One commenter stated that his refinery idled its process units in 2012 and is
presently operating as a terminal. However, because the refinery is idled rather than permanently
shut down, the refinery continues to apply the provisions of Refinery MACT 1 and 2 rules to the
facility. The commenter stated that it would be helpful for the EPA to exclude idled refinery
locations from the fenceline monitoring requirements because such facilities have an emissions
profile much more similar to a terminal operation.
Response 2: If the facility is subject to Refinery MACT 1, then implementation of the fenceline
monitoring standard would be required. If the facility no longer intends to operate as a petroleum
refinery, the facility owner or operator may elect to have their permits revised to indicate that the
refinery operations are permanently being discontinued and seek permitting solely as a terminal.
The idled refinery remains subject to Refinery MACT 1 and any revisions to the MACT until it
is no longer identified as being subject to Refinery MACT 1.
Comment 3: One commenter stated that the proposed rule would require every refinery to install
fenceline monitors at significant ongoing costs despite very low levels of risk. The commenter
stated that this one-size-fits-all approach will not improve public health, particularly when EPA's
analysis shows that a number of these refineries pose no health risk to the surrounding
communities. The commenter noted that one quarter of the refineries had maximum individual
risk levels below 1-in-l million, and two-thirds had risk levels below 10-in-l million. Another
commenter also claimed that risk does not support the fenceline monitoring requirement, noting
that 115 of the 142 facilities analyzed that have a cancer risk less than 10-in-l million. This
commenter asserted that fenceline monitoring should not apply to these 115 facilities.
Another commenter stated that the one-size fits all approach to fenceline monitoring fails to
comply with EO 13563, which mandates that agencies take a regulatory approach that reduces
burdens and maintains flexibility and freedom of choice. The commenter stated that the fenceline
240
-------
monitoring proposal fails to account for the specific conditions at the regional and local levels or
risk to surrounding communities. The commenter concluded that EPA should tailor the
requirement to those refineries in higher risk areas.
A commenter claimed that EPA should exempt Small Business Refineries which have inherently
lower emissions and disproportionately represent facilities located in remote locations. The
commenter stated that small refiners are concerned that this proposal would impose unnecessary
costs on small entities and would serve no useful purpose where there are no nearby downwind
off-site residential receptors. The commenter concludes that the "uncertainty" EPA relies on for
the requirement by narrowly tailoring monitoring to sources of uncertainty or gathering data on a
limited basis to better understand those sources of uncertainty.
Response 3: As an initial matter, the commenters erroneously assume that EPA is promulgating
the fenceline monitoring requirement to address risk. As EPA has noted in the preambles to both
the proposed and final rules, EPA is promulgating the fenceline monitoring requirement as a
cost-effective development in processes, practices or control technologies. We disagree that the
fenceline monitoring program cannot be implemented cost-effectively, even at smaller refineries.
As proposed, we are establishing requirements for the number of monitoring locations based on
the size of the refinery. We are providing facilities an option, after two years of data collection,
to submit site-specific monitoring plans to reduce the monitoring frequency as described in more
detail in the preamble and elsewhere in this document.
8.4.1 Require tiered levels of monitoring
Comment 1: One commenter was glad to see the fenceline monitoring proposal. The commenter
recognized that refineries might dispute the need for this monitoring based on a claim that they
are operating their refinery well. The commenter noted that those that are well managed will not
have violations of the 2 week average. The commenter supported adoption of the full fenceline
systems designed and built in the San Francisco Bay Area, which the commenter claimed has
upgraded community based technologies of auto-sampling hardware and particulate monitors
added to the fenceline lasers and chemical detection systems.
The commenter suggested that EPA could exempt from the full San Francisco Bay Area system
those refineries that do not exceed the EPA's minimalist detection system's limit for a year
unless a significant off-site impact occurs. Also, the commenter suggested, if a leak, explosion or
fire has off-site impacts, the requirement for the full system would be triggered. The commenter
realized that this is an after-the-fact monitoring system, but also that requiring all refineries to
build the full system may not be politically possible. The commenter further recommended that
if the EPA adopts this tiered approach to monitoring, the EPA should provide requirements for
the more complete system and the commenter detailed requirements that EPA should include.
The commenter noted that a full real-time system can warn the local public of an ongoing
release. The commenter further stated that the proposed systems would be a setback locally if
they replaced the current Bay Area systems or the systems about to be required by the local Air
District.
241
-------
Finally, the commenter suggested that the EPA consider adopting the BAAQMD's fenceline
system siting guidance which they claim is more robust than the limited fenceline systems
proposed.
Response 1: We appreciate the commenter's concerns, but for the reasons provided in the
preamble and elsewhere in the response to comments document, we disagree that the type of
system they recommend is as effective for improved management of fugitive emissions at the
refinery. We further note that state or local agencies may elect to require additional monitoring
techniques at some refineries and we do not anticipate that those areas with monitoring
requirements already in place will consider the fenceline monitoring approach in the final rule to
be an adequate substitute for the myriad purposes for which they are requiring monitoring.
8.4.2 Limit time or number of locations that need ongoing monitoring if
concentrations are low
Comment 1: A number of commenters indicated that the fenceline monitoring program should
provide relief for facilities that always achieve low monitored levels. For example, the
commenters stated, a refinery that consistently reports fenceline concentrations below the
corrective action level should be allowed to discontinue the program, if the facility is located
within a region that contains a rigorous monitoring network.
Several commenters suggested that if a facility reports less than half the action level for a period
of two years, then the frequency of monitoring be reduced to once every three years. One
commenter suggested that the EPA should consider waiving the fenceline monitoring program
for sites with modeled concentrations at or below the concentration action level or a threshold
used by a state to demonstrate protectiveness for a permit action.
Several commenters stated that, after an initial period is completed establishing that the normal
baseline of a particular refinery is below the action level, less frequent monitoring should be
allowed. These commenters recommended that, after two years of demonstrating a background
corrected maximum fenceline annual average concentration below the action level, monitoring
frequency be reduced to one 2-week period every quarter. Another commenter suggested two
consecutive 2-week periods per quarter. The commenters stated that if the background corrected
annual average benzene concentration based on the quarterly monitoring exceeds the action
level, a return to every two week monitoring would be required along with meeting the RCA/CA
requirement. The commenter further provided that reduced frequency should be available again
after one year of meeting the action level. Further, one commenter requested that a permanent
off-ramp be provided after five years of data below the background-corrected action level.
Commenters stated this is consistent with many U.S. EPA regulations that offer facilities the
opportunity to reduce the sampling frequency when measured values are consistently low or
below action levels. In this instance, it would further demonstrate to refineries that through good
housekeeping and generally good environmental management they are able to aggressively
identify and repair sources of fugitive emissions. Reducing their fugitive emissions will reduce
their contribution to fenceline monitoring levels and hopefully it will allow them to reduce their
benzene monitoring costs by keeping fenceline levels consistently below the action level.
242
-------
One commenter stated that while they continue to support use of Perimeter Boundary Monitoring
(PBM) systems for information gathering only, a more robust system could, nonetheless, provide
for future coordination of the dual monitoring systems (referring to other conventional fugitive
monitoring programs). EPA could structure the PBM requirements to allow facilities to
periodically review the effectiveness of the monitoring systems. For example, for an initial
period of two years, all refineries could use both conventional fugitive emissions monitoring and
a PBM program. After this initial period, the EPA should eliminate unnecessary monitoring by
allowing refineries the option of using the more effective compliance approach, if a single
monitoring approach would not leave unacceptable gaps in the monitoring protocol. If there is
not a more effective program at a given refinery, then the refinery would continue using both
methods, and re-evaluate the systems at two year intervals to determine whether either system
proves more effective in subsequent years.
Other commenters strongly supported that the fenceline monitoring remain in place at facilities
and that EPA not create an "off-ramp" for consistent compliance. These commenters were
concerned that a facility could have an unexpected event that would be undetected without the
monitors in operation. The commenter also contended that knowing that their fenceline
concentrations will be publicly available provides strong incentive for sources to maintain their
operations in good working order. The commenter stated that if the EPA does alter the fenceline
monitoring requirements for sources that consistently measure concentrations below the action
level, any subsequent change in the operations at the facility should require the facility to revert
to the full fenceline monitoring requirements.
Response 1: The goal of the fenceline monitoring program is to improve the management of
fugitive emissions by identifying emission sources quickly and reducing these emissions through
early detection and repair. Therefore, we believe that coverage of the perimeter and ongoing
monitoring are necessary elements of the fenceline monitoring work practice standard. However,
we do consider that there will be cases where certain "upwind" monitors may consistently
monitor low levels of benzene and may not be necessary in the long term. Therefore, we are
providing facilities an option, after two years of data collection, to reduce the monitoring
frequency for sampling locations that consistently read background concentrations at or below
0.9 ug/m3, which is 10 percent of the action level and approximates the benzene background
concentrations we observed from pilot studies. Specifically, we are allowing refinery owners or
operators to monitor every other two weeks (i.e., skip period monitoring) if over a two-year
period, each sample collected at a specific monitoring location is at or below 0.9 |ig/m3. If every
sample collected from that sampling location during the subsequent two-years is at or below 0.9
|ig/m3, the monitoring frequency may be reduced from every other two weeks to quarterly. After
an additional two years, the monitoring can be reduced to semiannually and finally to annually,
provided the samples continue to be at or below 0.9 |ig/m3 during all sampling events. If a
concentration greater than 0.9 |ig/m3 is measured, the sampling frequency reverts to the biweekly
monitoring frequency.
8.5 Benzene as target analyte
Comment 1: One commenter stated that while benzene is a common constituent in a number of
refinery feed stocks, its concentration may vary widely from refinery-to-refinery depending upon
243
-------
their configuration and crude slate. The commenter stated that benzene cannot be utilized for
other industrial facilities such as the chemical manufacturing industry because of the high level
of variability from process-to-process and facility-to-facility. The commenter cited a number of
reasons why they believe fenceline monitoring would be problematic for the other industries.
Response 1: This rule does not address requirements for the chemical manufacturing industry
and concerns about a potential fenceline monitoring program for those sources can be raised in
any future rulemaking in which the EPA proposes such a requirement.
Comment 2: Some commenters claimed that benzene is appropriate to monitor for refineries
because it is the most ubiquitous HAP, present in all refineries, and present in most HAP streams
common to basic refinery processes. They further stated that it is generally one of the higher
mass HAP emitted by refineries and thus would be among the easiest to monitor with samplers.
In comparison, they claimed, other HAPs would usually be present in lower concentrations and
therefore be harder to analyze and result in more non-detects. They also stated that longer sample
times and multiple samplers and multiple analyses would likely be required if sources are
required to monitor additional HAPs and that such additional monitoring would not lead to an
improvement in understanding refinery emissions though it would add to costs.
One commenter supported the use of benzene as an indicator for VOCs for two distinct reasons:
First, benzene is a largely ubiquitous class A carcinogen that likely represents the majority of the
human health risk associated with refineries; second, the use of benzene as a surrogate for the
other VOCs presents fewer sampling and analysis problems than those associated with
measuring semi-volatile PAH compounds that require different sampling media and analytical
methods. However, the commenter also suggested that total hydrocarbons be measured
concurrently with benzene to ensure that none of the samplers have been overloaded and
therefore benzene undercounted.
A number of other commenters supported monitoring of multiple pollutants.
One commenter stated that since the EPA is legally mandated to require the monitoring of each
and every toxic chemical and substance.
Another commenter stated that different fugitive sources at a refinery will have different
concentrations of the various HAPs and that readings on multiple pollutants can help profile and
identify the pollutant source.
Another commenter disagrees with EPA's statement that benzene "is ubiquitous at refineries,
and is present in nearly all process streams such that leaking components generally will leak
benzene at some level." The commenter raised three issues. First, the commenter contended that
the EPA has not provided any analysis of the ICR data showing that benzene is present in nearly
all leaking process streams. The commenter claimed that it has presented data from ambient air
quality monitors that shows that there are many instances when toluene and hexane are present in
the absence of benzene. Second, the commenter stated that the Act requires the EPA to set limits
on all hazardous air pollutants and thus the EPA needs to assure compliance with all of the
limits, not just benzene. Third, the commenter contended that EPA has not demonstrated that
244
-------
benzene may be used as a surrogate for all other relevant fugitive pollutants. The commenter
cited to Sierra Club v. EPA, 353 F.3d 976, 984 (D.C. Cir. 2004) (quoting Natl Lime Assn, 233
F.3d at 639), and claimed that a surrogate might be reasonable only if it meets, at least, the
following three conditions: (1) the target pollutant(s) must invariably be present in the surrogate
(2) the control technology used to control the surrogate must indiscriminately capture both the
surrogate and the target pollutant(s); and (3) control of the surrogate must be "the only means by
which facilities achieve reductions in the target pollutant." The commenter stated that the same
principle is applicable to monitoring for a surrogate and because benzene is not actually present
in all refinery pollution streams the EPA has failed to meet the surrogacy test to require
monitoring of only benzene.
One commenter stated that while benzene may be an excellent proxy for emissions of
organic HAPS generally, it is a dubious indicator of emissions of heavy metals such as lead,
mercury, cadmium and vanadium. These elements disperse differently in the atmosphere
than organic compounds, build up in the environment differently, and accumulate in people
differently. If the passive diffusive tube technology is chosen, the EPA should consider a
tube that absorbs mercury or lead as well.
Response 2: We reiterate that we did not intend to measure all pollutants, especially pollutants
that are emitted from point sources that are directly measurable through source tests and
continuous monitoring systems. These emissions sources and pollutants are subject to other
standards under these MACT.
We disagree that it is necessary to monitor for every HAP emitted from fugitive emission
sources at refineries. Petroleum refining streams can contain dozens of HAP and it is very
difficult for any method to detect every HAP potentially emitted from refineries. The fenceline
monitoring standard was proposed as part of the technology review to improve management of
fugitive emissions and not as a risk reduction measure. In order to meet that goal of improved
management of fugitive emissions, it is not necessary to obtain an accurate picture of the level of
all HAP emitted. In the proposal, we cited our reasons for using benzene as an appropriate HAP
to monitor to meet the goal of the fenceline monitoring program (79 FR at 36924). First, benzene
is present almost exclusively in emissions from fugitive emission sources (and not from point
sources) and is emitted from each of the different types of fugitive emission sources; thus it is a
very good indicator of fugitive emissions. It is easily measured and can be detected with good
precision by passive monitors.
Hexane was considered as a potential surrogate because it is generally present in higher
concentrations than benzene in many refinery process streams, but hexane is nearly insoluble in
water (solubility of 12.4 mg/Liter (L) versus 1,750 mg/L for benzene) and is therefore not
significantly emitted from wastewater collection and treatment systems. Therefore, we did not
consider hexane to be a good general surrogate for all fugitive emissions sources. Toluene is
another ubiquitous compound similar to benzene and commonly present in concentrations
similar to or higher than benzene. Its water solubility is about one-third that of benzene but is
still soluble enough to be regularly emitted from wastewater treatment systems. As a secondary
consideration, we expected neighboring communities to be more interested in ambient benzene
concentrations than other HAP because benzene is a known carcinogen and was a significant
245
-------
contributor to the health risks associated with refinery emissions. Therefore, benzene was
selected over toluene as the surrogate compound.
Although it is technically feasible to require further speciation of HAP collected within a single
sorbent tube, increasing the analyte list does increase slightly the analytical costs because
additional calibration standards are required. We considered it more direct to use a single HAP as
a surrogate for all fugitive HAP. It is easier to establish an appropriate action level, reduces
analytical costs and simplifies the determination of compliance for refinery owners and
operators. We did not include other HAP when determining modeled fenceline concentrations,
and had we, there would be questions as to whether each HAP would have its own action level or
if we would set a single action level for the sum of HAP measured. Of the potential HAP
considered for use as a surrogate in the fenceline monitoring standard, only toluene and benzene
were considered to be reasonable surrogates that were emitted from all fugitive sources
(including wastewater). Given the similarities in these compounds and the similarities in
concentrations across different refinery process streams, we determined it was unnecessary to
require the quantification of both in the fenceline standard. We then considered our secondary
purpose, which was to ensure our risk decisions were based on correct and accurate emissions
inventory information. Because benzene was also a key contributor to the MIR associated with
refinery emissions, we determined that monitoring of benzene as a surrogate for fugitive HAP
emissions was appropriate.
8.6 Adjusting for background emissions
Comment 1: Several commenters believed that interferences and confounders and the inability
of the methodology to segregate individual regulated emission sources or even emission types,
suggests that fenceline benzene levels cannot be used to determine compliance with any of the
applicable Refinery MACT 1 standards. One commenter also noted that Refinery MACT 2 does
not address fugitive emissions, so any impact on the fenceline results from Refinery MACT 2
would be an interference with the stated purpose.
Commenters stated that throughout the preamble and proposed new regulations, the EPA has
stated that refineries can collect background samples, duplicate samples, or increase the total
number of sampling sites along the fenceline to rule out contributions of near-field sources of
benzene. Many refineries are located in recognized industrial zones or adjacent to major
highways with significant vehicle emissions. Adequately characterizing background sources of
benzene will require the development and implementation of sophisticated monitoring programs
that include monitoring above and beyond the proposed minimum monitoring required by 40
CFR part 63 subpart CC. Refineries will have no choice but to rule out contributions of near-
field sources of benzene because the modeling performed by the EPA to establish the benzene
action level did not include benzene contributions from sources outside of refineries that
potentially influence fenceline benzene concentrations, and the time allowed for this analysis
under the proposed rule is short, as both the root cause analysis and the corrective action analysis
must be completed with 45 days after determining there is an exceedance.
On the other hand, one commenter stated that the proposal's method to identify background
sources will cause the fenceline concentration attributed to the regulated facility to be biased
246
-------
low. Under the proposal, each sampling period will have its own background concentration that
is defined as the lowest reading during the period. Of course, as EPA recognized, a facility's own
emissions will contribute to the lowest reading at the fenceline, especially during periods with
calm or variable winds. This will result in the background concentration being overestimated and
the facility's contribution being underestimated. One commenter echoed this concern, stating
that the proposed sampling design produces refinery-attribution data which are non-conservative.
Another commenter also believes that the proposal will produce low bias of facility contribution
because it allows facilities to propose site-specific monitoring options when the facility believes
that a near-field source is underestimated by the general method. There is no equivalent option
for the EPA, the public, or refinery communities to require an alternate method when the general
methodology overestimates background concentrations.
Response 1: First, several commenters appear to have a misconception of what a Refinery
MACT 1 source is and some commenters appear to suggest that Group 2 streams are not
Refinery MACT 1 sources. As provided in 40 CFR 63.640(a), Refinery MACT 1 "applies to
petroleum refining process units and to related emissions points that are specified in paragraphs
(c)(1) through (9) of this section that are located at a plant site and that meet the criteria in
paragraphs (a)(1) and (2) of this section." Paragraphs (a)(1) specifies that the facility is a major
source of HAP emissions and paragraph (a)(2) specifies the equipment contain or contact a HAP.
Paragraphs 63.640(c)(1) through (c) (9) clarify that the rule applies to all miscellaneous process
vents, all storage vessels, all wastewater streams and treatment operations, all equipment leaks,
all gasoline loading racks, all marine vessel loading operations, all equipment associated with
bulk gasoline terminal or pipeline breakout stations, all heat exchange systems, all releases
associated with decoking operations. There are exclusions to Refinery MACT 1 applicability in
paragraphs (d) and (g) of section 63.640 which exclude Refinery MACT 2 vent sources, ethylene
processes and HON units, among other sources. However, it is clear that Refinery MACT 1
specifically applies to all refinery sources and not only to Group 1 sources, which is clearly seen
in the emissions averaging equations in 40 CFR 63.642 and 63.652. So our first point is that
Refinery MACT 1 applicability is broad and covers essentially all HAP sources at the facility
that are not explicitly excluded. The fenceline monitoring program helps to ensure that the
emissions from all of these sources, including Group 2 sources are managed well.
With respect to on-site contributing sources, we note that we included all emission sources from
the submitted emissions inventories, which included co-located operations, such as HON units.
As such, we consider that the action level should be achievable from petroleum refineries while
including these sources. We provide a mechanism by which refinery owners or operators can
develop and implement a site-specific monitoring plan to account for the contribution of these
non-refinery MACT 1 sources to the measured concentrations at specific monitoring sites. We
provided this primarily due to legal considerations as to what we have the authority to cover
under the Refinery MACT 1 standard given the clear exclusions in Refinery MACT 1 for certain
sources and not because we expect facilities will need this provision in order to comply with the
AC concentration action level.
Finally, with respect to off-site sources such as highways or neighboring facilities, if these
facilities are downwind based on the prevailing wind direction, these sources will likely have
247
-------
minimal impact on the fenceline concentrations measured by the monitor. If this source is
upwind of the refinery, the monitors near these sources are likely to be affected but would then
generally result in a higher correction when determining AC due to higher background
concentrations. Furthermore, based on the API study, there was little evidence of off-site sources
significantly impacting the refinery concentrations. The facilities monitored in the API study
were blinded, but as this issue has been expressed previously by API, one would expect that at
least one or two of the sites selected would have been in industrial locations with the types of
issues the commenters suggest will require facilities to develop and use site-specific monitoring
plans. We see no evidence from any of the studies to date that suggest emissions from off-site
sources will somehow overwhelm the emissions contributions from the refinery when
determining AC.
We disagree with the commenters suggesting that the time for the root cause analysis is short
because the facilities must first assess background sources. We have provided up to two years to
begin official monitoring for determination of AC. We fully anticipate that facility owners or
operators will begin monitoring before this time to determine if they do have any significant
issues with off-site source contributions and need to develop and submit a site-specific
monitoring plan. Even if they do not have issues with off-site source contributions, the AC action
level is determined on an annual average basis. We fully expect that facilities will investigate
high monitor readings throughout the year to help identify and correct the source of elevated
concentrations to prevent having an exceedance of the AC action level. Based on the time we
provide to start official monitoring and that facilities will collect an entire year of monitoring
data before calculating an actionable AC value, we intend that facilities will have time to
determine if a site-specific monitoring plan is needed.
In the proposal we discussed the theoretical concern that a refinery's own sources of benzene
could cause elevated background concentrations that, when subtracted from the highest measured
concentration, would bias low the contribution of the refinery. Since the proposal, we conducted
more studies and reviewed industry study data that indicate that background concentrations at
refineries are low and fairly stable. In the API 12-refinery study, the average background does
not change significantly from urban to rural setting, and remains relatively constant (from 0.1 to
0.3 ppb). Further, most every instance of an elevated fenceline reading is associated with a
known refinery emission source located upwind and we do not see any cases where the all the
readings are high, and the AC is very low. Therefore, we continue to believe this simple
background correction is still valid and provides an accurate method to correct the highest
fenceline concentration for background. While we are retaining the provisions for refinery
owners or operators to develop a site-specific monitoring plan, the level of monitoring needed to
quantitatively account for the contribution of an off-site source for each monitoring location (for
which an adjustment is sought) is not trivial. We do not expect many refinery owners or
operators will elect to develop a site-specific monitoring plan based on the monitoring studies
that have been conducted to date.
8.6.1 Delta C approach
Comment 1: One commenter believed that the proposed rulemaking has appropriately
recognized that ambient concentrations can be affected by many factors and other sources,
248
-------
although the proposed approach to adjust for background concentration is significantly flawed
because in subtracting the lowest observed concentration from each of the other monitoring
stations, the agency has created an artificial average concentration for the facility that may be
much higher than a level that is actually measured. It could inflate the facility average by as
much as 40 to 50%. The commenter believed that a more appropriate method is to use a simple
arithmetic average of all readings to establish a facility wide concentration for a given period.
Further the commenter believed that neighboring or co-located sources could be above the
corrective action thresholds and that the procedures for adjusting the background concentrations
are not sufficient and that a higher corrective action threshold may be necessary to account for
these sources.
Response 1: We do not agree with the commenter's suggestion that refinery owners or operators
should calculate the average concentration from all fenceline monitors to compare against the
action level. The approach is to determine the facility's resulting fenceline concentration and the
AC approach, which requires subtracting the lowest observed concentration at any monitoring
location from the highest concentration over any given period, provides a better means to assess
the facility's impact than simply averaging the concentrations around the refinery fenceline.
Refineries with co-located sources may elect to comply with the fenceline AC action level value
for the entire facility. However, as in the proposal, the final rule allows facilities to prepare and
submit a site-specific monitoring plan by which they can quantify the impact of these co-located
sources on the concentrations measured at each monitoring location.
Comment 2: One commenter recommended that the rule should include provisions for
excluding outliers in 40 CFR 63.658(f)(1). A record of excluded data and the reason for the
exclusion should be required. Such an allowance is alluded to in the reporting requirements in
proposed 40 CFR 63.655(h)(8)(iii), where reports of such exclusions are required, but the fact
that such exclusions are allowed needs to also be included in 40 CFR 63.658(f). The commenter
also recommended that the rule provide guidance on how to handle biased results when
calculating averages from data that is flagged when field blanks contain greater than 1/3 of the
measured target analyte or compliance limit for field samples. Since a root cause analysis of an
average exceedance that uses such data is likely to find the exceedance is due to this biased data,
this would avoid that unnecessary burden.
Commenter also recommended that section 63.658(f)(3) should clarify that the annual rolling
average is defined as the 26 sampling period rolling average, rather than the 12-month rolling
average to avoid confusion about the period of time reported for each rolling average point.
Response 2: We note that Method 325A at Section 9.2 includes a discussion of potential
outliers. Specifically, a potential outlier is a result for which one or more sampling tube do not
agree with the trend in results shown by neighboring sampling tubes or with previous and
subsequent results for that sampling location. Section 9.2 (Method 325A) specifies that:
"Accidental contamination by the sample handler must be documented before any result can be
eliminated as an outlier. Rare but possible examples of contamination include loose or missing
storage caps or contaminated storage/shipping containers." Documented missing or loose caps
could be immediately flagged as a potential outlier.
249
-------
An unusually high value may be an indication of an unexpected emissions source, like a new
large leak. If a facility immediately investigates the high value by using an optical imaging
camera and identifies a leak or other emissions source that is corrected, then the high
concentration would not be an outlier, but an accurate indication of unusual emissions. If no
emissions source is identified, and the subsequent 14-day sample is again consistent with the
previous (pre-outlier) measurements, then the facility owner or operator would have a basis for
expecting the unusual measurement is an outlier and could cite this as part of a root cause
analysis. We note that some variability can be expected due to changes in temperatures, changes
in refinery operations, and changes in wind speed and direction.
Finally, we agree the use of the term "rolling 12-month average" is imprecise. For convenience,
we have revised 40 CFR 63.658(f)(2) to describe the calculation of "the annual average AC"
shall be determined as the average of the 26 most recent sampling intervals. In 40 CFR
63.658(f)(3), we now consistently use the phrase "annual average." With the clarification in 40
CFR 63.658(f)(2), it is clear that the annual average is a rolling average of 26 14-day sampling
period results.
Comment 3: One commenter stated that because the term ACi is used for the individual
corrected sample results, proposed paragraph (iii) seems to say AC for this sampling episode is
the maximum value of the corrected sampling rather than the difference between the maximum
and minimum corrected values, which is how AC is defined. That is illogical, since the site-
specific sampling is intended to correct for backgrounds and confounders in addition to the
benzene reaching the site from upwind. The definition of uniform background (UB) in paragraph
(i) makes this clear by limiting that term to the additional measurements specified included in the
site-specific monitoring plan and thus does not include the basic adjustment for upwind sources.
The commenter recommended the term ACi be replaced with the term Ci to represent the
individual corrected sample concentrations and that (A)C for each sampling episode then be
calculated as prescribed in 63.658(f) using the corrected values rather than the as measured
values.
The commenter also noted that there is a wording error in the definition of UB in paragraph (i)
that should be corrected here and in the Fugitive Emissions section of the rule preamble (page
36924). The subscripts are missing from the equation in the preamble and the multiplication
symbol should be deleted, i.e. "HFC = Maximum x (MFC-OSC)" should be "HFC = Maximum
(MFCi - OSCi)" The terminology and symbols that EPA decides on for 63.658(i)(2) should be
used consistently in the fugitive emissions section as well to avoid confusion.
Response 3: We disagree. We expect that the site-specific plan will assess the near-field source
contribution to a specific monitoring location and the UB. Subtracting these terms from the
measured fenceline concentration is the ACi for that monitoring location. One cannot correct the
fenceline concentrations for near-field sources and UB and then subtract out the lowest ACi from
the highest ACi. We agree that the preamble should not have the multiplication sign, but we see
no error in the equation for ACi in the proposed rule. We appreciate the comment regarding the
definition of UB in the rule and we have deleted the word "included" from that definition (in the
proposed definition of UB, the wording was "... measurements specified included in...").
250
-------
8.6.2 Site-specific approach for near-field sources
Comment 1: One commenter stated that it is arbitrary and unreasonable to limit site-specific
monitoring plans to dealing with only off-site upwind sources or on-site sources excluded under
63.640(g). Since winds are constantly changing "upwind" has no meaning and there is no value
in including this criterion. Similarly, it is unreasonable to limit on-site concerns to only sources
excluded under 63.640(g). Many on-site sources are not addressed by 63.640(g) and they may
need to be addressed in this sampling program. For instance, on-site mobile sources (e.g.,
automobiles, trucks, and locomotives), on-site laboratories, on-site and off-site spills, and
operations subject to NESHAPs other than the HON (which is called out in 63.640(g)), such as
the Hazardous Waste Incineration NESHAP, the Boiler and Process Heater NESHAP, or any of
the multitude of Chemical NESHAPs. Therefore the commenter recommended that the language
"off-site upwind sources or on-site sources excluded under 63.640(g)" be revised to "any sources
not regulated under this subpart."
Further the commenter indicated that the language in 63.658(i)(l) refers to the term "Near-field
source", which is an undefined term and its use suggests some limitation on what interfering or
confounding sources can be addressed through this additional monitoring. There is no sound
technical or legal reason for excluding any non-Refinery MACT 1 emissions source that has the
potential to bias the measured fenceline benzene level from being addressed through a site-
specific monitoring plan and, therefore, the commenter recommended that the term "near-field"
be replaced with "interfering or confounding" in this paragraph and at every other occurrence in
the proposal. The commenter also stated that it may be impossible to identify specific sources.
Sources may be multiple, transient and/or vary with wind direction. Nor should it matter whether
the correction is being made to address a particular, identifiable source or to address an observed
confounding action level impact. Thus, paragraph (i) should be generalized to require only a
general description of the sources or emission types that are expected to be addressed by the
additional monitoring.
Response 1: We disagree that most refineries will have to develop site-specific monitoring
plans. We established the action level including all emission sources at the refinery facility,
which included co-located operations such as HON units. We also find that AC approach works
well based on our pilot studies and the API study and that most facilities will use this simple and
direct approach.
We previously outlined that we expect all sources at the refinery are subject to the Refinery
MACT fenceline standard unless explicitly excluded. We suspect that the commenter believes
"regulated under this subpart" would limit the requirements to Group 1 sources, but we repeat
that Refinery MACT 1 effectively includes "all" sources including Group 2 sources. We
recognize that Refinery MACT 1 also specifies exclusions in paragraphs (d) of section 63.640
but note that these sources are still refinery emission sources.
We are unsure how "interfering or confounding sources" is better and more clear than "near-field
sources." We use the term NFS in the equation for near-field sources and we consider that
63.658(i) effectively defines what is considered a near-field source, so no further clarification is
needed.
251
-------
We are retaining the requirement that the site-specific monitoring plan must identify the near-
field sources for which the monitoring plan seeks to quantify; otherwise, the monitoring plan
cannot be evaluated for quantitative corrections for specific monitoring sections. With respect to
the uniform background concentration, as stated in the rule, if an owner or operator uses the AC
approach, the lowest monitored fenceline concentration is used as a proxy for the UB
concentration. If an owner or operator elects to use a site-specific monitoring plan, they must
separately determine the UB concentration (i.e., benzene concentration far from any industrial
facility, highway, or other significant source of benzene emissions). As noted in 40 CFR
63.658(i)(l)(ii)(A), a facility owner or operator may elect to assume that UB=0 rather than site-
specific, remote, UB monitoring station. Facility owners or operators cannot elect to use the
lowest fenceline concentration as the UB concentration if they elect to use a site-specific
monitoring plan.
Comment 2: One commenter recommended that site-specific monitoring plans be valid from
submission and that sites must revert to their previous plan or submit within 30 days revisions if
the new plan is denied. The commenter objected to the proposed language that requires approval
by the Administrator, who has 90 days to approve because sites are forbidden to use the plans
until they are approved. If the need for this plan resulted from changes in the on-site or off-site
confounding facilities or because of an impact found through a root cause analysis (RCA), many
additional 2-week results could be high or otherwise questionable while the site-specific plan
was being developed, submitted, and reviewed.
Response 2: We have provided 2 years to begin official monitoring specifically to provide
adequate time for refinery owners or operators to assess the need for a site-specific monitoring
plan and, if needed, time to develop the plan, submit it for approval with enough time to allow
the EPA adequate time to review and comment on the plan. If we had not provided an additional
year specifically to allow time to develop and implement a site-specific plan, we would consider
allowing owners or operators to follow a site-specific monitoring plan prior to approval.
However, the site-specific monitoring plan may be developed and submitted before exceeding
the annual average AC value. While we will accept and review site-specific monitoring plans
after the action level has been exceeded, poor planning by the refinery owner or operator does
not mean facilities can modify their compliance approach (must use the high-low AC differential
versus a site-specific monitoring plan) until they receive approval for a site-specific monitoring
plan.
Comment 3: One commenter stated that section 63.658(f) should specifically indicate that the
requirements of Section 12 of Proposed Draft Method 325A are superseded and do not apply to
the refinery fenceline monitoring program. Section 12 of Draft Method 325A calls for
calculating averages of the fenceline data, rather than the AC value that is the basis for the action
level determination, and calls for other analyses of the data that are not specified by section
63.658, discussed or justified in the preamble or backup documents and appear not to be
included in the Information Collection Request Supporting Statement. Thus, Section 12 of the
Draft Method is superfluous for this purpose and should be clearly over-ridden by 63.658(f) to
avoid future claims that it applies.
252
-------
Response 3: Section 12 of Method 325A does not require any calculations whatsoever. For
example, Section 12.1 states that the average concentration "can be calculated for any specified
period at each PS location using Equation 12.1." If we had intended the use of Equation 12.1, we
would have had to specify the time period and we would have used "you must" or "you shall."
Sections 12.2 and 12.3 use "you may..." As such, while we agree Section 12 provides additional
suggestions on how one might use or analyze the data, we do not consider it possible that the
EPA or local agency could enforce any part of Section 12, particularly for facilities using the
high-low AC approach. This section is primarily provided as guidance for things that could be
considered in developing and implementing a site-specific monitoring plan and the calculations
that might be needed to determine near-field source contributions. We revised Section 12.1 of
Method 325A to say "you may" rather than "you can" and we revised the section title to be
"Optional Data Calculations, Analysis and Documentation". We have also specified in section
63.658(i) that the procedures in Section 12 are not required, but may be used, if applicable, when
determining near-field source contributions if a site-specific monitoring plan is used.
8.7 Action level
Comment 1: One commenter stated that the long-term corrective action threshold the EPA
proposed would allow for shorter-term spikes in HAP emissions that endanger the health and
welfare of refinery neighbors. The commenter requested that the EPA also set a short-term
corrective action threshold to protect communities, including the most-exposed person under
112(f)(2), from those spikes. The commenter cited the agency's enforcement division
requirements for fenceline monitoring programs to help identify sources of illegal fugitives in the
Shell, Deer Park, consent decree, where the EPA is requiring corrective action based on a five
minute standard and an hourly standard. Any five-minute period, where the fenceline monitor
picks up benzene concentrations above 50 ppb triggers a corrective action requirement.
Additionally, corrective action is required if the benzene level exceeds 15 ppb for three five
minute periods during a single hour. At Flint Hills Resources in Port Arthur, the EPA's consent
decree has corrective action requirements for fenceline readings of 1,3-butadiene or benzene that
average above 25 ppb for an hour. By comparison, under the proposed rule, a facility could have
a significant spill of benzene causing one two week sample to spike above 100 |ig/m3. Yet at all
but two refineries, that accident would not cause the facility to exceed the annual corrective
action requirement according to the EPA's modeling of maximum off-site ambient benzene
concentrations. This is based on the EPA's model showing that only two refineries are expected
to have fenceline concentrations above 4 |ig/m3. Assuming these facilities only measured
benzene levels at 4 |ig/m3 for 25 out the 26 monitoring episodes and 100 |ig/m3 for the final one,
the annual average would be 9.33 |ig/m3 and not subject to corrective action. Thus the EPA's
proposed average will ignore the health risks of short-term exposure to high pollution levels.
Therefore, to be consistent with its enforcement actions, reflecting the expertise of its
enforcement division, the EPA should set a short-term corrective action level to help identify and
reduce the significant health risks created by refinery malfunctions.
Another commenter disagreed, stating that the Proposed Rule should be revised to more
explicitly provide that short-term, high benzene readings are not actionable, and do not require
investigation or response.
253
-------
Response 1: We established the fenceline monitoring requirement pursuant to the technology
review provision, CAA section 112(d)(6), and not the risk review provision, CAA section
112(f)(2). While we recognize that there is some possibility for a short, larger release of HAP
that might occur and that will not be detected until the end of the 14-day passive monitoring
interval, we did not identify a cost-effective means for monitoring for the type of short, large
releases of concern to the commenters. Furthermore, we note that this approach is more feasible
on a source-by-source basis, such as was done in the individual consent decrees referred to by
the commenters. We could not identify a single monitored value that would apply for all sources
and that would serve the purpose advocated by the commenter.
Comment 2: One commenter stated that the environmental agency in Delaware has operated one
fenceline monitor near the Delaware City Refinery. The commenter explained that the unit has
not been very useful in providing real-time air quality information. Historically, releases from
the refinery have been more likely to be identified by citizen complaints. It is likely that a ring of
12 or so monitors would be more effective, if the action levels are set appropriately. The
commenter stated that the action level of 9 |ig/m3 of benzene (12-month rolling average) seems
to be arbitrary. The commenter also asserted that such a level is very high relative to what would
be expected during "compliant" operations and would not serve as an appropriate means for
detecting "significant increases in emissions." Therefore, the commenter recommended that an
action level be set at 125 percent of a "background" level measured or modeled for each source.
The commenter recognized that in some cases this could result in action levels being reached for
reasons, such as inversions, not directly associated with a source release or malfunction. But, the
objective should be to identify concerning concentrations of hazardous air pollutants.
One commenter stated that the EPA should require each facility to construct control charts from
their data and use a lower value (2 sigma and 3 sigma) to identify situations where the emissions
are moving toward an out of control or exceedance state. If a facility can demonstrate, through
control charting of their data, that its emissions are consistently below the action level target
concentrations and varies below two sigma then it might be appropriate to consider allowing it to
reduce its monitoring frequency, especially if it has a consistent pattern of compliance even
during startup and shut down operations.
One commenter stated that the EPA cannot promulgate a system that penalizes a source for
"violating" an inaccurate emissions inventory. Moreover, the commenter stated that the proposed
action level does not create a level playing field for regulated refineries. By selecting an action
level with which the EPA believes that all facilities can comply, the EPA has, in fact, created
significant disparity in the stringency of the requirement for different facilities. This is because
the level does not account for the facility size and shape, volume of production, or the number of
fugitive emissions components at any particular refinery. This means that small facilities likely
have a large "compliance" margin before any corrective action measures might be triggered,
while large facilities may have little to no compliance margin. The action level also does not
account for inherent differences in material processed from different geographical locations.
Moreover, the level does not consider future expansions or production increases, or changes in
the type or nature of raw material processed in the future.
254
-------
Response 2: The commenters have some intriguing ideas for establishing statistical based limits
and control charts for identifying "out of control" periods. This approach would work well
provided the baseline or initial emissions are well controlled. If emissions are high and
concentrations are at 9 [j,g/m3 or above during the initial monitoring period, this approach would
not cause these facilities to trigger a corrective action. While we see benefits from a composite
approach (baseline emissions approach but cannot exceed 9 (j,g/m3), we have only included a
trend analysis in the optional calculation and data analysis section (Section 12 of Method 325A).
However, we did not elect to mandate this approach because we are concerned with mandating
corrective action based on the trend analysis when the fenceline concentration is very low.
However, by including this approach in Method 325A, we hope to encourage refinery owners or
operators to use this or similar proactive analytics to use the fenceline monitoring data to
improve management of fugitive emissions.
We disagree with the commenter that the fenceline limit puts large refineries at a disadvantage.
Most large refineries have large property boundaries, so the "fenceline" is often much further
from the sources. We are not moved by other arguments presented regarding refinery expansions
or new crude slates. We based our risk decisions on the current emissions inventory, and with the
fixed fenceline monitoring standard, we have confidence that our risk decisions are based on
accurate information.
8.8 Non-compliance and Corrective Action Plan
Comment 1: Several commenters identified concerns with sources being able to identify the
source for purposes of corrective action. Among the concerns raised were the wind vector, (i.e.,
if the wind is blowing toward the fenceline) and locations near waterways or roads where vehicle
and marine traffic can contribute to fenceline levels. These commenters contended that elevated
readings should not be considered compliance deviations. They also recommended that while an
elevated level can trigger the obligation to perform an investigation, the investigation should be
sufficient even if a definitive root cause has not been identified. One commenter also
hypothesized that exceedances could be due to short-term episodic incidents or exceedances due
to unusual weather conditions or sample contamination and that working with an annual average
reduces the likelihood of such a situation, but does not eliminate the possibility. The commenter
claimed that an annual average makes it more difficult to identify the root cause because the
cause could be the result of several different unrelated events occurring over the year.
Commenters also stated that the regulations should be clear that RC/CA is triggered only if there
is an exceedance of the annual average benchmark value and not in response to individual
canister sampling events.
Response 1: Based on the action level we established, we see very little evidence that the roads
and waterways will significantly contribute to fenceline exceedances. In addition, while the
regulations do not require sources to perform RC/CA of short term high readings, we
nevertheless encourage refinery owners or operators to conduct investigations of unusually high
readings immediately. If a source of emissions is found and quickly remediated, then it is
unlikely that the annual average action level will be exceeded.
255
-------
We do recognize that an annual average that exceeds the action level could be due to more than
one event. However, we note that the facility will have the 14-day monitoring data from each of
the passive samplers and that information should assist the facility in identifying the root cause
or causes of any annual exceedance of the action level. Reviewing the 14-day sampling data as it
becomes available will aid this effort.
We disagree with the commenters that suggest that corrective action is not needed if the source
(or sources) of the elevated benzene readings cannot be identified. As provided above, the data
gathered through the monitoring should help the refinery owner or operator identify the cause or
causes. Furthermore, providing the relief requested would eliminate the incentive for facilities to
take the actions necessary to improve management of fugitive emission sources.
Comment 2: One commenter noted that 40 CFR 63.658(g) includes examples of data collections
that could be done as part of an RCA/CA in response to an action level exceedance. The
commenter believes that these are all highly expensive and time consuming alternatives and it is
likely any identifiable source will be identified more quickly and through other, less costly and
burdensome means. Furthermore, it is unlikely these examples could be completed in the 45 days
allowed and, since no burdens are included in the Information Collection Supporting Statement,
they are not authorized under the PRA. Thus, these examples serve no purpose and only add
confusion and they should be deleted. If the examples in paragraph (g) are not deleted as
recommended, the commenter suggested the following wording changes. First, the reference to
paragraph (g) seems to require that the four example corrective actions at the end of paragraph
(g) had to be included in the initial corrective action and have been completed. That phrase
should be reworded to refer to the initial corrective actions rather than to paragraph (g). Second,
the action level is an annual average and thus cannot be used to determine if the initial corrective
actions were successful. Thus, the trigger for follow-up corrective action should be a one or two
round action level evaluation and the appropriate trigger should be specified in place of the term
"action level" in its first use in section 63.658(h).
Response 2: First, we consider the examples of methods to investigate and correct high emitting
sources to be instructive and we are including paragraphs (1) through (3) in 40 CFR 63.658(g),
as proposed. We are revising paragraph (4) to focus on identifying on-site sources. If a facility
owner or operator wants to account for off-site sources, they must submit a request for a site-
specific monitoring plan and cannot attempt to adjust the concentrations as a corrective action.
Next, we agree that the reference to paragraphs (1) through (4) in 40 CFR 63.658(g) is poorly
worded and appears to require all of these examples as part of the RCA/CA. That was not our
intent. Therefore, in the final rule, we reworded the phase "... as described in paragraphs (1)
through (4) of this section" to "... such as those described...". Facility owners or operators have
significant latitude in determining appropriate RCA/CA and may use different methods than
those listed in order to identify sources that are causing the benzene level to exceed the action
level.
Finally, we agree that we intended facilities to evaluate the measured AC value for the next
sampling period to determine if the next 14-day average AC remained over the 9 [j,g/m3 action
level threshold. It would be difficult for the results of one 14-day sample to be reduced so low
256
-------
that the annual average AC value is immediately below 9 [j,g/m3. Therefore, we reworded the first
sentence in 40 CFR 63.658(h) to read as follows: "If, upon completion of the corrective actions
described in paragraph (g) of this section, the AC value for the next 14-day sampling period for
which the sampling start time begins after the completion of the corrective action is greater than
9 [j,g/m3 or if all corrective action measures identified require more than 45 days to implement,
the owner or operator shall develop a corrective action plan that describes the corrective action(s)
completed to date, additional measures that the owner or operator proposes to employ to reduce
fenceline concentrations below the action level, and a schedule for completion of these
measures."
Comment 3: One commenter recommended that rather than being open-ended, the RCA consist
of a review of the nearby fugitive emission sources that the fenceline monitoring program is
designed to enhance; if no deviations are found at those nearby sources, then no further action
should be required. The commenter also recommended that the site should note any nearby, off-
site sources that may be contributing to the threshold from a broad list of possible sources
provided by the EPA. The commenter stated that when there are no deviations found, then there
should be no requirements for further review until the following year.
One commenter stated that the root cause analysis under the proposed rule contains no specific
requirements, only suggestions for leak inspection or "visual inspection." The commenter
claimed that EPA should, at minimum, require sources to inspect for leaks and repair all leaks
found. The commenter asserted that a root cause analysis with no actual requirements is not
likely to produce meaningful corrective action. The commenter also stated that an exceedance of
the action level should clearly be deemed a violation of the emission standards, such that all
applicable CAA penalties will apply until the facility ends and corrects the problem. The
commenter claimed that anything less is a malfunction exemption, and thus is unlawful under the
Act and D.C. Circuit precedent.
Response 3: With respect to the comments that assert that the RCA should consist of a review of
the "nearby" fugitive emissions sources, we disagree. As an initial matter, it is unclear what the
commenter means by "nearby." Regardless, we do not agree that there should be any distance
limit placed on the sources that would need to be evaluated pursuant to the root cause analysis.
We disagree that corrective action should be limited to instances in which a fugitive emission
source is "violating" the underlying emission standard. In general, fugitive emission sources are
subject to work practice standards. One purpose of the fenceline monitoring program is to
identify instances where the work practice standards in place are not effectively managing
fugitive emissions. Thus, in that instance, a source may not technically be in violation of the
MACT, but the evidence indicates that it needs to adjust or modify the work practice standards
being implemented to provide better management of emissions. For sources subject to an LDAR
program, we note that is not a violation of the standard to have an emissions leak. Thus, under
the scenario suggested by the commenter, no corrective action would be required under the
fenceline monitoring program if the source identifies a leak that leads to elevated benzene levels.
While it is true that the source would eventually be required to repair the leak under the quarterly
or semi-annual repair requirement of the LDAR program, we disagree that fenceline monitoring
257
-------
program should be structured to allow sources to defer addressing a leak if it has identified it as
causing elevated levels of benzene at the fenceline.
Finally, we disagree with the suggestion that because there are no mandated steps for how to
perform a root cause analysis that the analysis will not produce results. We expect that the cause
of an elevated fenceline benzene level could vary significantly from facility-to-facility and even
from event-to-event. We believe that allowing the refinery owner or operator flexibility in
performing the root cause analysis will result in better identification of the source of the elevated
benzene level including, if so desired by the refinery owner or operator, initially targeting a
review of sources "nearby" the monitoring location with the highest benzene concentration.
Mandating a detailed list of actions that a refinery owner or operator must do, regardless of what
the emissions source may be, would be burdensome and could be ineffective. In addition, we
disagree that an exceedance of the benzene action level must be deemed a violation. The benzene
action level is not an emissions standard. The Refinery MACT 1 regulations establish the
applicable emission standards that must be met on a continuous basis for fugitive emission
sources. The fenceline monitoring program, including the benzene action level, is a tool for
enhancing the existing standards and is not a replacement for them.
Comment 4: One commenter suggested a number of changes to the provisions of 40 CFR
63.658(h) to clarify when corrective action has been successful.
First, because of method variability and the potential that an exceedance is due to a
short-term episodic emission or confounder, it is likely that in many cases the next
rolling average will no longer indicate an exceedance. The proposed 40 CFR
63.658(g) language would appear to still require the RCA/CAA and corrective action
(CA) in such cases. Language should be added to 40 CFR 63.658(g) removing the
RCA/CAA requirement if the rolling average returns to below the action level prior to
completing the RCA/CAA.
Second, up to 45 days is provided for completing a RCA/CAA in the event of an
action level exceedance. CA may take longer than that, depending on the cause.
During that time, at least two more rounds of samples will be completed and two
additional action level determinations made. If the cause of the original exceedance is
not yet resolved, each of these additional exceedances may also exceed the action
level. Unless there is data to indicate these exceedances are due to a separate cause
(e.g., occur in a different portion of the property), RCA/CAA analysis should be
deferred until the initial cause can be identified and only reinstated if the initial cause
is determined not to be the cause of the later occurrences. This will be apparent from
the results of the follow-up monitoring after the CA is complete.
Third, the commenter suggested that the basis for determining whether the corrective
action has been successful should be the average background corrected maximum
benzene concentration for two successive sampling rounds following completion of a
corrective action. Two rounds are desirable to provide an allowance for the
anticipated meteorological and analytical variability of these measurements. If this
average is below the action level, the corrective action should be considered
successful and any future annual average action level breech would be considered a
new occurrence and would not trigger the second corrective action requirements in 40
258
-------
CFR 63.658(h). Further, because an exceedance during the follow-up monitoring
could be due to a totally different cause than the original occurrence, the conclusion
that additional corrective action is needed should only be reached if the follow-up
maximum benzene value is in the same area of the property boundary as the original
maximum value and due to the same root cause.
Finally, because an unusually high two-week maximum AC value could cause the
rolling average to stay high for as long as a year, even though emissions have long
since returned to normal, the commenter requested that provisions are needed to
remove these high values from the average calculation in such situations and from the
follow-up sampling to demonstrate that CA has been successful, so any new
occurrences are not missed and so multiple deviations are not accrued because of the
same, one time incident.
Response 4: The action level is based on an annual average consisting of 26 14-day samples. A
previous commenter noted how one high reading would not greatly affect the annual average.
Therefore, an exceedance of the action level is unlikely to surprise anyone and refinery owners
or operators will have plenty of time to evaluate and complete corrective action as they approach
an exceedance of the annual AC action level. As we have stated before, we expect and encourage
refinery owners or operators to proactively use the monitoring data to identify and correct high
emitting sources in order to comply with the annual average AC limit (i.e., to prevent an
exceedance of the action level in the first place).
When the annual average AC exceeds the action level, we recommend a proactive approach,
such as immediate screening using an optical imaging camera. If no source is identified and the
subsequent 14-day concentration is back to normal without any corrective action, while a RCA is
still required, the source can simply report these results as part of their RCA. If an emissions
source is identified, and the issue can be corrected before it affects too many 14-day sampling
period results, the owner or operator would also have records of performing the corrective action.
We do not agree the RCA/CAA should be deferred or that it is necessary to have two 14-day
periods for the evaluation of the effectiveness of the corrective action, but we agree that the
sampling period by which to judge the effectiveness of the corrective action must be completely
after the corrective action is completed. Therefore, we revised 40 CFR 63.658(h) to clarify that
the test to determine if the corrective action was effective would be based on "the next 14-day
sampling period for which the sampling start time begins after the completion of the corrective
action." If the next sampling period AC exceeds 9 [j,g/m3 at a different monitoring location
(however unlikely that is), we consider that immediate corrective action should be pursued to
correct this "new" emissions source. If the new emissions source is identified and corrected, this
can be documented in the corrective action plan with the statement that subsequent corrective
actions appear to have been effective and that no further corrective action is needed. The owner
or operator must retain records of these corrective actions whenever the action level is exceeded,
so documenting these in the corrective action plan is not an excessive burden.
While Method 325A includes provisions to exclude outliers and sample results with documented
issues (e.g. missing or loose storage caps on tubes received for analysis), we have not included
provisions to exclude data caused by known emissions events at the refinery. We recognize that a
259
-------
given 14-day sample (after the first year of monitoring) will remain in the average for one year.
We also recognize that during the first year of monitoring, it is possible to have one high
monitoring reading for an emissions event that could cause the annual average to be exceeded
and to have corrected this issue by the time the first complete annual average could be
calculated. Although the company would need to retain records of the corrective action
performed, no additional corrective action is required at the time the annual action level is
officially exceeded (i.e., once the first full year of sampling is achieved).
Comment 5: One commenter stated that the EPA is allowed 90 days to approve CA plans, while
facilities are only allowed 60 days to develop those plans. If the approval requirement is
maintained facilities should also be allowed at least 90 days to develop their CA plan.
Developing the plan is certainly more time consuming than reviewing it. One commenter
suggested that the EPA should either extend the period for submitting a plan to 60 days or more,
or allow owners and operators to submit a plan within 30 days after receiving lab results of an
exceedance occurring during the next two-week sampling episode following the completion of
the initial round of corrective action.
Response 5: First, in most cases, we expect that the corrective action analysis will suggest that
more extensive corrective action measures may be needed and that the facility already has over
90 days in most cases, including the initial time period provided to perform the corrective action
analysis to start preparing the corrective action plan if the initial corrective actions, if any are
identified, are unsuccessful in lowering the fenceline AC values. Second, the EPA did not
necessarily intend to use the full 90 days for plan review and approval. We primarily included
this time frame as a means to ensure that corrective action plan does not remain unexamined and
we provided default approval of the plan if the Administrator fails to act (i.e., disapprove the
plan in writing). We now consider it important for refinery owners or operators to submit their
plans and begin corrective action as soon as possible, without having to wait for EPA approval,
and therefore we are not finalizing the requirement that EPA must approve the plan. We believe
that 40 CFR 63.658(h) already provides that the 60-day period for developing and submitting a
corrective action plan starts after receiving the results from the subsequent 14-day sampling
period. However, we are revising the second sentence in 40 CFR 63.658(h) to further clarify this
point to read as follows: "The owner or operator shall submit the corrective action plan to the
Administrator within 60 days after receiving the analytical results indicating that the AC value
for the 14-day sampling period for which the sampling start time begins after the completion of
the corrective actions is greater than 9 [j,g/m3...".
Comment 6: One commenter stated that the relatively short period of 35 days for completion of
RCA does not allow a reasonably sufficient time to complete all of these steps for multiple
sampling locations at refinery facilities. The commenter asserted that EPA must carefully
consider the time requirements of the actions necessary to properly conduct the analysis and
evaluation of results of its proposed fenceline benzene monitoring and allow a reasonably
sufficient time for regulated facilities to perform those actions prior to requiring subsequent
actions and responses. The commenter stated that at the end of sampling events the passive
benzene monitors would need to be collected, transported to a laboratory, undergo laboratory
analysis, quality assurance and reporting of results. The reported laboratory results would then
need to be incorporated into a facility's 12-month rolling average benzene concentration
260
-------
calculations, which is a task that would likely require a review of meteorological data collected
during the sampling event to attempt to account for any background contribution.
Another commenter provided a table summarizing their recommended corrective action
timeframes and requirements for events and stated that the EPA should strengthen these
requirements in the following ways:
- Initial Corrective Action: The EPA must require facilities to complete the corrective action
within 5 days of initiating the root cause analysis.
- Further Corrective Action: A facility should have no longer than 14 days to develop a new
corrective action plan and begin to implement it. A facility also should not wait to implement
that plan until receiving EPA approval.
- Specific Reporting and Action: The rules should require immediate reporting and specific
corrective action, such as automatic shutdown and additional higher-quality monitoring (such as
UV-DOAS), with oversight such as an inspection and audit by an EPA expert staff or an
independent expert, until the problem has been fully resolved to prevent its repetition.
One commenter recommended that the final rule require refiners to submit the data at the end of
each 2-week period that the passive sample tubes are replaced and not allow for 6-month or year-
long intervals to go by without actionable data being relayed to state and federal enforcement
divisions. The frequency of data availability to regulators will help ensure regulators can better
correlate exposure data regarding increased toxic air emissions to health outcomes in response to
these emissions. One commenter believes it is important that sources be required to act if their
fenceline monitoring shows benzene concentrations above a certain level, and support the EPA
in requiring a corrective action plan. However, the commenter is concerned about the vagueness
in the proposal regarding what must be contained in a corrective action plan, the possible delays
in taking corrective action that could result, and the lack of detail about the enforcement actions
that will follow an exceedance. The commenter suggested that those provisions be strengthened
and made more specific. Additionally, with respect to the proposal to use a one-year rolling
average, the commenter recommended that the EPA call upon sources to identify problems as
they are developing and take action before exceeding the action levels whenever possible
Response 6: We disagree with the commenters that suggest up to 90 days are needed to verify
and validate the fenceline monitoring data before the action level can be determined. For
facilities using the high-low AC approach, there is no further adjustment allowed based on
meteorological data. If a facility has a site-specific monitoring plan, the calculation used to
determine the monitoring location specific adjustments are to be included in the monitoring plan
and only the calculation methods included in the approved plan may be used. Therefore, the
calculations can be either automated using a computer or performed by hand with minimal
personnel hours.
We added an additional 15 days to the 35 day period provided in 40 CFR 63.568(g) for a
maximum of 50 days after the end of the sampling period to initiate corrective action. We
consider 45 days sufficient to complete the analyses and to incorporate the results of the 14-day
261
-------
sampling period into the annual average AC value and an additional 5 days to initiate corrective
action.
We provide up to 45 days from the date the action level was exceeded to complete the root cause
and corrective action analysis. For corrective actions that require more time to implement, for
example, if the corrective action analysis suggests the most effective corrective action would be
to retrofit an external floating roof tank with a geodesic dome cover (i.e., converting the external
floating roof tank into an internal floating roof tank), this action will require more than 45 days
to complete. In this case, we considered refinery owners or operators would prepare and submit a
corrective action plan. We disagree with commenters that suggest much shorter corrective action
completion time periods are needed or appropriate. While some large leaks may be readily
identifiable using optical imaging cameras and readily repaired, we consider it to be unlikely that
a refinery owner or operator would be able to routinely complete a root-cause analysis within 5
days, let alone complete the corrective actions, particularly if the repair requires more than
tightening a bolt or replacing a cap.
We have previously addressed comments suggesting UV-DOAS systems should be required and
again note that these systems currently are not capable of detecting the concentrations needed to
assess and manage refinery fugitive emissions. We do agree that facilities should act proactively
to prevent an exceedance of the annual average action level and we anticipate most refineries
will do so. While we do not mandate this early action, we retain the proposed time tables for
corrective actions in part to encourage this proactive compliance approach.
Comment 7: Several commenters stated that it is extremely inappropriate to establish a standard
time frame for compliance with actions listed in a corrective action plan. A corrective action plan
is specific to the event, the refinery, and the root cause of the event. Time required to complete
actions is dependent on the complexity of the action; permitting and permit approval; whether
equipment can be taken out of service immediately or only during a turnaround, availability of
repair equipment, materials and skilled labor to complete the action, etc. Thus, no one-size-fits-
all time frame can be established. Even the consent decrees applicable to refineries do not
include such a time limit.
Response 7: Actually, we only specify the time to complete the corrective action analysis. In the
proposed rule we did not specify a specific time frame by which to complete the initial corrective
actions; although, we intended that initial corrective actions must be completed within 45 days.
We do consider that there may be instances that the corrective actions themselves may take
months to complete, depending on the emissions source and the scope of a project. If the initial
corrective action measures do not remedy the high fenceline concentrations, we proposed
provisions to complete a corrective action plan. We are revising the second sentence in 40 CFR
63.658(g) to clarify that initial corrective action measures are also to be completed within the 45-
day time period. Specifically, "The root cause analysis and corrective action analysis shall be
completed and initial corrective actions taken no later than 45 days after determining there is
an exceedance." We are clarifying 40 CFR 63.658(h) to account for cases where the owner or
operator could not identify any initial corrective action measures that could be completed within
45 days. "If, upon completion of the corrective action analysis and corrective actions described
in paragraph (g) of this section, the AC value for the next 14-day sampling period for which the
262
-------
sampling start time begins after the completion of the corrective actions is greater than 9 |ig/m3
or if all corrective action measures identified require more than 45 days to implement, the
owner or operator shall develop a corrective action plan that describes the corrective action(s)
completed to date, additional measures that the owner or operator proposes to employ to reduce
fenceline concentrations below the action level, and a schedule for completion of these measures.
The owner or operator shall submit the corrective action plan to the Administrator within 60
days after receiving the analytical results indicating that the AC value for the 14-day
sampling period following the completion of the initial corrective action is greater than 9
jig/m3 or, if no initial corrective actions were identified, no later than 60 days following the
completion of the corrective action analysis required in paragraph (g) of this section..."
Comment 8: One commenter objected to proposal language that would not delegate authority
for the approval of the corrective action plan under 40 CFR 63.658(h) to State, Local, or Tribal
agencies or that would automatically approve the plans if the EPA failed to approve them within
90 days.
Several commenters believe while some state and local agencies may not wish to take delegation
of the approval of corrective action plans, they should be provided this option, while other
commenters encouraged the EPA not to delegate the responsibility to review and approve CA
plans to the states.
Another commenter stated that EPA also must require corrective action plans to be submitted
for notice and comment. That can occur at the same time as the EPA is reviewing the plans,
without causing delay. Sources should be required to begin implementing corrective action
while receiving input from the public and the EPA, and the EPA must then decide whether to
add or modify corrective action requirements after considering public comment.
Response 8: We appreciate the varied comments on this issue. Upon further review of these
requirements, we have finalized the provisions not to require EPA approval of the corrective
action plans as proposed to avoid delays in the implementation of corrective actions. We have
also elected not to delegate the approval of the site-specific monitoring plans in 40 CFR
63.658(i).
8.9 Recordkeeping and reporting requirements
Comment 1: Commenters noted that the EPA did not propose how they will take the data
gathered at the fenceline and transform it into information that is useful for communities
concerned about risks. The commenter recommended that the EPA should provide, on every
page of the data, an explanation of the uncertainties and issues associated with this information.
The commenter also stated that the EPA should explain on the website with the reported data that
the action level is not based on a health risk to the public and that any individual data points
above the action level are not an indication that the facility is out of compliance with any
applicable requirement or permit. One commenter also recommended that the EPA provide
perspective by posting on the website comparative values from the EPA and World Health
Organization studies, such as the fact that indoor air levels of benzene in a home with a smoker
263
-------
(10.5+ |ig/m3) and outdoor concentrations of benzene from house fires, nearby agricultural
burning and from wood burning fireplaces (11 |ig/m3 or more).
Two commenters stated that the EPA is aware that any monitoring data can be subject to
interpretation, and there is significant interest in communities near refineries in evaluating
benzene concentrations in their communities and neighborhoods. Another commenter urged the
EPA to make the information available in a form that is easy for the public to access and
understand.
Response 1: We appreciate the comments and understand the concerns related to appropriate
presentation of the data. We intend to provide information on the website that explains what the
action level is and that it is evaluated on an annual average.
Comment 2: Commenters who advocated that EPA require open-path monitors instead of
passive monitors also suggested that the data from the "real-time" open-path monitors could
provide immediate feedback to the community with one commenter noting that there is a
mechanism for sending this data wirelessly from the stations using 3G or 4G.
Two commenters stated that the EPA should require each facility to post data collected from the
use of passive samplers on a continuous basis so that it is available as promptly as possible to
fenceline communities. Other commenters suggested that it would be easy and inexpensive for
refiners to make available all monitoring data, including but not limited the daily maximum
readings from each fenceline monitor, CEMS and continuous opacity monitoring system
(COMS) data, and other important emissions information, including unauthorized release
reports, meteorological data, rain fall totals and wind directions. Another commenter stated that
reporting "through EPA's electronic reporting and data retrieval portal" is not sufficient; the
EPA must provide a public website that makes these data accessible to community members.
One commenter suggested that another way to provide the public with real-time, transparent
information is to leverage social media. Another commenter suggested mechanisms such as
"robocalls" to provide real-time notice of exceedances of the action level.
Response 2: For the reasons provided in the preamble and earlier in this document, we are not
requiring use of real-time monitors. We disagree that information from passive monitors should
be made available "immediately." As discussed previously, refinery owners or operators will
need time to analyze samples, review the data, and assess QA/QC sample results. We do not
think the public interest will be served by making data available before these steps can be taken
and before it can be made available in a manner that is easily understandable by the general
public, as requested by one commenter. We believe that the website will be able to provide
certainty regarding the information and in a manner that the public can understand. After
considering the strong interest from commenters about the availability of data, we are finalizing
a requirement that monitoring data be submitted on a quarterly basis, rather than on a 6-month
semiannual basis, as proposed.
Comment 3: One commenter suggested that proposed 40 CFR 63.655(h)(8) be modified to
clarify that the fenceline monitoring semiannual reporting period is the same as the semiannual
periodic report reporting period. The commenter stated that the proposal calls for reporting
264
-------
results "after the end of each semiannual reporting period." However, 40 CFR 63.655(h) covers
various notice and report requirements, so it is unclear what "semiannual reporting period" is
referenced.
Response 3: We intended that the semi-annual reporting period is the period covered by the
periodic report. In response to other comments, we are requiring that data be reported on a
quarterly basis. We revised 40 CFR 63.655(h)(8) to clarify "For fenceline monitoring systems
subject to §63.658, within 45 calendar days after the end of each quarterly reporting period
covered by the periodic report.
Comment 4: One commenter stated that proposed 40 CFR 63.658(d)(1) requires hourly average
records for meteorological data, including wind speed, wind direction and temperature and
proposed 40 CFR 63.655(i)(8)(iii) requires a record of daily unit vector wind direction,
calculated daily sigma theta, daily average temperature and daily average barometric pressure
measurements. However, the commenter stated that neither daily unit vector wind direction or
daily sigma theta are defined or explained, so it is unclear what is required. Nor does the rule use
any of this information and it is therefore unclear how this burden is justified. Another
commenter recommended that the EPA explain how the burdens associated with these
recordkeeping requirements are justified, rather than just generating the information if needed for
source apportionment or that these recordkeeping requirements be deleted. If they are
maintained, the terms "daily unit vector wind direction" and "daily sigma theta" should be
defined and their units specified.
Response 4: We agree that the recordkeeping requirement in 40 CFR 63.655(i)(8)(iii) should be
consistent with the requirements outlined in 40 CFR 63.658(d). We also agree that if data is not
needed to demonstrate compliance with the standards, then we should not require it. The only
meteorological data required for refinery owners or operators opting to comply using the direct
AC approach is the average temperature and barometric pressure over the sampling period.
Therefore, in the final rule, we are requiring that only these data be collected in this
circumstance. For near-field source corrections or if alternative test methods using time resolved
measurements are used, then the final rule requires collection of hourly average meteorological
data, including temperature, barometric pressure, wind speed and wind direction and calculation
of daily unit vector wind direction and daily sigma theta. These data provide better information
regarding the origin of the air mass moving over the refinery for given time periods and we find
that these data are necessary to conduct the more robust monitoring approaches refinery owners
or operators may likely implement under site-specific monitoring or alternative test methods.
Comment 5: One commenter stated that proposed 655(i)(8)(v) requires, for samples that will be
adjusted for background, a record of the location of and the concentration measured
simultaneously by the background sampler, and the perimeter samplers to which it applies. The
word "simultaneously" could be interpreted to mean that all samplers must start and stop at the
same time; a physical impossibility and not a significant issue for two week samples that provide
concentration results. Thus, this wording should be changed to call for the background
concentration to be measured over "approximately" the same two week sampling period as the
samples it is used to correct.
265
-------
Response 5: We find that the two monitors with similar start times and end times will be
measuring the same time period "simultaneously." The language suggested by the commenter
could be interpreted to suggest these samples do not have to have the same start and end days,
which is not acceptable for the purpose of a background correction. We are retaining the
proposed language in 655(i)(8)(v).
8.10 Cost of Fenceline Monitoring
Comment 1: One commenter pointed out that the cost of open-path monitoring is insignificant
when weighed against the profits made by the average refining company or that the cost of such
monitoring can be easily recouped with the shoring up of fugitive emissions (representing lost
product) that is possible with open-path monitoring. The commenter stated that the EPA's cost
benefit analyses included in the proposed rule determined that open path UV monitors like those
used at the Chevron refinery in Richmond, California are too expensive, as are other types of gas
chromatographs and remote sensing technology and requested that the final rule consider an
updated cost-benefit analysis on monitoring technologies. Another commenter stated that the
EPA has inflated the costs of real-time systems by relying on costs of the most expensive UV
systems and underestimating the labor intensive costs of passive samplers. Several commenters
provided details about the costs of different systems. One commenter was concerned that in
rejecting real-time monitors, EPA was placing more importance on the bottom line of industry as
opposed to the health of communities.
Response 1: While we believe that open-path systems are more costly than the passive monitors
required in the final rule, our primary concern with open-path monitoring systems is that they are
not capable of detecting and quantifying benzene at concentrations at or below AC action level.
Comment 2: Numerous commenters claimed that EPA underestimated the cost of the fenceline
monitoring program in the proposed rule by underestimating the number of samplers required,
unrealistically assuming the analyses are done in-house, not including any costs or burdens for
the required root cause and corrective action analyses and underestimating the costs of
implementing a site-specific monitoring protocol.
One commenter stated that based on the number of sample locations identified in the EPA's and
the commenter's pilot studies and the significant number of additional samplers required by the
revised requirements for subgrouping in the proposed Draft EPA Method 325A, the number of
samplers used as the basis for the EPA's estimates is significantly understated. The commenter
asserted that several refineries will likely have to have multiples of the minimum number of
samplers, because they have disjointed property with the minimum number of samplers on each
parcel. In addition, a commenter claimed that 63.658(c)(3) requires one duplicate sample for
every ten samples and 2 field blanks per sampling round. A few commenters also claimed that
many sites will also add samples to develop site-specific background correction data and in order
to demonstrate that higher emissions within the refinery come from sources not subject to this
rule. Thus, 40 samplers per site and 48 samples every 14 days for a large site is a much more
reasonable estimate than the EPA's assumption of 24 samplers and samples per site. One
commenter claimed EPA's assumption that small entities would need only 12 monitors was
incorrect.
266
-------
Commenters stated that analysis costs will be at least double the EPA estimates because most
analyses will be done by outside laboratories in order to meet the QA/QC requirements specified
in Draft Method 325B and to minimize inter-laboratory variability. One commenter focused on
costs for small entities, claiming it was unreasonable for EPA to assume that small entities can
hire the manpower and build the facilities necessary to conduct in-house testing.
Another commenter claimed EPA underestimated costs by not realistically assessing the process
for establishing a site-specific monitoring protocol. The commenter asserted that such a process
will be needed by many facilities because of their location near other industrial sources and
transport corridors.
Finally, one commenter stated that EPA did not consider the costs that will be incurred for
installing and maintaining the required weather station as well as for the fencing, lighting, and
other infrastructure associated with locating these samplers at the property boundary.
Response 2: We agree with the commenters that stated that we need to account for the fact that
facilities may need to site additional monitors and will have to analyze a few additional QA
samples. However, we disagree that a typical refinery will have twice the number of monitoring
stations as the number we estimated. The commenters noted that the siting guidance contains
requirements to place additional monitors at locations where known emissions sources are near
the fenceline and/or for irregularly-shaped sites where additional monitors are required based on
the shape of the refinery's property. We agree with these comments. As the rule requires
additional monitors be sited under certain circumstances, the projected costs of the monitoring
program should reflect these required costs. We reviewed the number of monitoring locations
used in the API pilot studies as reported in Table A-l of the pilot study report (API, 2014). On
average, 6 additional monitoring locations were used for the small facilities and 8 additional
monitoring locations were used for the medium-sized facilities. Only one large facility (greater
than 1,500 acres) was included in the pilot study and this facility did not require any additional
monitoring locations. However, as large facilities may also be non-uniform in shape, we assumed
that large facilities would require, on average, 8 additional samples (similar to the medium-sized
facilities). Regarding the allegation that more monitors will be needed because facilities have
segregated property, the commenter did not identify that there were a significant number of
facilities for which this is the case and we do not have any evidence to suggest this is true.
However, we note that the final rule places a limit on the number of extra monitoring locations
needed for segregated areas, so we do not believe our costs estimates need to be revised further
in response to the concern about segregated areas.
As we see no evidence that the basic AC approach will not work well for any given refinery, we
continue to assume that refineries will use this method for compliance.
Regarding the cost concern for a weather station, the final rule does not include a requirement for
an on-site meteorological station but allows for use of a USWS within 40 kilometers (25 miles)
of the facility.
We disagree with the commenters' assertion that we must assume refinery owners or operators
will outsource the sample analyses. The analysis requires use of gas chromatography (GC) with
267
-------
mass spectrometry (MS), flame ionization detectors (FID), photoionization detectors (PID)
analysis, or other similar detection systems. As GC/FID and GC/PID systems are commonly
used by refineries for product quality analysis, and we expect all refineries to have qualified
personnel at the facility familiar with this equipment. Additionally, we included costs for
automated sample feed systems, which greatly minimizes the level of manual effort needed to
analyze a batch of diffusive tube samples. Therefore, we continue to project the costs of the
program based on in-house analysis. A refinery owner or operator may elect to outsource the
analysis, if they so choose, but as the rule does not require out-sourcing of the analysis, we do
not consider it appropriate or necessary to over-estimate the analytical costs assuming all
refinery owners or operators will outsource the tube analyses.
Comment 3: One commenter believes the EPA underestimated the cost of the proposed on-site
meteorological monitoring requirements, which include calibration and standardization
procedures incorporated by reference from the EPA's "Quality Assurance Handbook for Air
Pollution Measurement Systems Volume IV: Meteorological Measurements" (EPA-454/B-08-
002 March 2008). These requirements are similar to stringent and burdensome requirements for
meteorological data collected for air quality modeling and ambient monitoring programs, which
also require semi-annual audits of the equipment minimum instrumentation accuracy criteria.
The commenter is not currently required to conduct on-site meteorological monitoring.
Furthermore, most "off-the-shelf' meteorological monitoring systems that may currently be in
use at other refineries may not meet these proposed required criteria. In its cost analysis the EPA
did not consider the cost to regulated facilities for adding or upgrading existing on-site
meteorological monitoring systems that will meet the proposed requirements, or the ongoing
costs to calibrate, maintain, and operate those systems in compliance with the proposed
requirements. The operation of a meteorological monitoring system adds no value to correlating
short term (i.e., 14- day) ambient benzene data to refinery operations and therefore the
imposition of costs related to installing or upgrading and operating a meteorological monitoring
system is unreasonable.
Response 3: We included costs for monitoring stations for all refineries and although we did not
expressly include costs for semi-annual calibration of these systems, we included costs for
maintaining the equipment (10 percent of capital costs). In our revised cost estimate, we
considered adding additional costs specifically for calibrating these systems. However, in
response to another similar comment, we agreed that having an on-site meteorological station is
not critical for refinery owners or operators using the basic AC approach if representative
meteorological data are available from a USWS within 25 miles of the refinery. This provision is
expected to significantly reduce the number of meteorological stations that would need to be
installed, more than offsetting the added costs of semi-annual calibration. Therefore, we elected
not to revise the cost assumptions for meteorological monitoring stations and we consider the
costs assumptions used (i.e., every facility installing a meteorological station with annual costs
for maintenance of the equipment of 10 percent of capital costs) to be a conservatively high
estimate of the actual nationwide monitoring costs that will be incurred as a result of the final
rule.
268
-------
8.11 Other
Comment 1: A commenter raised a concern that one lab in the API study consistently reported
higher benzene uptake and therefore, would have consistently calculated higher benzene
concentrations had it been used for the study samples. The commenter noted that an investigation
during the course of this work did not identify a reason for this difference. The commenter asked
the pilot study contractor, to conduct a laboratory proficiency test (PT) on the two laboratories
used in the API/AFPM pilot fenceline monitoring study in order to confirm whether a systematic
bias is occurring. The commenter is concerned that systemic bias is an issue and will affect the
142 refineries that are sending samples to different laboratories. The commenter provided
detailed information about the different test results and analyses.
Response 1: We reviewed the information submitted by the commenter and asked for the
detailed laboratory analyses described by the commenter. In the course of this follow up, the
commenter indicated that they now realize that the primary laboratory did not follow the
procedure in proposed paragraph 9.13.1 of Method 325B that requires continuous calibration
verification through the development of response factors determined from the daily calibration
standard and provided a revised set of results that correctly calculated the benzene concentrations
in accordance with proposed Method 325B. The corrected concentrations resulted in an increase
in the average AC for all 12 refineries increasing from 4% to over 14%. Based on these results,
the inter-laboratory bias described by the commenter appears to be resolved and the overall
variability in the results found to be significantly lower.
269
-------
9.0 Other Refinery MACT 1 Provisions
Comment 1: One commenter stated that the term "regulated pollutant" or "regulated material"
in the proposal is used in subparts that contain generic requirements that are widely referenced.
The commenter believes its use in these subparts is ambiguous, since it can easily be interpreted
to include substances that are not regulated by these subparts and/or are not regulated by the
particular paragraph using the term. The commenter recommended the term be replaced in every
use with a Refinery MACT 1 or Refinery MACT 2 specific term, such as organic HAP, metal
HAP, nickel, PM, etc. Where a generic term is needed such as in certain recordkeeping and
reporting requirements the commenter recommended "HAP regulated by the applicable standard
in this subpart." Another commenter specifically cited proposed 40 CFR 63.655(g)(12)
(Reporting) and (i)(l 1) (Recordkeeping) for malfunctions resulting from the removal of the
historical SSM approach. The commenter stated that the term "regulated pollutant" should be
replaced with a more precise term.
Response 1: We disagree that these terms should be replaced. In the "Applicability and
designation of affected source" provision of section 63.640(a)(2), it states that subject sources
are those that, "Emit or have equipment containing or contacting one or more of the hazardous
air pollutants listed in table 1 of this subpart." We believe that this provisions makes clear that
the regulated pollutants are those HAP listed in table 1 of the subpart. Similarly, in Refinery
MACT 2, the provision stating the purpose of the subpart in section 63.1560 "establishes
national emission standards for hazardous air pollutants (HAP) emitted from petroleum
refineries." HAP is a term defined by the general provisions of part 63 in section 63.2.
The term "regulated material" is used primarily in reference to the proposed flare control device
requirements. While the flare control requirements are included in Refinery MACT 1, we also
proposed that flares used to control Refinery MACT 2 streams would need to comply with the
flare provisions in Refinery MACT 1. Thus, the term "regulated material" as used in the
requirements for flare control devices is intentionally general to accommodate the process
streams for which both Refinery MACT subparts require control. To improve the clarity of the
term "regulated material" we have included a definition of regulated material in subpart CC.
Specifically, "Regulated material means any stream associated with emission sources listed in
§63.640(c) required to meet control requirements under this subpart as well as any stream for
which this subpart or a cross-referencing subpart specifies that the requirements for flare control
devices in §63.670 must be met."
Comment 2: One commenter stated that the proposed recordkeeping and reporting requirements
in proposed 40 CFR 63.655(g)(12) (Reporting) and (i)(l 1) (Recordkeeping) for malfunctions are
unclear and should be clarified. The commenter explained that they do not understand what
"number of failures" refers to in these paragraphs, since the failure to meet an applicable
standard is by its nature a single occurrence. If this is referring to the number of failures of a
particular emission limitation over some time period, the time period should be indicated.
Furthermore, the record of the time, date, and duration of each failure provides that information
and would seem to provide a count of such occurrences. Thus, we believe the "number of
failures" should be deleted from this recordkeeping requirement.
270
-------
The commenter added that it is unclear how to apply these requirements to work practice
deviations because in those cases there are likely no excess emissions to estimate and no
information as to when the failure started, only when it was identified. The commenter provides
the example of a failure to perform a required inspection which would have no emission impact
and which may only be discovered after the time the inspection was due and the inspection
requirement may have been to perform the inspection within some calendar period (e.g., an
annual inspection). The commenter asserted that the proposed requirements should only be
applicable to failures involving excess emissions and, for other failures, the requirement should
be to report the failure to follow the work practice and identification of the time period when the
work practice action should have been performed.
Response 2: We have thoroughly reviewed the existing and proposed recordkeeping and
reporting requirements in subpart CC (including those which refer to other subparts i.e., FF, R,
Y, VV, and H). During the review, we noted that reporting requirements need to be added for
delayed coking units, and the requirement for reporting from bypass lines in 40 CFR
63.655(g)(6)(iii) needs clarification. With respect to delayed coking units, we are requiring
owners and operators to report any 60-cycle average (for existing sources) or any direct venting
event (for new sources) where the pressure of the DCU exceeds the applicable limit. We are also
including a requirements to report the total number of double quenching events during the
reporting period and each instance (drum, date, time and maximum temperature) of double-
quench draining activity when the drain water temperature exceeds 210 °F. The reporting
provision for bypass lines contained in section 63.655(g)(6)(iii), has been updated to now more
clearly identify the required reporting elements. The provision in the final rule provides: "(in)
For periods in closed vent systems when the vent stream flow was detected in the bypass line or
diverted from the control device, or a bypass of the system was indicated, report the date, time,
duration, estimate of the volume of gas, the concentration of organic HAP in the gas and the
resulting emissions of organic HAP that bypassed the control device. For periods when the flow
indicator is not operating, report the date, time, and duration." We have determined that proposed
40 CFR 63.655(g)(12) and (i)(l 1) are not needed because the recordkeeping and reporting
requirements for the individual emission sources included in 40 CFR 63.655(g) and (i)
effectively cover the "failures" expected to be reported and thus we are not included the
proposed overarching "failure" recordkeeping and reporting requirements in the final rule. The
recordkeeping and reporting provisions we are finalizing contain the necessary information for
the EPA to assess compliance with the applicable standards for each regulated source category in
Refinery MACT 1.
9.1 Applicability and Affected Sources
9.1.1 Applicability dates and Table 11 (other than for PRD)
Comment 1: One commenter stated that some required records are difficult to digitize or were
created before digital records became prevalent and may not be readily available within 24
hours. The commenter requested the 24 hour availability requirement be limited to 2 year old
records and up to 72 hours provided for records older than 2 years or that the current provisions
in 40 CFR 63.10(b) be maintained and applied to both Refinery MACT rules. The commenter
added that this would also provide better consistency with 40 CFR 63.1576(i) which references
271
-------
40 CFR 63.10(b)(1), which requires records to be maintained onsite for 2 years and then allows
them to be maintained offsite.
Response 1: EPA disagrees that additional time is needed to retrieve records older than 2 years.
Digital record storage including the ability to scan hard copy documents is standard operating
procedure and has been for arguably the last decade. Further, the commenter has not provided a
reason why a record (even a non-digital record) would take more than 24 hours to retrieve and
we cannot identify a reason.
9.1.2 Overlap provisions
Comment 1: One commenter noted a concern that subpart CC overlaps with the Marine Vessel
and Gasoline Distribution NESHAPs because proposed sections 40 CFR 63.650(d) and 40 CFR
63.651(e) would impose new flare requirements on flares dedicated to gasoline distribution and
marine facilities. The commenter stated that the new combustion control flare requirements
cannot be applied to dedicated part 63 subparts R and Y flares (flares that do not also receive
refinery gas) even though those facilities are part of the Refinery MACT 1 affected source
because those flares are not similar to refinery flares in their use and design.
The commenter also noted that these flare types were not considered in developing the proposed
combustion control requirements, and the EPA's risk and technology reviews of subparts R and
Y concluded there was no technology improvement that should be applied to such facilities
under section 112(d)(6). The commenter stated there is no reasonable basis for overturning that
conclusion and stated that the proposed 40 CFR 63.670 flare requirements should not apply to
flares that are dedicated to subpart R and/or Y operations and these two proposed new
paragraphs should not be finalized.
Response 1: We disagree that these sources should not be subject to the proposed flare
provisions in §63.670, because these units, when co-located at a refinery, are part of the affected
source. With this rulemaking, the EPA is posing new requirements for petroleum refinery
sources subject to part 63 subpart CC and UUU, including flares used to control emissions from
gasoline loading and marine vessel loading operations at petroleum refineries, which, as the
commenter noted, have been included as part of the "affected source" for Refinery MACT 1
since that standard was promulgated. Simply because we chose to harmonize the subpart CC
requirements with those in subparts R and Y does not alter the fact that these units, when co-
located at a petroleum refinery, are subject to subpart CC and are only subject to the control
requirements in subparts R and Y indirectly through requirements in subpart CC and the
emissions from these sources at a petroleum refinery are regulated refinery gas streams. We note
that the new flare requirements do not apply for separate facilities that are directly subject to
subparts R or Y, i.e., those that are not co-located at a petroleum refinery.
The commenters provided no basis for the statement that flares used for marine vessel loading
operations and gasoline loading are substantively different in design or use that other flares at the
refinery. As discussed in the preamble of the final rule, we find that the final flare requirements
are applicable to steam-assisted, air-assisted, and non-assisted flares. If the commenters are
referring to flares in dedicated service with a consistent gas composition, we have included
272
-------
specific monitoring provisions in the final rule for these types of flares; see the preamble to the
final rule for more detail on and our rationale for these requirements. For these reasons, we
disagree with the commenter that these flares are substantively different from those we
considered in developing the final requirements for flares in Refinery MACT 1.
Further, the technology reviews for subparts R and Y (gasoline loading racks and marine vessel
loading) did not evaluate or consider that flares may not be achieving 98 percent control
efficiency; therefore, we disagree the technology review performed for those subparts somehow
precludes adoption of these requirements for flares that are part of the Refinery MACT 1
affected source.
Comment 2: One commenter stated that the proposal attempted to address benzene tankage that
is subject to both part 61 subpart Y and Refinery MACT 1. The commenter noted that meeting
Refinery MACT 1 storage vessel requirements is more than equivalent to meeting the subpart Y
requirements and it is an efficiency to only have deal with Refinery MACT 1 burdens in such
situations.
Response 2: We appreciate the support of the proposed overlap provisions with part 61 subpart
Y and note that we are finalizing those provisions as proposed.
Comment 3: One commenter stated that some refinery units are subject to subpart CC and part
60 subpart VVa, without being subject to subpart GGGa. As is done for part 60 subpart
GGGa, the commenter recommended that proposed 63.640(p)(2) provide that if both Refinery
MACT 1 and subpart VVa apply, compliance with subpart VVa is only required.
Response 3: First, the commenter does not explain and we are unclear on when a refinery source
subject to CC would be also subject to VVa instead of GGGa. Based on the applicability of
GGGa, we expect that all equipment leaks subject to subpart CC would only be subject to GGGa.
Second, the overlap provision for subpart GGGa was included in previous revisions to subpart
CC and we did not receive comments to include VVa in this paragraph at that time. In the
revisions to §63.640(p)(2) that we proposed in June 2014, we simply added specific overlap
provisions related to the OGI provisions we proposed at §63.661. We expect the commenter was
seeking the use of the OGI provisions at §63.661 for units subject to VVa. However, we are not
finalizing the OGI provisions at §63.661 and, therefore, we are not finalizing the proposed
revisions to the overlap provisions at §63.640(p)(2).
Comment 4: One commenter stated that with the addition of expanded flare requirements to
subpart CC, many refinery flares will be subject to the subpart CC requirements and the General
Provisions requirements of sections 60.18 and/or 63.11 for flares. The commenter requested that
in order to avoid conflicts and confusion and eliminate unnecessary and wasteful burdens,
compliance with the Refinery MACT 1 flare requirements should be considered compliance with
60.18 and 63.11 for flares, where 63.670 also applies. The commenter suggested the addition of a
paragraph (s) to 63.640 as follows.
273
-------
(s) After the dates specified in Table 11 of this subpart for compliance with 63.670, compliance
with 63.670 is considered compliance with 60.18 and 60.11 flare requirements imposed by any
part 60, 61, or 63 subpart, for flares subject to 63.670 and 60.18 and/or 63.11.
Response 4: We agree that compliance with the sector specific requirements contained in
§63.670 for flares should be considered compliance with the general provisions for flares in 40
CFR 60.18 and 63.11 and have included regulatory text similar to that suggested by the
commenter.
9.2 Wastewater
Comment 1: One commenter stated that the proposed 40 CFR 63.647(c) applies new flare
provisions to Group 1 refinery wastewater streams where a flare is used to comply with 40 CFR
61.340 through 61.355 of part 61 subpart FF. However, 40 CFR 61.349(a)(2) establishes 95%
removal as the requirement for Refinery MACT waste management unit control devices and the
requirements in section 60.18 as the removal criterion for flares. The commenter stated that
imposing 40 CFR 63.670 and it's presumptive 98% destruction efficiency is a change in the
standard established in the Refinery MACT 1 rulemaking, that has not been justified under CAA
section 112(d)(6). Nor is there any evidence that 95% removal is not being achieved by flares
complying with section 60.18. The commenter also noted that the proposed 40 CFR 63.647(c)
appears to require flares at non-refinery facilities to meet the proposed 40 CFR 63.670
requirements which will drive those facilities to refuse to accept and treat refinery subpart FF
streams and impose large costs on refineries to install additional wastewater treatment facilities
and even hazardous waste incinerators leading to high costs with negligible benefits. The
commenter also stated that as discussed in the preamble, there is no risk driver or technology
improvement that would justify revisions to the Refinery MACT 1 wastewater provisions and
application of part 61 subpart FF already imposes stringent treatment and control requirements
on refinery wastewater. Therefore the commenter recommends maintaining the existing 40 CFR
63.647(c).
Response 1: We disagree that the existing 40 CFR 63.647(c) should be maintained. While part
61 subpart FF establishes the 95% reduction of organic HAP and section 60.18 establishes the
necessary criterion to meet this standard, neither regulation provides monitoring requirements to
assure the reduction is being achieved. Further, we believe all flares should be able to meet a
98%) destruction efficiency as proposed in the Refinery MACT 1 rulemaking. Finally, if a flare is
located on the refinery's property and is processing refinery vent streams, the flare would be
considered a refinery flare and be part of the affected source and thus subject to the provisions in
the rulemaking. We do not believe that these requirements would result in non-refinery facilities
to refuse to treat refinery wastewater streams.
Comment 2: One commenter stated that the EPA has acknowledged developments have
occurred in controlling emissions from wastewater treatment (79 FR at 36,918-19) and claimed
EPA has no valid basis for rejecting requirements for drain and tank controls, specific
performance parameters for enhanced biological units (EBU), and requiring wastewater streams
to be treated with stream-stripping. The commenter states that section 112(d)(6) does not allow
EPA to ignore "developments". Additionally the commenter believes the EPA also has no lawful
274
-------
basis to refuse to update these standards based on cost since the statute contains no authorization
to place cost above the statutory objectives of section 112(d).
Response 2: We disagree that EPA ignored these processes in its technology review. The
"developments" described by the commenter are all existing requirements or treatment methods
in part 61, subpart FF, with which many petroleum refinery wastewater treatment systems must
comply. In our technology review, we evaluated these technologies for wastewater controls for
small treatment systems and specific types of wastewater streams and we analyzed potential
emission reductions and associated costs. Consistent with our long-standing practice, we rejected
these requirements as not necessary for small treatment systems and specific wastewater streams
because of the high cost. As noted elsewhere in this response to comments, EPA has broad
discretion in deciding the factors to consider to determine whether it is "necessary" to revise an
existing MACT standard in light of an identified "development."
Comment 3: A commenter claimed EPA used incorrect data that is more than two decades old
instead of data from the 2011 ICR (which shows refinery wastewater treatment systems have
higher benzene and HAP concentrations).97 The commenter also raised a concern that EPA's use
of the Locating and Estimating Air Emissions from Sources of Benzene document was improper
because it omits several refinery processes that produce significant sources of wastewater and
include benzene.98 The commenter stated that there are inconsistencies between EPA's estimate
of the control effectiveness of biological treatment process systems and EPA's estimate of
allowable emissions because the Technology Review estimated an EBU achieves -88% control
efficiency, but EPA assumed that EBUs achieve 92% control efficiency in the allowable
analysis.99 Finally, the commenter claimed EPA must review the information collected as part of
the 2011 ICR on EBU control efficiency to determine the correct pollution control potential.
The commenter claimed that these inconsistencies either underestimate the benefits of better
controls or underestimates exposure.
Additionally, the commenter stated that because the EPA has recognized that wastewater is
driving a significant amount of health risk, the EPA must set standards to reduce these emissions
under section 112(f)(2) to provide the requisite "ample margin of safety." The commenter
believes the EPA should find risk from these emission points unacceptable and reject its cost-
based determination not to require more control from these sources, to assure the "ample margin
of safety" requirement is met.
Response 3: Refinery-specific wastewater generation rates, benzene, and organic HAP
concentrations for the model in the technology review were taken from the EPA's Locating and
97 ICR Component 1, supra note 103, at 64 tbl. 15-2
98 See T-4 Table C.
99
See infra EPA's Underlying Assumptions About Wastewater Treatment Systems Are Factually Incorrect. II.A.3.C
275
-------
Estimating Air Emissions from Sources of Benzene (L&E study).100 These data were used to
estimate the total flow rate and flow-weighted average benzene and HAP concentrations for each
of the modeled facilities in the technology review. Although, more recent data regarding
wastewater treatment operations was obtained during the 2011 ICR, some of the data was only
partially complete. Additionally, the ICR data was highly variable and contained significant
outliers. We were unable to quality assure the data within the time constraints of this rulemaking
process, and thus felt it was more appropriate to utilize the L&E study data. Since proposal, we
were able to review the ICR data in more detail. We performed an analysis of the facilities that
submitted a complete data set (meaning values for flow, benzene, and total organic HAP), and
calculated the average flow factor, flow-weighted average benzene concentration, and flow-
weighted average HAP concentration. We applied these factors to the model plants used in the
technology review and for 5 of the 6 model plants, the benzene loading rates determined from
the L&E study factors were higher than the loadings derived from the 2011 ICR factors. While
there are still issues with the 2011 ICR data for wastewater, this analysis indicates that the L&E
document factors provided a reasonably conservative estimate for the technology review of
petroleum refinery wastewater treatment systems, and that the use of the ICR data would not
change the outcome of the technology review. The analysis of the ICR wastewater data has been
added to the refinery docket EPA-HQ-OAR-2010-0682.
Contrary to the commenter's claim, the EPA's L&E study does not omit significant sources of
wastewater from refineries. It appears the commenter has attempted to match the ICR process
unit type description and the process units in Table 6-10 of the L&E study. However, the list of
process units the commenter attributes to Table 6-10 of the L&E study does not include all of the
units in the study (i.e., tank drawdown, catalytic reforming, full range distillation, thermal
cracking/vis breaking, hydrocracking). Additionally, the process units from the ICR list which
are not matched to Table 6-10 in the L&E study (i.e., Aromatics production, coke calcining,
ethylene production, fuel gas treatment, fuels solvent deasphalting, other petrochemical or
organic chemical production, oxygenate plant - methyl tert-butyl ether (MTBE), petroleum coke
storage, propylene production, product loading for container/ marine vessel/ rail car/ truck, and
other (specify)) are not significant sources of wastewater at a refinery and thus we believe the list
of sources from the L&E study is representative of the sources of wastewater at a refinery.
Regarding the claimed difference in the control efficiency for the technology review and that
identified for allowable emissions, we surmise from the comment that the commenter back-
calculated an average overall control efficiency of 88% for control option 4 for each model plant
in the technology review memorandum. It appears the commenter used the baseline emissions
and emission reduction quantities to calculate an overall average EBU control efficiency. This
calculation method effectively attributes the emission reduction solely to the EBU, when other
control options (with varying degrees of efficiency) are included (i.e., control of drain system
components), thus skewing the calculated efficiency of the EBU. This data cannot be used to
simply calculate the expected EBU's efficiency. Further, as described in the technical
100 U.S. Environmental Protection Agency. Office of Air Quality Planning and Standards,
Research Triangle Park, NC. EPA-454/R-98-011. June 1998. Available as Docket Item No.
EPA-HQ-OAR-2010-0869-0016.
276
-------
memorandum, model facilities 1 and 2 are below the Benzene Waste Operations NESHAP
(BWON) threshold (i.e., have a total annual benzene (TAB) less than 10 Mg/yr), and generally
would not be included in the calculation of the expected EBU control efficiency as these
facilities are not required to have add-on controls for wastewater. Although it is not accurate to
do so, if one calculates the expected average control efficiency relative to emission reductions
from control option 4 for model facilities 3, 4A, 4B, 5, and 6 and attributes it to the EBU only,
the control efficiency is approximately 92%, the value we used for allowable emissions.
Finally, we reviewed the EBU data collected as part of the ICR effort. We requested that 5
refineries test their EBU. Two of these facilities either did not have control systems meeting the
definition of EBU or reported sending their wastewater offsite for treatment. Three facilities did
submit test data. One facility (LA3C0690) reported an EBU control efficiency in excess of 99%.
The second facility reported having a steam stripper installed upstream of the EBU, therefore
greatly reducing the EBU load. In this configuration, the benzene emissions were reduced by
more than 95%, but the control efficiency of the EBU itself was lower. The remaining facility's
Refinery Wastewater Emission Tool (RWET) model projected an efficiency of 99%.
Based on the explanations given above, we disagree that we either underestimated the benefits of
better controls or underestimated exposure. Further, although these emissions drive risk, the risk
was found to be acceptable and thus there is no basis for adding additional controls for
wastewater irrespective of costs. Finally, although we are not adopting additional requirements
specific to wastewater collection and treatment as part of the technology review, as noted in
other responses, the fenceline monitoring program is designed with the specific intent to ensure
proper management of fugitive emissions sources such as wastewater collection and treatment
sources.
Comment 4: One commenter took issue with the EPA's proposal to continue to use benzene as a
surrogate for all other pollutants emitted by wastewater treatment (79 FR at 36,918) based on an
outdated consideration of surrogacy from 1994. The commenter noted that the D.C. Circuit
requires the EPA to justify the use of a surrogate based on facts in the record and that the EPA
has neither attempted to meet nor has it met the test to use a single pollutant to control other
emitted hazardous air pollutants. A surrogate is reasonable only if it meets three conditions: (1)
the target pollutant(s) must invariably be present in the surrogate; (2) the control technology used
to control the surrogate must indiscriminately capture both the surrogate and the target
pollutant(s); and (3) control of the surrogate must be the only means by which facilities achieve
reductions in the target pollutant.101
Response 4: We disagree that we have not justified the use of benzene as a surrogate for HAP
for wastewater systems, benzene is present in nearly every refinery stream, has a relatively high
volatility and water solubility so that it partitions into the wastewater more readily than most
other organic chemicals present at petroleum refineries. The few HAP present at refineries that
have higher solubility and lower octanol-water partition coefficients than benzene, such as
phenol or isobutyl ketone, have relatively low Henry's low constants (a measure of the relative
volatility in dilute aqueous solutions). Thus, these compounds are not readily emitted in the
wastewater collection system and are effectively removed from the wastewater during biological
101 Sierra Club v. EPA, 353 F.3d 976, 984 (D.C. Cir. 2004) (quoting Nat'l Lime Ass'n, 233 F.3d at 639).
277
-------
treatment. Benzene, on the other hand, has a relatively high Henry's law constant and will
readily volatilize to the atmosphere from open wastewater collection systems. Further, majority
of the total HAP loading in wastewater consists of compounds that are very similar to benzene
both in terms of chemical structure and volatility (from water phase to the air phase). Due to the
correlation between organic HAP and benzene, it is reasonable for benzene to serve as a
surrogate for HAP in wastewater. We also note that the emissions controls used to reduce
benzene emissions will also reduce emissions of other organic HAP and that there is no means
by which facilities can control organic HAP from wastewater without controlling benzene
emissions. Finally, in our technology review, we evaluated using different wastewater stream
applicability options and specifically considered using VOC as the surrogate. We concluded that
the existing requirements were reasonable and that is was not necessary to revise the standards or
to alter the surrogate used.
Comment 5: One commenter questioned how emissions from open waste water treatment tanks
and API separators will be reduced under the proposed Refinery Rules?
The commenter gives two examples of instance when wastewater controls did not work:
In October 2012 the local API separator emitted 32,654 pounds of pentane. This was
traced to the periodic openings of the pressure relief valves on the covered bays of the
API separator. Louisiana Department of Environmental Quality (LDEQ) SERC Incident
12-06692 T143678 http://vvvvvv.louisianarefinervaccidentdatabase.oru/pdfs/inc5670.pdf
In December 2009, the refinery in St Bernard could not handle a large, but normal
rainfall, and the wastewater treatment plant and the API separator did not function; the
result was a massive amount of oily wastewater in the neighborhood canal (photo
provided).
Response 5: It appears that these emissions events were unauthorized under the existing
standards and they would be unauthorized under the proposed rule. The proposed rule also
includes a fenceline monitoring program which will require facilities to monitor and take
corrective action to minimize fugitive emissions, such as those from wastewater treatment, where
the monitored emissions are above a specified action level. The fenceline monitoring program is
aimed at ensuring refineries manage sources of fugitive emissions and we believe that it will
incentivize refineries to ensure wastewater controls (among other things) are operating correctly.
9.3 Gasoline loading racks
Comment 1: One commenter agreed with the EPA's conclusion during the risk and technology
review of subpart R (71 FR 17352 (April 6, 2006)) that there have been no "developments in
practices, processes, and control technologies" that would justify revising the requirements for
gasoline loading at refinery subpart R facilities. Furthermore, this conclusion also applies to
flares dedicated to subpart R facilities.
Response 1: To the extent the commenter is suggesting that our conclusion for the technology
review for subpart R should extend to our review for the Refinery MACT rules we disagree and
stand by our analysis in this rule. Alternatively, it appears that the commenter may be suggesting
278
-------
that certain gasoline loading operations should be subject to the requirements of subpart R rather
than subject to the requirements of the Refinery MACT rules. We note that gasoline loading
racks at petroleum refineries are subject to the subpart CC. Simply because we generally
coordinate the control requirements between subpart R and subpart CC does not make the
gasoline loading racks at petroleum refineries "subpart R facilities." To determine whether a
specific operation is subject to the Refinery MACT rules, a person would need to refer to the
applicability provisions of the different subparts of the CFR. We have discussed in more detail in
Section 9.1.2 of this RTC document why we have not extended the same determination made for
subpart R facilities to facilities subject to subpart CC in this action.
9.4 Marine Vessel Loading Operation Provisions
Comment 1: One commenter noted that it had submitted comments in 2010 on proposed
revisions to subpart Y, which continued to exempt existing offshore loading terminals from
vapor recovery requirements but did require submerged loading (40 CFR 63.650(d)(6)). The
commenter supported the EPA's action in the RTR proposal continuing this exemption for
refining facilities and imposing submerged loading requirements.
Response 1: We appreciate the commenter's support of the requirements for existing offshore
loading terminals and for the proposed submerged loading requirements that we are including as
part of the final rule, as proposed.
Comment 2: Two commenters questioned the need for section 63.560(a)(4) of subpart Y since
the EPA stated in the preamble that the new requirement does not reduce HAP emissions or
residual risk since the U. S. Coast Guard already requires marine vessels to utilize submerged
loading and ensures compliance through vessel inspection and enforcement. The commenters go
on to say that the EPA stated that all marine vessels are therefore presumed to already be in
compliance and that this new requirement will only create duplicative recordkeeping and
reporting burdens (e.g., any deviations would have to be reported to both the Coast Guard and
the EPA) and duplicative enforcement burdens (i.e., both the EPA and Coast Guard will inspect
for compliance.)
Response 2: While the changes to 40 CFR 63.560(a)(4) will likely not affect actual emissions,
this change will reduce the allowable emissions from sources affected by the change. Though
there may be some duplicity in recordkeeping or reporting requirements, it should be
insignificant as facilities subject to the Refinery MACT already routinely report deviations to the
EPA, and thus it seems this would just require an update to the report to include this information.
Further, these are two separate programs with different drivers, and thus the information
submitted will be used in different ways by each Agency.
Comment 3: Two commenters stated that the EPA says $3,900 per ton to reduce HAPs is an
excessive amount, and they do not agree. The commenters also stated that the EPA should not
rely upon industry-reported emissions inventories, as the EPA does, to determine cost
effectiveness, because industries have shown time after time, they underestimate their emissions.
These commenters argued that this is a conflict of interest.
279
-------
Response 3: We disagree that EPA has determined that $3,900 per ton of HAP emissions
reduced is "excessive" and we did not reject any control option with that cost-effectiveness. It
appears the commenter is referring to a control option evaluated for marine vessel loading of
gasoline into barges that had a cost-effectiveness of $3,900 per ton of VOC reduced. The cost-
effectiveness of this option was estimated to be $77,000 per ton of HAP reduced and we did
determine that it was not necessary to revise the requirement under section 112(d)(6) based on
the cost per ton of HAP reduced. With respect to using emissions data provided by the industry,
we note that the refinery owners and operators have access to the necessary process information
to estimate the refinery emissions. We also note that we prepared an Emissions Protocol
document that refinery owners and operators were required to follow in developing the emissions
inventory. The fenceline monitoring requirement we are establishing in the final rule, in addition
to providing improved management of fugitive emissions, will also provide additional data that
will allow better verification of emission estimates in the future.
280
-------
10.0 Refinery MACT 2 Amendments
Comment 1: One commenter stated that the Operating, Monitoring, and Maintenance Plan
(OMMP) required by Refinery MACT 2 will need to be updated due to the new rule
requirements (e.g. stack testing requirements, new monitoring requirements, etc.). The
commenters asserted that currently any changes to an approved OMMP must be submitted and
cannot be followed until the EPA approves the new version of the OMMP. The commenter
believed this approval process puts the refiner at great risk of not being able to comply with the
rule by the compliance date (in many cases in this proposal the publication date) and
recommended that the EPA remove the OMMP approval requirements as follows:
§63.1574(f)(1): You must submit the plan to your permitting authority for review and approval
along with your notification of compliance status. While you do not have to include the entire
plan in your part 70 or 71 permit, you must include the duty to prepare and implement the plan as
an applicable requirement in your part 70 or 71 operating permit. You must submit any changes
to your permitting authority for review and begin complying with the revised plan once
submitted.
Response 1: Generally, we disagree that we did not provide adequate time for monitoring
revisions in Refinery MACT 2. We provide up to 18 months to comply with any new monitoring
requirements in Refinery MACT 2, which we consider adequate time to complete the necessary
steps to comply with the new provisions, including, if necessary, receiving approval of changes
to the OMMP. Additionally, we are not finalizing the revisions for FCCU metal HAP emissions
limits for units subject Refinery NSPS subpart J as we had proposed. Based on the requirements
we are finalizing, we consider it unlikely that refinery owners or operators will have to submit
changes to their OMMP. However, given that Table 41 now outlines minimum QA/QC
requirements for the monitoring systems, we consider it reasonable to allow refinery owners or
operators to submit revisions to the OMMP, if necessary, and to comply with the requirements of
the revised plan upon submission (as requested by the commenter). Therefore, we have revised
this requirement accordingly in the final rule. We are also providing an 18 month transition
period to comply with the requirements in Table 41. Together, these revisions will eliminate
compliance issues noted by the commenter if new monitoring systems are needed and/or
approval of the revised OMMP is delayed.
Comment 2: A commenter stated that they have not had the time prior to the close of the
comment period to comprehensively and systematically review the subpart UUU revised tables
in detail. The commenter believed these tables are complicated, detailed, and difficult to
interpret, and therefore requested that the EPA ensure that they are accurate, clear, and consistent
with the rule text in the promulgated rule.
Response 2: We appreciate the commenter's suggestion, and we extended the comment period
in efforts to allow commenters sufficient time to review the rule in its entirety. We have
reviewed the various tables in subpart UUU, and we have made every attempt to ensure that the
final revisions to these tables are consistent and accurate.
281
-------
Comment 3: A commenter believed that the proposed §63.1570(d) requirements ("During the
period between the compliance date specified for your affected source and the date upon which
continuous monitoring systems have been installed and validated and any applicable operating
limits have been set, you must maintain a log detailing the operation and maintenance of the
process and emissions control equipment.") is unclear, ambiguous, and unnecessary and should
be deleted. The commenter went on to say, that there can be no gap between the compliance date
and the availability of continuous monitoring systems because they must be installed and
validated to meet applicable operating limits no later than the compliance date. Even if that time
period is clarified, the commenter questioned what information and how much detail will be
required to maintain in a log and what the justification for having a log is if there is no applicable
emission limit or the existing emission limit and applicable compliance provisions apply.
Response 3: This phrase was previously included in §63.1570(c), and we acknowledge that
moving this requirement to a separate paragraph created ambiguity. In its previous context within
§63.1570(c), it is clear that this requirement refers to the general duty to minimize emissions in a
manner consistent with safety and good air pollution control practices. While certain provisions
of the rule may not become effective until 18 to 36 months after promulgation, owners or
operators of an affected facility must still comply with this general duty to minimize emissions at
all times and must operate any control equipment already installed in a manner consistent with
safety and good air pollution control practices for minimizing emissions. Therefore, we are
clarified in the final rule that the requirement to maintain a log in §63.1570(d) is to document the
procedures used to minimize emissions according to the general duty in §63.1570(c).
Comment 4: One commenter requested that, under §63.1573, EPA add an alternative to
measuring liquid flow, similar to that provided in §63.670(i)(4).
Response 4: The newly proposed flare provisions are specific to gaseous flows, understanding
the limited flow restrictions or equipment that would alter the back pressure in the
specific application to flares. We do not expect that this alternative would be suitable for
determining liquid flow rate in all applications. For example, depending on where the flow must
be monitored relative to other equipment and the possibility of blockage of the liquid flow lines,
the application of this method may not be as accurate and reliable as in the application of
gaseous flow to a flare. Therefore, we are not revising 40 CFR 63.1573 to provide the requested
alternative. However, we do note that section 63.1573(e) and (f) provides a means for a refinery
owner or operator to apply for alternative parameter monitoring. Thus, if a refinery owner or
operator expects that this alternative is applicable and accurate for their specific application, then
they may apply to use this alternative parameter monitoring approach.
10.1 Catalytic Cracking Units (fluid and other)
Comment 1: A commenter asserted that the proposed rule can pose significant challenges to
sources that are regulated under existing state implementation plan approved rules. For example,
the commenter stated, facilities regulated by the SCAQMD already have many prohibitory rules
that regulate unit operations. The commenter claimed that SCAQMD Rule 1105.1 regulates the
operations characteristics including flow rate, power requirements and emissions from FCCU.
The commenter believed that the EPA's proposed NESHAPs could superimpose additional
282
-------
requirements that may be incompatible with facility operations designed to comply with Rule
1105.1 and recommended that the EPA consult with stakeholders to modify the requirements
appropriately for units that are already regulated under local or state air quality rules.
Response 1: We appreciate the commenter's perspective, but the SCAQMD Rule 1105.1
requires a facility to develop a monitoring plan and operating limits so it is impossible to fully
evaluate potential differences. We note that as a general rule, it would be difficult or impossible
for EPA to ensure that its rules are consistent with rules from the many states and sometimes
local agencies that also have authority to regulate emissions. Regarding the specific issue raised
by the commenter concerning the SCAQMD rule, we reviewed the general suggested parameters
to be monitored in Attachment A of Rule 1105.1 and do not find that the revisions in subpart
UUU would be incompatible with these suggested requirements. Both the monitoring plan in
Rule 1105.1 and the provisions in §63.1573(e) and (f) provide a means for a refinery owner or
operator to harmonize the federal and local agency monitoring requirements and we find no
reason to modify the subpart UUU monitoring requirements as a result of this comment.
Comment 2: One commenter stated that the standards on coke burning and sulfur emissions
assume that each facility has the same catalytic cracking unit under the same air pressure and
using the same O2 enrichment in order to calculate the pollutants from the facility. However, the
commenter stated that many of these facilities are old and are calibrated differently and therefore
assuming that all facilities are under the same conditions is incorrect and gives facilities huge
margins for compliance.
Response 2: The equations used to determine the coke burn-off rate for the FCCU require flow
rate corrected to standard conditions (1 atmosphere pressure) and account for oxygen enrichment
and the amount of excess oxygen in the exhaust gas. Similarly, the revisions to the SRU
provisions were included specifically to account for the use of oxygen enrichment. The SRU
emission limits are provided on a concentration basis corrected to 0% excess air so these limits
are not subject to issues related to differences in pressure or excess air use rates. Therefore, the
commenter's assertion that we are assuming all facilities are operated under the same conditions
and that we provide facilities with huge margins for compliance is inaccurate.
10.1.1 Inclusion of NSPS Ja compliance option
Comment 1: A commenter requested that the amendments to §63.1564 and Tables 1-7 allowing
the use of PM CPMS as the compliance demonstration for FCCUs that are also subject to NSPS
subpart Ja should be made available to all FCCUs whether or not they are subject to NSPS Ja
(including those required to comply with NSPS J as their PM compliance option) since it has
already been demonstrated in the NSPS subpart Ja rulemaking to be an equivalent compliance
assurance approach for FCCUs. The commenter stated that it may be a preferable alternative for
some unit configurations to demonstrate compliance during periods of startup and shutdown.
Another commenter agreed that PM CEMS should be an optional compliance approach since
currently PM CEMS have not been fully demonstrated as reliable in refinery applications. The
commenter believed that by allowing this optional compliance approach the EPA can promote,
voluntarily, the further development of this technology which will allow future evaluations to
283
-------
determine whether enhanced compliance occurred in practice. Accordingly, this technology
forcing effect can yield far greater benefits than mandating subpart Ja requirements. The
commenter requested that the EPA not only revise Refinery MACT 2 in this manner, but it
amend subpart Ja to apply a similar, optional approach for these yet to be fully demonstrated
methods of compliance. The commenters supported an annual performance test requirement for
all sources.
Response 1: We did not intentionally intend to prevent an owner or operator of an FCCU subject
to subpart J from using the PM concentration/PM CEMS option. We note that subpart J at 40
CFR 60.100(e) allows units subject to subpart J to demonstrate compliance using the provisions
in subpart Ja, so we intended to allow all FCCU to elect to use the PM concentration limit and
PM CEMS monitoring option. To clarify this, we are revising these tables (primarily Item 1) to
allow units subject to NSPS subpart J the option to meet the 0.040 grains (gr)/dry standard cubic
foot (dscf) PM concentration limit if a PM CEMS is used, as allowed under 40 CFR 60.100(e).
With respect to annual performance test requirements, we proposed to require that the FCCU be
re-tested at least once every 5 years. Combined with the enhancement in the monitoring
requirements, we concluded that testing once every five years was sufficient to ensure
continuous compliance with the emission limits. However, for the reasons provided in Section
10.1.2 of this document, we are requiring owners or operators of FCCU that elect to comply with
the 20 percent opacity operating limit that also have emissions greater than 0.80 lbs/1,000 lb
coke burn-off to conduct performance tests annually.
Comment 2: A commenter stated that two compliance options are provided in §63.1565 and
Tables 8-14 for CO emissions (as a surrogate for organic HAP) and stated that is unclear what
the compliance options are for FCCUs that will become subject to NSPS subpart Ja. The
commenter also noted that NSPS subpart J includes language that allows for compliance with Ja
as a NSPS J compliance option and that it is confusing not to have it spelled out in Refinery
MACT 2, particularly relative to changes in NSPS J requirements spelled out in Refinery MACT
2. Similar to PM, the commenter believed that replacing existing requirements for CO options
cannot be done without proper justification and analysis being provided for comment and such
changes should be strenuously avoided. However, referencing NSPS Ja §60.102a everywhere
§60.103 is referenced in §63.1565 and Tables 8-14 would provide a clear compliance option for
FCCUs subject to NSPS Ja, without changing the requirements for other FCCUs.
Response 2: Both subpart J and subpart Ja have the same emission limits for CO (hourly average
less than 500 ppmv, dry basis) and both require use of a CO CEMS (unless it is demonstrated
that all hourly average CO concentrations are less than 50 ppmv, dry basis, corrected to 0 percent
excess air for a 30 day period). The Refinery MACT 2 emission limitations are identical to the
subpart J/Ja limits, but Refinery MACT 2 provides CPMS compliance options as alternatives to
the CO CEMS. We did not intend to limit or duplicate compliance requirements for units subject
to subpart Ja and we agree that Item 1 of Table 8 to subpart UUU should include units subject to
40 CFR 60.102a and that Item 2 would exclude units subject to either NSPS J or Ja, and have
revised these tables accordingly.
284
-------
Comment 3: A commenter stated that a delay-of-repair provision is needed with the proposed 12
hour repair requirement in §60.105a of NSPS Ja and §63.1573(b)(3) for faulty atomizing spray
wet scrubber air and water lines. The commenter believed that while evaluation of the problem
and identification of possible fixes can be initiated within 12 hours, depending on the severity
and location of the leak it may take a few days before the repair can be completed. The
commenter recommended that the EPA require the site to initiate the evaluation of the repair or
replacement within 12 hours, with the repair to be completed as soon as practicable but not later
than 15 days. The commenter provided recommended rule text as follows:
63.1573 What are my monitoring alternatives?
(b) What is the approved alternative for monitoring pressure drop? You may use this alternative
to a continuous parameter monitoring system for pressure drop if you operate a jet ejector type
wet scrubber or other type of wet scrubber equipped with atomizing spray nozzles. You shall:
(1) Conduct a daily check of the air or water pressure to the spray nozzles;
(2) Maintain records of the results of each daily check; and
(3) Initiate evaluation of the repair or replacement of Repair or replace faulty (e.g., leaking or
plugged) air or water lines within 12 hours of identification of an abnormal pressure reading. The
repair or replacement shall be completed as soon as practicable but not later than 15 days after
identification of the abnormal pressure reading. As an alternative, the owner or operator may
demonstrate via stack testing that the inorganic HAP limit is being met at these new conditions.
Response 3: We do not consider it appropriate to allow continued operation of the wet scrubber
for 15 days while there are issues with the spray nozzle system. In most cases when the water
spray system is not operating properly, there will be inadequate contact between the water
and PM particles, significantly reducing the PM removal efficiency of the system. We do agree
that facility owners or operators can demonstrate via a stack test that the inorganic HAP emission
limit is being met at the new conditions, but these provisions are already included in Refinery
MACT 2 at 40 CFR 63.1571(e). We are not adopting the revisions suggested by the commenter.
Comment 4: A commenter stated that the EPA set lower PM emission levels for new, modified
or reconstructed sources under subpart Ja but then pronounced that it is necessary to revise
Refinery MACT 2 to incorporate this limit for new sources, because FCCUs would already be
required to meet this limit under subpart Ja, and therefore there is no added cost associated with
the additional requirements. The commenter stated that this reasoning is significantly flawed
because a Refinery MACT 2 new source is not the same as a NSPS subpart Ja affected facility.
All FCCUs constructed after 1998 are subject to new source standards under Refinery MACT 2,
but few, if any, of these sources are subject to subpart Ja, which does not apply to affected
facilities constructed before 2007. Thus, the commenter believed that the EPA's statement that
there would be no cost for new sources to comply with Refinery MACT 2 is incorrect.
285
-------
Moreover, the commenter noted, even if all new sources under the Refinery MACT 2 standard
must already comply with subpart Ja requirements, this does not support the EPA's assumption
that revisions to Refinery MACT 2 are then automatically necessary. In fact, arguably, revision
would be unnecessary, because the revision would be a paperwork exercise with no benefits. The
EPA must do more than simply profess that it is possible, and therefore it is necessary.
Response 4: First, we only established more stringent PM emissions in Refinery NSPS subpart
Ja for newly constructed FCCU (after May 14, 2007); modified and reconstructed FCCU (after
May 14, 2007) must still comply with the 1.0 lb/1,000 lb coke burn-off emission limit. Second,
we recognized that the new source applicability dates in Refinery MACT 2 are different than in
the Refinery NSPS Ja; therefore, we did not require all new FCCU based on the Refinery MACT
2 applicability date to comply with the 0.5 lb/1,000 lb coke burn-off Refinery NSPS Ja new
source PM emission limitation. As seen in Table 1 to subpart UUU, the new and existing FCCU
requirements (according to Refinery MACT 2 definitions) are not segregated and only FCCU
that are subject to NSPS PM standards in 40 CFR 60.102a(b)(l)(ii), i.e., the newly constructed
FCCU after May 14, 2007, according the Refinery NSPS Ja, must meet the 0.5 lb/1,000 lb coke
burn-off PM emissions limit. Therefore, we maintain that the revisions to Refinery MACT 2 to
align with the subpart Ja requirements for subpart Ja newly constructed sources will not result in
additional costs based on the proposed Refinery MACT 2 revisions.
We agree that the inclusion of these provisions in Refinery MACT 2 does not result in additional
emissions reductions; however, we find that there is benefit in coordinating the compliance
requirements between subpart Ja and Refinery MACT 2. Therefore, we are retaining the NSPS Ja
new source emissions limit, if applicable, in Refinery MACT 2.
Comment 5: One commenter stated that without certainty that the bag leak detection (BLD) will
provide better assurance that the PM emissions limit is being met than existing baghouse
compliance methods, the EPA should only provide this compliance approach as an optional
method under Refinery MACT 2. Further, the commenter stated that proposed Table 2 for
subpart UUU should clarify that the BLD is only an option for new units subject to NSPS and
that these units do not have to meet both that and the opacity standard. Finally, the commenter
stated that given the EPA's admission that there is no evidence that BLD is more effective than
existing baghouse compliance methods, the EPA should also revise subpart Ja to provide
multiple compliance monitoring options rather than mandating the bag leak detection system.
Response 5: We disagree with the statement that the EPA admitted that there is no evidence that
BLD are more effective than existing baghouse compliance methods. BLD are proven
technologies and have been required for baghouse control systems in many EPA rules for the
past decade. We clearly stated that we consider BLD to be superior to opacity monitors for
ensuring fabric filter control systems are operating efficiently, with our primary point being that
there is only one baghouse currently used to control FCCU emissions and we did not consider it
necessary for this one unit to upgrade their monitoring system. As discussed in the preamble to
the final rule and in the next section of this document, we received numerous comments on the
proposed rule that argued that opacity was not well correlated with PM emissions. In a meeting
between representatives from the EPA and from API, one representative noted that the
extrapolation allowed in the site-specific opacity limit results in an opacity limit of greater than
286
-------
100 percent for their unit controlled with a baghouse. One could never exceed such an
operating limit regardless of whether significant holes develop in the bags, leading to high PM
emissions well exceeding the 1 lb/1,000 lb coke burn-off Therefore, having investigated this
issue more fully in response to comments received, we are convinced that a BLD system is the
best monitoring system for baghouse control systems. In general, we are finalizing provisions for
the use of BLD systems in Refinery MACT 2 and we are not revising subpart Ja to include an
opacity limit alternative for FCCU controlled with a baghouse.
While we are convinced that the BLD system is superior monitoring system than a continuous
opacity monitor for FCCU controlled with a baghouse, we are not revising the NSPS J
compliance option in Refinery MACT 2 to the extent proposed. For the reasons provided in the
preamble of the final rule and in the following sections of this document, we are finalizing a 20
percent opacity operating limit for units subject to subpart J. If an FCCU that is controlled with a
baghouse is subject to Refinery NSPS J, then an opacity operating limit is provided as a
compliance option. If an FCCU that is controlled with a baghouse is subject to Refinery NSPS
Ja, then the owner or operator would be required to install and use a BLD system.
10.1.2 Phase-out of NSPS J compliance option
Comment 1: Several commenters requested a minimum of 18 months to comply with the
revised FCCU standards (except for PM) to revise their procedures, permits, AMPs and
to perform training. Several of these commenters also requested at least 3 years for changes that
require a project to install new equipment, including monitoring changes. One commenter
requested 3 years to comply with the new requirements of the Refinery MACT 2 if promulgated
as proposed in order to coordinate compliance with ongoing plant modifications, or turnarounds.
Response 1: The proposed rule provides an 18-month period to comply with the revised
standards for FCCU and 36 months to comply with the revised flare requirements, if
applicable. For the revised FCCU standards, the primary requirement would be to schedule and
perform a performance test, if necessary, to determine the new operating limits. As the primary
emission limits are not being revised, we do not consider it necessary to provide more time to
comply with the revised FCCU operating limits. We also reviewed the requirements in Table 41
for CPMS and consider the requirements provided there are reflective of industrial monitoring
systems and that additional time is not needed in order to comply with the requirements in Table
41 as revised in the final rule (Note: the O2 CPMS requirements were revised, see Section 10.4
of this document).
Comment 2: Commenters stated that FCCUs, particularly those controlled using tertiary
cyclones, cannot meet the proposed opacity limitations at all times. Commenters believed that
available data show that a performance test based limit would likely result in an opacity limit in
the range of 10% for cyclones and such a limit cannot be maintained under normal operating
conditions because variability in operations of both the FCCU and/or PM control equipment may
be reflected in opacity changes with no impact on permitted PM.102 Commenters stated that a
10% opacity limit would cause FCCUs to exceed the opacity indicator, when the actual PM
102 API discussed these performance issues in a meeting with Brenda Shine and staff on August 7, 2014.
287
-------
emissions are well below the standard and that the EPA has not demonstrated that units with
higher opacity are exceeding the 1 pound per 1,000 pound limit.
Commenters stated that parameter monitor limits established through performance testing are
appropriate if there is a reasonably direct correlation between control device operating
parameters and emissions, but because FCCU regenerator opacity measurements are less precise
than other parameter measures and are strongly influenced by particle size (which is inversely
related to mass), opacity is only a qualitative indicator for PM emission rate or mass in this
service. One commenter stated that the EPA conceded in the NSPS subpart Ja (page 76 of Public
Comments and Responses Dated April 2008) rulemaking that opacity does not correlate well
with PM emissions but rather is an indicator of control device performance.
Two commenters included data that they believe show that opacity is not a strong indicator of
PM emissions and demonstrates the difficulty in establishing an opacity limit via performance
testing. One of these commenters included figures that show the R2 statistic of PM emissions
(lb/1,000 lb coke burn-off or lb/hr) versus opacity for individual FCCU that use third stage
separators for PM control. The other commenter shows a summary of performance test data from
Marathon Petroleum Corporation's Canton refinery from which it states Marathon Petroleum
Corporation, EPA, and the Department of Justice concluded opacity was not a strong indicator of
PM emissions
Commenters recommended keeping the existing 30% opacity limit but increasing the stack test
frequency from every 5 years to an annual PM stack test. If EPA changes the opacity limit for
these units, a minimum of 3 years is needed to comply.
Response 2: As discussed in the preamble to the final rule, studies have shown that the 30
percent opacity limit does not correlate well with the 1 g/kg coke burn-off limit and that an
FCCU can comply with the 30 percent opacity limit while its emissions exceed the 1 g/kg coke
burn-off PM emissions limit. This same study indicated that 20 percent opacity limit provided a
better correlation with units meeting the 1 g/kg coke burn-off PM emissions limit. We also
reviewed the data submitted by the commenters regarding PM emissions and opacity
correlation. While the data may be used to suggest that there is variability/uncertainty in the
PM/opacity correlation, the data certainly do not support that a 30 percent opacity limit would
ensure compliance. We note that 4 of the 6 individual test runs that exhibited an opacity of 12
percent or higher had PM emissions greater than the 1.0 g/kg coke burn-off PM emissions
limit. Even when considering the uncertainty associated with the PM/opacity correlation, the data
provided would suggest an opacity limit of approximately 15- or 20-percent would be
appropriate to ensure compliance for this unit and that the 1 g/kg coke burn-off PM limit would
be exceeded well before 30 percent opacity is reached. Therefore, we are not including a flat 30
percent opacity limit option to demonstrate compliance with the PM (or metal HAP) emissions
limit.
In our review of the PM/opacity data submitted by the commenters, we agree that the
PM/opacity correlation is not perfect, however, the correlation appears to be reasonable.
Some correlations presented by the commenter showed PM emissions on a lb/hr rate rather than
a lb/1,000 lb (or g/kg) coke burn-off basis, and the mass emission rate can decrease due to flow
288
-------
reductions even though the PM concentrations (i.e., opacity and PM emissions in g/kg) are
constant. We also note that the comments generally required the correlation to go through the x/y
axis intersection (0% opacity and 0 PM emissions). While we agree that the correlation should
go through the x/y axis intersection, we also consider it appropriate to consider this as a
measured value (based on calibration and instrument drift checks) and including this
point when determining the R2 of the correlation leads to much higher correlation (0.62
compared to 0.07; see memorandum entitled "Analysis of FCCU Opacity Data" included in
Docket ID No. EPA-HQ-OAR-2010-0682). Therefore, while we recognize that this correlation is
not perfect, it is reasonable and it certainly suggests that opacity readings above 20 percent (for
the FCCU/cyclone tested) would most likely indicate that the PM emissions would exceed
1.0 g/kg PM emissions limit.
We note that when analyzing the PM/opacity data, we did not use the test data from the first day
of testing (April 2004). The emissions from the unit during this test well exceeded the
1.0 g/kg PM emissions limit. The tertiary cyclone had just been installed at the time of the test. It
appears some adjustments were made to the system after the April 2004 tests. All subsequent test
data were used. While some individual test runs exceeded 1.0 g/kg emissions limit, all three run
averages met the 1.0 g/kg emissions limit. For each combination of three consecutive runs (not
necessarily performed on the same month), we evaluated the variability in the 3-run average
opacity limit as well as the site-specific opacity limit that would be calculated based on the
source test runs. The variability analysis was predicated on the idea that the opacity limit was an
indicator that the control device was operating similarly to when the performance test was
conducted. Based on the variability of the 3-run average opacity limits, we determined that, if the
3-hour average opacity exceeded 20 percent, then it was highly likely (98 to 99% confident) that
the FCCU emissions are exceeding the 1.0 g/kg emissions limit.
Based on our review of the comments and data submitted by the commenters, we are adding an
opacity operating limit for units subject to subpart J (or that elect compliance with subpart J). To
further ensure units that are required to or elect to use the subpart J option, and that are operating
near the PM emissions limit during the performance test, are complying with the mass emissions
limit at all times, these units will be required to conduct a performance test annually rather than
once every 5 years. The commenter suggested this option be specific to units using a 30 percent
opacity, but we determined that a 20 percent opacity limit determined on a 3-hour average is
much more indicative of good PM emissions performance and that re-testing is only necessary
when the mass PM emissions are near the emissions limits. Therefore, we are revising the on-
going testing requirement to be annual when the PM emissions measured during the source test
is greater than 0.8 g/kg coke burn-off for units using the fixed 20 percent opacity 3-hour average
on-going compliance option. The testing frequencies for units with measured emissions at
or below these levels will be once every 5 years.
Comment 3: One commenter stated that the proposed Table 2 has led to confusion regarding the
types of controls that are subject to site-specific opacity limits. The commenter believed the
EPA's intent was to give units equipped with fabric filters or ESPs the option of demonstrating
compliance using either the specified parametric monitoring approach or the site-specific opacity
monitoring approach, until such time as the site triggered NSPS subpart Ja for that unit. The
commenter stated that the table does not make this clear and that the EPA should clarify that
289
-------
units equipped with ESPs can choose to demonstrate compliance by coke burn-off or flow rate
limit and total power limit and secondary current, without the need to demonstrate compliance
with the site-specific opacity limit.
The commenter suggested revising the monitoring requirements for ESPs in Table 2, item 4(b)(i)
as follows, to be consistent with existing regulations and Table 2, items 1(c) and 2(c), as follows:
maintain the daily average gas coke burn-off rate or daily average flow rate no higher than the
limit established in the performance test; and maintain the daily 3-hour rolling average voltage
total power and secondary current (or total power input) above the limit established in the
performance test.
Response 3: We note that the owner or operator of an FCCU is only required to comply with a
specific row in Table 2 based on the applicability of the options and the type of monitoring
system and control device used. We understand that some emissions limit/control device pairs
may have an option to elect one of two different monitoring options. A facility may elect to
comply with whichever one option they choose (provided they have the appropriate monitoring
systems), and must continue to comply with that option until he notifies the EPA (or the
administrative authority) that a different option will be used to demonstrate compliance. While it
is true that an owner or operator of an FCCU that has both ESP parameter monitoring systems
and an opacity monitoring system is not required to comply with both sets of operating limits, it
is also true that the owner or operator must elect which compliance option they will meet at all
times. To clarify this, we are revising §63.1564(a)(2) to state that if two separate monitoring
system options are provided for a specific control device, the owner or operator may select the
monitoring option with which the facility will comply at all times. The owner or operator must
provide notice to the Administrator (or designated authority) prior to changing the monitoring
compliance option for the facility.
The primary reason certain requirements were divided into numbered requirements in the last
column was to clarify what parameters were provisionally available while the transition to the
newly proposed requirements was being made. Since this transition period is considered
essential, we are maintaining the separate numbered listing of some of these requirements.
10.1.3 Startup shutdown provisions
Comment 1: One commenter was concerned about the lack of a mechanism to obtain case-by-
case review and approval of separate alternative SSM limits. While the commenter
recommended some alternative limits in their discussions on FCCUs and SRPs, they claimed that
other alternatives will likely be needed because of 1) advancements in technology, 2) process
improvements, and 3) the possibility that not all situations may be readily foreseen at this
time. The commenter urged the EPA to allow a process for companies to submit an application
for case-by-case alternative SSM limits to be approved by either the EPA or delegated States,
similar to the case-by-case alternate NOx limits for process heaters operated under certain
conditions as provided in subpart Ja.
The commenter also stated that Table 1 of Refinery MACT 2 tabulates the applicable Metal HAP
emission limits and Table 8 the applicable Organic HAP emission limits. The proposed tables
290
-------
only include the normal operating limits. The commenter suggested, that for clarity, the
alternative limits applicable during other time periods should be included in these tables.
Response 1: With respect to the application for case-by-case alternative limits to be approved by
the agency, we note that this option exists for nonopacity emissions standards through the
General Provisions in 40 CFR 63.6(g). We disagree that the process for establishing alternative
limits for MACT standards should be modeled after that approach in NSPS Ja because section
111 allows consideration of costs in establishing limits while MACT standards are based on the
best performing facilities are without consideration of cost. We note that if a source wishes to
seek an alternative limit under 40 CFR 63.6(g), the owner or operator would need to demonstrate
that the source would be achieving at least as much HAP emissions reductions as obtained via
the SS requirements in the final rule.
We are not revising Tables 1 and 8 to specifically include the startup and shutdown (SS)
emissions limitations, because we consider these provisions to be operating limits. We have
specifically included these limits in Tables 2 and 9 and we are adding monitoring and
compliance requirements to other tables in subpart UUU to specify the monitoring and
compliance requirements associates with these SS emissions limitations.
10.1.3.1 Opacity for metal HAP
Comment 1: Several commenters stated that although some FCCUs with WGS are designed to
be able to safely keep those controls on-line during periods of SS, including hot standby, there
are situations where some sites are unable to meet their WGS operating parameter requirements
during these periods.
One commenter stated that maintaining the 3-hour rolling average pressure drop above the limit
(established in the performance test) for scrubbers utilizing Agglo-Filtering Modules I Cyclone
Systems (or similar equipment) during the process unit startup is not achievable. The
commenters stated that typically, the required flue gas rates necessary to achieve the pressure
drop established in the performance test does not occur until the unit is near the end of the
process unit startup. This startup time can last 30 to 60 hours depending on the size and design of
the unit, and there is no practical monitoring alternative during this period. The commenter
requested that EPA provide relief on this monitoring point during startup until such time as the
process rates established in the performance test are obtained.
Another commenter stated that FCCU equipped with WGS cannot consistently satisfy the 1.0 lb
PM/1,000 lb coke burn limit during all phases of the startup sequence, since there may be
insufficient coke burn against which to compare the particulate matter emissions. Indeed, the
coke burn rate may not even support a meaningful calculation for comparison against the
standard. Instead, FCCU equipped with WGS typically demonstrate compliance during stack
testing by achieving performance consistent with the desired operating parameters for the
scrubber system; the commenter requested this compliance demonstration apply for both startup
and shutdown scenarios.
291
-------
A different commenter stated that units with WGS can also experience operational and safety
issues during SS activities. Refineries do not operate the LoTOx portion of the unit used for NOx
control because sulfites are needed to minimize the ozone slip. The commenter noted that
unsteady state operations cause the WGS to achieve a variable rate of emissions control
efficiency making it ineffective as a continuous emissions control device. Operational variability
also can result in a need to bypass the wet scrubber altogether. The commenter pointed out that
FCCU with scrubbers that operate under AMP may not operate within monitored parameters
during these periods because of the fluctuations that are seen in the gases being processed
through the unit. Accordingly, the commenter stated, the Refinery MACT 2 level of control does
not represent the MACT floor for SS activities at FCCUs with WGS.
Response 1: We understand that during startup, shutdown and hot standby, gaseous flow rates
and pressure drop may not be as high as during normal operation and that the emissions limit is
not relevant when there is no coke burn-off As discussed previously we established operating
limits with respect to the primary cyclones that is applicable regardless of the control device
used. However, we consider it feasible for that facilities that use wet scrubbers to maintain the
liquid to gas ratio operating limits during periods of startup, shutdown and hot standby.
Therefore, we are providing an alternative compliance option for FCCU equipped with wet
scrubbers to meet only the liquid to gas ratio operating limits during startup, shutdown and hot
standby. That is, during periods of startup, shutdown and hot standby, the pressure drop
operating limit is not applicable. This option provides a method for owners or operators of FCCU
equipped with wet scrubbers to use their existing monitoring parameters to demonstrate
compliance with the emission limitations rather than using completely different parameters.
Regarding the operation of the LoTOx system, we note that NOx control is not required by the
MACT standards, so there is no requirement in the MACT rule requiring that the LoTOx system
be operated during startup or shutdown. Furthermore, the Refinery NSPS subpart Ja
requirements have fairly large averaging periods, which we expect will allow refineries to meet
the emissions limitations at all times. However, we are not establishing alternative emission
limits for periods of startup and shutdown in NSPS subpart Ja at this time.
Comment 2: One commenter noted that issues can arise during the shutdown process of an
FCCU when the process unit is steamed out. During these events, high volumes of steam are sent
through the reactor and out the regenerator stack. This high volume of steam at high velocities
has been known to cause the opacity monitors to read opacity levels well above the standard. The
commenter believed these steam out periods in the FCCU shutdown process should be excluded
from the opacity standards.
Response 2: We note that high steam content in the stack gas can interfere with opacity monitor
readings, so it is unclear if the high opacity from these events is caused by increased PM
emissions or steam interference. In any event, we are not finalizing the proposed opacity limit in
favor of a minimum inlet velocity for the internal cyclones of 20 feet/sec. We expect facilities
can comply with the inlet velocity limit during periods of steam out.
Comment 3: One commenter stated that the prohibition on the use of control device bypasses
during start-up, shutdown, and malfunction, regardless of the safety risk associated with sending
292
-------
those streams to those controls, will increase the risk of fire and explosion. In particular,
operating ESPs on FCCUs during SSM has caused injuries and equipment damage.
Response 3: We specifically provided emissions limits for SS events that would allow
compliance with the emissions limitations during these events, acknowledging the fire hazards
associated with these events. We also note that we did not revise the bypass line provisions in
Refinery MACT 2 so nothing in the proposed or final rule alters the existing FCCU bypass
provisions.
10.1.3.2 O2 concentration for organic HAP
Comment 1: One commenter believed that the EPA's proposed rule requiring a minimum of 1%
O2 in the regenerator exhaust gas during startup is not appropriate or achievable because it forces
refineries to choose between emissions violations. See proposed 40 CFR 63.1565(a)(5)(ii). The
commenter stated that the proposed rule identifies a potential standard that refiners could use to
minimize CO emissions and its surrogate organic HAP emissions, by requiring the use of excess
oxygen to ensure complete combustion of CO during startup. However, by requiring refiners to
add additional oxygen to the regenerator during startup, the EPA overlooks the unintended
consequences associated with this practice. For example, while adding excess oxygen to the
regenerator may reduce CO emissions, this practice is also likely to significantly increase NOx
emissions as a result of the additional O2 reacting with nitrogen in the regenerator. As a result,
refiners will be forced to choose between violating the CO limits proposed in this rule and
violating their corresponding FCCU NOx limits.
The commenter believed that a more appropriate approach to SSM events would be to require
sources to craft a site-specific plan that details the best available work practices for balancing all
emissions during SSM. This approach would allow the EPA and refiners to determine an
achievable limit that becomes enforceable in the refinery's title V permit, while ensuring that
standards are in place continuously to address all periods during which there are emissions.
Response 1: First, we note that most commenters supported the 1% oxygen limit. Second, we
specifically selected this limit at 1% rather than a higher excess oxygen limit specifically to
allow FCCU owners or operators to still operate with relatively low levels of excess oxygen to
minimize NOx emissions during these periods. That is, based on the data we reviewed, refinery
owners and operators are expected to be able to comply with the Refinery NSPS subpart Ja NOx
limits at excess oxygen concentrations between 1 and 2 percent. We also note that the NOx
emissions limits in subpart Ja or elsewhere generally have relatively long averaging times (7-day
rolling average, in the case of NSPS subpart Ja) that would allow refinery owners or operators
ability to comply even though NOx emissions might be slightly elevated during startup or
shutdown. Finally, we find that the commenters' recommendation for a site-specific plan to be
identical to the SSM plan provisions that the courts considered illegal. Therefore, we are
retaining the 1% excess oxygen concentration limit and, as described previously, broadening it to
all FCCU startup and shutdown events.
Comment 2: One commenter stated that for FCCU which elect to comply with the alternative
organic HAP limit during periods of SS and hot standby, the proposed rule requires the site to
293
-------
also install, operate, and maintain an O2 CPMS to measure and record the oxygen content in the
catalyst regenerator vent. Most FCCUs already have oxygen analyzers in the regenerator vent as
required by the EPA's GHG Mandatory Reporting Rule and sites should be allowed to use these
analyzers and not be forced to replace them with expensive CPMS.
The commenter stated that the EPA has not provided any technical justification for requiring
sites to replace their existing FCCU regenerator vent oxygen analyzers with an O2 CPMS. There
is no basis to suggest that the current analyzer's accuracy is not adequate for Refinery MACT 2.
This is especially true given the fact that the alternative excess oxygen limit will only apply
during a very small percentage of the total operating time, i.e., only during periods of SS and hot
standby which, on average, may only occur a few hours a year.
The commenter noted that the requirement in Proposed Table 41 for the O2 CPMS is twice as
stringent as the current PS 3 requirement for O2 CEMS. It requires that an O2 CPMS must have
an accuracy of at least ± 1 percent of the range (0.25% O2), while part 60 Appendix B
Performance Specification 3, Section 13 specifies that an O2 CEMS have a calibration drift
(accuracy) of 0.5% O2.
The commenter recommended that the EPA allow sites to demonstrate compliance with the
excess oxygen alternate standard using oxygen analyzers meeting GHG requirements and that
the Table 41 accuracy specification be deleted. Recommended language was provided as follows
(shown as revisions to the proposed new 63.1565(b)(l)(iv)):
63.1565 What are my requirements for organic HAP emissions from catalytic cracking units?
(b) * * *
* * *
(iv) If you elect to comply with the alternative limit for periods of startup in paragraph (a)(5)(ii)
of this section, you must also install, operate, and maintain an oxygen monitor to continuously
measure and record the oxygen content (percent, dry basis) in the catalyst regenerator vent.
Response 2: While we understand the commenter's concern regarding the specific accuracy
requirements for the O2 CPMS, we see no reason to have no requirements at all for the O2
CPMS. While we understand that the direct O2 limit may only be applicable during periods of
startup and shutdown, we expect many refinery owners or operators currently use the
CO2/CO/O2 monitoring alternative to a flow CPMS in §63.1573(a)(2). We consider the gas
analyzers used for this purpose to be CPMS. If you elect to comply with a PM compliance
alternative that requires determination of the coke burn-off rate, then the gas analyzers used in
this case would also be considered CPMS.
We find that at least some quality control requirements are necessary for these instruments that
are used to demonstrate continuous compliance. Therefore, we are clarifying these continuous
gas analyzers are considered CPMS. We understand that many oxygen analyzers may be spanned
using air (maximum span value of 21 or 25%), so that the requirements in Table 41 for an O2
294
-------
CPMS may be more stringent than for an O2 CEMS. Therefore, we have revised the entry in
Table 41 to be "gas analyzer sensors". The gas analyzer sensors must be located to provide a
representative measurement of gas composition of the exit gas stream and the gas analyzers must
be accurate to ± 1 percent of the range of the sensor or to a nominal gas concentration of ± 0.5
percent, whichever is greater.
10.1.4 Performance testing every 5 years
Comment 1: One commenter agreed with the EPA's proposal to include new performance
testing requirements at a frequency of once every five years for existing sources. The
commenter supported an annual performance test to enhance compliance for new and existing
sources, if the EPA adjusted or eliminated the proposal to lower emission limitations and change
averaging times for new and existing sources, and the proposal to require subpart Ja monitoring
for all Refinery MACT 2 sources.
Response 1: We appreciate the commenter's support of the requirement to test FCCU at a
frequency of once every five years for existing sources. As noted in the preamble of the final
rule, we are not making any changes to the requirements to establish a 3-hour average operating
limit to demonstrate continuous compliance with the emissions limit. For most compliance
options, we determined that the 3-hour average operating limits provided adequate assurance of
compliance without more frequent compliance testing. However, as noted in the preamble to the
final rules, we are note finalizing the proposed revisions to the NSPS subpart J compliance
option in Refinery MACT 2; instead, we are adding a 20 percent opacity operating limits. For
this compliance options, we considered it necessary to include a requirement to test annually if
PM emissions are near the emissions limit during the performance test. See the preamble to the
final rule more detailed discussion of these changes.
10.2 Catalytic Reforming Units
Comment 1: One commenter argued that the limiting the applicability of the 5 psig exclusion to
prohibit active CRU purging releases to the atmosphere is not authorized by CAA section
112(d)(2) or (d)(6), significantly impacts fuel production, increases net hydrocarbon emissions
(particularly methane) and should not be finalized.
Two commenters argued that the 5 psig criterion represents the control floor for this vent and any
revision requires a CAA section 112(f)(2) or (d)(6) evaluation and determination, neither of
which was made in this proposal. Furthermore, the record for the original rulemaking identifies
this exception as applying to purging and depressuring the reactor, but does not indicate that the
depressuring and purging cannot be active. In fact, according to one commenter, it is clear that
the purpose of this provision was to allow release to the atmosphere when there was inadequate
pressure to reach control, a situation that can occur regardless of whether the purging is passive
or active. One commenter stated that the EPA has not presented any evidence to demonstrate that
the EPA's original intent was to apply the provision only to passive operations. Moreover, the
commenter was concerned that by terming this action as "clarifying," it implies that facilities
should have been complying with the interpretation all along. Clearly, the existing language does
295
-------
not provide fair notice of such an interpretation, and if the EPA finalizes its proposal, it must
affirmatively state that this is a new interpretation that applies prospectively only.
According to the commenters, the impact of this proposal is 1) to extend purge times and thereby
significantly reducing fuels production, or 2) to force sources to install facilities to compress
these streams and then supplement them with natural gas to allow sending them to combustion
controls, a wasteful action that increases net emissions and incurs substantial costs.
One commenter added that notwithstanding the EPA's mischaracterization of its action, if the
EPA proceeds to finalize its proposal, then the EPA must provide an extended period for
facilities to comply with this requirement. Section 112(i) compliance dates are inapplicable to
this proposed "clarification" or re-interpretation because such an action falls outside the types of
rulemaking procedures addressed by section 112(i) compliance schedule. Accordingly, the EPA
has broad authority to impose reasonable compliance deadlines, and the commenter requested
that the EPA provide facilities with seven years to comply. This time frame is roughly equivalent
to the typical turnaround time, and industry will need at least one turnaround to re-engineer the
CRU process. On the other hand, one commenter stated that three years is required if the
proposed revisions to §63.1566(a)(4) are finalized, and facility costs and burdens must be added
to the rule cost analyses (i.e., costs for additional compression, natural gas addition facilities, and
other facilities to allow these purge streams to be controlled) and information collection
supporting statement burden estimates. The commenter also suggested that the Agency should
specifically allow these purge streams to be flared, despite the flare gas minimization
requirements of NSPS subpart Ja and to specify that these purges are part of the base flare flow
for purposes of the NSPS subpart Ja flow RCA/CAA trigger (i.e., these flows do not count
towards the RCA/CAA flare flow trigger).
Response 1: As we noted in the preamble of the proposed rule, the 5 psig exclusion from the
emission limitations was provided based on state permit conditions, which recognized that
depressurization to an APCD is limited by the back pressure of the control system. Without
active purging, the amount of gas that can be released is limited. It is clear from the emissions
tests conducted on the CRU purge vents that the emissions limitations achieved by units using
active purging are not at all similar to the emissions limitations achieved by units that use a
series of re-pressurizations, which the industry trade organization and industry representatives
stated to be the common industry practice. HAP emissions per regeneration event using active
purging appear to be 100 times greater than the emissions that most CRU achieve. The 5
psig depressurization exclusion was never intended to allow vent streams with HAP
concentrations averaging 1,000 ppmv for an active venting period lasting over 1 hour at much
greater volumes (and therefore mass emissions) than possible for passive depressurization. Given
the common industry practice was to use a series of re-pressurizations and depressuring to a
control device and to depressure to the atmosphere only after this series of "purging" events, it is
evident that the exemption from the emissions limitations was too broad and allowed for
emissions that greatly exceed the emission limitations achieved by the best performing 12
percent of sources at the time the MACT floor was determined. Therefore, we maintain that this
correction to the exclusion provision is necessary to correct this error and to require all existing
CRU sources to achieve the emission levels consistent with the MACT floor units.
296
-------
We do recognize that the 5 psig exclusion, as worded in the original MACT standard, does
appear to allow active purging. We did not intend for our correction of this mistake to be applied
retrospectively to CRU owners or operators. We also recognize that some CRU owners or
operators will need to alter their purging procedures and/or divert these emissions to a control
device. This may require additional time to assess the control options and, if necessary,
install the systems needed to control these emissions. We consider 3 years to be adequate to
complete these activities; we do not find that 7 years are needed to start complying with these
MACT requirements.
Comment 2: One commenter stated that there are several MSS situations besides regenerator
purging, where nitrogen purges are used in CRU operations. For instance, for safety reasons, a
nitrogen purge is kept on the reformer catalyst to keep out air and moisture during unit
turnarounds, including when the regenerator vent caustic scrubber is being maintained, during
catalyst dumping operations. These high nitrogen purges are vented to the atmosphere for several
reasons: they are low pressure, they have inadequate heating value to be routed to fuel or to the
flare, there is little hydrocarbon, and control devices may be also shutdown. The commenter
stated that the rule needs to clarify that such Maintenance, Startup and Shutdown nitrogen purges
can be routed to the atmosphere without all of the burdens associated with MPVs being imposed.
One commenter similarly requested that the EPA recognize that an atmospheric discharge during
startup of a CRU remains essential also to avoid unsafe conditions. Nitrogen purging during
startup is required to assure that the parts reconnected to the reactor are air-free so that a reaction
with any pyrophoric material in the reactor is avoided. The nitrogen purge needs to discharge to
the atmosphere for a short period during startup because air in the flare lines would result in the
same fire/explosion hazard that air in the reactors would cause. This purge may contain some
VOC emissions because hydrocarbon can remain in the reactor when the catalyst is not changed
before startup. To reiterate comments above, industry does not currently control the emissions at
CRUs during startup operations; the EPA did not consider these activities in defining the affected
facility; and the Refinery MACT 2 requirements do not represent the MACT floor.
Response 2: We note that only "catalytic reformer regeneration vents" are excluded from the
definition of MPV and Refinery MACT 2 covers only the vents associated with catalytic
reformer catalyst regeneration vents. It appears that the commenters consider any vent associated
with the CRU to be subject to the Refinery MACT 2 requirements or are excluded altogether;
however, this is not the case. For cyclic or continuous reforming units, the CRU continues to
process feed, so the unit is definitely not being shut down. We recognize that semi-regenerative
CRU essentially "shutdown" in order to regenerate the catalyst, which may create some
confusion, but opening of reactor vessels or other process equipment is not needed to accomplish
catalyst regeneration, so we consider it reasonably clear in §63.1562(b)(2) that only CRU vents
associated with catalyst regeneration are specifically subject to the Refinery MACT 2 provisions.
For CRU startup and shutdowns that are not associated with catalyst regeneration or for vessel
openings for maintenance that are conducted during a regeneration cycle, these "vessel
openings" are included in the definition of "miscellaneous process vents" and must meet the
requirements for MPV. We note that we have added special startup and shutdown requirements
for equipment openings (i.e., maintenance MPV) to the MPV requirements to use 10% LEL (or
297
-------
alternative limits under specific circumstances). See Section 4.2 of this document for more detail
on these requirements.
For vents associated with catalyst regeneration, §63.1566(a)(3) makes it clear that "the emission
limitations in Tables 15 and 16 of this subpart apply to emissions from catalytic reforming unit
process vents associated with initial catalyst depressuring and catalyst purging operations that
occur prior to the coke burn-off cycle. The emission limitations in Tables 15 and 16 of this
subpart do not apply to the coke burn-off, catalyst rejuvenation, reduction or activation vents, or
to the control systems used for these vents)." We do not expect any organics to remain in the
reactor after coke burn-off and catalyst rejuvenation, so reduction purges to remove air from the
reactor vessels are not subject to the emission limitations in Tables 15 and 16.
10.3 Sulfur Recovery Units
10.3.1 Inclusion of NSPS Ja compliance option
Comment 1: One commenter argued that since this proposal incorporates the requirements from
NSPS subparts J and Ja, compliance with this rule would be equivalent to compliance with one
of the NSPS rules. It would be duplicative and wasteful for an SRP to have to demonstrate
compliance and separately report under both this rule and the NSPS rule. Therefore, the
commenter requested language be added to make compliance with this rule compliance with
NSPS subparts J or Ja, as applicable.
Response 1: We made every attempt to coordinate compliance alternatives for the NSPS and
MACT when coordination made sense. However, this is a MACT rule targeted to reduce sulfur
HAP emissions and there are some differences in the compliance requirements, e.g., for small
sulfur recovery plants with reduction systems not followed by an incinerator. Furthermore,
there are discrepancies in what emissions may be allowable under startup, shutdown, and
malfunction events. Thus, while we strived to minimize conflicting or duplicative monitoring
requirements, we maintain that there is a need for specific compliance requirements in Refinery
MACT 2 and we are retaining the requirement to report deviations from the requirements in
Refinery MACT 2 semi-annual report.
Comment 2: One commenter stated that the EPA should ensure consistency in regulation
between subpart Ja and the Refinery MACT relative to the alternative to the oxygen CEMS for
determining allowable sulfur rates. Otherwise, this subpart Ja provision would be unavailable to
refiners under the Refinery MACT, and thereby rendered a meaningless addition to subpart Ja.
Another commenter stated that the EPA acknowledges within the Proposed Rule that, during the
development of NSPS subpart Ja, the practice was common of using "oxygen-enriched air" in
Claus units to improve operational performance and reliability. The commenter argued that
the EPA therefore provided compensative equations (equation 1 within 40 CFR 62.102a(f)(l)(i)).
The Proposed Rule reflects the EPA's intent to revise the Refinery MACT to reference the same
equation 1 from subpart Ja; the commenter supported the intent of this aspect of the Proposed
Rule.
298
-------
However, the Proposed Rule does not merely incorporate into the Refinery MACT the existing
relevant equations from subpart Ja. Instead, the Proposed Rule also states that the EPA intends to
address "incomplete provisions" from subpart Ja in this regard. The Proposed Rule manifests
implementation of this statement through the proposed revisions found at §60.102a(f)(l)(i),
§60.106a(a)(5), and §60.106(a)(6). Tables 31 and 32 of subpart UUU would be similar to
§60.106a(a)(5) relative to the required installation of a CEMS to measure and record the oxygen
concentration of the enriched air. However, the commenter explained that Tables 31 and 32
should be modified to include this option to measure the flow rate of inlet oxygen and air,
consistent with subpart Ja. The commenter noted that purchased enriched air must meet exacting
specification and the oxygen content of air does not vary. Therefore, the oxygen content of the
mixture of these gases is known provided the quantity of each stream is measured and the
addition of an oxygen CEMS does not improve upon the accuracy of the overall emission
calculation. Consequently, there is no scientifically valid or environmentally beneficial
justification in deploying CEMS to measure oxygen content under the Refinery MACT and
doing so would result in significant costs which are not addressed among the economic impacts
of this rule. According to the commenter's preliminary estimates, this requirement would impose
a cost of approximately $1.5 million on a single refinery, with no attendant environmental
benefit.
Response 2: We intended to allow the oxygen enrichment provisions in subpart Ja for large
sulfur recovery plants; however, we determined that, for the purposes of HAP control, the
provisions in subpart Ja for smaller sulfur recovery plants were not appropriate. With respect to
the alternative monitoring provision to measure flow rates of air and oxygen and calculate the
oxygen concentration in the enriched air feed to the Claus burner, the omission of this alternative
in the proposed rule was an inadvertent error. As long as the flow rates of air and oxygen are
accurately monitored, the flow monitoring data provide an accurate means by which the oxygen
content of the mixed feed to the Claus burner can be determined. In the final rule, we have
revised Tables 31 and 32 to allow the flow monitoring option for determining the oxygen content
of the mixed air feed to the Claus burner.
10.3.2 Startup shutdown provisions
Comment 1: One commenter stated that the EPA's proposed SSM rule for SRUs is not
achievable. The proposed rule would allow diversion of purge gases during shutdown of a SRU
to a flare meeting the requirements of 40 CFR 63.670 (or temporarily §63.11), a measure that the
EPA believes will allow refineries to comply with the 300 ppmv reduced sulfur compound limit
in 40 CFR 60.102a(f)(l)(i) during periods of shutdown. According to the commenter, the sulfur
loading to the flare is increased significantly during periods of shutdown and proposing to
impose the 300 ppmv reduced sulfur compound emissions limit during these periods is not
achievable. According to the commenter, the only way to comply with the 300 ppmv limit would
be to completely eliminate flaring during periods of shutdown, a practice that is unachievable in
practice. The commenter has attempted to shut down its SRU without flaring on several
occasions, without success.
One commenter did not argue that the flare could not achieve the 300 ppmv reduced sulfur
compound limit, but stated that the flare may not be a viable option for the proposed SRU
299
-------
shutdown because the resulting emissions might exceed the 500 lb/day total sulfur RCA action
level. The commenter requested language be added to allow temporary MSS purges from SRPs
to be flared without including the contained sulfur in the NSPS subpart Ja RCA sulfur trigger
calculation.
The commenter also requested clarification in [§63.1568](a)(4)(iii) that the 1,200°F minimum
temperature specified is the firebox temperature, rather than the stack temperature. On the other
hand, one commenter stated that the 1,200°F minimum temperature is an excessively high
average temperature obligation for the characteristics of the emission stream to be
controlled. According to the commenter, this thermal level is consistent with the objective of
controlling emission streams with high VOC concentrations. However, emission streams
generated during the shutdown of the SRU will instead consist of elevated concentrations of H2S.
The auto-ignition temperature for H2S is 500°F, and the auto-ignition temperature for carbon
disulfide is 257°F. The commenter stated that the EPA's proposed limits were based on control
system operations during normal operating conditions. Therefore, the commenter stated
that the proposed temperature standard included in the Proposed Rule is excessive for this
application. The commenter concluded that to the extent that any alternative standard is applied
during the shutdown of an SRU, the Proposed Rule should be modified to authorize the affected
source to demonstrate compliance through emission testing of incinerators or thermal oxidizers
to reflect adequate destruction efficiency at reduced temperatures.
Response 1: First, we are unaware of any data, and the first commenter has not provided any
data, to substantiate the assertion that a properly operated flare will not effectively reduce sulfur
HAP emissions to the required limits. We recognize the difficulties associated with determining
compliance with the Refinery MACT 2 emission limits when using a flare, and therefore we are
relying on the performance indicators developed for organic HAP to ensure adequate destruction
of the sulfur HAP. Given the lower auto-ignition temperatures of the sulfur HAP as noted by the
third commenter, we consider that the flare performance indicators included in Refinery MACT
2 are adequate to ensure combustion of the sulfur HAP. The proposed rule (and as being
finalized) allows the use of a flare during startup and shutdown and compliance would be
determined based on the flare operating limits. Therefore, it is a mischaracterization of the
proposed requirements to state that the only way to comply with the proposed rule is to
completely eliminate flaring during SRU SS events.
We are not including provisions to allow flaring of SRU gases under Refinery MACT 2 to be
exempt from the RCA/CAA requirements in subpart Ja. These regulations address different
pollutants. While we consider that a flare will achieve effective control of reduced sulfur
compounds, we note that this control method will generate SO2. If a flare is used and the
emissions exceed 500 lb/day, we maintain that it is reasonable under NSPS subpart Ja for
refinery owners or operators to investigate the procedures used and to identify means to
eliminate the event or minimize SO2 emissions during such an event. Thus, while we allow the
use of flaring in the case of an SRU SS event for control of HAP, we do not intend to allow
refinery owners or operators to emit large quantities of SO2 without performing an
RCA/CAA. Also, we do not consider the RCA/CAA requirement to be so excessive that refinery
owners or operators will not use a flare when needed for SRU SS events.
300
-------
Next, we agree that the proposed 1,200°F temperature requirement refers to the firebox
temperature of the thermal oxidizer or incinerator. We do note that, if a facility elects to monitor
the temperature only at the stack, then they must meet the 1,200°F limit at the selected
monitoring location. With respect to the stringency of the 1,200°F temperature limit, we
acknowledge that this temperature is based largely on thermal oxidizer and incinerator data for
SRU controls during normal operation; however, we see no reason why the thermal control
device cannot be operated at these temperatures during the SS event of an SRU. For SRU that are
not subject to NSPS subpart J (or certain provisions of Ja), SRU owners or operators must
conduct a performance test to demonstrate compliance with the 300 ppmv sulfur HAP limit and
set operating limits based on the performance test results. It is possible that these owners or
operators could establish a lower temperature operating limit during their performance tests and
they may elect to comply with these operating limits at all times rather than use the alternative
1,200°F limit provided for periods of startup or shutdown.
Comment 2: One commenter stated that although the EPA generally removes the unlawful SSM
exemption, which it is required to remove, the EPA also proposes special standards for startup or
shutdown periods for SRU during shutdown. The commenter believed that the EPA has failed to
provide a reasoned explanation in the record for why these special standards are required, and
why it is not unlawful, arbitrary, and capricious to authorize more toxic air emissions during
these periods for these emission points.
The commenter noted that The Act requires sources to comply at all times with emission
standards. There is no exemption for any time periods. Thus, special standards for emission
points during any time periods are plainly unlawful.
Even if an alternate limit is permissible under the act, the commenter stated emissions event data
from the TCEQ demonstrates that a special standard for SRU shutdowns is not justifiable. The
commenter noted that out of 62 SRUs at refineries in Texas according to the 2011 Petroleum
Refinery ICR, between 2012 and 2013, not one facility reported any reduced sulfur HAP
emissions during shutdown periods from these SRUs. Facilities are required to report these
emissions to TCEQ pursuant to 30 Tex. Admin. Code 101.201 and .211. The lack of any
reported reduced sulfur HAP emissions demonstrates that an alternate shutdown emission limit is
not necessary for SRUs.
Additionally the commenter stated that if an alternate reduced sulfur SRU limit for refinery
shutdowns is supported by other evidence not provided in the record and is also deemed legally
permissible, the EPA must clarify that the alternate limit only applies when a shutdown is
planned and the entire facility is shutdown. First, the EPA's proposal explains that facilities are
expected to run the SRU "continuously" and only shutdown its operation during a complete
turnaround or shutdown of the facility. Yet the proposed rule does not clearly circumscribe the
limit to complete facility shutdowns. Specifically, the proposed rule states "during periods of
shutdown only, you can choose from the three options" without any further limitations or
definition of a shutdown.
Second, the commenter stated that the EPA must clearly prohibit the availability of the alternate
limit to planned shutdowns. An alternate shutdown limit that is available beyond planned
301
-------
shutdowns would create an unlawful exception to the reduced sulfur HAP limit for malfunctions
and other unplanned events. Without narrowing its availability, an alternate limit for shutdowns
would permit compliance with a weaker limit caused by poor maintenance, planning, or other
preventable event. For example, many of the cited causes for SRU shutdowns, reported to the
ICR Component 1, identify the loss of power or the loss of third-party hydrogen supply. As
discussed in the context of inherently safer technologies, facilities can eliminate or reduce the
frequency of power outages through maintaining back-up power sources. Further, facilities are in
the best position to ensure a reliable source of hydrogen through bringing those processes on site
or through the terms of their contract. The commenter believed that providing an alternate
standard for all shutdowns reduces the incentive to take these measures that would ensure
reliable refinery operations and the utilities that supply them. Therefore an alternate limit that is
not narrowly drawn to only apply during planned shutdowns is an end-run around the CAA's
mandate that standards must apply at all times and must be removed from the final rule.
The commenter emphasized that to ensure that the alternate limit is only available for planned
shutdown events, the EPA's final regulation must require facilities to submit a shutdown plan for
public notice and comment and approval by the appropriate authority. The plan should
specifically explain the reason for the shutdown and the measures the facility will take to
minimize emissions. These measures will ensure that the alternate emission limit is only utilized
when absolutely necessary and that the resulting emissions are minimized to the maximum
extent possible.
The commenter concluded that providing special standards during some time periods will
increase the health risks during those times, and the EPA has failed to consider, much less show,
how these standards could be lawful under section 112(f)(2). The EPA must prevent
unacceptable risk and assure an "ample margin of safety to protect public health," under this
provision. The EPA does not discuss these requirements, much less show, how the special
standards will meet these, when they will allow additional toxic air pollution to be released
during regular periods of startup and shutdown.
Response 2: First, we have not removed the requirement to control sulfur HAP emissions from
the SRU during periods of startup or shutdown. However, we recognize that during these
periods, emissions may be routed to an alternative control system (e.g., a flare) or the surrogate
used to assess sulfur HAP reduction during normal operations is not applicable during startup
and shutdown. For startup and shutdown, removing sulfur from the vessels is different than SRU
tail gas during normal operations and the concentration of SO2 produced from the combustion of
this shutdown gas is different than the combustion of the tail gas produced during normal
operations. During normal operations, SO2 is an indicator of sulfur recovery efficiency and high
sulfur recovery efficiency limits the amount of sulfur HAP that can be emitted. During startup
and shutdown, sulfur recovery efficiency may be highly variable and potentially meaningless
(e.g., in the case of shutdown when no feed is added to the unit). However, in Refinery MACT 2
we did not limit the applicability of the HAP requirements to Claus sulfur recovery units greater
than 20 long tons per day as was done in Refinery NSPS subpart J, because sulfur HAP control
could still be accomplished through the use of thermal oxidizers or incinerators even when sulfur
recovery efficiency was not as high as required by NSPS J/Ja. That is, in the development of the
Refinery MACT requirements, we recognized as MACT the control efficiency of thermal
302
-------
oxidizers or incinerators for SRU regardless of the overall sulfur recovery efficiency. The
primary reason to provide the set temperature and oxygen content limits during startup and
shutdown was to provide a compliance option for units that might use an SO2 CEMS during
normal operations and would not have developed incinerator operating limits based on source
test data.
We reviewed reports submitted under 30 Tex. Admin. Code section 101.201 and found
numerous instances of emission reports associated with SRU startup or shutdown. We note that
none of the reports provide sulfur HAP emission estimates, but that appears to be due to the
emission calculation procedures (i.e., accounting methods appear to consider only H2S
and SO2). As discussed previously in this response to comment document, we expect that
combustion of these gases, while potentially leading to high SO2 emissions, is effective in
reducing sulfur HAP emissions. That is, we do not consider that the alternative compliance
options provided for SRU are allowing for greater sulfur HAP emissions and therefore we
disagree with the commenter's assertion that these alternative limits are authorizing "more toxic
air emissions during these periods for these emission points."
Because these thermal devices are effective at controlling sulfur HAP emissions, we also do not
feel compelled to put further restrictions on the use of the startup/shutdown provisions to planned
shutdowns only. We find it reasonable to allow this compliance alternative during a shutdown
caused by power outages or other "unplanned" events. In fact, in response to comments received,
we are expanding the allowance for this alternative to startups as well as shutdowns. We
understand that using incineration during these events may lead to potentially high SO2 emission
releases, but they will not result in high sulfur HAP emissions, so we are not further limiting the
applicability of startup and shutdown alternatives in Refinery MACT 2 in response to these
comments. As mentioned in a previous response, owners or operators using these provisions may
have higher SO2 emissions and are still required to conduct an RCA/CAA under subpart Ja if the
SO2 emissions exceed the allowable emissions by 500 pounds per day or more, so the owner or
operator has an incentive to limit the potential for excess SO2 emissions.
Sulfur HAP are not considered cancerous so they do not contribute to maximum exposed
individual cancer risk or population cancer incidence. The highest hazard quotient for emissions
from sulfur recovery plants are several orders of magnitude less than 1, so we have no evidence
that there are adverse health effects associated with sulfur plant HAP emissions. Furthermore,
since we maintain that the SS provisions we have provided for SRU do not yield higher HAP
emissions, we conclude that further emissions reductions are not required or warranted under
CAA section 112(f)(2).
10.4 CPMS Requirements in Table 41
Comment 1: One commenter stated that the gas flow rate accuracy requirements in proposed
Table 41 are even less feasible for FCCU than they are for flare gas flow measurement. FCCU
regenerator stacks are typically larger in diameter than almost any flare header (8 to 12 foot
diameters are typical) and it is infeasible to measure 10 cfm in such large diameter stacks. Nor
are instruments available that can measure the entire flow regime with ±5% accuracy. For
303
-------
FCCUs, a typical and reasonable accuracy requirement is ±5% over the normal flow range, if an
instrument is used.
The commenter stated that because of the difficulty in measuring flow in such large ducts, the
blower speed or power is often monitored and the air flow is based on the blower design curve.
This monitoring alternative should be provided to avoid forcing the installation of questionable
flow monitors. If flow monitoring is required and the Table 41 specification is not changed, the
commenter stated that all new FCCU flow instrumentation involving multiple instruments will
be required. In addition to very high capital costs, this new instrumentation will require out-of-
sequence FCCU outages and associated production loss.103
Response 1: The accuracy requirement for air/gas flow rate is intended to apply to the normal
operating flow rate range and we do not consider that flows anywhere near 10 cfm would be
considered normal operating range for an FCCU. Additionally, we note that 40 CFR
63.1573(a)(1) and (2) contain approved alternatives to monitoring exhaust air flow rate. These
alternatives allow the use of "control room instrumentation" for determining air flow rate and
other gaseous flow rates to the regenerator. We expect that these air flow rates would be
commonly determined using blower speed/power curves or similar methods. Therefore, if blower
speed/power curves are used by the control room for determining FCCU air flow rate, Refinery
MACT 2 at 40 CFR 63.1573(a)(1) and (2) already contains the alternative requested by the
commenter. We also note that owners or operators have the ability to submit a request for an
AMP on a case-by-case basis, if needed. Based on the provisions already included in Refinery
MACT 2, we disagree with the commenter's assertion that the proposed requirements will have
high costs and/or result in production losses.
Comment 2: One commenter recommended that the EPA should not reference NSPS subpart Ja
in Table 41 for coke burn monitors. The commenter stated that in Table 41, the EPA proposed to
require that CO2, O2, and CO monitors for coke burn-off rate meet the requirements of 40 CFR
60.105a(b)(2). Yet, Table 41 already contains requirements for CO and O2 monitors and there
are differences between the two sets of requirements, including a significant increase in the
frequency of performance testing the monitors. Furthermore, it is unclear that existing monitors
will meet the new requirements and, thus, some may have to be replaced.
The commenter stated that since the EPA provides no justification for imposing NSPS
requirements on existing units and the new requirements conflict with existing requirements, the
changes should not be finalized. The commenter recommended that if the EPA believes these
changes are necessary, they develop a justification including cost estimates for any required
replacements, provide three years compliance time, and publish their justification for comment.
103 FCCU outages are massive undertakings, involving years of planning, scheduling large cranes and armies of
workers. These outages require extensive scheduling and lead times to redirect feeds and provide for replacement
product supplies and usually many other process units must be shutdown coincidently because of the loss of the
FCCU steam production. Every such shutdown also risks equipment damage due to the thermal cycling of the
equipment.
304
-------
Response 2: In the proposed rule, the CO2, O2 and CO monitor requirements for coke burn-off
were listed in Table 40 (CEMS table) rather than Table 41 (CPMS table). This was an
inadvertent error on our part as we consider these systems to be control device CPMS rather
than CEMS. As such, these requirements should have been included in Table 41 rather than in
Table 40. Therefore, we are moving these requirements to Table 41 in the final rule and we are
clarifying that all concentrations need to be determined on a dry basis. We are allowing facilities
that follow the requirements in 40 CFR 60.105a(b)(2) to use those requirements in order to
harmonize the requirements for units subject to Refinery NSPS subpart Ja, but we are not
requiring sources to meet the NSPS performance specifications for these monitors for existing
FCCU (i.e., units not subject to Refinery NSPS subpart Ja).
10.5 General provisions applicability (Table 44)
Comment 1: One commenter stated that there are discrepancies in the proposed QA/QC
requirements set forth in proposed Tables 40 and 44. Table 40 in subpart UUU references part 60
Appendix F in several places both directly and indirectly. Table 44, meanwhile, references 40
CFR 63.8(c)(7) as applicable to all sources subject to subpart UUU. However, section 63.8(c)(7)
and part 60 Appendix F define "out of control" (OOC) periods differently. Appendix F
references back to the last good validation to start the OOC period, whereas section 63.8(c)(7)
starts the OOC period beginning with a "bad" validation. The commenter requested the EPA
resolve this discrepancy in the regulations, perhaps by specifying which OOC definition should
be used in Table 44.
Response 1: We appreciate the commenter pointing out the discrepancy in the QA/QC
requirements. We reviewed the requirements and when owners or operators are required to
comply with part 60 for NSPS subpart J/Ja, it makes sense to require owners or operators to use
the QA/QC procedures in Appendix F to part 60 so they have a consistent set of QA/QC
requirements between the NSPS and Refinery MACT 2. In order to clarify this in the final rule,
we have added a note to the entry for section 63.8(c)(7) in Table 44 to indicate "Except when
subpart UUU specifies use of 40 CFR part 60, Appendix F, out of control periods are to be
defined as specified in part 60 Appendix F "
305
-------
11.0 General Compliance Requirements for MACT standards
Comment 1: Commenters raised concern with the structure of the rule, noting that compliance
requirements for Refinery MACT 1 are indicated in Table 11, whereas recordkeeping and
reporting requirements are indicated in section 63.655, and they suggested revisions to the
regulatory structure. Other commenters noted that the revisions proposed are so comprehensive
that compliance dates can only be clarified by establishing new subparts (i.e., subparts CCa and
UUUa)..
Response 1: We appreciate the commenters' suggestions on how to clarify the rule requirements.
We have considered these comments as we developed the final rules to make sure that they are
clear as to who has to comply with which provisions and by when. We added Table 11 to
Refinery MACT 1 (subpart CC) because there are several dates that the owners or operators must
consider in determining when construction/ reconstruction occurred, and therefore, by when they
must comply with applicable rule requirements. However, since we are amending the subpart CC
rule already in place, we retained the structure of section 63.655, which indicates when the
NOCS Reports, Periodic Reports and Other Reports are due.
Comment 2: Some commenters argued that the compliance deadlines for many of the proposed
requirements are simply not feasible and significantly more time will be required. One
commenter stated that 10 years are needed for changes requiring flare outages. One commenter
stated that a process to grant additional compliance extensions should be in place to handle
delays that may occur due to reasons outside of the control of a refinery. Another commenter
stated that the EPA has not met the test under CAA section 112(i)(3) showing that the 3-year
compliance period for all of the existing source standards being proposed under CAA section
112(d) [79 FR 36950-51] is "as expeditious as practicable," and that the EPA only has included
conclusory statements about the proposed compliance time frames, unsupported by evidence. :
This commenter supported the compliance date the EPA proposed for the new storage vessel
standards, claiming that the EPA appropriately recognized that it does not have authority to
extend this compliance date beyond 90 days, as section 112(f) provides. [See^i? v. EPA, 716
F.3d 667, 672 (D.C. Cir. 2013); 79 FR at 36,950.] In support, the commenter quoted section
112(f)(3) ("[a]ny emission standard established pursuant to this subsection [section 112(f)] shall
become effective upon promulgation)" and section 112(f)(4) establishes a prohibition on
violating the new standard, stating that ("[n]o air pollutant to which a standard under this
subsection applies may be emitted from any stationary source in violation of such standard,
except that in the case of an existing source . . . such standard shall not apply until 90 days after
its effective date.)"
Response 2: As explained in the proposal preamble (see 79 FR36950, June 30, 2014),
amendments to Refinery MACT 1 and 2 for adoption under CAA section 112(d)(2) and (3) and
112(d)(6) are subject to the compliance deadlines outlined in the CAA under section 112(i). For
existing sources, CAA section 112(i) provides that the compliance date shall be as expeditiously
as practicable, but no later than 3 years after the effective date of the standard. For new sources,
compliance is required by the effective date of the final amendments or upon startup, whichever
is later. We are finalizing the compliance dates, as proposed, for the reasons given at proposal,
with the exceptions of the fenceline monitoring standard, for which we are requiring the owners
306
-------
or operators to have the monitors in place within 2 years of the effective date of the rule instead
of 3 years, as proposed, and the work practice standards for PRDs and emergency flaring, as
explained in the final rule preamble and earlier in this document. In addition, as proposed, we are
retaining the requirement that storage vessels comply with the requirements in this final rule 90
days from the effective date of this rule since those requirements are being finalized under
authority of both CAA section 112(d)(6) and 112(f)(2). As provided in CAA section 112(f)(4),
risk standards shall not apply to existing sources until 90 days after the effective date of the rule,
and we are not granting the waiver for up to 2 years, as explained in the proposal preamble.
However, we disagree with the commenter's statement that the EPA does not have the authority
to extend this compliance date beyond 90 days.
Comment 3: One commenter stated that 18 months to 3 years is required to implement new and
revised monitoring requirements. The commenter stated that where new, replacement, relocated
or significantly revised instrumentation is required to meet the new requirements, 3 years is
necessary. Where existing instrumentation is able to meet the new requirements with little
revision, 18 months, at a minimum, is needed. Commenters also stated that 18 months will be
needed to put most of the new recordkeeping and reporting systems in place, to develop the
required monitoring plans, revise procedures, obtain approval for AMP, to implement
maintenance procedures, QA/QC and other requirements specified, and to obtain permits and
approvals for the changes.
Response 3: Where new monitoring requirements were added to the rule, we included time for
the owners or operators to implement the new monitoring requirements. Therefore, the refinery
owner or operator would have time to implement any recordkeeping and reporting requirements
associated with that monitoring system (and/or to request an AMP) because the recordkeeping
and reporting requirements would not apply until the monitoring system is in place. In general,
where we significantly revised the monitoring requirements, particularly when new operating
limits need to be established, we did not propose to require compliance within 90 days of the
effective date. Instead, for new monitoring requirements that require sources to develop
monitoring plans, develop specifications for the CPMS, receive and analyze vendor quotes,
select and schedule CPMS installation, and test the monitoring system (and software), we have
provided 18 months for compliance. This should provide the facilities ample time to work
through any state or local permits, although owners or operators should reach out to their
permitting authorities as soon as possible. Under the federal operation permit program under
Title V of the CAA, these final MACT requirements would be added to their operation permit as
part of their five-year permit renewal, but compliance on or before the dates specified in the final
rule must be met without regard to when the Title V permit is reopened. Other CPMS
requirements generally allow 90 days from the effective date of this final rule, but in most of
those cases, the added CPMS requirements follow generally established manufacturers'
operating specifications, and should not require the longer timeframe.
Comment 4: Several commenters expressed concern that the rule requires refiners to comply
with revised SSM and some other requirements as soon as the rule becomes effective. The
commenters stated that EPA provides no justification for making such changes in the existing
regulations effective immediately. These commenters argued that it would be arbitrary and
capricious to require compliance immediately, when CAA section 112(i) allows a compliance
307
-------
deadline of up to three years. The commenters stated that elimination of the SSM provisions of
the existing rules and the prohibition on atmospheric release of PRDs will require plants to make
significant changes to their facilities and procedures, which would take significant time to
execute and to permit. The commenters more specifically identified four actions that would need
to be taken prior to implementation: 1) extensive permit revisions; 2) identification and
evaluation of hundreds of new emission sources; 3) extensive procedural changes in light of the
revised SSM requirements, including all new equipment preparation procedures; and 4) revision
of all monitoring procedures. The commenters stated that while some activities can be started
based on the proposal, procedures cannot be finalized, permit revisions applied for, and
retraining of site personnel cannot begin until the rule is final. Furthermore, the commenter
stated changes that require permit revisions and OMMP revisions cannot legally be implemented
until the revisions are approved. One commenter stated the burden will be heavier for small
business refiners due to smaller staffs and less capability to adjust work priorities. Conversely,
another commenter opposed what they referred to as a 3-year compliance delay that applies to
the EPA's new prohibition on currently uncontrolled emissions from pressure relief devices and
bypass lines. The commenter stated that all uncontrolled emissions of this kind are unlawful and
allowing them to continue for three additional years is equivalent to allowing the unlawful
malfunction exemption to continue for another 3 years. For the same reasons that the EPA must
require and assure that emission standards apply "at all times," it may not allow an exemption to
continue for an additional three years, such that they will apply at no time until three years from
the final rule date. See Sierra Club v. EPA, 551 F.3d at 1028.
Response 4: First of all, although the proposed rule eliminated all exemptions associated with
SSM events, separate standards were proposed for startup and shutdown conditions associated
with certain Refinery MACT 2 vents, such as the FCCU and the SRU. In those cases, we
proposed immediate compliance upon the effective date of this final rule. For all other situations,
we did not establish separate standards during periods of startup and shutdown and required
immediate compliance with the MACT standards upon the effective date of the final rule. The
commenter is incorrect in stating that we proposed a 3-year delay for PRDs. In fact, we proposed
that the removal of the malfunction exemption was immediate; however, we did propose
additional time necessary for the refinery owners and operators to install equipment, as
necessary, sufficient to detect a release from a PRD. As discussed later in this section, although
we do not interpret the Sierra Club decision on SSM to require the EPA to set separate standards
during periods of malfunction, and in most cases, we are not able to do so and do not deem it
appropriate or necessary, we do not interpret the decision to preclude the EPA from doing so if
sufficient information is available and separate standards are appropriate. For this final rule, we
are setting work practice standards for PRD releases and emergency flaring. As explained
previously, we are doing so because we have a somewhat unique situation with this source
category in that we collected a substantial amount of information via our comprehensive 2011
Refinery ICR, information submitted by commenters during the comment period and precedents
for standards for these types of events issued by SCAQMD and California Air Resources Board
that provided us information on PRDs and emergency flaring such that we are able to estimate
what the best performing facilities are achieving. As explained further in the preamble, we are
providing 3 years for facilities to comply with these new standards.
308
-------
We disagree with the commenters that the SSM requirements should not be immediately
effective. Where we determined that no separate standards are necessary for periods of SSM,
owners or operators should already be in compliance with the applicable MACT standards under
subparts CC and UUU and are required to comply with these standards at all times. In addition,
for malfunctions other than those associated with PRD releases and emergency flaring,
accounting for the myriad of potential malfunctions within a refinery would be difficult, if not
impossible, and the courts have generally given the EPA wide latitude in determining the extent
of data gathering necessary to solve a problem and nothing in section 112 requires the EPA to
consider malfunctions in establishing standards. The alternate work practice standards for certain
startup and shutdown situations and the work practices standards for PRDs and emergency
flaring in this final rule should resolve most of the other issues raised by the commenters
regarding compliance timing (see further discussion at 79 FR 36944-36945, June 30, 2014).
11.1 Removal of SSM plan/exemptions
Comment 1: Several commenters seemed to suggest that the decision in Sierra Club was wrong
or does not apply to the Refinery MACT rules. Some commenters argued that the Sierra Club
decision interpreted the NESHAPs General Provisions, but did not address what the EPA may or
may not include in category-specific MACT standards, such as Refinery MACT 1 and 2. The
commenters noted that in reviewing source-category-specific MACT standards, the court in
Sierra Club emphasized the need for those standards to recognize and accommodate higher
emission levels that occur at times other than normal operations.
Commenters stated that when Congress enacted the "continuous basis" language in section
302(k) in 1977 the EPA's emissions standards under section 111 exempted SSM periods. The
commenters argued that there is nothing in the legislative history of the 1977 amendments to the
CAA that suggests Congress intended to overturn that practice and instead the history suggests
the language was meant to address intermittent controls. Moreover, the commenters noted that
court decisions both before and after the CAA Amendments of 1977 affirmed the
appropriateness of including special SSM provisions in standards issued under section 111
despite the "continuous basis" language in the definition of "emission limitation." The
commenters stated that there is nothing in the legislative history of the Clean Air Act
Amendments of 1990 that suggests Congress meant something completely different when it used
the same defined terms, "emission standard" and "emission limitation," in directing the EPA to
establish MACT standards.
Response 1: We disagree with the commenter's suggestion that Sierra Club v. EPA, 551 F.3d
1019 (D.C. Cir. 2008) is not relevant because it addressed the SSM exemption in the General
Provisions rather than source category-specific MACT standards. The holding in Sierra Club
that emissions limitations under section 112 must apply continuously and meet minimum
stringency requirements, even during periods of startup, shutdown and malfunction, is clearly
applicable to source category-specific MACT standards. The decision was based on the
definition of emission limitation in section 302(k) and the commenters have not provided and we
see no basis for interpreting that term differently for the refinery MACT source category than for
source categories which cross-reference the General Provisions.
309
-------
The EPA disagrees with the commenter's suggestion that the existence of an SSM exemption in
rules implementing section 111 in 1977 when Congress enacted the "continuous basis" language
in the definition of "emission standard" is evidence that Congress approved of that regulatory
SSM exemption. We do not believe the legislative history cited supports that Congress was
aware of or approved that exemption. Regardless, the legislative history does not alter or trump
the court's decision in Sierra Club regarding the meaning of that term.
Comment 2: Multiple commenters stated that the EPA has misinterpreted the Sierra Club
decision and that the decision did not say that the same emission limits the EPA has derived for
normal operations must also apply during SSM events. Some commenters contended that while a
blanket, open-ended exemption from any standard under section 112 is inconsistent with that
decision, the decision does not preclude the EPA from applying different standards during SSM
events than apply during normal operations. Commenters further noted that the court
acknowledges that the broad phrase "any requirement relating to the operation or maintenance of
a source to assure continuous emission reduction" in the definition of "emission standard"
suggests that the EPA can establish MACT standards consistent with CAA section 112 "without
necessarily continuously applying a single standard." Commenters claimed that the implication
of Sierra Club is that alternative SSM standards, including the inapplicability during SSM events
of otherwise applicable MACT emission standards, can be authorized when based upon
consideration of the relevant MACT criteria to specific source types.
Commenters stated that the Sierra Club decision did not address whether the EPA could use a
"design, equipment, work practice, or operational standard," as authorized under CAA section
112(h) and included in the definition of "emission limitation" and "emission standard" in CAA
section 302(k), in lieu of a numerical emission limitation during SSM events. One commenter
noted that the decision stands for the simple proposition that emission limits under section 112
must be apply during all periods of operation but that the emission limit may be numerical or,
where the statutory criteria are met, a work practice standard.
Several commenters stated that there is ample precedent for the EPA applying a different
standard during SSM events. In particular, the commenters claimed that the EPA has not
required sources to meet NSPS emission limitations established for normal operations during
SSM events.
Commenters claimed that the Agency expressly concluded that the SSM provisions of the
Refinery MACT were not affected by the Sierra Club decision and referred to the Letter from
Adam M. Kushner, Director, Office of Civil Enforcement for the Agency, to American
Chemistry Counsel, et al, July 22, 2009. The commenter stated that, even if the EPA's current
position with respect to the scope of the holding in Sierra Club had merit, the Agency must
justify the legal basis for its change in policy in the context of this rulemaking proceeding.
Response 2: To the extent that the commenters are suggesting that EPA took the position in the
proposed rule that section 302(k) and the Sierra Club decision bar EPA from establishing
alternative limits for periods of SSM, we disagree. In fact, in the proposed rule, where we
determined it was appropriate, we proposed specific alternative emission limitations that would
apply during certain startup and shutdown events. As explained in the preamble to the final rule,
310
-------
we have included alternative emissions limitations for startup and shutdown for additional
emission sources beyond those in the proposal where we have determined that the standard that
applies during normal operations cannot be met and where the startup and shutdown standards
are achievable by the best performing units.
In the preamble to the proposal, the EPA explained that it was not proposing alternative limits to
apply during periods of malfunction because it was not required to do so under CAA section
112(d)(2) & (3) and because of the difficulties in identifying the myriad types of malfunctions
and in predicting the frequency, degree and duration of such events. 79 FR at 36944. While we
continue to believe that is true for malfunctions in general, for the reasons provided in the
preamble to the final rule, we are establishing emission limitations for PRD releases to the
atmosphere and emergency flaring events.
The Adam Kushner Letter referred to by the commenters identified rules that were directly
affected by the court's decision in Sierra Club because those rules only cross-referenced the
SSM exemption in the General Provisions and did not contain any category-specific standards or
exemptions for periods of SSM. Thus, for those rules, once the mandate issued in Sierra Club,
the SSM exemption in §§63.6(f)(1) and 63.6(h)(1) was rendered null and void. Thus, at that time,
there was no longer an SSM exemption available for standards which incorporated those
provisions by reference rather than including separate language establishing an SSM exemption.
The Adam Kushner Letter also identified rules, such as the two Refinery MACT standards that
included specific language establishing an SSM exemption (see 63.655(g)(6)(iii) and 63.1560(g)
for subparts CC and UUU, respectively) instead of only cross-referencing the General
Provisions. For these rules, the court's mandate did not have the effect of rendering the SSM
exemption null and void. EPA has not changed its position regarding the effect of the Sierra
Club decision. Rather, EPA is simply amending the Refinery MACT rules to remove language
that the court has found is inconsistent with the definition of emission limitation in CAA section
302(k).
Comment 3: Several commenters stated that the proposal does not account for the differences
between normal emissions and maintenance, startup and shutdown emissions and that the EPA
must revise many existing floor determinations to include maintenance, startup and shutdown
emissions or set separate maintenance, startup and shutdown standards. Commenters stated that
at the time the MACT standards were promulgated, startup, shutdown and malfunction activities
were exempt from the standard and not included in EPA's evaluation. Commenters asserted that
the EPA codified the affected source definitions with the understanding that SSM was an activity
separate from the "affected source" for which it set MACT standards. Commenters asserted that
the EPA must: 1) explain why its original definitions of activities included in the affected source
definition are no longer technically valid and new activities should be included in the defined
affected source; and 2) determine how this revised scope of the activities included within the
affected source definition affects its original MACT floor finding. Commenters suggested that
facilities minimize emissions differently during startup and shutdown and that it is likely that
including maintenance, startup and shutdown activities under the standard would lower the
MACT floor for the affected sources, or require a change in the averaging period to account for
emission fluctuations. Commenters cited NRDC v. EPA, 859 F.2d 156, 210 (D.C. Cir. 1988) and
argued that the default assumption must be that "special" provisions are needed for maintenance,
311
-------
startup and shutdown and EPA cannot conclude that special provisions for these emissions are
not needed based on "mere speculation." Commenters stated that the EPA previously
determined, when establishing the existing standards applicable to refineries, that the best
performers on which the MACT standards were based may not achieve those standards during
MSS and that the EPA cannot change its mind without providing a factual analysis supporting
the conclusion that MACT standards can be achieved as well during all MSS periods. [See, e.g.,
Transactive Corp. v. United States, 91 F.3d 232, 237 (D.C. Cir. 1996).] Commenters claimed
EPA must conduct a thorough analysis for maintenance, startup and shutdown and justify
applying the existing emission standards under CAA section 112(d)(2) and (3) or instead develop
an alternative numerical emission standard or 112(h) work practice during maintenance, startup
and shutdown.
Commenters took exception to the EPA's assertion that, "[w]e expect facilities can meet nearly
all of the emission standards in Refinery MACT 1 and 2 during startup and shutdown, including
the amendments we are proposing in this action," or that, for process vents and transfer
operations, it is common practice to start an APCD prior to startup. Commenters stated that there
are numerous reasons why during startup and shutdown a source might not be able to comply
with emission limitations established based on performance during steady-state operation, even if
the control devices used are started up before the process units and are operational during the
shutdown phase of a process. Commenters gave the example of a control that is less efficient
until it reaches its design operating temperature, or is less efficient when the pollutant
concentrations in the gases to be treated are lower than during steady-state operation. See, e.g.,
68 FR1276 at 1287-88 (Jan. 9, 2003). One commenter stated that until a manufacturing process
reaches steady-state operation, that process may generate substantially higher emissions, either
on a total-mass basis or on a mass-per-unit-of-production basis. See, e.g., 76 FR 63878, 63883
col. 2 (Oct. 14, 2011). Commenters claimed if flammable gases are involved, routing the gases to
a thermal destruction device before the concentration of the flammable compounds in the vent
gas stream has exceeded the Upper Explosive Limit can result in an explosion.
Response 3: As explained previously, in removing the SSM exemptions in the Refinery MACT
standards, we evaluated whether separate standards were necessary during periods of SSM, as
well as during maintenance activities. As such, we have included alternative emissions
limitations for startup and shutdown for additional emission sources beyond those in the proposal
where we have determined that the standard that applies during normal operations cannot be met
and where the startup and shutdown standards are achievable by the best performing units. In
addition, as discussed previously, we are also establishing work practice standards for PRDs and
emergency flaring, based on the best performing units. Also, we are finalizing requirements for
opening process equipment to the atmosphere during maintenance events after draining and
purging to a closed system. As explained further in the final rule preamble, these requirements
will ensure that the maximum amount of material in the process equipment is sent to control
before the equipment is opened for maintenance, consistent with practices at the best performing
units. Regarding malfunction emissions, as provided in the preamble and elsewhere in this
response to comments, exempting periods of malfunction from complying with emissions
standards is not consistent with CAA section 302(k). We do not consider malfunction emissions
in setting MACT standards, which are based on the best performers whether or not the affected
source is defined as included periods of malfunction.
312
-------
Comment 4: Several commenters supported the proposal to provide alternative emission limits
for certain startup and shutdown activities at petroleum refineries. Another commenter stated that
the EPA has offered no basis for the statement that control devices will be operating normally for
certain startup and shutdown activities. First, the commenter raised a concern that thermal
destruction control devices, such as catalytic oxidizers, often need to achieve a certain
temperature before optimal destruction occurs, and that specific provisions for startup might be
required for that type of equipment. Second, the commenter claimed that absorbers may need to
be operated at different liquid/gas ratios during times of startup or shutdown where gas flow
rates can vary and the commenter recommended that the final rule allow for the use of
engineering calculations to establish a different liquid/gas limit during start-up and shutdown if
the calculations demonstrate that the standard applicable during normal operations cannot be
met.
Some commenters urged the EPA to allow a process for companies to apply for case-by-case
limits to be approved by either the EPA or delegated States. The commenters suggested that
advancements in technology and process improvements may occur in the future as well as the
possibility that not all situations may be readily foreseen at this time. The commenters suggested
a process similar to the case-by-case alternate NOx limits for process heaters operated under
certain conditions as provided in Ja. Commenters noted that small refiners' equipment may be
more constrained in flexibility due to its smaller size, and thus these requirements may pose a
larger burden on small refiners.
Response 4: While we agree with the general proposition that sources may not be able to meet
the limit applicable during normal operations during some startup and shutdown activities, we
disagree that this is the case for the examples provided by the commenter. Thermal oxidizers can
and should be brought up to temperature using natural gas fuel (or other non-HAP containing
fuel) prior to being used as a HAP control device. There is no technical reason why the thermal
oxidizer cannot be brought up to temperature prior to the startup of the unit that requires HAP
control. With respect to absorbers, these units generally have a liquid-to-gas ratio that they
cannot fall below. During startup, we anticipate that the gas flow rate will be small, so a normal
(or even reduced) liquid injection/circulation rate would allow compliance with the operating
limit. Thus, we do not believe it is necessary to allow "engineering calculations" as suggested by
the commenter. We further note that we are concerned with such an approach since such
calculations cannot easily be verified. Absorbers may also have a pH requirement, which again
can be established and maintained during the startup of the absorber prior to introduction of the
gas stream requiring control.
As discussed previously in this document in chapter 7, there is already a mechanism in place by
which refinery owners or operators can request an alternative means of emissions limitation if
after notice and opportunity for public hearing, the owner or operator establishes to the
satisfaction of the Administrator that an alternative means of emission limitation will achieve a
reduction in emissions of any air pollutant at least equivalent to the reduction in emissions of
such air pollutant achieved under the MACT standards. It is unclear that there is a circumstance
in which it would be appropriate to set an alternative to a MACT standard for a single source
since MACT is set at a prescribed level under CAA section 112(d)(2) and (3). We anticipate that
the reason a source might wish to request an alternative limit is because for any number of
313
-------
reasons, they prefer to meet a less prescriptive requirement and not because they prefer to meet
an equivalent or more prescriptive requirement. We do not agree that it is appropriate to establish
a mechanism whereby small sources can request less prescriptive alternatives to the MACT limit.
Again, the mechanism for prescribing the MACT level is set by statute and there is no exception
for small sources.
We addressed the analogy to the NOx limits in Refinery NSPS Ja in chapter 10 of this document.
Comment 5: Commenters stated that the EPA needs to conduct a MACT floor analysis and
when the EPA does so it is likely to conclude that section 112(h) work practice standards
represent the only reasonable approach for regulating startup and shutdown activities. The
commenter stated that a work practice is appropriate for startup and shutdown activities because
it is not possible to safely measure compliance with numerical standards for startup and
shutdown activities. Specifically, commenters asserted, performance testing is required to show
compliance with applicable emissions standards if a continuous emissions monitor is not
installed. The commenters alleged that because the process units are not at a steady state during
startup and shutdown, it is not safe to perform a performance test. Commenters noted that that
the applicable regulations generally prohibit testing during startup and shutdown and require that
startup and shutdown data not be used for compliance purposes.
Response 5: We disagree with the commenter that a one-size-fits-all work practice standard
approach should apply to startup and shutdown for all emission sources. For each separate
emission source addressed in this rulemaking, we have evaluated and explained the basis for the
emissions limitation that applies for startup and shutdown.
Comment 6: Commenters stated that if the standard applies continuously, then the EPA must
evaluate emissions on a continuous basis, including emissions during malfunctions. The
commenters charged that applying the same emission standards during malfunctions is not
compelled by the statute or by applicable case law.
Commenters objected to the EPA assertions that "CAA section 112 does not require that
emissions that occur during periods of malfunction be factored into development of CAA section
112 standards" and that "[tjhere is nothing in CAA section 112 that directs the agency to
consider malfunctions in determining the level 'achieved' by the best performing or best
controlled sources when setting emission standards." The commenters countered that nothing in
CAA section 112 allows the EPA to ignore malfunctions and set MACT standards based on a
level of emissions that even best-performing sources only achieve part of the time. The
commenters stated that even the best performing units in the source categories covered by the
proposal are subject to a wide variety of potential malfunctions (e.g., power failures, equipment
breakdowns) and that the EPA cannot rationally defend the view that applying the concept of
"best performing" is inconsistent with a source experiencing a malfunction. The commenters
argued that the EPA cannot ignore the requirement that MACT floor standards reflect
performance actually achieved. Some commenters contended that EPA is going beyond the
MACT floor by proposing standards that even the best performing sources cannot achieve part of
the time and that EPA did not do a beyond-the-floor analysis.
314
-------
Other commenters stated that the absence of a reference to specific operating conditions in CAA
section 112 means that all operating conditions must be considered in setting the standards.
Commenters noted that the case cited by EPA for recognizing that the Agency has discretion in
setting MACT standards under section 112 makes clear that EPA's prior interpretation has been
to consider all operating conditions in setting emission standards. See Nat'/ Ass 'n of Clean Water
Agencies v. EPA, 734 F.3d 1115, 1143 (D.C. Cir. 2013) ("EPA believes that it must set MACT
floors 'that the best performing sources can meet' every day and under all operating
conditions..." (quoting 75 FR 63269 (emphasis added)).
Commenters argued that the courts have long recognized that a "technology based standard
discards its fundamental premise when it ignores the limits inherent in technology." [See NRDC
v. EPA, 859 F.2d 156, 208 (D.C. Cir. 1988] The commenters noted that the D.C. Circuit
recognized, in Portland Cement Ass'n v. Ruckelshaus, 486 F.2d 375, 398 (D.C. Cir. 1973), that
"start-up" and "upset" conditions due to plant or emission device malfunction, is an inescapable
aspect of industrial life and that allowance must be made for such factors in the standards that are
promulgated." Id. at 399. The commenters also cited Essex Chem. Corp. v. Ruckelshaus, 486
F.2d 427, 433 (D.C. Cir. 1973), cert, denied, 416 U.S. 969 (1974) in which the court held that
SSM provisions are "necessary to preserve the reasonableness of the standards as a whole."
Finally, the commenters cited National Lime Ass 'n v. EPA, 627 F.2d 416, 431 n. 46 (D.C. Cir.
1980) in which the court held that the CAA requirement that NSPS be "achievable" means that
the standards must be capable of being met "on a regular basis," including "under most adverse
circumstances which can reasonably be expected to recur," including during periods of SSM. In
addition to citing CAA cases addressing periods of SSM, the commenters cited similar cases
under the Clean Water Act.
Commenters stated that the only decision that has dealt directly with how the EPA should
address SSM issues in setting section 112 standards is consistent with these cases. The
commenters claimed that in Cement Kiln Recycling Coalition v. EPA, 255 F.3d 855, 872 (D.C.
Cir. 2001) the court vacated MACT standards in part because of concerns about the EPA's
failure to exempt hazardous waste combustors from numerical emission limits during SSM
periods and because the Court had "doubts about EPA's decision to require sources to comply
with standards even during openings of emergency safety valves caused by events beyond the
sources' control." The commenter noted that in response to that decision, the EPA revised the
rule to exempt facilities from the limitations during SSM events. See 40 CFR 68.1206(b)(1); 67
FR 6792, 6798, 6813 (February 13, 2002).
Commenters argued that the EPA's decision to disregard emissions during malfunction periods
in the proposal is unreasonable in light of NRDC v. EPA, 749 F.3d 1055 (D.C. Cir. 2014) in
which the court vacated the EPA's inclusion of an affirmative defense to emission violations
during periods of malfunction. The commenters cited to a March 2014 rulemaking, 79 FR
17,340, 17,347 (Mar. 27, 2014), in which EPA explained the importance of the affirmative
defense in the absence of an SSM exemption. The commenter noted that while the EPA
questioned whether the cases the Agency relied on remained legally binding, the EPA noted that
decisions "support the EPA's view that a system that incorporates some level of flexibility is
reasonable and appropriate."
315
-------
Response 6: Commenters assert that the EPA has failed to develop standards that are "achieved
in practice" if the best performing sources are expected to have malfunctions, and those
malfunctions result in emissions that exceed the standard. As an initial matter, the commenters
appear to concede that the term "achieved" as used in section 112(d) is ambiguous. They make
an argument that the statute as written clearly requires EPA to consider malfunctions in
establishing MACT standards or to set separate standards for malfunctions. In fact, the terms
"malfunction" is not used at all in section 112(d); rather that provision, when read in its entirety
focuses on setting standards that reflect "best performing sources," "maximum degree of
reduction," and "best controlled similar source."
We disagree that the term "achievable" or "achieved in practice" must be read to mean that the
standard must be capable of being met under any circumstance, such as during a malfunction.
The argument that this term means that EPA must account for any type of malfunction in setting
a MACT standard goes too far. Malfunctions of widely ranging severity and due to a wide
variety of causes and can occur at any facility. The fact that these events may occur at some time
during the life of a facility does not mean that a standard based on the operation of a best
performer is "unachievable" simply because it does not reflect emissions levels that can occur
during these events as a result of a lack of control. Hurricanes and malfeasance can occur at well-
maintained and well-managed sources and can cause upset conditions that result in violations of
emission standards, but this does not warrant factoring such unpredictable events into revised
emission standards. Even if malfunctions were inevitable for all sources, including the best-
performing sources, that does not make it possible to take them into account when establishing
MACT emission standards, because they are still unknown in frequency, length, magnitude and,
most importantly, effect on emission levels.
Rather, we interpret "achievable" as used in section 112 to mean that when a source is operating
as it should that the standard is capable of being achieved. The fact that there may be a
malfunction or other event that may cause an exceedance of the standard does not transform a
standard that is achievable by a best performing source into unachievable standard. Under such
reasoning, every MACT standard is potentially invalid because it does not reflect emission levels
that may result from any possible event that might occur. We continue to take the position that it
is reasonable to interpret section 112(d) as not requiring that standards take into consideration
such unpredictable events because EPA cannot anticipate the frequency, length, magnitude and,
most importantly, effect on emission levels. The EPA's approach both accounts for variability
associated with a reasonably foreseeable range of operating conditions and recognizes that
enforcement mechanisms can address emission exceedances due to unpreventable equipment or
process failures. While commenters may seek greater accommodation for malfunctions, such
accommodation is not compelled by the Act.
Commenters cite to several cases to support their argument that EPA must consider malfunction
emissions when setting standards under section 112. We disagree that those cases support an
argument that EPA must consider undefined "malfunctions" in setting standards. Rather, those
cases focused on more predictable events, such as startups and shutdowns and do not stand for
the broad proposition advocated by the commenters. For example, the National Lime decision
relied on by the commenters simply provides that the standards must be capable of being met "on
a regular basis," including "under most adverse circumstances which can reasonably be expected
316
-------
to recur." Yet, the types of events that the commenters are suggesting must be addressed go well
beyond events that happen "regularly" or that are likely to "recur." Other cases, such as
Weyerhaeuser Co. v. Costle, 590 F.2d 1011, 1057-58 (D.C. Cir. 1978) upheld case-by-case
enforcement discretion approach to addressing malfunctions because interjecting factors
dependent on "murky determinations concerning the sequence of events in the plant, whether
those events would have been avoidable" and similar factors, is inconsistent with the intent of
Congress to require numeric emission standards that are measurable and easily enforceable. We
further note that the cases relied on by the commenters (as well as the Weyerhaeuser case) are
from the 1970s and 1980s and we believe that the court has refined its view since those
decisions. As recognized by the court in NRDC, arguments that violation was caused by
unavoidable technology failure can be made to the courts if an enforcement proceeding is
initiated.
We disagree with the commenter's argument regarding the decision in Cement Kiln. Nothing in
Cement Kiln suggests that EPA has a duty to reflect in its emission standards emission levels that
might occur during unpredictable malfunction events. To the contrary, Cement Kiln cuts to the
heart of commenters' argument. The fundamental principle of the commenters' argument is that
due to malfunctions, the MACT emission standards may be "unachievable" if specific sources
experience a malfunction. However, the court rejected the notion that because a MACT standard
may not be achievable by some sources, it is invalid. Cement Kiln, 255 F.3d at 861 ("EPA may
not deviate from section 7412(d)(3)'s requirement that floors reflect what the best performers
actually achieve by claiming that floors must be achievable by all sources using MACT
technology."). See also Sierra Club v. EPA, 479 F.3d at 878; NACWA, 734 F.3d at 1150
(explaining that the argument that a MACT standard is invalid because it is not achievable by all
sources at all times has been "roundly rejected" by the Court).
Thus, unlike start-up and shut-down, which are foreseeable operations, no one can predict the
nature, scope, severity, timing, length, number or likely recurrences of malfunctions a source
may - or may not - experience. Although the EPA bases emission standards on different
manners of operations and circumstances, in setting MACT standards, the EPA only takes into
account conditions that are "foreseeable" and "which can reasonably be expected to recur."
Sierra Club v. EPA, 167 F.3d 658, 665 (D.C. Cir. 1999); see also Nat'/Lime Ass 'n v. EPA, 627
F.2d. 416, 431 n.46 (1980).
Although EPA may have authority to address malfunctions through different mechanisms -
albeit limited by the Sierra Club (striking down exemptions for malfunctions) and NRDC
(striking down an affirmative defense for malfunctions) decisions, EPA is not required by section
112 to account for malfunction emissions by resetting MACT standards (e.g., with long
averaging times) or by setting separate MACT standards (e.g., work practice standards) as
suggested by the commenter. We note that in the final rule, based on detailed information
submitted by industry for two types of malfunction events - PRE) releases to the atmosphere and
emergency flaring events - we established work practice standards.
Comment 7: Commenters raised a concern that EPA is not establishing alternative emission
limits for malfunctions involving flares, relief valves, and control bypasses, which the
commenters identify as critical safety devices. The commenters stated that these safety devices
317
-------
are designed to operate with emissions to the atmosphere under emergency or significant process
upset situations. Commenters asserted that prohibiting atmospheric releases or subjecting use of
this equipment to standards applicable during normal operations would place operators in the
position of operating their units in potential non-compliance in order to protect plant personnel,
the community, and equipment. One commenter stated that malfunctions are a "normal"
operating mode for these systems.
Response 7: As provided in a previous response, we do not believe that section 112 requires
EPA to consider malfunctions in setting MACT standards. However, for the reasons provided in
detail in the preamble to the final rule, we have determined to set work practice standards for
malfunction emissions vented to PRDs and flares, which serve as safety devices.
Comment 8: Several commenters stated that there are work practices - such as monitoring of
operating parameters to identify a malfunction and stopping or cutting back the process - that
represent the best practices for minimizing emissions during a malfunction. Commenters stated
that while the measures that represent these best practices will depend on facility-specific issues,
such as process design, pollution control train, and other factors, they nonetheless represent as
required by CAA section 112(d)(2) & (3) "the maximum degree of reduction in emissions of the
hazardous air pollutants...achievable...through application of measures, processes, methods,
systems or techniques" and reflect "the emission control that is achieved in practice by the best
controlled similar source[s]." Other commenters stated that a reasonable approach would be to
gather data from the best performing sources on what operational "best practices" they have put
in place to mitigate any excess emissions during a malfunction event and return to normal
operating mode as soon as practicable. This could include things like: (1) preparing a plan or
checklist to address possible malfunctions of equipment which would include operational
procedures to be taken in the event of a malfunction; identification of back-up controls, monitors,
etc.; reporting and recordkeeping requirements and (2) prompt investigation or root cause
analysis to help prevent a recurrence of the malfunction.
Some commenters recommended that to address emissions during SSM the EPA could require
every MACT source to prepare an SSM plan for minimizing emissions as close to the normal
limit as practicable, and submit that plan to the EPA. The commenter stated that upon
submission to the EPA, the plan would be an enforceable term of the applicable permit, and
would provide for limits that are "continuously applicable." The commenter suggested that the
EPA would not need to approve the plans, but could reserve the right to comment on them, and
to request changes where needed. Commenters claimed these SSM Plans would provide
enforceable requirements to follow when unforeseeable startups or shutdowns or malfunctions,
make it impractical or unsafe to monitor or meet numerical limits. Commenters claimed that an
SSM plan requirement would be consistent with the Sierra Club decision, and noted that case
addressed a challenge to the 2002, 2003 and 2006 changes to the SSM plan rules that gradually
stripped away compliance requirements and public access to SSM plans to the point that the
court concluded the EPA had eliminated the enforceability of their content altogether.
Response 8: We disagree that we should require sources to develop an SSM plan to address all
emissions during SSM. As an initial matter, we believe that startup and shutdown are part of
planned and expected operation of equipment and that an emissions limit can be established for
318
-------
those periods of operation. The emission limit will vary by emissions source and for different
sources may be the same standard as applies during normal operations or may be an alternative
standard that is a numerical limit, work practice standard or operational standard. We see no
basis under section 112(d)(2) and (3) for allowing sources to establish in a separate planning
document not subject to EPA approval how each emission source will address startup and
shutdown emissions.
Regarding malfunctions, we believe that these events are generally varied and unpredictable and
it is difficult or impossible to anticipate what actions a source might need to take during the
different types of malfunction events that might occur. This is true both in terms of EPA
developing emission limitations that would apply during these events and also in terms of a
source identifying the actions it might take if such an event occurs. For that reason, we do not
believe that a generic work practice standard or an SSM plan would be an effective tool for
regulating emissions during a malfunction. Moreover, these approaches could possibly lead to a
situation where significant HAP emissions are released and the public has no recourse because
such emissions are "allowed" under the MACT standard.
Comment 9: One commenter urged the EPA to clarify that activation of an automatic shutdown
system does not automatically mean that the event is a malfunction. The commenter claimed that
automatic shutdown systems meet the definition of shutdown and/or startup. The commenter
noted that there are conditions that occur within the process units that may necessitate that the
unit shut down in a quick manner without significant planning beforehand, such as actions that
will prevent or mitigate significant equipment damage, protect personnel safety, or protect the
environment. The commenter contended that shutdowns that are triggered by an automatic
shutdown system, including Safety Instrumented Systems (SIS) meet the definition of
"shutdown" and "startup" as defined in the MACT general provisions (40 CFR 60.2).
Response 9: Shutdown means the cessation of operation of an affected source or portion of an
affected source for any purpose. Therefore, if a safety instrumented system causes a cessation of
operation of an affected source or portion of an affected source, then the phrase "for any
purpose" covers that event and it would be considered a shutdown.
Comment 10: Several commenters agreed that EPA should not establish an "affirmative
defense" to civil penalties for malfunction emissions that exceed the standards. One commenter,
citing NRDC v. EPA, 749 F.3d 1055, 1062-63 (D.C. Cir. 2014), claimed the EPA has no
authority to create such a defense by rule under the Act. Commenters asserted that affirmative
defenses have been widely abused in Texas and claimed one refinery assumes that every single
one of its emissions events will qualify for an affirmative defense.
Other commenters raised a concern that the proposal suggested that the NRDC decision
somehow precludes the EPA from including an affirmative defense to penalties in an
administrative enforcement action. The commenters asserted that the court specifically
recognized that the EPA has authority to limit the situations in which it will impose
administrative penalties in that manner. See 749 F.3d at 1063 ("By contrast, the EPA's ability to
determine whether penalties should be assessed for Clean Air Act violations extends only to
administrative penalties, not to civil penalties imposed by a court."). The commenters concluded
319
-------
that the EPA should affirmatively state in the final rule that the EPA will not seek administrative
penalties for excess emissions caused by malfunctions.
Response 10: We appreciate the support for not proposing to include an affirmative defense for
malfunctions. As the commenters note the court in NRDC vacated an affirmative defense in one
of the EPA's section 112(d) regulations. The court found that the EPA lacked authority to
establish an affirmative defense for private civil suits and held that under the CAA, the authority
to determine civil penalty amounts in such cases lies exclusively with the courts, not the EPA.
Specifically, the Court found: "As the language of the statute makes clear, the courts determine,
on a case-by-case basis, whether civil penalties are 'appropriate.'" See NRDC, 749 F.3d at 1063.
Further, as the D.C. Circuit recognized, in an EPA or citizen enforcement action, the court has
the discretion to consider any defense raised and determine whether penalties are appropriate. Cf.
NRDC, 749 F.3d at 1064 (arguments that violation was caused by unavoidable technology
failure can be made to the courts in future civil cases when the issue arises). The same is true for
the presiding officer in EPA administrative enforcement actions.
Comment 11: Commenters contended that the EPA has provided no explanation for changing its
position that it is inappropriate to rely on enforcement discretion, rather than establishing
regulatory language addressing the inability of sources to comply with technology-based
standards during SSM events. Commenters claimed courts have adopted the same view and cited
Portland Cement, 486 F.2d at 398 n.l; National Lime, 627 F.2d at 431 n.46 ("the flexibility
appropriate to enforcement will not render 'achievable' a standard which cannot be achieved on
a regular basis, either for the reasons expressly taken into account in compliance determination
regulations (here startup, shutdown and malfunction), or otherwise."); Marathon Oil Co. v. EPA,
564 F.2d at 1273 (EPA's statement that it would not take enforcement action against sources that
exceeded effluent limitations because of upset events is "not an adequate response" to the
argument that standards that cannot be met during unavoidable upsets fail to reflect available
technology). Also, the commenters contended that the EPA's statements that the EPA will
"determine an appropriate response" to reported exceedances based on, "among other things, the
good faith efforts of the source to minimize emissions during malfunction periods, including
preventative and corrective actions, as well as root cause analyses to ascertain and rectify excess
emissions" (79 FR 36945), are not in any way a substitute for the EPA setting the standards at an
achievable level in the first place. Commenters argued that the EPA should address the issues
raised by the inherent conflict between continually applicable emission standards and the
capability of the identified technology by promulgating some sort of alternative standard for
SSM events.
One commenter stated that the EPA's statement that it would "use its case-by-case enforcement
discretion" is woefully inadequate and questioned when and why would it ever be appropriate for
the EPA not to use its enforcement discretion if the source is unable to comply with emission
standards because of a malfunction, which the EPA defines as an event the source could not have
avoided through better design or operation and maintenance. Commenters also raised a concern
that the EPA's exercise of its enforcement discretion does nothing to prevent a source from
having to defend itself from a citizen suit or state enforcement action for the same malfunctions.
320
-------
Response 11: We addressed in a previous response the comments regarding that a standard
needs to be "achievable" and our decision not to either consider malfunctions in establishing the
MACT standard or set an alternative emissions limitations for most types of malfunctions.
Regarding the concerns that enforcement discretion is "woefully inadequate" from a facility's
perspective and that sources may be required to defend against a citizen suit or state
enforcement, as the court recognized in NRDC, in an EPA or citizen enforcement action, the
court has the discretion to consider any defense raised and determine whether penalties are
appropriate. NRDC, 749 F.3d at 1064 (arguments that violation was caused by unavoidable
technology failure can be made to the courts in future civil cases when the issue arises). Under
section 113(e) of the CAA, the Administrator or the court considers a wide variety of factors in
determining what penalty to assess, including compliance history and good faith efforts to
comply.
Comment 12: Multiple commenters supported elimination of the SSM exemption and EPA's
proposal to prohibit the release to the atmosphere from PRDs and stated that the CAA requires
emission standards to apply continuously. Commenters asserted that exemptions for emissions
during SSM does not encourage better control technology because it is easier to use a relief vent
as a control device than to design and install equipment to give facilities better control of the
process. Commenters stated eliminating the SSM exemption for flaring should drive a better
maintenance program at refineries. Commenters stated that some refineries have frequently
relied on the SSM exemptions in the past to avoid reducing their emissions and that it is
important to "remove this unlawful loophole."
Commenters provided examples of where they believe emissions from SSM events can have
significant air quality impacts. For example, commenters contended that emissions from a single
flare incident, even with 99% destruction efficiency, could potentially equal or exceed routine
emissions for the year, because of the large flows. Commenters also provided examples of SSM
events that led to SO2 emissions over the permit level and led to ambient air monitor readings
exceeding the SO2 NAAQS.
Some commenters asserted that all emissions, planned and unplanned, must be subject to
emission limits. Commenters stated that the EPA has failed to limit flaring emissions from
unplanned SSM events and asserted that Texas emission data shows a significant percentage of
flaring occurs as unplanned events. One commenter requested that the EPA include requirements
to prevent uncontrolled leaks, flaring, and explosions.
A few commenters stated that emissions from "emergency flaring" and other SSM emissions
should be accounted for in regulatory permits and that the EPA should reject permit applications
that fail to include all emissions.
Response 12: We appreciate the support for the elimination of the provisions that exempted
sources from applicable emission limits during SSM. Regarding startup and shutdown, we are
including in the final rule provisions to make the rule more practical and reflective of the
emission limitations achieved by the best performing facilities.
321
-------
As provided in previous responses, we disagree that section 112 requires EPA to set standards
that apply to all malfunction events. However, in the final rule, we are establishing work practice
standards that would apply to certain emergency flaring events and atmospheric releases from
pressure relief devices (PRDs) because we believe that these types of releases are easily
identifiable because flares, in part, and PRDs are devices that are designed to handle emissions
from malfunctions in order to prevent safety risks to personnel and equipment.
The comments regarding what should be included in federal and state permits are beyond the
scope of this rulemaking; those comments concern the requirements of the permit programs.
However, in general, we note that the purpose of permits is to specify the regulatory
requirements applicable to a source and to the extent that malfunctions result in emissions in
excess of what is required by regulation, such emissions likely would not be identified as a
regulatory requirement that is enforceable as part of a permit.
Comment 13: One commenter stated that history of the SSM exemption and compliance
problems for refineries demonstrate the need for stronger and more frequent monitoring, testing,
and reporting requirements, and additional enforcement provisions. The commenter stated that
the EPA must implement strong enforcement provisions to prevent and remedy emission spikes,
malfunctions and other violations in a way that will be enforceable by citizens in the Title V
permits for refineries.
A commenter stated that the EPA must promulgate specific public reporting and notification
requirements for malfunctions, or any emission exceedance that occurs. The commenter noted
that while EPA is requiring reporting for certain releases in excess of the emissions standards,
EPA proposed to delay reporting for much of these until the "periodic report," which may occur
as long as 8 months after the alleged malfunction. The commenter suggested EPA adopt a list of
reporting requirements focused on providing information to the community within 24 hours and
providing a more formal report to EPA within 7 days.
The commenter stated that a prohibition on malfunctions or other exceedances and quick
reporting when malfunctions do occur are not enough to protect the most exposed people from
refineries. The commenter stated that the EPA also must promulgate additional requirements that
apply in the event of a malfunction or violation of the emission standards to protect local
communities from health risk caused by refineries, and that facilities must not be able to emit in
an unlimited manner for an unlimited period of time. The commenter provided a list of actions
that it claimed EPA must require when a malfunction occurs, including automatic shut-off of
equipment, specific corrective measures, and written authorization from EPA to restart in certain
circumstances.
Another commenter encouraged the EPA to review best practices like near-miss incident
reporting to ensure that malfunctions are prevented and minimized as much as possible.
Response 13: This final rule does not broadly address the issue of catastrophic events and safer
practices, but as discussed previously, does establish standards for emergency releases to the
atmosphere via PRDs and emergency flaring. Under a separate program, the EPA implements
the mandates of CAA section 112(r) through its Risk Management Program (RMP). Among
322
-------
other things, this program requires facilities to develop and follow risk management plans so that
there are plans in place setting out how a facility will respond to and mitigate emergency releases
and how they will contact their local and state authorities so that in turn, the communities can be
notified in a timely manner, if necessary in order to take further actions and precautions. The
EPA, in coordination with the Department of Homeland Security and the Occupational Safety
and Health Administration, has been working to solicit public input through notices for
comment, community engagement and listening sessions on how to improve facility safety and
security requirements in response to the President's August 2013 Executive Order 13650
"Improving Chemical Facility Safety and Security." On July 31, 2014, the EPA issued a Request
for Information (see 79 FR 44604) to solicit public comments on potential revisions to its Risk
Management Program (RMP) regulations and is currently evaluating those comments.
11.2 Testing and monitoring requirements
Comment 1: One commenter stated that the EPA has not justified eliminating application of the
NESHAP General Provisions at 63.7(e) regarding performance tests, noting that the General
Provisions require that performance tests be conducted under "representative performance"
conditions. The commenter noted that, rather than reference the General Provisions
requirements, the EPA proposed to amend requirements already in Refinery MACT 1 and 2, and
in doing so for Refinery MACT 2, included regulatory language indicating that performance tests
can be conducted during startup or shutdown if "specified by the Administrator." The commenter
stated that such a regulatory change would allow the EPA to direct a source to conduct
performance testing during abnormal operations, including startups and shutdowns. The
commenter stated that, as the EPA and the courts have recognized, the manner prescribed for
performance testing affects the stringency of the emission standard, and that requiring sources to
demonstrate compliance using performance tests while the source is in startup or shutdown mode
can have the effect of making those existing MACT standards more stringent. The commenter
claimed that the EPA must provide a justification for such an amendment in light of EPA's
previous judgment about the appropriate level of the standards and the appropriate conditions for
performance testing. The commenter stated that the EPA must also demonstrate that the revised
standards would still meet the criteria of CAA section 112(d). The commenter stated that it
would only be appropriate for EPA to make this change if it first collects data during periods of
startup and shutdown that are used to re-establish the MACT standard, and only after proposal
and opportunity for public comments.
Response 1: As the commenter points out, rather than reference the performance testing
requirements in the General Provisions at 63.7(e), we included requirement directly in Refinery
MACT 1 and 2. We note that requirements were already in Refinery MACT 1 and 2, but as the
commenter notes, we proposed amendments to that regulatory language. We did not directly
reference the General Provision requirements. First, we wanted to be specific in the Refinery
MACT standards that while performance testing should be conducted under normal operating
conditions, those test should encompass the most challenging set of circumstances that occur
under normal operating conditions. Second, the phrase in the General Provisions requirements
stating, "emissions in excess of the level of the relevant standard... [are not] considered a
violation of the relevant standard..." is inconsistent with the Sierra Club decision. However, we
do not consider that the performance testing requirements that we added in Refinery MACT 1
323
-------
and 2 are otherwise significantly different than those provided in the General Provisions. We
note that we had intended to amend the requirements in both Refinery MACT 1 and 2 in a
similar manner, consistent with the intent of the General Provisions; however, we inadvertently
included the language about testing during periods of startup and shutdown in Refinery MACT
2. We have revised that language to remove that reference in this final rule.
Comment 2: A number of commenters disagreed with the quality assurance/quality control
(QA/QC) requirements specified in subparts CC and UUU. Commenters claim that despite the
logic and proven results of relying on manufacturer's recommendations and field experience for
QA/QC, the proposal specifies instrument QA/QC requirements in Table 13 of Refinery MACT
1 and Tables 40 and 41 of Refinery MACT 2. The commenters stated the proposed requirements
reflect an outdated and theoretical understanding of process instrumentation and do not reflect
the capabilities of modern instrumentation or refinery experience. Because of the diversity of
situations, the vast array of instrument technologies now available, the experience industry has
developed in assuring high quality measurements, and the most efficient match with current
practices and other applicable QA/QC requirements, a monitoring plan approach is the best way
to deal with CPMS QA/QC. Commenters stated that other rules and permits applicable to this
same equipment commonly include a requirement that manufacturers' recommendations be met.
For example, Refinery MACT 2 already requires such CPMS monitoring plans; they are already
in place and reflected in site permits and have been providing successful compliance assurance
for decades. These plans require monitoring equipment to be "installed, calibrated, maintained,
and operated according to manufacturer's specifications or other written procedures that provide
adequate assurance that the equipment will monitor accurately." The commenter contend where
manufacturer's specifications are available and appropriate to the actual installation, sources
should be allowed to use these methods in place of the generic QA/QC requirements included in
the proposed rule.
Additionally, commenters stated that the EPA has put forth no justification for overlaying the
proposed new prescriptive requirements on top of the existing plans or for imposing the burdens
on sources and regulators for modifying existing plans or for imposing the costs and burdens for
additional QA/QC. The EPA has made no demonstration to justify these changes under section
112(d)(6) or (f)(2), including evidence that compliance assurance problems justify these
extensive changes.
Commenters generally believed that these QA/QC requirements should be deleted from the
tables in favor of their inclusion in a CPMS monitoring plan. However, if the EPA believes there
is any justification for these high burden requirements in a modern facility, it should estimate
costs and burdens, identify the justification, and publish those analyses for comment, in order to
provide the analyses as part of the rulemaking record. If the EPA finalizes the proposed Refinery
MACT 1 Table 13 and Refinery MACT 2 Tables 40 and 41 QA/QC requirements, then
monitoring should be removed from the current operation, maintenance, and monitoring plan
provisions throughout the current Refinery MACT 2, and the proposed monitoring plan
requirements for Refinery MACT 1 should not be finalized, since these plans would only be
documenting the requirements already specified in these rule tables. Additionally, commenters
stated that 3 years are needed for compliance to allow evaluation of the many monitors currently
in use; to design and install upgrades and/or replacement monitors as needed; to revise
324
-------
procedures, OMMPs and permits to reflect these changes; or to obtain alternate monitoring
approvals where the requirements are infeasible or not appropriate. The commenter also stated
that delay provisions to the next scheduled shutdown should be provided for those situations
where a process or control device shutdown is required to complete a QA/QC activity.
Response 2: While we believe that it is essential for facilities to maintain a monitoring plan, we
disagree that a monitoring plan overrides the need for minimum standards for instrumentation.
Both subparts CC and UUU require that facilities develop monitoring plans. While it is often
necessary and prudent to rely on manufacturer recommendations, the minimum standards
outlined in the CPMS QA/QC requirements are meant to supplement manufacturer's
recommendations in order to ensure the monitoring systems maintain the required levels of
accuracy and reliability needed for continuous compliance. We believe that the general QA/QC
requirements provided in these tables are very reasonable and achievable. If these
requirements are more stringent than manufacturer's recommendations, we maintain that these
minimum requirements should be followed and can be implemented fairly quickly. If a
manufacturer recommends further QA/QC, more stringent QA/QC or more frequent QA/QC than
is required by the applicable subpart, the regulations in no way inhibit a facility from performing
the extra QA/QC. We also agree that each facility is unique and that diversity exists among the
source category, which is another reason why site-specific monitoring plans are crucial. While a
site-specific monitoring plan must still incorporate the minimum QA/QC requirements in the
applicable subpart, it allows facilities based upon their own knowledge of their operations to
determine what additional QA/QC is necessary and whether the QA/QC should be performed
more frequently. For these reasons, we also disagree that because we have maintained the
QA/QC requirements in Table 13 and Table 41 that we should remove the requirements for a
site-specific monitoring plan. In addition to the possible need to supplement the QA/QC
requirements specified by this rule, the site-specific monitoring plan should include information
beyond what is included in these tables. For example, subpart CC explicitly lays out what should
be in a monitoring plan, including a description of the equipment, the location, program of
corrective action, how data acquisition systems handle out-of-control periods, etc. None of this is
covered in Table 13, yet it is all integral to how the monitoring system operates and is
maintained.
We do not believe that outlining minimum QA/QC expectations for CPMS will conflict with
other rules or permits that have been issued. While these other rules and permits may specify that
the facility should follow manufacturer's specifications, the requirements in this rule do not
prohibit a facility from performing any QA/QC recommended by a manufacturer, although the
QA/QC outlined in this rule should also be implemented if it differs from that recommended by
the manufacturer as we believe that the requirements outlined in this rule are the basic QA/QC
necessary to ensure the accuracy and reliability of the monitors. While commenters assert that
relying on facility knowledge and manufacturer's recommendations have provided successful
compliance assurance for decades, we believe that it necessary to ensure a consistent level of
QA/QC across the industry. This is particularly important in order to decrease instances of
monitor downtime and to ensure continuous compliance. Commenters stated that it was
unnecessary to check monitor connections and to check for corrosion because loss of signal to
the control room indicates when there is an issue with the connections; likewise commenters
stated that normal industry practice is run-to-failure for certain temperature indicators. We
325
-------
believe that these are poor QA/QC practices that lead to preventable instances of monitor
downtime, which in turn lead to periods where there is no assurance of compliance. Therefore,
we believe it is appropriate to set minimum expectations for QA/QC in this rule.
We do not believe that the outlined QA/QC requirements pose a large burden on sources, as
these are the types of procedures that facilities should be performing to ensure the reliability and
accuracy of their instrumentation. We expect that facilities will have been performing these types
of procedures for some time. We expect that if a facility has not been performing these types of
procedures then they should see a decrease in downtime of their monitors by implementing these
basic QA/QC procedures. A decrease in monitor downtime will increase compliance assurance
and decrease periods of noncompliance, thereby benefitting the facility, the delegated authority
and the public.
We disagree that 3 years are needed for sources to comply with the accuracy requirements we
proposed (and are finalizing) in the tables. These accuracy requirements are very achievable for
the different monitoring systems, with the exception of the proposed oxygen sensor monitor
requirements in Table 41, which we have revised based on our review of these tables. The
specified QA/QC requirements do not require installation of new monitors or alternative
monitoring plans. Because manufacturer specifications should still supplement the specified
QA/QC, we do not expect that facilities will be performing less stringent QA/QC than they were
previously performing, and these new procedures should be able to be implemented while the
facility updates its monitoring plan. We also do not believe that any of the specified QA/QC
should require a unit shutdown. We have revised some of the proposed QA/QC requirements that
were identified as problematic (e.g., checking pressure taps). We have also added in allowances
for use of redundant sensors; redundant sensors would allow facilities to reduce the amount of
QA/QC that is required.
We maintain that the requirements in the final rule are reasonable for existing monitoring
systems; however, we reviewed these requirements to determine which, if any, may need
additional time to meet. The provisions in Table 13 of Refinery MACT 1 apply predominately to
new flare or DCU monitoring systems and 3 years are provided for those requirements. The
temperature QA/QC requirement would apply immediately to temperature monitoring systems
on MPV certain combustion control systems (other than flares). We expect any continuous
temperature monitoring system would be able to meet the QA/QC requirements, so we are not
providing additional time to comply with the requirements in Table 13. The only revision to
Table 40 is the addition of QA/QC requirements for PM CEMS. These are identical to the PM
CEMS requirements in Refinery NSPS Ja. We do not expect many units will have a PM CEMS
and only FCCU subject to NSPS Ja are expected to have a PM CEMS, so we expect any FCCU
with a PM CEMS would already be complying with the requirements in Table 40. Therefore,
more time is not being provided for compliance with Table 40. However, there are numerous
existing CPMS requirements in Refinery MACT 2 and new QA/QC requirements were proposed
for flow monitors, composition monitors (for determining coke burn-off rates) in addition to new
requirements for pressure and temperature sensors. While we expect that most monitors should
be capable of meeting the QA/QC requirements in Table 41, we consider it likely that as least
some monitoring systems may need to be upgraded in order to comply with the requirements in
326
-------
Table 41. Therefore, we are providing an 18-month transition period for compliance with the
requirements in Table 41.
Comment 3: One commenter stated that the EPA must require continuous monitoring of
emissions from refineries. The commenter provides a weblink to the Institute of Clean Air
Companies and stated that EPA is aware there is technology available to perform CEMS for Hg,
HC1, HF, PM, and opacity. The commenter asserted that the EPA should require the use of all
developments in continuous emission monitoring under section 112(d) and section 112(f) to
assure continuous compliance with emission standards, to implement the clean air act's enhanced
monitoring requirements, to take into consideration new developments in monitoring
technologies, and to assure an ample margin of safety. The commenter claimed stronger
emission monitoring is particularly needed due to the problematic compliance and exceedance
history. Finally, the commenter claimed EPA should require immediate reporting on the Internet
of all monitoring reports.
Response 3: We disagree that it is necessary to require continuous monitoring of all refinery
emissions as part of the risk and technology review. The monitoring requirements in this final
rule consist primarily of a combination of periodic compliance emission testing and continuous
parameter monitoring systems which we believe to be appropriate and sufficient for assuring
ongoing compliance with these standards. While we have provided an alternative to use a PM
CEMS for the FCCU regenerator, we also note that these systems have not been used or verified
on the FCCU source. Therefore, we are not requiring it as the only monitoring alternative at this
time. We also note that the web page cited by the commenter does not suggest that there are
CEMS for PM and mercury for refinery sources. The PM and mercury links from this page are
for control systems for these pollutants, not CEMS. Although we disagree that it is necessary to
require CEMS for all refinery emissions as part of the risk and technology review, we note that
we are enhancing the monitoring requirements for certain provisions, such as finalizing
additional continuous opacity monitoring requirements for FCCU complying with NSPS J,
revising averaging times for monitoring data, and finalizing continuous monitoring of pressure
relief devices to identify releases, as part of the final rule.
Regarding the request that we should require immediate reporting on the Internet of all
monitoring results, we disagree that "immediate" reporting of results is necessary or useful.
Although we are moving towards providing compliance data to the public through WebFIRE
these data must be quality assured, which requires time to conduct necessary checks to maintain
the integrity and accuracy of the information we make available to the public. In general, public
access to electronic data submitted to the EPA would be available in WebFIRE within 60 days of
submission.
Comment 4: Commenters stated that, if retained, Refinery MACT 1 Table 13 and Refinery
MACT 2 Table 41 should be made consistent based on the least burdensome requirement in each
table, and the language should be changed such that the same terminology is used in both tables.
The commenters noted that the inconsistencies range from fairly minor points (e.g., using metric
units in one table and English units in the other) to major differences (e.g., frequency of
instrumentation checks). The commenters contend that the differences are confusing and lead to
compliance and enforcement problems.
327
-------
A commenter noted that some of the QA/QC requirements specified in the tables in part 63
subparts CC and UUU differ from the QA/QC requirements specified for the same flare
instrumentation in part 60 subpart Ja. The specifications for the flow monitors in NSPS Ja do not
coincide clearly with the specifications of available instrumentation or the proposed
requirements in Table 13 of subpart CC. Ultrasonic flow meters are the bulk of the new and
existing flow instrument in this service. The commenter recommended the flow meter
specification should mimic the Shell Deer Park Flare consent decree.
Additionally, the commenter stated there are similar issues in the specifications for FCCU
Regenerator flow meters and for sulfur recover plants (SRP) flow meters, through the proposed
specifications in Table 41 of subpart UUU and the proposed additions of section 60.106a(a)(6)
and (7) to NSPS subpart Ja. Relative to FCCU, the commenter recommended an accuracy
requirement of ±5% over the normal flow range. For SRPs, the commenter requested that the
provisions be made consistent with those in Table 13 of subpart CC. The commenter requested
that the EPA clarify all NSPS Ja flow monitor specifications and provide clear specifications that
are consistent with available instrumentation specifications and that avoid unjustified
replacement of existing instruments.
Two commenters noted concern with the term "measurement sensitivity" in proposed sections
60.106a(a)(6)(i)(B) and 60.106a(a)(7)(i)(B) in both the sulfur recovery unit oxygen (O2)
alternative and flare flow standards in existing regulations 40 CFR 60.107a(f)(l )(ii). One
commenter believes this term requires clarification since "sensitivity" is not a term found on
typical data sheets and requested that the terminology used in these paragraphs be amended to
match the wording found in Table 13 of subpart CC.
Response 4: While we agree that the requirements for CPMS in Table 13 of subpart CC and
Table 41 of subpart UUU should be made consistent for the same type of CPMS, we do not
agree that the reconciliation must be made using the least burdensome requirement. We have
reviewed the requirements in these tables as well as the requirements for existing CMS in NSPS
subpart Ja and part 60 Appendix F to identify those minimum QA/QC requirements needed to
ensure accurate data are available for the CPMS. Based on this review, we have made revisions
to the requirements in these subparts and have reconciled the differences in the requirements to
address the concern that differing requirements are burdensome and lead to compliance issues.
For flare flow meters, specifications in Table 13 of subpart CC and NSPS Ja have been updated
to ±20% for a velocity range of 0.1-1 ft/s and ±5% of reading for a velocity > 1 ft/s in order to
better align with what is achievable, considering the widespread use of ultrasonic flow monitors.
The requirements in Table 41 of subpart UUU and the requirements for the flow sensors for SRP
in NSPS Ja have been updated to be consistent with those in Table 13 of subpart CC (i.e., ± 5%
over the normal range of flow measured or 280 liters per minute (10 ftVmin) whichever is
greater, for gas flowrate). Additionally, we have updated the requirements for flare and SRP flow
meters in NSPS subpart Ja to refer to monitor accuracy rather than measurement sensitivity.
Comment 5: Commenters were concerned about the new QA/QC requirements for CMS being
applied retroactively based on the EPA's claim in the proposed rule that these requirements
reflect what the EPA "always intended". Clearly identified applicability dates are needed to
328
-------
ensure that the changes are not interpreted to apply retroactively. This could be achieved by
adding a paragraph below the title for Refinery MACT 1 Table 13 and Refinery MACT 2 Tables
40 and 41 to indicate that these tables only apply after the compliance date for these
amendments. Another commenter suggested the changes to Table 40 should be included in a new
Table (e.g., Table 40a) with a clear applicability date for the new requirements specified.
Additionally, commenters stated the EPA should make clear the applicability of the tables.
Refinery MACT 1 Table 13 should clearly indicate that the table only applies to monitoring
instrumentation for flares as specified in 40 CFR 63.671, for Group 1 MPV combustion controls
as specified in 40 CFR 63.644(a), and for monitors used to monitor for flow in potential bypasses
around Group 1 MPV control devices as specified in 40 CFR 63.644(c).
Response 5: The revisions to the CMS QA/QC procedures are new requirements that apply to
the facility owners and operators only after the applicable dates provided in the final rule. While
we have not listed the applicability dates in the header of the tables, we have indicated the
applicability dates in the referencing regulatory text. For flare monitoring systems for example,
the requirements of Table 13 in subpart CC are referenced in §63.671(a)(1) and (e)(1). §63.670
clearly indicates when the requirements of §63.671 become effective.
The changes in Table 40 to subpart UUU are minor clarification changes or changes that address
compliance options that were added to the rule. For example, a PM CEMS option was provided
in the rule so QA/QC requirements for this is a new CEMS needed to be added to Table 40 to
ensure that the monitor is calibrated for the appropriate range over which it will be used. Because
the changes in the table are minor and the changes are linked to the changes in the emission limit
alternatives, the applicability dates are better suited to be placed on the emission limits, and we
do not see a need to create a new Table 40a. As noted in a previous response in this section, we
are providing an 18-month transition period for Table 41 because the revisions in Table 41 were
significant and impact a wide variety of existing monitoring systems.
Additionally, we believe that the regulatory text is clear on where these tables apply. Adding
applicability notations can become extremely cumbersome, especially if the list grows in the
future, and adds additional room for error and confusion if a specific reference is inadvertently
left out. We believe that it is most clear to let the regulatory text reference the tables when they
are applicable.
Comment 6: One commenter stated that Table 13 of subpart CC allows relief from certain
QA/QC flow meter requirements if "redundant" sensors are available; this provision should be
clarified and extended to Table 41 of subpart UUU. It is unclear what this term means,
particularly relative to flow monitors. In response to a similar question relative to a similar
requirement in NSPS subpart Ja, the commenter noted that the EPA stated the following:
Any meter or collection of meters that can provide a continuous measure of the
cumulative flow at the location of the required flare flow meter would qualify as a
"redundant flow sensor" in 40 CFR 60.107a(f)(l)(iv).
329
-------
[Reference: Letter from P. Tsirigotis, EPA to Matt Todd, API/AFPM, NSPS Ja
Clarifications and Corrections, August 27, 2013, Response to Question 14.]
The commenter requested this same clarification be made in the response to comments for this
rule or that the regulatory text be revised to make clear what a redundant flow monitor is.
Additionally, the commenter requested clarification that a redundant temperature or pressure
monitor would be one that measures the required temperature or pressure, even if it is not located
on the same pressure tap or in the same thermowell. Alternatively, to alleviate the need for
clarification, the commenter stated that the EPA should replace the word "redundant" with
"alternative" in Table 13 of subpart CC.
Response 6: We have reviewed the requirements in the tables to improve the consistency in the
CPMS requirements between subparts CC and UUU. The redundant flow monitor provision in
Table 13 of subpart CC (which we are also finalizing in Table 41 of subpart UUU) is analogous
to the requirement in 40 CFR 60.107a(f)(l)(iv), and the clarification provided in the letter from
P. Tsirigotis also applies here. As stated in that letter, the purpose of the inspection requirement
is to ensure that the flare flow measurements are not lost due to physical or operational integrity
problems, especially for reasons that could be avoided with appropriate preventive maintenance.
A redundant monitor can be a single meter, or if the flow measurement can be determined by
summing the flow from a series of other meters, the series of meters could serve as a redundant
flow monitor. The determination for whether a redundant flow monitor exists is whether the
required flow measurement is lost when the primary flow meter is down. If the required
measurement is not lost, then a redundant monitor exists.
We also agree that the same situation applies for the pressure or temperature monitor. The
redundant sensor does not need to be on the same pressure tap or thermowell, as long as the
required measurement is not lost when the primary meter is down. The redundant sensor must
meet the CPMS requirements in the applicable subpart. The owner or operator must also comply
with the operating limits using this redundant monitor. Thus, if an owner or operator elects to use
a temperature monitor at the incinerator exit as a redundant sensor to a temperature monitor in
the incinerator firebox, then the owner or operator must comply with the temperature operating
limit at the redundant monitoring location (i.e., at the incinerator exit) if the signal for the firebox
temperature sensor is lost.
We are retaining the word redundant in the tables, as we believe the intent is clearer with the
word redundant than with the word alternative. We are also concerned that the use of the word
alternative may cause confusion between a redundant monitor and a monitor approved under an
alternative monitoring plan.
Comment 7: Several commenters disagreed with specific QA/QC requirements for CPMS:
Pressure tap pluggage would be highly unusual in a refinery flare system or other
combustion based control device. Where pluggage is a problem, engineering fixes (such
as adding blow back steam or nitrogen) are usually preferable to daily or other routine
checks because they avoid environmental releases and potential personnel exposure. The
330
-------
requirement in Table 13 of MACT 1 and Table 41 of MACT 2 to check for pressure tap
obstructions should be replaced with the requirement to check that the pressure
instrument is responding (i.e., review straight-line readings), as a check of instrument
response covers the situation of a plugged pressure tap
Checking for pluggage of a pressure tap generally requires purging the pressure sensor,
thereby releasing VOC and HAPs to the atmosphere. If the requirement to check for
pressure tap obstruction is maintained, it should be changed to a weekly check, and the
resulting emissions to the atmosphere or to the flare should be specifically authorized.
The costs and burdens for checking for pressure tap obstructions have not been
considered in the rulemaking record and are significant. The commenter estimated a
typical incremental burden of 30 minutes per day of operator time for each pressure
CPMS. For delayed cokers, there are an average of 3.2 drums per unit and approximately
80 units resulting in an annual industry burden of over 46,000 hours per year for this one
set of pressure instruments. Since no emission reduction is associated with this activity,
the cost per ton of emissions reduction is infinite.
The requirements to check continuity on wiring and electrical connections and visually
check for corrosion are pointless, as are daily, weekly, monthly, and quarterly instrument
inspections for integrity issues. For most instrument transmitters, if the continuity fails on
any wiring the signal will no longer show in the control computer, and a work
notification will be created to fix the problem. For most noncritical temperature
indications, run-to-failure is the normal industry practice. The replacement of critical
temperature indicators is typically done on a turn-around cycle to prevent failure during
normal operation.
Manometers are unsafe to use and should not be required for pressure instrument
calibration. NIST traceable digital instrumentation is generally used instead of
manometers for most pressure instrument types. Because calibration practices change
over time, facilities should be allowed to determine the appropriate technology to be used
for calibration, based on recommendations from instrument manufacturers.
Annual calibration is a reasonable calibration frequency and all that should be required
for most monitors. Currently, annual calibration is performed in most cases, so moving to
more frequent calibrations is burdensome.
Response 7: We have reviewed the QA/QC requirements in subparts CC and UUU in response
to these comments. Based upon our review and consideration of the comments received we have
updated certain CPMS QA/QC requirements.
We agree that weekly checks of system response are sufficient to ensure the pressure monitoring
system is not plugged. We have revised the requirements in Table 13 of subpart CC and Table 41
of UUU to remove the requirement to specifically check for pressure tap obstructions and to
require weekly checks that the system is responding and to check for straight line (unchanging)
pressure.
We disagree that it is unnecessary to perform physical inspections of wiring and connections
associated with CPMS for integrity issues (e.g., corrosion, continuity, leakage, etc.). Although it
is true that when the connection is lost a signal will no longer show in the control room, the point
of the inspection is to catch issues before they reach the point of causing a lost signal. Once the
331
-------
CPMS reaches the point of losing signal to the control room, the problem may be substantial
such that the monitor will need to be taken out of service in order to fix it. It is prudent to inspect
the connections prior to this point in order to minimize monitor downtime. However, in the final
rule we have changed the requirements from monthly checks of pressure, temperature, oxygen
and flow rate sensors for leakage, galvanic corrosion, and continuity to quarterly checks. We
believe that quarterly checks will provide sufficient inspection of these systems for potential
issues prior to the point of failure of the signal to the control room.
We agree that some systems may operate at pressure ranges that are not suitable for
measurements with monometers and that NIST traceable instruments are acceptable alternatives
for performance evaluation of these pressure CPMS. We have removed references to the use of a
manometer for the performance evaluations.
In reviewing the QA/QC requirements for pressure CPMS in subparts CC and UUU, we noted
that the proposed rule contained an annual performance evaluation in Table 13 of subpart CC but
quarterly performance evaluation in Table 41 of subpart UUU. We agree that an annual
performance evaluation should be frequent enough to ensure adequate operation of the pressure
monitors covered by these subparts. As part of the reconciliation review of these requirements,
we have revised the calibration check for these devices to annually in both tables. However we
retain, with modifications, a requirement to review the pressure monitor output at least once per
week to verify that it is still operating properly and to perform corrective action if the pressure
readout suggests that there is blockage of the pressure tap (e.g., straight-line readings). We have
also revised the flow rate sensor calibration check requirements from semiannual to biennial
(every two year) and the oxygen content sensor calibration check requirements from quarterly to
annually. We believe that these frequencies provide adequate assurance that the monitors are
operating appropriately.
Comment 8: One commenter stated that approval should not be required for data compression
systems meeting rule requirements. The commenter stated that 40 CFR 63.655(h)(5)(iii) and
63.1573(d) specify requirements for use of digital data compression systems. [40 CFR
63.1573(d) is currently 40 CFR 63.1573(c), but proposed to be renumbered.] While such systems
were fairly new in 1994, they are standard now and data compression meeting the requirements
listed in these paragraphs is standard. Data compression is critical to assuring adequate data
storage and response time, because of the vast amount of data obtained by digital process data
systems and the large number of parameters monitored in a modern process operation. Many
requests for approval to use such systems have been submitted under Refinery MACT 1 and 2
and approved and we believe the burdens associated with obtaining such approval is no longer
justified if the criteria listed in each paragraph is met. The commenter therefore requested that
the requirement for obtaining approval to use these standard systems be changed to maintaining a
record that such a system is being used.
Response 8: We agree with the commenter and we have removed the requirement to obtain
approval to use these systems. We are not removing the requirements that the data compression
system must meet; we are only revising the requirement to obtain approval with a recordkeeping
requirement that the system meeting those criteria is being used.
332
-------
11.3 General duty provisions
Comment 1: One commenter stated that the EPA has not justified adding new "General Duty"
language to the existing subparts CC and UUU and should not finalize those proposed revisions.
The commenter noted that the EPA proposes to replace the NESHAPs General Provisions
section that establishes a "general duty" to operate a source consistent with safety and good air
pollution control practices for minimizing emissions, 40 CFR 63.6(e)(1), with somewhat
different "general duty" language in proposed sections 63.642(n) and 63.1570(c). The
commenter charged that this change is not being proposed under CAA sections 112(d)(6) or
112(f), and the EPA, therefore, lacks authority to make this change to the existing NESHAPs.
The commenter stated that the EPA's only explanation for these changes is that "[s]ome of the
language ... is no longer necessary or appropriate in light of the elimination of the SSM
exemption." See 79 FR 36945.
The commenter stated that D.C. Circuit did not vacate the provision the EPA proposes to replace,
40 CFR 63.6(e)(1), in Sierra Club v. EPA and, contrary to the EPA's assertion in the preamble to
the Proposed Rule, 40 CFR 63.6(e)(1) does not reference provisions that were vacated in Sierra
Club v. EPA. [Compare Table 6, 79 FR 36990, and Table 44, 79 FR 37043 of the Proposed Rule
with 79 FR 36945.] The commenter stated that even if the EPA had authority to change the
existing MACT standards in ways not required to address residual risk or new technology, the
Agency would have to provide a cogent explanation of why the old rule was unacceptable and
the new rule is necessary. The commenter stated that the EPA has not done so here.
The commenter stated that the EPA should not include proposed sections 63.642(n) and
63.1570(c) in the final rule. Alternatively, the commenter stated that the EPA needs to re-
propose the provision with some explanation of the basis and purpose for the provision, to allow
the public an opportunity to provide meaningful comments, as required by CAA section
307(d)(3). The commenter noted that the general duty language contained in the General
Provisions and the proposed general duty language are inconsistent with the EPA's insistence
that "releases" from relief valves are a violation of the proposed standards. The commenter
asserted that it would be arbitrary and capricious for the EPA to promulgate regulations that
require operation of relief valves (because they are necessary for the operation of the source
consistent with safety and good air pollution control for minimizing emissions) and
simultaneously state that it is a violation of the standards when those relief valves perform their
intended function.
Response 1: We did not propose these revisions pursuant to either CAA section 112(d)(6) or
112(f)(2). Rather, we proposed these revisions pursuant to our general authority to revise
regulations previously promulgated by the Agency. As we explained in the preamble to the
proposed rule, in order to ensure consistency with the court's decision in Sierra Club, we
proposed to remove provisions that exempted sources from compliance with emission standards
during periods of SSM. Because we are requiring that emission standards be met during periods
of SSM, we also proposed to remove the applicability to refinery MACT 1 and 2 sources of the
"general duty" provision, which provides: "The general duty to minimize emissions during a
period of startup, shutdown, or malfunction does not require the owner or operator to achieve
emission levels that would be required by the applicable standard at other times if this is not
333
-------
consistent with safety and good air pollution control practices." We are including all of the
requirements for periods of SSM for Refinery MACT 1 and 2 sources in subparts CC and UUU
rather than continuing to cross reference the General Provisions.
Regarding the concern that a general duty provision is inconsistent with the proposed ban on the
atmospheric releases from PRDs, we note that we are not finalizing the PRD provision as
proposed; instead, we are establishing work practice standards that apply to PRD releases.
11.4 Electronic reporting requirements
Comment 1: A few commenters stated that the proposed use of the Electronic Reporting Tool
(ERT) is not appropriate because the costs and burdens imposed are additive to the costs of
producing and submitting the written report, and there is no benefit that justifies the additional
cost. One commenter also stated that the EPA has not developed or articulated a reasonable
approach to using information that would be uploaded to the ERT. The commenters
recommended that the EPA remove this portion of the proposal until the ERT is demonstrated to
handle all the information from refinery performance tests (rather than only portions), thereby
eliminating the need for both written and electronic reporting and until the Agency demonstrates
that it is using the electronic data to develop improved air quality emission factors.
The commenters stated that there is essentially no likelihood the ERT systems will be able to
handle entire part 63 performance tests in the foreseeable future, and the added ERT burden
makes submission of reports within the short, required time (60 days) more difficult. One
commenter noted that the EPA proposed (in 40 CFR 63.655(9)(i)(B)) to require submittal of
performance test results to the EPA for tests conducted that are not compatible with the ERT.
The commenter stated that this added requirement complicates the cost and logistics of test
reporting for the industry and testing contractors by having to submit a report to the state agency,
enter partial information into the ERT, and then perhaps also having to send a full copy of the
test report to the EPA regional office. Another commenter similarly stated that the ERT
requirement does not supersede or replace any state reporting requirements and thus the
regulated industry will be subject to dual reporting requirements. Both commenters disagreed
with the preamble claim that eliminating the recordkeeping requirements for performance test
reports is a burden savings, and state that it may duplicate burdens already borne by the regulated
community.
The commenters expressed further concern that duplicative reporting requirements will strain the
regulated industry to comply with deadlines established by the rules for report submittals. One
commenter stated that there is no mechanism for obtaining extensions for special circumstances.
Under proposed 40 CFR 63.655(h)(9)(i), all reports are due in 60 days. By not referencing
reporting requirements to the General Provisions in 40 CFR 63.10(d)(2), there is no allowance
for obtaining additional time due to unforeseen circumstances or due to the difficulties involved
with completing particularly complex reports.
One commenter stated that the primary performance test method (Method 18) required for
determining compliance is not currently included in the list of methods supported by the ERT.
The commenter stated that the regulated community's experience with Method 18 is that it is a
334
-------
very broad methodology and can be exceptionally complex to execute and to report. The
commenter stated that the EPA is aware that Method 18 reporting is complex; that it may be
difficult to incorporate into the ERT; and that no time schedule has been defined for
development or implementation for this method.
The commenter also stated that without formal notice of changes to the ERT, the regulated
community is at risk of non-compliance. The only way for the regulated community to know that
changes have occurred within the ERT is to directly monitor the web site. Changes made to the
ERT are not formally announced by the EPA in the Federal Register so it will be possible for a
regulated entity to be unaware of changes made such as the incorporation of Method 18. The
commenter expressed concern that the proposal language is an open-ended commitment subject
to change without notice. The commenter stated that the EPA should clearly indicate when
facilities would be required to use the ERT when new test methods are included in the ERT.
Response 1: We disagree that use of the ERT for completing stack test reports is an added cost
and burden. To the contrary, based on the analysis performed for the Electronic Reporting and
Recordkeeping Requirements for the New Source Performance Standards Rulemaking
(ERRRNSPS) (80 FR 15100), electronic reporting results in an overall cost savings to industry
when annualized over a twenty year period. The cost savings is achieved through means such as
standardization of data, embedded quality assurance checks, automatic calculation routines, and
reduced data entry through the ability to reuse data in files instead of starting from scratch with
each test. As outlined in the ERRRNSPS, there are many benefits to electronic reporting. These
benefits span all users of the data - the EPA, state and local regulators, the regulated entities, and
the public. We note that in the preamble to the proposed rules we provided a number of reasons
why the use of the ERT will provide benefit going forward and that most of the benefits we
outlined were longer-term benefits (e.g., reducing burden of future information collection
requests). Additionally, we note that in 2011, in response to Executive Order 13563, the EPA
developed a plan104 to periodically review its regulations to determine if they should be
modified, streamlined, expanded, or repealed in an effort to make regulations more effective and
less burdensome. The plan includes replacing outdated paper reporting with electronic reporting.
In keeping with this plan and the White House's Digital Government Strategy105, in 2013 the
EPA issued an agency-wide policy specifying that new regulations will require reports to be
electronic to the maximum extent possible. By requiring electronic submission of stack test
reports in this rule, we are taking steps to implement this policy. We also disagree that we have
not developed or articulated a reasonable approach to using information that would be uploaded
to the ERT. To the contrary, we have discussed at length our plans for the use of stack test data
collected via the ERT. In 2009, we published an advanced notice of proposed rulemaking (74 FR
52723) for the Emissions Factors Program Improvements. In that notice, we first outlined our
intended approach for revising our emissions factors development procedures. This approach
included using stack test data collected with the ERT. We reiterated this position in our
Recommended Procedures for the Development of Emissions Factors and Use of the WebFIRE
104 EPA's Final Plan for Periodic Retrospective Reviews, August 2011. Available at:
http://www.epa.gov/regdarrt/retrospective/documents/eparetroreviewplan-aug2011 .pdf.
105 Digital Government: Building a 21st Century Platform to Better Serve the American People, May 2012. Available
at: https://www.whitehouse.gov/sites/default/files/omb/egov/digital-government/digital-government-strategy.pdf
335
-------
Database (http://www.epa.gov/ttn/chief/efpac/procedures/procedures81213.pdf). which was
public noticed before being finalized in 2013. Finally, we discussed uses of these data in the
preamble to the proposed rule and at length in the preamble to the ERRRNSPS.
We think that it is a circular argument to say that the Agency should eliminate the use of the
ERT until it demonstrates that it is using the electronic data. It would be impossible for the
agency to use data that it does not have. We can only use electronic data once we have electronic
data. We do note that we are nearing completion of programming the WebFIRE database with
our new emissions factor development procedures and anticipate running the routines on existing
data sets in the near future.
We continue to improve and upgrade the ERT on an ongoing basis. The current version of the
ERT supports 41 methods, including EPA Methods 1-4, 5, 5B, 5F, 25A 26, and 26A. We note
that the ERT does not currently support EPA Method 18, and for performance tests using
Method 18, the source will still have to produce a paper report. However, we are aware of the
need to add Method 18 to the ERT, and we are currently looking at developing this capability.
As noted in the ERRRNSPS, when new methods are added to the ERT, we will not only post
them to the website; we will also send out a listserv notice to the Clearinghouse for Inventories
and Emissions Factors (CHIEF) listserv. Information on joining the CHIEF listserv can be found
at http://www.epa.gOv/ttn/chief/listserv.html#chief. We are requiring the use of the ERT if the
method is supported by the ERT, as listed on the ERT website
(http://vvvvvv.epa. gov/ttn/chi ef/ert/ert info.html) at the time of the test. We do not agree that it is
overly burdensome to check a website for updates prior to conducting a performance test.
While the requirement to report the results of stack tests with the ERT does not supersede state
reporting requirements, we are aware of several states that already require the use of the ERT,
and we are aware of more states that are considering requiring its use. We note that where states
will not accept an ERT submittal, the ERT provides an option to print the report, and the printed
report can then be mailed to the state agency. We believe that the time savings in the ability to
reuse data elements within reports is more than equivalent to the cost incurred by printing out
and mailing a copy of the report.
Comment 2: Two commenters expressed interest in 24/7 emissions information. One
commenter asked for a public website reporting daily air pollution and flaring events, while
another commenter asked a public website posting real-time fenceline monitoring data. The
commenter described LACEENonline.org which gives citizens an opportunity to call or enter
online notification of episodes (e.g., flare-ups, explosions, fume releases). The commenter noted
that LACEEN is being used as a demonstration project all over California through the
Department of Toxic Substances Control (DTSC) agencies. The commenter recommended a
comprehensive plan with the EPA overseeing the AQMD, DTSC, and ARB, and all of our non-
profits working together to build a communication and data sharing system. The commenter
suggested they would be able to give each agency pertinent information and ensure data are
displayed in libraries, churches or city halls on a monthly basis. The commenter contended that
the public is not getting enough information.
336
-------
Response 2: The commenter is requesting monitoring data to be posted that goes beyond the
scope of the monitoring requirements in Refinery MACT 1 and 2, as amended by this final,
which we have determined is sufficient to demonstrate compliance with the standards. We note
that we are providing a website where the fenceline monitoring data will be available to the
public. However, as we have discussed previously in chapter 8 of this document and in the
preamble to this final rule, those fenceline data are not real-time data. In terms of the
commenter's suggestion that the EPA provide a system where citizens can report episodes, we
note that the EPA already has a mechanism for individuals to report what appears to be a
possible violation of an environmental regulation at http://www2.epa.gov/enforcement/report-
environmental-violations. However, many environmental programs have been delegated to the
states and they have primary responsibility for them. Often, it is most appropriate for an
individual to contact their local city, county, or state environmental agency (or health
department) rather than the EPA. We also note that much of what the commenter is requesting
fall within the scope of the EPA's RMP, which is discussed in more detail earlier in this chapter.
337
-------
12.0 Refinery NSPS Subpart Ja
Comment 1: One commenter stated that if the proposed rule is finalized many atmospheric RVs
will be routed to flares that will have to meet the new section 63.670 flare requirements, but
making the additional connections to the flare header for RVs will also trigger NSPS Ja
requirements for that flare and any interconnected flares because of the unique modification
definition in NSPS Ja for flares. The commenter believes it is unreasonable to force flares into
NSPS Ja sulfur and other requirements, through this rulemaking, which deals with organic HAPs
and assures their destruction and of any associated VOC. The commenter recommended EPA
add language to NSPS Ja to exclude the tie-in of atmospheric RVs to an existing flare header as
an NSPS Ja modification.
Response 1: First, we note that we have revised the proposed requirements such that the final
rule includes a requirement for refinery owners or operators to provide a minimum of 3
prevention measures to ensure atmospheric PRDs do not release rather than requiring all
atmospheric PRDs to be vented to the flare. We note that some refinery owners or operators may
elect to connect these PRDs to a flare header, but the final rule does not require all PRDs to be
vented to a flare or other control system. Second, we estimated that all flares would be subject to
subpart Ja within the time frame associated with the compliance with the final MACT
requirements. As such, we see no need to exclude any new connections made to a flare header as
a result of the MACT rule revisions.
Comment 2: One commenter notes that with respect to the flow measurement QA/QC standards
in proposed sections 60.106a(a)(6)(i)(D), 60.106a(a)(7)(i)(D), 63 subpart CC Table 13 and 40
CFR 60.107(f)(l )(iv), most electrical connections associated with flow meters are in NEC Class
I, Division 2 areas. This, the commenter states, is to provide a degree of protection from
corrosion (versus nonsealed applications). The commenter states that conducting intrusive
inspections could add risk which would not otherwise occur. Examples of intrusive inspection
risk include the introduction of material into the connection, improper re-sealing of the
connection enclosure, or loosening of connections. The commenter requests that due to the low
risk of connection failure and the burden of quarterly inspections, the inspection requirement be
changed to a requirement for an initial inspection, followed by an annual (or longer) inspection
schedule thereafter.
Response 2: As noted in our response to similar comment on the MACT provisions, the purpose
of the inspection requirement is to ensure that the required CPMS measurements are not lost due
to physical or operational integrity problems with the monitoring system. We find that this
requirement is effective in identifying problems that could prevent monitoring system failures
and are reasonable maintenance requirements. The commenter can provide training to inspection
personnel to prevent the issues outlined by the commenter. We also note that we have provided
provisions not to conduct the quarterly inspections if you have a redundant flow sensor. Since we
desire to have CMPS data available at all times to assess compliance with the rule provisions, we
have determined that the proposed requirements are reasonable and we are finalizing these
requirements as proposed.
338
-------
Comment 3: One commenter notes that Table 8 requires that for FCCUs subject to NSPS J,
compliance is demonstrated by meeting the NSPS requirements. NSPS J specifies that the
measured CO value be adjusted to 0% O2. The commenter states that the equation used for
correcting measured FCCU regenerator CO values to 0% O2, as required by the NSPS J, does not
work properly for FCCUs that use oxygen enrichment, a common practice. Particularly, the
equations for correction measured CO in NSPS J and Ja to 0% O2 fails if the O2 concentration in
the regenerator off-gas is near or above 20.9, which occurs as FCCUs transition from normal
operation to startup and shutdown or into and out of hot standby, where O2 enrichment is in use
by the FCCU. Additionally the commenter stated that with O2 enrichment the O2 concentration
will exceed the normal air concentration during these transitions and result in large negative CO
concentrations being calculated.
Response 3: We did not propose to revise the CO limits in subparts J or Ja, so the CO limits in
those rules are not open for revision in these amendments. While we understand oxygen
enrichment is used, it is rarely used to the extent that the correction to the oxygen correction
equation would be significant. Additionally, we expect most FCCU to be operated with limited
excess oxygen in order to comply with NOx limits (either in subpart Ja or due to state/local or
consent decree requirements), so the oxygen correction is expected to be quite small even if it is
slightly overestimated because it does not consider less nitrogen entrainment due to oxygen
enrichment. With respect to transitioning during startup, shutdown, or hot standby, we see no
reason why a refinery would use oxygen enrichment in these cases. Oxygen enrichment is
primarily used to help increase the capacity of the unit. It would seem to be a complete waste of
money to use oxygen enrichment during these periods where air alone will provide enough
oxygen to combust the fuel/coke in these transitions. Thus, if the refinery owner or operator uses
oxygen enrichment when needed and not during these transitions, they can save money and
the correction equation would work fine. For all of these reasons, we deem it unnecessary to
revise the CO limits and oxygen correction terms in subparts J and Ja. Finally, in Refinery
MACT 2, we proposed and we are finalizing, with some modification from what was proposed,
requirements specific to startup and shutdown where the oxygen concentration must be
maintained above 1 percent. Therefore, the oxygen correction to the CO concentration is not
applicable for owners or operators electing to comply with the oxygen concentration alternative
for organic HAP control during periods of startup, shutdown, or hot standby.
Comment 4: One commenter states that the performance test in paragraph (ii) of sections
60.107a(e)(l)(iii) and (e )(2)(iii) allows the use of Cylinder Gas Audits (CGA) in lieu of relative
accuracy test audits (RATA) and specifies that "it is not necessary to include as much of the
sampling probe or sampling line as practical." The commenters requests that for CGAs required
under Appendix F (60.107a(e)(l)(iii) and (e)(2)(iii)), the same language be included to state: "It
is not necessary to include as much of the sampling probe or sampling line as practical." If it is
not required for the performance test (RATA-CGA), it should not be required for the quarterly
CGA under Appendix F. EPA should use this opportunity to make this correction.
Response 4: Thank you for the comment. We have had other questions (both internal within
EPA and from outside groups) about this provision and the primary reason we stated that it is not
necessary to use as much of the sampling probe or sampling line as possible were primarily
motivated by safety reasons, not wanting cylinder gases too near the flare, where radiant heat
339
-------
might be an safety issue. However, as a QA/QC matter, cylinder gas audits should generally
include as much of the sampling probe and sampling line as practical. We are clarifying in this
response that safety considerations when operating near a flare were our primary motivation for
providing this exemption within Refinery NSPS Ja. We were concerned that issues may arise
when determining how much of the sampling probe and sampling line is "practical" to include
when considering safety issues related to sampling lines near a flare. We note that Appendix F
provisions apply to a wide variety of sources that do not have this issue and should be required to
use as much of the sampling probe and sampling line as practical. Therefore, we do not want to
edit Appendix F to remove this requirement. Also, we did not propose to modify Appendix F or
the paragraphs the commenter requested that we revise, so we could not at this time make the
edits the commenter is requesting. We also considered removing this exclusion in Refinery
NSPS Ja and require the CGA for flares to use as "much of the sampling probe and sampling line
as practical considering worker safety." However, since we did not propose these revisions, we
do not consider it appropriate to finalize the removal of this exemption in Refinery NSPS Ja at
this time. However, as a practical matter, refinery owners or operators may and generally should
use as much of the sampling probe as practical, but the owner or operator may fully consider
worker safety for sampling lines near a flare when considering how much of the sampling line to
include.
Comment 5: One commenter states that the annual interval defined in proposed section
60.104a(b), 79 FR at 36,957, differs from the proposed interval contained in proposed section
65.280(b) of EPA's proposed National Uniform Emissions Standards/or Heat Exchange Systems.
77 FR 960, 976 (January 6, 2012). The commenters state that the EPA should maintain
consistency between these provisions or explain the reason for its decision to apply different
interval definitions.
Response 5: The Uniform Emission Standards proposed a definition of "reasonable interval" for
tasks to be conducted periodically; the proposed "reasonable interval" for monitoring required
annually was that the activities must be separated by at least 120 calendar days. First we note that
this "reasonable interval" was provided in the proposed general provision to the Uniform
Emission Standards in part 65 (and they were not proposed in parts 60 or 63). Furthermore, it is
always possible for the EPA to elect to override general provisions requirements in certain
subparts, so even if this definition was in the part 60 general provisions, the EPA has the
authority to specify alternative intervals that apply to a specific provision in subpart Ja. On
considering a reasonable interval for annual performance tests, we considered it better to provide
a 120 day (4 month) allowance from the exact annual PM testing date in NSPS subpart
Ja. Therefore, we proposed an 8 to 16 month interval for annual PM testing in NSPS subpart
Ja. We note that this interval would be compliant with the "reasonable interval" requirements
proposed in part 65. We find this 8 month window provides sufficient time to coordinate the
testing activities with plan operations, accounting for turnaround and other maintenance
activities. Therefore, we are finalizing this requirement as proposed. If the EPA does move
forward with the Uniform Emission Standards we may consider the commenters suggestion for
consistency and promulgate different requirements for the definition of "reasonable interval."
340
-------
12.1 Sulfur Recovery Plant Provisions
Comment 1: One commenter requested that the EPA define the term SRU "release points" in
proposed 40 CFR 60.102a(f) (79 FR at 36,956-957).
Response 1: Release points can differ for different types of sulfur recovery units or
different sulfur recovery plant configurations. The primary release points from a Claus sulfur
recovery unit are the tail gas stack and, if not diverted to the tail gas stream, purge gas from
sulfur recovery pits. As noted in the definition of sulfur recovery plant, the plant can consist of
numerous sulfur recovery trains and each train may have a separate tail gas stack and some
refineries may have to employ a separate release point(s) for their sulfur pits. For LoCat systems,
the primary release point is the off gas from the oxidation chamber. If a refinery owner or
operator is unsure what release points are associated with their specific sulfur recovery plant,
we recommend that they contact EPA or the delegated state authority for their assistance in
resolving their questions regarding their specific sulfur recovery plant configuration.
Comment 2: One commenter requested that the EPA clarify how to calculate allowable
emissions against the RQ of 500 pounds of sulfur dioxide over 24 hours. The commenter stated
that there are inconsistencies in determining exceedances of the RQ. The commenter stated that
some permits have a limit of average hourly emission rates plus the RQ as their allowable
emissions, meaning the actual emissions could exceed the RQ without requiring reporting.
Response 2: RQs are generally determined as emissions above permitted levels. For example,
permitted emissions from a sulfur recovery plant can exceed 500 lbs per day of SO2 for plants
recovering 300 long tons of sulfur per day. If the permitted emissions for a sulfur recovery plant
is 600 lbs per day, then the sulfur recovery plant would have to emit 1,100 lbs per day SO2 to
exceed the reportable quantity (or trigger an RCA for the sulfur recovery plant).
12.2 Performance Test Requirements (for flares for H2S)
Comment 1: One commenter stated that the proposed amendment to 60.104a(a) to require a
flare performance test to demonstrate initial compliance with the H2S limit on flare gas was
purposely not included in the NSPS Ja rule, because it is unnecessary in light of the requirement
for continuous H2S monitoring. The commenter believes it is not reasonable to require such
testing since H2S levels in flare gas vary substantially and a representative sample is not likely
for many flares during the short three hour performance test period. The commenter noted that
for most flares there is no continuous flow to test, and it would be environmentally imprudent to
require flare flows just to test.
Response 1: The commenter is mistaken that the H2S performance test for flares was
purposefully not included in the NSPS Ja rule. The performance test requirement has always
been a part of the NSPS subpart J requirements and was included in the original final subpart Ja
rule (6/24/2008). In the final amendments on 9/12/2012, when we separated flares from fuel gas
combustion devices, we specifically included flares as requiring performance tests. However,
late in the process of finalizing the NSPS Ja amendments, we moved the H2S limits for flares
from section 60.102a to 60.103a and inadvertently did not make the change in 60.104a(a)
341
-------
because the reference to section 60.102a in that section was general. We note that our intent was
to maintain this long-standing requirement and had we intentionally meant to exclude this
requirement, section 60.104a(a) would not have a reference to flares and section 60.104a(j)
would not have a specific reference to the concentration limit for flares in 60.103a(h).
We find that it is essential that, prior to the use of the H2S CEMS, a performance test be
conducted to ensure the CEMS is operating accurately. Typically, the H2S monitor will be
located somewhere in the flare header system and there will be some gas flow through the header
(purge or sweep gas) to prevent oxygen intrusion into the system. Therefore, the owner or
operator does not need to supplement this flow with additional gases just to perform the initial
test, but can perform the test using the gases present in the header system at the time of the
performance test.
Comment 2: One commenter stated that the limits in NSPS Ja imposing H2S and TRS
concentration limits for flare gas are specified on a dry basis, but flare gas is wet and any sample
system that removes water will also remove at least some H2S and TRS, making the values
measured using such a sample system questionable at best. The commenter recommended that
the NSPS Ja specify alternatives for the measurement of moisture content for adjusting the wet
readings for flare gas measurements, as follows to clarify how these corrections are to be made,
rather than leaving a source in limbo as to whether a particular approach is acceptable.
Incorporating these alternatives would also clarify that measurement of moisture contents is not
required.
Option 1: Monitor the temperature of the waste gas exiting the knock out drum and assume the
waste gas is saturated with moisture at that temperature. Use that volume % moisture to correct
the wet gas concentration to a dry gas concentration, or
Option 2: Determine the moisture content of the waste gas (under typical flaring conditions),
through bomb samples, wet and dry bulb temperatures, or similar means, to determine the
moisture content of the flare gas and use that to correct the wet basis correction factor, or
Option 3: Develop a flare gas moisture content correction (based on typical to conservatively
high waste gas temperature from the knock out drum and saturated conditions) for applicable
scenarios (e.g., flare gas recovery compressor outage, steaming equipment to the flare, individual
safety valve release). This would be a conservative moisture correction factor, but would not
require any additional monitoring or calculations) by the facility.
Response 2: First we note that the H2S concentration/emission limit has always been on a dry
basis (in subpart J, the limit was expressed as 230 grams per dry standard cubic meter). While we
express the limit in different measurement units in subpart Ja (162 ppmv), the actual emissions
limit did not change and the need to assess the emissions on a dry basis has always applied for
subpart J, so this is not a new issue that only pertains to subpart Ja. Second, we did not propose
to amend this requirement in subpart Ja, so this comment is out of scope. Third, we note that the
requirement in 60.107a(a)(2) is to "install, operate, calibrate and maintain an instrument for
continuously monitoring and recording the concentration by volume (dry basis) of H2S in the
fuel gases before being burned in any fuel gas combustion device or flare." Thus, it is unclear if
342
-------
a wet gas monitoring system is compliant with this requirement and, at the very least, if a wet gas
system is used, the wet gas concentration must be corrected for the moisture content. We
generally agree that the options noted by the commenter are reasonable methods for determining
the percent moisture in fuel gas discharged to the flare and therefore find these methods to be
acceptable for correcting to a dry basis. Since the system must be corrected to a dry basis on a
"continuous" basis, the saturated assumption is likely the easiest to implement while ensuring the
correction is not underestimated.
Comment 3: One commenter stated that section 60.102a(g)(iii) of NSPS Ja concludes that sulfur
limits for combustion of fuel gas do not apply to portable generators combusting gas from tank
degassing and/or cleaning. The commenter however, noted that "Portable generator" is not
defined and the explanation for this exception in Section 2.1 of the Response to Comments is
unclear whether thermal incinerators and diesel engines, the commonly used controls, are
considered portable generators. For these short term control situations, portable engines, thermal
oxidizers, and flares are typically used, depending on availability, gas composition, utility
availability and other factors. Additionally, portable flares are sometimes used. The commenter
believes EPA should address flares used for this purpose and revise NSPS Ja to allow temporary
use of flares, thermal oxidizers, and engines for tank degassing with spot sulfur sampling. The
commenter stated that given the small amount of gas generated from this type of activity, the low
concentration of H2S in these gases, and their infrequent occurrence, there is no justification for
limiting the types of VOC controls by imposing extensive compliance assurance requirements.
The commenter believes that the use of portable thermal oxidizers should be encouraged, since
these are the most energy and environmentally efficient choice. While some AMPs have been
approved for these situations, the burdens of obtaining AMPs on owners/operators and regulators
are excessive and can introduce delays in tank maintenance or lead to the use of less than
optimum controls.
The commenter suggested the following revision to section 60.102a(g)(iii):
(iii) The combustion in a portable generator of fuel gas released as a result of tank degassing
and/or cleaning is exempt from the emissions limits in paragraphs (g)(l)(i) and (ii) of this
section. The combustion of fuel gas in a portable engine or thermal oxidizer released as a result
of tank degassing and/or cleaning is exempt from the requirements of section 60.107a.
Compliance with the emissions limits in paragraphs (g)(l)(i) and (ii) of this section shall be
demonstrated through an initial sample analysis or the average of three hourly sample analyses
using either an H2S colorimetric tube or a portable H2S meter to determine the concentration of
H2S in gases entering the portable unit. A record of the date and time of each sample and the
sample results shall be maintained.
Response 3: We did not propose to amend section 60.102a(g)(iii), so this comment is out-of-
scope for these amendments. The provision was provided based on comment received on the
initial subpart Ja rule because these gases could be discharged to a flare and we deemed it
preferable to burn the gases in a device that could provide useable energy. With that
understanding, a diesel engine could qualify as a portable generator.
343
-------
Comment 4: One commenter stated that the limit of 162 ppm for H2S for fuel drums seems high
given the state of Louisiana has a high ambient air standard for H2S (237 ppb 8 hour). The
commenter inquired if other refineries in the USA have a lower limit for H2S fuel drums and
whether the limit is contingent on the total number of drums.
Response 4: The concentration of H2S in the fuel gas is designed to limit SO2 emissions from
fuel gas combustion devices. These streams are not released directly to the atmosphere.
12.3 Other Revisions
Comment 1: One commenter believes the EPA should address the API NSPS Ja reconsideration
petition item relative to the applicability of fuel gas combustion device requirements to flares
that only handle wastewater treatment unit off gas. The commenter states that by removing the
"fuel gas combustion device" description from the definition of flares when finalizing the 2012
NSPS Ja reconsideration amendments there is no longer a tie to the definition of fuel gas which
specifically says that "Fuel gas does not include vapors that are collected and combusted in a
thermal oxidizer or flare installed to control emissions from wastewater treatment units ..." As a
result flares that are used to comply with NSPS QQQ and/or part 61 subpart FF, and flares that
combust other gases that are not considered fuel gases under subpart Ja are no longer excluded
from subpart Ja fuel gas combustion device requirements. The commenter believe this change
was not proposed and is presumably inadvertent. The needed rule revisions to make NSPS Ja
consistent with NSPS J flare applicability should be included in the Refining Sector Rule final
rule. The commenter recommends adding a paragraph (2) to section 60.103a(g) as follows (based
from the definition of fuel gas in section 60.101a) and renumbering the existing section
60.103a(g) as section 60.103a(g)(l): (2) Flares that are only used to control emissions from
wastewater treatment units other than those processing sour water, marine tank vessel loading
operations or asphalt processing units (i.e., asphalt blowing stills) are excluded from the
requirements of section 60.103a.
Response 1: We did not propose to revise subpart Ja based on the issues noted in the
11/13/12 petition for reconsideration. As such, this comment is out-of-scope for these
amendments. Furthermore, we disagree with the commenter that the flares suggested by the
commenter should be excluded from the flare minimization and root cause analysis
requirements.
During the development of subpart Ja, we determined that it was not cost-
effective for combustion devices used to control vapors from marine vessel loading or
wastewater operations to install and operate sulfur removal (amine treatment) equipment in these
locations because they are generally remote and not easily piped to existing sulfur removal
systems. Thus, the exclusion from the definition of fuel gas was specifically targeted to the FhS
concentration limits and was not restricted to flares. The rationale to exclude combustion devices
from the FhS limits does not apply to the other flare gas work practice requirements.
We also disagree that the definition of modification of a flare does not limit the connections to
those meeting the definition of fuel gas. Thus, we contend that the definition of modification and
the other flare work practice standards were not necessarily restricted to "fuel gas." As such, we
344
-------
find that it is appropriate for all new or modified flares at a refinery to prepare a flare
minimization plan and conduct RCA if the flares exceed the RCA thresholds. Also, we provided
additional compliance time for flares requiring the installation of monitoring equipment to
comply with the work practice standards. Therefore, even though refinery owners or operators
may have expected that certain flares were excluded from all work practice standards, it was not
our intent to exempt these flares, and we provided time for the refinery owners or operators to
comply with these provisions.
Comment 2: One commenter states that in the proposed revisions to 40 CFR 60.100a, (79 FR at
36,956), EPA proposes to remove the phrase "and delayed coker units" from section 60.100a(b).
However, the compliance date for both flares and delayed coker units are given separately in the
same paragraph. The commenter believes EPA should explain the reason for and implications of
the removal of this phrase.
Response 2: The removal of the phrase "and delayed coking units" from the first sentence in
60.100a(b) was an inadvertent error. The only revision that we intended to make in 60.100a was
to allow owners or operators subject to subpart J to elect to comply with the requirements
in subpart Ja. In the final amendments, we have included the phrase "and delayed coking
units" in the first sentence in 60.100a(b).
345
-------
13.0 Economic Impacts
13.1 Economic Impact Analysis
Comment 1: One commenter stated that significant man hours will be needed to collect samples,
appropriately maintain all the data, analyze the data and perform required calculations, and to
perform required reporting and the commenter expects to have to add at least one additional
employee to manage responsibility of these increased requirements. This cost will be in addition
to the required equipment installation and maintenance, sampling resources and laboratory fees
incurred for compliance with the proposed rulemaking. The commenter argued that as a small
refinery, the regulatory costs and burdens in the Proposed Rule threaten to impose a
disproportionate economic hardship. The commenter urged the EPA to consider whether the
proposed regulation imposes a disproportionate economic hardship on small refineries and, if so,
to provide exceptions or mitigations from the Proposed Rule in order to mitigate that harm.
Response 1: As part of our economic impact analysis for the proposed and final rules (for the
economic impact analysis for the proposed rule see Docket Item No. EPA-HQ-OAR-2010-0682-
0228 and for the economic impact analysis for the final rule see Docket ID No. EPA-HQ-OAR-
2010-0682), we specifically evaluated impacts on small businesses. In the analyses for the
proposed and final rules, we found that no small business would be subject to costs greater than 1
percent of sales. In the analysis for the final rule, the average cost-to-sales ratio of 0.16 percent
for small businesses is slightly higher than the average cost-to-sales ratio for large firms of 0.01
percent, and the maximum cost-to-sales ratio of 0.80 percent for small businesses is slightly
higher than the maximum cost-to-sales ratio for large firms of 0.64 percent (not considering
recovery credits). In the analyses for both the proposed and final rules, we determined
that, because no small business/firms are estimated to have cost-to-sales ratios greater than one
percent, the cost impacts for the risk and technology reviews for existing Refinery MACT 1 and
MACT 2 standards will not have a significant economic impact on a substantial number of small
entities.
Comment 2: One commenter stated that the EPA should finalize the rules, but allow industry to
charge $8.00 per gallon to compensate for the costs of rule compliance. On the other hand,
another commenter stated that the rule will have no noticeable effect on the price of produced
petroleum.
Response 2: As part of our economic impact analysis for the proposed rule (see Docket Item No.
EPA-HQ-OAR-2010-0682-0228), we specifically evaluated prices for five different refinery
products (motor gasoline, jet fuel, diesel fuel, residual fuel oil and liquefied petroleum gas). We
performed a similar analysis for the final rule (see Docket ID No. EPA-HQ-OAR-2010-0682). In
the analyses for both the proposed and final rules, we projected that the prices for these products
would rise by two one-hundredths of a penny or less per gallon. We do note that the EPA has no
jurisdiction on what prices a refinery or gasoline retailer charges for their products. However, in
a competitive, free-enterprise market, we agree with the commenter that suggested that this rule
will not have a noticeable effect on the price of gasoline or the other petroleum products
evaluated.
346
-------
Comment 3: One commenter discussed specific circumstances in the community of Wilmington
California. The commenter explained that the local refinery has made significant strides towards
pollution control and is a good steward in the community by providing high paying jobs and
donations to charity. The commenter also stated that they do not believe some of the rates of
illness or monitoring data presented by environmental justice organizations. The commenter
referenced a specific air quality monitoring study which is published on http://www.laceen.org/.
The commenter questioned whether the data presented in this study is reflective of the emissions
from the refinery alone and whether it excludes other sources of pollution such as the highway.
Finally, the commenter discussed how importing oil from countries such as China is under
consideration in their community and they do not believe that the pollution emitted from
producing the same refined product in China is better for the environment. Additionally,
importing fuel would do harm to the local community's economy and possibly result in higher
fuel prices which would disproportionally affect the poor.
Response 3: As part of our economic impact analysis for the proposed and final rule (for the
economic impact analysis for the proposed rule see Docket Item No. EPA-HQ-OAR-2010-0682-
0228 and for the economic impact analysis for the final rule see Docket ID No. EPA-HQ-OAR-
2010-0682), we include a discussion of a conceptual framework for considering the potential
influence of environmental regulation on employment in the U.S. economy and a discussion of
the limited empirical literature that is available. We concluded that deriving estimates of how
environmental regulations will impact net employment is a difficult task, requiring consideration
of labor demand in both the regulated and environmental protection sectors. Economic theory
predicts that the total effect of an environmental regulation on labor demand in regulated sectors
is not necessarily positive or negative. Peer-reviewed econometric studies that use a structural
approach, applicable to overall net effects in the regulated sectors, converge on the finding that
such effects, whether positive or negative, have been small and have not affected employment in
the national economy in a significant way. Based on our economic analysis for the final rule, it is
unlikely that this rule will materially affect the "good stewardship" of refineries in Wilmington,
CA or elsewhere.
13.2 ICR Burden Estimates
Comment 1: One commenter stated that none of the costs and burdens associated with the
massive effort required to revise the procedures and permits to reflect the proposed changes are
reflected in the Information Collection Supporting Statements or other analyses. Sites and
permitting authorities will spend countless hours implementing these changes and that burden
should not be ignored. In fact, it is likely the required Prevention of Significant Deterioration,
New Source Review and Permit reviews for the required added facilities will be extended and
highly burdensome.
Response 1: It is unclear what aspect of the rule the commenter is suggesting will require such a
massive re-permitting issue. We anticipate that much of this concern was related to identifying
additional Group 1 MPV that were previously excluded due to the "periodically discharged"
exemption or developing procedures to ensure vessel openings for maintenance purposes
remained below the Group 1 MPV threshold. We have developed specific requirements for
347
-------
maintenance MPV during start-up and shutdown that we expect to resolve this issue and reduce
or eliminate the need to individually permit these sources.
Comment 2: One commenter stated that the production impacts, costs and burdens due to the
revised averaging time and startup and shutdown restrictions on catalytic crackers and the new
restriction on catalytic reformer operation imposed by the proposed Refinery MACT 2 revisions
are not included in the record. Nor are any of the costs or burdens associated with the extensive
monitoring revisions proposed for catalytic crackers reflected in the Information Collection
Supporting Statement.
Response 2: At proposal, we included costs for performing performance tests for each
FCCU. New operating parameter limits would be established during these tests, so the cost of
developing the new operating limits are included in the costs to conduct the source test. Since the
parameters to be monitored remain the same, we projected no additional costs for these
parameters. We established specific operating limits for FCCU startup and shutdown using
parameters that we anticipate all FCCU owners or operators would use to operate the process, so
we did not project any additional costs to comply with these provisions. In the final rule, we are
adding a 3-hour average 20% opacity operating limit for refineries complying with subpart J for
PM. Because these facilities are already required to comply with the 30% opacity emissions, no
new monitoring systems are required, limit to a control device operating limit, we contend that
the operating parameters that we require to be monitored are routine parameters that are used to
operate these control systems, so no additional monitors would be required. While the records
that need to be maintained may change, we do not project any net increase or decrease in the
recordkeeping and reporting burden, so we did not include these unchanging burdens in the
Supporting Statement. Finally, with respect to the operational changes imposed for catalytic
reformers, we anticipated that an appropriate control device would be available, since the initial
depressurization is sent to a control system and only minor modification to the purging protocols
would be needed to comply with this requirement. However, we do acknowledge that these
changes will require some engineering evaluation and operator training to implement. Therefore,
we have revised the supporting statement to include these one-time costs. We do not anticipate
that this requirement would have on-going recordkeeping and reporting burdens beyond those
already imposed.
348
-------
14.0 Statutory and Executive Orders
Comment 1: Several commenters stated that this rulemaking is a significant Agency action that
must meet all of the requirements imposed on such rulemakings by a variety of laws and
Executive Orders because they claim that it will exceed the $100 million per year threshold. One
commenter estimated the required cost for the industry to be in excess of $20 - 40 billion in new
investment, and in excess of $1 billion if new flare investments can be avoided, all for a very
minimal reduction in risk and HAP emissions. The commenter argued that the EPA has not
provided statutory analysis and record support to justify the new requirements, nor has the EPA
met its other legal obligations. The commenters recommended that the EPA complete reviews
under Executive Order 12866, the Regulatory Flexibility Act, the Congressional Review Act,
and the Paperwork Reduction Act for comment because they consider the reviews done to
support the current proposal to be incomplete and unrepresentative of the precedents, issues,
costs and burdens imposed. Two commenters added that reviews under Executive Orders 13211
and 13650 are also necessary. Finally one commenter commented that a Regulatory Impact
Analysis (RIA) be completed consistent with Executive Order 13563, Executive Order 12866,
and Circular A-4.
Two commenters highlighted the Congressional Review Act, pointing out its stipulation of
providing at least 60 days between the final publication date of the rule and its effective date for
Congressional Review. The commenters stated that all references to the compliance date as the
date of publication in the Federal Register must be changed to the rule Effective Date, which
must be at least 60 days after publication of the final rule.
Response 1: We note that we did not anticipate the requirements proposed would not have a
significant impact on refinery operations. Based on the comments received, we have made a
number of revisions to the proposed rule to address issues that commenters suggested would
cause the need to build hundreds of new flares and to pipe all atmospheric PRDs to a flare. These
revisions are consistent with our proposed cost impacts and our original intent to improve the
performance of these systems without the need to build hundreds of new flares. We have revised
our cost estimates based on comments received, considering the requirements in the final rule,
and we continue to project that the annualized cost of this rule will be well below $100-
million/yr. However, due to the high capital costs of the rule, we are classifying the rule as a
major rulemaking. Therefore, we agree that this rulemaking is subject to the Congressional
Review Act and we have revised the effective date of the rule to be 60 days after publication of
the final rule in the Federal Register.
Comment 2: Several commenters asserted that the proposed rule is a "significant energy action"
as defined under Executive Order 13211 (66 FR 28355, May 22, 2001), because of its potential
to have a significant adverse effect on the supply, distribution or use of energy. They stated that
not only is the current timeframe for compliance too narrow (currently the date of rule
promulgation), but that the three years provided for compliance in NESHAP rules is as well,
estimating that up to a decade is needed to carry out the flare modifications and new flare
construction necessary for compliance with the proposed rule. Coupled with this, they both
explained that flare outages are typically scheduled to coincide with large planned maintenance
outages, but this narrow timeframe does not allow for such coordination and will result in many
349
-------
"out-of-cycle" production outages. These production outages will likely be extended due to
reduced fuels production resulting from the lack of alternative limits for startup, shutdown, and
equipment preparation venting and bypassing.
Response 2: We disagree with the commenters that this rule will have a significant adverse
effect on the supply, distribution or use of energy. We have provided additional alternative limits
for startup and shutdown for certain emission sources based on the public comments received
and we consider up to 3 years an appropriate time interval to install the necessary control and
monitoring systems for flares. Based on our economic impact analysis, there will be a very
minimal increase in the price of refined petroleum products (0.02 cents/gallon) and only an
unperceptively small (0.0001 percent) decrease in the supply of refined petroleum
products. Consequently, we maintain that this rule is not a "significant energy action."
Comment 3: One commenter stated that significant policy issues raised by this rulemaking
require review under Executive Order 12866. The commenter outlined several issues such as the
proposed ban on PRDs, requirements for fenceline monitoring, DCU provisions, and new flare
requirements. Another commenter stated that the proposed fenceline monitoring is not a
technology development for equipment leak, storage vessel or wastewater sources. According to
the commenter, addition of fenceline monitoring on top of the existing Refinery MACT
requirements is contrary to the Executive Order 12866 mandate to avoid redundant, costly
regulatory requirements that provide no emission reductions.
Response 3: We note that this rulemaking is considered a significant regulatory action under EO
12866 because it raises novel legal and policy issues as noted in the preamble. Office of
Management and Budget (OMB) was provided a 90-day comment period for the proposed rule
and comments received from OMB were addressed prior of the proposal of the Sector rule. The
OMB also reviewed the final rule as required under EO 12866 for significant regulatory actions.
The specific issues noted by the commenter are addressed in the appropriate emission source-
specific sections of this Response to Comment document.
Comment 4: One commenter argued that the proposal impacts safety systems that protect
workers and the community. The commenter stated that the proposal must be fully vetted, and,
under Executive Order 13650, coordinated with the OSHA and EPA's Risk Management Office.
The commenter stated that releases required for safety should not be prohibited or declared
violations. According to the commenter, significant safety concerns include:
- Requiring control of all atmospheric safety valves increases the chance of a catastrophic
equipment failure due to inadequate relieving capacity, particularly during the period between
promulgation of these amendments and completion of the new flare systems and installation of
the additional safety valves that will be required.
- Imposing the normal operations velocity limits on flares designed to be used at their hydraulic
limit during emergencies will place operators in an untenable position until additional flares can
be installedmanaging the emergency safely while risking enforcement for violating
environmental requirements.
350
-------
- Proposing to treat relief valves on vapor collection systems as potential bypasses, requiring
them to be car-sealed closed or otherwise blocked, will expose the system to potentially
catastrophic failure.
- The prohibition on the use of control device bypasses during start-up, shutdown, and
malfunction, regardless of the safety risk associated with sending those streams to those controls,
will increase the risk of fire and explosion. In particular, operating ESP on FCCUs during SSM
has caused injuries and equipment damage.
- Requiring routine instrument QA/QC for unsafe to access instruments puts workers at risk (e.g.,
the proposed QA/QC requirements on flare pilot thermocouples).
Response 4: We disagree that further review of this final rule is needed under EO 13650. We
have made a number of revisions to the proposed rule to address public comments
received. Most notably, we have revised the provisions for atmospheric PRE) to require a
minimum of three prevention measures. This requirement is similar to requirements to improve
the safety of chemical processes, and will improve refinery safety rather than cause safety
issues. We have also provided emergency shutdown provisions for flares which ensure the safe
destruction of gases needing disposal at these times without the need to build new flares or cause
operators to choose between plant safety and environmental violations. We specifically proposed
alternative limits to avoid unsafe operations during startup of an FCCU and we have generalized
and extended that provision to include startup as well, regardless of control device
configuration. Therefore, we find that the final rule will not have significant adverse impact
on chemical facility safety and should improve chemical facility safety by incorporating
prevention measures to reduce the likelihood of a chemical release. As such, further review
under EO 13650 is not required.
351
-------
15.0 Other Comments
15.1 General Support
Comment 1: Many commenters expressed general support for the rulemaking including removal
of SSM exemptions, stricter limits for delayed coking units, flares, storage tanks, and the
implementation of a fenceline monitoring program.
Response 1: We appreciate the commenters support for the proposed rule to reduce air emissions
from petroleum refineries.
15.2 Editorial Corrections
Comment 1: Some commenters stated that section 63.670(c) incorrectly references paragraph
(b) when it should reference paragraph (h) of this section.
One commenter stated that the reference in section 63.670(b) to paragraph (h) should be
paragraph (g).
One commenter also stated that the reference in section 63.660(i) to (c)(l)-(3) should be (i)(l)-
(3).
Another commenter stated that the reference to (f)(1), (2), or (3) should be (e)(1), (2), or
(3) in proposed section 63.670(e).
A third commenter stated that the flare section (k)(2)(ii) should say "owner or operator", but
currently reads "owner of operator".
Response 1: We appreciate the editorial corrections noted by the commenters and these
corrections have been made in the final rule.
15.3 Miscellaneous Other Comments
Comment 1: Many commenters suggested the EPA adopt practices to reduce emissions and
increase safety using its 112(d) and (f) authority including:
Phase-out of HF
Requiring back-up power system to reduce emissions from power failure
Stronger leak monitoring including use of best available fence-line monitoring techniques
Adoption of an anonymous work reporting system to encourage workers who are closest
to the issues a confidential way to protect their neighbors and families. Some of the
commenters cited a few examples of these types of systems as a proof of concept
including the Confidential Incident Reporting and Analysis System developed by the
Federal Aviation Administration.
Requiring mandatory maintenance and parts replacement plans to prevent breakdowns
and malfunctions.
352
-------
Response 1: Please see Comment 3 in this chapter for our response regarding HF alkylation and
Sections 6.1 and 8.3 for our response regarding stronger leak monitoring and "best available"
fenceline monitoring techniques.
Requiring back-up power is a very significant capital expenditure. Facilities that produce more
fuel gas than they consume may elect to build a co-generation unit to utilize their excess fuel gas
and produce electricity for sale to the grid (and for internal use). However, refineries that do
not have excess fuel gas would need to use natural gas for their "back-up power" plant. Due to
economies of scale, it will generally be cheaper to purchase electricity than to produce it, so the
"back-up power systems" requirement would cause most facilities to make large capital
investments to build an electricity generation plant that would be used only once or a few times
per year, for only a few hours at a time. For the plant to generate power to be used immediately
upon a power failure, the unit would likely need to be operated in hot standby, resulting in
additional combustion (CO2) emissions on a continual basis in hopes to reduce or eliminate very
infrequent emission episodes from power interruptions. This is not an environmentally efficient
or beneficial solution.
Facilities should already have confidential reporting systems in-place and we believe "whistle
blower" laws are in-place to protect employees that report incidences of violations.
We are not specifically requiring detailed mandatory maintenance and parts replacement
programs. However, we are requiring redundant prevention measures for PRV and root cause
analysis for PRV releases and emergency flaring, which may result in such measures being
implemented. We note that, if an event recurs that is based on the same root cause or if an event
is caused due to poor maintenance, those events are considered violations of the emissions
standards. As such, refinery owners or operators have significant incentive to develop and
implement effective maintenance and parts replacement programs rather than a one-size fits all
federal requirement that would be either ineffective or unnecessarily burdensome.
Comment 2: A few commenters stated that a mandatory phase-out plan for hazardous and toxic
chemicals should be developed and would begin with the identification and annual public
notification of all hazardous and toxic chemicals, the publication of health impact data and risk
for every toxic chemical and substance released by oil refineries, and the identification of
alternative safe and acceptable chemicals and processes.
Response 2: Refineries produce petroleum products largely through the distillation of crude oil,
which is primarily the separation of naturally occurring mixture of chemicals by different boiling
point ranges. While some processes "crack" the larger hydrocarbon compounds into smaller
hydrocarbon compounds (e.g., C18 compound broken into two C9 compounds) for the "reform"
hydrocarbon compounds (e.g., converting hexane into benzene), nearly all of the basic
components of petroleum products are naturally occurring chemicals. Many of the issues that
have arisen from petroleum products has been associated with chemical additives, such as
tetraethyl lead and MTBE. We have phased these chemicals out of gasoline due to environmental
impacts of these chemicals. We have also required refineries to produce fuel products with lower
sulfur content to minimize emissions associated with the use of petroleum products.
353
-------
Petroleum products have been widely studied and we have a good understanding of the key risk
drivers associated with crude oil refining and petroleum refining products. While we have phased
out certain gasoline additives and establish other requirements for petroleum products, those
standards are beyond the scope of this rulemaking. Outside of the alternatives to HF alkylation,
which are addressed in Comment 3 of this section, we are not aware of any means to reduce or
eliminate specific HAP emissions through product or process substitution.
Comment 3: Two commenters stated that the EPA must set standards that will limit and assure
the phase-out of HF. One commenter made recommendations for requiring plants and refineries
to switch to cleaner and safer alternatives than HF during the listening sessions on improving
chemical safety as related to Executive Order 1350.
Another commenter explained that HF is a very corrosive and toxic inorganic acid. It can cause
severe health problems including death, permanent organ damage to the eyes, skin, nose, throat,
respiratory and bone systems. It can cause lung congestion, inflammation, and severe burns of
the skin and digestive tract.
The commenter cited an EPA report to congress detailing the "Goldfish Studies" in which certain
facilities and the Lawrence Livermore National Laboratory conducted a series of experiments
involving the atmospheric release of HF in discussing the risks to populations living near the
refineries (EPA Report at 1-2, http://www.epa.gov/oem/docs/chem/hydro.pdf). The commenter
also stated that 50 refineries across the nation still use the toxic chemical HF including some
which are near major cities including Houston, Philadelphia, Salt Lake City, and Memphis
(USW Report, supra note 617, at xvi). The commenter also added that a joint investigation of the
Center for Public Integrity and ABC News found that "[a]t least 16 million Americans" live in an
area where they would be exposed to HF from a refinery, if it were to be released in an accident
or a terrorist attack, according to refinery owners' worst case scenario reports (which the Center
analyzed but are not publicly released)106. And the Center also found that 23 refineries using HF
had at least 29 fires since the start of 2009. A majority - 32 of the 50 refineries using HF - were
cited "for willful, serious or repeat violations of rules designed to prevent fires, explosions and
chemical releases, according to U.S. OSHA data analyzed by the Center," over a recent 5-year
period. The United Steelworkers' (USW) survey found that over a five year period, there were
about 131 HF releases or near misses (USW Report, supra note 617, at vi-vii). Additional
evidence shows violations at other refineries.
The commenter stated that the EPA requires refineries using HF to estimate lives at risk from a
release, and half have a radius endpoint greater than 20 miles. The commenter argued that the
EPA has these reports, which it should consider and use in this rulemaking to phase out HF.
The commenter explained that the best available information shows that refineries can reduce the
catastrophic impacts of HF if it is used, and avoid the resulting threats to community and worker
106 Jim Morris & Chris Hamby, Ctr. for Pub. Integrity, Use of toxic acid puts millions at risk (Feb. 28, 2011),
http://www.publicintegrity.org/2011/02/24/2118/use-toxic-acid-puts-millions-risk ("CPI HF Report")
354
-------
health and safety, with various mitigation techniques or simply by not using HF at all.107 For
example, refineries using HF can:
Change the alkylation process to use a solid acid catalyst;
Convert the HF alkylation unit into a sulfuric acid unit; or
Add modifiers to the HF that decrease the gaseous nature of hydrofluoric acid and install
108
mitigation systems.
The commenter asserted in view of these facts, the EPA must set a limit on HF that protects
public health, under section 112(f)(2), and addresses developments in reducing HF emissions,
under section 112(d)(2)-(3) and (d)(6). There is no current emission standard for refineries that
directly controls HF. The commenter argued this violates the Act, as the D.C. Circuit explained
in National Lime Association (233 F.3d at 642.). Moreover, the commenter asserted that the EPA
has not satisfied the surrogacy test for HF {Sierra Club v. EPA, 353 F.3d 976, 984 (D.C. Cir.
2004)). Even if it had done so, the developments in the ability to reduce HF emissions would
require at least the following actions. First, the EPA must set as the standard for new sources an
emission limit of zero, to satisfy section 112(d)(3)'s requirement that the EPA set a standard
based on the best-performing single source. There are refineries that do not use HF and the EPA
has data on HF emissions from only 24 refineries, showing that many refineries are not currently
emitting HF (Draft Risk Assessment (-0225) at 33.). The EPA should also follow the Act's floor
and beyond-the-floor requirements in section 112(d)(2), (3) to set an HF limit for existing
sources that assures that if there is a catastrophic release, this will plainly violate those standards,
to provide additional incentive for sources to take steps needed to protect public health and
safety, and encourage more refineries to use safer alternatives to pure HF (section
112(d)(2)). Finally, as the use of HF is associated with such a high health threat, EPA has strong
grounds to require ultimate phase-out of its use. The EPA has done this under section 112 in
another rule, and should follow the same approach here. Specifically, in the chrome plating rule,
the EPA recognized that the health hazards associated with perfluoroctane sulfonic acid-based
fume suppressants were substantial enough that it should be phased out, and ultimately
banned.109 Section 112(d)(2) requires the EPA to assure the "maximum achievable degree of
emission reduction," and this includes consideration of whether emissions can be "eliminated."
Moreover, phasing out the use of HF would dramatically reduce the health threats associated
with potential releases of this dangerous pollutant at refineries, and thus would satisfy the section
107 See USW Report, supra note 617 at 3, 8; Meghan Purvis & Margaret Herman, U.S. PIRG Education Fund,
Needless Risk at 16-20 (Aug. 2005), http://www.uspirg.org/reports/usp/needless-risk (?U.S. PIRG Report?); CPI HF
Report, supra note 660
108 U.S. PIRG Report, supra note 667, at 16.
109 Final Rule, Chromium Electroplating, 77 FR 58,220, 58,230, 58,243-44 (Sept. 19, 2012) (40 CFR 63.342(c)(1))
(finalizing phase-out of PFOS-based fume suppressants, also described as perfhiorooctyl sulfonate); Proposed Rule,
75 FR at 65,068, 65,094 (Oct. 21, 2010) (citing report, EPA-HQ-OAR-2010-0600-0072).
355
-------
112(f)(2) requirement to assure an "ample margin of safety to protect public health.(quoting
Nat'/ Lime Ass'n, 233 F.3d at 639)."
Response 3: While HF is a HAP, our assessment of the remaining (or residual) risk after
application of the MACT standards does not indicate that there is unacceptable risk from HF
emissions. As one commenter notes, HF releases as a result of accidents and other catastrophic
events are among the scenarios evaluated under EPA's risk management program worst-case
release scenario in 40 CFR 68.25, and as a result of the concern about HF releases, refinery
owners and operators employ a number of redundant prevention measures to protect against a
significant release of HF. Based on the data we have evaluated, we do not find it necessary to
impose additional controls (or require refinery owners or operators to use alternative alkylation
processes) under CAA section 112(f)(2) because these emissions are adequately reduced under
section 112(r) of the CAA was amended in 1990. We currently have a petition for rulemaking to
address 1ST under section 112(r) and part 68. Also, pursuant to Executive Order 13650, EPA has
issued a "Request for Information" soliciting public views on the appropriateness of 1ST
regulations and is considering pursuing rulemaking as part of the regulation, standard and
guidance modernization effort called for by the Executive Order. Historically, EPA's authority to
address catastrophic releases under the NESHAP program was viewed as limited under the pre
1990 CAA. Congress added section 112(r) to address this gap. In light of the extensive history
and efforts of the agency on 1ST specifically and catastrophic accidents generally under the
section 112(r) program, and in light of the statutory structure of section 112, we view the request
to enact 1ST provisions in this rule to be outside the scope of section 1112(d)(2), section
112(f)(2) and section 112(d)(6). Therefore, this comment is outside the scope of the current
rulemaking.
Comment 4: One commenter stated that current standards only require testing for pollutants
every 5 years for air quality standards and 8 years for water quality standards. The commenter
added that there has been an increase in pollution in north central South Dakota related to the
discovery and extraction of oil. Based on the increase in pollution, the commenter recommended
that air and water quality studies be conducted semi-annually throughout the country and they
should be paid for by the industries or companies. Another commenter stated that the EPA
should require monitoring of all chemicals released by refineries on an ongoing basis.
Response 4: While specific performance tests are only required periodically, we generally have
continuous emission or parameter monitoring systems that are used to ensure the emission
limitations are achieved on a continuous basis. In many cases, we use surrogate compounds, such
as PM for metal HAP or total hydrocarbons or benzene as a surrogate for organic HAP, to cost-
effectively test and monitor for the large variety of pollutants potentially released from
petroleum refinery process units. By using these continuous emissions, parameter, and
surrogate monitoring systems, we effectively require monitoring of all chemical released by
refineries on an ongoing basis.
With respect to upstream (oil production) emissions and water quality issues, these are beyond
the scope of this rulemaking.
356
-------
Comment 5: One commenter asserted that there is a need for public health impacts research data
on all chemicals. This would include determining and reporting all carcinogenic, neurological,
non-carcinogenic, bio-cumulative, physical and developmental public health impacts for all
chemicals.
The commenter further stated that there should be public health and safety limits for all
chemicals per the National Academy of Sciences. The commenter added that mandatory public
health exposure limits on all chemicals and substances released by oil refineries should be
established, as well as public safety limits for all potential public hazard scenarios on-site and
off-site. The commenter also stated that Worst Case Scenario Disaster Studies including
cascading and cumulative scenarios should be mandatory.
Response 5: We appreciate the desire to have health impacts data and health-based limits for
all HAP. As seen in our draft residual risk assessment for this source category (Docket Item No.
EPA-HQ-OAR-2010-0682-0225), we do have health effects benchmark levels for nearly all
HAP emitted from petroleum refineries and we are confident that we have fully characterized the
key risk drivers for the petroleum refining industry. While we have not established "public health
and safety limits for all chemicals" we have evaluated the projected ambient concentrations of
HAP near refineries based on refinery HAP emissions and compared these concentrations with
the health benchmark levels (which can be thought of public health limits) and refinery
emissions do not cause an exceedance of these levels (Hazard Index or HQ greater than 1).
Comment 6: Two commenters stated while the EPA included costs to the industry, a
cost/benefit analysis relating to health benefits was omitted. One commenter stated that this is a
serious omission. The commenter argued that the EPA went to great length to conclude that the
health effects of the releases after the rule is in place are essentially negligible based on current
information and added surely there is information available to determine the health benefits that
arise from this rule, particularly since the EPA states that the rule will result in a reduction of
5,600 tons per year of toxic air pollutants, and 52,000 tons per year of VOCs.
Response 6: We did not state that the health effects are negligible, but that the risks were
acceptable. We also discussed in the preamble of the proposed rule that the proposed rule is
expected to reduce the number of people exposed to risks exceeding 1-in-l million by 1 million
people and reduce the cancer incidence by 0.05 cases per year (a 15 percent reduction in the
overall cancer incidence associated with refinery emissions). Therefore, we agree that there are
health benefits associated with these HAP emissions reductions. Due to methodology and data
limitations, we were unable to estimate the benefits associated with HAP or VOCs that would be
reduced as a result of this rule. This is not to imply that there are no benefits of this rule; rather, it
is a reflection of the difficulties in modeling the direct and indirect impacts of the reductions in
emissions for this industrial sector with the data currently available.
Comment 7: One commenter inquired if there are calculations supporting the statement in the
Fact Sheet regarding the reduction of 5,600 tons per year of toxic air pollutants and 52,000 tons
per year of VOC.
357
-------
Response 7: In the preamble to the proposed rule, we projected 1,760 tons per year of HAP and
18,800 tons per year of VOC reductions achieved by the storage vessel and delayed coking
unit. We also projected significant emission reductions associated with the new flare operating
limits and monitoring parameters designed to ensure all flares achieve 98 percent destruction
efficiency. These reductions of 3,810 tons per year of HAP 33,100 tons per year of VOC are
documented in the memo "Petroleum Refinery Sector Rule: Flare Impact Estimates", which was
included in the docket (see Docket Item No. EPA-HQ-OAR-2010-0682-0209). The Fact Sheet
considered both the "new" reductions from storage vessels and delayed coking units and the
"old" reductions that we expected to be achieved when flares were used as a MACT control
device, with the emissions reductions rounded to two significant figures.
Comment 8: One commenter asserted that the new major proposed rulemaking amendments
justify having a Health Impact Assessment (HIA) prepared, and suggested that HIA be a
requirement in Title V permits in addition to Health Risk Assessments (HRAs). The commenter
suggested including a public health survey requirement in the HIA to establish the Public Health
Baseline. The commenter also recommended that HRAs be updated every 4 years and use the
worst year annual emissions as the baseline (not allowing the selection of the best or lowest
emissions year). The commenter also requested that the HRA show 5 consecutive years of data
for trending of the results. The commenter also stated that the EPA has failed to adequately
assess and underestimated the health impacts to pregnant women and prenatal fetuses. And
finally, the commenter suggested that the EPA sponsor health research on oil refineries.
Response 8 Section 112 of the CAA mandates a single risk review for MACT standards.
Ongoing risk assessment requirements suggested by the commenter are not authorized under
section 112 of the CAA.
Comment 9: One commenter stated that the proposed amendments to the MACT standards will
place an onerous and unnecessary burden on refineries that are not major sources of HAP
emissions, but are subject to the MACT standards as a result of the agency's 1995 "once in
always in" policy. The commenter argued that the proposed amendments should not apply to a
minor source of HAPs and the EPA could revoke the "once in always in" policy as considered in
2007 or more narrowly, the proposed rule could apply to refineries that were a major sources on
the effective date of the rule modifications, rather than the first substantive compliance date of
the original rule. Alternatively, the commenter added that if the EPA is concerned about
backsliding, there could be an exemption for non-major sources of HAP in Refinery MACT 1
and 2 standards.
The commenter is an asphalt refinery and stated that due to the proposed amendments their
storage tanks would have to comply with the Generic MACT standards even though these
standards were never meant to apply to sources that are not major sources of HAPs. Further, the
facility's benzene emissions are very low and thus fenceline monitoring is nonsensical. The
commenter added that their greatest concern are the proposed changes to subpart A affecting
flares. The commenter argued that the proposed flare provisions are wholly unnecessary and
unintended for a facility that is not a major source of HAP emissions.
358
-------
Response 9: We disagree with the commenter that this would apply to area sources of HAP
emissions. We note that we identified 7 refineries that applied for synthetic minor permits (to
reduce their federally enforceable PTE to less than 10 tpy of a single HAP and less than 25 tpy of
total HAP). If a facility did not obtain a synthetic minor permit prior to the applicability date of
the MACT standard, then the refinery is a major source of HAP emissions regardless of whether
the facility reduces its emissions to below the major source threshold once it complies with all of
the MACT provisions. The "once in always in" policy requires facilities to continue to comply
with the MACT standard in this case.
Comment 10: One commenter requested information on how the EPA will ensure refineries are
in compliance with the new rule. The commenter suggested that the rule identify and define the
type of oversight (i.e., inspections, monitoring, penalties) the EPA will perform for compliance
assurance. Another commenter suggested that a third party be responsible for oversight. A third
commenter suggested that there is a correlation between higher fines and the number of
violations, and suggested that the EPA enforce higher fines to assure compliance. One
commenter suggested a similar correlation between the number of lawsuits and resulting
emission reductions, and recommended that stricter regulations be put in place to require these
types of reductions without the use of litigation. Another commenter agreed that compliance
assurance and transparency of requirements is key in the providing the public with the ability to
understand what is being gained through this proposed rule.
Response 10: The new rule requirements will apply at all times after the applicability date of the
requirement (although we are finalizing reduced fenceline monitoring requirements under special
cases). We generally require continuous emission or continuous parameter monitoring systems to
ensure refinery owners or operators are complying with the rule requirements at all
times. Facility owners and operators with deviations of the emission limitations in the final rule
may be subject to enforcement actions, including the imposition of fines ($25,000 per day).
Comment 11: One commenter expressed concern that the EPA does not have the ability to
evaluate and adopt current technologies or address technology-related comments on this
rulemaking with the closure of its Environmental Technology Verification program.
Response 11: The Environmental Technology Verification (ETV) program concluded operations
in early 2014, but the ETV program only evaluated very specific technologies under specific
conditions. Most new and innovative control technologies that have been developed and
employed in the petroleum refining industry were not developed through or tested in the ETV
program. Therefore, we do not agree that the closure of the ETV program significantly limits
EPA's ability to evaluate current or innovative technologies or to address technology-
related comments on this or other rulemakings.
Comment 12: Two commenters stated that the public should be notified of incidents at a
refinery directly and one of these commenters added that such a system could be analogous to
those used for floods or missing children.
Another commenter stated that there are inconsistencies with how incidents are reported to the
NRC. Some refineries report all incidents and others repot only those to the LDEQ unless the
359
-------
Coast Guard is required to be involved. The commenter added that residents have more access to
the reports sent to the NRC.
Response 12: We have reviewed the emergency notification procedures currently required to be
provided by refinery owners and operators to local safety officials. We find these requirements to
be timely (typically within 15 minutes of the onset of an occurrence) and appropriate (including
the agencies contacted). If the release is significant enough to notify the surrounding
communities to implement an evacuation of other activity, the local safety officials would use
available means (sirens, emergency broadcast signal, etc.). The EPA considers the current
reporting requirements for emergency releases adequate and sufficient to protect the
communities near a petroleum refinery. We do note that local agencies can implement different
reporting thresholds than those specified for reporting to the NRC, so there may be instances
where the local agency may be contacted without a required NRC report.
Comment 13: One commenter stated that the EPA should oversee funds that are given to
Community Benefits Trust Fund to ensure that they go to children with conditions such as
respiratory illness caused by the refineries, such as the one implemented and utilized in the South
Bay.
Similarly, another commenter stated that industries should be mandated to put money into an
escrow account to provide environmental health professional services in those communities that
incur pollution as a result of industry operation. These environmental health professionals would
also examine patients and report data from conditions related to pollution.
Response 13: We disagree with the commenters that petroleum refineries (or other
manufacturing industries) should be mandated to put money in a fund to pay for the health care
of people living near the facilities and we certainly do not have the authority under the CAA to
mandate such a fund.
Comment 14: One commenter requested clarification on whether the Federal or state
government would have presiding authority over the rule. The commenter recommended that the
proposed rule be monitored from the Federal level.
Response 14: Generally, the delegated authority is the state agency (assuming they have an
approved program). There are some items, such as requests for alternative emissions limitations
for flares and site-specific monitoring plans for fenceline monitoring, where we also require
submission to the federal agency because we would like to provide technical assistance and
ensure consistency across these site-specific alternatives. Except for these cases, we consider the
state agency to be the appropriate authority for permitting and MACT enforcement.
Comment 15: A commenter requested clarification on whether the emission limits in the
proposed regulation would be subject to further reductions in the future. The commenter
recommended that if this is the case, the EPA should impose stricter standards now.
Response 15: The EPA is required to review MACT standards for developments in practices,
processes and control technologies every 8 years. Therefore, subsequent technology reviews may
360
-------
lead to stricter standards for certain sources. However, there is no on-going requirement to
perform risk reviews beyond this rulemaking.
Comment 16: One commenter inquired as to whether companies who comply or exceed the
reduction of emissions requirement will be given incentives. The commenter stated they believe
in incentives because history has shown allowing a company to receive incentives or grants
would entice compliance to the new rule.
Response 16: There are no specific incentives (beyond not being subject to enforcement actions)
for sources to comply with the requirements in the final rule. We have addressed suggestions that
we should reduce the burden of fenceline monitoring if refiners can demonstrate that portions of
the fenceline are significantly and consistently below action levels by allowing them to apply for
alternative monitoring approaches, as discussed in the fenceline monitoring chapter. We believe
this allowance will provide some incentive to achieve and maintain emission reductions.
Comment 17: One commenter stated that there should be a limit on total emissions from a
facility regardless of other sources within the facility. The commenter argued that the health
effects are not based on process chemistry or how the benzene was released into the local
community. The community is affected by cumulative exposure and regulatory efforts should be
to limit that exposure wherever it is economically practical to do so.
The commenter stated that all emissions within a particular facility, even when tied to product
shipping operations, are still emissions to which their surrounding communities are exposed.
Allowing a facility to deduct such emissions without a cap reduces their incentive to improve
those processes and prevent unnecessary exposure within the surrounding communities. The
commenter stated that allowing such emissions to go unabated is counter to the stated goals of
the new proposals.
Response 17: In Refinery MACT 1, most emissions sources that are co-located at the refinery,
such as gasoline loading operations and marine vessel loading operations, are part of Refinery
MACT 1 source category. We acknowledge that refineries may also have units subject to the
HON, but these are not unregulated sources. Thus, it is unclear to us what emission sources we
are allowing to go unabated. We expect that this may be due to provisions in the fenceline
monitoring corrective action requirements that allow facilities to correct for contributions from
non-refinery MACT sources. So, if a benzene storage tank subject to the HON provision is
leading to high fenceline concentration, the facility is allowed to correct for the contributions of
this HON tank. However, that tank would still have to meet the abatement criteria required by
the HON. If the HON requirements are not being met, then that facility owner or operator would
still be subject to compliance violations. In summary, we do not believe that there are any
"unabated sources" at a refinery facility that we are exempting from any MACT requirements.
Comment 18: One commenter asserted that the EPA has failed to evaluate the effect these rules
will have on the cumulative employment impact of EPA's rules. This requirement is set out in
section 321(a) of the CAA, which states that "the Administrator shall conduct continuing
evaluations of potential loss or shifts of employment which may result from the administration or
enforcement" of the law. The commenter argued that the EPA has failed to comply with this
361
-------
section and should amend the rules to account for (1) the cumulative employment impact of all
its existing regulations, and (2) how these rules will add to that.
Response 18: We disagree with this comment. Section 321(a) does not require the EPA to
evaluate the cumulative employment impacts of its rules, nor does it require the EPA to conduct
evaluations as part of the rulemaking process or as a prerequisite to the issuance of a rule. The
broad language of section 321(a) does not address the scope, frequency, or timing of
employment evaluations, leaving these matters to the discretion of the EPA. In the exercise of its
discretion, however, the EPA does routinely evaluate employment impacts when it conducts new
rulemaking. For example, in the EIA accompanying this final rule, the EPA provided a
qualitative assessment of employment impacts, describing the current state of knowledge based
on the peer-reviewed literature and the analytical challenges associated with determining the
direction of net employment effects of regulation on both the regulated sector and other
industries. In addition, the EPA has conducted similar assessments for other rulemakings and
continues to evaluate employment impacts generally and to refine its employment-analysis
methods. Together, these efforts constitute "continuing evaluations of potential loss or shifts in
employment" for the purposes of section 321(a). Finally, to the extent that the commenter is
suggesting that section 321(a) requires the EPA to amend the substantive requirements of the
rule to account for cumulative employment impacts, section 321(d) clearly states that nothing in
section 321 shall be construed to "require or authorize" the EPA "to modify or withdraw any
requirement" proposed under the Act.
Comment 19: One commenter stated that the purpose of the CAA is to protect public health by
preventing air pollution and that the statutory test for the health risk rulemaking under section
112(f)(2) is two-fold: (1) the EPA must prevent all unacceptable health risks; and (2) the EPA
must assure an "ample margin of safety to protect public health" and "prevent.. .an adverse
environmental effect (Id. section 112(f)(2)(A))." The commenter stated that the D.C. Circuit has
recognized that the "aspirational goal" of this provision includes reducing lifetime cancer risk to
the most-exposed person to be one-in-one million or lower. The commenter notes that EPA has
recognized that this provision directs the EPA to "protect the greatest number of persons possible
to an individual lifetime risk level no higher than approximately 1-in-l million" and "limit to no
higher than approximately l-in-10 thousand [i.e., 100-in-l million] the estimated risk that a
person living near a plant would have if he or she were exposed to the maximum pollutant
concentrations for 70 years."
The commenter asserted that the technology review required by section 112(d)(6) aims to
continue to reduce toxic air pollution and exposure to air toxics as greater emission reductions
become achievable.
The commenter claimed that under section 112(d)(6), when "developments" exist, EPA must
update the standards, as developments are the "core requirement" of section 112(d)(6). The
commenter added that revised standards must be governed by the plain text and statutory test
provided for all section 112(d) emission standards in section 112(d)(2)-(3). The commenter
argued that EPA must follow the plain text of section 112(d)(2)-(3) because those provisions
govern "emissions standards promulgated under this subsection," i.e., subsection 112(d), and that
EPA must set floor and beyond-the-floor standards.
362
-------
The commenter argued that although the D.C. Circuit rejected this argument in 2013, neither the
EPA nor any court has addressed the plain text of the statutory provisions. The commenter
concluded that because neither EPA or the court has addressed the statutory interpretation or
reasoning they believe that the EPA and the court wrongly decided this issue.
The commenter asserted that the EPA has a responsibility to provide a statutory interpretation
supporting its policy decision. For that reason and all of the reasons provided in the attachments
to the original comment, the EPA should set revised technology standards based on CAA section
112(d)(2) &(3). The commenter contended that refusing to recognize that revised air toxics
standards follow the same test as other section 112(d) standards would conflict with the EPA's
policy and interpretation in the electric utility MACT standards. The commenter asserted that for
the utility MACT standards, EPA recognized that section 112(n)(l)(A) provides the test for
whether standards are needed and then section 112(d)(2)-(3) provides the test for the actual
standards. The commenter stated that similarly section 112(d)(6) provides the test for whether
standards are needed and section 112(d)(2)-(3) provides the test for the standards.
The commenter also claimed that in White Stallion v. EPA, Both the court and EPA rejected the
same policy the EPA uses here. The commenter asserted that in that case, industry petitioners
argued that section 112(n)(l)(A) governed both the decision to set standards for coal-fired power
plants and the stringency of those standards, just as the EPA does for 112(d)(6), and that EPA's
approach for 112(d)(6) should fail for the same reason that argument failed in White Stallion
that is not what section 112 says and the interpretation disregards Congress's careful structure.
The commenter also argued that even if the Act were ambiguous, the EPA should interpret it to
apply the same test to revised air toxics standards as other emission standards because doing so
would better serve congressional intent and provide more health benefit which is the core of the
Act's objective.
The commenter cross-referenced previous comments they had submitted on this issue and their
briefs in NASF v. the EPA). The commenters stated that if the EPA does not follow the
commenter's interpretation of the Act, communities around the nation living in air toxics hot
spots will suffer. The commenter asserted that these communities are more likely to be
communities of color and lower income communities. The commenter cited EPA's recently
released Second Integrated Urban Air Toxics Report (Aug. 21, 2014) as recognizing that there
continue to be elevated areas with high cancer and other health risks and that the EPA would
"continue to address urban air toxics.. .through regulations called for under the CAA." The
commenter concluded that rulemakings where the EPA finds developments in pollution control
have occurred, and there is significant disparity in risk and health impact distribution - are
exactly the context in which the agency should take action to make its policy mirror what federal
law actually requires.
Response 19: As recognized by the commenter, the court in Association of Battery Recyclers v.
EPA, 716 F.3d 667 (ABR), rejected the statutory interpretation they make here. That court also
rejected a petition for rehearing en banc that the commenters filed on this same issue. More
recently, the court also rejected the argument, when raised again, in NASF v. EPA (D.C. Cir. No.
12-1459). As provided in previous rules and in the briefs in the cited cases, we continue to
363
-------
disagree with the position that the EPA must recalculate MACT floors under CAA section
112(d)(2)-(3) as part of the section 112(d)(6) review.
We read section 112(d)(6) as providing the EPA with substantial latitude in weighing a variety of
factors and arriving at an appropriate balance in considering revisions to standards promulgated
under section 112(d)(2) & (3). Nothing in section 112(d)(6) expressly or implicitly requires that
EPA recalculate the MACT floor as part of the section 112(d)(6) review. The interpretation that
the commenters urge is far from clear based on the statutory text. Nothing in the language
requires EPA to apply the criteria in section 112(d)(2)-(3) every time it reviews and revises
MACT standards. See NRDC, 529 F.3d at 1084. Although the language in section 112(d)(2)-(3)
may appear to be broad it cannot be read to apply to govern revised standards addressed in a
separate provision. See ABR, 716 F.3d at 672 ("It is a well-established principle of statutory
construction that '[gjeneral language of a statutory provision, although broad enough to include
it, will not be held to apply to a matter specifically dealt with in another part of the same
enactment.'") (citation omitted). Importantly, section 112(d)(2)-(3) apply to promulgation of
standards while section 112(d)(6) applies to a revision of standards that were already
"promulgated" within the meaning of section 112(d)(2)-(3). See Natural Res. Def. Council v.
EPA, 902 F.2d 962, 982 (D.C. Cir. 1990) (concurring opinion of Wald, C.J., vacated on other
grounds, 921 F.2d 326 (D.C. Cir. 1991), noting CAA's distinction between "promulgation" and
"revision"). For this reason, EPA's interpretation in the utility MACT, which was challenged in
White Stallion Energy Center, LLC v. EPA, 748 F.3d 1222 (D.C. Cir. 2014), is different from
EPA's interpretation in this and other rules issued under section 112(d)(6). EPA's promulgation
of the Utility MACT under CAA section 112(n)(l)(A) is precisely that - promulgation of a
MACT standard - not revision of an already promulgated standard. Further interpretation of this
provision in light of the arguments that the commenters make in their cross-referenced briefs can
be found in EPA's brief in the NASF case, which is located in the docket for this rule.
To the extent that the commenters raise health and risk concerns to advocate that EPA adopt their
preferred interpretation, we note that EPA has taken the position that it could consider those
factors as part of its analysis whether it is "necessary" to revise the standard under section
112(d)(6). However, in this rulemaking, we simultaneously performed the risk review required
under CAA section 112(f)(2), and we considered residual risk and health effects as part of that
review. We see no reason to duplicate that effort as part of the technology review in this
rulemaking.
Comment 20: One commenter stated that the agency has a long-standing federal CAA duty to
promulgate greenhouse gas emissions standards for refineries under section 111 of the CAA, and
has missed statutory and settlement deadlines to do so and argued that the EPA should take
advantage of the information it has collected for this rule to accelerate its progress towards
greenhouse gas controls for this very large pollution source. Another commenter stated that there
should be regulations to limit greenhouse gases from refineries and influence the use of carbon
sequestration. A third commenter also agrees that greenhouse gases should be limited and that
the U.S. should work towards total divestiture from fossil fuels. The third commenter expresses
concerns over global warming from a human rights perspective and makes other comments
related to limiting poverty and the development of projects like Keystone XL Pipeline or Trans-
Pacific Partnership.
364
-------
Response 20: We did not propose nor are we promulgating GHG standards as part of the
Refinery Sector Rule. We did collect data on energy use and other operating characteristics as
part of the information collection request performed to support the Sector rulemaking efforts that
may be useful if the EPA decides to propose and promulgate GHG standards for petroleum
refineries in future rulemakings.
15.4 Out-of-Scope
Comment 1: One commenter stated that there has been a large increase in the cost of fuel
especially for Ethanol Free gasoline which is used for recreational fishers and for lawn
equipment. Recreational fishing is a large portion of the St. Bernard's economy and the increased
fuel cost is negatively affecting economic development in this area. The commenter argued that
the surrounding refineries make supplies of Ethanol Free gasoline, but do not support the local
fuel needs while the community incurs the pollution associated with the refining process. The
commenter suggested that the refineries enter into a community benefits agreement which would
include meeting the local community's need for Ethanol free gasoline along with access to real
time monitor readings electronically, public admittance to monthly Community Advisory Panel
meetings, reinvestment of Louisiana DEQ penalty payments back into the community, and
funding for community identified projects, such as: sidewalks, bicycle paths, including the
Connect the Nine improvements to the Industrial Canal Bridge in New Orleans and the
Mississippi River Trail bike path along the river in Meraux and Violet (St Bernard); skate-parks,
dog parks, other parkways with added trees to help clear the air; public health clinics, work force
initiatives, and educational and cultural programs for youth.
Response 1: This comment is beyond the scope of this rulemaking.
Comment 2: One commenter argued that petroleum refineries are not significant source of
HAPs and that the EPA's efforts should be directed at the more significant sources of HAP
emissions - nonpoint and mobile sources.
The commenter supported their argument by stating that Texas accounts for about 15 percent of
the nation's HAP emissions since 1988. (See David Adelman, Environmental Federalism When
Numbers Matter More Than Size, 32 UCLA J. ENVT'L L. & POL'Y at 290) Harris County's
HAP emissions exceeded statewide emissions for all states other than Louisiana and Texas. Id.
Yet even in the Houston area, industrial HAP emissions are dwarfed by emissions from nonpoint
and mobile sources, representing only 25% of all HAP emissions. Id. at 280, 292. Further, HAP
emissions from petroleum refineries represent less than five percent of all industrial HAP
emissions. Id. at 288, Figure 11. While petroleum refineries will continue to work to further
decrease their resulting 1.25% contribution of HAP emissions in Harris County, EPA's efforts
should be directed at the more significant sources of HAP emissions - nonpoint and mobile
sources.
Response 2: Control of non-point or mobile sources is beyond the scope of this
rulemaking. CAA section 112 require MACT standards from all major sources of HAP
emissions. Because refineries are major sources of HAP emissions, the CAA requires us to
develop MACT standards for petroleum refineries.
365
-------
Comment 3: One commenter stated that the emission factor update proposal for flares and
petroleum refinery emissions should not be allowed to prejudice this rulemaking and any
emission factor changes must be delayed until after this rulemaking. The commenter pointed out
that the EPA posted an updated version of the Emissions Estimation Protocol for Petroleum
Refineries concurrently with its proposed AP-42 revisions, incorporating those proposed AP-42
emission factor revisions in it. The commenter claimed that for flares, the EPA's proposal of new
emission factors that are the same as those used in the Refinery Sector rulemaking prejudices the
evaluation of comments on the Refinery rulemaking since the Emission Factor Consent Decree
requires the new emission factors to be finalized prior to finalizing the Refinery Sector rule. The
commenter stated that the emission factor evaluation is prejudiced because the EPA would not
want to finalize anything different from the refinery rule basis, otherwise it would have to redo
the refinery emissions analyses and thus be unable to meet the deadlines in that Consent Decree.
The commenter stated that the Emission Factor Review schedule is unreasonable and provides
inadequate time for the EPA and the public to conduct a proper and thorough review of the
emission factors for the emission sources affected by the proposed revisions. The commenter
suggested that the EPA allow at least one year for completing an emission factor review for these
sources after the Refinery Sector Rule is finalized as well as at least a 180-day comment period
to provide adequate time for the multitude of interested parties to file sound comments on the
new emission factor proposal, including consideration of the comments received on the Sector
Rule proposal.
Response 3: This comment is beyond the scope of this rulemaking. We note that we did provide
an extension to the comment period for the proposed updates to the refinery emission factors. We
also note that we did not propose a correlation equation or other means to adjust control
efficiency based on flare operating parameters, so we disagree that the proposal and subsequent
finalization of new flare emission factors prejudiced the Refinery Sector rulemaking.
Comment 4: Two commenters stated that the EPA should review noise and vibrations from a
national performance standards perspective. The commenters added that currently noise
ordinances are regulated by local governments and they lack the resources and knowledge to
protect human health from the negative effects of noise pollution. One of the commenters
explained that high noise levels will cause hearing loss, sleep deprivation, and prevent children
from being able to read, study, or finish school assignments. The commenter also stated that the
vibrations have caused structural damage claims to houses. The commenter also explained that
are reported more frequently with "emergency flaring" and SSM activities as well
as "temporary" boilers, lowNOx burners, and even compressors, steam pressure relief valves,
and flare tip atomizers.
Response 4: We appreciate the concerns expressed by the commenters, but sections 111 and 112
of the CAA are specific to criteria and hazardous air pollutants. As such, noise and vibration
standards are beyond the scope of this rulemaking.
366
------- |