I
55
\
0*
Particulate Matter
Urban-Focused Visibility Assessment
Second External Review Draft
January 2010
-------
DISCLAIMER
This second draft document has been prepared by staff from the Office of Air Quality
Planning and Standards, U.S. Environmental Protection Agency. Any opinions, findings,
conclusions, or recommendations are those of the authors and do not necessarily reflect the
views of the EPA. This document is being circulated to obtain review and comment from the
Clean Air Scientific Advisory Committee (CASAC) and the general public. Comments on this
draft document should be addressed to Vicki Sandiford, U.S. Environmental Protection Agency,
Office of Air Quality Planning and Standards, C504-06, Research Triangle Park, North Carolina
27711 (email: sandiford.vicki@epa.gov).
Elements of this report (see Acknowledgements below) have been provided to the U.S.
Environmental Protection Agency (EPA) by Abt Associates Inc. and Stratus Consulting Inc. in
partial fulfillment of Contract No. EP-D-08-100, Work Assignment 0-11. Any opinions,
findings, conclusions, or recommendations are those of the authors and do not necessarily reflect
the views of the EPA or its contractors.
ACKNOWLEDGEMENTS
In addition to EPA staff, personnel from Abt Associates Inc. and Stratus Consulting Inc.
contributed to the writing of this document. The specific chapter where Abt Associates Inc. and
Stratus Consulting Inc. made contributions is chapter 2 (Urban Visibility Preference Studies).
-------
EP A 452/P-10-002
January 2010
Paniculate Matter
Urban-Focused Visibility Assessment
Second External Review Draft
U.S. Environmental Protection Agency
Office of Air and Radiation
Office of Air Quality Planning and Standards
Research Triangle Park, North Carolina 27711
-------
This page left intentionally blank.
-------
TABLE OF CONTENTS
List of Tables iv
List of Figures v
List of Acronyms/Abbreviations vii
1 Introduction 1-1
1.1 PM NAAQS BACKGROUND 1-3
1.2 VISIBILITY EFFECTS SCIENCE OVERVIEW 1-6
1.3 GOALS AND APPROACH 1-9
1.4 SCOPE OF URBAN-FOCUSED VISIBILITY ASSESSMENT 1-11
1.4.1 Background 1-11
1.4.2 Selection of Alternative Scenarios for First Draft Assessments 1-12
1.4.3 Selection of Alternative Scenarios for Second Draft Assessments 1-13
1.5 ORGANIZATION OF DOCUMENT 1-16
2 Urban Visibility Preference Studies 2-1
2.1 METHODS USED IN PREVIOUS STUDIES 2-1
2.2 DENVER, COLORADO 2-3
2.3 VANCOUVER, BRITISH COLUMBIA, CANADA 2-7
2.4 PHOENIX, ARIZONA 2-11
2.5 WASHINGTON, DC 2-14
2.5.1 Washington, DC 2001 2-15
2.5.2 Washington, DC, 2009 2-17
2.6 SUMMARY OF PREFERENCE STUDIES AND SELECTION OF CANDIDATE
PROTECTION LEVELS 2-25
3 Estimation of Current PM Concentrations and PM Light Extinction 3-1
3.1 SUMMARY OF PREVIOUS CHARACTERIZATIONS OF PM CONCENTRATIONS AND
LIGHT EXTINCTION 3-1
3.1.1 PM2.5 and PMio-2.5 3-1
3.1.2 PM light extinction 3-4
3.2 OVERVIEW OF APPROACH AND DATA SOURCES FOR URBAN STUDY ANALYSIS 3-9
3.2.1 Study Period, Study Areas, Monitoring Sites, and Sources of Ambient PM Data 3-10
3.2.2 Use of CMAQ Model Validation Runs for 2004 to Augment Ambient Data 3-16
3.2.3 Use of Original IMPROVE Algorithm to Estimate PM light extinction 3-17
IMPROVE ALGORITHM 3-19
3.3 DETAILED STEPS 3-19
3.3.1 Hourly PM2.5 Component Concentrations 3-19
3.3.2 Hourly PMio-2.5 Concentrations 3-26
3.3.3 Hourly Relative Humidity Data 3-27
3.3.4 Calculation of Daylight 1-HourPM Light Extinction 3-27
-------
3.3.5 Exclusion of Hours with Relative Humidity Greater than 90 Percent from PM Light
Extinction NAAQS Scenarios and Most Results 3-28
3.3.6 Calculation of Daily Maximum 1-Hour PM Light Extinction 3-32
3.4 SUMMARY OF RESULTS FOR CURRENT CONDITIONS 3-32
3.4.1 Levels of Estimated PM2.5, PM2.5 Components, PMio-2.5, and Relative Humidity 3-32
3.4.2 Levels of Estimated PM light extinction 3-35
3.4.3 Patterns of Relative Humidity and Relationship between Relative Humidity and PM light
extinction 3-41
3.4.4 Tile Plots of Hourly PM Light Extinction 3-44
3.4.5 Extinction Budgets for High PM Light Extinction Conditions 3-61
3.5 POLICY RELEVANT BACKGROUND 3-64
4 PM light extinction under "What If Conditions of Just Meeting Specific Alternative
Secondary NAAQS 4-1
4.1 ALTERNATIVE SECONDARY NAAQS BASED ON MEASURED PM LIGHT
EXTINCTION AS THE INDICATOR 4-1
4.1.1 Indicator and Monitoring Method 4-1
4.1.2 Alternative Secondary NAAQS Scenarios based on Measured PM light extinction 4-1
4.1.3 Monitoring Site Considerations for Alternative Secondary NAAQS Based on Measured
PM light extinction 4-2
4.1.4 Approach to Modeling "What If Conditions for Alternative Secondary NAAQS Based
on Measured PM Light Extinction 4-3
4.2 ALTERNATIVE SECONDARY PM2.5 NAAQS BASED ON ANNUAL AND 24-HOUR PM2.5
MASS 4-10
4.2.1 Secondary NAAQS Scenarios Based on Annual and 24-hour PM2.5 Mass 4-10
4.2.2 Approach to Modeling Conditions If Secondary PM2.5 NAAQS Based on Annual and 24-
hour PM2.s Mass Were Just Met 4-11
4.3 RESULTS FOR "JUST MEETING" EACH ALTERNATIVE SECONDARY NAAQS
SCENARIO 4-12
5 References 5-1
Appendices 5-1
Appendix A - PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the Analysis of
PM light extinction in the 15 Study Areas A-l
Appendix B - Distributions of Estimated PM2.5 And Other Components B-l
Appendix C - Development of PRB Estimates of PM2.5 components, PMio-2.5, and PM light
extinction C-l
Appendix D - Relationships between PM Mass Concentration and PM light extinction under
Current Conditions D-l
Appendix E - Differences in Daily Patterns of Relative Humidity and PM light extinction between
Areas and Seasons E-l
Appendix F - Distributions of Maximum Daily and Hourly Daylight PM light extinction - Under
"Just Meet" Conditions F-l
-------
Appendix G - ADditional Information on the Exclusion of Daylight Hours with Relative Humidity
Greater Than 90 Percent G-l
APPENDIX H-Inter-Year Variability H-l
APPENDIX I-Daylight Hours 1-1
in
-------
LIST OF TABLES
Table 2-1. VAQ of Denver photos substantively misclassified by majority of participants 2-5
Table 2-2. Summary of photographs used in British Columbia study 2-9
Table 2-3 Logit Analysis Results 2-28
Table 2-4 Logit model estimated VAQ values corresponding to various percent acceptability
values for the four cities 2-29
Table 3-1. Annual Mean Reconstructed 24-hour Light Extinction Estimates 3-7
Table 3-2. Urban Visibility Assessment Study Areas 3-11
Table 3-3. PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the Tacoma
Study Area 3-14
Table 3-4. Number of days per quarter in each study area 3-15
Table 3-5 Assumed daylight hours by season (Local Standard Time) 3-28
Table 3-6 Comparison of Meteorological Parameters for Daylight Hours with Relative Humidity
Greater than 90 Percent and Other Daylight Hours, During 2005 -2007 3-31
Table 3-7 Percentage of daily maximum hourly values and individual hourly values of daylight
PM light extinction exceeding CPLs (excluding hours with relative humidity
greater than 90 percent) 3-39
Table 4-1. Alternative Secondary NAAQS Scenarios for PM Light Extinction 4-2
Table 4-2. Current Conditions PM light extinction design values for the study areas 4-5
Table 4-3. Percentage reductions in non-PRB light extinction required to "just meet" the
NAAQS scenarios based on measured light extinction 4-9
Table 4-4. Percentage reductions required in non-PRB PM2.5 mass to "just meet" NAAQS
scenarios based on annual and 24-hour PM2.5 mass 4-12
Table 4-5. PM light extinction design values for "just meeting" secondary NAAQS scenarios
based on measured PM light extinction (excluding hours with relative humidity
greater than 90 percent) 4-17
Table 4-6. PM light extinction design values for "just meeting" secondary NAAQS scenarios
based on PM2.5 mass (excluding hours with relative humidity greater than 90
percent) 4-18
Table 4-7. Percentage of days across three years (two in the case of Phoenix and Houston) with
maximum 1-hour daylight PM light extinction above CPLs when "just meeting"
the NAAQS scenarios 4-19
January 2010 iv DRAFT - Do Not Quote or Cite
-------
LIST OF FIGURES
Figure 1-1 Progression from PM characteristics to PM light extinction that shows the modeling
approach (shaded light green) as well as the use of direct measurements (shaded
blue) as alternative ways to estimate PM light extinction 1-7
Figure 1-2 Progression from PM light extinction to value of visual air quality (VAQ) 1-9
Figure 2-1 Percent of Denver participants who considered VAQ in each photograph
"acceptable." 2-4
Figure 2-2 Photograph time of day information 2-6
Figure 2-3 Denver photograph time of day results (9:00 a.m. photographs eliminated), with the
broader range (17.7 dv and 24.6 dv) of the 50% acceptability criteria shown.... 2-7
Figure 2-4 Composite Chilliwack, BC photograph shows VAQ of 14.1 dv and 34 dv 2-8
Figure 2-5 Percent of BC participants who consider VAQ in each photograph "acceptable." 2-11
Figure 2-6 Reproduction of image with the best VAQ (15 dv) used in the Phoenix study 2-13
Figure 2-7 Percent of Phoenix participants who consider VAQ in each image "acceptable." .2-14
Figure 2-8 Reproduction of the image with the best VAQ (8.8 dv) used in the Washington, DC
study 2-16
Figure 2-9 Percent of 2001 Washington participants who considered VAQ acceptable in each
image 2-17
Figure 2-10 Percent of 2009 Test 1 study participants who considered VAQ acceptable in each
image, showing the range of the lower and upper bound of 50% acceptability
criteria 2-19
Figure 2-11 Combined results of two Washington preference studies (showing 50%
acceptability criteria from 2009, Test 1) 2-20
Figure 2-12 Comparison of results from Test 1 and Test 2 (Smith and Howell, 2009) 2-21
Figure 2-13 Average quality of visibility ratings for the Washington, DC WinHaze images by
participants in Tests 1 -3 conducted by Smith and Howell (2009) 2-22
Figure 2-14 Comparison of results from the Smith and Howell (2009) Test 1 and Test 3 2-24
Figure 2-15 Composite results from Smith and Howell (2009) Tests 1 and 3, and Abt (2001)
Washington, DC pilot study 2-25
Figure 2-16 Summary of results of urban visibility studies in four cities, showing the identified
range of the 50% acceptance criteria 2-26
Figure 3-1. Annual average and 24-hour (98* percentile 24-hour concentrations) PM2.5
concentrations in ug/m3, 2007 3-2
Figure 3-2. Reconstructed 24-hour light extinction in U.S. urban areas in 2003 3-6
Figure 3-3. Isopleth map of annual total reconstructed particulate extinction based on 2000-2004
IMPROVE data 3-8
Figure 3-4. January and August monthly average diurnal profiles of PM2.s components derived
from the 2004 CMAQ modeling platform, for the Detroit study area 3-18
Figure 3-5. Sequence of steps used to estimate hourly PM2.5 components and PM light extinction
3-22
Figure 3-6. Example from Detroit study area 3-23
Figure 3-7. Distribution of PM parameters and relative humidity across the 2005-2007 period,
by study area 3-33
January 2010 v DRAFT - Do Not Quote or Cite
-------
Figure 3-8. Distributions of estimated daylight 1-hour PM light extinction and maximum daily
daylight 1-hour PM light extinction across the 2005-2007 period, by study area
(excluding hours with relative humidity greater than 90 percent) 3-37
Figure 3-9. Distributions of 1-hour PM light extinction levels by daylight hour across the 2005-
2007 period, by study area (excluding hours with relative humidity greater than
90 percent) 3-40
Figure 3-10. Distributions of 1-hour relative humidity levels by daylight hour across the 2005-
2007 period, by study area (excluding hours with relative humidity greater than
90 percent) 3-42
Figure 3-11. Scatter plot of daylight 1-hour relative humidity (percent) vs. reconstructed PM
light extinction (Mm"1) across the 2005-2007 period, by study area (excluding
hours with relative humidity greater than 90 percent) 3-43
Figure 3-12 Tile Plots of Hourly PM Light Extinction 3-46
Figure 3-13 Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily 1-
hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours
3-66
Figure 4-1 Comparison of Daily Max and All Daylight Hour Design Values 4-6
Figure 4-2 Comparison of Required Percentage Reductions in Non-PRB PM Light Extinction
Needed to Meet NAAQS Scenarios 4-10
Figure 4-3 Distributions of daily maximum daylight 1-hour PM light extinction under two "just
meeting" secondary NAAQS scenarios (excluding hours with relative humidity
greater than 90 percent) 4-15
January 2010 vi DRAFT - Do Not Quote or Cite
-------
LIST OF ACRONYMS/ABBREVIATIONS
AQS
BAM
BC
CAA
CAIR
CASAC
CBS A
CCN
CDPHE
CMAQ
CONUS
CPL
CRA
CSA
CSN
CTM
ORE
dv
EPA
FEM
FRM
GEOS
IMPROVE
ISA
Km
LCD
LOESS
Mm
MSA
N
NAAQS
NARSTO
NCEA
NOAA
EPA's Air Quality System
Beta Attenuation Mass Monitor
British Columbia
Clean Air Act
Clean Air Interstate Rule
Clean Air Scientific Advisory Committee
Consolidated Business Statistical Area
Cloud Condensation Nuclei
Colorado Department of Public Health and Environment
Community Multiscale Air Quality
CMAQ simulations covering continental US
Candidate Protection Level
Charles River Associates
Consolidated Statistical Area
Chemical Speciation Network
Chemical Transport Model
Direct Radiative Effects
deciview
United States Environmental Protection Agency
Federal Equivalent Method
Federal Reference Method
Global Scale Air Circulation Model
Interagency Monitoring of Protected Visual Environment
Integrated Science Assessment
Kilometer
Liquid Crystal Display
Locally weighted Scatter Plot Smoothing
Megameter
Metropolitan Statistical Area
Nitrogen
National Ambient Air Quality Standards
North American Research Strategy for Tropospheric Ozone
National Center for Environmental Assessment
National Oceanic and Atmospheric Administration
Ianuary2010
vn
DRAFT - Do Not Quote or Cite
-------
NOx
NFS
NRC
NWS
OAQPS
OAR
OMB
ORD
PA
PM
PM2.5
PMiQ-2.5
PRB
REA
RF
RH
SANDWICH
SEARCH
SMOKE
S
S02
sox
STP
TEOM
UBC
UFVA
VAQ
Nitrogen oxides
National Park Service
National Research Council
National Weather Service
Office of Air Quality Planning and Standards
Office of Air and Radiation
Office of Management and Budget
Office of Research and Development
Policy Assessment
Particulate Matter
Particles with a 50% upper cut-point of 2.5 um aerodynamic
diameter and a penetration curve as specified in the Code of
Federal Regulations.
Particles with a 50% upper cut-point of 10± 0.5 um aerodynamic
diameter and a penetration curve as specified in the Code of
Federal Regulations.
Particles with a 50% upper cut-point of 10 um aerodynamic
diameter and a lower 50% cut-point of 2.5 um aerodynamic
diameter.
Policy Relevant Background
Risk and Exposure Assessment
Radiative Forcing
Relative Humidity
Skilfate, Adjusted Nitrate, Derived Water, Inferred Carbonaceous
mass approach
Southeastern Aerosol Research and Characterization Study
Sparse Matrix Operator Kernal Emissions
Sulfur
Sulfur Dioxide
Sulfur Oxides
Standard Temperature and Pressure
Tapered Element Oscillating Microbalance
University of British Columbia
Urban-Focused Visibility Impact Assessment
Visual Air Quality
January 2010
Vlll
DRAFT - Do Not Quote or Cite
-------
1 1 INTRODUCTION
2 The U.S. Environmental Protection Agency (EPA) is presently conducting a review of
3 the national ambient air quality standards (NAAQS) for particulate matter (PM). Sections 108
4 and 109 of the Clean Air Act (Act) govern the establishment and periodic review of the NAAQS.
5 The NAAQS are to be based on air quality criteria, which are to accurately reflect the latest
6 scientific knowledge useful in indicating the kind and extent of identifiable effects on public
7 health or welfare that may be expected from the presence of the pollutant in ambient air. The
8 EPA Administrator is to promulgate and periodically review, at no later than five-year intervals,
9 "primary" (health-based) and "secondary" (welfare-based) NAAQS for such pollutants. Based
10 on periodic reviews of the air quality criteria and standards, the Administrator is to make
11 revisions in the air quality criteria and standards, and to promulgate any new standards, as may
12 be appropriate. The Act also requires that an independent scientific review committee advise the
13 Administrator as part of this NAAQS review process, a function performed by the Clean Air
14 Scientific Advisory Committee (CASAC).
15 The current NAAQS for PM are a suite of identical primary and secondary standards
16 established to provide protection from health and welfare effects related to fine and coarse
17 particles, using PM2.5 and PMio as indicators, respectively (71 FR 61144, October 17, 2006).
18 With regard to the primary standards for fine particles, in 2006 EPA revised the level of the 24-
19 hour PM2.5 standard to 35 ug/m3 (calculated as a 3-year average of the 98 percentile of 24-hour
20 concentrations at each population-oriented monitor), retained the level of the annual PM2.5 annual
21 standard at 15 ug/m3 (calculated as the 3-year average of the weighted annual mean PM2.5
22 concentrations from single or multiple community-oriented monitors), and revised the form of
23 the annual PIVb.s standard by narrowing the constraints on the optional use of spatial averaging1.
24 With regard to the primary standards for PMio, EPA retained the 24-hour PMio standard at 150
25 ug/m3 (not to be exceeded more than once per year on average over 3 years) and revoked the
26 annual standard because available evidence generally did not suggest a link between long-term
27 exposure to current ambient levels of coarse particles and health effects. The 2006 primary
28 standards were based primarily on a large body of epidemiological evidence relating ambient PM
29 concentrations to various adverse health outcomes.
30 The 2006 secondary standards for PM2.5 and PMio were set to be identical to the primary
31 standards, on the basis that these standards would, in conjunction with the Regional Haze
1 In the revisions to the PM NAAQS finalized in 2006, EPA tightened the constraints on the spatial averaging option
limiting the conditions under which some areas may average measurements from multiple community-oriented
monitors to determine compliance (see 71 FR 61165-61167, October 17, 2006).
January 2010 1-1 DRAFT - Do Not Quote or Cite
-------
r\
1 Program , provide appropriate protection to address PM-related welfare effects, including
2 visibility impairment, effects on vegetation and ecosystems, materials damage and soiling, and
3 effects on climate change. (As noted below, this judgment was reversed and remanded by the
4 Court of Appeals for the District of Columbia Circuit.)
5 The next periodic review of the PM NAAQS is now underway.3 In the Integrated
6 Review Plan for the National Ambient Air Quality Standards for Paniculate Matter, March 2008
7 (US EPA, 2008a), EPA outlined the science policy questions that will frame this review, outlined
8 the process and schedule that the review will follow, and provided more complete descriptions of
9 the purpose, contents, and approach for developing the key documents that will be developed in
10 the review.4 EPA has recently completed the process of assessing the latest available policy-
11 relevant scientific information to inform the review of the PM standards. The final assessment is
12 contained in the final Integrated Science Assessment for Particulate Matter (ISA, US EPA,
13 2009a) which was released in December 2009. The final PM ISA includes a summary of the
14 scientific evidence for the relationship of PM to visibility effects, remote area and urban haze
15 conditions, the PM components responsible for visibility impacts, and studies of public
16 preference with respect to urban visibility conditions.
17 Building upon the visibility effects evidence presented in the PM ISA, as well as CAS AC
18 advice (Samet, 2009a and b) and public comments on the plan for and first draft of the UFVA
19 (US EPA, 2009b, c), EPA's Office of Air Quality Planning and Standards (OAQPS) has
20 developed this second draft Urban-Focused Visibility Assessment (UFVA) describing the
21 quantitative assessments conducted by the Agency to support the review of the secondary PM
22 standards. This draft document presents the methods, key results, observations, and related
23 uncertainties associated with the quantitative analyses performed. Revisions to this second draft
24 UFVA draw upon the final ISA and reflect consideration of CAS AC and public comments on the
25 first draft UFVA, as described in section 1.2 below.
26 The final ISA and final UFVA will inform the policy assessment and rulemaking steps
27 that will lead to final decisions on the secondary PM NAAQS. A draft Policy Assessment (PA)
28 is now being prepared by OAQPS staff to provide a transparent staff analysis of the scientific
2 See http://www.epa.gov/air/visibilitv/program.html for more information on EPA's Regional Haze Program.
3 See http://www.epa. gov/ttn/naaqs/standards/pm/s_pm index,html for more information on the current and
previous PM NAAQS reviews.
4 On November 30, 2007, EPA held a consultation with the Clean Air Scientific Advisory Committee (CASAC) on
the draft IRP (Henderson, 2008). Public comments were also requested on the draft plan and presented at that
CASAC teleconference. The final IRP incorporated comments received from CASAC and the general public on the
draft plan as well as input from senior Agency managers. CASAC is an independent scientific advisory committee
established to meet the requirements of section 109(d)(2) of the Clean Air Act See
http://vosemite.epa.gov/sab/sabpeople.nsf/WebCommittees/CASAC for more information, and, in particular,
information on the CASAC PM Review Panel activities.
January 2010 1-2 DRAFT - Do Not Quote or Cite
-------
1 basis for alternative policy options for consideration by senior EPA management prior to
2 rulemaking. The PA is intended to help "bridge the gap" between the Agency's scientific
3 assessments, presented in the ISA and UFVA, and the judgments required of the Administrator
4 in determining whether it is appropriate to retain or revise the secondary PM standards. The PA
5 will integrate and interpret information from the ISA and the UFVA to frame policy options and
6 to facilitate CASAC's advice to the Agency and recommendations on any new standards or
7 revisions to existing standards as may be appropriate, as provided for in the Clean Air Act. A
8 very preliminary draft PA was released in September 2009 to facilitate discussion on the overall
9 structure, areas of focus, and level of detail to be included in a first external review draft PA
10 document, which EPA plans to release for CAS AC review and public comment in February of
11 2010. This preliminary draft PA was discussed in conjunction with CAS AC review of and
12 public comment on the second draft ISA, first draft UFVA, and first draft health risk assessment
13 documents produced in support of this PM NAAQS rulemaking.
14 1.1 PM NAAQS BACKGROUND
15 In the review of the secondary PM NAAQS completed in 2006, EPA took into account
16 that the Regional Haze Program, authorized under sections 169 A and 169B of the CAA, was
17 established to address all human-caused visibility impairment in federal Class I areas. The
18 national goal of this program is to prevent any future, and remedy any existing, impairment of
19 visibility in mandatory class I Federal areas (Class I areas) which impairment results from
20 manmade air pollution. This program also mandates that states develop SIPs to ensure that
21 reasonable progress is made towards meeting those goals. Because Congress explicitly targeted
22 Class I areas for this pristine level of protection, it can be concluded that Congress did not
23 envision such a stringent goal in non-Class I areas. See American Trucking Ass'n v. Browner,
24 175 F. 3d 1027, 1056-57 (D. C. Cir. 2002) (upholding this position). However, Congress
25 recognized that visibility impairment can and often does occur in areas outside federal Class I
26 areas, including urban areas and judged that protection from visibility impairment was important
27 in those areas as well. In this regard, Congress included visibility effects in the definition of
28 public welfare effects that should be protected under the national ambient air quality standards
29 (NAAQS) program authorized in sections 108 and 109 of the CAA. As a result, EPA may
30 establish secondary standards addressing visibility impairment notwithstanding existence of the
31 Regional Haze Program. Under the NAAQS program, it is up to the Administrator to judge what
32 is the requisite level of public welfare visibility protection.
33 Recognizing that efforts were underway to provide increased protection to Class I areas
34 under the Regional Haze Program, EPA focused the 2006 PM NAAQS review on visibility
35 impairment in non-Class I areas. Because most of the available non-Class I PM data came from
January 2010 1-3 DRAFT - Do Not Quote or Cite
-------
1 PM monitoring sites located primarily in urban areas, the assessments took on an urban focus. In
2 addition, EPA considered available information on people's preferences for different levels of
3 visual air quality which came from studies conducted in urban areas and from existing urban
4 visibility programs and goals.
5 In an effort to minimize the factors that historically had complicated efforts to address
6 visibility impairment nationally, given the substantial East/West differences observed in Class I
7 areas, EPA staff noted that with respect to fine particles, East/West differences are substantially
8 smaller in urban than in rural areas. Further, relative humidity levels, though generally higher in
9 eastern than western areas, are appreciably lower in both regions during daylight as compared to
10 nighttime hours. The PIVb.s data available at that time in urban areas were obtained using a filter
11 -based Federal Reference Method (FRM) which captures ambient PM2.5 on a filter and then dries
12 it to get the dry PM2.5 mass concentration. By drying the sample, most water and to some extent
13 other labile PM compounds evaporate so that the original characteristics (e.g., particle size and
14 composition) of the ambient PM are altered. Using PM and meteorological data from 161 cities,
15 EPA staff assessed the correlations between PM2.5 levels and reconstructed light extinction (RE)
16 during daylight hours for different regions of the country. This assessment showed that the
17 strongest correlation in the relationship of ambient PM light extinction to dry PM2.5 mass
18 concentration was during afternoon periods when lower relative humidity conditions generally
19 prevailed in all regions of the country and ambient PM was drier (US EPA, 2005). While EPA
20 recognized that the effect of ambient PM on visibility results from the ambient particle
21 characteristics of size, concentration, and composition (including associated water) present in the
22 air in the sight path of the observer, given the data availability at the time, EPA viewed the FRM
23 altered PM2.5 mass concentration as a permissible indicator for addressing ambient PM-related
24 visibility effects at the national scale during afternoon hours. Thus, the 2005 Staff Paper chose
25 to address the issue in terms of averaging time rather than indicator, discussing the use of a sub-
26 daily afternoon dry PM2.5 standard, because the generally lower afternoon relative humidity
27 tended to produce a more uniform relationship between light extinction and dry PM2.5 mass
28 concentration throughout the country, therefore providing a more uniform level of visibility
29 protection nationwide. This more uniform level of visibility protection, however, was limited to
30 the afternoon hours of the day when relative humidity and visibility impairment tend to be the
31 lowest.
32 Based on the above, in the 2005 PM Staff Paper, EPA staff recommended a separate sub-
33 daily secondary standard to address visibility impairment using dried PM2.5 mass concentration
34 as the indicator, a recommendation endorsed by CASAC. In the 2006 proposal notice, however,
35 EPA proposed to revise the secondary standards by making them identical to the suite of
36 proposed primary standards for fine and coarse particles, to provide protection against PM-
January 2010 1-4 DRAFT - Do Not Quote or Cite
-------
1 related public welfare effects including visibility impairment, effects on vegetation and
2 ecosystems, materials damage and soiling, and climate, while soliciting comment on adding a
3 new sub-daily PM2.5 secondary standard to address visibility impairment primarily in urban areas
4 (71 FR 2620). CAS AC provided additional advice to EPA in a letter to the Administrator
5 requesting reconsideration of CASAC's recommendations for both the primary and secondary
6 PM2.5 standards as well as standards for thoracic coarse particles (Henderson, 2006). With
7 regard to the secondary standard, CAS AC reaffirmed "... the recommendation of Agency staff
8 regarding a separate secondary fine particle standard to protect visibility.... the CASAC wishes
9 to emphasize that continuing to rely on primary standards to protect against all PM-related
10 adverse environmental and welfare effects assures neglect, and will allow substantial continued
11 degradation, of visual air quality over large areas of the country" (Henderson, 2006).
12 On September 21, 2006, EPA announced its final decisions to provide increased
13 protection of public welfare by making the secondary NAAQS identical to the revised primary
14 standards (71 FR 61144, October 17, 2006). This suite of secondary standards was designed to
15 address both visibility and other non-visibility welfare related effects. Specifically, with regard
16 to the secondary welfare effect of visibility impairment, the Administrator believed that revising
17 both the 24-hour and annual PM2.5 secondary standards to be identical to the revised suite of
18 PM2.5 primary standards was a reasonable policy approach to address visibility impairment
19 primarily in urban areas. In particular, EPA revised the level of the 24-hour PM2.5 standard to 35
20 ng/m3, retained the level of the annual PM2.5 standard at 15 |ig/m3, and revised the form of the
21 annual PM2.5 standard by narrowing the constraints on the optional use of spatial averaging.
22 With regard to the other non-visibility PM-related welfare effects such as vegetation and
23 ecosystems, materials damage and soiling, and climate, the Administrator concluded that it was
24 appropriate to address these effects by revising the current suite of PM2.5 secondary standards,
25 making them identical in all respects to the suite of primary PM2.5 standards, while retaining the
26 current 24-hour PMio secondary standard and revoking the current annual PMio secondary
27 standard. In particular for coarse particles, EPA retained PMio as the indicator for purposes of
28 regulating the coarse fraction of PMio and retained the 24-hour secondary PMio standard at 150
29 |ig/m3 and revoked the annual secondary PMio standard.
30 Several parties filed petitions for review following promulgation of the revised PM
31 NAAQS in 2006. These petitions addressed a number of issues, including the decision to set the
32 secondary PM2.s standards identical to the primary standards. On judicial review the court
33 remanded the secondary PM2.5 NAAQS to EPA because the Agency failed to adequately explain
34 why setting the PM2.5 secondary standards equal to the primary PM2.5 standards provided the
35 required protection from visibility impairment. In particular, the Agency failed to identify a
36 target level of visibility impairment that would be requisite to protect the public welfare, and
January 2010 1-5 DRAFT - Do Not Quote or Cite
-------
1 improperly relied on a misleading comparison of the number of counties which would be in
2 nonattainment for the revised primary NAAQS compared to one alternative secondary standard
3 under consideration. Among other things, this equivalence analysis failed to address the issue of
4 regional differences in humidity-related effects on visibility. American Farm Bureau Federation
5 v. EPA, 559 F. 3d 512, 530-31 (D.C. Cir. 2009).
6 1.2 VISIBILITY EFFECTS SCIENCE OVERVIEW
7 Light extinction is the loss of light per unit of distance and occurs when light is scattered
8 and/or absorbed. Particulate matter and gases can both scatter and absorb light. Light scattering
9 by gases (e.g., nitrogen, oxygen, etc.) that comprise the pollutant free or clean atmosphere (also
10 known as Rayleigh or clean-air scattering) is related to the density of the air, which is
11 sufficiently constant with elevation that it can be taken to be a time invariant constant that
12 depends principally on elevation above sea level. NC>2 is the only atmospheric pollutant gas that
13 absorbs light appreciably and its effects are generally small (i.e., less than 5%) compared to PM
14 light extinction. Hereinafter the phrase "PM light extinction" indicates that the Rayleigh
15 contribution to light extinction (nominally considered 10 Mm"1) has been subtracted out and the
16 NC>2 contribution is considered negligible or is simply excluded due to the measurement
17 approach used. By contrast, the term "light extinction" or "total light extinction" is meant to
18 include both the Rayleigh and NO2 contributions.
19 Visual air quality is defined as the visibility effect caused solely by air quality conditions
20 and excluding those associated with meteorological conditions like fog and precipitation. It is
21 commonly measured as either light extinction (in terms of inverse megameters, Mm"1) or the
22 haziness index (in terms of deciview, dv) (Pitchford and Malm, 1993). The haziness index
23 measured in deciview units was developed for use in visibility perception studies because it has a
24 more linear relationship to perceived changes in haze compared with light extinction. It is
25 defined as ten times the natural logarithmic of one tenth of the light extinction in inverse
26 megameter units (Mm"1) (Pitchford and Malm, 1993). Light extinction and haziness are physical
27 measures of the amount of visibility impairment (e.g., the amount of "haze"), with both
28 increasing as the amount of haze increases.
29 PM is a heterogeneous mixture of particles of different sizes and chemical compositions.
30 While visibility impairment has been associated most often with PM2.5, larger particles such as
31 those found in PMio may be a significant contributor in some areas. Thus, UFVA considers the
32 visibility impairment caused by all particles 10 microns or smaller. As stated above, the degree
33 of visibility impairment caused by a given mass of PM depends in large part on the size, density
34 and chemical composition of the PM. If the ambient PM has a large number of hygroscopic
35 particles, and also occurs when the relative humidity of the air is higher, those particles will be
January 2010 1-6 DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
larger in size so that the PM will have a larger haze effect than if PM with the same
concentration and composition minus the water was present and the ambient air had lower
relative humidity.
As shown in Figure 1-1, the ambient PM light extinction can be estimated from dry PM
mass and composition data and relative humidity using an algorithm that accounts for water
present in hygroscopic PM components and uses assumed light extinction efficiencies for each of
Figure 1-1 Progression from PM characteristics to PM light extinction that shows the
modeling approach (shaded light green) as well as the use of direct measurements (shaded
blue) as alternative ways to estimate PM light extinction.
Measured Quantities
Derived Characteristics
12
13
14
Dry PM Mass
Concentration
Dry PM Composition
^
DryPM
Characteristics
>
t
Ambient Relative
Humidity
Hygroscopic
Growth Model
Ambient PM
Characteristics
Light Extinction
Model
Ambient PM Light
Extinction
^
*
Ambient PM Light
Extinction
the major PM species. Ambient PM light extinction is most accurately determined by direct
measurements. However, because there is limited ambient PM light extinction data available in
January 2010
1-7
DRAFT - Do Not Quote or Cite
-------
1 urban areas, the assessments below will principally use monitored and modeled dry PM mass
2 and species estimates, along with relative humidity measurements as input to a simple algorithm
3 for estimating ambient PM light extinction.
4 The extent to which any amount of light extinction affects a person's ability to view a
5 scene depends on both scene and light characteristics. For example the appearance of a nearby
6 object (i.e., a building) is generally less sensitive to a change in light extinction than the
7 appearance of a similar object at a greater distance. For a scene with known characteristics, the
8 degradation in the scene associated with a change in light extinction can be determined and the
9 resulting appearance can be realistically displayed on a digital photograph of the scene using the
10 WinHaze system. Figure 1-2 below shows the progression from PM light extinction to perceived
11 visual air quality impacts to the valuation of those perceived impacts.
12 Survey studies have used sets of photographs or computer simulated images developed
13 from a base photo depicting a range of visibility conditions on urban scenes to assess the
14 individual's opinion on the acceptability of conditions. For the specific scenes used in such
15 studies there is a known or predetermined one-to-one correspondence between the computer
16 generated haze in the photographs and the associated amount of ambient PM light extinction.
17 For visibility preference studies, visibility levels are generally characterized using the haze index
18 in units of deciview (similar to the decibel scale for sound).
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
January 2010 1-8 DRAFT - Do Not Quote or Cite
-------
1 Figure 1-2 Progression from PM light extinction to value of visual air quality
2 (VAQ)
4
5
6
7
8
9
10
11
12
13
14
Relationship Steps from PM
Light Extinction to VAQ
Non-Air Quality
Information
Ambient PM Light
Extinction
WinHaze
Modeling
Scene and Lighting
Characteristics
Perceived Visual Air
Quality (Images)
f Valuation Studies \<-
Public-Scene
Contextual
Information
Value of Improved
VAQ
1.3 GOALS AND APPROACH
The principal goal of the UFVA is to characterize recent levels of visibility impairment in 15
urban areas, as well as "just meet" scenarios for both the current secondary PM2.5 standards, as
well as various alternative standards, including those which utilize a different indicator, and a
range of forms that may better reflect the relationship between PM and visibility impairment. In
particular, this UFVA focuses on the use of a PM light extinction-based indicator for a possible
secondary PM NAAQS (see Figure 1-1 and 1-2). This is done by comparing estimates of hourly
PM light extinction in 15 major U.S. urban areas over the three-year period 2005-2007 to the
candidate protection levels (CPLs), which are a range of light extinction values beyond which
half of the participants in assessed urban visibility preference studies indicated the haze
conditions were unacceptable (see discussion in chapter 2 below and Stratus Consulting Inc.,
January 2010
1-9
DRAFT - Do Not Quote or Cite
-------
1 2009). In addition, this second draft UFVA includes additional characterizations of the
2 effectiveness of a sub-daily PM2.5 mass concentration indicator, which was explored in the 2005
3 PM staff paper and which was considered a viable option by EPA staff and CAS AC in the 2006
4 review. These latter assessments are summarized in Appendix D.
5 The previous PM NAAQS review used the results of visibility preference survey studies
6 conducted in Denver (1990), Phoenix (2003), and British Columbia (1993) as the basis for
7 suggesting that a standard set to protect visibility conditions to a level within a visual range from
8 between about 40 km to about 60 km (corresponding to light extinction from -100 Mm"1 to -67
9 Mm"1) could represent an appropriate degree of welfare protection from PM5. With the
10 exception of a small pilot study conducted in Washington, DC in 2001 (9 participants; Abt
11 Associates Inc., 2001), and a replicate study also conducted for Washington, DC in 2009 (26
12 participants; Smith and Howell, 2009), there are no additional visibility preference survey studies
13 upon which to base the selection of CPLs.
14 The EPA staff, with contractor support, has conducted a more detailed, in-depth
15 assessment of the results from these studies, including the two Washington, DC studies. This
16 assessment includes an analysis that combines data from across all studies using graphical and
17 logit model analysis to examine the consistency of the results between the surveys (Stratus
18 Consulting Inc., 2009). Based on the results of this analysis, we have been able to refine the
19 range of visibility conditions that could represent an appropriate degree of public welfare
20 visibility protection that was put forth in the 2006 review, and to determine a central tendency
21 value for the CPLs. These analyses and results are described below in chapter 2.
22 In the previous PM NAAQS review, the characterization of urban visibility conditions
23 were based on IMPROVE algorithm estimates using the 2001 to 2003 PM2.5 mass and speciation
24 data from 161 urban areas by assuming a constant composition for every hour of the day equal to
25 the 24-hour measured composition and by using either actual or monthly average (10-year mean)
26 hour of the day relative humidity. Statistical relationships between hourly light extinction
27 estimates and concurrent hourly PM2.5 mass concentrations were used to show that daytime and
28 especially afternoon relationships are relatively strong with a similar linear relationship for both
29 eastern and western urban areas (i.e. R2>0.6, slope -6 m2/g).
30 The current assessment of urban visibility conditions (as described in chapter 3) uses a
31 modeling approach to estimate hourly light extinction using PM2.5 mass and speciation data with
32 measured relative humidity. However, it differs by replacing the unrealistic assumption of
33 constant composition for PM2.5, with composition that is made to vary during the day using
34 urban-specific monthly mean diurnal variations of species concentrations determined from
5 Light extinction is inversely related to visual range.
January 2010 1-10 DRAFT - Do Not Quote or Cite
-------
1 regional air quality model results, while constraining the means of the hourly species
2 concentration for each day to closely match the 24-hour duration measured species
3 concentrations.
4 1.4 SCOPE OF URBAN-FOCUSED VISIBILITY ASSESSMENT
5 This section provides an overview of the scope and key design elements of the UFVA,
6 including the process that has been followed to design the analyses. Following initiation of this
7 PM NAAQS review in 2007, we began the design of the assessments in the UFVA by revisiting
8 the analyses completed during the previous PM NAAQS review (Abt Associates Inc., 2001; US
9 EPA, 2005, chapter 6) with an emphasis on considering key limitations and sources of
10 uncertainty recognized in that review.
11 1.4.1 Background
12 As an initial step in this review, EPA invited a wide range of external experts as well as
13 EPA staff, representing a variety of areas of expertise to participate in a workshop titled,
14 "Workshop to Discuss Policy-Relevant Science to Inform EPA's Integrated Plan for the Review
15 of the Secondary PM NAAQS" (72 FR 34005, June 20, 2007). This workshop provided an
16 opportunity for the participants to broadly discuss the key policy-relevant issues around which
17 EPA would structure the PM NAAQS review and to discuss the most meaningful new science
18 that would be available to inform our understanding of these issues. One session of this
19 workshop centered on issues related to visibility impacts associated with ambient PM.
20 Specifically, the discussions focused on the extent to which new research and/or improved
21 methodologies were available to inform how EPA evaluated visibility impairment in this review.
22 Based in part on these workshop discussions, EPA developed a draft IRP outlining the
23 schedule, the process, and the key policy-relevant science issues that would guide the evaluation
24 of the air quality criteria for PM and the review of the primary and secondary PM NAAQS,
25 including initial thoughts for conducting quantitative assessments (US EPA, 2007, chapter 6).
26 On November 30, 2007, CAS AC held a teleconference with EPA to provide its comments on the
27 draft IRP (72 FR 63177, November 8, 2007). Public comments were also presented at that
28 teleconference. A final IRP incorporating comments received from CASAC and the general
29 public on the draft plan was issued in March 2008 (US EPA, 2008a).
30 In articulating a rationale for the urban focus of this assessment, we reviewed the
31 available information and found the following information compelling: 1) PM levels in urban
32 areas are often in excess of those of the surrounding region since urban haze typically includes
33 both regional and local contributions (US EPA, 2009a; sections 9.2.3.3 and 9.2.3.4), suggesting
34 the potential for higher levels of PM-induced visibility impairment in urban areas; 2) the
January 2010 1-11 DRAFT - Do Not Quote or Cite
-------
1 existence of numerous urban visibility protection programs and goals demonstrating that urban
2 VAQ is noticed and considered an important value to urban residents (US EPA, 2009a; section
3 9.2.4); and 3) the existence of large urban populations means that potentially more people are
4 routinely affected by poor VAQ than in rural areas. These features of urban areas have led EPA
5 staff to conclude that urban dwellers represent a susceptible population group for adverse PM-
6 related effects on visibility. However, this conclusion is not meant to imply that there are not
7 other susceptible populations or individuals living in other non-urban and non-Class I areas that
8 are currently adversely impacted by ambient PM-related visibility conditions. Unfortunately,
9 visibility preferences and PM levels in these areas have not been well characterized. Although
10 this visibility assessment focuses only on selected urban areas, a new secondary PM standard would
11 apply to all non-Class I areas of the country.
12 On October 6-8, 2008 the EPA sponsored an urban visibility workshop in Denver,
13 Colorado to identify and discuss methods and materials that could be used in "next step" projects
14 to develop additional information about people's preferences for reducing existing impairment of
15 urban visibility, and about the value of improving urban visibility. Invited individuals came
16 from a broad array of relevant technical and policy backgrounds, including visual air quality
17 (VAQ) science, sociology, psychology, survey research methods, economics, and EPA's process
18 of setting NAAQS. The 23 people who attended the workshop (including one via teleconference
19 line) came from EPA, the National Oceanic and Atmospheric Administration (NOAA), National
20 Park Service, academia, regional and state air pollution planning agencies, and consulting firms.6
21 The information discussed at this Workshop was useful in informing subsequent steps in the
22 process.
23 1.4.2 Selection of Alternative Scenarios for First Draft Assessments
24 In designing the quantitative assessments to include in the first draft UFVA, EPA staff
25 developed a planning document outlining the initial design for the PM NAAQS visibility
26 assessment - Paniculate Matter National Ambient Air Quality Standards: Scope and Methods
27 Plan for Urban Visibility Impact Assessment, henceforth Scope and Methods Plan (US EPA,
28 2009b). This planning document was released for CASAC consultation and public review in
29 February 2009. Based on consideration of CASAC and public comments on the Scope and
30 Methods Plan, along with ongoing review of the latest PM-related literature, several aspects of
31 the original scope of the urban visibility conditions assessment, as depicted in Figure 1-1 of
32 section 1.3 of the Scope and Methods document (US EPA, 2009b), were modified in the first
33 draft UFVA (US EPA, 2009c). Taking into account the nature of urban versus more remote area
6 To view the complete report from the October 2008 urban visibility workshop, see:
http://vista.cira.colostate.edu/improve/Publications/GrayLit/grav literature.htm
January 2010 1-12 DRAFT - Do Not Quote or Cite
-------
1 PM composition, and input received at the April 2, 2009 CAS AC meeting, EPA staff concluded
2 that it was unnecessary to develop a new urban-optimized algorithm at this time and that it
3 remained appropriate in the context of this assessment to use the original IMPROVE algorithm
4 to relate urban PM to local haze (PM light extinction). One of the primary reasons for initially
5 considering an urban-optimized algorithm was a concern that the organic components of PM in
6 urban areas, being generally nearer their emission sources, would have a lower ratio to the
7 measured organic carbon mass than the ratio of organic component mass to measured organic
8 carbon mass currently used for the more aged PM organic components found in remote areas.
9 As described below in chapter 3, this concern has been addressed by using the SANDWICH
10 mass balance approach to estimate the PM organic component mass, which negates the need to
11 estimate organic component mass from measured organic carbon mass.
12 With regard to the urban visual air quality preference assessment described in the Scope
13 and Methods document (US EPA, 2009b, section 1.3), more significant modifications occurred.
14 EPA staff decided to conduct a reanalysis of the urban visibility preference studies available at
15 the time of the 2006 PM NAAQS review, rather than conduct new public preference studies, as it
16 has become apparent that the results of these studies would be unlikely to be completed in time
17 to inform this review. Recognition that the initial plans described in the Scope and Methods
18 document were possibly overly ambitious was also shared by members of CAS AC (see
19 individual member comments; Samet, 2009a). The analysis, therefore, relied on pre-existing,
20 rather than new, urban visibility preference studies and was designed to explore the similarities
21 and differences (comparability) among these studies. Information drawn from these results
22 informed the selection of VAQ candidate protection levels (CPLs) (described in chapter 2 below)
23 to be used in subsequent impact assessments. Further, information presented during the public
24 comment phase of the April 2, 2009 CAS AC meeting and later provided to EPA staff, led to the
25 inclusion of a recent study by Smith and Howell (2009) for Washington, DC in the reanalysis.
26 1.4.3 Selection of Alternative Scenarios for Second Draft Assessments
27 The first draft UFVA was reviewed at an October 2009 CASAC meeting, and a CASAC
28 letter providing its advice and recommendations was submitted to the Administrator in
29 November 2009 (Samet, 2009b). In its letter, the CASAC indicated support for EPA staffs
30 approach to evaluating the nature and degree of PM-related visibility impairment, including
31 EPA's focus on non-Class I areas, including in particular, urban areas as an "effective
32 complement" to the Regional Haze Rule. In this regard, CASAC expressed support for
33 consideration of a new PM light extinction indicator, a one hour averaging time, and for the
34 range of selected candidate light extinction levels.
35 • Indicator: PM Light Extinction
January 2010 1-13 DRAFT - Do Not Quote or Cite
-------
1 There are a number of different ways to measure ambient PM: particle counts, surface
2 area, volume, mass concentrations, and concentration of components. Each of these different
3 characteristic of ambient PM can be important in the context of different effects. For example,
4 particle count may be important from the perspective of cloud formation or to characterize the
5 abundance of ultrafme PM, which is of interest for health effects. In a similar way PM light
6 extinction measures the characteristic of ambient PM most relevant and directly related to the
7 effect of PM visibility impairment. Thus, as described in the Scope and Methods document (US
8 EPA, 2009b) and first draft UFVA, EPA staff is continuing to focus assessments in this second
9 draft document in terms of ambient PM light extinction as the indicator for PM visibility
10 impairment, instead of the traditional PM2.5 mass concentration. Unlike PM mass concentration,
11 which generally changes the composition and size of the particles by driving off most of the
12 water, ambient PM light extinction captures the PM-induced visibility impairment of the
13 particles as they exist in the atmosphere. PM light extinction, like conventional PM mass
14 concentration, is a measurable physical characteristic of atmospheric PM.
15 Section 109 (b) (2) of the CAA states that "Any national secondary ambient air quality
16 standard prescribed under subsection (a) of this section shall specify a level of air quality the
17 attainment and maintenance of which ... is requisite to protect the public welfare from any
18 known or anticipated adverse effects associated with the presence of such air pollutant in the
19 ambient air...." (emphasis added). In addition, section 108 (a) (2) states that the air quality
20 criteria "for an air pollutant shall accurately reflect the latest scientific knowledge useful in
21 indicating the kind and extent of all identifiable effects on public health or welfare which may be
22 expected from the presence of such pollutant in the ambient air, in varying quantities. The
23 criteria ... shall include information on (A) those variable factors (including atmospheric
24 conditions) which of themselves or in combination with other factors may alter the effects on
25 public health or welfare of such air pollutant;..." (emphasis added). Thus, EPA staff believes
26 that the visibility effects of PM important to the public welfare are precisely the visibility effects
27 of PM occurring in the ambient air, which necessarily entails association with ambient
28 atmospheric conditions that affect the nature or magnitude of the PM visibility effect. These
29 ambient conditions lead to constant changes in the size and composition of particles as these
30 particles come in contact with other pollutants or natural components, become oxidized/age as
31 they are transported great distances, and shrink or grow in the absence or presence of water
32 vapor, or other atmospheric gases. The combined effect of all these interactions of ambient PM
33 with real time atmospheric conditions and chemistry on the public welfare effect of visibility
34 impairment depends on factors other than dry PM mass concentration alone. Use of PM light
35 extinction as the indictor for a secondary PM NAAQS is thus a more appropriate and direct
January 2010 1-14 DRAFT - Do Not Quote or Cite
-------
1 measure of the relationship between ambient PM and the public welfare effect of visibility
2 impairment than any dry PM mass concentration (either PM2.5 or any other dry mass fraction).
O
4 • Averaging times: Daylight Daily Max. 1 Hour or All Daylight Hours
5 It is necessary to also identify an averaging time to apply along with the CPLs in the
6 assessments described in chapters 3 and 4. Because the nature of visibility impairment and its
7 impact on the public welfare is sufficiently different and less well understood at night, this
8 assessment only considers daylight hour visibility. Though not directly supported by preference
9 or other studies, EPA staff believes that a short averaging time (e.g. an hour) may be more
10 appropriate than longer time periods (e.g. multiple hours) since VAQ impacts are instantaneously
11 perceived. This is also consistent with staffs belief that most individuals in an urban setting
12 experience urban VAQ in relatively short-term incidental and intermittent periods when they
13 have the opportunity to be outdoors (e.g. during commutes to work, school, shopping, etc.).
14 Since this fraction of the public may experience poor VAQ during a relatively small time period
15 and not have the opportunity to see it improve later during the same day, it seems appropriate to
16 EPA staff to consider assessing the current and projected conditions in chapters 3 and 4 by
17 comparing the 1-hour daily maximum light extinction to each of the three CPLs supported by the
18 preference studies. There is uncertainty associated with predicting the duration of the effect
19 associated with such brief periods of exposures, i.e., it is not known how long the person
20 remembers the poor VAQ conditions once he/she goes indoors and is removed from the sight.
21 Alternately, a complementary fraction of the public may have multiple or continuing
22 opportunities to experience visibility throughout the day. People in this situation can experience
23 a variety of conditions ranging from improvement, maintenance, or diminished VAQ throughout
24 the day. For them, a day with several hours that exceed acceptable VAQ levels may represents a
25 greater impact on their wellbeing than on a day with only one such hour. To assess impacts
26 more related to this portion of the population, in which the degree of impact depends upon the
27 conditions present across multiple hours of exposure, EPA staff has also considered all daylight
28 hours which have light extinction levels beyond the three CPLs, as well as the 1-hour daily
29 maximum light extinction in the assessments described in chapters 3 and 4.
30
31 • Level: Candidate Protection Levels (CPLs)
32 In order to identify a range of light extinction levels associated with acceptable VAQ to
33 compare to current and projected conditions in the assessment in chapters 3 and 4 of this
34 document, CPLs have been selected in a range from 20 dv to 30 dv (74 Mm"1 to 201 Mm"1) based
35 on the composite results and the effective range of 50th percentile acceptability across the four
36 urban preference study areas shown in Figure 2-16. A midpoint of 25 dv (122 Mm"1) was also
January 2010 1-15 DRAFT - Do Not Quote or Cite
-------
1 selected for use in the assessment. These three values provide a low, middle, and high set of
2 light extinction conditions that are used in subsequent sections of the UFVA to provisionally
3 define daylight hours with urban haze conditions that have been judged unacceptable by the
4 participants of these preference studies. As discussed in greater detail in section 1.2 above, PM
5 light extinction is taken to be light extinction minus the Rayleigh scatter (i.e. light scattering by
6 atmospheric gases is about 10 Mm"1), so the PM light extinction levels that correspond to low,
7 middle and high CPLs are about 64 Mm"1, 112 Mm"1 and 191 Mm"1, respectively.
8
9 • Forms: Percentiles and Relative Humidity Constraints
10 In considering an appropriate range of forms to consider in the analyses of alternative PM
11 light extinction visibility standards analyzed in chapter 4 of this second draft UFVA, staff
12 considered what frequency of conditions at or below the CPLs should be considered acceptable.
13 Again, none of the preference studies provided insight into this aspect of acceptability. Because
14 the nature of the public welfare effect is one of aesthetics and/or on feelings of wellbeing and not
15 directly related to a physical health outcome, EPA staff believes that it is not necessary to
16 eliminate all such exposures and that some number of hours/days with poor VAQ can reasonably
17 be tolerated. In the first draft UFVA, staff selected the 90 and 95* percentiles to assess. In the
18 CAS AC letter following the review of the first draft UFVA, CAS AC recommended that other
19 percentiles be considered, up to and including the 98thpercentile used for the current 24-hour primary
20 and secondary standards. EPA staff is therefore considering the 90th, 95th and 98th percentiles per
21 year in this document. Due to inter-annual variability in meterology and other circumstances
22 that affect air quality, EPA staff is recommending using a three year average form of the
23 standard for purposes of consistency and stability, as is the current 24-hour primary PM standard.
24 By considering all of the combinations of the two hourly forms (i.e. each daylight hour and
25 daylight 1-hour daily maximum), the three CPLs and the three frequencies, a total of 18 separate
26 alternative secondary PM NAAQS scenarios are generated for use in the assessments described
27 below in chapters 3 and 4 (See table 4-1). An additional CAS AC recommendation, that the
28 relative humidity (RH) limit be lowered from 95% to 90% and used as a screen (i.e., hours above
29 it should be discarded) rather than as a cap, to more clearly exclude weather events like fog or
30 precipitation and to minimize effects of measurement error and spatial variability, has also been
31 incorporated in this draft.
32 1.5 ORGANIZATION OF DOCUMENT
33 The remainder of this document is organized as follows: Chapter 2 includes an analysis
34 of the urban visibility preference studies with a discussion of similarities and differences
35 regarding the approaches and methods used and results obtained for each study. This chapter
January 2010 1-16 DRAFT - Do Not Quote or Cite
-------
1 also includes a summary discussion of the results of a composite assessment of the combined
2 results from the four urban areas (Denver, Phoenix, British Columbia, and Washington, DC, an
3 accompanying logit (statistical) analysis, and use of these results in the selection of the
4 alternative levels evaluated in the remainder of the assessment. Chapter 3 describes the
5 analytical approach, methods, and data used in conducting the assessment of recent urban
6 visibility conditions, both in terms of PM2.5 and PM light extinction indicators for the set of
7 urban case studies included in this analysis. Selected results are presented in chapter 3, with
8 additional results found in the Appendices. Chapter 4 presents estimates of PM2.5 and PM light
9 extinction conditions generated for the urban case studies for six alternative PM2.5 and light
10 extinction scenarios. Additional information regarding approaches, results, method validation
11 studies and uncertainty assessments for both chapters 3 and 4 are presented in Appendices A-I).
12
January 2010 1-17 DRAFT - Do Not Quote or Cite
-------
1 2 URBAN VISIBILITY PREFERENCE STUDIES
2 The purpose of this chapter is to describe the reanalysis of available urban visibility
3 preference studies conducted by EPA staff with contractor support. The goal of this reanalysis
4 was to provide information useful in selecting a range of light extinction CPLs for use in
5 subsequent UFVA assessments of current and alternative VAQ conditions. The available urban
6 visibility preference studies all used a similar group interview type of survey to investigate the
7 level of visibility impairment that participants described as "acceptable. While each study asked
8 the basic question, "What level of visibility degradation is acceptable?", the term "acceptable"
9 was not defined, so that each person's response was based on his/her own values and preferences
10 for VAQ.
11 The reanalysis included three completed urban visibility preference survey studies plus a
12 pair of smaller focus studies designed to explore and further develop urban visibility survey
13 instruments. The three western studies included Denver, Colorado (Ely et al., 1991), one in the
14 lower Fraser River valley near Vancouver, British Columbia (BC), Canada (Pryor, 1996), and
15 one in Phoenix, Arizona (BBC Research & Consulting, 2003). A pilot focus group study was
16 also conducted for Washington, DC (Abt Associates Inc., 2001). In response to an EPA request
17 for public comment on the Scope and Methods Plan (74 FR 11580, March 18, 2009), Dr. Anne
18 Smith provided comments (Smith, 2009) about the results of a new Washington, DC focus group
19 study that had been conducted using methods and approaches similar to the method and approach
20 employed in the EPA pilot study (Smith and Howell, 2009). When taken together, these studies
21 from the four different urban areas included a total of 852 individuals, with each individual
22 responding to a series of questions answered while viewing a set of images of various urban
23 VAQ conditions.
24 2.1 METHODS USED IN PREVIOUS STUDIES
25 In all but one7 of the visibility preference studies assessed in this document, participants
26 were shown a series of different VAQ conditions projected on a large screen using a slide
27 projector. In the earliest two studies (the Denver and lower Frazer River Valley British
28 Columbia studies) the range of VAQ conditions were presented by projecting photographs
29 (slides) of actual VAQ conditions. The photographs were taken on different days from the same
30 location, and presented the same scene. Photographs were selected to avoid depicting significant
7 Smith and Howell (2009) used digital projection technology not used by the other studies to present the series of
VAQ conditions. Some of the participants in the Smith and Howell study were shown images using a LCD
projector connected to a laptop computer. In other sessions, participants in the Smith and Howell study were shown
images on a computer monitor connected to the computer.
January 2010 2-1 DRAFT - Do Not Quote or Cite
-------
1 weather events (e.g., rain, snow, or fog), and where measured extinction data were available
2 from the time the photograph was taken.
3 The Phoenix study and the Washington, DC projects used computer generated
4 photographic-quality images to present different VAQ conditions. Using an original near-
5 pristine base photograph, additional images representing a range of VAQ conditions were
6 generated using the WinHaze software program, which is based on a technique described in
7 Molenar et al. (1994). The Phoenix study and the 2001 Washington, DC project projected slides
8 of digital images prepared by WinHaze. The 2009 Washington, DC project presented images
9 directly from the desktop version of WinHaze using either a liquid crystal display (LCD)
10 proj ector or a computer monitor.
11 WinHaze analysis synthetically superimposes a uniform haze on a digitized, actual
12 photograph. The WinHaze computer algorithm calculates how a given extinction level would
13 impair the appearance of each individual portion of the photograph. A major advantage of
14 presenting WinHaze-generated images is that they provide viewers depictions of alternative
15 VAQ levels, with each image containing exactly the same scene, with identical light angle, time
16 of day properties, weather conditions, and specific scene content details (e.g., the amount of
17 traffic in a intersection). Additional details about WinHaze, and a discussion of the applicability
18 of WinHaze images for regulatory purposes, is in the 2004 PM Criteria Document (U.S. EPA,
19 2004). The desktop version of WinHaze is available online (Air Resources Specialists, 2008).
20 The first urban visibility preference study (Denver, CO; Ely et al., 1991) developed the
21 basic survey method used in all the subsequent studies. Although there are variations in specific
22 details in each study, all the studies use a similar overall approach (key variations are discussed
23 in the section on each study later in this chapter). This approach consisted of conducting a series
24 of group interview sessions, where the participants were shown a set of photographs or images of
25 alternative VAQ conditions and asked a series of questions. The group interview sessions were
26 conducted multiple times with different participants. Ideally the participants will be a
27 representative sample of the residents of the metropolitan area. While all studies agree that this
28 is the preferred approach, due to the high cost of organizing and conducting a series of in-person
29 group interviews with a large, statistically representative sample, only the Phoenix study was
30 able to fully meet this objective. During the group interview sessions, the participants were
31 instructed to consider whether the VAQ in each photograph or image would meet an urban
32 visibility standard, according to their own preferences and considering three factors:
33
34 1. The standard would be for their own urban area, not a pristine national park area
35 where the standards might be stricter.
January 2010 2-2 DRAFT - Do Not Quote or Cite
-------
1 2. The level of an urban visibility standard violation should be set at a VAQ level
2 considered to be unreasonable, objectionable, and unacceptable visually.
3 3. Judgments of standards violations should be based on visibility only, not on
4 health effects.
5 The photographs (images) were not shown in order of ascending or descending VAQ
6 conditions; the VAQ conditions were shown in a randomized order (with the same order used in
7 each group interview session). In order to check on the consistency of each individual's
8 answers, the full set of photographs (images) shown during the group interview included
9 duplicates with the identical VAQ conditions.
10 The participants were initially given a set of "warm up" exercises to familiarize them
11 with how the scene in the photograph or image appears under different VAQ conditions. The
12 participants next were shown 25 randomly ordered photographs (images), and asked to rate each
13 one based on a scale of 1 (poor) to 7 (excellent). They were then shown the same photographs or
14 images again (in the same order), and asked to judge whether each of the photographs (images)
15 would violate what they would consider to be an appropriate urban visibility standard (i.e.,
16 whether the level of impairment was "acceptable" or "unacceptable").
17 2.2 DENVER, COLORADO
18 The Denver urban visibility preference study (Ely et al., 1991) was conducted on behalf
19 of the Colorado Department of Public Health and Environment (CDPHE). The study consisted
20 of a series of focus group sessions conducted in 1989 with participants from 16 civic
o
21 associations, community groups, and employees of state and local government organizations.
22 The participants were not selected to be a fully representative sample of the Denver metropolitan
23 population but were instead selected to take advantage of previously scheduled meetings.
24 During the 16 focus group sessions, a total of 214 individuals were asked to rate
25 photographs of varying visibility conditions in Denver. The photographs were taken November
26 1987 through January 1988 by a camera in Thornton, Colorado. Thornton is suburb of Denver,
27 located approximately six miles north of downtown Denver. The photographs were taken as part
28 of a CDPJrtE study of Denver's air quality. The scene in the photographs was toward the south
29 from Thornton and included a broad view of downtown Denver and the mountains to the south.
30 Each group was shown one of two sets of 20 randomly ordered unique photographs (13 of the
31 sessions included 5 duplicate slides, for a total of 25 photographs, to evaluate consistency of
32 responses). The two sets of different slides were used to investigate whether the responses
33 between the two sets of photographs were different (no differences were found). Approximately
8 No preference data were collected at a 17th focus group session due to a slide projector malfunction.
January 2010 2-3 DRAFT - Do Not Quote or Cite
-------
1 100 participants viewed each photograph. Projected color slides were used to present the
2 photographs to focus group participants, and were projected on a large screen
3 The VAQ conditions in each Denver photograph were recorded when the photograph was
4 taken and measured by a transmissometer yielding hourly average light extinction, bext- The
5 transmissometer was located in downtown Denver, approximately eight miles from the camera
6 and in the middle of the camera's view path. Ely et al. (1991) provide the time of day and
7 measured extinction level for each photograph. The extinction levels presented in the Denver
8 photographs ranged from 30 to 596 Mm"1. This corresponds to 1 Idv to 41dv, approximating the
9 10* 1090* percentile of wintertime visibility conditions in Denver in the late 1980s.
10 The participants first rated the VAQ in each photograph on a 1 to 7 scale, and
11 subsequently were asked if each photograph would violate an urban visibility standard. The
12 individual's rating on the 1 to 7 scale and whether the photograph violated a visibility standard
13 were highly correlated (Pearson correlation coefficient greater than 80%).
14 The percent of participants who found a photograph acceptable to them (i.e., would meet
15 an appropriate urban visibility standard) was calculated for each photograph. Figure 2-1 shows
16 the results of the Denver participants' responses, with VAQ measured in deciviews.
17
Figure 2-1 Percent of Denver participants who considered VAQ in each
photograph "acceptable."
18
19
20
21
100%
n :
i- o>
« s
c n
n :
a.
50%
0%
10
15
20
25 30
Deciview
35
40
45
% acceptable Denver standard
Ely et al. (1991) introduce a "50% acceptability" criteria analysis of the Denver
preference study results. The 50% acceptability criteria is designed to identify the VAQ level
that best divides the photographs into two groups: those with a VAQ rated as acceptable by the
January 2010
2-4
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
majority of the participants, and those rated not acceptable by the majority of participants. While
no single VAQ level creates a perfect separation between the two groups, the CDPHE identified
a VAQ of 20.3 dv as the point that best separates the Denver study responses into "acceptable"
and "not acceptable" groups. Based in part on the findings of the Denver visibility preference
study, the CDPHE established a Denver visibility standard at bext = 76 Mm"1 (dv = 20.3).
Using 20.3 dv as the 50% acceptability criteria led to six photographs being
inconsistently rated by the majority of the viewers. A photograph was inconsistently rated for
two possible reasons; either the photograph's VAQ was at least 1 dv better than the Denver
standard (i.e., dv < 19.3) but was judged to be "unacceptable" by a majority of the participants
rating that photograph, or the VAQ was at least 1 dv worse than the standard (> 21.3 dv) but
found to be acceptable by the majority of the participants. This definition of inconsistent rating
helps evaluate the robustness of the study results to support the selection of the Denver urban
visibility standard at 76 Mm"1 (20.3 dv) by identifying photographs with VAQ a minimum of 1
dv above or below the standard and ignoring "near misses" involving photographs within 1 dv of
the standard. A change of 1 or 2 dv in uniform haze under many viewing conditions will be seen
as a small but noticeable change in the appearance of a scene, regardless of the initial haze
condition (U.S. EPA, 2004).
Table 2-1 presents information about the six photographs that were inconsistently rated.
All six of the inconsistently rated photographs were taken at 9:00 a.m. The five inconsistently
rated photographs with a VAQ better than the Denver standard have a VAQ at least 2 dv below
the standard. The VAQ in the only inconsistently rated photograph with air quality worse than
the standard (Photograph #6) is 1.1 dv above the standard. The study used 18 photographs from
9:00 a.m., so a third of the 9:00 a.m. photographs were inconsistently rated. Conversely, none of
the 32 photographs taken at noon or 3:00 p.m. were inconsistently rated.
Table 2-1. VAQ of Denver photos substantively misclassified by majority of participants
Photograph #
14
18
19
20
24
36
VAQ in photograph
in extinction
(Mm1)
44
54
54
55
60
85
VAQ in
photograph (dv)
13.8
16.9
16.9
17.0
17.9
21.4
% of participants
who rated the photo
"acceptable"
43%
43%
31%
42%
13%
72%
Time of day of
photograph
9:00 a.m.
9:00 a.m.
9:00 a.m.
9:00 a.m.
9:00 a.m.
9:00 a.m.
28
January 2010
2-5
DRAFT - Do Not Quote or Cite
-------
1 Figure 2-2 shows the same data results about percent of participants who rated each
2 photograph acceptable as in Figure 2-1, but with the time of day of each photograph indicated by
3 different colors. The time of day colors clearly indicate how inconsistently participants rated
4 some of the 9:00 a.m. photographs.
5 Eliminating the 9:00 a.m. photographs creates a "hole" in the range of remaining
6 photographs; there are no photographs with a VAQ between 17.7 dv and 20.3 dv. As seen in
7 Figure 2-3, this is a critical range in evaluating the responses. All of the photographs with a
8 VAQ equal to or better (i.e., a lower dv value) than 17.7 dv are rated acceptable by the majority
9 of the participants, and all photographs with a VAQ at or above 20.3 dv are rated not acceptable.
10 After eliminating the 9:00 a.m. photographs, any VAQ level between 17.7dv and 20.3 dv would
11 completely divide the photographs into two groups with no inconsistent ratings.
12
Figure 2-2 Photograph time of day information
13
14
15
16
17
18
19
20
21
22
100%
D)
c
H :
Ss
I "
.!&
'o o
k. CU
o%
••.
10
15
20
25 30
Deciview
35
40
45
9:00 AM
12:OOPM
3:OOPM
Denver standard
A modestly broader range of VAQ conditions provides an even more unambiguous
interpretation of the Denver study results. Every photograph with a VAQ of 17.7 dv or lower
was rated acceptable by 89% or more of participants, and every photograph with a VAQ of 24.6
or higher was rated not acceptable by 84% or more of the participants. The 17.7 dv to 24.6 dv
range separating the results is shown in Figure 2-3, which also eliminates the 9:00 a.m. results.
January 2010
2-6
DRAFT - Do Not Quote or Cite
-------
Figure 2-3 Denver photograph time of day results (9:00 a.m. photographs
eliminated), with the broader range (17.7 dv and 24.6 dv) of the 50% acceptability criteria
shown.
IUU /O
~ -i I I r->
d
2 2.3 VANCOUVER, BRITISH COLUMBIA, CANADA
3 The BC urban visibility preference study (Pryor, 1996) was conducted on behalf of the
4 BC Ministry of Environment following the methods used in the Denver study. Participants were
5 students at the University of British Columbia, who were in one of four focus group sessions
6 with between 7 and 95 participants. A total of 180 participants completed the surveys (29 did
7 not complete the survey).
8 The BC study used photographs (projected as slides) depicting various VAQ conditions
9 in two cities (Chilliwack and Abbotsford) in the lower Fraser River valley in southwestern BC.
10 Abbotsford is located approximately 75 miles east of Vancouver, BC, and had a 2006 population
11 of 159,000 (Statistics Canada, 2009a). Abbotsford has a diverse and successful economy, with
12 approximately 25% of the labor force working in the Vancouver metropolitan area. Chilliwack
13 is adjacent to Abbotsford to the east. Both cities have experienced rapid population growth,
14 growing faster than the Vancouver metropolitan area, and are considered suburbs (or exurbs) of
15 Vancouver.
16 The survey was conducted at the University of British Columbia (UBC) in 1994. The
17 participants were 206 undergraduate and graduate students enrolled in classes in UBC's
18 Department of Geography. Information about student demographics and where they lived prior
19 to enrolling at UBC (which potentially influences their knowledge of, and preferences for,
20 Vancouver area visibility) is not available.
January 2010
2-7
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
The BC survey showed 20 unique photographs to the participants in random order. Ten
photographs were from Chilliwack, and 10 were from Abbotsford. The Chilliwack photographs
were taken at the Chilliwack Hospital, and the scene includes a complex foreground with
downtown buildings, with mountains in the background up to 40 miles away. Figure 2-4 is a
composite of two of the Chilliwack photographs used in the preference study, showing the scene
with a good visibility day (14.1 dv) in the middle and a significantly impaired day (34 dv) around
the border (Jacques Whitford AXYS, 2007). The Abbotsford photographs were taken at the
Abbotsford Airport. The Abbotsford scene includes fewer man-made objects in the foreground
and is primarily a more rural scene with the mountains in the background up to 36 miles away.
Figure 2-4 Composite Chilliwack, BC photograph shows VAQ of 14.1 dv and 34 dv.
12
13 The photographs were taken in July and August 1993 as part of a VAQ and fine
14 particulate monitoring project sponsored by the BC Ministry of Environment, Lands and Parks
15 (REVEAL, the Regional Visibility Experimental Assessment in the Lower Fraser Valley). All of
16 the photographs were taken at either 12:00 p.m. or 3:00 p.m. VAQ data were available for each
17 photograph from visibility monitors near the location of each camera. The types of VAQ
18 measurement data available from the two locations were not identical. The Chilliwack location
19 used both an open-chamber nephelometer and a long path transmissometer and collected hourly
20 average data on both aerosol light scattering (bsp) and total extinction (bext\ respectively. The
January 2010
2-8
DRAFT - Do Not Quote or Cite
-------
1 visibility monitoring at the Abbotsford location had only a nephelometer and collected only bsp
2 data.
3 As explained in section 1.3, total light extinction is the sum of scattering by gases (bsg)
4 and particles (bsp) plus light absorption by gases (bag) and particles (bap). In order to present the
5 preference results from the BC study in comparable terms, bext for the Abbotsford photographs is
6 estimated by assuming that the average of the ratios of PM light extinction (i.e., bap + bsp) to PM
7 light scattering (bsp) for all ten of the Chilliwack photographs can be multiplied by the
8 Abbotsford nephelometer determined bsp values corresponding to each of its photographs to
9 estimate its PM light extinction value. By assuming that absorption by gases (bag) is zero, total
10 light extinction is equal to the PM light extinction (i.e., bap + bsp) plus particle scattering by gases
11 (i.e., bsg that is approximately equal to 10 Mm"1). Table 2-2 presents the data from the
12 photographs used in the BC study, including the estimated bext for the Abbotsford photographs.
13 There are two caveats to be noted about the extinction data for the photographs reported
14 in Pryor, 1996. First, in Table 2 of the original article, two of the Abbotsford photographs are
15 listed with the same date and time (12:00 p.m., 7/26/1993). There is no information provided for
16 a 3:00 p.m., 7/26/1993 Abbotsford photograph, although there is a Chilliwack photograph from
17 that time. The preference and VAQ data are presumed to be correct for both photographs and
18 one of the two identical date/time labels is assumed to be a typographic error. The second caveat
19 is that bsp levels from the same date and time can differ substantially between Abbotsford and
20 Chilliwack, and the relative levels can change rapidly, even though the two cities are only 25
21 miles apart. For example, at 12:00 p.m. on 8/19/1993, the bsp level in Chilliwack was about one-
22 third of the Abbotsford bsp level. By 3:00 p.m. the situation was reversed, with the Chilliwack
23 bsp level 50% higher than Abbotsford. In those three hours the Chilliwack bsp level had more
24 than doubled (from 46 Mm"1 to 105 Mm"1), and the Abbotsford level had fallen by over half
25 (from 145 Mm"1 to 67 Mm"1). Such substantial changes in measured bsp levels occurring across a
26 relatively short period of time and short distance, may reflect an inherent uncertainty introduced
27 by using a single measure of light extinction from a portion of visual scene (where the
28 nephelometer or transmissometer was operating) to assess visibility conditions throughout an
29 actual photographs of a complex scene. Spatial and temporal non-uniformity of visibility
30 conditions within a scene are an atmospheric condition known to occur on some days, and may
31 contribute to the variability in participant responses in preference studies utilizing actual
32 photographs.
33
34
35
36 Table 2-2. Summary of photographs used in British Columbia study
January 2010 2-9 DRAFT - Do Not Quote or Cite
-------
Date
Time
bsp
bext
Ratio
(bexrbsg)/bsp
Estimated
bext
Deciview
Chilliwack
7/26/93
7/26/93
7/27/93
7/27/93
8/2/93
8/2/93
8/5/93
8/5/93
8/19/93
8/19/93
12:00 p.m.
3:00 p.m.
12:00 p.m.
3:00 p.m.
12:00 p.m.
3:00 p.m.
12:00 p.m.
3:00 p.m.
12:00 p.m.
3:00 p.m.
Average
86
67
63
119
18
20
45
51
46
105
62
128
112
105
185
37
36
70
96
81
170
102
1.372
1.522
1.508
1.471
1.5
1.3
1.333
1.686
1.543
1.524
1.476
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
25.49
24.16
23.51
29.18
13.08
12.81
19.46
22.62
20.92
28.33
21.96
Abbotsford
7/26/93
7/26/93
7/27/93
7/27/93
8/2/93
8/2/93
8/5/93
8/5/93
8/19/93
8/19/93
12:00 p.m.
12:00 p.m.
12:00 p.m.
3:00 p.m.
12:00 p.m.
3:00 p.m.
12:00 p.m.
3:00 p.m.
12:00 p.m.
3:00 p.m.
Average
39
82
104
132
24
25
62
75
67
145
76
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
68
131
205
164
45
47
121
102
224
109
122
19.17
25.73
30.20
27.97
15.04
15.48
24.93
23.22
31.09
23.89
23.67
2
3
4
5
6
7
Figure 2-5 presents the results of the BC study. The division corresponding to the
Denver "50% acceptable" criteria occurs between 22.6 dv and 23.2 dv. All of the photographs
with a VAQ better than 22.6 dv were rated acceptable by the majority of the participants with
one exception (47% of the participants judged the 19.2 dv photograph to be acceptable). All
photographs with a VAQ better than 19.2 dv were rated acceptable by over 90% of the
participants. All photographs with a VAQ worse than 22.6 dv were rated not acceptable by the
January 2010
2-10
DRAFT - Do Not Quote or Cite
-------
2
O
4
5
6
7
8
9
10
11
12
13
14
majority of the participants, and all photographs with a VAQ worse than 28.3 dv were rated not
acceptable by over 90% of the participants.
Figure 2-5 also suggests that there may be some difference between the preferences
expressed for the Chilliwack scene and those for the Abbotsford scene. All photographs were
rated by the same individuals (students at UBC), but the summary of the responses indicate that
the participants may have rated as acceptable a worse level of impaired VAQ impairment (e.g.,
higher dv levels) in photographs showing more of a downtown area (Chilliwack) than in less
congested scenes (Abbotsford). The strongest evidence for this hypothesis, however, is the
preference for a single photograph (the 19.0 dv photograph from Abbotsford, rated as acceptable
by 47%), previously identified as an outlier observation.
Figure 2-5 Percent of BC participants who consider VAQ in each photograph
"acceptable."
15
16
17
18
Results of British Columbia Visiblity Study
100%
_o
.0
$
a.
o
u
u
•*:
n
0£
c
n
Q.
'u
'•£
n
Q.
50%
10
15
20
25 30
Deciview
35
40
45
• Chilliwack • Abbotsford
The BC Ministry of the Environment is considering the BC urban visibility preference
study as part of establishing urban and wilderness visibility goals in BC.
19 2.4 PHOENIX, ARIZONA
20 The Phoenix urban visibility preference study (BBC Research & Consulting, 2002),
21 which was conducted on behalf of the Arizona Department of Environmental Quality, used
January 2010
2-11
DRAFT - Do Not Quote or Cite
-------
1 group interviews based on the methods used in the Denver study, with two major exceptions: (1)
2 the focus group participants were selected as a representative sample of the Phoenix area
3 population, and (2) the pictures presented in the focus groups were computer-generated images
4 to depict specific uniform haze conditions.
5 The Phoenix study included 385 participants in 27 separate focus group sessions.
6 Participants were recruited using random digit dialing to obtain a sample group designed to be
7 demographically representative of the larger Phoenix population. During July 2002, group
8 interview sessions took place at six neighborhood locations throughout the metropolitan area to
9 improve the participation rate. Participants received $50 as an inducement to participate.
10 Three sessions were held in Spanish in one region of the city with a large Hispanic
11 population (25%), although the final overall participation of native Spanish speakers (18%) in
12 the study was below the targeted level. The age distribution of the participants corresponded
13 reasonably well to the overall age distribution in the 2000 U.S. Census for the Phoenix area
14 (BBC Research & Consulting, 2002). Participants slightly over-represented the middle-income
15 range ($50,000 to $74,999), compared with 2000 Census data, and slightly under-represented
16 very low-income ranges (under $24,999). The distribution of participant education levels was
17 fairly consistent with the education distribution in the 2000 Census.
18 Photographic-quality slides of the images were developed using the WinHaze software
19 (Molenar et al., 1994). The scene used in the Phoenix study images was taken at a water
20 treatment plant. The view is toward the southwest, including downtown Phoenix, with the Sierra
21 Estrella Mountains in the background at a distance of 25 miles. Figure 2-6 shows the image with
22 thebestVAQ(15dv).
23
January 2010 2-12 DRAFT - Do Not Quote or Cite
-------
Figure 2-6 Reproduction of image with the best VAQ (15 dv) used in the Phoenix
study.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
The study used a total of 21 unique WinHaze images. Four of the 21 unique images were
randomly selected and used twice to evaluate consistency; participants viewed a total of
25 images. The 25 images were randomly ordered, with all participants viewing the images in
the same order. The WinHaze images used in the Phoenix study do not include layered haze, a
frequent and widely recognized form of visibility impairment in the Phoenix area.
The VAQ levels in the 21 unique images ranged from 15 dv to 35 dv (the extinction
coefficient bext ranged from 45 Mm"1 to 330 Mm"1). As in the Denver study, participants first
individually rated the randomly shown slides on the same VAQ scale of 1 to 7. Participants were
instructed to rate the photographs solely on visibility and to not base their decisions on either
health concerns or what it would cost to have better visibility. Next, the participants individually
rated the randomly ordered slides as "acceptable" or "not acceptable," defined as whether the
visibility in the slide is unreasonable or objectionable.
Figure 2-7 presents the percent acceptability results from the Phoenix study. The
combination of the use of WinHaze images and the larger number of participants than in the
Denver study may account for the "smoother" backwards S-shaped pattern of preferences.
January 2010
2-13
DRAFT - Do Not Quote or Cite
-------
Figure 2-7 Percent of Phoenix participants who consider VAQ in each image
'acceptable."
100%
«
.a
t
o
u
B>
n
Q.
'u
'•£
n
Q.
50%
0%
10
15
20
25
30
35
40
45
Deciview
2
3 Ninety percent or more of the participants rated a VAQ of 20 dv or better as acceptable, and 70%
4 rated a VAQ of 22 dv or better as acceptable. The "50% acceptable criteria" was met at
5 approximately 24.3 dv (with 51.3% of the participants rating that image as acceptable). The
6 percent acceptability declines rapidly as VAQ worsens; only 27% of the participants rated a
7 26 DV image as acceptable, and fewer than 10% rated a 29 dv image as acceptable.
8 The Phoenix urban visibility study formed the basis of the decision of the Phoenix
9 Visibility Index Oversight Committee for a visibility index for the Phoenix metropolitan area
10 (Arizona Department of Environmental Quality, 2003). The Phoenix Visibility Index establishes
11 an indexed system with 5 categories of visibility conditions, ranging from "Excellent" (14 dv or
12 less, which was a better VAQ than any of the images used in the Phoenix study) to "Very Poor"
13 (29 dv or greater, which less than 10% of the study participants rated as acceptable). The
14 "Good" range is 15 dv to 20 dv (more than 90% of the participants rated images in this VAQ
15 range as acceptable). The environmental goal of the Phoenix urban visibility program is to
16 achieve continued progress through 2018 by moving the number of days in poorer quality
17 categories into better quality categories.
18 2.5 WASHINGTON, DC
19 One of the Washington, DC urban visibility pilot studies was conducted on behalf of
20 EPA (Abt Associates Inc., 2001). It was designed to be a pilot focus group study, an initial
January 2010
2-14
DRAFT - Do Not Quote or Cite
-------
1 developmental trial run of a larger study. The intent of the pilot study was to refine both focus
2 group method design and potential survey questions. Due to funding limitations, only a single
3 focus group session took place, consisting of one extended session with nine participants. No
4 further urban visibility focus group sessions were held in Washington, DC on behalf of EPA.
5 In March 2009, Dr. Anne Smith conducted a separate study of Washington urban
6 visibility, using the same photographs and similar approach as the 2001 study (Smith and
7 Howell, 2009). On behalf of the Utility Air Regulatory Group, Dr. Smith presented comments
8 (Smith, 2009) to the CAS AC at a public meeting held on April 2, 2009 to review EPA's plan
9 (US EPA, 2009b) for conducting further urban visibility studies in support of PM NAAQS
10 reviews. Dr. Smith submitted the Smith and Howell (2009) report to the CAS AC as part of the
11 public comment process. The Smith and Howell study conducted three study variations of a
12 Washington, DC, preference study, including one experiment involving 26 participants designed
13 to replicate the EPA 2001 preference study.
14 Both the Abt Associates Inc. (2001) study results and the results of the Smith and Howell
15 (2009) study are discussed below.
16 2.5.1 Washington, DC 2001
17 The EPA's Washington, DC study (Abt Associates Inc., 2001) adopted the general study
18 methods used in the Denver, BC, and Phoenix studies, modifying them appropriately to be
19 applicable in an eastern urban setting. Washington's (and the entire East's) current visibility
20 conditions are typically substantially worse than western cities and have different characteristics.
21 Washington's visibility impairment is primarily a uniform whitish haze dominated by sulfates,
22 and the relative humidity levels are higher compared with the western study areas. In addition,
23 the relatively low-lying terrain9 in Washington, DC provides substantially shorter maximum
24 sight distances. Many residents are not well informed that anthropogenic emissions impair
25 visibility on hazy days.
26 The Washington, DC focus group session included questions on valuation, as well as on
27 preferences. The focus group content dealing with preferences for an urban visibility standard
28 was similar to the focus group sessions in the western studies.
29 A single scene of a panoramic photograph taken from Arlington National Cemetery in
30 Virginia was used, and included an iconic view of the Potomac River, the National Mall, and
31 downtown Washington, DC. All of the distinct buildings in the scene are less than four miles
32 from the camera, and the higher elevations in the background are less than 10 miles from the
33 camera. Figure 2-8 presents the photograph used in the study.
34
9The maximum elevation in Washington, DC is 409 feet.
January 2010 2-15 DRAFT - Do Not Quote or Cite
-------
Figure 2-8 Reproduction of the image with the best VAQ (8.8 dv) used in the
Washington, DC study.
1
2 The Washington, DC study used 20 unique images generated by WinHaze, each prepared
3 from the same original photograph. Humidity and gaseous light scattering was held constant in
4 preparing the WinHaze images, as was the relative chemical mix of aerosol particulates in the
5 photos (i.e., only the aerosol concentrations were increased to create the images with worse
6 VAQ). Five of the images were repeated as a consistency check, so participants viewed a total
7 of 25 slides. The range of VAQ in the images ranged from 8.8 to 38.3 dv.
8 Figure 2-9 presents the percent acceptability results from the 2001 Washington study.
9 Because only nine participants were involved in the study, the possible values of "percent
10 acceptable" are limited to multiples of 1/9. Figure 2-9 also shows an anomalous result involving
11 one of the five repeated images. Three of the repeat images had the same ranking each time they
12 were presented (i.e., all nine participants rated them acceptable or not acceptable both times they
13 rated that slide). One of the images (the image with 8.8 dv, the best VAQ image used in the
14 study) was rated acceptable by all nine participants the first time it was used, but the repeat of
15 that slide was rated not acceptable by one participant. Another image, however, had a
16 substantially different result. The 30.9 dv image was rated acceptable by five of the nine
17 participants the first time it was presented, but the repeat of the slide was only rated acceptable
18 by one of the nine participants. The responses for all five pairs of repeated images are shown in
January 2010
2-16
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
red on Figure 2-9, including the images which were identically rated both times they were
presented.
Figure 2-9 Percent of 2001 Washington participants who considered VAQ
acceptable in each image
Results of 2001 Washington DC Preference Study
100%
B>
C
'^ =
OL «
$ «
if 50%
O U
%*
Q.
0%
*• «•
4 • »
5 10 15 20 25 30 35 40 45
De civ Jew
Unique Images • 5 Repeated Images
In the 2001 Washington, DC study, all images with a VAQ below 25.9 dv were rated
acceptable by the majority of the participants, and all images with a VAQ below 29.2 dv were
rated acceptable by at least four of the nine (44%) participants. All images with a VAQ above
30.9 dv were rated not acceptable. The "50% acceptability criteria" division occurs in the range
of 25.9 dv to 30.9 dv, with the anomalous result of the inconsistent responses to the repeated
image with 30.9 dv effectively broadening this range and adding uncertainty to identifying a
clear division.
2.5.2 Washington, DC, 2009
The Smith and Howell (2009) study conducted additional focus group sessions based on
the methods and materials used in the 2001 Washington, DC study. Smith and Howell recreated
the WinHaze images used in the 2001 Washington, DC urban visibility preference study, using
the description in the report on the 2001 study (Abt Associates Inc., 2001), and created images
using currently available desktop computer version of WinHaze (Version 2.9.0). Smith and
January 2010
2-17
DRAFT - Do Not Quote or Cite
-------
1 Howell used a shortened version of the same question protocol as the 2001 study. The WinHaze
2 images were presented to a total of 64 participants who were all employees of Charles River
3 Associates (CRA International, Inc). (Smith and Howell also are CRA International employees).
4 The CRA employees were based at the firm's Washington, DC and Houston, Texas offices (44
5 and 20 participants, respectively). The Houston participants were included to explore whether
6 familiarity with Washington, DC VAQ conditions developed from currently living in the
7 Washington region noticeably influenced the responses. As noted by Smith and Howell, the
8 participants were not a representative sample of either metropolitan area's population; all
9 participants were employed, and the participant group included a higher proportion of college
10 educated individuals and higher household incomes than the general population.
11 Eight of the Washington-based participants and all of the Houston participants viewed the
12 WinHaze images on a desktop computer monitor. The remaining Washington participants
13 viewed the images projected on a screen.
14 The stated purpose of the Smith and Howell study was to explore the robustness of the
15 2001 results. To investigate this issue, Smith and Howell conducted three different tests
16 concerning urban visibility preferences. Each participant was involved with only one test. The
17 three tests were:
18 * Test 1 - replicated the Abt Associates Inc. (2001) study
19
20 * Test 2 - reduced the upper end of the range of VAQ by eliminating the 11 images
21 used in Test 1 with a VAQ above 27.1 dv
22
23 * Test 3 - increased the upper end of the range of VAQ by including two new images
24 of worse VAQ; the two new images had a VAQ of 42 dv and 45 dv
25
26 Sixteen employees from the Washington, DC office and 10 participants from the Houston
27 office took Test 1 (a total of 26 participants). All the participants viewed the same unique 20
28 Washington, DC WinHaze images as the 2001 study (plus repeated images for a total of 25
29 images shown to participants). Images were presented in the same random order as in the 2001
30 study. Figure 2-10 presents the results of Test 1. The results for the 16 Washington participants
31 are indicated in blue and results for the 10 Houston participants in red. Although all images used
32 in the study were of Washington, DC, the results suggest that there is not a significant difference
33 in the preferences of participants based in the two offices. The scene in the images is an
34 immediately recognizable iconic view of the National Mall and downtown Washington, DC,
35 which may influence the similarity of responses by residents of the two cities.
36
37
January 2010 2-18 DRAFT - Do Not Quote or Cite
-------
Figure 2-10 Percent of 2009 Test 1 study participants who considered VAQ
acceptable in each image, showing the range of the lower and upper bound of 50%
acceptability criteria.
100%
_o
.a
a
*•
a.
o
o
o
50%
a
a.
'o
'•E
a
a.
0%
_
A
— i
L A A
H
•
• A
•
A ff.A
•
A
'"••A
5 10 15 20 25 30 35 4
Deciview
• Washington A Houston Lower Bound Upper Bound
2
3 Using the combined Test 1 results from the two CRA offices (26 total participants), the
4 majority of participants in the 2009 study rated all VAQ images with 25.9 dv or less as
5 acceptable and all VAQ images with 29.2 dv or greater as not acceptable. The image of 27.1 dv
6 was rated as acceptable by 50% of the total participants (56% of the Washington-based and 40%
7 of the Houston-based participants). All images with a VAQ less than 22.9 dv were rated
8 acceptable by at least 90% of the participants, and all images with a VAQ greater than 32.3 dv
9 were rated not acceptable by 88% of the participants.
10 Figure 2-11 presents the Abt 2001 study and Smith and Howell 2009 (Test 1) study
11 results on a single graph, representing the results of 35 total participants of preferences for urban
12 visibility in Washington, DC. The results from the 2009 study on Figure 2-11 combine the Test
13 1 responses from the two CRA offices. Figure 2-11 also shows the 50% acceptability criteria
14 range (22.9 dv to 32.3 dv) from the 2009 Test 1. In comparison, the 2001 study 50%
15 acceptability range was 25.9 dv to 30.9 dv. Inspection of the points in Figure 2-11 indicates that
16 the results from the 2009 study (Test 1) are not appreciably different than the results of the 2001
17 Washington study.
January 2010
2-19
DRAFT - Do Not Quote or Cite
-------
1 In Test 2, Smith and Howell reduced the range of VAQ images to images with a VAQ of
2 27.1 dv or less. The 26 participants in the Test 2 study were different people than the Test 1
3 participants. Test 2 presented only the nine unique clearest WinHaze images from the full Test 1
4 set of 20 images, along with 3 duplicates for a total of 12 images. This constricted the VAQ
5 levels presented to the range that the majority of participants in the 2001 study rated as
6 acceptable and reduced the upper end of the VAQ range by 11.2 dv.
7
Figure 2-11 Combined results of two Washington preference studies (showing 50%
acceptability criteria from 2009, Test 1).
9
10
11
12
13
14
15
16
17
18
19
20
.a
ns
+j
a.
10 15 20 25 30 35 4
Deciview
• 2001 Study A 2009, Test 1 Lower Bound Upper Bound
Figure 2-12 presents the Test 1 and Test 2 results. Test 2 found a substantial shift in the
responses regarding which VAQ levels are considered acceptable. The smaller number of
images used in Test 2 made identifying the range of the 50% acceptability criteria more difficult
than in Test 1. The lower bound of the range occurs between 15.6 and 18.7 dv, and the upper
bound occurs between 24.5 and 27.1 dv. Smith and Howell conclude that the shift in the
acceptability responses between Test 1 and Test 2 suggests that the VAQ levels identified as
acceptable in an urban visibility preference study conducted using the general approach
previously used in the all the studies may be influenced by the range of VAQ images presented.
January 2010
2-20
DRAFT - Do Not Quote or Cite
-------
Figure 2-12 Comparison of results from Test 1 and Test 2 (Smith and Howell, 2009).
Results of Smith and Howell Test 1 and Test 2
100%
£1
f
I 50%
c
ro
o.
'o
•a
ro
o.
0%
10 15 20 25 30 35 40
Deciview
I Test 2 » Test 1
1 In order for the range of images shown to be able to influence the acceptability ratings,
2 participants would need to be aware of the upper and lower bounds of the range prior to the
3 judging acceptability. However, since they were shown images randomly with respect to the
4 VAQ levels, asked to rate each one before going to the next image, and were not given a chance
5 to revise their acceptability ratings, this was not possible during the acceptability exercise itself.
6 The only other opportunity participants could have to learn the VAQ range is during the VAQ
7 rating exercise done just prior to the acceptability rating. However, in the VAQ rating exercise
8 where the participants were asked to rate the quality of visibility for the shown images on a scale
9 from 1 to 7, the images were also shown in a random order, participants were not aware how
10 many photographs would be shown or the range of conditions, they were asked to rate each one
11 using a value from 1 to 7 before going on to the next image and they did not have the opportunity
12 to revise the ratings of earlier viewed images.
13 Figure 2-13 shows the average visibility rating on the 1 to 7 scale for each image used in
14 each of the three tests conducted by Smith and Howell (2009). The consistency observed in the
15 relationship between VAQ deciview levels and the average scores assigned across the three tests
16 demonstrates that the participants come to the survey with the capability to consistently rate the
17 haze levels shown in the images, regardless of the breadth of the range used or the order or
18 number of slides shown, and that they are aware of a full range of conditions, even when, as was
19 the case in Test 2, they were not shown the worst haze images.
January 2010
2-21
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
Figure 2-13 Average quality of visibility ratings for the Washington, DC WinHaze
images by participants in Tests 1-3 conducted by Smith and Howell (2009).
3
2
Test 1 (sample of 26)
Test 2 (sample of 26)
Test 3 (sample of 12)
8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44
DectvFew
-}
6 Why then did Test 2 participants in the subsequent part of the survey rate images of haze
7 levels as unacceptable that were rated acceptable by participants in the other tests and the earlier
8 Washington, DC pilot study? In a three sentence script10 that constituted the only instructions
9 read prior to the acceptability rating, the participants were told that they would see the same set
10 of slides that they had just rated (i.e., on the 1 to 7 scale), and they were asked to rate them
11 according to whether the VAQ depicted were acceptable or unacceptable to them. Apparently by
12 directing then to rate the same images for acceptability, the participants understood that their
13 choices of visibility conditions were restricted to a range of conditions shown in the 1 to 7
14 ratings that they had just completed. For participants in Test 2 this would mean that by their own
15 1 to 7 ratings the range was restricted to include no poor visibility conditions (i.e. only scenes
16 rated from 3 to 7).
17 Smith and Howell (2009) concluded that the effects of a changed range on the
18 acceptability ratings results demonstrates that VAQ preference studies results are not robust and
19 do not reflect an enduring view on the "unacceptability" of different levels of VAQ degradation.
10 The complete script for the acceptability/unacceptability part of the study is as follows. "Now you will
be shown the same set of slides that you just rated. Again each image will illustrate the effects of a different level of
visibility. This time, rate the slides according to whether the visibility is acceptable or unacceptable to you. "
January 2010
2-22
DRAFT - Do Not Quote or Cite
-------
1 However, there is an alternative explanation. It seems more likely that the use of such a severely
2 truncated range of VAQ conditions in Test 2, which did not include any of the images of VAQ
3 that previous studies identified as unacceptable, in effect fundamentally changed the implied
4 instructions for the participants. Instead of conveying that they were to identify VAQ levels that
5 they found acceptable among a full range of VAQ conditions from very poor to very good, the
6 implied message was that they should identify the VAQ levels that they found acceptable among
7 a curtailed range of VAQ conditions that only included average to very good VAQ. By this
8 reasoning, it would be inappropriate to include Test 2 results with those of the other tests as a
9 measure of VAQ preference for Washington, DC.
10 In Test 3, Smith and Howell expanded the VAQ range of WinHaze images shown to the
11 participants, including two new images with a worse VAQ. The new images had a VAQ of 42
12 dv and 45 dv, raising the upper end of the VAQ range by 6.7 dv. Test 3 also reduced the total
13 number of images shown to participants to 19 images by eliminating the use of the five repeat
14 images in Test 1, and also eliminated three additional images in order to reduce the participants'
15 time burden. The three deleted images had a VAQ of 11.1, 15.6, and 24.5 dv. The best VAQ
16 image shown to Test 3 participants was 8.8 dv (same as the best VAQ image in Tests 1 and 2).
17 However, in Test 3 there were no images with VAQ between 8.8 dv and 18.7 dv, creating a
18 significant "hole" in the distribution of VAQ conditions presented to the Test 3 participants.
19 Test 3 was conducted with 12 participants from the CRA Washington office (none of whom
20 participated in Test 1 or Test 2). No Houston participants were involved with Test 3. Figure 2-
21 13 shows that the Test 3 average ratings from 1 to 7 during the VAQ rating exercise increased
22 the average participant rating by about 1 at the low end of the scale (very poor VAQ). The
23 results of Test 3 are shown in Figure 2-14, along with the results of Test 1.
24 Test 3 resulted in an overall increase in the percent of respondents rating as acceptable
25 the VAQ images used in both tests. In Test 3 all images with a VAQ below 22.9 dv were rated
26 acceptable by 100% of the participants (similar to the Test 1 results), implying there was no
27 general change in the acceptability of the images with good VAQ. However, for all VAQ
28 images (that were used in both studies) between 25.9 dv and 33.6 dv, a noticeably larger
29 percentage of the participants in Test 3 rated the image as acceptable than in Test 1. At VAQ
30 levels worse than 33.6 dv, the majority of the participants found the VAQ level not acceptable in
31 both tests. While the differences are noticeable, the small number of participants in Test 3 (i.e.
32 12) makes the significance of the difference unclear.
33
34
35
36
January 2010 2-23 DRAFT - Do Not Quote or Cite
-------
Figure 2-14 Comparison of results from the Smith and Howell (2009) Test 1 and
Test 3.
100% i
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
3
a.
o
o
o
n
o:
.52
E
n
n
o_
50% -
0%
10
15
20
25
Deciview
30
35
40
45
I Test 3 » Test 1
The range limitations identified with Test 2 (i.e. an overly restrictive range) that resulted
in its results being eliminated from consideration in the selection of appropriate CPLs, do not
apply to Test 3 results due to its somewhat more complete coverage of the 1 to 7 rating range in
the VAQ rating exercise. In that sense, Test 3 results may be considered somewhat more reliable
than those from Test 1 and the original Washington, DC pilot study. However the number of
participants in Test 3 (i.e., 12) is small enough that the statistical uncertainty of the results may
be an issue if used alone. Given that most of the same images of VAQ conditions were used in
all of the tests, composite acceptability ratings (i.e., from the original pilot study and from Tests
land 3) of each image were developed to increase the number of participant ratings for each
image. Figure 2-15 shows the composite results of these three groups involving a total of 47
participants. The 50% acceptability criteria value for this composite dataset lies unambiguously
between the 30.1 dv (at 51.1%) and the 30.9 dv points (at 46.3%). This analysis does not address
the question of whether a significant "hole" in the Test 3 VAQ distribution between 8.8 dv and
18.7 dv potentially had an effect on participant acceptability responses.
January 2010
2-24
DRAFT - Do Not Quote or Cite
-------
1 Figure 2-15 Composite results from Smith and Howell (2009) Tests 1 and 3, and
2 Abt (2001) Washington, DC pilot study.
4
5
6
7
9
10
11
12
13
14
15
16
17
18
19
20
100%
50%
0%
Washington DC Studies Combined
2001 and Smith Tests 1 & 3
t * _ • • *
10 15 20 25 30
Deciview
35
40
45
50
2.6 SUMMARY OF PREFERENCE STUDIES AND SELECTION OF
CANDIDATE PROTECTION LEVELS
Each of the studies reviewed in this assessment investigates the common question, "What
level of visibility degradation is acceptable?" The approaches used in the four studies are similar
and are all derived from the method first developed for the Denver urban visibility study. As a
result, EPA staff has concluded that it is reasonable to compare the results to identify overall
trends in the study findings and that this comparison can usefully inform the selection of CPLs
for use in further analyses. However, because variations in the specific materials and methods
used in each study introduce uncertainties, direct comparison of the study results should take
these factors into account. Key differences between the studies include:
* Image presentation methods (e.g., projected slides of actual photos, projected images
generated using WinHaze (a significant technical advance in the method of presenting
VAQ conditions), use of computer monitor screen
January 2010
2-25
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
* Number of participants in each study,
* Participant representativeness of the general population of the relevant metropolitan
area, and
* Specific wording used to frame the questions used in the group interview process.
Figure 2-16 presents a graphical summary of the results of the studies in the four cities
and draws on results previously presented in Figures 2-3, 2-5, 2-7 and 2-11.
Figure 2-16 Summary of results of urban visibility studies in four cities, showing the
identified range of the 50% acceptance criteria .
n
11
12
13
14
15
16
17
2O Mm-1
100% i •
SO Mm1
1OO Mm-1 2OO Mm * 4OO Mirr1 8OO Mm *
0%
Denver
Denver Logil
Phoenix
PnoenixLogit
BC
BC Logit
* Washington
DC Logit
For clarity in Figure 2-16, the Denver results omit the 9:00 a.m. photograph results, the
Chilliwack and Abbotsford photographs appear as a single set of data for the BC study, and the
results from 2001 and 2009 (Test 1) studies of VAQ preferences in Washington, DC are
presented as a single combined set of data. The results from the 2009 Washington, DC study
Tests 2 and 3 are not included on Figure 2-16; Test 2 is not a comparable study because it
restricted the range of VAQ conditions to only those rated average to best (e.g., 3-7) visibility in
11 Top scale shows light extinction in inverse megameter units; bottom scale in deciviews. Logit analysis estimated
response functions are shown as the color-coded curved lines for each of the four urban areas
January 2010
2-26
DRAFT - Do Not Quote or Cite
-------
1 the VAQ ratings, and Test 3 is not included because of concerns for the effects on having a
2 significant "hole" in the VAQ distributions shown to participants between 8.8 dv and 18.7 dv.
3 Figure 2-16 shows the results of a logistical regression analysis using a logit model of the
4 greater than 19,000 ratings of haze images as acceptable or unacceptable. The logit model is a
5 generalized linear model used for binomial regression analysis which fits explanatory data about
6 binary outcomes (in this case, a person rating a VAQ image as acceptable or not) to a logistic
7 function curve.
8 In the context of the preference studies, the logit model determines a function that best
9 estimates the percentage of respondents that rate an image acceptable based on a set of
10 explanatory variables. The observations on the dependent variable have one of two discrete
11 values: 1 (the person rated the image acceptable) or 0 (unacceptable). For this application, the
12 logit model determines an equation estimating the proportion of participants who will find any
13 particular deviciew level acceptable. There were two basic types of explanatory (independent)
14 variables used: one continuous numerical variable (the image's haziness level or VAQ in
15 deciviews), and a set of discrete variables that identify which city the observation is from.
16 The fundamental form of a logistic function is;
17 probability•(" yes") = f(z) = —-—
l + e z
18 where the variable z, known as the logit, is the influence of all the explanatory variables;
19 z = 00 + 0lxl+02x2+...
20 In this analysis the estimated logistic function f(z) is the estimated probability of the
21 participants in the study rating an image as acceptable, given the dv value of the image and what
22 city the observation came from. In this application the logit is
23 z = Intercept + ftdv + J32BC + J33 (dv x BC) + J34DC + J35 (dv x DC) + j36Phoenix + J37 (dv x Phoenix)
24
25 The variables BC (British Columbia), DC (Washington, DC), and Phoenix are "dummy"
26 variables. For example, the BC variable is set equal to one if the observation is from the BC
27 study, and set to zero if that observation is from a study for a different city. Denver is used as
28 the omitted city dummy variable, allowing the estimated coefficients on the other three city
29 dummy variables to estimate if the response function is different in those cities than in Denver.
30 For example the estimated total intercept for Washington becomes Intercept + f}4, and the
31 estimated slope of the function is ft'4 + ft'5. A statistically significant estimate of the interaction
January 2010 2-27 DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
term coefficient (/?j, /?j, or/?/,) for a particular city implies that the response function has a
different slope than the Denver function.
The logit analysis was conducted using STATA® Data Analysis and Statistical Software
(Release ES 10.1), using the LOGIT procedure. Table 2-3 presents the parameter estimates from
the logit analysis, which investigates whether both slope and the intercept of the estimated
response function differ between cities. The pseudo-R2 estimate was 0.4756 and the
r\
loglikelihood chi test also strongly rejects the null hypothesis that there is no effect of the
explanatory variables on the probability that a respondent would find a image acceptable
(Pr(chi2)=0 < 0.000). In other words, the acceptability ratings depend both on the deciview value
and city.
Table 2-3 Logit Analysis Results
Variable
Dv
British Columbia
Washington, DC
Phoenix
BCxdv
Wash.Xdv
Phoenix xdv
Constant
Coefficient
-0.3862
1 .0496
2.9450
3.5682
-0.0029
0.0200
-0.0797
7.6844
Standard
Error
0.0094
0.3589
0.8458
0.3015
0.0162
0.0293
0.0136
0.1830
z-statistic
-41.16
2.92
3.48
11.84
-0.18
0.68
-5.88
41.99
Pr/z/ = 0
~0
0.003
~0
~0
0.86
0.495
~0
~0
5% confidence
estimate
-0.4045
0.3463
1.2873
2.9773
-0.0345
-0.0374
-0.1063
7.3257
95% confidence
estimate
-0.3678
1.7530
4.6026
4.1591
0.0288
0.0774
-0.0531
8.0431
The city intercept coefficients are all positive and statistically significant indicating that
the response functions for different cities shifted right relative to the function for Denver.
However, only the Phoenix interaction term is insignificant, indicating that the Phoenix response
function has a different slope than the other three cities, as can be seen in Figure 2-16. The
negative estimated coefficient on the Phoenix interaction term results in the Phoenix response
function being steeper than the other cities' functions. Figure 2-16 also shows the Washington,
DC function is modestly less steep than the others, but the decrease in the slope is not
statistically significant.
The model results can be used to estimate the VAQ deciview values where the estimated
response functions cross the 50% acceptability level, as well as any alternative criteria levels.
Selected examples of these are shown in Table 2-4.
January 2010
2-28
DRAFT - Do Not Quote or Cite
-------
1
2
3
Table 2-4 Logit model estimated VAQ values corresponding to various percent
acceptability values for the four cities.
90% Acceptability criteria
75% Acceptability criteria
50% Acceptability criteria
25% Acceptability criteria
10% Acceptability criteria
Denver
14.21
17.05
19.90
22.74
25.59
British
Columbia
16.80
19.63
22.45
25.28
28.10
Phoenix
24.15
21.80
24.15
26.51
28.87
Washington,
DC
23.03
26.03
29.03
32.03
35.03
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Figure 2-16 also contains lines at 20 dv and 30 dv that effectively and pragmatically
identify a range where the 50% acceptance criteria occur across all four of the urban preference
studies. Out of the 114 data points shown in Figure 2-16, only one photograph (or image) with a
VAQ below 20 dv was rated as acceptable by less than 50% of the participants who rated that
1 9
photograph. Similarly, only one image with a VAQ above 30 dv was rated acceptable by more
than 50% of the participants who viewed it.13 These upper and lower range values are also
supported by the logit model data which estimates 50* percentile acceptability values near 20 dv
for Denver and near 30 dv for Washington, DC (see Table 2-4).
There are several hypotheses that may explain why the VAQ acceptability response
curves for the four cities are different and why some study results have greater variability than
others.14 First, as mentioned, the use of photographs (Denver and BC surveys) versus WinHaze-
generated images (Phoenix and Washington, DC surveys) may play a significant role in
preference studies, perhaps introducing bias (such as suggested by the responses to the 9:00 a.m.
Denver photographs) as well as variability. Further, the use of photographs from different days
and times of day that rely on associated ambient measurements of light extinction to characterize
their VAQ level can introduce two other types of uncertainty. The intrinsic appearance of the
scene can change due to the changing shadow pattern and cloud conditions, and spatial variations
in air quality can result in ambient light extinction measurements not being representative of the
sight-path-averaged light extinction. WinHaze has neither of these sources of uncertainty
12 Only 47% of the BC participants rated a 19.2 dv photograph as acceptable.
13 In the 2001 Washington, D.C. study, a 30.9 dv image was used as a repeated slide. The first time it was shown
56% of the participants rated it as acceptable, and 11% rated it as acceptable the second time it was shown. The
same VAQ level was rated as acceptable by 42% of the participants in the 2009 study (Test 1).
14 Variability here refers to the degree of scatter of the average acceptability ratings for each image around the logit
curve for that city.
January 2010
2-29
DRAFT - Do Not Quote or Cite
-------
1 because the same base photograph is used (i.e. no intrinsic change in scene appearance) and the
2 modeled haze that is displayed in the photograph is determined based on uniform light extinction
3 throughout the scene.
4 Second, variation in the degree of representativeness of the participants and the sizes of
5 the participant samples involved may also be important factors. The small sample size and fairly
6 uniform population of respondents is a plausible explanation for the noisiness of the combined
7 Washington, DC results (35 participants, including 26 from a single consulting firm and 10 of
8 those from a different city) compared with the larger and more representative population of
9 responders from Phoenix (385 participants, carefully selected to be representative of the Phoenix
10 population).
11 A third hypothesis promoted by Smith and Howell (2009) is that the range of VAQ
12 images presented in the survey may influence the results. As discussed above, a more plausible
13 explanation it that the range of haze images shown to participants in the VAQ 1 to 7 rating
14 exercise was interpreted by participants as a restriction on acceptability rating exercise to confine
15 their rating to the range VAQ conditions shown, which for Test 2 was curtailed to only average
16 to good VAQ conditions. When other evidence is taken into account, the Smith and Howell
17 hypothesis seems an even more unlikely explanation for the differences in results between the
18 four urban preference studies. For example the Denver study included photographs with the
19 haziest conditions among the four studies, but resulted in the lowest haze condition for the 50
20 percentile preference ratings among the four, not the highest as might be expected if the range of
21 haze levels were a significant factor influencing the results of preference studies. Also,
22 inspection of the average VAQ 1 to 7 ratings for the Phoenix and Denver studies showed that
23 they spanned the full ratings range of values similar to those for the Smith and Howell Test 1 and
24 3, so the participants in those studies were not presented with a restricted range within which to
25 select acceptable VAQ conditions, suggesting that the range itself was not an important factor
26 influencing their results. Values for the British Columbia 1 to 7 VAQ rating exercise were not
27 readily available.
28 A fourth major hypothesis is that urban visibility preferences may differ by location, and
29 the differences may arise from inherent differences in the cityscape scene used in each city. The
30 key evidence to suggest this hypothesis is that the apparent differences between the Denver
31 results (which found the 50% acceptance criteria occurred in the best VAQ levels among the four
32 cities) and the Washington, DC results (which found the 50% acceptance criteria occurred at the
33 worst VAQ levels among the four cities). This hypothesis suggests that these results may occur
34 because the most prominent and picturesque feature of the cityscape of Denver is the clearly
35 visible snow-covered mountains in the distance, while the prominent and picturesque features of
January 2010 2-30 DRAFT - Do Not Quote or Cite
-------
1 the Washington, DC cityscape are buildings relatively nearby without prominent and/or valued
2 scenic features that are more distant.
3 Finally, and perhaps of significant importance is that the sensitivity of individual scenes
4 to perceived changes in VAQ under changing light extinction levels can be quite different. As in
5 the fourth hypothesis, this may in part explain why the Denver study scene, with its long distance
6 to the mountain backdrop, resulted a preference for the best VAQ level, with a 50% criteria value
7 of about 20 dv, while the Washington, DC study scene, with much shorter sight paths yielded a
8 50% criteria VAQ value at a substantially worse level of about 30 dv. The distinction between
9 the last two hypotheses are that the earlier one speaks to the desirability of seeing distant
10 mountains versus this hypothesis which concerns the ability to perceive changes in haze at lower
11 light extinction levels. Additional studies, including directly comparable studies using similar
12 methods in diverse cities, would be useful to gain further understanding of preferences for urban
13 visibility.
14 Based on the composite results and the effective range of 50* percentile acceptability
15 across the four urban preference studies shown in Figure 2-16, CPLs have been selected in a
16 range from 20 dv to 30 dv (74 Mm"1 to 201 Mm"1) for the purpose of comparing to current and
17 projected conditions in the assessment in chapters 3 and 4 of this document. A midpoint of 25
18 dv (122 Mm"1) was also selected for use in the assessment. These three values provide a low,
19 middle, and high set of light extinction conditions that are used in subsequent chapters of the
20 UFVA to provisionally define daylight hours with urban haze conditions that have been judged
21 unacceptable by the participants of these preference studies. As discussed earlier (section 1.2)
22 PM light extinction is taken to be light extinction minus the Rayleigh scatter (i.e. light scattering
23 by atmospheric gases is about 10 Mm"1), so the low, middle and high CPL levels correspond to
24 PM light extinction levels of about 64 Mm"1, 112 Mm"1 and 191 Mm"1.
January 2010 2-31 DRAFT - Do Not Quote or Cite
-------
1 3 ESTIMATION OF CURRENT PM CONCENTRATIONS AND PM
2 LIGHT EXTINCTION
3 The goals of the "current conditions" portion of this urban-focused visibility impact
4 assessment are to characterize hourly PM light extinction conditions in a set of urban study areas
5 during 2005-2007, in order (1) to improve understanding of the levels, patterns, and causes of
6 daylight hours PM light extinction; (2) to provide the starting point for projections of PM light
7 extinction levels under "what if scenarios; and (3) to examine the correlation between PM light
8 extinction and potential alternative indicator(s) based on PM2.5 mass concentration. This chapter
9 addresses the first goal. Chapter 4 addresses the second goal regarding "what if scenarios.
10 Appendix D addresses the third goal. A number of other appendices address related topics of
11 particular interest in more detail.
12 3.1 SUMMARY OF PREVIOUS CHARACTERIZATIONS OF PM
13 CONCENTRATIONS AND LIGHT EXTINCTION
14 3.1.1 PM2.5 and PM10.2.5
15 Chapter 2 of the 2005 Staff Paper from the previous review and chapters 3 (especially
16 section 3.5) and 9 (especially section 9.2.3) and Annex A of the final ISA (US EPA, 2009a) from
17 the current review present extensive characterizations of the levels, composition, and temporal
18 and spatial patterns of PM2.5 in U.S. urban areas. Both documents present data summaries based
19 on the approximately 1000 PM2.5 monitoring sites in the U.S. The characterizations in the 2005
20 Staff Paper were based on 2001-2003 data. The characterizations in the ISA are based on 2005-
21 2007 data, which is the same time period used in this visibility assessment. While there
22 generally have been reductions in the concentrations of PM2.s in many areas as a result of
23 emission reductions of PM2.5 and its precursors, the general patterns, and the diversity of patterns
24 across areas, noted in the 2005 Staff Paper still prevailed in the 2005-2007 period.
25 Using 2005-2007 air quality data, 38 urban areas violated the annual PM2.5 NAAQS set at
26 a level of 15 |ig/m3 in 1997 and retained in the last review completed in 2006. Seventy-six areas
27 violated the 2006 24-hour NAAQS level of 35 |ig/m3. There is considerable but not complete
28 overlap in the areas not meeting the two NAAQS. It should be noted that in many parts of the
29 U.S., PM2.5 concentrations in 2005 were high relative to the next three years. Figure 3-1
30 illustrates PM2.5 air quality in 2007 by representing each monitor by a symbol whose color
31 reflects the annual mean of the concentration at that site or the 98* percentile 24-hour
32 concentration, in both cases in that one year.
33
January 2010 3-1 DRAFT - Do Not Quote or Cite
-------
1
2
3
th
Figure 3-1. Annual average and 24-hour (98 percentile 24-hour concentrations) PMi.5
concentrations in ug/m3, 2007.
4
5
6
7
Annual
Concentration Range (ng/m3)
• 3.4-12.0(418 Sites)
O 12.1 - 15.0 (366 Sites)
O 15.1 -18.0 (86 Sites)
• 18.1 -22.5 (14 Sites)
Puerto Rico
24-hour
Concentration Range (pg/m3)
• 7-15 (38 Sites)
O 16-35 (662 Sites)
O 36 - 55 (166 Sites)
• 56-73 (18 Sites)
Puerto Rico
January 2010
3-2
DRAFT - Do Not Quote or Cite
-------
1 Each urban area exhibits its own detailed patterns of observed concentration levels,
2 temporal and spatial variation, and composition. These differences are due to differences in local
3 and transported emissions and in meteorology. Because of differences in the placement of PM2.5
4 monitoring sites in each urban area, the actual levels and spatial pattern of PM2.5 and PM2.5
5 species concentrations may not be consistently discernable in all areas. This variability and
6 limited monitoring network make it difficult to offer concise generalizations, although some
7 broad similarities can be drawn among areas.
8 Midwestern, southeastern, and eastern urban areas have much higher sulfate levels than
9 do more western areas, attributable to the much higher emissions of SC>2 in and upwind of them.
10 Upper midwestern areas and to a lesser extent upper eastern areas have notable nitrate
11 concentrations in winter but not in summer, while southeastern areas generally lack notable
12 nitrate even in winter. Many western urban areas have notable nitrate year round. In all areas,
13 carbonaceous material is an important component of PM2.s and is attributable to many emission
14 sources of organic material in PM form and of organic PM precursor gases. In some areas with
15 high local use of wood for residential heating carbonaceous material is dominant during the
16 heating season. PM2.s derived from crustal sources is generally a small fraction of total mass,
17 except during local high wind events or due to brief periods of intercontinental transport of dust
18 from Africa or Asia.
19 Comparison of PM2.5 species concentrations within and outside urban areas leads to the
20 conclusion that, in the eastern areas with high sulfate concentrations, the large majority of the
21 sulfate affecting any given urban area originates outside that area. Inward transport and local
22 generation of nitrate and carbonaceous material are more evenly balanced in eastern areas, with
23 some differences among areas. In western areas, local sources dominate for carbonaceous
24 material and nitrate, with the origins of the small sulfate component being more balanced. See
25 Figure 9-24 of the final ISA (US EPA, 2009a).
26 Southeastern areas have their highest PM2.5 concentrations in the summer, when
27 conditions are most conducive to sulfate formation. More northern areas, being affected by a
28 more balanced mix of contributors, tend not to have such a strongly seasonal pattern. The
29 seasonal patterns in western areas are individual and varied, related to differences in local
30 sources and formation and dispersion conditions. In all areas, inversion conditions with low
31 wind speeds are conducive to high concentrations due to the trapping of emissions from local
32 sources. Some western areas, especially those with valley or bowl-like topography, are
33 especially affected.
34 There is at present no systematic monitoring network in place for PMio-2.5, as states have
35 until January 1, 2011, to implement required monitoring sites for PMio-2.5. Consequently,
36 estimates of PMio-2.5 must be developed using data from PM2.5 and PMio monitoring sites and
January 2010 3-3 DRAFT - Do Not Quote or Cite
-------
1 equipment, which are not always collocated and consistent. The 2005 Staff Paper presented such
2 estimates in section 2.4.3. The final ISA presents such estimates in Figure 3-10 and Table 3-9 of
3 section 3.5.1.1. The 2005 Staff Paper used a data-inclusive approach in which the best available
4 data on PM2.5 and PMio concentrations - in some cases not very robust data - were used to
5 estimate 2001-2003 PMi0-2.5 concentrations for 351 metropolitan area counties. For these
6 counties, the annual mean PMio-2.5 concentrations were generally estimated to be below 40
7 ug/m3, with one maximum value as high as 64 ug/m3 and a median of about 10-11 ug/m3. The
8 ISA used a much more data-restrictive approach based only on paired (collocated) low-volume
9 filter-based samplers for both PMio and PM2.5. The ISA reports that only 40 counties have such
10 paired samplers. Using these available co-located PM measurements from 2005-2007, the mean
11 24-hr PMio-2.5 concentration in these 40 counties was 13 ug/m3. This urban visibility assessment
12 has used a data-inclusive approach to estimating PMio-2.5 concentrations, similar to that used for
13 the 2005 Staff Paper, where needed to obtain hourly PMio-2.5 estimates for 15 study areas, which
14 are reported below in section 3.3.2.
15 Additional detail on PM2.5, PMio, and PMio-2.5 concentrations, composition, and patterns
16 appears in section 3.5.1.1 of the ISA. Also, chapter 6 of the 2004 PM Assessment by NARSTO
17 contains more detailed characterizations of PM in different parts of the U.S.
18 3.1.2 PM light extinction
19 While total light extinction is directly measurable using a transmissometer and PM light
20 extinction can be measured with other instruments, there are very few regularly operating
21 monitoring sites measuring either form of light extinction in urban areas, and generally those that
22 do operate do not submit data to AQS.15 Consequently, any characterization of PM light
23 extinction conditions based on actual measurements is necessarily less comprehensive than for
24 PM2.5 and PMio-2.5. Many monitoring sites that employ nephelometers, which measure light
25 scattering, operate that equipment in a heated mode for purposes of tracking "dry" PM2.s mass
26 concentrations, and actual light scattering due to ambient PM is not reportable. There are many
27 more filter-based Aethalometers® and similar instruments for measuring light absorption in
28 operation and reporting to AQS, but light absorption is typically a small fraction of total PM
29 light extinction, so these data alone are not a good indicator of overall PM light extinction in
15 EPA is aware of routine, long-term direct measurement of light extinction using transmissometers only in the
Phoenix, AZ, Denver, CO, and Washington, DC urban areas, none of which submit data to AQS, although the site in
Washington submits data to the IMPROVE program data system. Also, there is a large network of "visual range"
monitors in operation at U.S. airports, aimed at providing information to determine landing and takeoff safety. Due
to their locations and to the lack of data resolution (values of visual range above the level needed for unlimited
airport operations are not individually reported) the data from these monitors are not suitable for use in this
assessment. The final PM ISA discusses these monitors in section 9.2.2.3.
January 2010 3-4 DRAFT - Do Not Quote or Cite
-------
1 urban areas. Also, there are unresolved issues of data corrections and comparability for the light
2 absorption data from these instruments now residing in AQS.
3 PM light extinction can be "reconstructed" from measurements of PM2.5 mass
4 components and PMio-2.5 concentrations, in combination with relative humidity values, using
5 either of two versions of the formula known as the IMPROVE algorithm but excluding its term
6 for Rayleigh scattering by gases in clean air. (Section 9.2.2.2 of the ISA gives an overview of
7 the algorithm and its basis. Section 3.2.3 of this document discusses the application of the
8 original version of the IMPROVE algorithm in this assessment. PM2.5 component measurements
9 are generally available only on a 24-hour average basis, so it generally is possible to estimate
10 only 24-hour average PM light extinction, unless additional information on hourly patterns is
11 brought to bear.16 Because EPA's Regional Haze Rule (RHR) currently requires states to
12 address visibility problems in Class I visibility protection areas, which are nearly all rural and
13 remote, there is a large body of literature characterizing light extinction in remote rural areas,
14 based on data from the IMPROVE network's 24-hour samplers and on special studies. Sections
15 9.2.3.2 and 9.2.3.4 of the ISA summarize this literature. Section 9.2.3.3 of the ISA contrasts
16 concentrations of PM and PM components between rural and urban areas using data from the
17 rural IMPROVE network and the urban Chemical Speciation Network (CSN), but does not
18 present estimates of light extinction in urban areas.
19 The CSN network provides 24-hour PM2.5 species measurements at about 200 urban
20 sites, from which mass components can be derived. These sites have a mix of daily, one day in
21 three, and one day in six sampling schedules. The 2005 Staff Paper (and its references) may be
22 the only readily available prior assessment to use these urban PM2.5 speciation monitoring data,
23 along with estimates of PMio-2.5 concentrations and data on relative humidity, to reconstruct daily
24 24-hour average light extinction in urban areas, for the year 2003.1? One presentation of the
25 results was in the form of a scatter plot of daily 24-hour reconstructed light extinction versus 24-
26 hour PM2.5 concentration. This graphic appears here as Figure 3-2. (For the immediate purpose
27 of this section, it is the distribution of the data points along the y-axis that is of interest, not the
28 relationship between light extinction and PM2.5 concentrations; the latter subject is addressed in
16 When the IMPROVE algorithm is used to estimate 24-hour light extinction from 24-hour PM2 5 species and
PM10-2 5 concentrations, an assumption is made that every hour has the same PM concentrations but its own relative
humidity value. Hourly estimates of light extinction, including the strongly non-linear effect of relative humidity,
are then averaged to get the 24-hour light extinction estimate.
17 Estimates of light extinction in the 2005 Staff Paper include Rayleigh scattering of 10 Mm-1 and thus represent
"total" light extinction (excluding NO2 absorption). Adjustment for consistency must be made before any close
comparisons to PM light extinction values in this document.
January 2010 3-5 DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
Appendix D.) Generally, most days have light extinction below 200 inverse megameters (Mm" ),
but a small percentage of values were as high as about 750 Mm"1.18
Figure 3-2. Reconstructed 24-hour light extinction in U.S. urban areas in 2003
Source: Schmidt et al., 2005
Output D.3
(Relationship RE & PM2.5; Diurnal RE; Timeframe)
2 of 30
Heccn, Extinctian
m
70C
GOC
50C
IOC
Significant relationship (low p-yalue)
M:RE=y+8.2*PM15,RH7i
, , I , , :
9 19 EC ::u 4* M
PM.5
r,o
Relationship benveen reconstructed light eitmction (RI) and 24-bour average PM,:, 2003. U?ing acraaljfJZfl)
Unfortunately, the file of paired data used to create this scatter plot is no longer available, so the actual
distribution of light extinction values cannot be described more specifically.
January 2010
3-6
DRAFT - Do Not Quote or Cite
-------
1 In addition to this scatter plot, a table developed for the previous PM NAAQS review
2 presented the annual average of estimates of 24-hour reconstructed light extinction values,
3 averaged across 161 urban areas grouped into seven regions (Schmidt, et al., 2005). Table 3-1
4 reproduces these estimates. For regions excluding Southern California, annual average 24-hour
5 light extinction ranged from 73 to 118 Mm"1. The estimate of the annual average 24-hour light
6 extinction for Southern California was 168 Mm"1. These estimates were based on 10-year
7 average 1-hour relative humidity values and 2003 PM monitoring data.
9
10
11
12
13
14
Table 3-1. Annual Mean Reconstructed 24-hour Light Extinction Estimates
by Region (Mm"1)
Region
Northeast
Southeast
Industrial Midwest
Upper Midwest
Southwest
Northwest
Southern California
Reconstructed 24-hour Light
Extinction in 2003
108
98
118
80
73
76
168
Source: Output D.3, Schmidt etal., 2005. We note these regions were used to summarize PM2 5 patterns for the PM
NAAQS review 1997 (US EPA, 1996).
January 2010
3-7
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Figure 3-3 is a contour map of annual average reconstructed 24-hour total light extinction
based on IMPROVE monitoring sites in 2000-2004, nearly all of which are remote and rural (the
three urban sites in Phoenix, AZ, Washington, DC, and Puget Sound, WA are indicated by
square symbols). A comparison of the mean urban light extinction levels by region listed in
Table 3-1, with this map of rural light extinction indicates that in most parts of the U.S., light
extinction levels in urban areas are notably higher than in the surrounding remote rural area, with
the northeast and the southeast regions having the most similarity between rural and urban light
extinction levels. This is consistent with observations of an "urban excess" of PM2.5 and
Figure 3-3. Isopleth map of annual total reconstructed particulate extinction based
on 2000-2004 IMPROVE data.
t •'",
— 105
B823
73.1
84.0
549
45.7
36.6
274
18.3
9.14
0.00
Mm-'
i
.">
• IMPROVE Site
• IMPROVE Urban Site
Puerto Rico t
Virgin Islands
(Source: Spatial and Seasonal Patterns and Temporal Variability of Haze and its Constituents in the
United States Report IV, November 2006.)
PMio-2.5, and with the known high regional concentrations of sulfate in these eastern areas.
One-hour light extinction values of course vary above and below the 24-hour average,
due to diurnal variations in PM2.5 component concentrations, PMio-2.5 concentrations, and
relative humidity. Although light extinction was formally reconstructed on an hourly basis in the
2005 Staff Paper analysis for the last review cited above, the actual full strength of the diurnal
January 2010
DRAFT - Do Not Quote or Cite
-------
1 pattern could not be discerned in that analysis because component mix was assumed not to vary
2 from hour to hour. Under the unverified assumption of constant component mix and using actual
3 hourly relative humidity data, the daily maximum daylight 1-hour light extinction values were
4 roughly 50 percent higher than the 24-hour average light extinction values.19 The new analysis
5 presented in this document includes a closer look at diurnal patterns, for 15 study areas.
6 3.2 OVERVIEW OF APPROACH AND DATA SOURCES FOR URBAN STUDY
7 ANALYSIS
8 As explained above, there are limited data from direct measurements of light extinction in
9 urban areas. Consequently, this assessment has reconstructed hourly PM light extinction levels
10 for daylight hours from values of hourly PM2.5 components, PMio-zs, and relative humidity.
11 Hourly monitoring data for PM2.5 components and PMio-2.5 are also generally lacking, so the
12 estimates of these parameters necessarily in turn have been developed from a combination of
13 other available ambient monitoring data and air quality modeling results from a chemical
14 transport model (CTM) run. Specifically, the ambient monitoring data starting points are 24-
15 hour PM2.5 mass measured by filter-based Federal Reference Method (FRM) or Federal
r\f\
16 Equivalent Method (FEM) monitors , 24-hour PM2.5 components measured by the filter-based
17 monitors of the Chemical Speciation Network, and hourly PM2.5 mass measured by continuous
18 instruments such as the Tapered Element Oscillating Microbalance (TEOM), beta attenuation
19 monitors (BAMs), and nephelometers, which were used at different sites. The CTM-based
20 diurnal profiles for individual components, in conjunction with hourly PM2.5 measurements, are
21 used to adjust and allocate the 24-hour PM2.5 components measurements to individual hours of
22 each day, as described in detail below. In addition, levels of hourly PMio-2.5 mass are calculated
23 from separate measurements of hourly PMio and hourly PM2.s if both are available, or by
24 applying PMio-2.5 to PM2.5 ratios to hourly PM2.5 data if both types of hourly measurements are
25 not available. The ambient data are from 2005-2007 and were all obtained from AQS in the first
26 half of 2009.
27 The CTM run was the "actual emissions" or "validation" run of the 2004 CMAQ
28 modeling platform with boundary conditions provided by GEOS-Chem global scale CTM.21 The
29 CTM modeling is used as one element in the development of realistic diurnal variations for each
19 These observations on diurnal patterns come from examination of "Output D.3 (Relationship RE & PM25; Diurnal
RE; Timeframe) 8 of 30" and "Output D.3 (Relationship RE & PM25; Diurnal RE; Timeframe) 17 of 30", Analyses
of Particulate Matter (PM) Data for the PM NAAQS Review, Schmidt et al., 2005.
20 Filter-based Federal Reference Method samplers and filter-based Federal Equivalent Method samplers will both
be referred to as FRM samplers in the remainder of this document.
21 GEOS-Chem is the NASA Goddard Earth Observing System-CHEMistty (global 3-D CTM for atmospheric
composition). This modeling platform, with an appropriately different emissions scenario, is also the basis for the
estimates of policy relevant background concentrations of PM25 presented in section 3.6 of the ISA (US EPA,
2009a).
January 2010 3-9 DRAFT - Do Not Quote or Cite
-------
1 of the major PM2.5 components used to estimate PM light extinction, anchored to site-specific,
2 day-specific measurements of 24-hour concentrations. That is, monthly averaged diurnal
3 profiles for the five major components were generated using the CTM results, which were then
4 combined with hour-specific measurements of PM2.5 to generate hourly concentration variations
5 for each of the 24-hour CSN sample days during the 2005-2007 period.
6 3.2.1 Study Period, Study Areas, Monitoring Sites, and Sources of Ambient PM Data
7 At the time this assessment began, the ambient monitoring data from 2005-2007, but not
8 from 2008, had been certified as accurate and complete by the state/local monitoring agencies
9 that collected them, and the data had been extensively summarized and presented in the first draft
10 ISA. The EPA staff aimed to develop estimates of daylight hours PM light extinction for a
11 reasonably representative number of days in each year of 2005-2007, to allow the application of
12 statistical forms based on three years of data. However, as explained in more detail below, in
13 several study areas the limited availability of starting data for these estimates resulted in estimate
14 sets that do not cover all three years. Also, even in areas with some data in all three years, the
15 number of days with valid estimates differs by year and is in some cases not large by typical
16 standards of monitoring data completeness.
17 For efficiency in the analysis, this visibility assessment uses the same 15 urban study
18 areas selected for the health risk assessment. These areas are listed in Table 3-2, along with the
19 area-wide (maximum) FRM-based 2005-2007 annual and 24-hour PM2.5 design values for each
20 study area based on the highest-reading monitor in each area, and for the specific site used in this
99
21 assessment. (See below for an explanation of the "site-specific" columns in Table 3-2.)
22
2005-2007 PM2 5 design values were taken from the information posted at
http://www.epa.gov/airtrends/values.html. and are consistent with the design values used in the health risk
assessment to "roll back" current concentrations to represent achievement of alternative annual and 24-hour PM25
NAAQS. Except in Dallas and Fresno, the area-wide design values are the highest design values of any monitoring
site in the designated (1997 NAAQS) nonattainment area that has sufficiently complete data to allow the calculation
of a design value according to the provisions of 40 CFR 50 appendix N. For Dallas, the design values come from a
site with nearly complete data, and are somewhat higher than the highest values from a site with complete data (see
the draft PM Risk Assessment, US EPA, 2009c, section 3.2.3) For Fresno, the area-wide design value is for the
Fresno-Madera CSA, which is only a portion of the San Joaquin Valley nonattainment area. Also, note that there
are three cases in which the nonattainment area does not include certain areas sometimes thought of as being part of
the area named in Table 2; monitors in these non-included areas were not considered in this assessment. (1) The
design value shown for Pittsburgh is for the Pittsburgh-Beaver nonattainment area; the Liberty-Clairton
nonattainment area is within the Pittsburgh CBS A but is distinct for regulatory purposes, and was not considered in
this assessment. (2) Baltimore was treated separately, although part of a CSA with Washington DC. (3) Berks Co.,
PA is part of the Philadelphia-Camden-Vineland CSA, but not part of the Philadelphia-Wilmington nonattainment
area.
January 2010 3-10 DRAFT - Do Not Quote or Cite
-------
Table 3-2. Urban Visibility Assessment Study Areas
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Area-wide 2005-
2007
Annual
Design Value
Oig/m3)
10.2
17.4
19.6
12.6
11.6
12.8
15.8
16.5
18.7
16.2
17.2
16.5
15.6
15.0
15.9
Area-wide
2005-2007
24-hour
Design Value
Oig/ni3)
43
63
55
32
55
26
31
39
44
35
43
43
37
38
42
Site-specific
2005-2007
Annual
Design Value
Oig/m3)
Same
Same
Same
7.9
10.7
11.5
13.1
14.5
Same
15.7
Same
15.0
14.5
14.7
14.4
Site-specific
2005-2007
24-hour
Design Value
Oig/ni3)
Same
Same
Same
15
48
25
25
34
Same
33
Same
40
35
37
42
2005 Staff
Paper
Region
(See map in
Table 3-1)
Northwest
Southern
California*
Southern
California
Southwest
Northwest
Southeast
Southeast
Midwest
Southeast
Southeast
Midwest
Industrial
Midwest
Northeast
Northeast
Northeast
* While not generally considered to be part of Southern California as the term is commonly used, Fresno lies just
south of the line used in the 2005 Staff Paper (based on earlier work by others) to separate the Southern California
region from the Northwest region.
2
3 For time reasons and because it was anticipated that some study areas would not contain
4 more than one suitable study site, EPA staff sought to identify the single best study site in each
5 area. In identifying the single best study site in each study area first consideration was given to
6 the availability of collocated 24-hour data on PM2.5 and its components, because the contribution
7 of PM2.5 components to PM light extinction will typically dominate the contribution from PMio-
8 2.5- Ideally, within each study area the three types of PM2.5 data (FRM PM2.5, CSN PM2.5
9 components, continuous PlV^.s) would be available at a common site, and that site would be
10 located in a manner consistent with reliance on it to characterize visibility as it would be
11 perceived by a large number of area residents and visitors. As can be seen in Table 3-2, in 10 of
12 the 15 study areas the site providing FRM data for this assessment is not the area-wide design
13 value site, because the area-wide design value site did not have collocated CSN and/or
14 continuous PIVb.s data.
15 Appendix A provides details on the site(s) identified and used in each study area,
16 including information on the type of monitoring equipment that provided the data and other
17 information that may help interpret the results of the analysis. A portion of this table for a single
January 2010
3-11
DRAFT - Do Not Quote or Cite
-------
1 site - Tacoma - is presented here as Table 3-3 as an example. When viewing this document
2 electronically, the site IDs in these tables are active links and can be used to view the location of
9^
3 the site via GoogleMaps.
4 In 11 of the study areas, the three types of PM2.5 data were available at a common site. In
5 the remaining four areas, Phoenix, AZ, Pittsburgh, PA, Baltimore, MD, and St. Louis, MO-IL,
6 two types of data were available at one site, but the remaining type of data had to be taken from
7 another site and treated as being representative of the former site.
8 The monitoring agencies described all but one of these sites as neighborhood or urban
9 scale, indicating those agencies' opinion that the sites represent concentrations in an area at least
10 0.5 to 4 km across. An aerial view of the remaining site (in Phoenix) which did not have a scale
11 characterization recorded in AQS suggests that it may be middle or neighborhood scale. As
12 already stated, selected sites are not necessarily the locations of the maximum measured annual
13 or 24-hour PM2.5 levels in their urban area.
14 Site days which were missing 1-hour PM2.5 concentration data points for more than 25
15 percent of daylight hours were excluded from the analysis, because such data gaps were judged
16 to result in too much uncertainty in estimates of 1-hour PM2.5 components, 1-hour light
17 extinction, and daily maximum light extinction. Days with fewer missing 1-hour PM2.5
18 concentration data points were retained, but no estimate of light extinction was made for hours
19 without 1-hour PM2.5 concentration data (see below for more explanation).
20 Hourly PMio-2.5 presented more varied challenges. In four areas (Birmingham, Detroit,
21 Baltimore, and Philadelphia) the site that provides the continuous PM2.5 data also hosts a
22 continuous FEM PMio monitor, and hourly PMio-2.5 could be calculated by difference for most
23 hours. In other areas, this was not the case, and either (1) hourly instruments at two different
24 sites were used in this subtraction (Tacoma, Los Angeles-South Coast Air Basin, Phoenix, St.
25 Louis, Atlanta, and New York-N. New Jersey) or (2) a single regionally applicable PMio-2.5 to
26 PM2.5 ratio calculated as part of the last review based on 2001-2003 24-hour FRM/FEM PMio
27 and PM2.5 samples was applied to 2005-2007 hourly PM2.5 data to estimate hourly PMio-2.5
28 (Fresno, Salt Lake City, Dallas, Houston, and Pittsburgh). In the case of Los Angeles-South
29 Coast Air Basin, the continuous PMio and PM2.5 sites were quite distant and separated by a range
30 of hills, so the estimates of PMio-2.5 and its contribution to PM light extinction are more uncertain
31 than if the monitors were clearly within the same air mass. Obviously, for the five study areas
32 for which 1-hour PMio-2.5 was estimated by application of ratios, PMio-2.5 estimates can only
23 Additional meta data on each monitoring site, and access to daily and annual data listings, can be conveniently
obtained using GoogleEarth and the PM25, PM10, and CSN monitoring network KML files that can be downloaded
from http://www.epa. gov/airexplorer/monitor kml.htm.
January 2010 3-12 DRAFT - Do Not Quote or Cite
-------
1 represent broad trends, not hour-specific conditions at the particular site. More description of the
2 methods used for estimating hourly PMio-2.5 appears in section 3.3.2.
3 The sampling schedule for CSN PM2.5 speciation monitoring was one-in-six days for
4 Tacoma, Phoenix, Houston, Detroit, and Philadelphia, and one-in-three days for the other study
5 areas. Not every scheduled CSN site day in 2005-2007 had data for all three types of PM2.s data,
6 due to missed or invalid samples. Also, for continuous PM2.5, values for a small number of hours
7 of an otherwise data-sufficient day were sometimes missing, due to equipment failure or
8 servicing. EPA staff retained only those days in which 75 percent or more of daylight hours had
9 measurements of PM2.5 (see section 3.3. for more details). If for isolated hours at a site (or site
10 pair) with collocated measurements, PMio-2.5 concentrations could not be estimated because of
11 gaps in the same-hour continuous PMio and/or PM2.5 data, EPA staff used the regional ratio
12 approach described above to estimate PMio-2.5 for those specific hours. Table 3-4 provides more
13 detailed information on the quarterly distribution of the successfully matched and sufficiently
14 complete data available for use. As described later, for some parts of this assessment EPA staff
15 substituted data for the single missing quarters of data in Phoenix and Houston, to achieve
16 seasonal balance.
17 In this assessment, we have not excluded PM concentration data that may have been
18 affected by exceptional events such as wildfires and wind storms. Under EPA's Exceptional
19 Events rule, for existing NAAQS states may request exclusion of such data from regulatory
20 determinations, and accordingly such data are not reflected in design values for existing NAAQS
21 once exclusion is approved by EPA. A similar arrangement presumably would apply to a new or
22 revised secondary PM NAAQS. Design values for PM light extinction under current conditions
23 (Table 4-2) and percentage reductions to "just meet" alternative secondary NAAQS based on PM
24 light extinction (Table 4-3), presented below, may thus be overestimates. Overestimation is
25 more likely for the western study sites than for the eastern study sites. However, PM2.5 design
26 values shown in Table 3-2, and associated estimates of the reductions needed from 2005-2007
27 PM2.5 level to just meet alternative secondary NAAQS based on PM2.5 mass (Table 4-4) do
28 reflect the exclusion of at least some data affected by exceptional events.
January 2010 3-13 DRAFT - Do Not Quote or Cite
-------
1
2
3
Table 3-3. PM2.s Monitoring Sites and Monitors Providing 2005-2007 Data for the Tacoma
Study Area
Study
Area
First PM2.5
Monitoring Site
Second PMi.s
Monitoring Site (if
applicable)
data source for
Tacoma
AQS ID 530530029
State: Washington
City: Tacoma
MSA: Tacoma, WA
Local Site Name: TACOMA -
LSTREET
Address: 7802 SOUTH L
STREET, TACOMA
0.5 miles east of 1-5
2005-2007 annual DV = 10.2
2005-2007 24-hr DV = 43
This is the highest 24-hour
PM2 5 DV site in the Seattle-
Tacoma-Olympia, WA annual
PM2 5 nonattainment area
Neighborhood Scale
Parameters taken from this
site:
* 24-hour FRMPM2 5 mass
(AQS parameter 88101;
one-in-three sampling
schedule)
* PM2 5 speciation (one-in-
six sampling schedule)
* l-hourPM25mass (AQS
parameter 88502,
Acceptable PM2 5 AQI &
Speciation Mass)
Correlated Radiance
Research M903
Nephelometry
No continuous PM10
monitoring at this site, see
right hand column.
N/A
AQS ID 530530031
State: Washington
City: Tacoma
MSA: Tacoma, WA
Local Site Name: TACOMA -
ALEXANDER AVE
Address: 2301 ALEXANDER AVE,
TACOMA, WA
6.4 miles NNE of PM2 5 site
Neighborhood Scale
Parameters taken from this site:
* 1 -hour PM{ o STP mass (AQS parameter
81102)
* Sample Collection Method:
INSTRUMENTAL-R&P SA246B-
INLET
* Sample Analysis Method: TEOM-
GRAVIMETRIC
7% of PM10.2.5 values were determined using
regional average PM10_2 s:PM2 5 ratios from
2005 Staff Paper
Additional Explanation
In this Table, the 1-hour concentration parameter "88502, Acceptable PM2.5 AQI & Speciation Mass" is the same as the ISA refers to
as "FRM-like" PM2 5 mass. An entry of "88501, PM2.5 Raw Data" indicates that the monitoring agency makes no representation as to
the degree of correlation with FRM PM2.5 mass. The latter type of continuous PM2 5 data were used only when the former were
unavailable.
Where PMio was reported in STP, it was converted to LC before PMio-is was calculated.
For convenience, continuous PM2.5 data was obtained through the AirNow website rather than from AQS, as an initial exploration
indicated that not all the desired 1-hour data had been submitted to AQS.
January 2010
3-14
DRAFT - Do Not Quote or Cite
-------
1
2
Table 3-4. Number of days per quarter in each study area
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Total Number
of Days
110
324
302
86
306
274
149
294
350
295
141
284
187
145
228
2005
Qi
0
19
28
0
27
22
21
27
30
22
12
26
19
15
22
Q2
0
24
28
13
28
24
20
27
30
25
12
23
17
11
23
Q3
0
27
22
11
30
26
10
24
29
25
10
25
15
13
13
Q4
0
27
28
14
26
22
14
27
30
24
11
23
11
10
15
2006
Qi
13
30
26
12
20
23
14
28
29
28
12
22
15
9
23
Q2
15
29
26
13
28
23
12
19
29
27
13
25
16
13
19
Q3
15
29
27
11
31
24
8
27
30
26
11
24
19
10
18
Q4
14
27
22
12
20
24
12
29
30
27
15
26
18
13
21
2007
Qi
13
26
21
0
23
18
15
29
30
25
11
22
12
13
19
Q2
13
28
26
0
25
23
14
25
30
19
11
22
12
14
15
Q3
14
30
24
0
19
24
9
22
27
26
12
23
17
12
19
Q4
13
28
24
0
29
21
0
10
26
21
11
23
16
12
21
Note: Only days with matched and sufficiently complete data were retained in the assessment.
January 2010
3-15
DRAFT - Do Not Quote or Cite
-------
1 3.2.2 Use of CMAQ Model Validation Runs for 2004 to Augment Ambient Data
2 Because systematic monitoring data on hourly PM2.5 component concentrations are not
3 available for most of the 15 study areas, EPA staff extracted and applied certain information
4 from the modeling platform for calendar year 2004 described in section 3.7.1.2 of the ISA, in
5 which the global-scale circulation model GEOS-Chem was paired with the regional scale air
6 quality model CMAQ.24 The main use of this platform in the ISA is to estimate policy-relevant
7 background concentrations of PM2.5. For the urban-focused visibility assessment described here,
8 however, we used results from the validation run of the platform, in which emissions for all
9 emission source types and countries are included, to develop realistic diurnal variations of the
10 major PM2.5 components.
11 The EPA staff identified the one or more 36 km-by-36 km CMAQ grid cells generally
12 corresponding to the urbanized area surrounding each study site, thus omitting grid cells
9S
13 dominated by rural land uses. We then extracted from the detailed model output for these grid
14 cells the day/hour-specific concentrations of sulfate, nitrate, elemental carbon, organic carbon,
15 and "crustal/unspeciated" PM2.5 during 2004, and then we averaged across grid cells and then
16 across days within the month for each individual hour of the day.26 Thus, for each species, EPA
17 staff obtained 24 hour-of-day values for a month, for each of the 12 calendar months. We then
18 averaged the 24 hour-of-day values in each monthly set for each component to obtain the
19 corresponding 24-hour average concentration for the month. We then divided each hour-of-day
20 value by the 24-hour value, to obtain a normalized diurnal profile for the pollutant, which was
21 taken as the initial representation of all days in that month for 2005, 2006, and 2007 (but further
22 adjusted day-by-day in a later step). In total, this resulted in 5 (components) x 12 (months) x 15
23 (study areas) = 900 profiles. Visual examination of a number of these showed them to be
24 reasonably smooth and generally to show morning (and sometimes also late afternoon/evening)
25 peaks which are the anticipated effect of higher vehicle traffic and lower mixing heights. The
26 peaks were generally moderate, as would be expected in light of the averaging of predictions for
27 multiple large grid cells, the averaging across days, and the generally moderate diurnal profiles
28 for SMOKE pre-processing of emissions in the CMAQ modeling platform. (Note, however, that
24 Similar modeling was not available for 2005, 2006, or 2007.
25 Urbanized area here refers to a specific land area identified by the U.S. Census Bureau based on population
density and other factors. Shape files for these areas were compared to the CMAQ grid to identify the grid cells to
be used.
26 For several of the listed components that are not direct CMAQ outputs, concentrations were estimated by post-
processing to aggregate the appropriate CMAQ outputs. The "crustal/unspeciated" CMAQ output results from non-
reactive dispersion of that portion of the PM25 emission inputs not assigned during SMOKE processing to a more
specific CMAQ species, and is considered in most EPA analyses to represent the same material as the "soil"
component reported for IMPROVE sampling.
January 2010 3-16 DRAFT - Do Not Quote or Cite
-------
1 as described below a later step in the estimation process reduces the smoothness in the diurnal
2 pattern of PM components.) Sulfate, as would be expected for a regionally transported pollutant,
3 generally had a flatter diurnal profile than for other components. Hourly nitrate concentrations
4 were low when expected: during warmer months and in warmer areas. Figure 3-4 shows
5 example diurnal profiles for the five PM2.s components, for the Detroit study area for the months
6 of January and August. Diurnal profiles like these were applied to 24-hour CSN measurements
7 of component concentrations, as explained in detail below.
8 3.2.3 Use of Original IMPROVE Algorithm to Estimate PM light extinction
9 The EPA staff used the original IMPROVE light extinction algorithm, rather than the
10 more recent revised version, because the original version is considered more representative of
11 urban situations, when emissions are still fresh rather than aged as at remote IMPROVE sites.27
12 To maintain consistency with the form of the candidate protection levels (CPLs) for PM light
13 extinction identified in chapter 2, EPA staff excluded from the IMPROVE algorithm for total
14 light extinction the term for Rayleigh scattering by gases in clean air. The formula for PM light
15 extinction using the traditional IMPROVE algorithm but without the Rayleigh scattering term is
16 shown below.
17 bextPM = 3 x f(RH) x [Sulfate]
18 + 3 xf(RH)x [Nitrate]
19 + 4 x [Organic Mass]
20 +10 x [Elemental Carbon]
21 + 1 x (Fine Soil]
22 + 0.6 x [Coarse Mass]
23 PM light extinction (bextPM) is in units of Mm"1, the mass concentrations of the
24 components indicated in brackets are in ug/m3, and f(RH) is the unitless water growth term that
25 depends on relative humidity. We refer to the first five terms in this algorithm as the five PM2.5
26 components. In this algorithm, the sulfate and nitrate components are to be expressed as fully
27 neutralized and as retained and measured in the IMPROVE sampling and laboratory methods.
28 Associated water is to be omitted from all bracketed terms since the water absorption effect is
29 reflected in the f(RH) term. The organic mass component is to include the mass of associated
30 elements in addition to carbon. As described below, we included steps in our development of
31 estimates of hourly component concentration to ensure consistency with these aspects of the
32
27 Other differences between the original and revised algorithms include estimates of sea salt contributions which
can be important for near-coastal locations, inclusion of site-elevation specific Rayleigh light scattering and
provision for calculating NO2 light absorption when NO2 data are available. Their exclusion in this assessment is not
expected to make any appreciable difference to the results or conclusions.
January 2010 3-17 DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Figure 3-4. January and August monthly average diurnal profiles of PM2.5 components
derived from the 2004 CMAQ modeling platform, for the Detroit study area.
CMAQ Relative Patterns (JANUARY)
-soil
-ec
no3
so4
-ocm
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
CMAQ Relative Patterns (AUGUST)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
January 2010
3-18
DRAFT - Do Not Quote or Cite
-------
1 IMPROVE algorithm.
2 3.3 DETAILED STEPS
3 3.3.1 Hourly PM2.5 Component Concentrations
4 The task of estimating hourly PM2.5 component concentrations is in a sense over-
5 determined, given the four types of available information: 24-hour PM2.5 mass by filter-based
6 FRM, 24-hour component concentrations by CSN, hourly PM2.5 mass by continuous instrument,
7 and diurnal profiles of components from the 2004 CMAQ run. There are multiple ways in which
8 two or three of these four data sources could be used to estimate hourly PM2.5 component
9 concentrations, and the result generally can be expected to be at least somewhat inconsistent with
10 the information in the remaining data source(s). For example, each 24-hour PM2.5 component
11 mass from CSN sampling can be apportioned to hours based on the monthly average diurnal
12 profile developed from the 2004 CMAQ run, but then in general the hourly values of PM2.5 mass
13 determined by summing the components in an hour would not exactly match the data from the
14 continuous PM2.5 instrument. EPA staff therefore used a sequence of steps which achieves a
15 prioritized compromise among the data sources. In this sequence, we have given greater weight
16 to the 24-hour FRM, CSN, and continuous PM2.s mass data because these are instrument-based
17 and location- and day-specific, than to the CMAQ-based profiles which are CTM-based,
18 averaged to the month, and extrapolated from 2004 to each of 2005, 2006, and 2007.
19 Because of differences in filter materials, sample collection, laboratory analysis, and data
20 reporting, there are differences between the contribution of some PM components to PM2.5 mass
21 as reported by a filter-based 24-hour FRM sampler, and the mass of the same components as
22 reported by CSN (or IMPROVE) sampling. The following summary of these differences may be
23 helpful in understanding the steps used to develop estimates of hourly PM2.5 components in this
24 analysis. In the IMPROVE algorithm for reconstructing light extinction, the light extinction
25 contribution multipliers per unit of mass concentration of components are not all the same for the
26 five principal components. Consequently, care is required to estimate these components as
27 consistently as possible with the IMPROVE sampling and analytical methods so that particle
28 mass is correctly assigned to the right component.
29 • Nitrate: CSN (and IMPROVE) sampling uses a Nylon filter for purposes of nitrate ion
30 quantification, while FRM sampling uses a Teflon filter for PM2.5 mass as a whole. The
31 Nylon filter limits the loss of nitrate in the form of nitric acid vapor which could
32 otherwise occur if the filter temperature rises above the temperature at the time of
33 collection, compared to the Teflon filter. The fine particle nitrate ion collected on nylon
34 and Teflon filters are assumed to be associated with ammonium ions, and for this analysis
January 2010 3-19 DRAFT - Do Not Quote or Cite
-------
r\f\
1 ammonium is assumed to evaporate at the same rate as nitrate on the FRM filters .
2 Hence, the nitrate ion and calculated ammonium nitrate concentrations reported by CSN
3 (and IMPROVE) sampling typically will be higher than the nitrate contribution to FRM
4 PM2.5 mass, particularly under warm ambient conditions. The latter steps make nitrate
5 mass as reported for a CSN (or IMPROVE) site higher than the nitrate contribution to
6 PM2.s mass reported by a FRM sampler at the same site. On the other hand, FRM
7 sampling may result in some water that is associated with nitrate being included in the
8 reported PM2.5 mass, while the nitrate mass reported by CSN (or IMPROVE) sampling
9 excludes all water. Continuous PM2.5 samplers employ a variety of methods for
10 measuring PM2.s mass, with correspondingly different behaviors regarding retention/loss
11 of nitrate. In this assessment's approach to estimating actual ambient concentrations and
12 PM light extinction, the FRM measurement of nitrate is used in the calculation of the
13 concentration of organic carbonaceous material, but not in estimating ambient
14 concentrations of nitrate or PM light extinction. The CSN-reported nitrate ion
15 concentration and corresponding ammonium nitrate mass is used for the latter purposes.
16 • Sulfate: Unlike nitrate, sulfate is not subject to loss once collected by a filter, so the
17 sulfate ion mass reported by a CSN (or IMPROVE) sampler will be about the same as the
18 contribution of sulfate ion to the mass reported by FRM sampling. In FRM sampling,
19 sulfate ion may not be fully neutralized. When IMPROVE data are used to estimate light
20 extinction, it is assumed that sulfate ion is fully neutralized. Even more important than
21 nitrate, FRM sampling results in water that is associated with sulfate being included in
22 the reported PM2.5 mass. While the water associated with the measured sulfate ion is
23 used in the calculation of the concentration of organic carbonaceous material, it is not
24 used in estimating ambient concentrations of sulfate or PM light extinction.
25 • Elemental and Organic Carbon: Only the mass of carbon atoms is included in the
26 reported elemental carbon and organic carbon for a CSN (or IMPROVE) sampler. In
27 addition, the assignment of carbon atoms between the reported elemental and organic
28 amounts is dependent on the specifics of the two different thermo-optical analytical
29 methods used in the CSN vs. the IMPROVE network.29 Also, the quartz filter used to
30 quantify carbonaceous material in CSN and IMPROVE sampling both absorbs and loses
31 organic vapors during sampling, while the Teflon filter in a FRM sampler does not
32 absorb organic vapors (although PM on the filter may do so). Therefore, some method
33 other than direct measurement must be used to estimate the total mass concentration of
34 organic carbonaceous material in ambient air. The IMPROVE program adjusts for
35 absorption of vapors by subtracting a monthly average backup filter value, and then
36 applies a standard adjustment factor (1.4 in the original IMPROVE method) to the
37 remaining organic carbon measurement to estimate organic carbonaceous material. In
38 contrast, the standard reports from CSN sampling submitted to AQS do not include these
39 two adjustments, but it is routine for EPA staff to apply adjustments for the same
28 EPA staff recognizes that fine particle nitrate may be in the form of calcium or sodium nitrate, but like the
IMPROVE program treats nitrate as ammonium nitrate.
29 While CSN carbon sampling and analysis methods have recently been harmonized with IMPROVE methods at
many CSN sites, it was not until mid-2007 that the first 57 sites were using the harmonized methods. Consequently,
most of the elemental and organic carbon data used in this assessment were obtained with the original CSN methods.
January 2010 3-20 DRAFT - Do Not Quote or Cite
-------
1 purpose, after reporting of CSN data to AQS. The latter are based on network-wide filter
2 field blanks and are judged as very approximate. For this assessment, the SANDWICH
3 approach to such adjustments (Frank, 2006) is used to estimate the organic mass through
4 a material balance of components measured on the CSN and FRM samplers.
5 • Hourly PMz.s: The continuous instruments used for measuring hourly PM2.5 mass were
6 different among sites (as listed in Appendix A). None of the instrument types that
7 provided hourly data for this assessment, when averaged over 24 hours, exactly matches
8 either the measurement of PM2.5 mass from a FRM sampler or the sum-of-components
9 reportable from CSN sampling. Differences can arise because of differences in water
10 capture and retention, inconsistent absorption and loss of organic vapors and nitric acid
11 vapor, etc. Furthermore, comparability between hourly and 24-hour integrated
12 measurements can only be made on a daily average basis. Consequently, the continuous
13 instruments providing data to this assessment can be assumed to have a range of
14 correlation performance versus the FRM. In light of these consistency issues, the hourly
15 data from the continuous instruments were taken to be most indicative of the relative
16 concentrations of PM25 from hour-to-hour, with less reliance on the absolute accuracy of
17 the continuous instruments.30
18 Taking into consideration the above information, EPA staff combined the four types of
19 available PM2 5 data in each study area using the following steps. Figure 3-5 provides a flow
20 chart to assist in understanding these steps.
21 1. The SANDWICH method (Frank, 2006) was used to subdivide the 24-hour PM2.5 mass
22 reported by the FRM for each day and site into sulfate (including associated ammonium
23 and residual water during filter equilibration and weighing), nitrate (including associated
24 ammonium, but not necessarily enough to fully neutralize the sulfate ion, and residual
25 water during filter weighing), elemental carbon, organic carbonaceous mass, and fine
26 soil/crustal mass. This is done using information from the CSN measurements, physical
27 models, and day-specific temperatures. The primary purpose of this SANDWICH step is
28 to estimate organic carbonaceous mass. Significantly, in the SANDWICH method, the
29 component referred to as organic carbonaceous mass is actually a residual whose value is
30 determined as the difference between the PM2 5 mass determined from weighing the FRM
31 filter and the sum of the estimated masses of the other four mass components as listed
32 above. Therefore, it is not necessary to adjust for organic carbon sampling artifacts or to
33 apply the 1.4 factor commonly used to estimate organic carbonaceous material from
34 IMPROVE measurements of organic carbon. The SANDWICH procedure did not
35 consider sea salt in the material balance, since this is generally a very small mass
36 constituent for the urban areas considered in this analysis. For the same reason, sea salt
37 was also not considered in the aerosol based light extinction algorithm.
38
30 In 2006, EPA developed and promulgated criteria for approval of continuous PM25 samplers as "federal
equivalent methods". These criteria assure a minimum level of correlation between approved continuous
instruments and the FPJVI method, when data from both are expressed as 24-hour average concentrations. However,
in 2005-2007 no commercially available instruments were yet approved under those criteria.
January 2010 3-21 DRAFT - Do Not Quote or Cite
-------
1 Figure 3-5. Sequence of steps used to estimate hourly PMi.s components and PM light
2 extinction
Consistent with FRM
FRM Data:
CMAQ: Diurnal
Profiles of PM25
Components
24-hour PM
Components by SANDWICH
2.5
Steps 2 and 3
Tentative 1-hour PM2.5 Components
and Sum-of-Components
Steps 5 and 6
1-hour PM2.5
Components Adjusted to be
Consistent With measured
1-hour PM2.5
Steps 7 and
i 1-hour Relative
: Humidity Data
Estimates of 1-hour ;
PM
'10-2.5
IMPROVE Light Extinction Algorithm
3
4 2. The CMAQ-derived monthly diurnal profiles for the sulfate, nitrate, elemental carbon,
5 organic carbon and fine soil/crustal components, like the examples for Detroit in Figure
6 3-4, were multiplied by the day-specific SANDWICH-based estimates of the 24-hour
7 average concentrations of these five PM2.5 components, to get day-specific hourly
8 estimates of these five components (including ammonium and water associated with
9 sulfate and nitrate ion).
10 3. The hourly concentrations of these five components (including ammonium and water
11 associated with sulfate and nitrate ion when the filter is weighed) were added together, to
12 get a sum-of-components estimate of hourly PM2.5 mass for the day of the FRM
13 sampling.
14 4. The hourly data from the continuous PM2.5 instrument on the day of the FRM sampling
15 were normalized by their 24-hour average, to get a diurnal profile. (Recall that days were
16 not used in this assessment if hourly PM2.5 mass data were missing for more than 25
17 percent of daylight hours.) This profile was applied to the 24-hour PM2.5 mass reported
18 by the FRM sampler, to get a preliminary, FRM-consistent estimate of hourly PM2.s mass
19 for the day of the FRM sampling. This is straightforward when all 24 values of 1-hour
January 2010
3-22
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
5.
PM2.5 mass were available for the day. However, for some (but not many) days, some
values for continuously measured hourly PM2.5 mass were missing. In such cases, EPA
staff used only the hours with valid 1-hour PM2.5 mass values to develop the diurnal
profile and then applied the profile to the FRM value as just described. This keeps the
average of the valid 1-hour PM2.5 values equal to the 24-hour value from the FRM
sampler.
The two estimates of hourly PM2.5 mass from steps 3 and 4 were compared, hour-by-
hour. By virtue of the way they were derived, the averages of these estimates across all
24 hours of the day will necessarily be the same (and will be equal to the 24-hour FRM
measurement). However, while the diurnal pattern of these two estimates of the same
physical parameter should also be generally similar, it can be expected (and it is
observed) that the hourly measurements from the continuous PM2.5 instruments (after
adjustment to be consistent with the FRM data) have more hour-to-hour variability.
Figure 3-6 gives an example of this comparison, for one day for the Detroit study area.
Figure 3-6. Example from Detroit study area.
.- 20 -
10 -
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
FRM adjusted Hourly PM2.5
Sum of diurnalized SANDWICHED species data using CMAQ profiles
Example comparison from the Detroit study area of hourly PM2.5 mass on March 24, 2006 as
estimated by applying CMAQ-based diurnal profiles to SANDWICH estimates of 24-hour
component concentrations versus applying a diurnal profile derived from continuous PM2.5
measurements to FRM PM2.5 mass.
Given that the continuous instrument is reacting to hour-specific local conditions that can
vary from hour-to-hour due to real variations in local emissions and dispersion/transport
conditions, while the CMAQ-based estimates contain much less specific information, the
diurnal pattern of PM2.5 mass observed by the continuous instrument (adjusted to be
January 2010
3-23
DRAFT - Do Not Quote or Cite
-------
1 consistent with the FRM value for 24-hour average PM^.s) was taken as more reliable.
2 Within each hour, the estimates of all five components from step 2 were increased or
3 decreased by a common percentage (referred to below as A; where the subscript i
4 indicates the hour) so that the sum of the five components after this adjustment was equal
5 to the estimate of the hourly PM2.5 mass from step 4. The adjustment percentage varied
6 from hour-to-hour. Necessarily, in some hours the adjustment is an increase in the
7 concentrations of all components, and in other hours it is a decrease. While this
8 adjustment preserves the consistency between the 24 values of hourly PM2.5 mass and the
9 24-hour FRM mass, it can disturb the consistency between the daily average of hourly
10 estimates of PM2.5 components and the SANDWICH-based estimates of 24-hour average
11 component concentrations. This disturbance was generally small, because the
12 adjustments necessarily go in one direction for some hours and the other direction for
13 other hours. For example, for the particular day in Detroit used for illustration purposes
14 in Figure 6, the effect of this step was to cause a discrepancy of 3 percent between the
15 SANDWICH-based values of 24-hour sulfate concentration and the average of the 24
16 estimates of 1-hour sulfate concentrations (the positive percent indicates a higher
17 concentration in the result of this step than the SANDWICH-based value). The
18 discrepancies were 1, 1,2, and 2 percent for nitrate, elemental carbon, organic carbon,
19 and fine soil/crustal, respectively.
20 7. Each hourly estimate of sulfate concentration from step 6 (which includes estimates of
21 associated ammonium and particle bound water) was adjusted so that it excludes water
22 and reflects full neutralization and therefore is consistent with the reporting practices of
23 the IMPROVE program and the IMPROVE algorithm. This was done via these sub-
24 steps:
25 a. The 24-hour CSN value for the dry mass of sulfate ion (not SANDWICHed, no
26 ammonium or water) was multiplied by 1.375 to reflect an assumption of full
27 neutralization of dry sulfate mass.31
28 b. The ratio of this fully neutralized 24-hour sulfate mass to the SANDWICH-based
29 24-hour sulfate value was calculated.
30 c. This ratio was applied to each individual hour's sulfate concentration from step 6.
31 As in Step 6, it is possible for the 24 final hourly sulfate estimates to no longer be
32 exactly consistent with the 24-hour CSN sulfate measurement, both reported as fully
33 neutralized sulfate ion.
34
35 8. A similar adjustment as in step 7 (for sulfate) was made to each hour's nitrate
36 concentration from step 6, so that the estimate of hourly nitrate would reflect actual
37 atmospheric conditions and be consistent with the IMPROVE algorithm. However, the
38 ratio approach used in step 7(b) for sulfate could not be applied for nitrate, so this
39 adjustment had to be more complicated. Because in warm weather the FRM Teflon filter
31 While it would have been possible to develop a more realistic estimate of partially neutralized sulfate, the
assumption of full neutralization was used to maintain consistency with the basis for the f(RH) term in the
IMPROVE algorithm.
January 2010 3-24 DRAFT - Do Not Quote or Cite
-------
1 does not retain nitrate, the initial FRM-consistent nitrate estimate derived by applying the
2 SANDWICH method to the FRM and CSN data can be zero. Such a zero value makes it
3 impossible to use the ratio approach in 7(b). Instead, the adjustment was made as
4 follows:
5 a. The 24-hour CSN value for nitrate ion (not SANDWICHed, no ammonium or
6 water) was multiplied by 1.29 to reflect an assumption of full neutralization by
7 ammonia.
8 b. This 24-hour value was then diurnalized using the CMAQ-based profile, similar
9 to step 2.
10 c. Each resulting hourly value of nitrate was further multiplied by the Ai factor from
11 step 6.
12 d. This new estimate of hourly nitrate was used to replace the initial nitrate value
13 that had resulted from step 6.
14 For cooler areas and days in which the 24-hour SANDWICH results include some nitrate,
15 the effect of these steps for nitrate are exactly the same as the effects of step 7 for sulfate
16 (except for the 1.29 vs. 1.375 neutralization factor). For warmer areas and days in which
17 the 24-hour SANDWICH results did not include any nitrate even though nitrate was
18 measured on the CSN Nylon filter, the effect of these steps is to assign the CSN nitrate to
19 each hour using a combination of the information in the CMAQ-based profiles and the
20 information provided by the continuous PM2.5 sampler. As in Step 6, it is possible for the
21 24 final hourly nitrate estimates to no longer be exactly consistent with the 24-hour CSN
22 nitrate measurement.
23 The net effect of these steps is believed by EPA staff to result in hourly PM light
24 extinction estimates with the following features with respect to some of the complicating aspects
25 of PM sampling:
26 • The 24-hour average of the hourly nitrate concentrations used to estimate hourly PM light
27 extinction agrees closely but not exactly with the 24-hour value provided by the CSN
28 sampling, and generally is higher than the contribution of nitrate to the FRM measure of
29 PM2.5 mass. In some mid-day hours in some areas, estimated hourly nitrate is zero which
30 is a more realistic approach than applying a 24-hour species mix to each hour.
31 • The 24-hour average of the hourly organic carbonaceous material concentrations used to
32 estimate hourly PM light extinction achieves FRM mass balance closure, taking into
33 account also the difference in nitrate and the possibly partial neutralization of sulfate ion
34 on the FRM filter. Because the Teflon filter used in FRM sampling is less subject to
35 positive artifacts for organic material, this approach sidesteps an area of uncertainty in the
36 IMPROVE sampling method. By relying on mass closure as the driving principle for
January 2010 3-25 DRAFT - Do Not Quote or Cite
-------
1 estimating organic material, it is not necessary to choose a multiplier to relate organic
2 carbon to organic carbonaceous material.32
3 • The 24-hour average of the hourly elemental carbon concentrations used to estimate
4 hourly PM light extinction agrees closely but not exactly with the 24-hour value provided
5 by the CSN sampling, and with the contribution of elemental carbon to the FRM measure
6 of PM2.5 mass. Elemental carbon is generally defined by the thermal optical transmission
7 method used in CSN, rather than the thermal optical reflectance method used in
8 IMPROVE.
9 3.3.2 Hourly PMio-2.5 Concentrations
10 Three different paths were used to estimate hourly PMio-2.5 concentrations depending on
11 data availability, in the following order of preference:
12
13 1. When hourly data from a collocated PMio instruments were available at the continuous
14 PM2.5 site in a study area, PM2.5 was subtracted hour-by-hour from PMio. Negative
15 values were reset to zero. This was the approach most often used in Birmingham,
16 Detroit, Baltimore, and Philadelphia. This method should result in reliable estimates of
17 actual PMio-2.5 at the study site. (How well the study site represents the study area
18 generally, or the most visibility-impacted portions, of the study area is a separate issue.)
19
20 2. When collocated continuous PMio data were not available at the continuous PM2.5 site in
21 a study area, but continuous PMio data were available at another site in or near the same
22 study area, PMio-2.5 was estimated by subtraction, implicitly assuming that the latter site
23 was also representative of PMio at the former site. This was the approach most often
24 used in Los Angeles, Phoenix, St. Louis, Atlanta, and New York. As a result, estimates
25 of PMio-2.5 for these areas could be affected by site-to-site differences. In particular, the
26 two sites in Los Angeles were a good distance apart, and the PMio site in Victorville may
27 represent influences from agricultural operations rather than typical urban influences. In
28 St. Louis, the PMio site may also have been influenced by particular local sources. In
29 both cases, very high estimates of hourly PMio-2.5 may not represent reality at the PM2.s
30 site, although they may be reasonable estimates for the PMio site.
31
32 3. If neither of the first two methods was possible, a regional average ratio of PMio-2.5 to
33 PM2.s determined from an analysis of 24-hour data for the 2005 Staff Paper was applied
34 to hourly PM2.5 from the continuous instrument associated with the study area. This was
35 the approach used for all hours in Tacoma, Fresno, Salt Lake City, Dallas, Houston, and
36 Pittsburgh. With this approach, it is not possible for there to be any particularly high
37 estimates of hourly PMio-2.5-
38
32 In other work, EPA staff has observed that when applied to urban sampling data together with CSN network-wide
field blanks applied to reported OC measured concentrations, the multipliers that can be back-calculated from the
results of the SANDWICH method tend to be nearer to 1.4 than to the higher value used in the new IMPROVE
algorithm.
January 2010 3-26 DRAFT - Do Not Quote or Cite
-------
1 The estimation of PMio-2.5 was further complicated because some types of data were
2 missing for isolated hours in the 2005-2007 period. As result, even for a single study area more
3 than one method sometimes had to be used to estimate hourly PMio-2.5. Appendix A gives more
4 specifics about the estimation of hourly PMio-2.5 in each study area.
5 The three-path approach described here is similar to that used for the visibility analysis
6 reported in the 2005 Staff Paper. While the second and third paths involve the use of data and
7 assumptions that are not robust compared to the use of paired, collocated, same-method
8 continuous instruments or compared to the use of paired low-volume filter-based samplers, in
9 most areas and periods the contribution to PM light extinction from the resulting PMio-2.5
10 concentrations was not large compared to the PM light extinction due to PM2.5 components.
11 3.3.3 Hourly Relative Humidity Data
12 Hourly relative humidity (RH) data for each study area's primary monitoring site were
13 obtained hour-by-hour from the closest available non-missing relative humidity measurement, as
14 reported by either an air monitoring station reporting such data to AQS or a National Weather
15 Service (NWS) station. For the AQS RH data, parameter 62201 values were utilized. RH data
16 from both sources are expressed as percentages.33
17 3.3.4 Calculation of Daylight 1-Hour PM Light Extinction
18 Because the interest in this analysis is on visibility during daylight hours, EPA staff
19 applied a scheme to denote those hours that would be considered daylight hours. For simplicity,
20 all the days within each "season" in all study areas were considered to have the same daylight
21 hours.34 Table 3-5 shows the dividing times used to denote daylight hours for the study areas.
22 Unless otherwise stated, all subsequent discussion of the results refers only to the values of
23 parameters during these daylight hours.
24 The original IMPROVE algorithm was applied hour-by-hour to estimate PM light
25 extinction in each study area for each daylight hour. When doing so, we capped the value of the
26 humidity adjustment factor in the IMPROVE algorithm ("f(RH)") at the value of 7.4 that it has
27 for a relative humidity of 95 percent. The effect of measurement errors in relative humidity at
28 values above 95 percent on the value of f(RH) and thus on reconstructed PM light extinction is
33 After release of the first public review draft of this assessment, and error was discovered in the data processing
that assembled the relative humidity data base, such that the nearest site was not used as described here. That error
has been corrected in this second public review draft. See also "Corrections to Relative Humidity Values Used in
the Draft UFVA, Corrected Graphics, Tables, and Availability of Detailed Data File for Current Conditions"., P.
Lorang, November 10, 2009.
34 This simple approach does not account for the effects of the actual date within a three-month season, latitude, or
east-west position within a time zone on the actual local hours that are entirely daylight. Appendix I examines the
possible impact of this simplification, concluding that it is unlikely to affect later answers to policy relevant
questions.
January 2010 3-27 DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
considerable because of the highly nonlinear form of the function in that range. This creates
uncertainty as to the representativeness of the extinction values calculated with high values of
relative humidity.
35
Table 3-5 Assumed daylight hours by season (Local Standard Time)
First hour that is
entirely daylight
Last hour that is
entirely daylight
Number of daylight
hours
November-January
8:00-9:00 AM
3:00-4:00 PM
8
February-April
7:00-8:00 AM
5:00-6:00 PM
11
May July
5:00-6:00 AM
6:00-7:00 PM
14
August-October
6:00-7:00 AM
5:00-6:00 PM
12
7 3.3.5 Exclusion of Hours with Relative Humidity Greater than 90 Percent from PM
8 Light Extinction NAAQS Scenarios and Most Results
9
10 As advised by CAS AC as part of its comments on the first public review draft of this
11 assessment, EPA staff considered whether to structure the PM light extinction NAAQS scenarios
12 so that ambient data obtained during daylight hours in which relative humidity was greater than
13 90 percent would play no role in the indicator/form of the NAAQS, i.e., so that those data would
14 not enter into the calculation of the design value. EPA staff obtained hourly meteorological
15 parameters from National Weather Service monitoring sites near 11 of the 15 study sites (usually
16 a major airport), for 2005 through 2007, for all days in this period including days for which PM
17 observations to support estimate of PM light extinction are not available36. For these sites, we
18 compared the occurrence of liquid precipitation, hail, other frozen precipitation, fog, and
19 haze/mist during daylight hours with humidity greater than 90 percent and during all other
20 daylight hours. These five conditions are generally considered natural causes of reduced
21 visibility. Table 3-6 presents this comparison. The percentages of hours with each of these five
22 conditions individually and for any one or more of the five conditions together are shown for the
23 two sets of daylight hours. NWS observations of these conditions are instantaneous, and are
24 generally made about 50 minutes after the hour. The relative humidity observations are made at
25 the same time. It should be noted that this analysis of the co-occurrence of high relative
26 humidity and these five conditions uses data from NWS sites other than the AQS sites that
35 The IMPROVE program also caps the value of f(RH) at its value for a relative humidity of 95% when reporting
visibility in deciviews.
36 Through an oversight, EPA staff did not obtain NWS data for Los Angeles, St. Louis, Houston, and Detroit in
time for processing and incorporation of results into Table 3-6. These data will be added in the final version of this
assessment.
January 2010
3-28
DRAFT - Do Not Quote or Cite
-------
1 provided the relative humidity value for the light extinction estimate. AQS sites could not be
2 used for this analysis because they generally do not report similar weather condition data.
3 The comparison for the 11 sites shows that in the set of hours with relative humidity
4 above 90 percent, the frequencies of liquid precipitation (rain), fog and haze/mist individually
5 and the frequency of any one or more of them together were considerably higher than in the set
6 of hours with lower relative humidity.3? The frequencies of hail and other frozen precipitation
7 were too low for meaningful comparisons. Moreover, except in Tacoma, the frequency of rain or
8 fog at the observation moments during the hours with relative humidity less than or equal to 90
9 percent was less than 6 percent. Also, a separate analysis (not shown) indicated that rainy hours
10 with lower relative humidity experience considerably less accumulation than rainy hours with
11 higher relative humidity. Based on this assessment, the 90% relative humidity cutoff criteria is
12 effective in that on average less than 6% of the hours are removed from consideration, yet those
13 hours have on average over twelve times the likelihood of weather conditions that directly reduce
14 visibility compared to hours with 90% or less relative humidity.
15 Rain, fog, and mist cause a natural reduction in visibility, independent of PM
16 concentrations. To reduce the likelihood of a secondary PM NAAQS based on an indicator/form
17 that could be affected by measurements made under natural weather conditions that reduce
18 visibility, for this assessment EPA staff eliminated the estimates of PM light extinction from any
TO
19 daylight hours with relative humidity above 90 percent from design value calculations. Also,
20 because PM light extinction during such hours is not as likely to be the primary cause of adverse
21 effects on the public, all figures and tables in the body of this document and in Appendices that
22 present PM light extinction values or statistics exclude values for such hours (unless explicitly
23 stated to include them), so that the patterns of PM light extinction during the remaining daylight
24 hours can be seen clearly. Figures and tables that present PM component concentrations and
25 relative humidity values are based on all daylight hours, however.
26 More information on this topic can be found in Appendix G, which reports by study area
27 the percentages of daylight hours that were excluded from design values, the distribution of the
28 excluded hours by time of day, and the percentage of days that had one or more daylight hours
37 The "haze/mist" category is not an original NWS reporting category. It is a combination of three original NWS
weather categories: mist, smoke, and haze that were prepared earlier by EPA staff for another purpose. EPA staff
was unable to separate the occurrence of these three conditions in time for this version of this assessment. Of these,
only mist, defined as fog-like conditions that do not impair visibility below 0.5 nautical miles, is clearly a natural
condition during which people would not consider limited visibility to be aesthetically undesirable. Consequently,
the columns of Table 3-6 for "haze/mist" and "any" must be interpreted accordingly.
38 Another consideration is that instruments used to measure light extinction could be adversely affected if allowed
to operate without heating or other protective method (such as diffusion drying of incoming air) when relative
humidity is very high. If protected, however, the measured light scattering would not reflect actual ambient
conditions.
January 2010 3-29 DRAFT - Do Not Quote or Cite
-------
1 eliminated. Appendix G also contains box plots which contrast the distributions of daylight 1-
2 hour PM light extinction values (and maximum daily daylight 1-hour PM light extinction, see
3 section 3.3.6) before and after this elimination step. The tile plots in Figure 3-12 also present
4 additional detailed information on the specific hours that had relative humidity values above 90
5 percent, and on the PM light extinction values during those and other daylight hours.
January 2010 3-30 DRAFT - Do Not Quote or Cite
-------
1 Table 3-6 Comparison of Meteorological Parameters for Daylight Hours with Relative Humidity Greater than 90
2 Percent and Other Daylight Hours, During 2005 -2007
Study Area
Tacoma
Fresno
Los
Angeles
Phoenix
Salt Lake
City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Daylight Hours with Relative Humidity <= 90%
Number
of
Hours
18293
24245
26045
24989
25519
23826
23696
22254
22867
24302
24963
Percentage of Hours with Weather Condition
Liquid
Precip.
12
3
1
4
3
4
5
5
4
6
6
Hail
0
0
0
0
0
0
0
0
0
0
0
Other
Frozen
Precip.
0
0
0
2
0
0
0
7
1
0
1
Fog
0
2
0
1
1
1
1
1
2
1
1
Haze/Mist
3
16
0
4
4
8
7
8
9
6
9
Any
13
18
1
8
6
11
10
17
12
11
13
Daylight Hours with Relative Humidity > 90%
Number
of
Hours
7987
1615
235
1291
761
2454
2584
4026
3413
1978
1317
Percentage of Hours with Weather Condition
Liquid
Precip.
24
12
50
21
47
30
39
36
36
44
52
Hail
0
0
0
0
0
0
0
0
0
0
0
Other
Frozen
Precip.
1
0
0
30
1
0
0
9
3
4
8
Fog
7
44
11
33
18
32
34
26
30
26
41
Haze/Mist
26
60
27
57
71
55
61
54
64
64
79
Any
45
79
56
69
80
61
71
69
74
80
89
January 2010
3-31
DRAFT - Do Not Quote or Cite
-------
1 3.3.6 Calculation of Daily Maximum 1-Hour PM Light Extinction
2 Daily maximum 1-hour PM light extinction is a statistic of interest in this assessment, as
3 briefly discussed in section 1.4.3. The daylight hour with the maximum value of PM light
4 extinction and the corresponding PM light extinction value were identified for each day for each
5 study area. As mentioned in section 3.2. 1, days which were missing 1-hour PM2.5 values for
6 more than 25 percent of daylight hours were not used in this analysis. No further completeness
7 requirement for 1-hour data during a day was applied when selecting the daylight hour with the
8 maximum value of PM light extinction.
9 3.4 SUMMARY OF RESULTS FOR CURRENT CONDITIONS
10 3.4.1 Levels of Estimated PM25, PM^s Components, PMio-2.5? and Relative Humidity
1 1 Figure 3-7 presents box-and-whisker plots to illustrate the distributions in each study area
12 of the estimates of 1-hour PM2.5 (the diurnalized FRM value, resulting from step 4 in section
13 3.4.1), PMio-2.5, and relative humidity over the entire 2005-2007 study period. In the plot for
14 each parameter, areas are ordered by longitude, to make it easier to see east-versus-west regional
15 differences. For these three parameters, the distributions are given for all the daylight 1-hour
16 estimates, including hours with relative humidity greater than 90 percent. Similar plots of the
17 daily maximum daylight 1-hour values of PM2.s and PMio-2.5 concentrations and relative
18 humidity are available in Appendix B, as are plots of all daylight 1-hour values for each of the
19 PM2.5 component species.39
20 From these plots we see that the distributions of PM2.5 generally trend toward higher
21 concentrations from west to east except for the two California urban locations which have PM2.5
22 concentrations more typical of eastern areas. The lowest median PM2.s concentrations are in
23 Tacoma, WA, and Phoenix, AZ. Median PMio-2.5 concentrations are highest in St. Louis, MO,
24 and Phoenix, AZ, and lower elsewhere. The highest outlier PMio-2.5 concentrations are in St.
25 Louis, MO, and Los Angeles, CA. Relative humidity is lowest for the western urban areas
26 except for Tacoma, WA, which is similar to the northeastern urban locations with respect to
27 humidity. These hourly daylight PM concentration and relative humidity box and whisker plots
28 are consistent with our expectations based on regional 24-hour PM concentration values and
29 humidity climatology.
30
39 In all box-and-whisker plots in this document, the box represents the 25th to 75th percentile range and the whiskers
represent the 10th and 90th percentile points of the data; individual data points below the 10th percentile and above
the 90th percentile are graphed as small circles (which may not all be visible because they may lie on top of one
another as is the case for relative humidity in Figure 3-7(c) because relative humidity is reported on as an interger.
January 2010 3-32 DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
Figure 3-7. Distribution of PM parameters and relative humidity across the 2005-2007
period, by study area
(a) Estimates of 1-hour PMi.5 mass, based on applying continuous instrument-based
diurnal profiles to 24-hour FRM PMi.s mass
PM 2.5 hourly (Daylight Hours)
6
7
9
10
8
^
/ ^
*"*
y ^
&
^' s
-------
1
2
3
4
Figure 3-7. (cont). Distribution of PM parameters and relative humidity across the 2005-
2007 period, by study area
(c) 1-hour relative humidity
5
6
s -
Relative Humidity hourly (Daylight Hours)
839 39JS 34
-------
1 3.4.2 Levels of Estimated PM light extinction
2 Figure 3-8 presents box-and-whisker plots to illustrate the distributions of the estimates
3 of daylight 1-hour reconstructed PM light extinction levels in each area in each year (excluding
4 hours with relative humidity greater than 90 percent). The distribution of (a) the daily maximum
5 1-hour values and (b) the individual 1-hour values are both shown. The horizontal dashed lines
6 in the plots represent the low, middle, and high candidate protection levels (CPLs) for PM light
7 extinction as discussed in section 2.6. These benchmarks for PM light extinction are 64, 112,
8 and 191 Mm"1, corresponding to the benchmark VAQ values of 20 dv, 25 dv and 30 dv. Table 3-
9 7 provides (a) the percentages of days (across all of 2005-2007, unweighted) in which the daily
10 maximum daylight 1-hour PM light extinction level was greater than each of the three candidate
11 protection levels (excluding hours with relative humidity greater than 90 percent), and (b) the
12 similar percentage based on all daylight hours (with the same exclusion).
13 As was also seen in the comparable PM2.5 concentration box and whisker plots in Figure
14 3-7, the high percentile hourly PM light extinction values in Figure 3-8 tend to be higher in the
15 eastern urban areas and lower in the non-California western urban areas. The distributions of
16 maximum daily PM light extinction values are higher (Figure 3-8b), as expected, than for all
17 hours (Figure 3-8a). Both Figure 3-8 and Table 3-7 indicate that all 15 urban areas have daily
18 maximum hourly PM light extinctions that exceed even the highest of the CPLs some of the
19 time. Again, the non-California western urban locations have the lowest frequency of maximum
20 hourly PM light extinction with values in excess of the high CPL for 8 percent or fewer of the
21 days. Except for the two Texas and the non-California western urban areas, all of the other
22 urban areas exceed that high CPL from about 20 percent to over 60 percent of the days. Based
23 on these estimated maximum hourly PM light extinction estimates, all 15 of the urban areas
24 exceed the low CPL for about 40 percent to over 90 percent of the days. As noted in section
25 3.2.1, in 10 of the 15 study areas the study site used in this assessment is not the site in the study
26 area with the highest concentrations of PM2.5. Thus, these estimates may not characterize
27 visibility in the worst-visibility portion of each study area.
28 In the last review of the secondary PM NAAQS, the pattern of light extinction during the
29 day was of particular interest. To illustrate the distributions of 1-hour PM light extinction levels
30 in specific daylight hours, Figure 3-9 shows the distributions of 1-hour PM light extinction
31 across the entire three-year study period, individually for the study areas (excluding hours with
32 relative humidity greater than 90 percent). (Appendix E provides additional graphics related to
33 temporal/spatial patterns of light extinction.) These plots show that high PM light extinction can
34 occur during any of the daylight hours, though for most of these urban areas the morning hours
January 2010 3-35 DRAFT - Do Not Quote or Cite
-------
1 have somewhat higher PM light extinction than in the afternoon.40 Urban areas without a
2 preference for morning high PM light extinction include Phoenix, AZ; Salt Lake City, UT;
3 Tacoma, WA; Fresno, CA; and Philadelphia, PA.
4
40 If hours with relative humidity greater than 90 percent were not eliminated, the tendency for higher PM light
extinction in the morning hours would be stronger.
January 2010 3-36 DRAFT - Do Not Quote or Cite
-------
1
2
3
4
Figure 3-8. Distributions of estimated daylight 1-hour PM light extinction and maximum
daily daylight 1-hour PM light extinction across the 2005-2007 period, by study area
(excluding hours with relative humidity greater than 90 percent).
(a) Maximum daily values
6
7
8
9
10
11
12
13
14
Daily Maximum Extinction (Daylight Hours)
'09 324 3CS 86 306 273 148 IK 349 279 '41 577 18' '43 2E5
-4~-4-
^
„
//
January 2010
3-37
DRAFT - Do Not Quote or Cite
-------
(b) Individual 1-hour values
2
3
4
5
6
Hourly Extinction (Daylight Hours)
S3S7 3019 I4« 3883 3759 2471 1533
\*
January 2010
3-38
DRAFT - Do Not Quote or Cite
-------
2 Table 3-7 Percentage of daily maximum
3 of daylight PM light extinction exceeding CPLs
4 greater than 90 percent).
hourly values and individual hourly values
(excluding hours with relative humidity
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Average
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Average
Number of Days with
Estimates
109
324
300
86
306
273
148
289
349
279
141
277
181
143
225
229
Number of Daylight Hours
with Estimates
1087
3533
3048
988
3366
3043
1504
3096
3763
2507
1547
2842
1873
1468
2296
2397
Candidate Protection Level
64MH11
112 Mm -1
191 Mm1
(a) Percentage of Daily Maximum Hourly Values
Exceeding CPL
52
75
90
42
44
80
79
98
89
91
87
85
80
86
83
77
22
52
83
7
17
41
45
78
65
75
68
57
50
64
59
52
4
30
62
1
8
10
11
40
34
31
43
26
23
31
28
26
(b) Percentage of Individual Daylight Hours
Exceeding CPL
14
41
68
11
17
33
35
66
57
60
62
53
55
55
53
45
4
21
43
1
7
10
8
36
25
28
36
25
24
28
28
22
1
10
20
0
3
2
1
11
8
5
14
7
7
9
9
7
January 2010
3-39
DRAFT - Do Not Quote or Cite
-------
1 Figure 3-9. Distributions of 1-hour PM light extinction levels by daylight hour across the 2005-2007 period, by study area
2 (excluding hours with relative humidity greater than 90 percent).
4
5
1000 -
600 -
400 -
200 -
0 -
-
-
-
1000 -
600 -
400 -
200 -
o -
Phoenix, AZ
Fresno, CA
1 « '
• * ' "
Atlanta, GA
..
Pittsburgh, PA
~ • " " '
Houston, TX
Baltimore, MD
• , '. •
Salt Lake City, UT
. , * f ," " ' f,
Los Angeles, CA
"" -
;«••..
Birmingham, AL
St. Louis, IL
New York, NY
, n .
Dallas, TX
' * ,
Tacoma, WA
Philadelphia, PA
Detroit, Ml
-
-
-
-
-
- 1000
- 600
- 400
- 200
- 0
-
-
-
-
-
05 0807 08 09 10 11 12 13 14 15 16 17 18 05 0607 08 09 10 11 12 13 14 15 16 17 18 05 060708 09 10 11 12 13 14 15 16 17 18 05 0607 0809 10 11 12 13 14 15 16 17 18 05 06 07 0809 10 11 12 13 14 15 16 17 18
January 2010
3-40 DRAFT - Do Not Quote or Cite
-------
1 3.4.3 Patterns of Relative Humidity and Relationship between Relative Humidity and
2 PM light extinction
3 Figure 3-10 shows the distribution of relative humidity values at each daylight hour, for
4 each study area across 2005-2007 (excluding hours with relative humidity greater than 90
5 percent).41 As expected, in every area relative humidity is lowest in the early afternoon,
6 typically the warmest part of the day. Relative humidity is most similar across areas in the early
7 afternoon, as observed in the 2005 Staff Paper. However, even in this period there are notable
8 differences among areas. This variation was not as evident in the information presented in the
9 2005 Staff Paper because only regionally averaged information was presented. In all areas, there
10 is considerable variation in hour-specific relative humidity during the three-year period.
11 To allow closer inspection of the relationship between PM light extinction values and
12 relative humidity values, Figure 3-11 is a scatter plot of actual 1-hour relative humidity and 1-
13 hour reconstructed PM light extinction (excluding hours with relative humidity greater than 90
14 percent). Horizontal lines are included in each of the individual plots corresponding to the three
15 benchmarks for PM light extinction and a vertical line in each for the 90 percent relative
16 humidity cutoff. There are many instances with PM light extinction greater than the candidate
17 protection levels when relative humidity is 90 percent or lower. Notice that in Figure 3-11 there
18 also are plenty of high humidity conditions for each urban area that correspond to low PM light
19 extinction values. This is because humid air does not by itself contribute to light extinction.
20 Particles composed of material that absorbs water in high relative humidity conditions (e.g.,
21 sulfate and nitrate PM) swell to larger solution droplets that scatter more light than their smaller
22 dry particle counterparts in a less humid environment. The magnitude of the relative humidity
23 effect on light extinction depends directly on the concentration of these hygroscopic PM
24 components. (Figure 3-11 reveals skips in reported relative humidity values for some but not all
25 the study areas. This is a result of calculations of relative humidity from dry and wet bulb
26 temperatures reported to the nearest whole Celsius degree.)
41 Similar information on diurnal patterns but broken out by season is given in Appendix E.
January 2010 3-41 DRAFT - Do Not Quote or Cite
-------
1 Figure 3-10. Distributions of 1-hour relative humidity levels by daylight hour across the 2005-2007 period, by study area
2 (excluding hours with relative humidity greater than 90 percent).
3
4
5
40
20
Phoenix. AZ
Fresno
o " o o
Atlanta
CA
GA
Pittsburgh, PA
Houston, TX
Baltimore. MD
Salt Lake City. UT
Los Angeles
Birmingham
CA
AL
St. Louis. IL
New York. NY
Dallas
TX
Tacoma. WA
Philadelphia, PA
Detroit. Ml
- 60
- 40
- 20
- 0
05 06 07 08 09 10 11 12 13 14 15 16 17 18 05 06 0708 09 10 11 12 13 14 15 16 17 18 05 06 07 0809 10 11 12 13 14 15 16 17 18 0506 07 08 09 10 11 12 13 14 15 16 17 18 05 06 07 08 09 10 11 12 13 14 15 16 17 18
January 2010
3 -42 DRAFT - Do Not Quote or Cite
-------
1 Figure 3-11. Scatter plot of daylight 1-hour relative humidity (percent) vs. reconstructed PM light extinction (Mm"1) across
2 the 2005-2007 period, by study area (excluding hours with relative humidity greater than 90 percent).
3
4
5
1000
800
600
400
200
0
1000
800
600
400
200
o
Phoenix, AZ
Fresno. CA
Atlanta, GA
0 20 40 60 80
I | | ] I
Pittsburgh, PA
Houston, TX
Baltimore, MD
0 20 40 60 80
I | 1 I L_
Salt Lake City, UT
=-:,
Los Angeles. CA
Birmingham, AL
St. Louis, IL
New York, NY
Dallas, TX
I T
0 20
60 80
0 20 40 60 8C
Percent Relative Humidity
Tacoma, WA
i
Philadelphia, PA
Detroit, Ml
1000
- 800
-600
- 400
- 200
-o
0 20 40 60
January 2010
3-43
DRAFT - Do Not Quote or Cite
-------
1 3.4.4 Tile Plots of Hourly PM Light Extinction
2 Figure 3-12 consists of "tile plots" that show the estimated levels of 1-hour PM light
3 extinction for each daylight hour for each study area. These plots assist in understanding the
4 times of the year and hours of the day in which high relative humidity and high PM light
5 extinction occur, both separately and together.
6 Time runs horizontally with each row of tiles representing a single day from midnight
7 (left site) to midnight (right side), and vertically from January (top) to December (bottom). Each
8 tile represents one hour of the year for which data to estimate PM light extinction were
9 sufficient. Sites with 1:3 speciation sampling have more (and smaller) tiles than sites with 1:6
10 speciation sampling. The tick marks on the vertical axis identify the first available sample day of
11 each month identified by its month number.
12 PM light extinction is presented in terms of four ranges or bins defined by the two
13 intervals between the three CPLs, a bin above the high CPL, and a bin below the low CPL. For
14 the hours with relative humidity of 90 percent and below (referred to as "Low RH bext" in the
15 figure legend), shades of green are used to indicate the CPL range. Contrasting blue color scales
16 are used for the tiles representing hours with relative humidity greater than 90 percent (referred
17 to as "High RH bext" in the shading legend), so that the hours excluded from the PM NAAQS
18 scenarios (see section 3.3.5 and Chapter 4) can be distinguished. Hours with missing PM2.s data
19 from the continuous instrument have no estimates of PM light extinction and are white. Such
20 cases are rare, following the prior complete exclusion of days in which more than 25 percent of
21 daylight hours were missing such data.
22 Note that for Tacoma and Phoenix there are plots for only two years because the third
23 year did not have suitable data, and for Phoenix and Houston only 9 months are shown for one of
24 the available years because suitable data were not available for the remaining quarter (the
25 available 9 months of results are stretched over the same vertical distance as the 12 months in the
26 other cases).
27 One observation that can be made in looking at these tile plots is that in very many cases,
28 days which have one or more hours with high PM light extinction excluded because of high
29 relative humidity have other hours with high PM light extinction which are not excluded.
30 Although none of the PM light extinction NAAQS scenarios considered in Chapter 4 are
31 based on a averaging period longer than one hour, these tile plots can be used to get a rough
32 sense of whether hours with high PM light extinction tend to be isolated, such than average
33 values over several hours would be considerably lower, or tend to occur together, such that a
34 longer averaging period would produce roughly the same design value. A number of the eastern
35 urban areas have numerous day-long haze episodes throughout the year (e.g. St. Louis, Detroit,
36 Pittsburgh, Philadelphia and New York) or seasonally (e.g. Fresno and Salt Lake City, in the
January 2010 3-44 DRAFT - Do Not Quote or Cite
-------
1 winter, and Los Angeles and Atlanta in the summer). Some of the urban areas have morning
2 haze levels that diminish later in the day on a year-around basis (e.g. Dallas) or seasonally (e.g.
3 Los Angeles, Birmingham and Atlanta in winter and Tacoma, Fresno, and St. Louis in the
4 summer). This type of information may be useful in this regard during the subsequent
5 preparation of the Policy Assessment Document.
6
January 2010 3-45 DRAFT - Do Not Quote or Cite
-------
Figure 3-12 Tile Plots of Hourly PM Light Extinction
Tacoma, WA
2005
2006
2007
2
3
4 -
5 -
6-
7
8 -
9 -
10-
11 -
12 -
12
Hour
18 24
1 -
5 -
6 -
7 -
8 -
9 -
10 -
11 -
12 -
f
Low RH bext
Inf
190
111
- 64
0
High RH bext
Inf
190
111
64
0
January 2010
3-46
DRAFT - Do Not Quote or Cite
-------
Figure 3-12. Tile Plots of Hourly PM Light Extinction, continued
Fresno, CA
2005
2006
2007
1 -i
2 -
3-
6 -
7 -
8
10 -
11 -
12 -
2
1 -,
4 -
9 -
10-
11 -
12 -
12
Hour
18
I
24
1 -
2 -
3 -
4 -
5-
6-
7 -
8-
Low RH bext
n
Inf
190
111
64
High RH bext
Inf
190
111
64
January 2010
3-47
DRAFT - Do Not Quote or Cite
-------
Figure 3-12. Tile Plots of Hourly PM Light Extinction, continued
Los Angeles, CA
2005
2006
2007
6 -
7 -
8
9 -
10 -
11 -
12 -
2
5-
9-
10 -
11 -
12 -
12
Hour
18
I
24
1 -
2 -
3 -
4 -
5-
6 -
7 -
8 -
9 -
10 -
11 -
12 -
Low RH bext
n
Inf
190
111
64
High RH bext
Inf
190
111
64
January 2010
3-48
DRAFT - Do Not Quote or Cite
-------
Figure 3-12. Tile Plots of Hourly PM Light Extinction, continued
Phoenix, AZ
2005
2006
2007
4 -
5 -
6 -
10 -
11 -
12 -
^B •
i^^^l
1 I
r,
•i
i -
2 -
3 -
4-
5 -
6 -
7 -
8-
9 -
10-
11 -
12 -
fr.
^^l
*"
•
• •
* "
•
•
2
"1I
7 12 18 24
Hour
12
Hour
18
Low RH bext
n
Inf
190
111
64
High RH bext
January 2010
3-49
DRAFT - Do Not Quote or Cite
-------
Figure 3-12. Tile Plots of Hourly PM Light Extinction, continued
Salt Lake City, UT
2005
2006
2007
1 -i
5-
6 -
7 -
10 -
11 -
12 -
2
12
Hour
18
!
24
1
2
3-
4 -
5
6
7-
8 -
10 -
11 -
12 -
1
12
Hour
18
I
24
1
2
3-
4 -
5
6
8
10 -
11 -
12 -
I \
12
Hour
18 24
Low RH bext
n
Inf
190
111
64
High RH bext
Inf
190
111
64
January 2010
3-50
DRAFT - Do Not Quote or Cite
-------
Figure 3-12. Tile Plots of Hourly PM Light Extinction, continued
Dallas, TX
2005
2006
2007
1 -,
7 -
10 -
11 -
12 -
2
12
Hour
18
!
24
7 -
8 -
9 -
10 -
11 -
12 -
12
Hour
18
I
24
1 -,
2 -
3 -
4 -
5 -
6 -
7 -
8-
9 -
10 -
11 -
12 -
1
12 18
Hour
24
Low RH bext
n
Inf
190
111
64
High RH bext
Inf
190
111
64
January 2010
3-51
DRAFT - Do Not Quote or Cite
-------
Figure 3-12. Tile Plots of Hourly PM Light Extinction, continued
Houston, TX
2005
2006
2007
1 -
2 -
3 -
4 -
5-
7 -
10 -
11 -
12 -
r
1 -i
2-
3 -
4-
5-
7-
9-
10-
11 -
12-
1 -i
2 -
3 -
4 -
5 -
6 -
r-
Low RH bext
n
Inf
190
111
64
High RH bext
Inf
190
111
64
1 7 12 18 24
Hour
7 12 18 24
Hour
12
Hour
18 24
January 2010
3-52
DRAFT - Do Not Quote or Cite
-------
Figure 3-12. Tile Plots of Hourly PM Light Extinction, continued
St. Louis, IL
2005
2006
2007
1 -
2-
3 -
4 -
5-
6 -
7 -
8 -
9-
10 -
11 -
12 -
1 -i
2
7 12 18
Hour
24
4
5-
6-
7 -
8-
9 -
10-
11 -
12 -
12
Hour
18
I
24
1 -i
2 -
3 -
4 -
5-
6-
7 -
8 -
9 -
10 -
11 -
12 -
Low RH bext
n
Inf
190
111
64
High RH bext
Inf
190
111
64
January 2010
3-53
DRAFT - Do Not Quote or Cite
-------
Figure 3-12. Tile Plots of Hourly PM Light Extinction, continued
Birmingham, AL
2005
2006
2007
1 -
2 -
3-
4 -
5 -
6 -
7 -
8 -
9 -
10 -
11 -
12 -
2
—i—
12
Hour
18 24
1 -i
2 -
3-
4-
5 -
6 -
7 -
8 -
9-
10 -
11 -
12 -
12
Hour
18
I
24
1 -
2 -
3 -
4 -
8-
9 -
10 -
11 -
12 -
12
Hour
18
24
Low RH bext
n
Inf
190
111
64
High RH bext
Inf
190
111
64
January 2010
3-54
DRAFT - Do Not Quote or Cite
-------
Figure 3-12. Tile Plots of Hourly PM Light Extinction, continued
Atlanta, GA
2005
2006
2007
3-
4 -
9-
10 -
11 -
12 -
2
12
Hour
18 24
1 -i
8-
9-
10 -
11 -
12 -
12
Hour
18
24
1 -
2 -
3 -
4 -
5-
6 -
7 -
8-
10 -
11 -
12 -
12
Hour
18 24
Low RH bext
n
Inf
190
111
64
High RH bext
Inf
190
111
64
January 2010
3-55
DRAFT - Do Not Quote or Cite
-------
Figure 3-12. Tile Plots of Hourly PM Light Extinction, continued
Detroit, Ml
2005
2006
2007
3-
4 -
5 -
6 -
7-
8-
9 -
10 -
11 -
12 -
2
1 -
2 -
3-
4 -
5 -
6-
7 -
8 -
9 -
10 -
11 -
12 -
1 -
2 -
3 -
4 -
5 -
6 -
7 -
8 -
9 -
10 -
11 -
12 -
Low RH bext
n
Inf
190
111
64
High RH bext
Inf
190
111
64
January 2010
3-56
DRAFT - Do Not Quote or Cite
-------
Figure 3-12. Tile Plots of Hourly PM Light Extinction, continued
Pittsburgh, PA
2005
2006
2007
2
1 -i
4 -
B -
6-
7 -
8-
9 -
10 -
3-
4 -
8 -
9-
10 -
11 -
12 -
12
Hour
18
24
1 -
2 -
3 -
4 -
5 -
6 -
7 -
8 -
10 -
11 -
12 -
12
Hour
18 24
Low RH bext
n
Inf
190
111
64
High RH bext
Inf
190
111
64
January 2010
3-57
DRAFT - Do Not Quote or Cite
-------
Figure 3-12. Tile Plots of Hourly PM Light Extinction, continued
Baltimore, MD
2005
2006
2007
1 -
2-
3 -
4 -
5 -
6 -
7 -
8-
9 -
10 -
11 -
12 -
2
12
Hour
18 24
1
2
3
4
5
6
7
8
9 -
10
11
12 -
~i i i i
7 12 18 24
Hour
2
3
4
5
6 -
7
8
10 -
11 -
12 -
Low RH bext
n
Inf
190
111
64
High RH bext
Inf
190
111
64
January 2010
3-58
DRAFT - Do Not Quote or Cite
-------
Figure 3-12. Tile Plots of Hourly PM Light Extinction, continued
Philadelphia, PA
2005
2006
2007
2 -
3-
4 -
5-
6-
7 -
8-
9 -
10 -
11
12 -
.L
„
2
i
7 12 18
Hour
1 n
2 -
3 -
4 -
7 -
8-
9 -
10-
11 -
12 -
!
24
ffi
7 12 18
Hour
1 -
2 -
3 -
4 -
5 -
6 -
7 -
8 -
10 -
i
24
t£
u-
Low RH bext
n
Inf
190
111
64
High RH bext
Inf
190
111
64
January 2010
3-59
DRAFT - Do Not Quote or Cite
-------
Figure 3-12. Tile Plots of Hourly PM Light Extinction, continued
New York, NY
2005
2006
2007
1 -i
2-
5 -
6-
7 -
8 -
9 -
10 -
11 -
12 -
1 -
2 -
3-
4 -
5-
6-
7 -
8 -
9-
10-
11 -
12 -
1 -
2 -
3-
4 -
5-
6-
7 -
8-
9 -
10-
11 -
12 -
Low RH bext
Inf
190
111
64
High RH bext
Inf
190
111
64
January 2010
3-60
DRAFT - Do Not Quote or Cite
-------
1 3.4.5 Extinction Budgets for High PM Light Extinction Conditions
2 An extinction budget for a single period shows the contribution that each PM component
3 makes to PM light extinction via the additive terms of the IMPROVE algorithm. It can be
4 expected that the pattern in the extinction budgets will vary by time of year and by study area.
5 Examination of extinction budgets allows initial insights into what pollutants cause poor urban
6 visibility and what emission reduction approach may be most effective in reducing PM light
7 extinction.
8 Figure 3-13 presents (a) day-specific maximum daylight 1-hour light extinction budgets
9 for the 10 percent of the days in each study area that have the highest daily maximum 1-hour PM
10 light extinction levels (excluding hours with relative humidity greater than 90 percent), and (b)
11 similar but more aggregated information based on all individual daylight hours. For the
12 maximum daily budgets, the day and hour of each hourly budget are indicated on the horizontal
13 axis, and the hours are arranged chronologically. There are too many individual daylight hours
14 within the top 10 percent group to display separately, so component concentrations for all days
15 within 10 one-percentile-point-range "bins" have been averaged together for display.42 Note that
16 the vertical scale differs from figure to figure, to accommodate the wide variation in PM light
17 extinction values. The pattern of results shown in Figures 3-13 is generally as expected in light
18 of emissions and climate differences among study areas. Except for the PM2.5 soil component,
19 each of the components of PM light extinction is a major contributor to extreme light extinction
20 events at some time and location. In the West, carbonaceous PM2.5 (i.e., organic mass and
21 elemental carbon), nitrate, and/or coarse mass (especially in Phoenix) tend to be most
22 responsible for these high haze hours. In the East it tends to be sulfate, nitrate, and the
23 carbonaceous PM2.5 components that are the large contributors to PM light extinction. From the
24 sample period dates we can determine the seasonal variations in major components. Nitrate and
25 carbonaceous PM2.s contribute more to the extreme light extinction periods during winter, while
26 sulfate contributes more in the summer. In many of the more northerly eastern urban areas, a
27 combination of sulfate and nitrate contributes to high light extinction year-round.
28 Looking at individual urban areas, the following are some highlights:
29 • Tacoma has its highest light extinction hours in the colder months and primarily
30 due to carbonaceous PM2.5 components. Because coarse PM was estimated by
31 applying a regional factor to the local PM2.s mass value, it would not have been
32 possible for the results to indicate a significant coarse PM contribution to PM
33 light extinction even if one existed at this site. However, from what EPA staff
42 Note that this binning approach may combine days with dissimilar extinction budgets into one bin because their
PM light extinction values are similar, obscuring some of the heterogeniety among hours.
January 2010 3-61 DRAFT - Do Not Quote or Cite
-------
1 know of the area, it is unlikely that there is a significant contribution from coarse
2 PM.
3 • Extreme haze hours in the two California urban areas are primarily caused by
4 high nitrate PM2.5, though Los Angeles has two extreme hours associated with
5 coarse PM and several other hours with moderate contribution from coarse PM.
6 Recall that estimates of coarse PM in Los Angeles are based in part on hourly
7 PMio measurements in Victorville, and may not represent coarse PM at the PM2.5
8 mass and speciation site in Rubidoux or in the larger South Coast Basin. Also,
9 such high coarse PM values may indicate influence from exceptional winds in
10 Victorville. Figure B-l(b) in Appendix B shows that next several other days with
11 high daily maximum PM coarse concentrations had concentrations only about 60
12 percent or less than on the two days appearing in Figure 3-13; the fact that these
13 other days do not appear among the top 10 percent indicates that other
14 contributors to PM light extinction were low on those days. Whether or not the
15 PMio measurements in Victorville represent the PM2.5 mass and speciation site in
16 Rubidoux, it can be concluded that nitrate and to a lesser extent sulfate dominate
17 PM light extinction on the days likely to be above the CPLs. Because coarse PM
18 for Fresno was estimated by applying a regional factor to the local PM2.5 mass
19 value, it would not have been possible for the results to indicate a significant
20 coarse PM contribution to PM light extinction even if one existed at the Fresno
21 site. However, given the presence of agricultural operations and occasional high
22 winds in the San Joaquin Valley, the possibility of a significant contribution from
23 coarse PM in some hours cannot not be ruled out.
24 • Phoenix is unique among the 15 urban areas in having most of its extreme light
25 extinction caused by coarse PM, though there are a few top-10-percent days
26 where the maximum hourly haze is dominated by carbonaceous, sulfate, and
27 nitrate PM2.5. Unlike for Los Angeles, this domination by coarse PM is no doubt
28 correct. PMio measurements for Phoenix come from a site near the center of the
29 metro area, while the PM2.5 measurements are from a more peripheral site (see
30 Appendix A) and are probably underestimates of PM2.5 at the PMio measurement
31 site, this would have only a small effect on estimates of coarse PM. While it is
32 quite possible that the very highest coarse PM concentration (indicated in Figure
33 B-l(b) to be about 500 |ig/m3) reflects the effect of exceptional winds, and might
34 be excluded under the Exceptional Event rule, the next-highest non-excludable
35 values of PM light extinction almost certainly would also be dominated by coarse
January 2010 3-62 DRAFT - Do Not Quote or Cite
-------
1 PM concentrations in the range of 150 to 200 |ig/m3 and many might not be
2 excludable.
3 • Salt Lake City has extreme haze hours caused mostly by nitrate in the winter with
4 some periods with carbonaceous PM2.5 being the major contributor. Because
5 coarse PM in Salt Lake City was estimated by applying a regional factor to the
6 local PM2.5 mass value, it would not have been possible for the results to indicate
7 a significant coarse PM contribution to PM light extinction even if one existed at
8 this site. However, from what EPA staff know of the area, it is unlikely that there
9 is a frequent large contribution from coarse PM. The area typically has at most a
10 few days per year with measured 24-hour average PMio as high as of 150-200
11 Hg/m3. If this were all coarse PM, the contribution to 24-hour average light
12 extinction would be 90-120 Mm"1, with the possibility of much higher hourly
13 contributions by coarse mass during these few days.
14 • Dallas and Houston have high contributions to PM light extinction by sulfate
15 PM2.5, but Dallas has some winter hours with extreme PM light extinction with
16 substantial contributions from nitrate and organic carbonaceous material, while
17 Houston seems to have less contribution by nitrate. Because coarse PM in both
18 Dallas and Houston was estimated by applying a regional factor to the local PM2.5
19 mass value, it would not have been possible for the results to indicate a significant
20 coarse PM contribution to PM light extinction even if one existed at this site.
21 However, from what EPA staff know of the areas, it is unlikely that there is a
22 frequent large contribution from coarse PM. Houston typically has at most a few
23 days per year with measured 24-hour average PMio as high as of 150-200 |ig/m3.
24 If this were all coarse PM, the contribution to 24-hour average light extinction
25 would be 90-120 Mm"1. Dallas typically does not have PMio as high as 150 |ig/m3
26 • Sulfate in the summer and nitrate in the fall and winter are responsible for most of
27 the extreme light extinction at St. Louis, though there are several maximum
28 hourly periods where coarse PM is a major component. Recall that estimates of
29 coarse PM in St. Louis may be affected by a very local source (see Appendix A),
30 and thus the instances of high PM light extinction due to coarse PM may be
31 limited in geographic scope.
32 • Birmingham and Atlanta are similar in having sulfate year-round and winter
33 carbonaceous PM2.5 as major contributors to their extreme light extinction periods.
34 Coarse PM for Birmingham was estimated using data from a single site, and the
35 estimates should be reasonably representative. Coarse PM for Atlanta was
January 2010 3-63 DRAFT - Do Not Quote or Cite
-------
1 estimated using data from two fairly close sites and the estimates should be
2 reasonably representative.
3 • Detroit has frequent large light extinction contributions from nitrate PM2.s, mostly
4 in the winter, as well as some contributions from sulfate PM2.5 year-round and
5 several fall and winter days with high contributions from carbonaceous PM2.5.
6 Coarse PM makes a notable contribution on a few days. Coarse PM for Detroit
7 was estimated using data from a single site near an automobile plant, and the
8 estimates should be reasonably representative for that site.
9 • The remaining four urban locations (Pittsburgh, Baltimore, Philadelphia, and New
10 York) are similar in that most of their extreme light extinction is from year-round
11 combinations of sulfate and nitrate. New York also has some winter elemental
12 and organic carbonaceous contributions to its extreme light extinction. Recall that
13 the PM2.5 site representing the New York area is actually in Elizabeth, NJ;
14 emissions from diesel trucks on nearby interstate highways and/or diesel engines
15 associated with port activities might explain the carbonaceous contributions.
16 Coarse PM for Baltimore and Philadelphia was estimated using data from a single
17 site in each area, and the estimates should be reasonably representative. Coarse
18 PM for New York was estimated using data from two fairly distant sites and the
19 estimates may not be representative of both sites. Because coarse PM was
20 estimated for Pittsburgh by applying a regional factor to the local PM2.5 mass
21 value, it would not have been possible for the results to indicate a significant
22 coarse PM contribution to PM light extinction even if one existed at this site.
23 However, exceedances of the PMio NAAQS are rare in Pittsburgh suggesting that
24 coarse PM likely is not a frequent significant contributor to PM light extinction.
25 3.5 POLICY RELEVANT BACKGROUND
26 Policy relevant background levels of PM light extinction have been estimated for this
27 assessment by relying on outputs for the 2004 CMAQ run in which anthropogenic emissions in
28 the U.S., Canada, and Mexico were omitted, as described in the second draft ISA. Estimates of
29 PRB for PM light extinction were calculated from modeled concentrations of PM2.5 components
30 using the IMPROVE algorithm. The necessary component concentrations were extracted from
31 the CMAQ output files, as they were not summarized in the final ISA. More detail is provided in
32 Appendix C.
33 It is also necessary to have estimates of PRB for PMio-2.5, as input to the IMPROVE
34 algorithm. The final ISA for this review does not present any new information on this subject.
35 The approach used in the two previous reviews was to present the historical range of annual
January 2010 3-64 DRAFT - Do Not Quote or Cite
-------
1 means of PMio-2.5 concentrations from IMPROVE monitoring sites selected as being least
2 influenced by anthropogenic emissions (US EPA, 2004, Table 3E-1). For this assessment, EPA
3 staff estimated PRB for PMio-2.5 using a contour map based on average 2000-2004 PMio-2.5
4 concentrations from all IMPROVE monitoring sites, found in a recent report from the
5 IMPROVE program (DeBell, 2006). More detail is provided in Appendix C.
6 The outcome of the procedures for estimating PRB consists of hour-specific estimates of
7 PRB for PM2.5 components and annual average estimates for PRB for PMio-2.5- Thus, hour-
8 specific estimates of PM light extinction are possible, using the same hour-specific relative
9 humidity values as for the estimate of current conditions PM light extinction.
10 The PRB estimates play a role in this assessment (other than allowing confirmation of the
11 obvious fact that current conditions PM light extinction values are generally well above PRB
12 conditions) only in the estimation of "what if scenarios representing compliance with alternative
13 NAAQS scenarios based on PM light extinction. This role is described in section 4.1.4.
14
15
January 2010 3-65 DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
10
11
Figure 3-13 Light Extinction Budgets for the Top 10 Percent of Days for Maximum
Daily 1-hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours
Tacoma
(a) Top 10 Percent of Days for Maximum Daily 1-hour PM light Extinction
Top 10% Daylight Hour Daily Maxima (RH<«90%): Tacoma, WA
9 (b) Top 10 Percent of Individual Daylight Hours (Aggregated)
Top 10% Daylight Hours Pereentile Average (RH<-90%): Tacoma, WA
90-91
B6-M Wf*^1
91 -92
>88-91 Wr^l
92-93
>»1-9T fcti^l
93-94
94-95 95-96
Pereentile Range
96-97 97-98 98-99 99-100
soil • EC • N03 n SO4 n OCM a PMc
January 2010
3-66
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours,
continued
Fresno
8
9
10
11
12
13
(a) Top 10 Percent of Days for Maximum Daily 1-hour PM light Extinction
Top 10% Daylight Hour Daily Maxima (RH<=90%) Fresno. CA
(b) Top 10 Percent of Individual Daylight Hours (Aggregated)
Top 10% Daylight Hours Percenlile Average (RH<=90%): Fresno, CA
91-92 92-93 93-94 94-95 95-96 96-97 97-98 98-99
Percentile Range
soil • EC • N03 n S04 n OCM n PMc
January 2010
3-67
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours,
continued
Los Angeles
(a) Top 10 Percent of Days for Maximum Daily 1-hour PM light Extinction
Top 10% Daylight Hour Daily Maxima (RH<=90%): Los Angeles. CA
soli "EC • NO3 n soi n OCM n PMC
(b) Top 10 Percent of Individual Daylight Hours (Aggregated)
Top 10% Daylight Hours Perceniile Average (RH<=90%): Los Angeles. CA
Percentile Range
• soil "EC • HO3 n S04 n OCM n PMc
11
12
January 2010
3-68
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours,
continued
Phoenix
(a) Top 10 Percent of Days for Maximum Daily 1-hour PM light Extinction
Top 10% Daylight Hour Daily Maxima (RH<=90%): Phoenix, AZ
• soil • EC • H03 n S04 n OCM n PMc
(b) Top 10 Percent of Individual Daylight Hours (Aggregated)
Top 10% Daylight Hours Percentile Average (RH<=90%) Phoenix. AZ
Percentile Range
• soil "EC • NO3 n SO4 n OCM n PMc
12
January 2010
3-69
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours,
continued
Salt Lake City
(a) Top 10 Percent of Days for Maximum Daily 1-hour PM light Extinction
Top 10% Daylight Hour Daily Maxima (RH<=90%): Salt Lake City. UT
8 -
(b) Top 10 Percent of Individual Daylight Hours (Aggregated)
Top 10% Daylight Hours Percentile Average (RH<=90%): Salt Lake Cily, UT
Percentile Range
NO3 n SO4 n OCM n PMc
January 2010
3-70
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours,
continued
Dallas
(a) Top 10 Percent of Days for Maximum Daily 1-hour PM light Extinction
Top 10% Dayligh* Hour Daily Maxima (RH«90%) Dallas, TX
O> -Cjf O> <£> O^ Of <%.
I • soil • EC • NO3 D SO4 O OCM n PMc |
(b) Top 10 Percent of Individual Daylight Hours (Aggregated)
Top 10% Daylight Hours Petcentile Average (RH<=90%) Dallas. TX
90-91
135-130 Wrt"-1
Percentile Range
soil "EC • NO3 n SO4 n OCM n PMc
January 2010
3-71
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours,
continued
Houston
(a) Top 10 Percent of Days for Maximum Daily 1-hour PM light Extinction
Top 10% DayligM Hour Daily Maxima (RH<=90<*): Houston TX
• soil • EC • H03 n S04 n OCM n PMc
\
(b) Top 10 Percent of Individual Daylight Hours (Aggregated)
Top 10% Daylight Hours Percentite Average (RH<=90%): Houston. TX
Percentile Range
NO3 n SO4 n OCM n PMc
January 2010
3-72
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours,
continued
St. Louis
(a) Top 10 Percent of Days for Maximum Daily 1-hour PM light Extinction
Top 10% Daylight Hour Daily Maxima (RH<=90%): SI. Louis, IL
12
13
• soil • EC • N03 n S04 n OCM n PMc
10 (b) Top 10 Percent of Individual Daylight Hours (Aggregated)
11
Top 10% Daylight Hours Percentile Average (RH<-90%) SI. Louis, IL
Percentile Range
soil • EC • N03 n S04 n OCM n PMc
January 2010
3-73
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours,
continued
Birmingham
(a) Top 10 Percent of Days for Maximum Daily 1-hour PM light Extinction
Top 10% Daylight Hour Daily Maxima (RH<=90%); Birmingham, AL
• soil • EC • NO3 D SO4 O OCM n PMc |
(b) Top 10 Percent of Individual Daylight Hours (Aggregated)
Top 10% Daylight Hours Percentite Average (RH<=90%) Birmingham. AL
Percentile Range
NO3 n SO4 n OCM n PMc
January 2010
3-74
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours,
continued
Atlanta
(a) Top 10 Percent of Days for Maximum Daily 1-hour PM light Extinction
Top 10% Daylight Hour Daity Maxima (RH<=90%): Atlanta, QA
• soli • EC • NO3 n SO4 a OCM n PMc
8
9
10 (b) Top 10 Percent of Individual Daylight Hours (Aggregated)
11
Top 10% Daylight Hours Rercenlile Average (RH<=90%) Atlanta. GA
Percentile Range
• soil "EC • NO3 n SO4 n OCM n PMc
12
13
January 2010
3-75
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours,
continued
Detroit
(a) Top 10 Percent of Days for Maximum Daily 1-hour PM light Extinction
Top 10% DayligM Hour Daily Maxima <£«& Ji % &
VJ> >>. "?=.
• soil • EC • H03 n S04 n OCM n PMc
(b) Top 10 Percent of Individual Daylight Hours (Aggregated)
Top 10% Daylight Hours Percentile Average (RH«90%) Detroit, Ml
90-91
235-229 Wn"-1
99-100
>376-1«1 Mrrv-1
Percentile Range
NO3 n SO4 n OCM n PMc
January 2010
3-76
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours,
continued
Baltimore
(a) Top 10 Percent of Days for Maximum Daily 1-hour PM light Extinction
Top 10% Daylight Hour Daily Maxima (RH<=90%) Baltimore. MD
\ \ \ \ \ \ \
•&. °t ^
• soil • EC • NO3 D SO4 a OCM n PMc
*t ' fo
(b) Top 10 Percent of Individual Daylight Hours (Aggregated)
Top 10% Daylight Hours Percenlile Average (RH«90%): Baltimore. MD
90-91 91 -92
DO-ierWn"-1 >1B7-19* Wv^
94-95 95-96
Percentile Range
soil "EC • NO3 n SO4 n OCM n PMo
January 2010
3-77
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours,
continued
Pittsburgh
(a) Top 10 Percent of Days for Maximum Daily 1-hour PM light Extinction
Top 10% Daylight Hour Daily Maxima (RH<=90%); Pittsburgh. PA
(b) Top 10 Percent of Individual Daylight Hours (Aggregated)
Top 10% Daylight Hours Percenlile Average (RH<=90%): Pittsburgh. PA
Percentile Range
soil "EC • NO3 n SO4 n OCM n PMc
January 2010
3-78
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
8
9
10
11
12
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours,
continued
Philadelphia
(a) Top 10 Percent of Days for Maximum Daily 1-hour PM light Extinction
Top 10% Daylight Hour Daity Maxima (RH«90%): Philadelphia, PA
(b) Top 10 Percent of Individual Daylight Hours (Aggregated)
Top 10% Daylight Hours Percentile Average (RH<=90%) Philadelphia. PA
90-91
1W-K1 wn"-1
Percentile Range
NO3 n SO4
OCM n PMc
January 2010
3-79
DRAFT - Do Not Quote or Cite
-------
1
2
3
4
5
6
7
9
10
Figure 3-13. Light Extinction Budgets for the Top 10 Percent of Days for Maximum Daily
1-hour PM light Extinction and for the Top 10 Percent of Individual Daylight Hours,
continued
New York
(a) Top 10 Percent of Days for Maximum Daily 1-hour PM light Extinction
Top 10% Daylight Hour Daily Maxima (RH<=90%): New York. NY
-S- '* °f 'o>
-------
1 4 PM LIGHT EXTINCTION UNDER "WHAT IF" CONDITIONS OF
2 JUST MEETING SPECIFIC ALTERNATIVE SECONDARY
3 NAAQS
4 4.1 ALTERNATIVE SECONDARY NAAQS BASED ON MEASURED PM
5 LIGHT EXTINCTION AS THE INDICATOR
6 4.1.1 Indicator and Monitoring Method
7 The indicator considered in this section is PM light extinction, assumed to be measured
8 by a continuous instrument, or instrument pair, capable of reporting both light scattering and
9 light absorption. For example, the measurement method could be an Aethalometer® or similar
10 instrument for measuring light absorption paired with a nephelometer, with both instruments
11 using a PMio inlet so that PM light extinction due to PM2.s and PMi0-2.5 combined would be
12 measured. A measurement of light absorption using an Aethalometer® or similar instrument
13 based on optical analysis of collected PM would not be affected by ambient NO2 concentrations.
14 Also, if a nephelometer is calibrated to zero using filtered or zero air and spanned using a light-
15 scattering span gas with a well characterized scattering coefficient, such as carbon dioxide,
16 SUVA 134A, Freon 12, or Freon 22, then subsequent measurements of light extinction would
17 reflect PM light scattering, without including the effect of Rayleigh scattering.
18 4.1.2 Alternative Secondary NAAQS Scenarios based on Measured PM light
19 extinction
20 Eighteen alternative NAAQS scenarios presented in Table 4-1 are analyzed in this
21 section. Nine are based on daily maximum daylight 1-hour PM light extinction and nine on light
22 extinction in all hours without the restriction to daily maxima. Within each set of nine, the
23 scenarios are ordered from least to most stringent.
24
25
January 2010 4-1 DRAFT - Do Note Quote or Cite
-------
Table 4-1. Alternative Secondary NAAQS Scenarios for PM Light Extinction
Level
Annual
Percentile
Form
Scenarios Based on Daily Maximum Daylight 1-hour PM Light Extinction
(a) 191 Mm'1
(b) 191 Mm'1
(c) 191 Mm'1
(d) 112 Mm'1
(e) 112 Mm'1
(f) 112 Mm'1
(g) 64 Mm'1
(h) 64 Mm'1
(i) 64 Mm'1
90
95
98
90
95
98
90
95
98
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
Scenarios Based on Daylight 1-hour PM Light Extinction (All Daylight Hours)
(j) 191 Mm'1
(k) 191 Mm'1
(1) 191 Mm'1
(m) 112 Mm'1
(n) 112 Mm'1
(o) 112 Mm'1
(p) 64 Mm'1
(q) 64 Mm'1
(r) 64 Mm'1
90
95
98
90
95
98
90
95
98
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3-year average of percentile value
3 4.1.3 Monitoring Site Considerations for Alternative Secondary NAAQS Based on
4 Measured PM light extinction
5 It is useful to think ahead tentatively to monitor siting aspects of NAAQS
6 implementation, so that the results presented in the remainder of this chapter based on the 15
7 specific study sites can be better interpreted in terms of how well they might represent later
8 findings if these (and other) areas were to deploy PM light extinction measurement instruments
9 as part of implementing a secondary NAAQS.
10 It is most likely that the instruments that would be used to implement a secondary
11 NAAQS with an indicator based on measured PM light extinction will be "closed path"
12 instruments that react only to air quality in their immediate vicinity. However, light paths that
13 matter to perceived visual air quality are likely to be several kilometers long. Therefore, a
14 monitoring site should be at least neighborhood in scale, i.e., its relationship to emission sources
15 and transport should be such that measurements made at the site reasonably reflect
16 concentrations in an area surrounding the site of at least about 0.5 to 4 kilometers in diameter.
17 It would be logical to require that in any urban area for which light extinction monitoring
18 is deemed a necessary requirement, at least one monitoring site would be placed in an area
January 2010
4-2
DRAFT - Do Note Quote or Cite
-------
1 expected to have the maximum PM light extinction conditions, subject to the above scale of
2 representation consideration and possibly also subject to the condition that the site be in an area
3 (or reasonably represent such an area) where valued urban scenes are able to be perceived by
4 people, i.e., that the site is "population oriented." It is difficult to imagine a neighborhood scale
5 monitoring location within the census-defined urbanized area of an urban area which would not
6 be "population oriented" for purposes of visual air quality, as "neighborhood" size land areas
7 typically would have residents, workers, etc. somewhere within them during daylight hours.
8 With regard to the monitoring sites used in this assessment, all are reported to be, or
9 appear to be, neighborhood or larger scale, and all are in areas where people are present during
10 daylight hours. The sites in Detroit (Dearborn) and New York (Elizabeth, NJ) are, however,
11 rather close to an industrial source and a major interstate highway interchange/turnpike exit,
12 respectively. Significantly, most of the study sites are not the highest PM2.5 concentration site in
13 their urban area, so a "what if scenario that manipulates the "current conditions" at these sites to
14 "just meet" an alternative secondary NAAQS might implicitly leave other parts of their urban
15 areas with PM light extinction above the NAAQS.
16 Probe height is another consideration. For purposes of a secondary NAAQS aimed to
17 protect visibility, monitoring probes logically should be placed so that the sampled air is
18 reasonably representative of the air along the sight path to the valued scene, which may be
19 different than the probe heights of the monitors that provided data for this assessment.43 We
20 have not yet studied this issue further.
21 4.1.4 Approach to Modeling "What If Conditions for Alternative Secondary
22 NAAQS Based on Measured PM Light Extinction
23 Before modeling "what if conditions, EPA staff augmented the data set described in
24 Table 4 so that the sets of study days for Houston and Phoenix were seasonally balanced despite
25 the lack of actual monitoring data for one quarter in each city. For the first quarter of 2005 in
26 Phoenix, we substituted the available 12 days from the first quarter of 2006. For the fourth
27 quarter of 2007 in Houston, we substituted 13 randomly drawn days from the fourth quarters of
28 2005 and 2006.
29 Also, Tacoma (originally) and Phoenix (after this augmentation) each have only two
30 calendar years of suitable data, while the form of the alternative NAAQS scenarios requires the
31 averaging of the 90th, 95th, or 98th percentile values from three years. In Tacoma and Phoenix,
32 for every step in the analysis at which a design value is used as an input or reported as an output,
33 we averaged the percentile values from the only two available years.
43 Probe height influence on measured PM light extinction might alternatively be taken into consideration in setting
the level of a NAAQS.
January 2010 4-3 DRAFT - Do Note Quote or Cite
-------
1 We modeled daylight and daily maximum daylight 1-hour PM light extinction under each
2 of the "what if scenarios (in which each study area "just meets" one of the 18 alternative
3 secondary NAAQS listed in section 4.1.2) via the following steps. These steps are essentially the
4 same as the "proportional rollback" steps that have been used in the health risk assessment
5 modeling of "what if conditions in several previous NAAQS reviews for PM and other criteria
6 pollutants. The steps are described here for the nine scenarios based on daily maximum daylight
7 1-hour PM light extinction; similar steps were followed for the nine scenarios based on
8 percentiles of all daylight 1-hour PM light-extinction. The referenced tables present results for
9 both sets of scenarios.
10
11 1. After excluding hours with relative humidity greater than 90 percent, identify the
12 appropriate percentile (90th, 95th, or 98th) daily maximum daylight 1-hour light
13 extinction value in each year, noting the day and hour each occurred, and average
14 these values across years to calculate the light extinction design value for each site
15 consistent with the percentile form of the NAAQS scenario.44 The three resulting
16 design values for each area (for the 90th, 95th, and 98th percentile forms) are shown in
17 Table 4-2. (Note that in a number of cases, which are identified by a footnote, the
18 study area meets one or more of the NAAQS scenario under current conditions. In
19 these cases, the "current conditions" PM light extinction values are not adjusted, i.e.,
20 PM light extinction values are never "rolled up.") Notice that the design values for
21 the 90* percentile maximum daily 1-hour for most cities are generally similar to the
22 design values for the 98* percentile of all daylight hours. On average there are about
23 ten hours defined as daylight per day, so if the light extinction were randomly
24 distributed among the daylight hours and days, the 90th percentile maximum daily 1-
25 hour would correspond to the 99* percentile of all hours; the fact that the point of
26 rough equivalency is the 98 percentile indicates a tendency for hours with higher
27 light extinction to cluster together in the same day. Figure 4-1 presents two scatter
28 plots that relate the design values based on daily maximum 1-hour PM light
29 extinction values and the design values based on all daylight 1-hour light extinction
30 values. In Panel A, design values for the daily maximum and all hours forms are
31 paired by the defining percentile, and colors are used to distinguish the 90th, 95th, and
32 98th percentile statistical forms. It appears from Panel A that the design values for the
33 two approaches to defining the NAAQS scenarios are highly correlated but with the
34 all hours approach resulting in numerically lower design values than the daily
35 maximum approach. The correlation breaks down for the 98th percentile form for the
36 few study areas with the highest levels of PM light extinction. Panel B compares the
44 Annual percentile values were picked from the set of day-specific or hour-specific estimates according to the same
scheme as used for the current 24-hour secondary PM2.5 standard, as explained in section 4.5(a) of 40 CFR 50
appendix N. For example, if there are 60 daily maximum values in a year, the second highest value is the 98th
percentile value. Note that this differs from the algorithm used by some spreadsheet and other statistical programs,
which may interpolate between sample values. Also, this is a different approach than in the Regional Haze program,
in which conditions in the best and worst 20 percent of days are averaged together, rather than focusing on
conditions on the specific day at the 80th and 20th percentile points.
January 2010 4-4 DRAFT - Do Note Quote or Cite
-------
1
2
3
4
5
6
7
90* percentile design values based on daily maximum PM light extinction with the
90th, 95th, and 98th percentile design values based on all daylight hours PM light
extinction. There is close agreement between the 90th percentile design values based
on daily maximum values and the 98 percentile design value based on all daylight
hours.
Table 4-2. Current Conditions PM light extinction design values for the study areas.
Study Area
Design Value for
90th Percentile
Form (Mm ~l)
Design Value for 95th
Percentile Form (Mm"1)
Design Value for
98th Percentile Form
(Mm1
Design Values Based on Daily Maximum Daylight 1-hour PM Light Extinction
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
140*
338
469
105*
164*
183*
194
307
357
249
310
278
246
286
306
157*
463
554
144*
252
239
234
381
483
288
473
313
286
339
355
210
533
624
266
410
302
291
467
562
331
644
364
328
393
457
Design Values Based on Daylight 1-hour PM Light Extinction (All Daylight Hours)
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
76*
190*
266
68*
93*
113*
105*
194
173*
166*
212
167*
172*
183*
186*
106*
266
349
79*
142*
143*
128*
235
227
195
251
209
227
222
244
136*
373
451
94*
225
188*
171*
290
309
238
315
264
265
279
300
* This design value meets one or more of the NAAQS scenarios based on PM light extinction.
January 2010
4-5
DRAFT - Do Note Quote or Cite
-------
1 Figure 4-1 Comparison of Daily Max and All Daylight Hour Design Values
2 (A) Comparison of design values matched by percentile form
4
5
6
7
7nn -,
^ finn
0)
— f^nn
re ouu
>
c
01 Ann
••$ 400
3
2 ^nn
^ ouu
o
** onn
£ zlUU
0)
">
-------
1 2. Using the same days and hours, find the three (or two, in the case of Phoenix and
2 Houston for which there were only two years of suitable data available)
3 corresponding values of PRB PM light extinction, and average these values across
4 years to calculate the PRB portion of the design value.
5
6 3. Subtract the value from step 2 from the value from step 1, to determine the non-PRB
7 portion of the design value.
8
9 4. Calculate the percentage reduction required in non-PRB PM light extinction in order
10 to reduce the design value to the PM light extinction level that defines the NAAQS
11 scenario, using the following equation:
12
13 Percent reduction required = 1 - (NAAQS level - PRB portion of the design value)/(non-PRB
14 portion of the design value)
15
16 The percentage reductions determined in step 4 are shown in Table 4-3. Figure 4-2
17 presents them graphically in the form of a scatter plot, comparing the required
18 reductions for scenarios based on daily maximum 1-hour daylight PM light extinction
19 values to scenarios with the same level and percentile form but based on all daylight
20 hours 1-hour PM light extinction values. For the NAAQS scenarios involving higher
21 levels and lower percentile forms, there are some notable differences in the
22 percentage reductions required for some area to attain. As was the case for the design
23 values, notice in Table 4-3 that there is generally similar percentage reductions for
24 each city and level for the 90th percentile maximum daily and 98th percentile of all
25 daylight hours.
26
27 As already stated, if the study area is meeting a NAAQS scenario in the current
28 conditions case, no adjustments were made to represent the "just meeting" case. In
29 effect, negative values for the percent reduction required to meet the NAAQS
30 scenario calculated by the above equation were re-set to zero.
311.
32 5. Turning to the entire set of day/hour-specific actual and PRB daylight PM light
33 extinction values for the three (or two) year period, determine the non-PRB portion of
34 PM light extinction in an hour, reduce it by the percentage determined in step 4, and
35 add back in the PRB PM light extinction. The result is the "just meets" PM light
36 extinction value for that day and hour.
37
38 Note that in these steps, it is not necessary to make any explicit or implicit assumption
39 about what PM components would be reduced to allow the area to meet the NAAQS scenario, as
40 the NAAQS scenario's target design value is itself in units of light extinction. One path to
41 meeting a NAAQS scenario would be to reduce each of the five PM2.5 components (and thus the
42 annual and 24-hour design values shown in Table 3-2) and PMio-2.5 by the calculated "percent
43 reduction required". However, a lesser reduction in one or more of the six PM concentrations
44 could be offset by a greater reduction in or more of the remaining concentrations. Thus, it is not
January 2010 4-7 DRAFT - Do Note Quote or Cite
-------
1 possible to associate unique values of annual average and 24-hour average PM2.5 with the "just
2 meeting" NAAQS scenarios reported in Table 4-3.
January 2010 4-8 DRAFT - Do Note Quote or Cite
-------
1 Table 4-3. Percentage reductions in non-PRB light extinction required to "just meet" the
2 NAAQS scenarios based on measured light extinction.
Level (Mm-1)
Percentile
Form
Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Level (Mm-1)
Percentile
Form
Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
NAAQS Scenarios Based on Daily Maximum 1-hour Daylight PM light extinction, Average
of Percentile Value Over Three Years
(a)
191
90th
(b)
191
95th
(c)
191
98th
(d)
112
90th
(e)
112
95th
(f)
112
98th
(g)
64
90th
00
64
95th
(i)
64
98th
Percentage Reduction Required in
Non-PRB PM light extinction
0
45
61
0
0
0
2
39
48
25
39
32
23
34
39
0
60
67
0
25
21
20
51
61
35
61
40
35
45
48
10
65
71
0
54
39
36
61
67
44
71
48
43
52
59
22
69
78
0
33
42
45
66
71
58
66
61
57
63
65
34
77
82
23
57
55
57
73
78
64
78
66
64
68
70
53
81
84
60
74
66
64
78
82
69
84
70
68
73
77
61
84
88
42
63
70
71
82
85
78
82
79
77
80
81
70
88
90
59
76
76
79
86
88
81
88
81
81
83
84
80
90
91
78
85
83
81
89
90
84
91
84
83
85
88
NAAQS Scenarios Based on 1-hour Daylight PM light extinction, Average of Percentile
Value Over Three Years (All Daylight Hours)
(i)
191
90th
00
191
95th
(1)
191
98th
(m)
112
90th
(n)
112
95th
(o)
112
98th
(P)
64
90th
(q)
64
95th
(r)
64
98th
Percentage Reduction Required in
Non-PRB PM light extinction
0
0
29
0
0
0
0
1
0
0
10
0
0
0
0
0
29
47
0
0
0
0
19
17
2
25
9
16
15
22
0
50
59
0
15
0
0
35
39
21
40
28
29
32
37
0
43
60
0
0
1
0
44
37
34
49
34
37
41
42
0
60
70
0
22
23
13
54
53
45
57
48
52
52
56
22
72
77
0
51
42
38
63
66
55
66
59
60
62
64
19
70
78
6
33
46
42
70
66
64
73
64
66
68
69
46
79
84
20
56
59
53
76
75
71
77
71
74
74
76
67
85
88
34
73
69
70
80
82
77
82
78
79
79
81
January 2010
4-9
DRAFT - Do Note Quote or Cite
-------
1 Figure 4-2 Comparison of Required Percentage Reductions in Non-PRB PM Light
2 Extinction Needed to Meet NAAQS Scenarios
"«!" -inn -,
21 luu
5 on
an
ra
"^ 7n
UJ 7°
— i fin
+J
-= c;n
O) aU
"^. /in
« 40
Q on
o on
•o
a) m
8) 1U
iS o <
(
-
c__ *
*m.
_
0- • 3F
X, X** •*-
Afc
A Vk?A*
A A *
AB •*_"
• • BAX •
ft
) 20 40 60 80 100
Based on Daily Maximum Daylight LE
* 191790th • 191795th A 191798th * 112790th x 112795th • 112798th + 64790th
- 64795th 64798th
4 4.2 ALTERNATIVE SECONDARY PM2.5 NAAQS BASED ON ANNUAL AND
5 24-HOUR PM2.5 MASS
6 4.2.1 Secondary NAAQS Scenarios Based on Annual and 24-hour PM2.5 Mass
7 In this second draft version of the assessment, EPA staff have modeled two "what if
8 scenarios based on the same indicators and averaging periods as define the current secondary
9 PM2.5 NAAQS:
10 • 15 |ig/m3 weighted annual average PM2.5 concentration and 35 |ig/m3 24-hour average
11 PM2.5 concentration with a 98th percentile form, both averaged over three years. These
12 are the current secondary NAAQS for PM2.5.
13 • 12 |ig/m3 weighted annual average PM2.5 concentration and 25 |ig/m3 24-hour average
14 PM2.5 concentration with a 98th percentile form, both averaged over three years.
15 These are the highest and lowest alternative NAAQS scenarios considered in the health
16 risk assessment, and therefore encompass the full range of alternative primary PM2.5 NAAQS
17 being analyzed by EPA staff.
18
January 2010
4-10
DRAFT - Do Note Quote or Cite
-------
1 4.2.2 Approach to Modeling Conditions If Secondary PM2 5 NAAQS Based on
2 Annual and 24-hour PM2 5 Mass Were Just Met
3 Because these NAAQS scenarios are based on PM2.5 mass as the indicator, rather than
4 light extinction, the steps needed to model "what if conditions are somewhat different, and
5 involve explicit consideration of changes in PM2.5 components.
6
7 1. Apply proportional rollback to all the PM2.5 monitoring sites in each study area,
8 taking into account PRB PM2.5 mass, to "just meet" the NAAQS scenario for the area
9 as a whole, not just at the visibility assessment study site. The health risk assessment
10 document describes this procedure in detail. The degree of rollback is controlled by
11 the highest annual or 24-hour design value, which in most study areas is from a site
12 other than the site used in this visibility assessment. The relevant result from this
13 analysis is the percentage reduction in non-PRB PM2.5 mass need to "just meet" the
14 NAAQS scenario, for each study area. These percentage reductions are shown in
15 Table 4-4. Note that Phoenix and Salt Lake City meet the 15/35 NAAQS scenario
16 under current conditions, and require no reduction. PM2.5 levels in these two cities
17 were not "rolled up."
18
19 2. For each day and hour for each PM2.5 component, subtract the PRB concentration
20 from the current conditions concentration, to determine the non-PRB portion of the
21 current conditions concentration.
22
23 3. Apply the percentage reduction from step 1 to the non-PRB portion of each of the
24 five PM2.5 components. Add back the PRB portion of the component.
25
26 4. Re-apply the IMPROVE algorithm, using the reduced PIVb.s component
27 concentrations, the current conditions PMio-2.5 concentration for the day and hour, and
28 relative humidity for the day and hour. Include the term for Rayleigh scattering.
29
30 The results of these steps are shown in Table 4-6.
31
January 2010 4-11 DRAFT - Do Note Quote or Cite
-------
1 Table 4-4. Percentage reductions required in non-PRB PMi.s mass to "just meet" NAAQS
2 scenarios based on annual and 24-hour PMi.s mass
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Percentage Reduction Required
(s)
Annual PM2 5 NAAQS = 15 ug/m3
24-hour PM2 5 NAAQS = 35 ug/m3
19
45
37
0*
37
0*
6
10
22
8
19
19
6
8
17
(t)
Annual PM2 5 NAAQS = 12 ug/m3
24-hour PM2 5 NAAQS = 25 ug/m3
43
61
55
22
56
7
27
37
45
30
43
43
33
35
41
* These areas meet this NAAQS scenario under current conditions.
4 4.3 RESULTS FOR "JUST MEETING" EACH ALTERNATIVE SECONDARY
5 NAAQS SCENARIO
6 The modeling described in sections 4.1 and 4.2 resulted in estimates of PM light
7 extinction for each day and hour in each study area, for each NAAQS scenario. Four summaries
8 of these conditions are presented here.
9 Figure 4-3 shows two box-and-whisker plots of daily maximum daylight 1-hour PM light
10 extinction. The top panel (a) is for the single illustrative scenario of a NAAQS based on daily
11 maximum daylight 1-hour light extinction with a level of 112 Mm"1 and a 90* percentile form,
12 which was chosen for this illustration because it is approximately mid-way among the nine
13 scenarios based on daily maximum light extinction in terms of stringency.45 The bottom panel
14 (b) is for the scenario of meeting the current secondary PM2 5 NAAQS of 15 |ig/m3 for the annual
15 average and 35 |ig/m3 for the 98th percentile 24-hour average. A notable feature of this
16 comparison is that in the top panel, all the study areas have a similar distribution of the daily
45 Plots of the distribution of daily maximum light extinction for all 18 NAAQS scenarios based on daily maximum
light extinction, and of individual hourly light extinction for all 18 NAAQS scenarios based on individual daylight
hours, are provided in Appendix F.
January 2010
4-12
DRAFT - Do Note Quote or Cite
-------
1 maximum daylight 1-hour PM light extinction, while in the bottom panel this is not the case.
2 This is expected, since a NAAQS based on a measured daily maximum PM light extinction
3 indicator will of course result in areas achieving similar daily maximum PM light extinction
4 patterns once each area reaches a "just meets" condition. In areas with generally higher relative
5 humidity conditions, concentrations of PM2.s components and/or PMi0-2.5 would need to be lower
6 to achieve the "just meet" condition. In contrast, in the NAAQS scenario represented by the
7 bottom panel, concentrations of PM2.5 mass will be similar across areas, but concentrations of
8 PM2.5 components may not be, and levels of PM light extinction will not be similar in areas with
9 dissimilar levels of relative humidity. The specific differences among areas in the bottom panel
10 are generally as expected, with the drier study areas having lower levels of PM light extinction.
11 Tables 4-5 and 4-6 summarize the "just meets" conditions in the NAAQS scenarios in
12 terms of the PM light extinction design values when just meeting. Table 4-5 addresses the 18
13 scenarios of NAAQS based on measured PM light extinction. When an area just meets a
14 NAAQS scenario, its design value in principle should exactly equal the NAAQS level, so
15 preparation of this table serves as a check against calculation errors. Note that the design values
16 in Table 4-5, resulting from the rollback steps described in section 4.1.4, in some cases do not
17 exactly equal the assumed level of the NAAQS, although all are quite close. Closer investigation
18 has revealed that this is mostly a result of hours switching their ranking in the rollback process.
19 Hours can switch rank because the level of PRB PM light extinction varies with each hour, so a
20 uniform percentage reduction in non-PRB light extinction (step 5) can result in non-uniform
21 percentage reductions in actual PM light extinction; a lower ranking hour can thereby move up in
22 the post-rollback ranking. In principle, rollback could be iterated to exactly achieve a design
23 value equal to the level of the NAAQS for each scenario. However, the discrepancies indicated
24 in Table 4-5 were judged too small to justify iterative rollback, given other uncertainties in the
25 analysis.
26 Table 4-6 addresses the two scenarios of NAAQS based on PM2.5 mass, with PM light
27 extinction design values shown for the 90* , 95* , and 98* percentile forms.
28 Table 4-7 summarizes all 20 scenarios in terms of the percentage of days (across 2005 to
29 2007, but after rollback) in which the daily maximum daylight 1-hour PM light extinction under
30 "just meeting" conditions exceeds each of the CPLs. Part A of the table applies to NAAQS
31 scenarios based on daily maximum 1-hour PM light extinction values. Part B of the table applies
32 to the scenarios based on 1-hour PM light extinction values during all daylight hours. Note that
33 the reported percentages in both Part A and Part B is the percentage of days in which the daily
34 maximum daylight 1-hour PM light extinction under "just meeting" conditions exceeds each of
35 the CPLs; this allows comparison of the "effectiveness" of the two NAAQS approaches using a
36 consistent metric. (The 15/35 and 12/25 NAAQS scenarios are the same in Part A and Part B,
January 2010 4-13 DRAFT - Do Note Quote or Cite
-------
1 and are repeated only for convenience in making comparisons.) Hours with relative humidity
2 above 90 percent have been excluded from consideration, consistent with the definition of the
3 NAAQS scenarios. Also shown at the bottom of the table in each column representing a
4 NAAQS scenario is the average of these percentages of time across the 15 study areas (this is the
5 simple column average, not weighted by the number of days available in each area).
6 Comparisons of these percentages allows a rough indication of how the two scenarios of a
7 NAAQS based on PIVb.s mass compare to the other 18 scenarios in terms of protecting visual air
8 quality. Notice that the most restrictive of the two NAAQS scenarios based on PM2.5 mass
9 would reduce the projected 1-hour maximum daily light extinction above the least restrictive
10 CPL (IQlMm"1) to less than 10 percent of the time for most of the urban areas (only L.A., St.
11 Louis, and Birmingham have values above 10 percent). However at the current PM NAAQS
12 level (i.e., 15/35) all of the eastern urban areas and Los Angeles exceed the least restrictive CPL
13 more than 10% of the time. Comparison of Parts A and B of Figure 4-7 indicates that basing a
14 PM light extinction NAAQS scenario on daily maximum 1-hour light extinction has a lower
15 percentage in excess of the 1-hour daily maximum versus the NAAQS scenario based on all
16 daylight hours light extinction for a given level and percentile form of the NAAQS. This is
17 consistent with the results presented in Table 4-2 and Figure 4-1, which indicated that current
18 conditions design values are generally lower for the all hours approach. Again there is near
19 equivalence between the 90* percentile daily maximum and 98* percentile all daylight hours in
20 terms of the percent of days exceeding the daily maximum CPL values in Table 4-7.
21
22
January 2010 4-14 DRAFT - Do Note Quote or Cite
-------
Figure 4-3 Distributions of daily maximum daylight 1-hour PM light extinction under two "just meeting" secondary NAAQS
scenarios (excluding hours with relative humidity greater than 90 percent)
(a) Secondary NAAQS based on daily maximum daylight 1-hour PM light extinction with a level of 112 Mm-1 and a 90th
percentile form
ExtRo!IbackDaiIyMaxNAAQS112Pctl90DVsFromdaily.max
o
o _
o
o
CM
108 324 300 i8 306 273 158 289 349 279 141 277 181 143 225
X O
LU o
, . <£)
O
O
e,*
January 2010
4-15
DRAFT - Do Note Quote or Cite
-------
Figure 4-3. Distributions of daily maximum daylight 1-hour PM light extinction under two "just meeting" secondary NAAQS
scenarios (excluding hours with relative humidity greater than 90 percent), continued
(b) Secondary NAAQS of 15 ug/m3 for the annual average and 35 ug/m3 for the 98th percentile 24-hour average
PMRollbackDailyMaxCasel NAAQS
•2 8
«— CD
x
January 2010
4-16
DRAFT - Do Note Quote or Cite
-------
Table 4-5. PM light extinction design values for "just meeting" secondary NAAQS
scenarios based on measured PM light extinction (excluding hours with relative humidity
greater than 90 percent)
Level (Mm'1)
Percentile Form
Tacoma, WA
Fresno, CA
Los Angeles, CA
Phoenix, AZ
Salt Lake City, UT
Dallas, TX
Houston, TX
St. Louis, IL
Birmingham, AL
Atlanta, GA
Detroit, MI
Pittsburgh, PA
Baltimore, MD
Philadelphia, PA
New York, NY
Level (Mm'1)
Percentile Form
Tacoma, WA
Fresno, CA
Los Angeles, CA
Phoenix, AZ
Salt Lake City, UT
Dallas, TX
Houston, TX
St. Louis, IL
Birmingham, AL
Atlanta, GA
Detroit, MI
Pittsburgh, PA
Baltimore, MD
Philadelphia, PA
New York, NY
Secondary NAAQS Scenarios Based on Daily Maximum
(a)
191
90th
(b)
191
95th
(c)
191
98th
(d)
112
90th
(e)
112
95th
(f)
112
98th
(K)
64
90th
(h)
64
95th
(0
64
98th
PM light extinction Design Value
(based on same percentile form as the NAAQS scenario)
140
191
191
105
164
183
191
191
191
191
191
191
191
191
192
157
191
191
144
191
191
191
191
192
191
191
191
191
191
191
191
191
191
185
191
191
191
191
191
191
191
191
191
191
191
112
112
112
105
112
113
115
113
113
112
112
112
111
112
113
112
112
112
112
112
113
112
112
114
111
112
112
112
112
112
108
112
112
112
112
112
112
112
112
112
112
112
112
112
112
66
64
65
64
64
64
67
65
64
64
64
64
63
65
65
70
64
64
64
64
66
61
64
66
63
64
64
64
64
64
60
64
64
64
64
66
67
64
64
65
65
64
65
64
64
Secondary NAAQS Scenarios Based on All Daylight Hours
(i)
191
90th
(k)
191
95th
®
191
98th
(m)
112
90th
(n)
112
95th
(o)
112
98th
(P)
64
90th
(q)
64
95th
(r)
64
98th
PM light extinction Design Value
(based on same percentile form as the NAAQS scenario)
76
190
192
68
93
113
105
191
173
166
191
167
172
183
186
106
191
191
79
142
143
128
191
191
191
191
191
191
191
191
136
191
192
94
191
188
171
191
192
192
191
191
191
191
191
76
113
113
68
93
112
105
113
113
113
112
113
112
112
112
106
112
112
79
112
112
113
112
112
111
113
113
112
112
112
112
112
112
94
112
113
110
111
113
113
112
112
112
113
113
63
65
66
64
64
66
65
65
65
66
64
65
64
64
64
64
64
64
64
64
64
66
65
64
63
65
66
65
64
66
59
64
64
64
64
67
61
65
65
65
64
64
64
65
66
January 2010
4-17
DRAFT - Do Note Quote or Cite
-------
Table 4-6. PM light extinction design values for "just meeting" secondary NAAQS
scenarios based on PMi.s mass (excluding hours with relative humidity greater than 90
percent)
Annual/1 -hour
PM2S NAAQS
City Name
Tacoma, WA
Fresno, CA
Los Angeles, CA
Phoenix, AZ
Salt Lake City, UT
Dallas, TX
Houston, TX
St. Louis, IL
Birmingham, AL
Atlanta, GA
Detroit, MI
Pittsburgh, PA
Baltimore, MD
Philadelphia, PA
New York, NY
Tacoma, WA
Fresno, CA
Los Angeles, CA
Phoenix, AZ
Salt Lake City, UT
Dallas, TX
Houston, TX
St. Louis, IL
Birmingham, AL
Atlanta, GA
Detroit, MI
Pittsburgh, PA
Baltimore, MD
Philadelphia, PA
New York, NY
(s)
15ng/m3 / 35ng/m3
90th %tile
Design Value
(Mm1)
95th %tile
Design Value
(Mm1)
98th %tile
Design Value
(Mm1)
(t)
12ng/m3 / 25ng/m3
90th %tile
Design Value
(Mm1)
95th %tile
Design Value
(Mm1)
98th %tile
Design Value
(Mm1)
Design Values Based on Daily Maximum Daylight 1-hour PM Light Extinction
120
195
323
105*
110
183*
185
286
285
230
257
229
233
264
255
131
267
365
143*
168
239*
222
355
394
266
389
258
272
313
296
177
306
436
265*
269
302*
276
441
464
307
536
299
310
364
381
94
144
239
97
83
172
148
253
213
181
189
167
169
194
183
102
197
263
135
125
224
178
289
300
208
278
188
202
226
213
136
225
360
250
198
282
220
364
365
243
401
218
222
269
272
Design Values Based on Daylight 1-hour PM Light Extinction (All Daylight Hours)
65
112
176
68*
63
113*
99
181
140
154
176
138
163
169
156
88
154
233
79*
95
143*
122
221
183
180
209
173
215
206
204
113
217
299
94*
150
188*
163
271
247
220
258
218
251
258
250
52
84
131
60
48
106
81
147
105
123
130
102
121
123
113
70
115
172
70
72
134
99
183
138
144
155
127
157
150
148
84
161
223
86
112
176
131
237
186
174
188
159
184
187
179
* Phoenix and Dallas meet 15 ug/m 735 ug/m under current conditions, so these entries are essentially the same as
for current conditions.
January 2010
4-18
DRAFT - Do Note Quote or Cite
-------
Table 4-7. Percentage of days across three years (two in the case of Phoenix and Houston) with maximum 1-hour daylight PM light
extinction above CPLs when "just meeting" the NAAQS scenarios
(A) NAAQS Scenarios Based on Daily Maximum 1-hour PM Light Extinction
NAAQS
Level Mm"1
NAAQS
Percentile
Form
Annual/24-
hour
Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake
City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Average
Days with max hour above
64 Mm -1
(a)
191
90
(b)
191
95
(c)
191
98
(d)
112
90
(e)
112
95
(f)
112
98
(0
64
90
(h)
64
95
(i)
64
98
(s)
157
35
(t)
127
25
Percentage of days
52
55
74
44
44
80
77
83
63
86
74
70
68
73
63
67
52
42
67
44
27
66
65
72
51
81
54
64
62
66
59
58
49
34
60
44
14
51
57
57
41
77
43
54
55
61
42
49
40
31
44
44
24
49
47
47
34
62
48
40
43
43
35
42
32
20
35
27
11
28
31
35
20
52
23
32
31
33
28
29
17
15
28
10
5
14
20
21
16
37
6
24
27
28
18
19
11
10
11
10
10
11
13
12
10
10
11
9
12
8
9
10
7
4
6
5
5
5
4
6
6
5
4
6
4
6
6
5
2
3
3
2
1
1
3
2
3
1
3
2
2
3
2
2
43
54
85
44
24
81
75
97
84
90
80
78
78
85
76
72
28
40
79
40
15
77
64
89
70
85
74
63
64
74
62
62
Days with max hour above
112 Mm -1
(a)
191
90
(b)
191
95
(c)
191
98
(d)
112
90
(e)
112
95
(f)
112
98
(P)
64
90
(h)
64
95
(i)
64
98
(s)
157
35
(t)
127
25
Percentage of days
22
30
41
6
17
41
43
45
31
59
45
38
40
39
32
35
22
16
32
6
11
23
28
30
18
47
18
29
29
30
25
24
14
12
26
6
5
13
16
19
12
31
6
22
23
25
16
16
10
10
10
6
9
10
12
12
10
11
10
9
12
8
10
10
7
5
6
5
5
5
4
6
5
5
4
6
5
5
6
5
2
3
3
2
1
1
3
2
3
1
3
1
2
3
2
2
1
1
0
2
4
1
1
1
1
0
4
0
1
0
1
1
1
0
0
1
1
0
0
0
0
0
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
35
10
30
69
6
9
41
41
74
55
71
61
48
48
61
45
45
5
17
53
6
6
37
23
57
38
54
50
28
31
38
30
32
Days with max hour above
191 Mm -1
(a)
191
90
(b)
191
95
(c)
191
98
(d)
112
90
(e)
112
95
(f)
112
98
(p)
64
90
(h)
64
95
(i)
64
98
(s)
157
35
(t)
127
25
Percentage of days
4
10
10
1
8
10
11
11
10
11
10
9
12
8
10
9
4
5
6
1
5
5
6
5
5
4
4
6
6
5
6
5
3
3
3
1
1
1
2
3
2
2
3
1
2
3
2
2
1
2
1
1
4
1
1
2
1
0
4
0
1
0
1
1
1
0
0
1
1
0
1
1
0
0
1
0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
10
39
1
4
10
11
36
24
25
33
16
19
28
19
18
0
5
20
1
2
8
3
21
13
8
10
5
8
9
8
8
January 2010
4-19
DRAFT - Do Note Quote or Cite
-------
(B) NAAQS Scenarios Based on PM Light Extinction During All Daylight Hours*
NAAQS
Level Mm"1
NAAQS
Percentile
Form
Annual/24-
hour
Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake
City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Average
Days with max hour above
64 Mm -1
(,j)
191
90
(k)
191
95
(1)
191
98
(m)
112
90
(n)
112
95
(o)
112
98
(P)
64
90
(q)
64
95
(r)
64
98
(s)
157
35
(t)
127
25
Percentage of days
52
75
86
44
44
80
77
98
89
91
84
85
80
86
83
77
52
65
83
44
44
80
77
93
85
91
78
83
72
83
74
74
52
52
76
44
34
80
77
85
73
87
73
74
66
74
63
67
52
57
76
44
44
80
77
78
74
82
67
69
61
68
62
66
52
42
61
44
28
64
70
69
59
76
57
55
45
60
46
55
40
29
47
44
15
48
53
52
43
64
47
44
38
45
36
43
40
30
44
44
24
43
51
40
42
52
41
36
29
34
31
39
24
18
27
30
11
21
35
28
26
30
29
23
15
23
19
24
10
9
13
17
6
11
13
15
15
15
11
11
9
9
11
12
43
54
85
44
24
81
75
97
84
90
80
78
78
85
76
72
28
40
79
40
15
77
64
89
70
85
74
63
64
74
62
62
Days with max hour above
112 Mm -1
(j)
191
90
(k)
191
95
(1)
191
98
(m)
112
90
(n)
112
95
(o)
112
98
(p)
64
90
(q)
64
95
(r)
64
98
(s)
157
35
(t)
127
25
Percentage of days
22
52
73
6
17
41
44
76
65
75
64
57
50
64
59
51
22
37
58
6
17
41
44
62
56
73
55
53
43
58
42
44
22
26
44
6
14
41
44
48
40
62
45
40
34
41
33
36
22
30
42
6
17
41
44
39
42
48
40
35
28
33
30
33
22
16
28
6
11
21
34
28
25
30
26
22
14
24
18
22
10
8
13
6
6
10
13
15
14
13
9
11
8
8
10
10
10
9
10
6
9
9
12
8
14
5
5
8
4
5
7
8
4
3
2
5
5
3
8
3
7
1
4
1
2
1
3
3
1
1
1
3
1
1
1
2
3
0
4
0
0
0
1
1
10
30
69
6
9
41
41
74
55
71
61
48
48
61
45
45
5
17
53
6
6
37
23
57
38
54
50
28
31
38
30
32
Days with max hour above
191 Mm -1
Q)
191
90
(k)
191
95
(1)
191
98
(m)
112
90
(n)
112
95
(o)
112
98
(P)
64
90
(q)
64
95
(r)
64
98
(s)
157
35
(t)
127
25
Percentage of days
30
42
1
8
10
11
39
34
31
40
26
23
31
28
24
16
27
1
8
10
11
27
26
30
26
22
14
24
18
18
8
12
1
5
10
11
15
14
12
9
11
8
8
10
9
10
12
1
8
10
11
9
15
6
6
8
5
7
8
8
5
3
1
5
4
7
4
9
1
4
2
2
1
3
4
1
1
1
1
2
1
2
2
3
0
4
0
1
0
1
1
1
2
0
1
4
1
1
1
3
0
1
0
0
0
0
1
0
0
0
1
1
0
1
0
1
0
1
0
0
0
0
0
0
0
0
1
1
0
0
0
0
0
1
0
0
0
0
0
1
10
39
1
4
10
11
36
24
25
33
16
19
28
19
18
0
5
20
1
2
8
3
21
13
8
10
5
8
9
8
8
* Note that the table reports results based on daily maximum daylight hour, while the NAAQS scenarios in Panel B are based on all daylight
hours (in both cases excluding hours with RH>90%).
January 2010
4-20
DRAFT - Do Note Quote or Cite
-------
5 REFERENCES
Abt Associates Inc., 2001. Assessing Public Opinions on Visibility Impairment due to Air Pollution: Summary
Report. Prepared for EPA Office of Air Quality Planning and Standards; funded under EPA Contract No.
68-D-98-001. Abt Associates Inc., Bethesda, MD. January. Available:
http://www.epa.gov/ttncaaal/tl/reports/vis rpt final.pdf Accessed 9/16/2008.
Air Resource Specialists. 2003. 2003 WinHaze Air Quality Modeler, Version 2.9. Air Resource Specialists, Inc.
Available: http://www.air-resource.com/downloads.php. Accessed 9/16/2008.
Arizona Department of Environmental Quality. 2003. Recommendation for a Phoenix Area Visibility Index.
Visibility Index Oversight Committee. Available: http://www.phoenixvis.net/PDF/vis 031403final.pdf.
Accessed 9/16/2008. BBC Research & Consulting. 2002. Phoenix Area Visibility Survey. Draft Report.
Available: http://www.azdeq.gov/environ/air/download/vis 021903f.pdf. Accessed 9/16/2008.
BBC Research & Consulting. 2003. Phoenix Area Visibility Survey. Draft Report. Available:
http://www.azdeq.gov/environ/air/download/vis 021903f.pdf. Accessed 9/16/2008.
DeBell L. (2006). Spatial and seasonal patterns and temporal variability of haze and its constituents in the United States: Report
IV. 1 Cooperative Institute for Research in the Atmosphere. Colorado State University Fort Collins, CO. 156388
Ely, D.W., J.T. Leary, T.R. Stewart, and D.M. Ross. 1991. The Establishment of the Denver Visibility Standard.
Presented at the 84th Annual Meeting & Exhibition of the Air and Waste Management Association. June
16-21.
Frank NH. (2006). Retained nitrate, hydrated sulfates, and carbonaceous mass in federal reference method fine
paniculate matter for six eastern U.S. cities. J Air Waste Manag Assoc, 56: 500-11
Henderson, R. (2006) Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee to the
Honorable Stephen L. Johnson, Administrator, US EPA. Clean Air Scientific Advisory Committee
Recommendations Concerning the Proposed National Ambient Air Quality Standards for Paniculate
Matter. March 21, 2006. Available: http://www.epa.gov/sab/pdf/casac-ltr-06-002.pdf
Henderson, R. (2008) Letter from Dr. Rogene Henderson, Chair, Clean Air Scientific Advisory Committee to the
Honorable Stephen L. Johnson, Administrator, US EPA. Clean Air Scientific Advisory Committee
Consultation on EPA's Draft Integrated Review Plan for the National Ambient Air Quality Standards for
Paniculate Matter January 3, 2008.
Jacques Whitford AXYS. 2007. The View Ahead; Identifying Options for a Visibility Management Framework for
British Columbia. Report for the British Columbia Ministry of Environment. Available:
http://www.env.gov.bc.ca/air/airqualitv/pdfs/view ahead.pdf. Accessed 8/5/2009.
Molenar, J.V., W.C. Malm, and C.E. Johnson. 1994. Visual air quality simulation techniques. Atmospheric
Environment 28(5): 1055-1063.
NARSTO (2004) Paniculate Matter Science for Policy Makers: A NARSTO Assessment. P. McMurry, M.
Shepherd, and J. Vickery, eds. Cambridge University Press, Cambridge, England. ISBN 0 52 184287 5
Pitchford, M. and W. Malm. 1993. Development and applications of a standard visual index. Atmospheric
Environment 28(5): 1049-1054.
January 2010 5-1 Do Not Quote or Cite
-------
Pryor, S.C. 1996. Assessing public perception of visibility for standard setting exercises. Atmospheric Environment
30(15):2705-2716.
RWDI AIR. 2008. Final Report: Establishing a Visibility Goal for Wilderness and Urban areas in British Columbia
and Canada. Report to the British Columbia Ministry of Environment. Available:
http://www.env.gov.bc.ca/air/airqualitv/pdfs/visibilitv goal report_final.pdf. Accessed 8/5/2009.
Samet, J. (2009a) Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee (CAS AC)
to The Honorable Lisa P. Jackson, Administrator, U.S. EPA. Consultation on EPA's Particulate Matter
National Ambient Air Quality Standards: Scope and Methods Plan for Urban Visibility Impact Assessment.
May 21, 2009. Available: http://vosemite.epa.gov/sab/sabpeople.nsf/WebCommittees/CASAC.
Samet, J. (2009b) Letter from Dr. Jonathan M. Samet, Chair, Clean Air Scientific Advisory Committee (CASAC) to
The Honorable Lisa P. Jackson, Administrator, U.S. EPA. Review of the Particulate Matter Urban-
Focused Visibility Assessment (External Review Draft); November 24, 2009. - Available:
http://vosemite.epa.gov/sab/sabpeople.nsf/WebCommittees/CASAC.
Schmidt. M; Frank, N.; Mintz, D.; Rao, T.; McCluney, L. (2005). Analyses of paniculate matter (PM) data for the
PM NAAQS review. Memorandum to PM NAAQS review docket EPA-HQ-OAR-2001-0017. June 30,
2005. Available: http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_cr_td.html
Smith, A. 2009. Comments on the First External Review Draft of EPA's "Integrated Science Assessment for
Particulate Matter." CRA International, Washington, DC. March 30. Prepared for the Utility Air
Regulatory Group. Submitted as public comment to the public meeting EPA Clean Air Science Advisory
Council. April 2.
Smith, A.E. and S. Howell. 2009. An Assessment of the Robustness of Visual Air Quality Preference Study Results.
CRA International, Washington, DC. March 30. Prepared for the Utility Air Regulatory Group. Submitted
as supplemental material to presentation by Anne Smith to the public meeting of the EPA Clean Air
Science Advisory Council. April 2.
Statistics Canada. 2009a. Population and Dwelling Count Highlight Tables, 2006 Census. Available:
http://wwwl2.statcan.gc.ca/census-recensement/2006/dp-pd/hlt/97-550/Index.cfm?Page = INDX&LANG
= Eng. Accessed 7/13/2009.
Statistics Canada. 2009b. Population: Chilliwack, British Columbia (Census Agglomeration). Available:
http://wwwl2.statcan.ca/census-recensement/2006/dp-pd/prof/92-591/details/page.cfm7Lang = E&Geol =
CMA&Codel = 930_&Geo2 = PR&Code2 = 59&Data = Count&SearchText = Chilliwack&SearchType
= Begins&SearchPR = 01&B1 = Population&Custom. Accessed 7/13/2009.
Stratus Consulting Inc., 2009. Review of Urban Visibility Public Preference Studies: Final Report. Prepared for EPA
Office of Air Quality Planning and Standards; funded under EPA Contract No. EP-D-08-100 with Abt
Associates Inc., Bethesda, MD. September. Available:
US EPA (1996a). Air Quality Criteria for Particulate Matter. Research Triangle Park, NC: National Center for
Environmental Assessment-RTF Office; report no. EPA/600/P-95/001aF-cF. 3v. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm_pr cd.html.
US EPA (1996b). Review of the National Ambient Air Quality Standards for Particulate Matter: Policy Assessment
of Scientific and Technical Information, OAQPS Staff Paper. Research Triangle Park, NC 27711: Office
of Air Quality Planning and Standards; report no. EPA-452/R-96-013. Available:
http://www.epa.gOv/ttn/naaqs/standards/pm/s pm pr sp.html.
January 2010 5-2 Do Not Quote or Cite
-------
US EPA (2004). Air Quality Criteria for Paniculate Matter. National Center for Environmental Assessment, Office
of Research and Development, U.S. Environmental Protection Agency, Research Triangle Park, NC 27711;
report no. EPA/600/P-99/002aF and EPA/600/P-99/002bF. October 2004. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm cr cd.html
US EPA (2005). Review of the National Ambient Air Quality Standards for Paniculate Matter: Policy Assessment
of Scientific and Technical Information, OAQPS Staff Paper. Research Triangle Park, NC 27711: Office
of Air Quality Planning and Standards; report no. EPA EPA-452/R-05-005a. December 2005. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm cr sp.html
U.S. Environmental Protection Agency. (2007). Draft Integrated Review Plan for the National Ambient Air Quality
Standards for Paniculate Matter. October 2007. U.S. Environmental Protection Agency, Research
Triangle Park, NC, EPA 452/P-08-006. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm 2007_pd.html
US EPA (2008a). Integrated Review Plan for the National Ambient Air Quality Standards for Paniculate Matter.
National Center for Environmental Assessment and Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agency, Research Triangle Park, NC. Report No. EPA 452/R-08-004. March
2008. Available at:
http://www.epa.gov/ttn/naaqs/standards/pm/data/2008 03 final integrated review_plan.pdf
U.S. EPA(2008b). Integrated Science Assessment for Paniculate Matter: First External Review Draft. National
Center for Environmental Assessment-RTF Division, Office of Air Quality Planning and Standards,
Research Triangle Park, NC. EPA/600/R-08/139 and 139A. December 2008. Available:
http://www.epa.gOv/ttn/maqs/standards/pm/s jm_2007_isa.html.
US EPA (2009a). Integrated Science Assessment for Paniculate Matter: Final Report. National Center for
Environmental Assessment-RTF Division, Office of Air Quality Planning and Standards, Research
Triangle Park, NC. EPA/600/R-08/139F. December 2009. Available:
http://www.epa.gOv/ttn/maqs/standards/pm/s jm_2007_isa.html.
US EPA (2009b). Paniculate Matter National Ambient Air Quality Standards: Scope and Methods Plan for Urban
Visibility Assessment. Office of Air Quality Planning and Standards, U.S. Environmental Protection
Agency, Research Triangle Park, NC. EPA-452/P-09-001. February 2009. Available:
http://www.epa.gov/ttn/naaqs/standards/pm/s_pm 2007_pd.html.
US EPA (2009c). Paniculate Matter Urban-Focused Visibility Assessment: First External Review Draft. Office of
Air Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC.
EPA-452/P-09-005. September 2009. Available:
http://www.epa.gOv/ttn/naaqs/standards/pm/s pm 2007 pd.html.
January 2010 5-3 Do Not Quote or Cite
-------
APPENDICES
A. PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the Analysis of PM
light extinction in the 15 Study Areas
B. Distributions of Estimated PMi.s and Other Components under Current Conditions
C. Development of PRB Estimates of PMi.s components, PMio-2.5? and PM light extinction
D. Relationships between PM Mass Concentration and PM Light Extinction under Current
Conditions
E. Differences in Daily Patterns of Relative Humidity and Light Extinction between Areas
and Seasons
F. Distributions of Maximum Daily Daylight PM light extinction under "Just Meets"
Conditions
G. Additional Information on the Exclusion of Daylight Hours with Relative Humidity
Greater than 90 Percent
H. Inter- Year Variability
I. Daylight Hours
January 2010 Do Not Quote or Cite
-------
APPENDIX A - PM2 5 MONITORING SITES AND
MONITORS PROVIDING 2005-2007 DATA FOR THE
ANALYSIS OF PM LIGHT EXTINCTION IN THE 15 STUDY
AREAS
PM2.5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PM light extinction in the 15 Study Areas
Study Area
Tacoma
First PM2.S Monitoring Site
AQS ID 530530029
State: Washington
City: Tacoma
MSA: Tacoma, WA
Local Site Name: TACOMA - L
STREET
Address: 7802 SOUTH L
STREET, TACOMA
0.5 miles east of 1-5
2005-2007 annual DV = 10.2
2005-2007 24-hr DV = 43
This is the highest 24-hour PM25
DV site in the Seattle-Tacoma-
Olympia, WA annual PM2.5
nonattainment area
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM25 mass
(AQS parameter 881 01 ; one-in-
three sampling schedule)
• PM2.5 speciation (one-in-six
sampling schedule)
• 1-hourPM2.5 mass (AQS
parameter 88502, Acceptable
PM2.s AQI & Speciation Mass)
Correlated Radiance Research
M903 Nephelometry
No continuous PM10 monitoring at
this site, see right hand column..
Second PM2 s
Monitoring Site (if
applicable)
NA
PM10 data source for PM10.2.5
AOS ID 530530031
State: Washington
City: Tacoma
MSA: Tacoma, WA
Local Site Name: TACOMA - ALEXANDER
AVE
Address: 2301 ALEXANDER AVE, TACOMA,
WA
6.4 miles NNE of PM2.5 site
Neighborhood Scale
Parameters taken from this site:
• 1 -hour PM10 STP mass (AQS parameter
81102)
• Sample Collection Method:
INSTRUMENTAL-R&P SA246B-INLET
• Sample Analysis Method: TEOM-
GRAVIMETRIC
7% of PM10-2.5 values were determined using
regional average PM10-2.5: PM2.5 ratios from
2005 Staff Paper
January 2010
A-l
Do Not Quote or Cite
-------
PM2 5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PM light extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Fresno
AQS ID 060190008
State: California
City: Fresno
MSA: Fresno, CA
Local Site Name: None given
Address: 3425 N FIRSTS!,
FRESNO
2.5 miles west of the airport, 3
miles NNE of central Fresno
2005-2007 annual DV = 17.4
2005-2007 24-hr DV = 63
This is not the highest annual or
24-hr PM2 5 DV site in the San
Joaquin nonattainment area.
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM25 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.s speciation (one-in-
three sampling schedule)
• 1-hour PM2.s mass (AQS
parameter 88501, PM2 5 Raw
Data) Met-One BAM
No continuous PM10 monitoring at
this site, see right hand column..
NA
PMio-2.5 values were determined using
regional average PMio-2s: PM25 ratios from
2005 Staff Paper
January 2010
A-2
Do Not Quote or Cite
-------
PM2 5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PM light extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Los Angeles
AQS ID 060658001
State: California
City: Rubidoux (West Riverside)
MSA: Riverside-San Bernardino,
CA
Local Site Name: None given
Address: 5888 MISSION BLVD.,
RUBIDOUX
Eastern SCAB, 0.4 miles from
Pomona Freeway.
2005-2007 annual DV = 19.6
2005-2007 24-hr DV = 55
This site is not the highest DV site
in the LA-South Coast
nonattainment area.
Neighborhood scale.
Parameters taken from this site:
• 24-hour FRM PM25 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.s speciation (one-in-
three sampling schedule)
• 1-hourPM2.5 (AQS
parameter 88502, Acceptable
PM2s AQI & Speciation Mass)
R&P1400TEOM
No continuous PMio monitoring at
this site, see right hand column..
NA
AOS ID 060710306
State: California
City: Victorville
MSA: Riverside-San Bernardino, CA
Local Site Name: MOVED FROM 060710014
Address: 14306 PARKAVE., VICTORVILLE,
CA
36 miles north of PM25 site, on the other side
of a range of hills. 0.4 miles from 1-15
Measurement Scale not given in AQS, but
appears Neighborhood by aerial image.
Parameters taken from this site:
• 1-hour PMio STP mass (AQS parameter
81102)
• Sample Collection Method:
INSTRUMENTAL-R&P SA246B-INLET
• Sample Analysis Method: TEOM-
GRAVIMETRIC
6% of PMio-2.s values were determined using
regional average PM10-2.5: PM2.5 ratios from
2005 Staff Paper
January 2010
A-2
Do Not Quote or Cite
-------
PM2 5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PM light extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Phoenix
AQS ID 040137020 (FRM
&CSN)
State: Arizona
City: Scottsdale
MSA: Phoenix-Mesa, AZ
Local Site Name:
Address: 10844 EAST OSBORN
ROAD SCOTTSDALE' AZ
Reporting Agency: Salt River
Pima-Maricopa Indian Community
of Salt River Reservation
Eastern edge of the metro area,
largely surrounded by agricultural
fields.
2005-2007 annual DV = 7.9
2005-2007 24-hr DV= 15
This site is not the highest DV site
in the Phoenix-Mesa CBSA.
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM25 mass
(AQS parameter 88101; one-in-
six sampling schedule)
• PM2.s speciation (one-in-
three sampling schedule)
No continuous PM10 monitoring at
this site, see right hand column.
AQS ID 040139998
(Continuous)
State: Arizona
City: Phoenix
MSA: Phoenix-Mesa, AZ
Local Site Name: Vehicle
Emissions Laboratory
Address: 600 N 40th St &
Fillmore St
Measurement Scale not
available; 0.75 miles from
intersection of two
freeways, 1 mile from
Phoenix airport.
Parameters taken from this
site:
• 1-hour PM2.s mass.
Nephelometer.
AQS ID 040133002
State: Arizona
City: Phoenix
MSA: Phoenix-Mesa, AZ
Local Site Name: CENTRAL PHOENIX
Address: 1645 E ROOSEVELT ST-CENTRAL
PHOENIX STN
1.8 miles NE of central Phoenix
Neighborhood Scale
Parameters taken from this site:
• 1-hour PMio STP mass (AQS parameter
81102)
Sample Collection Method: INSTRUMENTAL-
R&PSA246B-INLET
Sample Analysis Method: TEOM-
GRAVIMETRIC
2% of PMio-2.5 values were determined using
regional average PMio-2s: PM2s ratios from
2005 Staff Paper
January 2010
A-4
Do Not Quote or Cite
-------
PM2 5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PM light extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Salt Lake City
AQS ID490353006
State: Utah
City: Salt Lake City
MSA: Salt Lake City-Ogden, UT
Local Site Name: UTM
COORDINATES = PROBE
LOCATION
Address: 1675 SOUTH 600 EAST,
SALT LAKE CITY
2.5 miles SSE of central Salt Lake
City
2005-2007 annual DV = 10.7
2005-2007 24-hr DV = 48
This is not the highest DV site in
the Salt Lake City CSA.
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM25 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.s speciation (one-in-
three sampling schedule)
• 1-hour PM2.s mass (AQS
parameter 88501, PM2.5 Raw
Data) FDMS-Gravimetric
No continuous PM10 monitoring at
this site, see right hand column.
NA
PM10-2.5 values were determined using
regional average PMio-2s: PM25 ratios from
2005 Staff Paper
Dallas
AQS ID 481130069
State: Texas
City: Dallas
MSA: Dallas, TX
Local Site Name: DALLAS
HINTON
Address: 1415 HINTON STREET
4.5 miles NE of central Dallas
2005-2007 annual DV = 11.5
2005-2007 24-hr DV = 25
This is not the highest DV site in
the Dallas-Ft. Worth CSA.
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM25 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.5 speciation (one-in-
three sampling schedule)
• 1-hourPM2.5 mass (AQS
parameter 88502, Acceptable
PM2.5 AQI & Speciation Mass)
TEOM Gravimetric 50 deg C
No continuous PM10 monitoring at
this site, see right hand column..
NA
PMio-2.5 values were determined using
regional average PMio-2s: PM25 ratios from
2005 Staff Paper
January 2010
A-5
Do Not Quote or Cite
-------
PM2 5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PM light extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Houston
AQS ID 482010024
State: Texas
City: Not in a city
MSA: Houston, TX
Local Site Name: HOUSTON
ALDINE
Address: 4510 1/2 ALDINE MAIL
RD
10 miles NNE of central Houston
2005-2007 annual DV= 13.1
2005-2007 24-hr DV = 25
This is not the highest DV site in
the 'Houston-Baytown-Huntsville,
TX CSA.
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM25 mass
(AQS parameter 88101; one-in-six
day sampling schedule)
• PM2.s speciation (one-in-six
sampling schedule)
• 1-hourPM2.s mass (AQS
parameter 88502, Acceptable
PM2.5 AQI & Speciation Mass)
TEOM Gravimetric 50 deg C
No continuous PM10 monitoring at
this site, see right hand column.
NA
PMio-2.5 values were determined using
regional average PMio-2s: PM25 ratios from
2005 Staff Paper
January 2010
A-6
Do Not Quote or Cite
-------
PM2 5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PM light extinction in the 15 Study Areas
Study Area
St. Louis
First PM2.5 Monitoring Site
AQS ID 2951 00085
State: Missouri
City: St. Louis
MSA: St, Louis, MO-IL
Local Site Name: BLAIR STREET
CATEGORY A CORE SLAM
PM2.5.
Address: BLAIRS
2 miles north of central St. Louis
2005-2007 annual DV = 14.5
2005-2007 24-hr DV = 34
This is not the highest DV site in
the St. Louis nonattainment area.
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM25 mass
(AQS parameter 881 01 ; every
day sampling schedule)
• PM2.s speciation (one-in-
three sampling schedule)
• 1-hourPM2.s mass (AQS
parameter 88502, Acceptable
PM2.5 AQI & Speciation Mass)
TEOM Gravimetric 30 deg C
No continuous PM10 monitoring at
this site, see right hand column.
Second PM2.5
Monitoring Site (if
applicable)
NA
PM10 data source for PM10.2.s
AQS ID 295100092 (2005 and 2006
data)
State: Missouri
City: St. Louis
MSA: St, Louis, MO-IL
Local Site Name:
Address: 3 NORTH MARKET
0.7 miles ESE of PM2.5 site, across the street
from the eastern edge of what appears to be
a recycling/municipal works yard.
Middle Scale
Parameters taken from this site:
• 1 -hour PMio STP mass (AQS parameter
81102)
• Sample Collection Method:
INSTRUMENTAL-R&P SA246B-INLET
• Sample Analysis Method: TEOM-
GRAVIMETRIC Site was on the other
(western) side of the recycling/municipal
works yard as site 295100093, below.
295100093 (2007 data)
^tfltp' MiQQni iri
Old 1C. IVIIooUUI 1
Pity ^t 1 ni I!Q
wily . OL. l_UUIo
MSA: St, Louis, MO-IL
Local Site Name: None given
Address: Branch Street
0.6 miles ESE of PM2 5 site, across the street
from the western edge of what appears to be
a recycling/municipal works yard.
Middle Scale
Parameters taken from this site:
• 1 -hour PM10 STP mass (AQS parameter
81102)
• Sample Collection Method:
INSTRUMENTAL-R&P SA246B-INLET
• Sample Analysis Method: TEOM-
GRAVIMETRIC
•
4% of PMio-2.s values were determined using
regional average PM10-2.5: PM2.5 ratios from
2005 Staff Paper
January 2010
A-7
Do Not Quote or Cite
-------
PM2 5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PM light extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Birmingham
AQS ID 010730023
State: Alabama
City: Birmingham
MSA: Birmingham, AL
Local Site Name:
Address: NO. B'HAM.SOU R.R.,
3009 28TH ST. NO
2.3 miles north of central
Birmingham
2005-2007 annual DV = 18.7
2005-2007 24-hr DV = 44
This is the highest DV site in the
Birmingham nonattainment area
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM25 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.s speciation (one-in-
three sampling schedule)
• 1-hour PM2.s mass (AQS
parameter 88502, Acceptable
PM2.5 AQI & Speciation Mass)
TEOM Gravimetric 50 deg C
• 1 -hour PM10 STP mass (AQS
parameter 81102)
Sample Collection Method:
INSTRUMENTAL-R&P SA246B-
INLET
Sample Analysis Method: TEOM-
GRAVIMETRIC
NA
Same as PM2.5 site.
0.3% of PMio-2.s values were determined
using regional average PM10.2.5: PM2.5 ratios
from 2005 Staff Paper
January 2010
A-8
Do Not Quote or Cite
-------
PM2 5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PM light extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Atlanta
AQSID 130890002
State: Georgia
City: Decatur
MSA: Atlanta, GA
Local Site Name: 2390-B
WILDCAT ROAD, DECATUR, GA
Address: SOUTH DEKALB
About 7 miles SE of central Atlanta
2005-2007 annual DV = 15.7
2005-2007 24-hr DV = 33
This is not the highest DV site in
the Atlanta nonattainment area.
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM25 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.5 speciation (one-in-
three sampling schedule)
• 1-hour PM2.s mass (AQS
parameter 88502, Acceptable
PM2.5 AQI & Speciation Mass)
TEOM Gravimetric 30 deg C
No continuous PM10 monitoring at
this site, see right hand column.
NA
AQS ID 131210048
State: Georgia
City: Atlanta
MSA: Atlanta, GA
Local Site Name: Georgia Tech, Ford
Environmental Science and Technology Bldg,
roof
Address: GA. TECH., Ford ES&T Bldg, 311
Ferst St NW, Atlanta GA
8.6 miles NW of PM2.5 site
Neighborhood Scale
Parameters taken from this site:
• 1-hour PMio STP mass (AQS parameter
81102)
Sample Collection Method: INSTRUMENT
MET ONE 4 MODELS
Sample Analysis Method: BETA
ATTENUATION
8% of PM10.2.5 values were determined using
regional average PMio-2s: PM25 ratios from
2005 Staff Paper
January 2010
A-9
Do Not Quote or Cite
-------
PM2 5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PM light extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Detroit
AQS ID 261630033
State: Michigan
City: Dearborn
MSA: Detroit, Ml
Local Site Name: PROPERTY
OWNED BY DEARBORN PUBLIC
SCHOOLS
Address: 2842 WYOMING
About 0.2 miles from Ford River
Rouge auto plant
2005-2007 annual DV = 17.2
2005-2007 24-hr DV = 43
This is the highest annual and 24-
hr DV site in the Detroit
nonattainment area
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM25 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.s speciation (one-in-six
sampling schedule)
• 1-hourPM2.s mass (AQS
parameter 88501, PM2.5 Raw
Data) TEOM Gravimetric 50 deg C
• 1 -hour PM10 STP mass (AQS
parameter 81102)
Sample Collection Method:
INSTRUMENTAL-R&P SA246B-
INLET
Sample Analysis Method: TEOM-
GRAVIMETRIC
NA
Same as PM2.5 site.
2% of PMio-2.s values were determined using
regional average PM10-2.5: PM2.5 ratios from
2005 Staff Paper
January 2010
A-10
Do Not Quote or Cite
-------
PM2 5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PM light extinction in the 15 Study Areas
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
Pittsburgh
AQS ID 420030008
State: Pennsylvania
City: Pittsburgh
MSA: Pittsburgh, PA
Local Site Name: None given
Address: BAPC 301 39TH
STREET BLDG #7
3 miles NE of central Pittsburgh,
0.5 miles from Allegheny River
2005-2007 annual DV = 15.0
2005-2007 24-hr DV = 40
This site is not the highest DV site
in the Pittsburgh nonattainment
area.
Urban Scale
Parameters taken from this site:
• 24-hour FRM PM25 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.s speciation (one-in-
three sampling schedule)
• 1-hour PM2.s mass (AQS
parameter 88502, Acceptable
PM2.5 AQI & Speciation Mass)
TEOM Gravimetric 50 deg C
No continuous PM10 monitoring at
this site, see right hand column.
NA
PMio-2.5 values were determined using
regional average PMio-2s: PM25 ratios from
2005 Staff Paper
Baltimore
AQS ID 240053001 (FRM
&CSN)
State: Maryland
City: Essex
MSA: Baltimore, MD
Local Site Name: Essex
Address: 600 Dorsey Avenue
7 miles east of central Baltimore
2005-2007 annual DV = 14.5
2005-2007 24-hr DV = 35
This is not the highest DV site in
the Baltimore nonattainment area.
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM25 mass
(AQS parameter 88101; every
day sampling schedule)
• PM2.5 speciation (one-in-
three sampling schedule)
• 1 -hour PM10 LC mass (AQS
parameter 85101
AQS ID 245100040
(Continuous)
State: Maryland
City: Baltimore
MSA: Baltimore, MD
Local Site Name: Oldtown
Address: Oldtown Fire
Station, 1100 Hillen Street
1 mile NNE of Inner Harbor
Middle Scale
Parameters taken from this
site:
• 1-hour PM2.5 mass
(AQS parameter 88502,
Acceptable PM25 AQI &
Speciation Mass) TEOM
Gravimetric 50 deg C
Same as PM2.5 site.
5% of PMio-2.s values were determined using
regional average PM10-2.5: PM2.5 ratios from
2005 Staff Paper
January 2010
A-ll
Do Not Quote or Cite
-------
PM2 5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PM light extinction in the 15 Study Areas
Study Area
Philadelphia
New York
First PM2.5 Monitoring Site
AQSID1 00032004 (DE)
State: Delaware
City: Wilmington
MSA: Wilmington-Newark, DE-MD
Local Site Name: CORNER OF
MLK BLVD AND JUSTISON ST
2.5 miles NE of central
Wilmington, 0.25 miles from the
Delaware River, 22 miles SWfrom
central Philadelphia
2005-2007 annual DV = 14.7
2005-2007 24-hr DV = 37
This is not the highest DV site in
the Philadelphia nonattainment
area
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM25 mass
(AQS parameter 881 01 ; every
day sampling schedule)
• PM2.s speciation (one-in-six
sampling schedule)
• 1 -hour PM2.s mass (AQS
parameter 88501 , PM2.5 Raw
Data) Beta Attenuation
• 1 -hour PM10 STP mass (AQS
parameter 81 102)
AQS ID 340390004 (NJ)
State: New Jersey
City: Elizabeth
MSA: Newark, NJ
Local Site Name: ELIZABETH
LAB
Address: NEW JERSEY
TURNPIKE INTERCHANGE 13
1 .75 miles south of Elizabeth, at
the I-95 interchange with I-278
2005-2007 annual DV = 14.4
2005-2007 24-hr DV = 42
This is not the highest DV site in
the New York nonattainment area
Neighborhood Scale
Parameters taken from this site:
• 24-hour FRM PM25 mass
(AQS parameter 881 01 ; every
day sampling schedule)
• PM2.5 speciation (one-in-
three sampling schedule)
• 1-hourPM2.5 mass (AQS
parameter 88502, Acceptable
PM2.5 AQI & Speciation Mass)
TEOM Gravimetric 30 deg C
No continuous PMio monitoring at
this site, see right hand column.
Second PM2.5
Monitoring Site (if
applicable)
NA
NA
PM10 data source for PM10.2.s
Same as PM2.5 site.
3% of PMio-2.s values were determined using
regional average PM10-2.5: PM2.5 ratios from
2005 Staff Paper
AQS ID 36061 01 25
State: New York
City: New York
MSA: New York, NY
Local Site Name: PARK ROW
Address: 1 PACE PLAZA
Near the on-ramp to the Brooklyn Bridge,
Manhattan end
Measurement scale not stated.
Parameters taken from this site:
• 1 -hour PMio STP mass (AQS parameter
81102)
Sample Collection Method: INSTRUMENTAL-
R&PSA246B-INLET
Sample Analysis Method: TEOM-
GRAVIMETRIC
2% of PMio-2.s values were determined using
regional average PMio-2s: PM25 ratios from
2005 Staff Paper
January 2010
A-12
Do Not Quote or Cite
-------
Study Area
First PM2.5 Monitoring Site
Second PM2.5
Monitoring Site (if
applicable)
PM10 data source for PM10.2.s
PM2 5 Monitoring Sites and Monitors Providing 2005-2007 Data for the
Analysis of PM light extinction in the 15 Study Areas
Notes:
In this Table, the 1-hour concentration parameter "88502, Acceptable PM2 5 AQI & Speciation Mass" is the
same as the ISA refers to as "FRM-like" PM25 mass. An entry of "88501, PM25Raw Data" indicates that
the monitoring agency makes no representation as to the degree of correlation with FRMPM2 5 mass. The
latter type of continuous PM2 5 data were used only when the former were unavailable.
Where PM10 was reported in STP, it was converted to LC before PM10.2 5 was calculated.
All continuous PM2 5 data were obtained through the AirNow data system rather than from AQS, as an
initial exploration indicated that not all the desired 1-hour data from all sites had been submitted to AQS.
Data are submitted to the AirNow system within hours of collection and may not be subject to as much data
validation review as is typical for data in AQS, despite the opportunity offered by the AirNow system for
monitoring agencies to correct data after initial submission.
January 2010
A-13
Do Not Quote or Cite
-------
APPENDIX B - DISTRIBUTIONS OF ESTIMATED PM2 5
AND OTHER COMPONENTS
January 2010 B-l
DRAFT Do Not Quote or Cite
-------
Figure B-l - Distribution of daily maximum PM2 5 and PMio-2.5 across the 2005-2007 period, by study area
(a) Daily maximum daylight PM2.5
Daily Maximum PM2.5 (Daylight Hours)
o
108 304 288
-9-
300 263 143 274 346 276 131 268 187 142 219
-r -r T T
-r
January 2010
DRAFT Do Not Quote or Cite
B-2
-------
(b) Daily maximum daylight PMio-2.5
Daily Maximum Coarse (Daylight Hours)
o
304 288
300 263
274 346 276 131 268 187 142 219
^ *?'
i r
January 2010
DRAFT Do Not Quote or Cite
B-3
-------
Figure B-2 - Distribution of hourly PM2.5 components across the 2005-2007 period, by study area
(a) 1 -hour daylight sulfate (dry, fully neutralized)
Sulfate hourly (Daylight Hours)
o
1238 3643
3457 3106 1652 3273 3930 3262 1567 3179 2095 161S
^
V
&
-------
Figure B-2 - Distribution of hourly PM2.5 components across the 2005-2007 period, by study area, continued
(b) 1-hour daylight nitrate (dry, fully neutralized, CSN method consistent)
Nitrate hourly (Daylight Hours)
H)
o
c
o
O
1238 3643
0
3383
0
3457 3106 1652 3273 3930 3262 1567 3179 2095
J-
,v- ^
X X X /
January 2010
DRAFT Do Not Quote or Cite
B-5
-------
Figure B-2 - Distribution of hourly PM2.5 components across the 2005-2007 period, by study area, continued
(c) 1-hour daylight elemental carbon
Elemental Carbon hourly (Daylight Hours)
o
1238 3643 3383
3457 3106 1652 3273 3930 3262 1567 3179 2095 161S 2515
V
•£>' 4
o* *•*
"
^r -ft.-" N
•S' / ><*"
y X ^
January 2010
DRAFT Do Not Quote or Cite
B-6
-------
Figure B-2 - Distribution of hourly PM2.5 components across the 2005-2007 period, by study area, continued
(d) 1-hour daylight organic carbonaceous material (by SANDWICH method)
Organic Carbon hourly (Daylight Hours)
H)
o
c
o
O
3930
O
,v- ^
os 4?
-------
Figure B-2 - Distribution of hourly PM2.5 components across the 2005-2007 period, by study area, continued
(e) 1-hour daylight fine soil
Soil hourly (Daylight Hours)
o
1238 3643 3383 988 3457 3106 1652 3273 3930 3262 1567 3179 2095 1618 2515
8
I
I
I
&
s3>-
January 2010
DRAFT Do Not Quote or Cite
B-8
-------
APPENDIX C - DEVELOPMENT OF PRB ESTIMATES OF
PM2 5 COMPONENTS, PM10_2 5, AND PM LIGHT
EXTINCTION
Policy relevant background levels of PM light extinction have been estimated for this
assessment by relying on outputs for the 2004 CMAQ run in which anthropogenic emission
in the U.S., Canada, and Mexico were omitted, as described in the second draft ISA.
Estimates of PRB for PM light extinction were calculated from modeled concentrations of
PM2.5 components using the IMPROVE algorithm. The necessary component concentrations
were extracted from the CMAQ output files, as they were not summarized in the second draft
ISA.
More specifically, for each study area, EPA staff overlaid CMAQ grid cells over
shapes representing the Census-defined urbanized area for each study area, and visually
identified the CMAQ grid cells that had a substantial portion of their area coincident with the
urbanized area. For each such grid cell, for each of the 12 months of the year, we obtained
the 24 values of the hour-specific average concentrations of the five PM2.5 components. We
then averaged these across the selected grid cells. Thus, a given hour of the day has the same
PRB estimate for a component on all days within a month, but months and study areas differ.
We generally observed that PRB concentrations did not vary greatly across the several grid
cells overlaying the urbanized area of a given study area; this is reasonable given the
exclusion of local anthropogenic sources from this CMAQ model run. CMAQ estimates of
PRB for the five PM2.5 components averaged across grid cells and months were not adjusted
46
in any.
There too many values of PRB to present or illustrate them comprehensively in this
document. Table C-l presents annual average concentrations by study area to summarize
these PRB estimates for the PM2.5 components (including the specific form assumed for
sulfate, nitrate, and organic carbon). The right hand column of the table shows the PM2.5
mass calculated from the CMAQ-estimated components, including factors to fully neutralize
sulfate and nitrate (but with no water mass added). One notable feature of the annual average
of the PRB estimates is the relatively high values for elemental and organic carbon PRB for
the Tacoma study area. This area is often affected by wildfires for extended periods in the
46 This approach to estimation of PRB for PM2.5 shares the same information source but is more
disaggregated than the approach used in the health risk assessment for this review of the PM NAAQS. In the health
risk assessment, PRB estimates for PM2.5 mass concentration are taken from the same CMAQ model run, but is
averaged by calendar quarter and by region of the country.
January 2010 C-l
DRAFT - Do Not Quote or Cite
-------
autumn months, and such fires were included in the 2004 emissions scenario for the PRB
CMAQ run. A cursory review of information on fire events in 2005-2007 confirmed that the
fire situation in this part of the country in 2004 was not an anomaly.
Another notable feature of the PRB estimates is that the values for nitrate and fine
soil/crustal are low relative to previous estimates of natural background concentrations of
these fine PM components in Class I areas. These previous estimates by Trijonis, repeated in
the 2003 EPA guidance document "Guidance for Estimating Natural Visibility Conditions
Under the Regional Haze Program", are 0.10 ug/m3 for (neutralized) nitrate and 0.50 ug/m3
for fine soil. These estimates are based largely on data from the earliest of the IMPROVE
monitoring stations, and thus may include some influence from non-PRB emissions. On the
other hand, it is understandable that the unadjusted output from the PRB CMAQ scenario
would underestimate nitrate and fine soil. CMAQ is known to underestimate actual nitrate in
many situations when provided with a complete NOx inventory, and the nonanthropogenic
emission inventory for NOx itself has uncertainties. The non-anthropogenic emission
inventory for the PRB CMAQ run may also quite easily underestimate nonanthropogenic
emissions of fine soil. However, even if the estimates for PRB nitrate and fine soil were
increased to match the Trijonis estimates, the resulting values for PRB PM light extinction
would increase only a little. Even at 90 percent relative humidity, the contribution to PM
light extinction calculated from the Trijonis estimates is 1.7 Mm"1, versus the average of
about 0.5 Mm"1 using the estimates in Table C-l. The increment of 1.2 Mm"1 would be only
about 10 to 20 percent of the PRB PM light extinction estimates shown in Table C-4, and
would not significantly affect the calculation of PM light extinction values under the "what
if scenarios.
January 2010 C-2
DRAFT - Do Not Quote or Cite
-------
Table C-l. Summary of PRB estimates for the five PMi.5 components: average 1-hour
values across 2005-2007
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit-Ann
Pittsburgh
Baltimore
Philadelphia
New York City
Average
Average 1-Hour PRB Concentration Across 2005-2007 (ug/mj)
Sulfate
(dry, no
ammonium)
0.45
0.4
0.36
0.31
0.25
0.27
0.3
0.31
0.29
0.3
0.34
0.3
0.34
0.34
0.36
0.33
Nitrate
(dry, no
ammonium)
0.026
0.00062
0.0037
0.000052
0.00028
0.0022
0.0055
0.0027
0.007
0.016
0.00062
0.00052
0.0016
0.00097
0.0038
0.00
Elemental
Carbon
0.15
0.08
0.028
0.02
0.025
0.055
0.091
0.047
0.099
0.1
0.024
0.029
0.039
0.03
0.026
0.06
Organic
Carbonaceous
Material
1.3
0.74
0.3
0.26
0.26
0.59
0.86
0.53
1.1
1.1
0.32
0.36
0.44
0.36
0.31
0.59
Fine
Soil/Crustal
0.31
0.19
0.036
0.015
0.034
0.092
0.17
0.07
0.19
0.19
0.018
0.034
0.054
0.032
0.022
0.10
Calculated
PM25
2.4
1.6
0.9
0.7
0.7
1.1
1.5
1.1
1.8
1.8
0.8
0.8
1.0
0.9
0.9
1.20
January 2010
DRAFT - Do Not Quote or Cite
C-3
-------
It is also necessary to have estimates of PRB for PMio-2.5, to feed into the IMPROVE
algorithm. It is not EPA's practice to rely on coarse PM estimates from CMAQ modeling, so
other sources of PRB estimates were considered. The final ISA for this review does not
present any new information on this subject. The approach used in the previous two Criteria
Documents was to present the historical range of annual means of PMio-2.5 concentrations
from IMPROVE monitoring sites selected as being least influenced by anthropogenic
emissions. See Table 3E-1 of the 2004 Criteria Document (reproduced here as Table C-3).
For sites in the lower 48 states, these annual means ranged from a low of 1.8 ug/m3 to a high
of 10.8 ug/m3. No cross-year average or median values were provided that could be used as
the point estimates needed in this assessment. Therefore, for this assessment, EPA staff
estimated PRB for PMio-2.5 using a contour map based on average 2000-2004 PMio-2.5
concentrations from all IMPROVE monitoring sites, found in a recent report from the
IMPROVE program. (DeBell, 2006). We located each study area's position on this map,
and assigned it the mid-point of the range of concentrations indicated by the contour band for
that location. The contour map is reproduced here as Figure C-l. Stars show locations of the
15 study areas. In this reproduction, the midpoints of the contour ranges have been added to
the legend.
The results for PRB for coarse PM are shown in Table C-2. Lacking any other
information, these PRB values are taken to apply to every hour of the year. The contour map
and thus these values are influenced by data from IMPROVE sites that were not considered
in the 2004 Criteria Document because they are not sufficiently isolated from the influence of
anthropogenic emissions, including three IMPROVE sites in urban areas which clearly are
influenced by anthropogenic emissions, and thus may be overestimates of PRB for coarse
PM. Nevertheless, these values are generally within the range of values presented in the
Criteria Document for the more isolated sites. These values for the more isolated sites are
reproduced here in Table C-3 for ease of comparison. Further, these PRB values are low
enough that their exact values have little effect on the results of "what if estimation of PM
light extinction levels under possible secondary PM NAAQS.
Table C-4 presents the resulting 2005-2007 average PRB daylight PM light extinction
by study area, determined by using each daylight hour's f(RH),47 the hour-specific PRB
PM2.5 component estimates (summarized only as annual averages in Table C-l), the PRB
PMio-2.5 estimates in Table C-2, and the IMPROVE algorithm. The sulfate and nitrate
component values in Table C-l are multiplied by 1.375 and 1.29 to reflect full neutralization,
47 Hour-specific relative humidity for PRB conditions was assumed to be the same as measured for current
conditions.
January 2010 C-4
DRAFT - Do Not Quote or Cite
-------
before being used in the IMPROVE algorithm. While for conciseness Table C-4 presents
only the annual average PRM for PM light extinction for all daylight hours in 2005-2007
(excluding hours with relative humidity greater than 90 percent), in the rollback analysis of
"what if conditions hour-specific PRB values are retained and used.
The values of PRB PM light extinction in Table C-4 range between 5 and 11 Mm"1.
For comparison, the default estimates of natural visibility conditions in the 2003 EPA
guidance document for Class I areas range between about 15 and 20 Mm"1, including the
Rayleigh contribution of about 10 Mm"1. Thus, on an annual average basis the range of PRB
estimates for PM light extinction used for this assessment is very consistent with the range of
total light extinction values recommended in the guidance document.
January 2010 C-5
DRAFT - Do Not Quote or Cite
-------
Figure C-l. Selection of PRB values for PMi0-2.5 based on contoured IMPROVE
monitoring data
e
Puerto Rico /
Virgin Islands
January 2010
DRAFT - Do Not Quote or Cite
C-6
-------
Table C-2. Policy Relevant Background Concentrations of PMio-i.s Used in This
Assessment, Based on Measurements at IMPROVE Sites
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
PRB PM10_2.5Mass (ug/m3)
4.5
5.5
4.5
5.5
4.5
8.5
5.5
7.5
5.5
5.5
9.5
3.5
3.5
6.5
3.5
Table C-3. Ranges of 1990-2002 Annual Mean PM Concentrations at IMPROVE
Monitoring Sites (jiig/m3)
PM,, PMlt
Site
Acadia National Park. ME
Big Bend National Park, TX
Boundary Waters Canoe Area. MN
Bryce Canyon National Park. UT
Bndger Wilderness. WY
Cauyonlands National Park. UT
Denali National Park, AK
Gila Wilderness, NM
Glacier National Park, MT
La&sen Volcanic National Park. CA
Lone Peak Wilderness. UT
Lye Brook Wilderness, VT
Redwood National Park, CA
Three Sisters Wilderness. OR
Voyageurs National Park 1 , MN
Voyageurs National Park 2. MN
Yellowstone National Park 1, WY
Yellowstone National Park 2. WY
Noiisulfate
2.6-4.7
2.7-4.9
2.6-3.9
1.7-2.4
1.5-2.2
1.9-3.2
0.7-2.4
2.4-3.4
3.8-5.5
1.7-4.5
3.1-5.3
2.3-4.8
2.8-4.6
2.0-5.4
3.2-3.5
2.6-5.4
2.0-3.0
1.7-4.1
(Total)
(4.9-8.2)
(5.0-7.8)
(4.4-5.8)
(2.6-3.4)
(2.1-2.9)
(2.8-4.0)
(1.1-3.2)
(3.4-4.5)
(4.8-6.5)
(2.1-5.1)
(4.1-6.9)
(4.5-8.8)
(3.6-5.4)
(2.7-6.5)
(5.1-5.9)
(4.1-7.2)
(2.6-3.6)
(2.3-4.7)
Xonsulfate
4.6-11.3
8.8-1S.7
5.0-10.2
4.4-7.6
3.7-6.5
5.1-10.5
2.0-7.5
4.9-7.9
7.6-14.2
4.0-8.1
7.1-10.9
4.2-9.7
6.0-10.6
4.0-8.1
5.7-11.2
5.2-10.8
6.0-9.2
3.6-9.0
(Total)
(7.3-15.0)
(11.3-18.6)
(7.0-12.0)
(5.3-8.5)
(4.3-7.3)
(6.3-11.7)
(2.4-8.3)
(6.0-9.2)
(8.5-15.2)
(4.6-8.5)
(8.1-12.5)
(7.0-13.6)
(7.2-11.7)
(4.6-9.1)
(8.1-13.1)
(7.0-12.5)
(6.6-9.9)
(4.2-9.6)
Coarse PM
1.8-6.0
5.6-10.8
2.3-7.3
2.5-5.6
1.9-4.7
3.2-8.0
1.1-5.6
2.5-5.0
3.7-9.6
1.8-6.4
3.7-6.0
1.6-4.8
3.3-6.5
1.9-4.4
2.S-7.8
2.6-5.3
3.8-7.0
1.9-5.0
Source: Table 3E-1 of the 2004 Air Quality Criteria Document for PM
January 2010
DRAFT - Do Not Quote or Cite
C-7
-------
Table C-4. 2005-2007 Average Policy Relevant Background Daylight PM light extinction
(excluding hours with relative humidity above 90%)
Study Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
2005-2007 Average Policy Relevant
Background Daylight PM light extinction,
Mm1
11
11
9
8
5
8
10
9
9
10
7
7
8
8
8
January 2010
DRAFT - Do Not Quote or Cite
-------
APPENDIX D - RELATIONSHIPS BETWEEN PM MASS
CONCENTRATION AND PM LIGHT EXTINCTION UNDER
CURRENT CONDITIONS
In the last review, the 2005 Staff Paper examined the correlation between PM light
extinction and PM2.5 mass concentrations, each defined for various consistent time periods. The
2005 Staff Paper analysis assumed that the percentage mix of PM2.5 components was the same
in all 24 hours of each day, equal to that indicated by 24-hour CSN sampling. The modeling of
1-hour PM light extinction in this new assessment allows these correlations to be re-examined,
with the more realistic treatment in which the mix of PM2.5 components is modeled to vary
during the day, based in part on diurnal profiles from CMAQ modeling (see section 3.2.2).
Five scatter plot figures relating PM2.s mass concentrations and PM light extinction are
presented here for the individual study areas, using different time periods for the two parameters;
these time periods are not always matched. In each figure, the solid red curve was estimated by
applying locally weighted scatter plot smoothing (LOESS) to the data. LOESS is a form of
locally weighted polynomial regression (see http://support.sas.com/rnd/app/papers/loesssugi.pdf)
and is a convenient way to visualize whether a dense data cloud in a scatter plot reflects a more
linear or more nonlinear relationship. The LOESS results in each case indicate a generally linear
relationship as a central tendency but with considerable variability around that central tendency.
Table D-l presents squared correlation coefficients between observed and LOESS model-
predicted values for all five figures. Because the LOESS regressions are generally linear,
comparisons among these correlation coefficients should lead to the same qualitative conclusions
as if coefficients from linear regressions were compared. All values of PM light extinction
presented here are based on excluding daylight hours with relative humidity greater than 90
percent; hence, a nominally 4-hour period might have as few a one 1-hour PM light extinction
value, although this is rare in this data set (see the tile plots in Figure 3-12). However, values of
PM2.5 mass concentration do not exclude any hours within the time period specified. Note that
if several study areas were grouped by region and combined into a single scatter plot and LOESS
fit, similar to the analysis of this topic in the 2005 Staff Paper, the correlations would be weaker
than observed here for individual study areas.
Figure D-l compares 24-hour PM2.5 mass (as measured by the FRM/FEM filter-based
sampler) to daily maximum daylight PM light extinction. The scatter is due the variations in
PM2.s concentration, in the mix of PM2.s components, and in relative humidity during the day
and across days. Variations in PMio-2.5 concentrations also contribute to the scatter, in all five
comparisons presented here, since very high levels of PMio-2.5 substantially influence PM light
January 2010 D-l
DRAFT - Do Not Quote or Cite
-------
extinction. This source of variability in the scatter plots is particularly important for Los
Angeles, Phoenix, and St. Louis which have many (Phoenix) or some (Los Angeles and St.
Louis) hours with high PMio-2.5.
Because of the large scatter and low correlation coefficients when using 24-hour PM2.5
mass concentration to predict daily maximum daylight PM light extinction, it is natural to
investigate how much the correlation improves when the PM2.5 mass indicator is limited to
shorter periods of time. The next four figures investigate correlations during such shorter
periods, both matched and un-matched in time.
Figure D-2 compares hourly PM2.5 mass (as actually measured by the continuous
instruments) vs. same-hour daylight PM light extinction. Lack of agreement due to mismatch of
time period is not a factor in this comparison. However, there is still considerable scatter due to
variations in the mix of PM2.5 components and in relative humidity across hours and days. In
addition, continuous PM2.5 mass instruments do not register the mass of each component
consistently with FRM/FEM and CSN samplers and lab analysis methods. This affects the
scatter in this figure because the estimates of hourly light extinction are linked to the FRM/FEM
and CSN measurements more strongly than to the continuous PM2.s measurements. Note that
the correlation values in Table D-l for this comparison are better than those for the 24-hour
comparison in most but not all study areas. An implication of this figure and the information in
Table D-l is that a wide range of light extinction levels can prevail in hours that have the same
PM2.5 mass concentration, even at a single site. Additional variability no doubt exists across
areas.
Figure D-3 compares 12-4 pm average PM2.5 mass vs. 12-4 pm average PM light
extinction. The 2005 Staff Paper observed that because this time period is generally the time of
lowest relative humidity, the relationship between PM2.5 mass and PM light extinction (i.e., the
ratio of the two or the slope of the regression line) is more uniform across areas during this
period than the relationship for values of each averaged over all 24 hours in a day. In addition,
the longer averaging period might be expected to reduce the effect of variability in the
measurement of hourly PM2.5 mass. However, comparison of Figures D-2 (time-matched single
hours) and D-3 (time-matched 4 afternoon hours) and the corresponding columns of Table D-l
indicates that after exclusion of hours with relatively humidity greater than 90 percent the scatter
in Figure D-3 is about the same as in Figure D-2. This residual scatter is due to composition
differences from hour-to-hour, as well as to variations in relative humidity during hours with
relative humidity of 90 percent or less. It can also be observed by comparing Figures D-2 and D-
3 that the period between 12 pm and 4 pm generally has lower levels of PM light extinction than
for all daylight hours taken together, even after the exclusion of the hours with the highest
relative humidity. (Note the change in scale between these two figures.)
January 2010 D-2
DRAFT - Do Not Quote or Cite
-------
Figure D-4 compares 12-4 pm average PM2.5 mass vs. daily maximum daylight 1-hour
PM light extinction. This time-unmatched comparison tests the usefulness of a 12-4 pm PM2.5
mass indicator as a predictor of the daily PM light extinction metric of potentially greatest
interest. The scatter in Figure D-4 is typically more than in Figure D-3 (4 time-matched
afternoon hours), because daily maximum daylight 1-hour PM light extinction often occurs
earlier in the day than the 12-4 pm period used to average the PM2.5 mass, and the time period
mismatch introduces prediction errors due to changes in PM2.5 concentration and composition
and relative humidity. An implication is that while a secondary NAAQS based on 12-4 pm
average PM2.5 mass might achieve a given level of protection across days and areas in avoiding
high levels of PM light extinction between 12 and 4 pm, with some variation across areas due to
composition and relative humidity differences, there could be considerable additional variation in
the level of protection against PM light extinction during the earlier hours of the day when some
areas often have their highest PM light extinction levels.
Figure D-5 compares 8 am-12 pm average PM2.5 mass vs. daily maximum daylight 1-
hour PM light extinction. This comparison is of interest because it may reduce the number of
instances of time mismatch, versus the comparison made in Figure D-4, if the daily maximum
light extinction often occurs between 8 am and 12 pm. The scatter in Figure D-5 is typically less
than in Figure D-4 and the squared correlation coefficients larger, indicating that this earlier
averaging period for PM2.5 mass more often encompasses the period of maximum PM light
extinction. However, the scatter in Figure D-5 is greater than that in Figure D-3 (4 time-matched
afternoon hours).
Figure D-6 provides another perspective on the possible use of PM2.5 mass concentration
as an indicator for a secondary PM NAAQS aimed at protecting visual air quality. Figure D-6
shows in box-and-whisker plot form two versions of the ratios of PM light extinction to PM2.5
mass concentration, allowing a comparison across the 15 study areas of the central tendencies
and the distributions of these ratios. The Panel A version corresponds to the comparison in
Figure D-l (24-hour averages of PM2.5 mass and PM light extinction) and the Panel B version
corresponds to the comparison in Figure D-2 (time-matched single hour values). The data points
in Figure D-6 were prepared as follows. In each day for each study area, the value of the
indicated PM light extinction (24-hour average or 1-hour value) was divided by the indicated
PM2.5 concentration metric (24-hour average or 1-hour value). Ratios that reflect PM2.5
concentrations less than 5 |ig/m3 or PM light extinction less than 64 Mm"1 were eliminated
before plotting, as such data points represent days or hours that could not play any role in
determining compliance with any of the NAAQS scenarios considered in this assessment; also,
some of these low-concentration/extinction data pairs produced extreme ratios that obscured the
pattern for data pairs of most policy interest. The maximum ratio value for the vertical scale in
January 2010 D-3
DRAFT - Do Not Quote or Cite
-------
these plots is set at 40 to allow closer examination of the portion of the plot representing the bulk
of the data; this prevents a very small number of daily maximum data points for a few study
areas from appearing in Panel A and a very small percentage of 1-hour data points for a few
study areas (Los Angeles and St. Louis in particular) from appearing in Panel B. The notable
variation in the vertical positions of the 25-75 percentile boxes and the 90 percentile whiskers
representing the ratios in the 15 areas illustrates the point that because of differences in PM
composition mix and relative humidity (even after excluding hours with relative humidity greater
than 90 percent) across study areas, a secondary NAAQS based on PM2.5 mass concentration
would not give equal protection in terms of PM light extinction levels across cities, days, and
hours.
In the first public review draft of this assessment, it was notable that the correlation
values for St. Louis and Philadelphia were much lower than for other areas. In this version
(reflecting both corrections to relative humidity inputs and exclusion of hours with very high
relative humidity) the correlation value for Philadelphia is about that for other eastern areas. The
correlation values for St. Louis remain notably low relative to the average of all areas, for all five
scatter plots. This is likely due to the influence of the high estimated values for PMio-2.s. In
several other cases of notably low correlation, the small available range of PM^.svalues relative
to other areas contributes to the lower correlation values, e.g., in Phoenix, Dallas, and Houston.
January 2010 D-4
DRAFT - Do Not Quote or Cite
-------
Table D-l. Squared correlation coefficients between observed and LOESS
model-predicted values of PM light extinction
Area
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
AVERAGE
Figure D-l
24-hour
PM25
mass vs.
daily
maximum
daylight 1-
hour PM
light
extinction
0.48
0.76
0.57
0.22
0.88
0.45
0.46
0.40
0.61
0.54
0.62
0.73
0.78
0.60
0.69
0.59
Figure D-2
1-hour
PM2.s mass
vs. same-
hour PM
light
extinction
0.81
0.83
0.63
0.67
0.89
0.59
0.61
0.43
0.81
0.72
0.55
0.63
0.69
0.6
0.76
0.68
Figure D-3
12-4 pm
average
PM2.s mass
vs. 12-4 pm
average PM
light
extinction
0.78
0.9
0.67
0.73
0.95
0.53
0.62
0.2
0.78
0.8
0.6
0.65
0.69
0.57
0.76
0.68
Figure D-4
12-4 pm
average
PM2.s mass
vs. daily
maximum
daylight 1-
hour PM
light
extinction
0.29
0.69
0.53
0.18
0.8
0.2
0.2
0.18
0.34
0.4
0.11
0.52
0.58
0.38
0.5
0.39
Figure D-5
8 am-12pm
average PM2.5
mass vs. daily
maximum
daylight 1-
hour PM
light
extinction
0.65
0.83
0.7
0.2
0.89
0.35
0.3
0.36
0.44
0.7
0.3
0.62
0.71
0.49
0.62
0.54
January 2010
DRAFT - Do Not Quote or Cite
D-5
-------
Figure D-l. - Relationship between 24-hour PMi.5 mass vs. daily maximum daylight 1-hour PM light extinction.
0 20 40
0 20 40
Phoenix, AZ
Pittsburgh. PA
Salt Lake City, UT
St. Louis, IL
Tacoma. WA
600
400
200
0
Fresno, CA
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
LU
£
- 1000
- 800
- 600
- 400
- 200
- 0
Atlanta, GA
Baltimore, MD
Birmingham, At
Dallas. TX
Detroit, Ml
600 -
400 -
200 -
0 -
0 20 40 60
0 20 40 60
Pm2.5 Daily average
0 20 40 60
January 2010
DRAFT - Do Not Quote or Cite
D-6
-------
Figure D-2. - Relationship between daylight 1-hour PMi.5 mass vs. same-hour PM light extinction.
0 50 100 150 200
0 50 100 150 200
_j [_
Phoenix, AZ
Pittsburgh, PA
Salt Lake City, UT
St. Louis, IL
Tacoma, WA
600
400
200
0
Fresno, CA
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
- 600
- 400
- 200
Atlanta, GA
Baltimore, MD
Birmingham, AL
Dallas, TX
Detroit, Ml
50 100 150 200
50 100 150 200
PM2.5 (ug/m3)
0 50 100 150 200
January 2010
DRAFT - Do Not Quote or Cite
D-7
-------
Figure D-3. Relationship between 12-4 pm average PMi.s mass vs. 12-4 pm average PM light extinction.
0 20 40
0 20 40 60
Phoenix, AZ
Pittsburgh, PA
Salt Lake City, UT
St. Louis, IL
Tacoma, WA
• \f£
Fresno, CA
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Atlanta, GA
Baltimore, MD
Birmingham, AL
Dallas, TX
Detroit. Ml
60 80 100 120
0 20 40 60 80 100 120
Pm2.5 12-4 average
0 20 40 60 80 100 120
January 2010
DRAFT - Do Not Quote or Cite
D-8
-------
Figure D-4. Relationship between 12-4 pm average PMi.s mass vs. daily maximum daylight 1-hour PM light
extinction.
0 20 40 60 80 100 120
0 20 40 60 80 100 120
Phoenix, AZ
Pittsburgh, PA
Salt Lake City, UT
St. Louis, II
Tacoma, WA
400
200
0
Fresno, CA
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia. PA
- 1000
- 800
- 600
- 400
- 200
- 0
Atlanta, GA
Baltimore, MD
Birmingham, AL
Dallas, TX
Detroit, Ml
1000
800
600
400 -
200 -
0 -
0 20 40 60 80 100 120
0 20 40 60 80 100 120
Pm2.5 12-4 average
0 20 40 60 80 100 120
January 2010
DRAFT - Do Not Quote or Cite
D-9
-------
Figure D-5. Relationship between 8 am-12pm average PMi.5 mass vs. daily maximum daylight 1-hour PM light extinction
50
Phoenix, AZ
Pittsburgh, PA
Salt Lake City, UT
St. Louis, II
Tacoma, WA
400
200
0
Fresno, CA
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
- 600
- 400
- 200
- o
Atlanta, GA
Baltimore, MD
Birmingham, AL
Dallas, TX
Detroit, Ml
0 50 100
Pm2.5 8-12 average
January 2010
DRAFT - Do Not Quote or Cite
D-10
-------
Figure D-6. Distribution of ratios of 1-hour PM light extinction and PMi.s mass
concentration.
A - Ratios of daily maximum daylight 1-hour PM light extinction to 24-hour average PMi.5
concentration.
4-
T
4-
T
T T
B - Ratios of daylight 1-hour PM light extinction to same-hour PMi.5 concentration
i i ii
-
T
1
.
T T
-t-
T
ilii
B
1
January 2010
DRAFT - Do Not Quote or Cite
D-ll
-------
APPENDIX E - DIFFERENCES IN DAILY PATTERNS OF
RELATIVE HUMIDITY AND PM LIGHT EXTINCTION
BETWEEN AREAS AND SEASONS
In the last review of the secondary PM NAAQS, the pattern of PM light extinction
during the day was of particular interest. It was noted, using estimates of hourly PM light
extinction based on a simpler approach than described for this analysis, that both (1) mid-day
PM light extinction and (2) the slope of the relationship between PM light extinction and
PM2.s concentration varied less among regions of the country that at other times of the day.
This was attributed to greater homogeneity of relative humidity across regions in the mid-day
period. This is in contrast to the situation in the morning and later afternoon hours, when
more eastern areas typically experience higher relative humidity levels than the more arid
western and southwestern areas. The current analysis allows these patterns to be re-
examined.
Figures E-l through E-4 show the diurnal pattern of season-average, hour-specific
PM light extinction and relative humidity for the four "daylight seasons." These graphics
exclude hours with relative humidity greater than 90 percent. Light extinction and relative
humidity for a given clock hour are averaged across the days in the season, across all three
years. Daylight hours (per the simplified schedule of Table 3-5) are indicated by solid
circles. Average 1-hour PM light extinction generally is highest in the morning,
corresponding to higher relatively humidity (mostly due to lower temperature), higher
vehicle traffic, and less dispersive conditions than later in the day. As was observed in the
last review, there is more variation in average 1-hour PM light extinction among areas in the
morning than at mid-day, although the morning variation has been reduced (relative to same
information in the first public review draft of this assessment) by the exclusion of hours with
relative humidity greater than 90 percent.
January 2010 E-l
DRAFT - Do Not Quote or Cite
-------
Figure E-l. Diurnal and seasonal patterns of relative humidity (percent) and PM light extinction (Mm" ) for 2005-2007
(a) November-January
Phoenix, AZ
Pittsburgh, PA
Salt Lake City, UT
St. Louis, IL
Tacoma, WA
8-
8 1
(
1
to <
25 _
99
a £ -
8 -
<
3-
8.
s -<
(
^asa\ A
Q_T^^ ^-4r^
" «»*P
) 6 12 17 2
Hour
Fresno, CA
P o°
^w/3
XP°
) 6 12 17 2
Hour
Atlanta, GA
•-0
.
V*vS\^f*
1 6 12 17 2
Hour
- e a-
m 58
'"*• oj ~
-8 t -
>- 8 8 -
3 I
" i 3n
L R $ J
-a s -
-8 8 -;
3 I
— O JO —
>- e sg .
**• &
-a B -
>- 8 8 H
3 (
ODOJ*X^%£pcfntf>
?»B*PlXM«Xl£d
6 12 17 2
Hour
Houston, TX
N^f
rtvfvfP* O)
^ cy eo
3 C
8in
in —
r S 8-
• S b ""
r 3 S -
3 C
QXCrXIV t&^®
• cr
^^?^^39
6 12 17 2
Hour
.os Angeles, Ct
°2S3§
^iF «^^r
6 12 17 2
Hour
Birmingham, AL
^ X
o0."1*""
^^ *U ,tfsft^
e 12 17 2
Hour
-5 SS -
r R 1 t
-8 fc -
-8 8 J
3 I
i
- 1 3-
- R 1-
-8 fc -
-8 8-1
3 (
- 1 8-
10 jg _
1
s £
L 8 8-7
3 (
"•"N^
0^
EC50^ ^Hfrt^P^^
G 12 17 2
Hour
New York, NY
^Xw^5
0*.
p%rf>^T^^rjcrf™
6 12 17 2
Hour
Dallas, TX
Vv*X
1 6 12 17 2
Hour
-s a -
_ in g j
-8 £ -
"-8 8 -
3 (
8*4
in —
- irt S -
CJ t
- s S -
'- 8 8 J
3 1
8in
jn —
-R i^
- s B -
-8 8 H
3 (
<**W^moa*kH
m°
^ru - jf ^!
b ^V1"^
8 12 17 2
Hour
Philadelphia, Pt
0^+^"
*J->cpf*PV««^PocPJ
e 12 i? 2
Hour
Detroit, Ml
usmow
^^V^ff?0
«ftrf^L
^Vwitc
6 12 17 2
Hour
- s
"- R
-85
- 8
3
l
- 1
r R
- s £ • RH
- 8
• Ext
3
- 1
in
, h^
" S cc
r 8
3
January 2010
DRAFT - Do Not Quote or Cite
E-2
-------
Figure E-2. Diurnal and seasonal patterns of relative humidity (percent) and PM light extinction (Mm"1) for 2005-2007,
continued
(b) February-April
Phoenix, AZ
Pittsburgh, PA
Salt Lake City, UT
St. Louis, IL
Tacoma, WA
8-
8-
8 -<
1
(
a B -j
<
3-
zl ™ ~
% i
<
~-
^_ ^<&
QOooxA^fe^HpocP"
) 6 12 17 2
Hour
Fresno, CA
V%J** 0°
^T^ O
) 6 12 17 2
Hour
Atlanta, GA
• X-°
y&ff? *• jj3-*
^**^
1 6 12 17 2
Hour
§UJ
OT ~
- R a-
-a t -
L 8 8 -J
3 I
>- e $ J
j- a B -
3 I
81/J
irt —
- *> » .
f1- ft
-a B -
'-a s -i
3 I
arf«*.
%»«K^
CtP1^, ^Jl
"^^^^^
1 6 12 17 2
Hour
Houston, TX
• <^
^"'XlM**^
1 6 12 17 2
Ho'jr
Baltimore, MD
tfO^V ,*
^^W^rf5^
ttSO^^^^^^^^^Oi
1 6 12 17 2
Hour
O u"J
- R 8-
-a B -
jr s s -
3 C
I
8tfl
in —
L. m 8 -
fS- OJ
-a B -
3 C
I
8in
in —
- R 8-
• S b *"
I
>- a s -
3 C
X^
UJ^J*<1**^>rr
e 12 17 2
Hour
.os Angeles, Ct
%af\_ rfo*
occp • *^ dJ
^- ^P*
6 12 17 2
Hour
Birmingham, AL
ftf^b
•. ,s&
D^"|S*-
- § 1 "
. w g „
""" «
i
- S fc -
j- S3 S J
3 I
i
- 1 S-
~ fC S —
j- 8 5 J
w en
3 (
- 1 8-
, i
-a B -
1 S3 8 J
3 (
"^V^
X±L
tacP ^rt%iiUJ
6 12 17 2
Hour
New York, NY
^^_
J^k. ^»n
6 12 17 2
Hour
Dallas, TX
X*p
*»»*"%, _^»
^**p
6 12 17 2
Hour
8 in
to —
- 12 ^ J
-a B -
-a s -
«
3 (
- s 8-
- a B -
3 (
- § 3n
- m S -
^ CJ
-a B -
to Q> J
)T TO
3 (
******"• rtH
6 12 17 2
Hour
Detroit, Ml
tf^V
v %»^eftcB;
6 12 17 2
Hour
- §
- 8 5
- a
3
l
- R
-as • RH
in
• Ext
3
- §
_ IO
" S CC
3
January 2010
DRAFT - Do Not Quote or Cite
E-3
-------
Figure E-3. Diurnal and seasonal patterns of relative humidity (percent) and PM light extinction (Mm"1) for 2005-2007,
continued
(c) May-July
Phoenix, AZ
- §
- is
- 8
r S3
1 •" • i
0 6 12 17 23
Hour
Fresno, CA
Pittsburgh, PA
Salt Lake City, UT
r §
- s
r 8
St. Louis, IL
8 -
-8
0 6 12 17 23
Hour
Houston, TX
Los Angeles, CA
fc -
r *
'- K
S
a
£ -i
1 i' ' r
0 6 12 17 23
- I
r R
r 8
New York, NY
3n - 8
0 6 12 17 23
Hour
Baltimore, MD
Birmingham, AL
Dallas, TX
R -
rrtf'*
- s
"i i"" '
0 6 12 17 23
Hour
-
8-
- S
- s
0 6 12 17 23
Hour
• '
0 6 12 17 23
Hour
Tacoma, WA
8 -
- 8
^ K
- 8
0 8 12 17 23
Hour
Philadelphia, PA
- I
K
~ 8 EE
0 6 12 17 23
Hour
Detroit, Ml
R -
- S
r S
>- K
"i"" : i
0 6 12 17 23
Hour
RH
Ext
January 2010
DRAFT - Do Not Quote or Cite
E-4
-------
Figure E-4. Diurnal and seasonal patterns of relative humidity (percent) and PM light extinction (Mm"1) for 2005-2007,
continued
(d) August-October
Phoenix, AZ Pittsburgh, PA Salt Lake City, UT St. Louis, IL Tacoma, WA
^1
1-
8 -
i
VN^*
&xpM^l4%fBQ
§UJ
to -
*S^
O UJ
r* 1-
- S £ J
h S S -
1
\ 0^
"***^
•~»ini 11,1 r '
- 1 8-
>- S fc -
-S3 8 -
^*vX
2E_
•^^rfp3"
8ia
10 —
r R 8J
- s B -
-a s -
pXCf^ (3
^g^^f^r^
- §
- R
- «
- a
0 6 12 17 23 0 6 12 17 23 0 6 12 17 23 0 6 12 17 23 0 6 12 17 23
Hour Hour Hour Hour Hour
Fresno, CA Houston, TX Los Angeles, CA New York, NY Philadelphia, PA
§"
8 -
<
S -
3 T
*• _rf
\ /
pctf^^S^ „<{!
X»rf^
-§ §n
- R 8 J
L s B^
_ Irt O> _
r c*j co i
• 0
*^^^^f
U*i rf°^
8tfl
in —
L is 8 -
fs- OJ
L s £ ^
>- ^ § -
Pa/*X>" at
•W
o irt
- s 1 1
-a B -
-a a -
tf#\
^U ^r^
,<.
2 *«^^ftl
- 1 8-1
- R 1 J
- s B -
>- a s J
v <^
^tL. -**
)(i^ l^> xpj
- 1
*- R
-«
>- a
0 6 12 17 23 0 6 12 17 23 0 6 12 17 23 0 6 12 17 23 0 6 12 17 23
Hour Hour Hour Hour Hour
Atlanta, GA Baltimore, MD Birmingham, AL Dallas, TX Detroit, Ml
In
1-
K. —
S -
t***'^/»
* ^M
to ^ • <*9
° * ^ii?
8i/j
irt —
L R 8-
- s B -
-a s -j
""^^
,
8 in
in —
- R 8J
• S b ~"
r a s -
^V X
«^» N""'yD
*^^^r
- 1 8-
- K 8-
i
j- s B -
-S3 S -3
^•.
\«-p
-------
APPENDIX F - DISTRIBUTIONS OF MAXIMUM DAILY
AND HOURLY DAYLIGHT PM LIGHT EXTINCTION -
UNDER "JUST MEET" CONDITIONS
F-l
January 2010
DRAFT Do Not Quote or Cite
-------
(a) NAAQS Scenario
Daily Max
191 Mm-1
90th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90% RH)
ExiRallbackOail/MaxN AAQS191 Pctl90DVsFromdaily.max
8 -
109 324 3W 98
i*S 22S
90% RH)
F-2
January 2010
DRAFT Do Not Quote or Cite
-------
ExlFtoltbackLowRHDayHours.NAAQS191Pcll90DVsFromdaily.max
I 8
338? 3019 s*NS
o
i
January 2010
DRAFT Do Not Quote or Cite
F-3
-------
(b) NAAQS Scenario
Daily Max
191 Mm-1
95th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExtRollbackDailyMaxNAAQSI 91 PciBSDVsFromdaily.max
8 -
1*9 324 JM
SOS 2T3 Jit 289 349 279
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
ExiRollbackLowRHDayHoursNAAOS191PcH95DVsFromdaily.max
I 8
1688 SS! 1
*• ^
January 2010
DRAFT Do Not Quote or Cite
F-4
-------
(c) NAAQS Scenario
Daily Max
191 Mm-1
98th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExlRol!backDai!yMaxNAAQS191 PctBBDVsFromdaiiy.max
1*9 324 MS 38 SOfl 2« SS* 289 349
)43 S2S
-^ -IT
il '^
'V
r ?
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
ExlFtolibackLowRHDayHoursNAAOS191PclB8DVsFrofndailymax
*
/
F-5
January 2010
DRAFT Do Not Quote or Cite
-------
(d) NAAQS Scenario
Daily Max
112 Mm-1
90th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExiRolIb3ckDaiiyMaxNAAQS112Pcti9QDVsFromdaiIy.max
8 -
1*9 324 JM
SOS 2?3 Jit 289 349 279
-f~-g-
.4—~_&,
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
ExiRollbackLowRHDayHoursNAAOS112PcH90DVsFromdaily.max
I 8
143 22$
^ ./ «*
* «/ ^
F-6
*• ^
January 2010
DRAFT Do Not Quote or Cite
-------
(e) NAAQS Scenario
Daily Max
112 Mm-1
95th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExtRollbackDailyMaxNAAQSI12PciI95DVsFromdaiIy.max
8 -
1*9 324 JM 98 SOS 2T3 Jit 289 349 279
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
ExiRollbackLowRHDayHoursNAAOS112PcH95DVsFromdaily.max
±±33
..
*• ^
/
F-7
January 2010
DRAFT Do Not Quote or Cite
-------
(f) NAAQS Scenario
Daily Max
112 Mm1
98th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExlRol!b3ckDai!yMaxNAAQS112Pct(98DVsFromdaiiy.max
8 -
1*9 324 MS 98 SOS
1*1 289 349 2J9
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
ExiRollbackLowRHDayHoursNAAOS112PcH98DVsFromdaily.max
)43 S2S
F-8
January 2010
DRAFT Do Not Quote or Cite
-------
(g) NAAQS Scenario
Daily Max
64 Mm-1
90th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExtRollbackDailyMaxNAAQS64PcII90DVsFromdaily.max
1*9 324 MS 38 5M 271 S5* 289 349
i i
*y cT
90% RH)
ExtfiollbackLo«RHDayHoursNAAQS64Pctl90DVsFromdaily.max
I 8
2471 1S33 3816
F-9
January 2010
DRAFT Do Not Quote or Cite
-------
(h) NAAQS Scenario
Daily Max
64 Mm-1
95th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExiFtolIbackDaityMaxNAAQSS4Pctl95DVsFromdaily.max
8 -
1*9 324 MS
SOS 2T3 Jit 289 349 2J9
)43 32S
l' HI
'V
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
E xtnollbacfcLowRHOayHoursN AAQS64 Pctl95O VsFromdaily. max
I 8
2471 1S33 3816
F-10
January 2010
DRAFT Do Not Quote or Cite
-------
(i) NAAQS Scenario
Daily Max
64 Mm-1
98th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExtRoilbackDailyMaxNAAQS64PcII98DVsFromdaily.max
8 -
1*9 324 3M 98 SOfl 2« 1S4 289 349
•f CF
l' HI
'V
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
ExtFtollbackLcnvRHDayHoursNAAQS64Pctl98DVsfromdaily.max
I 8
s:m ae#s
«
F-ll
January 2010
DRAFT Do Not Quote or Cite
-------
(j) NAAQS Scenario
All hours
191 Mm-1
90th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExtftollbackDailyMaxNAAQSI 91 Pcll9f}DVsFrt>mall.hours
8 -
169 324 358 98 SOS 2?3 Jit 289 349 279
JTT 181 )43 22S
•-••p-jr—T
i .i i
AT ^w
/" y
^ * ,
/ -' /' / X
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQSI 91 PctraOOVsFiomall hours
I 8
1686 SS! S 063$ 1188 33S? SftlS
F-12
January 2010
DRAFT Do Not Quote or Cite
-------
(k) NAAQS Scenario
All hours
191 Mm-1
95th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExiRollba<*DailyMaxNAAOSi9lPcll95DVsFromall.hours
8 -
1*9 324 358 98 SOS 2T3 Jit 289 349 2J9
2?T 181 )43 aas
:;;:B{_
1 I I
4-
E3^
r -r
A I "-O
'V
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS191Pctl95DVsFromall.houts
I 8
8
^illiiUU
«
F-13
January 2010
DRAFT Do Not Quote or Cite
-------
(1) NAAQS Scenario
All hours
191 Mm-1
98th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExtflollbackDailyMaxNAAQSI 91 PcllSSDVsFromall.hours
8 -
1*9 324 MS 98 SOS 2M Jit 289 349
--&.-
=^p
i^-^-^frJ-T-^_r^_
J cr
' 90% RH)
ExtRollbackLowRHDayHoursNAAQSI 91 PctlSSOVsFiomall hours
I 8
o o <5
143 S2S
r
i 11:|1 1 iiili
.1. T" ^-^ _•_ -1- _—. ^^—. i i _i—, 1=1 r—— , 1 ^—, i
«
F-14
January 2010
DRAFT Do Not Quote or Cite
-------
(m) NAAQS Scenario
All hours
112 Mm-1
90th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExtflollbackDailyMaxNAAQSI12Pcll9f}DVsFrt>mall.hours
8 -
1*9 324 MS 98 SOS 2M Jit 289 349
27T 1*1 )43 22$
l...,i,..,t,.4.4.
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS112Pctl90OVsFromall.hours
s -
lose ssu
iifla sas*
2471 1B33 3816
!4*3 K8S
January 2010
DRAFT Do Not Quote or Cite
F-15
-------
(n) NAAQS Scenario
All hours
112 Mm-1
95th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExiRollbackDailyMaxNAAGSi 12Pcll95DVsFromall.hours
8 -
1*9 324 MS 38 SOfl 2« SS* 16$ 349
A
^
-f-
)43 S2S
A I "-O
'V
i ij-E ----- L
^ -r
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS112Pctl95DVsFromall.houts
I 8
o
o
o o §
44±F±±
t I * JL J, I
0
o o o
fi
11 g
tit
o
8 o 1
I till
£^=*=5=3-e:=j-E=jj:^r-
January 2010
DRAFT Do Not Quote or Cite
F-16
-------
(o) NAAQS Scenario
All hours
112 Mm-1
98th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExtftollbackDailyMaxNAAQSI12Pcll98DVsFr<>mall.hours
8 -
1*9 324 JM 98 SOS 2M Jit 289 349 279
O O
.i.-4-.X-J..
<*- ^+
'• / y
^ * ,
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
ExtRollbackLowRHDayHoursNAAQS112Pctl98OVsFromall.hours
8 8
F-17
January 2010
DRAFT Do Not Quote or Cite
-------
(p) NAAQS Scenario
All hours
64 Mm-1
90th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExlRolibackDai(yMaxNAAQS64Pctl90DVsFiomali.hours
1*9 324 JM 98 SOS 2?3 Jit 289 349
ti::r:l:::i:
~——, ** , , r ™
143 22S
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
ExlRoilbackLowRHDajfHoutsNAAQS84Pcll90DVsFromall.hours
I 8
1688 SS! 1
2471 1S33 3816
*•
January 2010
DRAFT Do Not Quote or Cite
F-18
-------
(q) NAAQS Scenario
All hours
64 Mm-1
95th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExtRolbackDailyMaxNAAQS64Pctl95DVsFiomall.hours
8 -
1*9 324 MS
SOS 2?3 Jit 289 349 2J9
90% RH)
ExlRoilbackLowRHDajfHoursNAAQS64Pcll95DVsFromall.hours
143 22S
// /
F-19
January 2010
DRAFT Do Not Quote or Cite
-------
(r) NAAQS Scenario
All hours
64 Mm-1
98th percentile
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
ExlRolibackDaityMaxNAAQS64Pctl98DVsFiomall. hours
8 -
1*9 324 MS 98 SOS m 14* 289 349
:i
<* X/ x / /
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
ExlRoilbackLowRHDajfHoursNAAQS64Pcll98DVsFromall.hours
I 8
2471 1S33 3816
)43 32S
•
F-20
January 2010
DRAFT Do Not Quote or Cite
-------
(s) NAAQS Scenario
15 ng/m3 annual
35 ng/m3 24-hour
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
PMRollbackDaityMaxCasel NAAQS
Tw m m si IB Ii5 i« in Hi K9 i3i !n ilT
0 e
8 g o
I * ! I i 1 I
" *•"" •* • 11—i——.-. 1 4 ^ — "| ~ :,i~:
r^rr -.
T T
B:
/••
- •»
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
PMRoIlbachCasaiNAAQS
F-21
January 2010
DRAFT Do Not Quote or Cite
-------
(t) NAAQS Scenario
12 ng/m3 annual
25 ng/m3 24-hour
Displayed: Daily Max Daylight Light Extinction (excluding hours >90%RH)
Displayed: Hourly Daylight Light Extinction (excluding hours >90% RH)
PM Roilba<*Case2 NAAQS
1MB Kn
8
14
o 8
•** c?
c/ /' V
<" <= o*
January 2010
DRAFT Do Not Quote or Cite
F-22
-------
APPENDIX G - ADDITIONAL INFORMATION ON THE
EXCLUSION OF DAYLIGHT HOURS WITH RELATIVE
HUMIDITY GREATER THAN 90 PERCENT
This appendix provides detailed information related to the exclusion of daylight hours
with relative humidity greater than 90 percent from the design value formula for the secondary
NAAQS scenarios based on PM light extinction, as discussed in section 3.3.5. As described in
that section, these hours have also been excluded from graphical displays of the distribution of
PM light extinction under current conditions and the various NAAQS scenarios, and from the
denominator of percentages of day or hours (as in Table 4-7).
Table G-l shows how many estimates of 1-hour daylight PM light extinction were
excluded, based both on individual hours and on days that were affected by the exclusion of one
or more daylight hours. Phoenix was not affected at all. Among the other areas, Detroit was the
least affected. For all areas, comparison of the percentage of hours affected to the percentage of
days affected indicates that several hours with high relatively humidity tend to occur in the same
day, rather than being evenly distributed across all days. For example, in Atlanta 24 percent of
daylight hours have relative humidity greater than 90 percent, which corresponds to about 876
hours per year (assuming there were data for every day of the year and given that on average
there are about 10 fully daylight hours per day). However, only 80 percent of the days
(corresponding to 292 days, if there were data for every day of the year) are affected. Thus, on
average, an affected day in Atlanta has about 3 affected hours. The tile plots in Figure 3-12 also
illustrate the tendency for hours with high PM light extinction to cluster in some days.
Figure G-l shows when during the daylight hours these hours with relative humidity
greater than 90 percent occurred, prior to the their exclusion. Some but not all areas have a
strong tendency for the affected hours to be in the morning. The counts in this figure are across
all the days in 2006-2008 that have estimates of PM light extinction, not all the actual calendar
days in that three year period. Given the regularity of the monitoring schedules, these results
should represent year-round conditions reasonably well. However, the estimates of PM light
extinction for Phoenix and Houston are not seasonally balanced due to one calendar quarter with
no data in each case (see Table 3-4), so the true year-round time-of-day distributions of excluded
hours for these two areas may be somewhat different than shown here.
Figure G-2 contrasts the distribution of daylight PM light extinction estimates before and
after the exclusion, based on both daily maximum values and all daylight hourly values
individually. The differences observable in the figure are consistent with the information on the
percentages of hours and day affected in the study areas. In most cases, the highest values of
G-l
January 2010
DRAFT Do Not Quote or Cite
-------
light extinction are notably lower after exclusion, on both a daily maximum basis and individual
hour basis, indicating that PM concentrations in some of the excluded hours are fairly high. If
only low-PM hours were excluded by the relative humidity screen, the highest values of light
extinction would not have been affected.
Finally, Table G-2 contrasts PM light extinction design values before and after the
exclusion, for the 90* and 95* percentile forms based on daily maximum daylight 1-hour PM
light extinction, for current conditions. (A similar comparison for the 98th percentile form was
not generated.) As expected, design values are notably lower after the exclusion. For both
percentile forms, the largest reduction is in Los Angeles (represented by the Rubidoux site in the
far eastern part of the South Coast Air Basin). Phoenix had no hours with relative humidity
greater than 90 percent, and accordingly Table G-2 shows that its PM light extinction design
values are not affected by the exclusion. Similarly, Detroit and Dallas had only a few hours with
relative humidity greater than 90 percent, and their design values are affected very little by the
exclusion.
Table G-l. Percent of daylight hours and days affected by the elimination of
hours with relative humidity greater than 90 percent
Study Areas
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Percent of Daylight Hours
Excluded
12.3
3.6
10.6
0.0
2.9
2.8
9.6
6.4
4.4
24.1
2.3
11.4
10.6
9.6
9.1
Percent of Days with at
Least One Daylight Hour
Excluded
49.1
15.7
49.7
0.0
13.7
12.8
40.9
21.1
19.1
80.7
7.1
41.2
33.2
31.7
22.4
G-2
January 2010
DRAFT Do Not Quote or Cite
-------
Figure G-l. Distribution by time of day of eliminated daylight hours with relative humidity greater than 90 percent.
.1
1
Fresno, CA
llll.-
Atlanta, GA
Illlllhl.
llllllll
Houston, TX
lull...
Baltimore, MD
•llllll.lll-.
.11 .
Los Anqeles, CA
llll,
Birmingham, AL
•••••••••••-•_
•llllllll. I. •_
New York, NY
.•••••••••••I-
Dallas, TX
..•III.. . _
Philadelphia, PA
Detroit, Ml
- 50
150 -
05 06 07 08 09 10 11 12 13 14 15 16 17 18 05 06 07 08 09 10 11 12 13 14 15 16 17 18 05 06 07 08 09 10 11 12 13 14 15 16 17 18 05 06 07 08 09 10 11 12 13 14 15 16 17 18 05 06 07 08 09 10 11 12 13 14 15 16 17 18
Hour of Day
G-3
January 2010
DRAFT Do Not Quote or Cite
-------
Figure G-2. Comparison of distributions of estimated daylight 1-hour PM light extinction
and maximum daily daylight 1-hour PM light extinction across the 2005-2007 period for
current conditions, by study area, before and after elimination of hours with relative
humidity greater than 90 percent.
(a) Maximum daily values
Before Elimination
Daly Maximum Extinction (Daylight Hours)
-S-
8
i
/^
After Elimination
Daily Maximum Extinction (Daylight Hours)
§-
ti--^
V" r
i
-s******,-
F^
G-4
January 2010
DRAFT Do Not Quote or Cite
-------
(b) Individual 1-hour values
Before Elimination
Hourly Extinction (Daylight Hours)
After Elimination
Hourly Extinction (Daylight Hours)
S
January 2010
DRAFT Do Not Quote or Cite
G-5
-------
Table G-2. Comparison of 90th and 95th percentile PM light extinction design values for the
2005-2007 period for current conditions based on maximum daily 1-hour daylight PM light
extinction, before and after elimination of hours with relative humidity greater than 90
percent
Study
Areas
Tacoma
Fresno
Los Angeles
Phoenix
Salt Lake
City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
PM Light Extinction Design Values Based on Daily Maximum 1 -Hour Values
90th Percentile
Before
Exclusion
244
381
919
105
176
189
253
359
366
380
313
368
399
382
339
After
Exclusion
140
338
469
105
164
183
194
307
357
249
310
278
246
286
306
Reduction
Due to
Exclusion
104
43
450
0
12
5
59
52
9
131
3
90
153
96
33
95th Percentile
Before
Exclusion
371
533
114
0
144
266
239
279
423
496
462
473
500
446
449
415
After
Exclusion
157
463
554
144
252
239
234
381
483
288
473
313
286
339
355
Reduction
Due to
Exclusion
215
70
586
0
13
0
44
42
13
174
0
187
159
110
61
January 2010
DRAFT Do Not Quote or Cite
G-6
-------
APPENDIX H - INTER-YEAR VARIABILITY
One aspect of a NAAQS is whether it is based on the level of the selected indicator for a
single year, or the average of the level of that indicator over multiple years. The NAAQS
scenarios examined in this assessment are all based on a three-year average approach. That is,
design values are based on the average of specified percentile values of PM light extinction from
2005, 2006, and 2007. Table H-l presents more detailed information on the variability of these
percentiles across these three years.
Figure H-l presents some of the information in Table H-l in graphical form, specifically
for the 90* percentile form for both the daily maximum and all hours approaches.
H-l
January 2010
DRAFT Do Not Quote or Cite
-------
Table H-l. Year-specific percentile values of PM light extinction for 2005, 2006, and
2007
Study
Areas
Tacoma
Fresno
Los
Angeles
Phoenix
Salt Lake
City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
Tacoma
Fresno
Los
Angeles
Phoenix
Salt Lake
City
Dallas
Houston
St. Louis
Birmingham
Atlanta
Detroit
Pittsburgh
Baltimore
Philadelphia
New York
90th Percentile Form
2005
2006
2007
2005-
2007
Average
95th Percentile Form
2005
2006
2007
2005-
2007
Average
98th Percentile Form
2005
2006
2007
2005-
2007
Average
Based on Daily Maximum 1-Hour Daylight PM Light Extinction
(Excluding hours with relative humidity greater than 90%)
NA
308
435
100
217
184
217
350
438
235
300
284
303
257
333
121
307
485
110
112
169
204
331
307
255
312
257
227
325
265
158
398
486
NA
163
197
161
239
325
257
318
292
208
276
320
140
338
469
105
164
183
194
307
357
249
310
278
246
286
306
NA
553
507
156
310
252
269
434
547
283
347
347
362
333
405
140
364
535
131
141
223
238
405
410
295
408
272
258
367
275
173
472
619
NA
306
242
196
303
493
286
663
320
239
318
384
157
463
554
144
252
239
234
381
483
288
473
313
286
339
355
NA
658
606
344
343
313
306
483
608
305
391
360
417
426
568
214
400
605
187
191
321
319
572
513
338
489
350
308
376
352
206
542
662
NA
696
271
248
347
565
351
1051
383
260
377
451
210
533
624
266
410
302
291
467
562
331
644
364
328
393
457
Based on 1-Hour Daylight PM Light Extinction (All Hours)
(Excluding hours with relative humidity greater than 90%)
NA
183
263
67
115
114
116
229
191
166
226
173
203
163
204
73
175
275
68
67
100
98
195
162
164
213
153
163
204
169
78
212
259
NA
96
125
100
157
166
169
198
176
150
183
186
76
190
266
68
93
113
105
194
173
166
212
167
172
183
186
NA
257
325
79
194
145
143
211
251
188
267
217
290
209
265
102
262
362
78
83
126
122
239
204
194
253
193
194
234
222
109
278
359
NA
148
158
119
188
226
202
234
218
196
224
244
106
266
349
79
142
143
128
235
227
195
251
209
227
222
244
NA
395
408
92
255
184
191
334
340
233
320
284
345
280
313
120
332
458
96
116
176
174
309
267
233
312
236
225
298
267
151
391
486
NA
303
204
148
226
319
248
314
272
225
258
320
136
373
451
94
225
188
171
290
309
238
315
264
265
279
300
H-2
January 2010
DRAFT Do Not Quote or Cite
-------
Figure H-l. Inter-year variability in 90th percentile 1-hour daylight PM light
extinction (excluding hours with relative humidity greater than 90 percent)
(a) Daily maximum approach
+J
.2>
'3
S
0. „
+J T
Si
> s
TO
0 1
X U
TO
S
_>
'(5
0
enn -.--——————————^^
DUU
cnn
ouu
I* Ann
„ 300
j
j 200
•i nn
1 UU
•v
^^^-^
fc — ^
*^= A- .
^~~-----~-^_____ ^r-=-^
r i i
2005 2006 2007
Year
—•—Salt Lake City -"—Dallas -A— Houston St. Louis
-*— Birmingham -•—Atlanta — i— Detroit Pittsburgh
Baltimore Philadelphia New York -A— Phoenix
Los Angeles Fresno Tacoma
(b) All daylight hours approach
£
.2>
Q. T
£ j
I1 *
i"
0
X
onn ^
ouu
ocn
-------
APPENDIX I - DAYLIGHT HOURS
Table 3-5 shows the simple scheme used in this analysis to denote hours as fully daylight
and thus eligible for consideration in the calculation of design values for the secondary NAAQS
scenarios based on PM light extinction. This scheme also has been used to select which hours to
to show in various graphics. The scheme is based on applying a fixed set of fully daylight hours
for each three-month season (November to January, etc.). In reality, the local time minutes of
daylight vary continuously during the year, with latitude, and with the east-west position of a city
within its time zone. The hours that are fully daylight will change in increments rather than
continuously. This appendix examines how well the simple scheme reflects actual conditions and
how disparities if any might affect the results presented and the answers to policy relevant
questions that may be addressed in the subsequent policy assessment document.
Six study areas were selected for this examination: Tacoma, Los Angeles, Phoenix,
Houston, Detroit, and New York. These areas cover the extremes with regard to latitude and to
east-west position within time zone. For each area, the times of sunrise (defined by the leading
or top edge of the sun appearing above the horizon) and of sunset (defined by the leading or
bottom edge of the sun disappearing below the horizon) were obtained for each day of the year.
It is several minutes after each of these times that the sun is fully visible in the morning and not
visible at all in the evening.
Figure 1-1 shows the relationship between these sunrise and sunset times and the simple
scheme used to denote hours as fully daylight. The vertical scale is in hours with zero
corresponding to local noon. The smooth curves represent the actual times of sunrise (top of
figure) and sunset (bottom of figure). The stepped lines represent the scheme used to select the
first and last hour denoted as fully daylight. Months are indicated on the horizontal axis. The
figure indicates that the simple scheme has the effect of treating some hours as daylight that in
fact contain minutes prior to sunrise or after sunset, and conversely treating some hours as not
daylight that include no such minutes. In particular:
• In February, the hours from 7 am to 8 am and from 5 pm to 6 pm are treated as
daylight but include non-daylight minutes in most of the example areas.
• In April, the hour from 6 am to 7 am is treated as non-daylight but in many areas
includes only minutes that are after sunrise.
• In most of June and most of July, for Detroit and Tacoma only, the hour of 7 pm to 8
pm is treated as non-daylight but in fact has no minutes after sunset.
1-1
January 2010
DRAFT Do Not Quote or Cite
-------
• In October, the hours of 6 am to 7 am and 5 pm to 6 pm are treated as daylight but
include non-daylight minutes in all of the example areas.
The tile plots in Figure 3-12 can be used to assess the significance of these disparities,
i.e., whether they are likely to significantly affect PM light extinction design values. Table 1-1
contains observations for each of the 24 combinations of the four time periods listed above and
the six example areas. Taken together, these observations make it likely that refining the scheme
for designating hours as fully daylight would not significantly change conclusions that can be
drawn from this assessment as it has been performed. Changing the scheme would involve
considerable effort in updating virtually every table and graphic in the assessment, however.
1-2
January 2010
DRAFT Do Not Quote or Cite
-------
Figure 1-1. Comparison of Actual Sunrise and Sunset Times to this Assessment's Scheme to Denote Hours as Fully
Daylight
o/inn
Z4UU
oonn
zzuu
onnn
zuuu
1 snn
1 oUU ~
CD -icnn -
— louu
• _ -i /inn
] 1 4-UU
1 onn
— I ZUU
(0
01 nnn
I UUU
• snn
_J oUU
ROD
DUU
^nn
4UU
onn
ZUU
n
Ja
1
Sunrise/Sunset for Six Sites
Jtlggj" I ' ~~* ^^""^^IL.^
^^r~ ^^^^
^~.T" ' — "
^^^^"*"^^fc- ^^^^^^^1 ^ — '
'~-^~— ~- — — —
n- Feb- Mar- Apr- May- Jun- Jul- Aug- Sep- Oct- Nov- De
0 10 10 10 10 10 10 10 10 10 10 1
Date
3C-
0
Assumed daylight start
Assumed daylight end
TAG Rise
TAG Set
- LA Rise
LA Set
-PHXRise
PHX Set
HOD Set
HPT r?icn
~ ut i Kise
ni — r °Lot
L/C 1 Ocl
-NY Rise
NY Set
1-3
January 2010
DRAFT Do Not Quote or Cite
-------
Table 1-1. Observations from Tile Plots for Hours with Questionable Daylight/Non-Daylight Status in Six Study Areas
Study Area
February (AM and PM)
April (AM)
June-July (PM)
October (AM and PM)
Tacoma
The morning hour with questionable
daylight status tends to have RH > 90%.
The evening hour in question tends to
either have low PM light extinction or
to have RH > 90%.
The tile plot does not show data for the
morning hour that may better be denoted
daylight, but the instances of high PM light
extinction that do appear typically last
multiple hours.
Very late afternoon typically is not a
period of high PM light extinction.
Instances of high light extinction involving
the questionable hours are multi-hour and/or
involve RH > 90%.
Los
Angeles
Instances of high light extinction
involving the questionable hours are
multi-hour and/or involve RH > 90%.
The tile plot does not show data for the
morning hour that may better be denoted
daylight, but the instances of high PM light
extinction that do appear typically last
multiple hours.
NA
Instances of high light extinction involving
the questionable hours are multi-hour and/or
involve RH > 90%.
Phoenix
Instances of high light extinction
involving the questionable hours are
multi-hour.
The tile plot does not show data for the
morning hour that may better be denoted
daylight, but early morning in April typically
is not a time of high PM light extinction.
NA
PM light extinction is usually low in
October; on those days with moderate levels
in the questionable hours, another hour in
the central part of the day has a similar level.
Houston
Instances of high light extinction
involving the questionable hours are
multi-hour and/or involve RH > 90%.
The tile plot does not show data for the
morning hour that may better be denoted
daylight, but the instances of high PM light
extinction that do appear typically last
multiple hours and/or involve RH >90%.
NA
The amount of information is limited due to
missing data. On those days with moderate
to high PM light extinction during the
questionable hours, another hour has a
similar level, or RH >90% plays a role.
Detroit
Instances of high light extinction
involving the questionable hours are
multi-hour.
The tile plot does not show data for the
morning hour that may better be denoted
daylight, but the instances of high PM light
extinction that do appear typically last
multiple hours.
July generally is a time of high PM
light extinction for the hours currently
considered daylight. Adding one
more late afternoon hour likely would
not affect design values.
Instances of high light extinction involving
the questionable hours are multi-hour.
New York
All but one instance of high light
extinction involving the questionable
hours are multi-hour.
The tile plot does not show data for the
morning hour that may better be denoted
daylight, but the instances of high PM light
extinction that do appear typically last
multiple hours.
NA
Instances of high light extinction involving
the questionable hours are multi-hour and/or
involve RH > 90%.
January 2010
DRAFT Do Not Quote or Cite
1-4
-------
January 2010
DRAFT Do Not Quote or Cite
-------
United States Office of Air Quality Planning and Standards Publication No. EP A-452/P-10-002
Environmental Protection Health and Environmental Impacts Division January, 2010
Agency Research Triangle Park, NC
January 2010
DRAFT Do Not Quote or Cite
------- |