Technical Support Document:
Intended Area Redesignation for the 2010 1-Hour Sulfur
Dioxide Primary National Ambient Air Quality Standard for
Portions of Westmoreland and Cambria Counties in
Pennsylvania
January 2023
TSD prepared by Tim Leon-Guerrero,
Air Quality and Analysis Branch
and
Megan Goold,
Planning and Implementation Branch
Air and Radiation Division,
U.S. EPA Region 3
Reviewed by:
0*^- "H-. CW
Alice Chow, Chief
Air Quality and Analysis Branch
Mike Gordon, Chief
Planning and Implementation Branch
1
-------
Contents
1. Summary 4
2. General Approach and Schedule 5
3. Definitions 6
4. Background 7
4.1. 2OIOSO2NAAQS 7
4.2. History of 2010 SO2NAAQS Designations 7
4.3. History of Westmoreland, Cambria, Indiana, PA 2010 SO2 Modeled violations 7
5. Technical Analysis 8
5.1. Overview 9
5.2. Air Quality Modeling Analysis for Westmoreland and Cambria Counties, PA 10
5.2.1. Modeling Selection and Components 13
5.2.2. Modeling Parameter: Rural or Urban Dispersion 15
5.2.3. Modeling Parameter: Area of Analysis (Receptor Grid) 17
5.2.4. Modeling Parameter: Source Characterization 21
5.2.5. Modeling Parameter: Emissions 22
5.2.6. Modeling Parameter: Meteorology and Surface Characteristics 53
5.2.7. Modeling Parameter: Background SO2 Concentrations 68
5.2.8. EPA Site-Specific Adjusted U-star Modeling Summary and Results 80
5.2.9. EPA Site-Specific Turbulence Summary and Results 87
5.3. Air Quality Monitoring Data for the Westmoreland and Cambria Counties, PA 92
5.4. Intended Designation Boundary Determination 93
5.4.1. Factor 1: Ambient Air Quality Data and Dispersion Modeling Results 94
5.4.2. Factor 2: Emissions-Related Data 94
5.4.3. Factor 3: Meteorology 95
5.4.4. Factor 4: Geography and Topography 95
5.4.5. Factor 5: Jurisdictional Boundaries 95
5.4.6. Intended Nonattainment Area Boundary 96
5.5. Modeling Analyses Provided by Other Parties 97
5.5.1. KEY-CON 100
5.5.2. Sierra Club 122
2
-------
6. Summary of EPA's Intended Revised Designation for the Westmoreland and Cambria Area
137
7. References 139
3
-------
1. Summary
Pursuant to section 107(d) of the Clean Air Act (CAA), the U.S. Environmental Protection
Agency (the EPA, we, or us) was required to designate areas as either "nonattainment,"
"attainment," or "unclassifiable" for the 2010 1-hour sulfur dioxide (SO2) primary national
ambient air quality standard (NAAQS) (2010 SO2 NAAQS). The CAA defines a nonattainment
area as an area that does not meet the NAAQS or that contributes to a nearby area that does not
meet the NAAQS. An attainment area is defined by the CAA as any area that meets the NAAQS
and does not contribute to a nearby area that does not meet the NAAQS. Unclassifiable areas are
defined by the CAA as those that cannot be classified on the basis of available information as
meeting or not meeting the NAAQS. See CAA section 107(d)(l)(A)(i)-(iii).
In previous final actions, the EPA issued designations for the 2010 SO2 NAAQS for the entire
country.1 Once an area has been designated, the EPA Administrator, under CAA section
107(d)(3), "may at any time" notify a state that a designation should be revised "on the basis of
air quality data, planning and control considerations, or any other air quality-related
considerations the Administrator deems appropriate." CAA section 107(d)(3)(A).
Based on recent modeling analyses described below, Table 1 identifies portions of two counties
in Pennsylvania that EPA intends to redesignate from "attainment/unclassifiable" to
"nonattainment," and from "unclassifiable" to "nonattainment," for the 2010 SO2 NAAQS. As
explained in the technical analysis below, modeled nonattainment area is centered around
impacts in portions of Westmoreland and Cambria Counties resulting from SO2 emissions from
the Conemaugh Power Plant and Seward Station located in Indiana County, PA and is smaller
than the presumptive county-wide boundary.
Table 1-1 identifies EPA's intended revised designations for portions of Westmoreland and
Cambria Counties in Pennsylvania. It also lists current designations.
Table 1-1. Summary of the EPA's Intended Designations and the Curreni
Designation
Area/('011 illy
Curreni
Designation
Boundary
Curreni
Designation
EPA's Intended
Area Definition
(Boundary)
EPA's Intended
Designation
Westmoreland
Westmoreland
Entire County
Attainment/
Unclassifiable
Portion of
Westmoreland
County that
includes St.
Clair Township,
including
Seward borough
and New
Nonattainment
1 All areas of the U.S. were previously designated for the 2010 SO2 NAAQS in actions published on August 5, 2013
(78 FR 47191), July 12, 2016 (81 FR 45039), December 13, 2016 (81 FR 89870), December 21, 2017 (83 FR 1098),
March 28, 2018 (83 FR 14597) and March 26, 2021 (86 FR 16055).
4
-------
Florence
borough
Cambria
Cambria
Entire County
Unclassifiable
Portion of
Cambria County
that includes
Lower Yoder
Township
Nonattainment
2. General Approach and Schedule
CAA section 107(d)(3) identifies the schedule for the redesignation process. Per CAA section
107(d)(3)(A) and (B), EPA will notify the Commonwealth of Pennsylvania of our intended
redesignation, establishing a 120-day period for the state to respond. If EPA deems any
modifications necessary to its intended redesignation, including modifications based on the
state's response, EPA will inform Pennsylvania of such modification at least 60 days prior to
issuing the redesignation. Although not required by the Act, EPA will also make our intended
redesignation decision and supporting documentation for Westmoreland and Cambria Counties,
PA available to the general public and announce a 30-day public comment period in the Federal
Register.
A final redesignation of portions of Westmoreland and Cambria Counties to nonattainment for
the 2010 SO2 NAAQS would impose certain planning requirements on the Commonwealth of
Pennsylvania to reduce SO2 concentrations. These include, but are not limited to, the requirement
per CAA section 191(a) to submit, within 18 months of redesignation, a revision to the
Pennsylvania state implementation plan (SIP) that provides for attainment of the SO2 standard as
expeditiously as practicable, but no later than 5 years after the date of redesignation to
nonattainment, per CAA section 192(a).
EPA issued a designations guidance document for the 2010 primary SO2 NAAQS on March 20,
2015, which identified factors that EPA uses to evaluate whether areas are in violation of the
2010 SO2 NAAQS.2 The document also contains the factors that the EPA intends to evaluate in
determining the boundaries for this area. These factors include: 1) air quality characterization via
ambient monitoring and/or dispersion modeling results; 2) emissions-related data; 3)
meteorology; 4) geography and topography; and 5) jurisdictional boundaries. EPA also issued
guidance documents for designations for the 2010 primary SO2 NAAQS on July 22, 2016 and
September 5, 2019.3
2 https://www.epa.gov/sites/default/files/2016-04/documents/20150320so2designations.pdf
3 https://www.epa.gov/sites/default/files/2016-07/documents/areadesign.pdf
https://www.epa.gov/sites/production/files/2019-09/documents/ronnd 4 so2 designations memo 09-05-
2019_final.pdf
5
-------
3. Definitions
The following are definitions of important terms used in this document:
1) 2010 SO2 NAAQS - The primary NAAQS for SO2 promulgated in 2010. This NAAQS is
75 parts per billion (ppb), based on the 3-year average of the 99th percentile of the annual
distribution of daily maximum 1-hour average concentrations. See 40 CFR 50.17.
2) Design Value - a statistic computed according to the data handling procedures of the
NAAQS (in 40 CFR part 50 Appendix T) that, by comparison to the level of the NAAQS,
indicates whether the area is violating the 2010 SO2 NAAQS.
3) Designated nonattainment area -an area that, based on available information including
(but not limited to) monitoring data and/or appropriate modeling analyses, EPA has
determined either: (1) does not meet the 2010 SO2 NAAQS, or (2) contributes to ambient
air quality in a nearby area that does not meet the NAAQS.
4) Designated attainment/unclassifiable area - an area that, based on available information
including (but not limited to) appropriate monitoring data and/or appropriate modeling
analyses, EPA has determined meets the 2010 SO2 NAAQS and does not likely
contribute to ambient air quality in a nearby area that does not meet the NAAQS.
5) Designated unclassifiable area - an area for which the available information does not
allow EPA to determine whether the area meets the definition of a nonattainment area or
the definition of an attainment/unclassifiable area.
6) Modeled violation - a modeled design value impact above the 2010 SO2 NAAQS
demonstrated by air dispersion modeling.
7) Violating monitor - an ambient air monitor meeting 40 CFR parts 50, 53, and 58
requirements whose valid design value exceeds 75 ppb, based on data analysis conducted
in accordance with Appendix T of 40 CFR part 50.
8) We, our, and us - these refer to the EPA.
6
-------
4. Background
4.1. 2OIOSO2NAAQS
On June 2, 2010, the U. S. Environmental Protection Agency (EPA) Administrator signed a final
rule establishing a new SO2 primary NAAQS as a 1-hour standard of 75 ppb, based on a 3-year
average of the annual 99th percentile of daily maximum 1-hour average concentrations. 75 FR
35520 (June 22, 2010), codified at 40 CFR 50.17. This action also provided for revocation of the
existing 1971 primary annual and 24-hour standards, subject to certain conditions. 40 CFR
50.4(e). Following promulgation of a new or revised NAAQS, EPA is required by the CAA to
designate areas throughout the United States as attaining or not attaining the NAAQS; this
designation process is described in section 107(d)(l)-(2) of the CAA.
4.2. History of 2010 SO2 NAAQS Designations
On August 5, 2013, EPA promulgated initial air quality designations for 29 areas for the 2010
SO2 NAAQS (78 FR 47191). These designations became effective on October 4, 2013 and were
based on violating air quality monitoring data for calendar years 2009-2011, where there was
sufficient data to support a nonattainment designation. The Indiana, PA area, which consists of
all of Indiana County and a portion of Armstrong County, was designated as nonattainment in
this initial (first) round of designations, (78 FR 47191, Aug. 5, 2013).
On June 30, 2016, EPA completed a second round of area designations (81 FR 45039). This
second round did not address Cambria and Westmoreland Counties. On December 21, 2017,
EPA completed the third round of SO; designations during which Cambria County, PA was
designated unciassiliable, and Westmoreland County was designated attainment/unclassifiable
(81 FR 89870). During Round 3, Pennsylvania submitted a modeling analysis for Cambria
County, but due to inadequacies, the modeling could not be used to determine if the county could
be designated as attainment, and therefore an unci assi fi able designation was determined.
Pursuant to a court-ordered deadline of December 3 1, 2020, the Round 4 2010 SO; N AAQS
designations action was signed by the EPA Administrator, Andrew Wheeler, on December 21,
2020. For administrative purposes only, and in compliance with requirements of the Office of the
Federal Register, Acting Administrator Jane Nishida re-signed the same action on March 10,
2021 for publication in the Federal Register (86 FR 16055). This fourth round did not revisit the
round 3 designations of Cambria and Westmoreland Counties.
4.3. History of Westmoreland, Cambria, Indiana, PA 2010 SO2 Modeled
violations
7
-------
During the public comment period for the proposed approval of the Indiana, PA SO2 attainment
plan (83 FR 32606, July 13, 2018), the Sierra Club (in conjunction with the National Parks
Conservation Association, PennFuture, Earthjustice, and Clean Air Council) submitted a
modeling analysis using actual emissions for Conemaugh (coal-fired power plant) and Seward
Station (Coal waste facility) which claimed to show violations of the SO2 NAAQS outside of the
nonattainment area, beyond the eastern border of Indiana county within nearby portions of
Westmoreland and Cambria Counties. On October 19, 2020 (85 FR 66240), EPA finalized
approval of the Indiana, PA Area SO2 attainment plan, noting that the modeled violations outside
the nonattainment area were not an independent reason to disapprove the attainment plan.
On December 18, 2020, the Sierra Club, Clean Air Council, and PennFuture filed a petition for
judicial review with the U.S. Court of Appeals for the Third Circuit, challenging that final
approval.4 On April 5, 2021, EPA filed a motion for voluntary remand without vacatur of its
approval of the Indiana, PA SO2 attainment plan.
In a short order without any commentary, on August 17, 2021, the U.S. Court of Appeals for the
Third Circuit granted EPA's request for remand without vacatur of the final approval of
Pennsylvania's SO2 attainment plan for the Indiana, PA Nonattainment area, and required that
EPA take final action in response to the remand no later than one year from the date of the
court's order (i.e., by August 17, 2022).
After reconsideration, on March 17, 2022, EPA proposed partial disapproval and partial approval
of the Indiana, PA SO2 attainment plan, and during the public comment period received air
quality modeling (including modeling files) from the Sierra Club (in conjunction with the
National Parks Conservation Association, PennFuture, Earthjustice, and Clean Air Council)
using updated emissions data showing modeled violations in Westmoreland and Cambria
Counties due to Conemaugh and Seward sources located in Indiana County, PA. EPA also
received an air quality modeling report from Keystone-Conemaugh Projects, LLC (KEY-CON),
the licensee of Keystone and Conemaugh power plants, which used updated emissions from
Conemaugh and Seward plants and modeled concentrations that are below the NAAQS in
Westmoreland and Cambria Counties. On April 20, 2022, KEY-CON emailed the modeling files
to EPA. On August 18, 2022 (87 FR 502778), EPA finalized the partial disapproval and partial
approval of the Indiana, PA SO2 Attainment Plan. EPA explained that, although the attainment
plan must be disapproved for other reasons, the modeled violations in Cambria and
Westmoreland Counties were not a reason for that disapproval and noted that EPA was
considering taking additional regulatory action to remedy the modeled violations.
EPA then conducted two modeling analyses, discussed in more detail in the following sections,
which focused on the portions of Cambria and Westmoreland counties near the Conemaugh and
Seward power plants.
5. Technical Analysis
4 Sierra Club, et. al. v. EPA, Case No. 20-3568 (3rd Cir.).
8
-------
5.1. Overview
This section presents all the available air quality information for portions of Westmoreland and
Cambria Area.
As seen in Figure 5-1 below, the Conemaugh and Seward facilities are located in Indiana
County, while Cambria Cogen, Colver Power and Ebensburg Power are located in Cambria
County. The figure also shows the location of the Laurel Ridge (shaded red) that lies east of the
Indiana, PA 1-hour SO: nonattainment area. Cambria and Westmoreland counties were formally
designated as Unclassifiable and Attainment/Unclassifiable, respectfully, during EPA's Round 3
designations (83 FR 1098, January 9, 2018).
Figure 5-1. Map of Point Sources Discussed in this Technical Support Document (TSD)
Westmoreland-Cambria, PA Source Overview with Laurel Ridge
Jefferson
Clearfield
Legend
S02Sources
^^Conemaugh
^^^Homer City Generating Station
^^^Keystone
^^Seward Power Plant
N
•J Cambria Cogen
S
•J Colver Power Project
* j Ebensburg Power
Laurel Ridge
50 Kilometers
ZEPA
EPA conducted two assessments of the Westmoreland and Cambria areas focusing on the Laurel
Ridge specifically at the county boundaries of Westmoreland, Cambria and Indiana, using air
dispersion modeling software, i.e., AERMOD, analyzing actual emissions, which resulted in a
peak modeled SO2 concentration of 117.6 ppb. EPA's focus for this analysis was directed to a
small portion of Cambria County near Conemaugh and Seward plants. Our analysis does not
attempt to recharacterize the entirety of Cambria County, which was previously designated in
Round 3 as Unclassifiable due to inconclusive modeling. After careful review of EPA's
Armstrong
Keystone
Westmoreland
Colver
¦
Cambria Cogen (Shut Down)
Cambria 0
* IEbensburg
9
-------
assessment, and other third-party assessments, supporting documentation, and all available data,
the EPA intends to redesignate portions of the Westmoreland and Cambria Counties
nonattainment. Our reasoning for this conclusion is explained in later sections of this TSD.
5.2. Air Quality Modeling Analysis for Westmoreland and Cambria Counties,
PA
The discussion and analysis that follows below will reference the "SO2 NAAQS Designations
Modeling Technical Assistance Document" 5 (Modeling TAD) and the factors for evaluation
contained in the EPA's September 5, 2019, guidance, July 22, 2016, guidance and March 20,
2015, guidance, as appropriate.
For this area, the EPA received and considered 2 different modeling assessments plus EPA
provided its own 2 assessments. To avoid confusion in referring to these assessments, the
following table lists them, provides an identifier for the assessment that is used in the discussion
of the assessments that follow and identifies any distinguishing features of the modeling
assessments. Table 5.2-2 summarizes EPA's modeling assessment inputs. These apply to both
EPA simulations; one using the adjusted u-star option (no turbulence) and the other using the
Ash Site #1 collected turbulence measurements. At this time, EPA is not endorsing the use of the
adjusted u-star option or the turbulence measurements (see section 5.5.3 for additional
discussion). In this case, the use of adjusted u-star or turbulence both result in air quality
modeled design values above the NAAQS.
5 https://www.epa.gov/sites/prodiiction/files/2016-04/dociHnents/so2modelingtad.pdf.
10
-------
Table 5.2-1. Modeling Assessments for the Westmoreland and Cambria Area
Assessment
Subin illed In
Identifier I sed in
litis I SI)
Distinguishing or
Otherwise Kev
l-'eatures
Sierra Club (87 FR
502778)
Sierra Club
Actual emission
(2019-2021),
Johnstown-Cambria
County Airport
Meteorology data;
2016 Land cover
KEY-CON (87 FR
502778)
KEY-CON
Actual Emission
(2019-2021), Site-
Specific Meteorology,
turbulence, 1992 Land
Cover
EPA
EPA Site-Specific
Adjusted U-star
Modeling
Actual emissions (1
July 2017 through 30
June 2020), 1-year
Site-Specific
Meteorology, Adjust
U-star, 2016 Land
cover
EPA
EPA Site-Specific
Turbulence
Modeling
Actual emissions (1
July 2017 through 30
June 2020), 1-year
Site-Specific
Meteorology
Turbulence, 2016 Land
cover
11
-------
Table 5.2-2: Summary of AERMOD Modeling Input Parameters for EPA's Modeling for
the Westmoreland and Cambria Area
1 iiput Parameter
Value
AERMOD Version
22112
Dispersion Characteristics
Rural
Modeled Sources
5
Modeled Stacks
7
Modeled Structures
33
Modeled Fencelines
None
Total receptors
10,705
Emissions Type
Actual
Emissions Years
1 July 2017 through 30 June
2020
1 September 2015 through 31
August 2016
Met Data transposed to fit
Meteorology Years
emission period as per
Modeling TAD
Site-Specific/ automated
surface observation system
(ASOS)
Ash Site #1 &
NWS Station for Surface
Johnstown/Cambria County
Meteorology
ASOS
NWS Station Upper Air
Meteorology
Pittsburgh, PA
NWS Station for Calculating
Surface Characteristics
Ash Site #1
Methodology for Calculating
Background SO2 Concentration
Season by Hour of Day,
Strongstown, PA
Calculated Background SO2
Concentration
Varies
12
-------
5.2.1. Modeling Selection and Components
The EPA's Modeling TAD notes that for area designations for the 2010 SO2 NAAQS, the
AERMOD modeling system should be used, unless use of an alternative model can be justified.
The AERMOD modeling system contains the following components:
- AERMOD: the dispersion model
- AERMAP: the terrain processor for AERMOD
- AERMET: the meteorological data processor for AERMOD
- BPIPPRM: the building input processor
- AERMINUTE: a pre-processor to AERMET incorporating 1-minute automated surface
observation system (ASOS) wind data
- AERSURFACE: the surface characteristics processor for AERMET
- AERSCREEN: a screening version of AERMOD (not used for this analysis)
EPA used AERMOD version 22112 in regulatory default mode for its analysis. This was the
most current regulatory version of the model available at the time of preparation. AERMOD was
promulgated with the publication of EPA's revisions to the Guideline on Air Quality Models,
which was published in the Federal Register on January 17, 20176. AERMOD platform
component versions will be noted as they are discussed in the following sections. Individual
AERMOD component versions were current at the time EPA prepared this modeling analysis.
EPA chose to utilize meteorological data processed with the adjusted u-star (ADJ U*) option
within the AERMET preprocessor, excluding the site-specific Ash Site #1 turbulence
measurements as instructed following EPA guidance7. Meteorological processing, for this
modeling analysis, is therefore consistent with the preprocessing steps completed for the
Supplementary Analysis done for the southeast portion of the Indiana, PA nonattainment area.
Many of the elements used in EPA's modeling analysis were taken from AECOM's September
2020 modeling protocol and additional reports, the Pennsylvania Department of Environmental
Protection (PA DEP) review and summary materials, electronic files that were included in
Pennsylvania's original and supplemental SIP submittals, as well as other exchanges between
EPA Region 3, PA DEP, the affected sources and AECOM.
A brief summary of modeling elements (and their sources/adjustments) are listed here:
• Indiana County Source Information
o Building and stack information as provided by Pennsylvania. Information was
checked versus information provided by the Armstrong/Indiana County sources from
the SIP and Supplemental Analysis submissions,
o Hourly emissions and stack parameters provided to Pennsylvania by
Armstrong/Indiana County sources. Some adjustments were made based on EPA
Clean Air Markets Division data.
6 https://www.epa.gov/scraiii/clean-air-act-permit-modeling-giiidance
7 In accordance with section 4.7.6.5 of EPA's AERMET User's Guide
13
-------
• Cambria County Source Information
o Stack information taken from Pennsylvania's Round 3 designation modeling. No
building information was considered given the distance between these 3 sources and
the Laurel Ridge (downwash would not be important at distances greater than 10 km),
o Hourly source emissions from CAMD. Stack velocity and temperatures based on
linear relationships from source loading information as modeled by Pennsylvania for
its Round 3 designation analysis.
• AERMOD Receptor Grid
o EPA determined (locations) using current NED input files
• Meteorological Data
o Ash Site #1: 1-year of site-specific 100-m tower and SODAR data submitted as part
of Pennsylvania's Supplemental Analysis,
o Pittsburgh International Airport: Upper-air data with additional EPA processing to
account for missing surface observations,
o Sector defined surface characteristics from AERSURFACE using 2016 land use-land
cover, impervious surface, and tree canopy data,
o Final processing for one of EPA's analyses excluded Ash Site #1 turbulence
measurements with adjusted u-star option (to counter AERMOD's known
overpredictions under some stable low-wind speed conditions). EPA chose to run this
option because it is consistent with Pennsylvania's Indiana, PA S02 Attainment Plan
submission.
o A second air quality model run including the Ash Site #1 turbulence measurements
without the adjusted u-star processing option. EPA has shown the use of turbulence
and the adjusted u-star processing biases the model towards underprediction.
Therefore, the EPA has determined that the ADJU* option should not be used in
AERMET in combination with use of measured turbulence data because of the
observed tendency for model underpredictions resulting from the combined
influences of the ADJ U* and the turbulence parameters within the current model
formulation. (FR 82, 5187, January 17, 2017).
EPA runs use the most current version of AERMOD/AERMET (version 22112)
14
-------
5.2.2. Modeling Parameter: Rural or Urban Dispersion
For any dispersion modeling exercise, the "urban" or "rural" determination of a source is
important in determining the boundary layer characteristics that affect the model's prediction of
downwind concentrations. For SO2 modeling, the urban/rural determination is important because
AERMOD invokes a 4-hour half-life for urban SO2 sources. Section 6.3 of the Modeling TAD
details the procedures used to determine if a source is urban or rural based on land use or
population density.
Section 7.2.1.1 of Appendix W Guideline on Air Quality Models, outlines 2 methods that can be
used to choose the rural or urban options within AERMOD. One utilizes a population density
survey surrounding a source and the other uses Auer land use classifications surveyed
surrounding a source.
EPA utilized a land use survey to establish if the modeling analysis should use AERMOD's rural
or urban dispersion coefficients. We utilized the same land use/land cover information used to
determine the surface characteristics for the Ash Site #1 meteorological tower. AERSURFACE
was rerun using a 3 km survey area with only 1 sector (encompassing 360°) from the
Conemaugh and Seward stacks. The U.S. Geological Survey (USGS) 2016 Land Use/Land
Cover (LULC) data from the AERSURFACE log file was then examined to calculate the
percentage of developed land categories versus the total number of parcel counts within 3-km of
the Conemaugh and Seward stacks.
Figure 5.2-1 shows the USGS 2016 LULC within 3-km of the Conemaugh and Seward stacks.
The pink and red parcels on the figure represent developed land use categories. EPA counted the
low, medium and high developed categories as "urban". The remaining parcel categories are
treated as "rural". Table 5.2-3 summarizes the parcel count for each LULC category within 3-km
of the Conemaugh and Seward stacks. Percentages for the rural and urban LULC categories were
then calculated by dividing these values by the total number of parcels within the 3-km buffer
around the Conemaugh and Seward stacks.
Less than 7% of the USGS 2016 LULC categories within 3-km of either Conemaugh's or
Seward's stacks fall within the defined urban categories. EPA's analysis, therefore, will use
AERMOD's rural dispersion coefficients.
15
-------
Figure 5.2-1. USGS 2016 LULC within 3-km of Ash Site #1
Auer 3-km Survey ¦ Conemaugh & Seward I 2016 USGS LULC
Legend
201G U SGS LULC Categories
Op«r YiMf | | fc&ad Fonast
| Qcr4C4c&aa. Opar Space | I Scrufc Brush
_JjJ Ocnexcpac, low triersit» | | jas and He
Oareiacea. Mcdu'i intercity | | FagLfl&'Hay
Sources
Conemaugh
Oed&ous Forts!
Seward
3-km Auer Analysis
Buffer
Indiana, PA
Nonattainment Area
mssA
®nni
0 1.75 3.5 7 Kilometers 5 CDAE™«.ii
1 1 1 1 1 1 1 1 1 LZlir\:.sVr':''r,'r'' '
16
-------
Table 5.2-3. 3-km Survey of USGS 2016 LULC Categories for Conemaugh and Seward
AERSURFACE 3-km 2016 LULC Survey Results
Category
2016 LULC Description
Conemaugh
Seward
11
Open Water
835
815
21
Developed, Open Space
1,607
2,000
22
Developed, Low Intensity
899
1,126
23
Developed, Medium Intensity
676
515
24
Developed, High Intensity
471
372
31
Barren Land (Rock/Sand/Clay)
653
302
41
Deciduous Forest
20,400
20,198
42
Evergreen Forest
37
41
43
Mixed Forest
2,840
3,097
52
Shrub/Scrub
81
107
71
Grasslands/Herbaceous
725
736
81
Pasture/Hay
1,623
1,752
82
Cultivated Crops
212
125
90
Woody Wetlands
310
223
95
Emergent Herbaceous Wetland
32
3
22, 23, 24
Urban
2,046
2,013
All Others
Rural
29,355
29,399
% Urban
6.52%
6.41%
5.2.3. Modeling Parameter: Area of Analysis (Receptor Grid)
EPA's Modeling TAD recommends that the first step towards characterization of air quality in
the area around a source or group of sources is to determine the extent of the area of analysis and
the spacing of the receptor grid. Considerations presented in the Modeling TAD include but are
not limited to the location of the SO2 emission sources or facilities considered for modeling; the
extent of significant concentration gradients due to the influence of nearby sources; and
sufficient receptor coverage and density to adequately capture and resolve the model predicted
maximum SO2 concentrations.
17
-------
A preprocessor program, AERMAP, was developed to process terrain data in conjunction with a
layout of receptors and sources to be used in AERMOD control files. The terrain elevation for
each receptor, and emission source was determined using USGS 1/3 arc second National
Elevation Data (NED). The NED, obtained from the U.S. Geological Survey (USGS), has terrain
elevations at approximately 10-meter intervals. A total of 4 NED files were downloaded and
processed following directions on EPA's Support Center for Regulatory Atmospheric Modeling
(SCRAM) website. NED files downloaded from USGS are not directly usable by AERMAP and
must be in an uncompressed format. The 4 NED files were converted to this uncompressed
format in accordance with instructions posted on SCRAM8. These uncompressed files served as
input data for AERMAP to determine the model receptor and source elevation heights.
AERMAP also assigns hill height scales to all receptors. Hill height scales are used to calculate
the critical dividing streamline height for each model receptor.
The model receptor grid used in EPA's modeling analysis was confined to portions of Cambria
and Westmoreland counties within approximately 15 km of the Conemaugh and Seward power
plants. It is not the same as the receptor grid described in AECOM's September 2020 modeling
protocol. The model receptor grid includes portions of the Chestnut Ridge to the west and the
Laurel Ridge, which lies just east of Conemaugh and Seward. Receptor spacing was initially set
at 360 meters creating a coarse Cartesian grid over the previously described area. A finer 90-
meter spaced grid was created to cover most of the Laurel Ridge which was then clipped to only
cover portions of the Laurel Ridge inside Westmoreland County. The county border between
Cambria and Westmoreland counties is roughly marked by the ridgeline of the Laurel Ridge.
Two additional 45-meter Cartesian grids were created within the 90-m grid to provide additional
model receptors near the areas of maximum modeled concentrations. The northern 45-m grid
was also confined to portions within Westmoreland County.
The initial 360-m Cartesian grid was produced using R9, filtered by distance from Conemaugh
and Seward then imported into GIS and clipped to be within either Cambria or Westmoreland
counties. The 90-m and 45-m grids were similarly constructed. EPA added three 22.5-m
cartesian receptor grids around the areas on the Laurel Ridge with the highest model values to
ensure that the final receptor grid captured the maximum modeled concentration as described in
section 9.2.2 (d) of Appendix W.
Each grid was run through AERMAP (version 18181) then combined removing any identical
model receptors the grids had in common. The final grid contains 10,705 individual model
receptor points and should be adequate to properly resolve the maximum model concentrations
from Conemaugh and Seward.
8 See Elevation Data Access section of https://www.epa.gov/scram/air~qnalitv-dispersion-modeling~related~model~
suppo rt~pro gra tns#ae rmap
9 R Core Team (2021). R: A language and environment for statistical computing. R Foundation for Statistical
Computing, Vienna, Austria. URL https://www.R-project.org/.
18
-------
AERMAP runs were broken down into smaller grid sections (within the same modeling domain)
to create more manageable processing times and prevent losses that could occur during extended
model run times. EPA notes that AERMAP run times can be exceptionally long, especially over
network connected computers. Any network interruptions during a model run would lead to loss
of data and necessitate a restarting of the simulation. Long run times on local computer drives
can also be interrupted by the computer's power saving settings. For these reasons, AERMAP
run times were generally kept to 8 hours or less. The final model receptor grid was a combination
of all of the smaller grids processed in AERMAP.
Figure 5.2-2 displays the area that contains the model receptor grid used in the EPA modeling
analysis. The figure also shows the sources included in EPA's modeling analysis and the
Strongstown, PA SO2 monitor. A close up of the actual model receptor locations along the
Laurel Ridge in both Cambria and Westmoreland counties is shown in Figure 5.2-3. Both figures
also display the local terrain elevations.
Both Conemaugh and Seward are located along the Conemaugh River in Indiana County and are
contained within the Ligonier Valley. The Chestnut Ridge lies to the west of these facilities and
the Laurel Ridge lies to the east. Both terrain features largely pinch out to the north but extend
many miles to the south. Water drainage is to the west, eventually becoming part of the Ohio
River Basin. The Conemaugh River bisects both ridges creating the Conemaugh River Gorge as
it passes through the Laurel Ridge. This pattern indicates the general drainage patterns were
established before the land experienced uplifting during the Cenozoic time period; the river
systems incised downward into the land as it was raised upwards.
There are no fence-line receptors included in EPA's modeling receptor grid. Both Conemaugh
and Seward, along with the 3 waste-coal units in Cambria County, do not contain a plant
footprint inside the formal model receptor grid. Therefore, each source's ambient air boundary
does not need to be delineated within EPA's model receptor grid.
19
-------
Figure 5.2-2. Model Receptor Area for EPA Grid in Cambria and Westmoreland Counties
Indiana
Seward
Westmoreland,
III!!!
EPA Modeling Analysis of Portions of the Laurel Ridge Near Conemaugh and Seward
JZambria£ ogenTtSlruflD.tmn )1
LfcDensp^!g|
Iteam (Trial
Legend
A strongstowi SOj Montior
B3EPA Modeling G rid
—Indiana, PA Nonattainment
Area
Elevation
Meters
High : 2035
20 Kilometers
—I
sera.;;—
Figure 5.2-3. EPA Model Receptor Grid Along Portions of the Laurel Ridge
m&mi
Analysis of Portions of the Laurel Ridge Near Conemaugh and Seward
Legend
SO2 Sources
^^^Conemaugh
~ Model Receptors
^—Indiana, PA Nonattainment
Elevation
Meters
^ High : 2035
Low: -99
6 Kilometers
AEPA
20
-------
5.2.4. Modeling Parameter: Source Characterization
Section 6 of the Modeling TAD offers recommendations on source characterization including
source types, use of accurate stack parameters, inclusion of building dimensions for building
downwash (if warranted), and the use of actual stack heights with actual emissions or following
GEP policy with allowable emissions.
The following are brief facility descriptions for each source included in EPA's modeling
analysis:
Conemaugh: A traditional coal-fired boiler power plant burning western Pennsylvania
bituminous coal. Two coal-fired units were commissioned in 1970 and 1971. A wet flue
gas desulfurization (FGD) system was installed in the 1990s to control SO2 emissions. A
new FGD stack was added to properly handle saturated plume conditions exhausted from
the FGD control system. Conemaugh can burn approximately 4 million tons of
bituminous coal per year to produce about 1,800 megawatts of electricity (for the PJM
managed electrical grid). The plant is located in Indiana County near New Florence, PA.
Seward: Utilizes waste coal feedstock, which is sometimes referred to as GOB or
garbage of bituminous. The plant burns waste coal in 2 circulating fluidized bed (CFB)
boilers, which were commissioned around 2004. SO2 emissions are controlled through
lime injection into the CFB units and into the flue gas prior to the facility's baghouse
unit. Waste coal can have highly variable british thermal unit or BTU values along with
percent sulfur values. Seward is the world's largest waste coal facility and can produce
approximately 525 megawatts of electricity. The facility's current stack was built for a
previous coal fired power plant; there has been a power plant operating on this site since
the early 1900s. The plant is located in Indiana County near Seward, PA.
Cambria Cogen: A 85 MW, base load, waste-coal fired power plant located near
Ebensburg, PA in Cambria County. The plant has 2 circulating fluidized bed (CFB)
boilers that began commercial operations in 1991. SO2 emissions are controlled via lime
injection into the CFB units. The facility was deactivated from the PJM electric grid10 in
2019. A Retirement Unit Exemption form was filed with EPA's Clean Air Markets
Division or CAMD notifying the units' deactivation. This became effective in September
of 2020. Emissions from this source ceased after the 2nd quarter of 2019 for EPA's
simulation.
Colver Power or Colver Green Energy: A 118 MW, base load, waste-coal fired power
plant located near Colver, PA in Cambria County. At full operation, the plant's CFB unit
can consume 700,000 tons of waste-coal per year. SO2 emissions are controlled via lime
injection into the CFB unit. The plant began operations in 1995.
10 https://www.pim.com/planning/services-reaiiests/gen-deactivations
21
-------
Ebensburg Power: A 50-megawatt waste coal plant constructed in 1991 in Cambria
County near Ebensburg, PA. The plant utilizes CFB combustion technology to force hot
air and limestone into the boiler to burn low quality refuse coal mined from abandoned
piles located throughout Western Pennsylvania. The limestone in the boiler captures SO2
emissions in a controlled fashion rather than the free release that occurs over time when
refuse piles spontaneously combust at the mine sites. All (7) stacks included in EPA's
modeling analysis were modeled as point sources in AERMOD. EPA's modeling analysis
largely borrowed previous building downwash analysis from Pennsylvania's original SIP
and Supplementary Analysis submissions. Both of these analyses used EPA's Building
Profile Input Program or BPIP (version 04274). Building positions for Seward were
adjusted based on a visual inspection completed by the PA DEP and shared with EPA in
March of 2022. No building downwash was included for the Cambia County sources.
The Cambria County sources are greater than 10 km from the EPA model receptor grid.
Any impact of building downwash from these sources within the model receptor grid was
therefore expected to be minimal.
Each source's stack(s) and building information were entered into BPIP to generate building
downwash information utilized in AERMOD. BPIP output also listed GEP formula height
calculations for each stack. EPA's modeling analysis only included downwash information from
Conemaugh and Seward.
BPIP GEP formula heights for Conemaugh's FGD stack came out higher than the actual stack
height; 160 m versus BPIP's GEP calculated value of 173.36 m. Conemaugh's stack, built during
the FGD installation in the mid-1990s, appears to comply with GEP.
One change was made to the Conemaugh stack for EPA's 3-year modeling analysis. As will be
explained in a following section, Conemaugh was modeled using 3 separate stack options: a
virtual merged stack when both Conemaugh units are operating simultaneously, and two single
stacks for each individual unit when only one unit is operating (when only one unit is operating,
each unit emits out of its own separate stack). When both units are operating, their plumes
become merged shortly after exiting their individual flues. This merged plume enhances lift
which can be accounted for by merging the stacks (in the modeling analysis) using an equivalent
diameter stack to enhance the exit velocity. Stack locations for each stack were identical for
modeling in BPIP. In reality, Conemaugh's FGD stack is a single dual-flue stack with each unit
having 1 flue. The exact location of each flue is not known while the merged stack (with
equivalent area diameter) is set at the central portion of the FGD stack. In reality, the individual
unit flues are probably several meters from the center of the FGD stack. Any discrepancy in the
exact locations of the unit flues within the stack is not expected to make any significant
differences in the BPIP downwash calculations.
5.2.5. Modeling Parameter: Emissions
22
-------
The EPA's Modeling TAD notes that for the purpose of modeling to characterize air quality for
use in designations, the recommended approach is to use the most recent 3 years of actual
emissions data and concurrent meteorological data. However, the TAD also indicates that it
would be acceptable to use allowable emissions in the form of the most recently permitted
(referred to as PTE or allowable) emissions rate that is federally enforceable and effective. The
EPA believes that continuous emissions monitoring systems (CEMS) data provide acceptable
historical emissions information when they are available. These data are available for many
electric generating units. In the absence of CEMS data, the EPA's Modeling TAD highly
encourages the use of AERMOD's hourly varying emissions keyword HOUREMIS, or through
the use of AERMOD's variable emissions factors keyword EMISFACT. When choosing one of
these methods, the EPA recommends using detailed throughput, operating schedules, and
emissions information from the impacted source(s).
In certain instances, states and other interested parties may find that it is more advantageous or
simpler to use PTE rates as part of their modeling runs. For example, where a facility has
recently adopted a new federally enforceable emissions limit or implemented other federally
enforceable mechanisms and control technologies to limit SO2 emissions to a level that indicates
compliance with the NAAQS, the state may choose to model PTE rates. These new limits or
conditions may be used in the application of AERMOD for the purposes of modeling for
designations, even if the source has not been subject to these limits for the entirety of the most
recent 3 calendar years. In these cases, the Modeling TAD notes that a state should be able to
find the necessary emissions information for designations-related modeling in the existing SO2
emissions inventories used for permitting or SIP planning demonstrations. In the event that these
short-term emissions are not readily available, they may be calculated using the methodology in
Table 8-1 of Appendix W to 40 CFR Part 51 titled, "Guideline on Air Quality Models."
EPA utilized actual hourly emissions for the Conemaugh and Seward power plants in its
modeling analysis of portions of Cambria and Westmoreland counties near these 2 power plants.
We also included actual emissions from 3 waste coal facilities in the northern part of Cambria
County east of the Indiana, PA nonattainment area. These latter sources were modeled as part of
EPA's Data Requirement Rule (DRR) Round 3 designations. The DRR was set up to better
characterize ambient air SO2 concentrations near large polluting sources. Cambria County
sources included Cambria Cogen, Colver Power and Ebensburg Power. All 3 of these sources
burn waste coal via CFB boiler units.
EPA's modeling analysis included actual hourly emissions over a 3-year period, 1 July 2017
through 30 June 2020. This was the period included in the September 2020 AECOM modeling
protocol submittal. Hourly SO2 emissions and other data from CAMD over the identical time
period were used to "check" these values. The remainder of this section will provide an overview
of the construction of Conemaugh and Seward's emissions profiles, along with the 3 Cambria
County sources, including actual hourly SO2 emissions, stack temperatures and stack velocities
used in EPA's modeling analysis.
23
-------
EPA compared actual hourly emissions for Conemaugh and Seward from the AECOM protocol
submittal with hourly emissions for each source as reported to EPA's CAMD database. Note that
the AECOM protocol only included hourly data for Conemaugh and Seward, not the other waste
coal sources in neighboring Cambria County. Hourly SO2 emissions for Conemaugh's 2 coal-
fired units and Seward's combined CFB units were largely identical between the protocol CEMS
data and CAMD. Discrepancies between the 2 data sets for each source amounted to fewer than
72 hours over the 3-year simulation period. In all instances, when actual hourly emission values
differed, the protocol hourly emissions were less than the values reported to CAMD. For some
hours, the protocol emission values were lower than the corresponding CAMD database values
even when the information in CAMD indicated hourly SO2 emissions were marked as valid.
For hours where SO2 emissions did not match, EPA substituted the higher CAMD hourly values
over using values in the modeling protocol submittal. We felt this would be conservative. The
infrequency of these differences makes it highly unlikely that the higher CAMD hourly emission
values will cause any significant changes to the model results. This would only be true if the
increased hourly emissions occurred during the worst-case meteorological conditions that
determine the model simulation's 1-hr SO2 design value.
In addition to hourly emission rates, EPA's modeling analysis needed hourly stack parameters
for Conemaugh and Seward. This information is generally not available from the CAMD
database, though CAMD does contain information on flow rate measurement validity. Hourly
stack exhaust velocity and temperature were largely taken from the AECOM September 2020
protocol submittal. Missing hourly stack exhaust values were substituted using a linear
relationship developed between each unit's heat input and corresponding stack velocity. This
largely follows the same procedures used in AECOM's modeling protocol to fill in missing stack
velocity data. Our analysis of valid flow rates taken from the CAMD database indicates each
source had between 50 to 300 hours of invalid flow rate measurements (unusable stack exhaust
rates). This indicates only a small fraction of the 3-year simulation period used stack exhaust
flows inferred from operational data.
There was one last difference between the AECOM protocol and EPA's modeling analysis that
will be described here. The AECOM protocol hourly emissions file utilized a merged stack for
Conemaugh. Both Conemaugh unit emissions, stack exhaust rates and stack temperatures were
combined into a (virtual) merged stack unit with an equivalent area diameter and flow weighted
temperature representing Conemaugh's actual dual-flued FGD stack. This is permitted under
EPA's October 10, 1985 memo entitled Questions and Answers on Implementing the Revised
Stack Height Regulation and is more specifically described in EPA's answer provided to
question 19, item 211.
11 https://www.epa.gov/sites/default/files/2015-07/documents/reinders.pdf
24
-------
EPA found some fault with the protocol's use of a merged stack for Conemaugh. More
specifically, the use of a merged stack when only 1 of Conemaugh's 2 coal-fired units are
operating. We note, however, that the use of a merged stack when both units are operating is
acceptable. EPA Region 3 made this comment when it reviewed AECOM's modeling protocol
submittal and included it as a formal comment that was provided to Pennsylvania in February of
2021. It's our understanding that EPA and PA DEP comments were both forwarded to AECOM.
Conemaugh's hourly emissions were modeled using the individual flues for each unit and a
merged stack when both units were operating simultaneously. This approach mimics what was
done for the Brandon Shores units in the DRR Round 2 modeling submitted by the State of
Maryland for its Round 2 SO2 designations modeling for Anne Arundel-Baltimore County12.
AECOM's modeling protocol did not address the Cambria County (waste-coal) sources so there
were no hourly SO2 emission rates, stack temperature or stack velocity information to utilize.
EPA downloaded the reported CAMD hourly SO2 emission rates for Cambria Cogen, Colver
Power and Ebensburg Power. Physical stack parameters (stack positions, stack heights and stack
diameters) for these sources were taken from Pennsylvania's Round 3 DRR modeling analysis.
Hourly stack velocity and temperature information is not available from the CAMD database.
EPA took the previous modeling analysis Pennsylvania did as part of their DRR Round 3
modeling analysis and established linear trendlines between these variables and the
corresponding CAMD source unit hourly heat inputs. Additional details regarding the
construction of the modeled hourly source profiles for Conemaugh, Seward and the Cambria
County sources are included in the following subsections.
5.2.5.1. Model Input Parameters for Conemaugh
Conemaugh Modeled Hourly Emission Rates: EPA downloaded (actual) hourly emissions for
both Conemaugh coal-fired units (units 1 & 2) from the CAMD database. Hourly SO2 emissions
from Conemaugh's combined units from 2010-2020 are shown in Figure 5.2-4. The green
shaded area of the graph is EPA's 3-year model period corresponding to Conemaugh's protocol
spreadsheet information. Figure 5.2-5 shows Conemaugh's combined unit emissions over the 3-
year simulation period (the green shaded area on Figure 5.2-4).
12 https://www.epa.gov/siilfur-dioxide-designations/so2-designations-roniKl-2-marvlaiKl-state-recommendation-aiKl-
epa-response See Maryland Round 2 State Recommendation Attachment 1 and EPA Response to Maryland Round 2
Recommendation Attachment.
25
-------
Figure 5.2-4. Conemaugh Hourly CAMD SO2 Emissions from 2010-2020
Conemaugh Hourly Emissions
CAMD Part 75 2010-20
| 10,000-
O
CO
Measured
Calculated
CEV
Jinlk.
2015 2016
Date
Figure 5.2-5. Conemaugh Hourly CAMD SO2 Emissions over 3-year Simulation Period
Conemaugh Hourly Emissions
CAMD Part 75 1 July 2017 to 30 June 2020
Measured
Calculated
CEV
E
ID
o
CO
&
&
26
-------
Conemaugh's hourly CAMD SO2 emissions shown in both figures include the facility's modeled
Critical Emission Value or CEV. This represents the maximum hourly emission rate that is
modeled within the Indiana, PA nonattainment area that does not exceed the 1-hr SO2 NAAQS
of 75 ppb. This was included for reference only and does not reflect what the CEV would be for
areas outside the Indiana, PA nonattainment area. Conemaugh's CEV determined in
Pennsylvania's Supplemental Analysis was 3,381 lbs/hr13. As noted in the hourly emissions
figures, Conemaugh's actual hourly SO2 emission rate rarely exceeds its model-defined CEV
threshold.
The figures showing Conemaugh's (combined) hourly emission rates also include information
regarding whether the (CAMD) hourly emission rate was measured or calculated. This is based
on method of determination codes (MODC). These codes are listed in Table 4a section 75.57(c)
of 40 CFR Part 75. Hourly SO2 emissions are based on monitor concentration and flow rates
measured by its CEMS. To have a valid measured value, both the monitor concentration and
flow rate instruments must be functioning properly. If either or both of the instruments
malfunction and there is not a redundant back up measurement available, the emission rate is
calculated based on a predefined methodology. Thus, hourly emissions are either "measured"
when all instruments are functioning in a given hour or "calculated" if there are instrument
malfunctions. Calculated emission estimates can assume worst-case conditions if instrument
down times are significant. This scenario ensures CEMS units are functioning most of the time;
otherwise, a source will be forced to purchase emission offsets if it exceeds its yearly budget.
Hours with "calculated" emission values could be much higher than what the actual emissions
are. As seen on the graph, nearly all of Conemaugh's hourly emissions are measured (have valid
MODC).
EPA used the protocol submittal information as a basis for the development of its (actual) hourly
emissions for Conemaugh. Hourly SO2 emission rates for units 1 and 2 were compared to
corresponding hourly emissions from the CAMD database. Nearly all of Conemaugh's hourly
emission rates from the protocol submittal matched the corresponding CAMD emission rates.
Approximately 1-2 days' worth of hourly SO2 emissions, however, were not the same over the 3-
year simulation period.
13 See Table ES-1 of AECOM (2019)
27
-------
Table 5.2-4 summarized the results of the comparison of hourly SO2 emission rates from the
protocol submittal, also referred to as CEMS, and CAMD database for both Conemaugh and
Seward. There are several dozen hours where the protocol and CAMD emission rates do not
match. In all cases, the CAMD hourly emission rates were higher than the corresponding
protocol hourly emission rates. Hours that did not match were divided between hours where the
CAMD emission rate was measured (valid MODC) and calculated (invalid MODC). EPA is
unsure why there are differences in hourly emission rates for the hours that have measured
values according to CAMD, though these times make up a small fraction of the 3-year simulation
period. EPA replaced Conemaugh's hourly SO2 emissions with CAMD values for any hours over
the 3-year simulation period where there were mismatches between the 2 databases. For all
mismatched hours, the CAMD emissions exceeded the CEMS values. Model concentrations,
therefore, would be higher using the CAMD values versus the CEMS values.
Table 5.2-4. Summary of Protocol and CAMD Emission Differences
Comparison of CEMS and CAMD Hourly S02 Emissions
1 July 2017 to 30 June 2020
Source
Total Operating
Hours
Hours CAMD > CEMS,
Measured
Hours CAMD > CEMS,
Calculated
Conemaugh Unit 1
21,135
30
7
Conemaugh Unit 2
22,892
18
9
Seward
20,856
39
11
Conemaugh Modeled Hourly Stack Parameters: Stack parameters including stack
temperatures and velocities for EPA's modeling analysis were largely taken from the AECOM
protocol submittal. These values are generally not reported to any public data system such as
CAMD. Flow rates for most CAMD reporting sources are submitted but these values are
typically reported in standard cubic feet per hour. A conversion needs to be applied to transform
these reported flow rates to actual cubic feet per hour, which could then be used to calculate
stack exhaust velocities. Flow rate MODC, however, can still be (and were) used to flag hours
where actual stack flow rates are not measured.
Stack temperatures were available for both Conemaugh units over the entire 3-year modeled
period. Stack velocities for each unit were available for most hours in the protocol submittal. We
note that there was an adjustment made to Conemaugh's stack flows due to differences in the
diameters between the flow rate measurement site and Conemaugh's stack top openings. The
flow measurement site's diameter was 33 feet while the stack top opening was 5 feet narrower.
This meant that stack top velocities had to be increased by a ratio of 33 feet divided by 28 feet
then raised to the second power (for both Conemaugh units).
28
-------
EPA used Conemaugh's flow rate MODC to flag hours with possible invalid stack velocities.
Table 5.2-5 summarizes the hours with invalid flow rate MODC over the 3-year simulation
period for both Conemaugh units and Seward. MODC descriptions from table 4a from Part 75
are also included in the table.
29
-------
Table 5.2-5. Summary of Flow Rate MODC over 3-Year Simulation Period (Hours)
Summary of Flow MODC Part 75 (Table 4a)
1 July 2017 to 30 June 2020
MODC
Part 75 Explanation
Conemaugh Unit 1
Conemaugh Unit 2
Seward
00
Off
5,169
3,412
5,448
01
Certified Primary Instrument
20,992
22,619
20,793
06
Average of the Prior and Following Hour
-
-
6
08
90th % Hourly Flow Rate
-
32
25
10
Maximum Hourly Flow Rate
-
-
2
11
Average of Hourly Flow Rates in
Applicable Lookback Period
143
70
30
20
200% of the Full-Scale Range Setting
-
171
-
CAMD flow rate MODC indicate several hundred hours of possible invalid stack velocities for
each of Conemaugh's coal-fired units. Several MODC indicate flow rates during some of the
hours with invalid MODC can be quite high, often on the extreme end of the unit's overall
distribution of measured flow rates or even exceeding them. For the Part 75 program, using
exaggerated flow rates will result in exaggerated hourly emission rates. CAMD hourly emission
rates are based on a concentration measurement and a flow rate measurement. For modeling
purposes, however, using an exaggerated flow rate can enhance stack dispersion characteristics
since higher stack velocities (especially from tall stacks) generally lower final model
concentrations by lofting the initial plume higher above the surface. Use of exaggerated flow
rates for invalid MODC hours, therefore, should be avoided for any modeling analysis.
To replace missing stack velocity values or values where exaggerated flow rates may be present,
EPA used the same method to generate a more realistic stack velocity that was used in
AECOM's September 2020 protocol submittal. A surrogate stack velocity was substituted for
hours with missing or invalid flow rate measurements. Surrogate values were based on the unit's
remaining (valid) measured heat input and stack velocities. In general, the higher the unit heat
input (the heat released when coal is burned in the boiler unit) the higher the flow rate. A
comparison of measured unit heat input verses stack velocity indicates an excellent linear
correlation.
Figure 5.2-6 displays scatter plots of each unit's measured heat input in millions of British
thermal units (mmBtu) versus its stack velocity (flow rate measurement). A linear trend line was
fitted to the data and shows an excellent linear correlation (R2 values very close to 1). The linear
fit equations were then used to fill in all hours with invalid flow rate MODC. As an example, a
missing stack velocity for unit 1 would be replaced by using the corresponding hour heat input
value in the linear fitted trend line equation:
Missing Unit 1 Stack Velocity (m/s) = 1.8 + 0.0024 * Unit 1 Heat Input (mmBtu/hr)
30
-------
Figure 5.2-6. Conemaugh Heat Input versus Stack Velocity and Linear Trend Lines
Conemaugh Unit 1 Heat Input versus Adjusted Stack Velocity
Select CEMS Data 1 July 2017 to 30 June 2020
Conemaugh Unit 2 Heat Input versus Adjusted Stack Velocity
Select CEMS Data 1 July 2017 to 30 June 2020
Heat Input (mmBtu/hr)
5,000 7,500
Heat Input (mmBtu/hr)
As mentioned previously, Conemaugh's emissions were passed through a virtual merged stack in
the protocol submittal's final AERMOD-ready hourly emission file. Merged stack parameters
were defined by considering each unit's stack temperature and final adjusted stack velocity (to
account for the diameter differences between the unit's flow rate measurement site and stack top
opening). Conemaugh's merged stack temperature was calculated using each unit's stack
temperature weighted by the unit's flow rate. If one unit's flow rate was higher than the other,
the merged stack modeled temperature would be slightly closer to the unit with the higher stack
velocity than the average of the 2 units' temperatures. The merged stack velocity was just the
average of each unit's stack velocity for Conemaugh in accordance with the September 2020
protocol.
EPA has expressed an issue with this approach. Mainly that using a merged stack may
misrepresent stack dispersion characteristics when only a single unit is operating at Conemaugh.
We do think, however, that using a merged stack is appropriate when both units are operating,
but the actual stack characteristics should be modeled when only 1 unit is operating.
Using a merged stack when a single unit is operating may introduce some errors in actual stack
dispersion. EPA processed both Conemaugh's protocol submittal stack information and its actual
hourly AERMOD input file to illustrate the impact of a unit shut down on merged stack
parameters. Table 5.2-6 shows a segment of modeled hours slightly before Conemaugh unit 1
shuts down and the hours after the unit ceases burning coal. In the table, unit 1, 2 and the merged
stack emissions are highlighted in yellow, temperatures in green and stack velocities in blue.
Unit 1 and 2 parameters are from the protocol spreadsheets and the merged stack parameters are
from the AERMOD ready hourly emissions file provided in the protocol submittal.
31
-------
Table 5.2-6. Merged Stack Parameters During Unit Shut Down
Conemaugh Hourly Emissions (by Unit) versus AERMOD Input File
Date
HR
Ul.lbshr
Ul.mmBtu
U1.F
m.fs
U2.lbshr
U2.mmBtu
U2.F
U2.fs
Q.lbshr
Temp.F
Vel.fs
2017-07-06
4
166.7
4,707.70
128.1
41.3
248.4
5,052.4
123.5
43.1
415.1
125.7
42.2
2017-07-06
5
177.9
4,736.10
128.5
41.1
242.8
5,050.5
123.5
43.1
420.7
125.9
42.1
2017-07-06
6
356.4
5,535.90
128.7
47.9
447.1
5,766.8
123.8
48.9
803.5
126.2
48.4
2017-07-06
7
1,242.3
8,102.10
130.2
67.6
1,339.6
8,112.2
124.4
68.2
2,581.9
127.2
67.9
2017-07-06
8
1,462.2
8,684.60
131.0
71.8
1,337.9
8,495.6
125.1
72.0
2,800.1
128.0
71.9
2017-07-06
9
364.5
2,219.76
131.2
57.2
1,364.3
8,532.5
125.1
72.2
1,728.8
127.7
64.7
2017-07-06
10
-
-
117.5
37.6
1,289.4
8,478.0
125.2
71.8
1,289.4
122.4
54.7
2017-07-06
11
-
-
112.4
41.7
1,291.9
8,451.1
125.3
71.0
1,291.9
120.2
56.3
2017-07-06
12
-
-
108.7
41.4
1,356.4
8,492.5
125.5
70.7
1,356.4
118.7
56.0
2017-07-06
13
-
-
105.8
41.0
1,159.0
8,535.3
125.0
71.7
1,159.0
117.3
56.3
2017-07-06
14
-
-
102.9
40.3
1,255.3
8,523.8
125.4
71.7
1,255.3
116.3
56.0
2017-07-06
15
-
-
100.4
39.4
1,334.0
8,573.8
125.1
72.8
1,334.0
115.1
56.1
2017-07-06
16
-
-
99.3
41.5
1,240.4
8,530.9
125.2
71.8
1,240.4
114.3
56.7
2017-07-06
17
-
-
97.4
41.5
1,330.1
8,479.2
125.2
71.4
1,330.1
113.3
56.5
2017-07-06
18
-
-
94.9
41.7
1,215.9
8,458.8
125.1
71.8
1,215.9
112.0
56.7
2017-07-06
19
-
-
92.2
41.3
1,385.1
8,553.4
125.0
72.1
1,385.1
110.7
56.7
2017-07-06
20
-
-
89.8
41.0
1,227.2
8,524.0
124.9
73.1
1,227.2
109.5
57.0
2017-07-06
21
-
-
88.0
45.1
1,034.2
8,241.0
125.0
70.6
1,034.2
107.4
57.9
2017-07-06
22
-
-
87.4
46.3
536.4
6,457.7
124.3
55.7
536.4
104.3
51.0
2017-07-06
23
-
-
87.8
46.4
314.9
5,448.1
123.7
47.9
314.9
103.0
47.2
2017-07-07
0
-
-
86.7
46.3
222.6
5,013.1
123.4
44.2
222.6
101.4
45.2
2017-07-07
1
-
-
85.1
45.8
216.9
4,980.7
122.9
44.7
216.9
100.3
45.3
2017-07-07
2
-
-
83.7
34.2
218.7
5,005.1
122.7
44.4
218.7
102.0
39.3
2017-07-07
3
-
-
83.1
1.8
264.5
5,017.3
122.9
44.6
264.5
120.7
23.2
2017-07-07
4
-
-
82.9
0.3
224.1
5,000.5
123.0
44.4
224.1
122.6
22.4
2017-07-07
5
-
-
87.2
1.4
275.9
5,187.2
122.9
44.9
275.9
121.4
23.1
2017-07-07
6
-
-
86.5
2.8
447.1
5,636.5
123.4
48.5
447.1
120.6
25.6
2017-07-07
7
-
-
86.6
2.9
489.4
6,051.8
123.5
52.4
489.4
120.8
27.6
2017-07-07
8
-
-
86.7
2.8
1,213.3
7,857.0
123.4
65.7
1,213.3
121.3
34.2
32
-------
Conemaugh unit 1 ceased burning coal after hour 9 on July 6, 2017 (SO2 emissions go to 0 after
this hour). While the unit is no longer burning coal based on its SO2 emission rate, the unit
continues to report significant exhaust velocities and the unit temperature remains elevated
though those values begin to decline after hour 9. Unit 1 stack exhaust velocities drop off
significantly after hour 2 of July 7, 2017, indicating the unit is more fully shut down. While it's
clear that unit 1 is shut down, the merged stack parameters are still being impacted by unit 1.
Merged stack velocities decline after the unit shuts down even though unit 2's actual stack
velocities remain high. A similar downward trend in the merged stack (modeled) temperature is
also observed after unit 1 begins to shut down. This indicates using a single merged stack could
contain a number of hours with "depreciated" stack parameters if one of Conemaugh's units
shuts down (no longer burning coal). Under this scenario, merged stack velocities and
temperatures would be reduced as the shutdown unit's temperature and exhaust velocity decline.
Because of this possibility and to better represent actual stack operations, EPA only used a
merged stack when both of Conemaugh's coal units were actively operating (SO2 emissions were
greater than 0 lbs/hr). Each unit was modeled using its actual stack diameter when both units
were not burning coal simultaneously. The final EPA AERMOD-ready file will show 3 stacks
for Conemaugh but there will only be emissions from one stack for any given hour during the 3-
year simulation when the plant is actually operating. Hourly SO2 emissions will be entered into
the model for unit 1, unit 2 or a merged stack when both units are operating.
Final emission rates for Conemaugh (and all other sources) were converted using the National
Institute of Standards and Technology (NIST) conversion from pounds to grams (453.59237 g
per pound). Stack parameters similarly used the NIST conversion from feet to meters (1 foot is
0.3048 meters). This conversion was necessary since AERMOD typically operates using metric
values. Enforcement and permitting typically use imperial units (pounds and feet).
Conemaugh Heat Input versus SO2 Emissions Analysis: EPA examined both Conemaugh
units' heat input versus SO2 emissions from the CAMD database for two 3-year periods. One for
2010-12 and another over the 3-year simulated period (1 July 2017 through 30 June 2020).
Generally, one would expect a good linear relationship between the boiler heat input and SO2
emissions. As more coal is burned in Conemaugh's boilers (higher boiler heat input) more SO2
should be emitted as sulfur in the fuel is converted to SO2. This assumes unit control efficiency
via Conemaugh's wet FGD system and the feedstock coal percent sulfur has been relatively
constant over the last decade.
33
-------
EPA has reviewed coal summary statistics for the area near the Conemaugh and Seward power
plants14 and found the percent sulfur of coal in this area to be quite similar (around 2% sulfur).
We therefore expect pre-control SO2 emissions to be relatively stable over time if Conemaugh's
coal continues to be mined from the locally available coal deposits. Figure 5.2-7 shows scatter
plots for both units over the last decade. The linear relationship between each unit's heat input
(coal consumed) and SO2 emissions seems to have declined over the last decade. R2 values, an
approximation of a linear fit in the data, are much lower over the more recent 3-year simulation
period than the earlier (2010-12) time period for both of Conemaugh's units. These values
indicate a somewhat weak correlation between heat input and SO2 emissions for the 2010-12
period (values closer to 1 indicate a good linear correlation). By the time of the 3-year simulation
period, correlations become very poor to nonexistent.
Figure 5.2-7. Heat Input versus SO2 Emissions and Linear Trend Lines for Conemaugh
3-Year Simulation Period versus 2010-12
Conemaugh Unit 1 Hourly Emissions versus Heat input Conemaugh Unit 2 Hourly Emissions versus Heat Input
CAMD Part 75 1 July 2017 to 30 June 2020 CAMD Part 75 1 July 2017 to 30 June 2020
2,000- y= -270 + 0.11 x 2,000- y = _390 + 0.14x
| R2 = 0.44 | R2 = 0.51
w 1,500- A w 1,500- A
® m ® A
c 1.000" c 1.000-
¦§ m ¦§
¦| 500 - '1 500 ~
CO , , , CO
0 5,000 10,000 15,000 0 5,000 10,000 15,000
Heat Input (mmBtu/hr) Heat Input (mmBtu/hr)
Conemaugh Unit 1 Hourly Emissions versus Heat Input
CAMD Part 75 2010-12
2,000-
1,500-
1,000-
500-
Conemaugh Unit 2 Hourly Emissions versus Heat Input
CAMD Part 75 2010-12
2,000"
1,500"
1,000-
500-
0-
0 5,000
Heat Input
10,000 15,000
(mmBtu/hr)
o-
0 5,000
Heat Input
10,000 15,000
(mmBtu/hr)
34
-------
EPA speculates that Conemaugh's FGD control efficiency may be dropping over time possibly
reflecting unit degredation. Records indicate Conemaugh's FGD units were installed in the mid-
1990s and therefore have been operating for almost 30 years. Changes in the linear relationship
between the Conemaugh unit's heat input versus SO2 emissions may, however, be due to other
factors, such as a change in coal characteristics or due to other unknown operational changes that
could impact the FGD control efficiencies.
5.2.5.2. Model Input Parameters for Seward
Seward Modeled Hourly Emission Rates: EPA downloaded (actual) hourly emissions for both
Seward's waste-coal fired units (units 1 & 2) from the CAMD database. Hourly SO2 emissions
from Seward's combined units from 2010-2020 are shown in Figure 5.2-8. The shaded area of
the graph is the 3-year period modeled by EPA. Figure 5.2-9 shows Seward's combined unit
emissions over the 3-year simulation period.
Figure 5.2-8. Seward Hourly CAMD SO2 Emissions from 2010-2020
Seward Hourly Emissions
CAMD Part 75 2010-20
25,000-
Measured
Calculated
— CEV
20,000-
w 15,000-
2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
Date
14 Geology and mineral resources of the New Florence quadrangle, Pennsylvania (Bolivar, New Florence, Wilpen,
and Rachelwood 7.5-minute quadrangles, Indiana, Westmoreland, Cambria, and Somerset Counties), (1958), from
PA DCNR: http://maps.dcnr.pa. gov/publications/Default.aspx?id=9
35
-------
Figure 5.2-9. Seward Hourly CAMD SO2 Emissions over 3-year Simulation Period
Seward Hourly Emissions
CAMD Part 75 1 July 2017 to 30 June 2020
20,000-
Measured
Calculated
— CEV
^ 15,000-
2017-07-01
Seward's hourly CAMD SO2 emissions shown in both figures include the facility's modeled
CEV. This represents the maximum hourly emission rate that is modeled within the Indiana, PA
nonattainment area that does not exceed the 1-hr SO2 NAAQS of 75 ppb. This was included for
reference only and does not reflect what the CEV would be for areas outside the Indiana, PA
nonattainment area. Seward's CEV determined in Pennsylvania's Supplemental Analysis was
4,500 lbs/hr13. As shown in the hourly emissions figures, Seward's actual hourly SO2 emission
rate does, at times, exceed its model-defined CEV threshold and can, at times, be several factors
higher than its modeled CEV.
The figures showing Seward's (combined) hourly emission rates also include information
regarding whether the (CAMD) hourly emission rate was measured or calculated. This is based
on MODC. These codes are listed in Table 4a section 75.57(c) of 40 CFR Part 75. Hourly SO2
emissions are based on monitor concentration and flow rates measured by its CEMS. To have a
valid measured value both the monitor concentration and flow rate instruments must be
functioning properly. If either or both of these instruments malfunction and there is not a
redundant back up measurement available, the emission rate is calculated based on a predefined
methodology. Thus, hourly emissions are either "measured" when all instruments are functioning
in a given hour or "calculated" if there are instrument malfunctions.
15 See Table 5-1 of AECOM (2019)
36
-------
Calculated emission estimates can assume worst-case conditions if instrument down times are
significant. This scenario ensures CEMS units are functioning most of the time; otherwise, a
source will be forced to purchase emission offsets if it exceeds its yearly budget. Keep in mind
hours with "calculated" emission values could be much higher than what the actual emissions
are. As seen on the graph, the vast majority of Seward's hourly emissions are measured (have
valid MODC). EPA is providing some additional explanation for Seward's sometimes extreme
hourly SO2 emission spikes that are observed in the CAMD database compared to its
neighboring Conemaugh power plant. Each of Conemaugh's coal fired boilers have an estimated
maximum heat input rating16 of 8,280 mmBtu/hr. This is significantly larger, roughly 3 times
larger, than Seward's estimated maximum heat input rating17 of 2,532 mmBtu/hr for each of its
CFB boilers. Even though Seward's fuel burning capacity is approximately one third of
Conemaugh, hourly SO2 emissions from Seward can at times be 4 times higher.
The discrepancy in hourly SO2 emission spikes between Conemaugh and Seward is mainly due
to the significant differences in the 2 power plants' fuel characteristics. Conemaugh is a
traditional coal-fired boiler while Seward utilizes waste coal feedstock, which is sometimes
referred to as GOB or garbage of bituminous. Due to differences in these materials, which are
related to the depositional environments when these materials were created, Seward is prone to
have much higher spikes in hourly SO2 emissions than its neighbor.
Waste coal or GOB has significantly different characteristics than its parent material, western
Pennsylvanian bituminous coal. We estimate that GOB has approximately one third the Btu heat
value of the local bituminous coal stock that feeds Conemaugh's boilers and possibly 2 to 3
times the sulfur content. To get the same heat input value, 3 times as much material must be
consumed in Seward's CFB units. Since GOB has a much higher sulfur content than bituminous
coal, Seward's hourly SO2 emissions can spike much higher than its neighbor Conemaugh if
there are any issues with the Seward's SO2 control efficiency even though the plant has a much
smaller electric production capacity.
EPA used the September 2020 protocol submittal information as a basis for the development of
its (actual) hourly emissions for Seward. Hourly SO2 emission rates already represent the
combined output of both of Seward's CFB units. This removed the need to combine unit
emissions as was done with Conemaugh. A direct comparison of Seward's protocol submittal
emissions and emissions from EPA's CAMD database could be made. Nearly all of Seward's
hourly emission rates from the protocol submittal matched the corresponding CAMD emission
rates. Approximately 1-2 days' worth of hourly SO2 emissions, however, were not the same over
the 3-year simulation period.
16 Taken from SECTION A. Site Inventory List of Conemaugh's October 17, 2019 Title V/State Operating Permit
17 Taken from SECTION A. Site Inventory List of Seward's July 29, 2021 Title V/State Operating Permit
37
-------
Table 5.2-4 summarized the results of the comparison of hourly SO2 emission rates from the
protocol submittal, also referred to as CEMS, and the CAMD database for both Conemaugh and
Seward. There are several dozen hours where the protocol and CAMD emission rates do not
match. In all cases, the CAMD hourly emission rates were higher than the corresponding
protocol hourly emission rates. Hours that did not match were divided between hours where the
CAMD emission rate was measured (valid MODC) and calculated (invalid MODC). EPA is
unsure why there are differences in hourly emission rates for the hours that have measured
values according to CAMD, though these times make up a small fraction of the 3-year simulation
period. EPA replaced Seward's hourly SO2 emissions with CAMD values for any hours over the
3-year simulation period where there were mismatches between the 2 databases. For all
mismatched hours (see Table 5.2-4), the CAMD emissions exceeded the CEMS values. Model
concentrations, therefore, would be higher using the CAMD values versus the CEMS values.
Seward Modeled Hourly Stack Parameters: Stack parameters including stack temperatures
and velocities for EPA's modeling analysis were largely taken from the September 2020 protocol
submittal. These values are generally not reported to any public data system such as CAMD.
Flow rates for most CAMD reporting sources are submitted but these values are typically
reported in standard cubic feet per hour. A conversion needs to be applied to transform these
reported flow rates to actual cubic feet per hour, which could then be used to calculate stack
exhaust velocities. Flow rate MODC, however, can still be (and were) used to flag hours where
actual stack flow rates are not measured. Seward's reported stack parameters, unlike
Conemaugh, already represent combined impacts from both CFB units. No manipulation, other
than conversion to metric units, was necessary for the Seward stack parameters.
Stack temperatures were available for Seward's CFB units over the entire 3-year modeling
period. Stack velocities for each unit were available for most hours in the protocol submittal.
Values for Seward, unlike Conemaugh, needed no significant manipulation for incorporation into
the 3-year simulation period.
EPA used Seward's flow rate MODC (from CAMD) to flag hours with possible invalid stack
velocities. Table 5.2-4 summarizes Seward's hours with invalid flow rate MODC over the 3-year
simulation period. MODC descriptions from table 4a from Part 75 are also included in the table.
38
-------
CAMD flow rate MODC indicate there were 63 hours of possible invalid stack velocities for
Seward. Several MODC indicate flow rates during some of the hours with invalid MODC can be
quite high often on the extreme end of Seward's overall distribution of measured flow rates. For
the Part 75 program, using exaggerated flow rates will result in exaggerated hourly emission
rates (note Seward's peak hourly emission rate over the 3-year simulation period was calculated).
CAMD hourly mission rates are based on a concentration measurement and a flow rate
measurement. For modeling purposes, however, using an exaggerated flow rate can enhance
stack dispersion characteristics since higher stack velocities (especially from tall stacks)
generally lower final model concentrations by lofting the initial plume higher above the surface.
Use of exaggerated flow rates for invalid MODC hours, therefore, should be avoided in the
modeling analysis.
To replace missing stack velocity values or values where exaggerated flow rates may be present,
EPA used the same method to generate a more realistic stack velocity that was used in the
September 2020 protocol submittal. A surrogate stack velocity was substituted for hours with
missing or invalid flow rate measurements. Surrogate values were based on Seward's remaining
measured heat inputs and stack velocities. In general, the higher the unit heat input (the heat
released when waste coal is consumed in the CFB units) the higher the flow rate. A comparison
of measured unit heat input verses stack velocity shows there is a good (and acceptable) linear
correlation.
Figure 5.2-10 displays scatter plots for Seward's measured heat input in mmBtu versus its stack
velocity (flow rate measurement). A linear trend line was fitted to the data and shows a good
linear correlation (R2 values close to 1). The linear fit equations were then used to fill in all hours
with invalid flow rate MODC. As an example, a missing stack velocity for would be replaced by
using the corresponding hour heat input value in the linear fitted trend line equation:
Missing Stack Velocity (m/s) = 11 + 0.0054 * Heat Input (mmBtu/hr)
Figure 5.2-10. Seward Heat Input versus Stack Velocity and Linear Trend Lines
39
-------
Seward Heat Input versus Adjusted Stack Velocity
Select CEMS Data 1 July 2017 to 30 June 2020
50 "
<
0 -
I I I I
0 2,000 4,000 6,000
Heat Input (mmBtu/hr)
Seward Heat Input versus SO2 Emissions Analysis: EPA examined Seward's hourly heat
input versus SO2 emissions from the CAMD database for two 3-year periods. One for 2010-12
and another over the 3-year simulated period (1 July 2017 through 30 June 2020). Generally, one
would expect a good linear relationship between the boiler heat input and SO2 emissions. As
more coal waste is burned in Seward's CFB boilers (higher boiler heat input) more SO2 should
be emitted as sulfur in the fuel is converted to SO2. This assumes control efficiency via Seward's
limestone injection system and the percent sulfur of the feedstock material has been relatively
constant over the last decade.
Figure 5.2-11 shows scatter plots for Seward over the last decade. The linear relationship
between the heat input (waste-coal consumed) and SO2 emissions seems to have significantly
improved over the last decade. R2 values, an approximation of a linear fit in the data, are much
higher over the more recent 3-year simulation period than the earlier (2010-12) time period.
These values indicate a very poor linear correlation between heat input and SO2 emissions for the
2010-12 period; values between -0.3 and 0.3 are generally understood as having no correlation.
By the time of the 3-year simulation period, R2 values indicate there is a weak to somewhat good
correlation between the units' heat input and SO2 emissions. The graph for the 3-year simulation
period shows points much more clustered around the linear trendline than the 2010-12 period
does. We also note that points are more clustered in the higher emission/higher heat input (upper
right quadrant) of the 2012 graph than during the 3-year simulation period (indicating fewer
emission spikes when the plant is operating near its intended capacity).
40
-------
Figure 5.2-11. Heat Input versus SO2 Emissions and Linear Trend Lines for Seward
3-Year Simulation Period versus 2010-12
^ 15,000-
sz
C/)
-Q
.2 10,000-
03
QZ
c.
o
'
(/)
5,000-
LU
CN
o
w
0-
One possible explanation for this change is improved control efficiency at Seward. These
changes, however, could also be due to changes in the feed-stock waste coal material. In
AECOM's (2019) report it includes a short description of more recent operation changes at
Seward. From the report:
"[I]n December 2016, Seward Station changed ownership. Operational changes daring
plant start-ups have been implemented. Seward is currently adding limestone to the
combustor during initial firing to reduce SO2 emissions. This is a change to the previous
operating strategy and is expected to continue with this practice going forward. The
distribution accounts for the frequency and duration observed during actual station
operations, and this operation is expected to continue in a similar manner for future
years. "
The linear correlation improvement EPA notes in Seward's heat input versus SO2 emissions over
the 3-year simulation period versus the 2010-12 appears to offer some tacit support that there
have been recent operational changes made at Seward that are leading to fewer hourly SO2
emission spikes. It would be helpful if the Commonwealth of Pennsylvania and/or Seward could
more fully document these changes and when they occurred.
Seward Hourly Emissions versus Heat input
CAMD Part 75 1 July 2017 to 30 June 2020
y = 45 + 0.
R2 = 0.74
55 x
LU
CN
o
C/5
0 2,000 4,000 6,000
Heat Input (mmBtu/hr)
Seward Hourly Emissions versus Heat Input
CAMD Part 75 2010-12
5,000-
y= 1
R2 =
10 + C
0.32
).39>
f
i
;,r> 0°
0-
0 2,000 4,000 6,000
Heat Input (mmBtu/hr)
41
-------
5.2.5.3. Overview of Model Input Parameters for Cambria County Sources
EPA downloaded information for the 3 Cambria County sources from its CAMD database.
Yearly and quarterly SO2 emissions tons per year (tpy) and operating hours for Cambria Cogen,
Colver Power and Ebensburg Power are summarized in Tables 5.2-7 and 5.2-8. Cambria Cogen
was deactivated in 2019 accounting for the drop off in emissions and operating hours in the
summary tables. Both tables indicate the Cambria County (waste coal) sources generally operate
on a fairly consistent basis over the 3-year simulation period. Collectively, these sources emitted
between 7,000 and 8,000 tpy of SO2 when they were all operating. With the closure of Cambria
Cogen, emissions appear to be on the order of 4,000 to 5,000 tpy or about 30% lower than
historic averages.
Table 5.2-7. Cambria County Source Yearly SO2 Emissions and Operating Hours
Cambria County Sources Yearly S02 Emissions and Operating Hours
1 July 2017 to 30 June 2020
Cambria Cogen
Colver Power
Ebensburg Power
Year
S02 (tpy)
Hours On
S02 (tpy)
Hours On
S02 (tpy)
Hours On
2017
1,207.91
4,416
1,317.75
3,803
776.85
4,174
2018
2,520.23
8,760
2,731.83
7,960
1,855.84
8,037
2019
491.36
4,344
2,265.65
7,600
1,145.33
6,919
2020
0.00
0
1,008.01
3,134
574.46
3,475
Table 5.2-8. Cambria County Source Quarterly SO2 Emissions and Operating Hours
Cambria County Sources Quarterly S02 Emissions and Operating Hours
1 July 2017 to 30 June 2020
Cambria Cogen
Colver Power
Ebensburg Power
Year
Quarter
S02 (tpy)
Hours On
S02 (tpy)
Hours On
S02 (tpy)
Hours On
3
596.46
2,208
731.68
2,091
396.55
2,208 ;
2017
4
611,45
2,208
586.07
1,712
380.30
1,966 |
1
568,64
2,160
717.41
2,075
501,78
__
2
626.57
2,184
703.38
2,029
416.13
1,850
2018
3
601.89
2,208
671.05
1,987
471.54
2,114
4
723.13
2,208
639.99
1,869
466.38
1,913 J
1
459,51
2,160
677.45
2,047
347.51
__
2
31.84
2,184
417.71
1,844
221.55
1,689
2019
3
0.00
0
573.54
1,858
430.80
2,091
4
0.00
0
596.96
1,851
145.47
1,108 j
1
0.00
0
692.95
2,133
336.55
2,184 I
2020
2
0,00
0
315.06
1,001
237.90
1,291 |
42
-------
The CAMD database includes hourly emissions as well as information on concentration and flow
instrument validity via MODC. A summary of Cambria Cogen's, Colver Power's, and
Ebensburg Power's SO2 concentration MODC and flow MODC are shown in Table 5.2-9 and
Table 5.2-10. Note there are fewer hours for Cambria Cogen since it ceased reporting to CAMD
prior to the end of the 3-year modeling period.
Table 5.2-9. Cambria County Source SO2 Concentration MODC
Summary of S02 MODC Part 75 (Table 4a)
1 July 2017 to 30 June 2020
MODC
Part 75 Explanation
Cambria Cogen
Unit 1
Cambria Cogen
Unit 2
Colver
Ebensburg
00
Off
3,307
3,766
3,807
3,699
01
Certified Primary Instrument
14,188
13,685
22,045
22,458
06
Average of the Hour Before and the Hour
Following a Missing Data Period
24
65
286
68
08
90th Percentile Hourly S02 Rate
0
0
28
0
12
Maximum Hourly S02 Rate Over Lookback
Period
1
3
1
0
16
S02 Concentration Value of 2.0 ppm During
Hours When Only Very Low Sulfur Fuel
0
0
87
0
21
Negative Hourly S02 Concentration Replaced
with 0
0
1
50
79
Table 5.2-10. Cambria County Source Flow MODC
Summary of Flow MODC Part 75 (Table 4a)
1 July 2017 to 30 June 2020
MODC
Part 75 Explanation
Cambria Cogen
Unit 1
Cambria Cogen
Unit 2
Colver
Ebensburg
00
Off
3,307
3,766
3,807
3.699
01
Certified Primary Instrument
14,166
13,711
22,359
19,441
06
Average of the Hour Before and the
Hour Following a Missing Data Period
0
0
27
541
08
90th Percentile Hourly Flow Rate
0
0
0
129
09
95th Percentile Hourly Flow Rate
0
0
0
142
10
Maximum Hourly Flow Rate Over
Lookback Period
0
0
0
876
11
Average of Hourly Flow Rates Over
Lookback Peiod
41
40
111
117
12
Maximum Potential Flow Rate
0
0
0
129
20
200% of the Full-Scale Range Setting
0
0
0
1,230
55
Other Substitute Data Approved
Through Petition
6
3
0
0
CAMD reported (mass) emissions rates are based on CEMS flow rates and concentration
instrumentation. As described earlier, measured mass flow rates represent hours with valid
instrument values and calculated mass flow rates are generated for hours with instrument
malfunctions. Since CAMD is part of an emissions trading program, calculated mass flow rates
can sometimes be intentionally over estimated. These hours should be noted as they have the
potential to impact hourly emission inputs into the 3-year modeling analysis.
43
-------
For the Cambria County sources, the MODC concentration and flow rate summaries indicate
most of the sources utilize measured mass emission values. Missing values for Cambria Cogen
and Colver are largely replaced with reasonable estimations. Ebensburg Power, however, appears
to have a significant number of hours with calculated flow rates that could lead to very high
calculated hourly SO2 emission rates. Exaggerated flow rates will return much higher hourly SO2
emission rates since the flow rate figures into the mass emission calculation.
Given the distance from the area of focus in Cambria and Westmoreland counties and the
average emission rates, the emissions of the Cambria County sources are not very impactful to
the violating receptors (see Table 5.2-15). Conemaugh and Seward being closer and having
higher hourly emission rates have significantly higher impacts. Additionally, the Cambria
County sources peak impacts on the area of focus are expected to occur under northerly wind
directions. Under those circumstances, emissions from Conemaugh and Seward would be pushed
south into the Ligonier Valley and away from the Laurel Ridge. Thus, the Cambria County
sources produce minimal impacts when Conemaugh and Seward emission are blown east
towards the controlling Laurel Ridge topographic feature.
5,2,5,3,1, Model Input Parameters for Cambria Cogen
Cambria Cogen Modeled Hourly Emission Rates: EPA downloaded Cambria Cogen's
(actual) hourly emissions for both its waste-coal fired units (units 1 & 2) over the 3-year model
simulation period (1 July 2017 through 20 June 2020). The combined units' hourly emission rate
was used in EPA's modeling analysis.
Similar to Seward, Cambria Cogen was fired with waste coal or GOB. Fuel is consumed in 2
CFB units with lime injection to control SO2 emissions. Cambria Cogen's units, however, are
much smaller than Seward's. Seward's listed Title V boiler ratings are 2,532 million British
thermal units per hour (mmBtu/hr) for both of its units. For comparison, Cambria Cogen's listed
Title V boiler ratings are 630 mmBtu/hr for each of its waste coal units. In total, Seward's
combined boiler rating is about 8 times higher than Cambria Cogen. We note that a Retirement
Unit Exemption form was filed with CAMD notifying the units' deactivation. This became
effective in September of 2020.
Figure 5.2-12 shows Cambria Cogen's combined hourly SO2 emission rate over the 3-year model
simulation period (highlighted in green). Similar to Conemaugh and Seward, EPA identified
which hours were "measured" (with valid flow and concentration MODC) and which hours were
"calculated" (either invalid flow and/or concentration MODC).
44
-------
Figure 5.2-12. Cambria Cogen's Hourly CAMD SO2 Emissions over 3-year Simulation
Period
Cambria Cogen Hourly Emissions
CAMD Part 75 2017-20
Measured
Calculated
£ 1,000-
LU
O
CO
Date
Cambria Cogen's hourly SO2 emissions generally average between 500 and 750 lbs/hr over the
simulation period. Some emission spikes do occur, at times exceeding 1,000 pounds per hour.
Like Seward, percent sulfur variability in the fuel source (coal waste or GOB) and control
efficiency drops may account for these emission spikes. As mentioned previously, hourly
emissions generally cease after the 1st quarter of 2019 marking the time these units were
deactivated.
Cambria Cogen's Modeled Hourly Stack Parameters: Stack parameters including stack
temperatures and velocities for EPA's modeling analysis were based on information that was
used in Pennsylvania's Round 3 DRR modeling analysis combined with CAMD information
pulled by EPA over the same time period (2013-15). Physical stack locations, stack heights and
stack diameters were taken from the Pennsylvania Round 3 DRR modeling file. No building
downwash was used since the Cambria County sources were located well away from the area of
interest and would therefore have little or no effect on the model results.
CAMD data does include flow rate information. This flow rate information, however, is reported
as an adjusted flow rate; standard cubic feet per hour (scfh). Actual flow rate information is not
reported to CAMD and is usually only available from the source's CEMS units. Additionally, no
information is available for hourly stack (emission) temperatures. For modeling purposes, the
CAMD database can provide hourly SO2 emission rates and information on which hours flow
rates may be invalid.
45
-------
To develop Cambria Cogen's hourly stack parameters, EPA utilized Pennsylvania's hourly
emission file for its Round 3 DRR modeling analysis covering 2013-15 coupled with
corresponding CAMD information. This information was used to establish relationships between
the combined units' heat input and modeled hourly stack temperature and stack velocity.
Relationships between Cambria Cogen's heat input, stack temperature and stack velocity from
the DRR modeling period were then used to determine modeled stack temperature and velocity
over EPA's 3-year simulation period.
Cambria Cogen's stack temperature versus combined unit heat input (for all operating hours)
over the 3-year DRR Round 3 modeling period is shown on Figure 5.2-13. A best fit linear
trendline was imposed on the data and is shown on the figure along with the linear equation and
R-squared value, which indicates how well the linear trend line fits the data. While the R-squared
value indicates a poor fit, EPA believes the developed linear fit line provides the best estimate of
hourly varying stack temperatures over the 3-year simulation period. This was done by plugging
Cambria Cogen's hourly heat input over the 3-year simulation period (1 July 2017 through 30
June 2020) into the liner fit line equation.
Figure 5.2-13 Cambria Cogen Linear Fitted Equation for Stack Temperature
Cambria Cogen Combined Heat Input versus Stack Temperature
2013-15 Round 3 DRR Modeling Information
500-
y = 430 + 0 025 x
-------
EPA utilized a similar process to develop model hourly stack velocities for Cambria Cogen.
Again, hourly stack velocities were coupled with hourly (combined) heat input to construct linear
fits to the data. Unlike the temperature construction, CAMD flow MODC were used to screen
out any hours with invalid (calculated) flow rates. Figure 5.2-14 shows Cambria Cogen's best
linear fit between its DRR Round 3 model stack velocity and corresponding hourly heat input. It-
squared values indicate a much better fit with the data than Cambria Cogen's stack temperatures.
Of the 2, modeled stack velocities probably have the greater impact on final model
concentrations.
Cambria Cogen's hourly stack velocities for the 3-year simulation period were determined by
plugging in the hourly heat input information into the linear fit line equations. This information
along with the corresponding hourly SO2 emission rate and calculated stack temperature was
then used to create an AERMOD hourly emission rate file for the 3-year simulation period.
Figure 5.2-14. Cambria Cogen Linear Fitted Equation for Stack Velocity
Cambria Cogen Combined Heat Input versus Stack Velocity
2013-15 Round 3 DRR Modeling Information
20-
15-
(/)
E
o
o 10-
d)
>
o
3
CO
y= 2.4 + 0.0092 x
R2 = 0.86
I*
1*
0 500 1,000 1,500
Heat Input (mmBtu/hr)
5.2.5.3.2. Model Input Parameters for Colver Power
Colver Power Modeled Hourly Emission Rates: EPA downloaded Colver Power's (actual)
hourly emissions for its waste-coal fired unit over the 3-year model simulation period (1 July
2017 through 20 June 2020). The CAMD hourly emission rate was used in the modeling
analysis.
47
-------
Colver Power is 1 of 3 waste coal or GOB fired facilities in Cambria County. Similar to the other
waste coal facilities, fuel is consumed in CFB units with lime injection to control SO2 emissions.
Colver Power's waste coal unit, like the other units in Cambria County, is much smaller than
Seward's. Colver Power's listed Title V boiler rating is 1,214.5 mmBtu/hr making it Cambria
County's largest single waste-coal unit. In boiler rating size, Colver Power is about one fourth of
Seward's combined listed boiler rating.
Figure 5.2-15 shows Colver Power's hourly SO2 emission rate over the 3-year model simulation
period (highlighted in green). Similar to other sources, EPA identified which hours were
"measured" (with valid flow and concentration MODC) and with hours were "calculated" (either
invalid flow and/or concentration MODC).
Figure 5.2-15. Colver Power's Hourly CAMD SO2 Emissions over 3-year Simulation Period
Colver Power Hourly Emissions
CAMD Part 75 2017-20
2,500-
Measured
— Calculated
2,000-
2017 2018 2019 2020 2021
Date
Colver Power's hourly SO2 emissions generally average between 500 and 1,000 lbs/hr over the
simulation period. Some emission spikes do occur, at times exceeding 1,500 pounds per hour.
Like Seward, percent sulfur variability in the fuel source (GOB) and control efficiency drops
may account for these emission spikes. Colver Power's hourly emissions profile indicates near
constant operations over the 3-year simulation period.
48
-------
Colver Power's Modeled Hourly Stack Parameters: EPA constructed Colver Power's
modeled hourly stack temperature and stack velocity using the same strategy described
previously. We used Pennsylvania's DRR Round 3 model information coupled with
corresponding CAMD heat input to produce a linear relationship that could be used to estimate
Colver Power's hourly stack temperature and stack velocity over the 3-year simulation period.
Figure 5.2-16 shows Colver Power's heat input versus stack temperature from CAMD and the
DRR Round 3 modeling model hourly emission file. Similar to Cambria Cogen, R-squared
values for the linear fit are poor. While the R-squared value indicates a poor fit, EPA believes the
developed linear fit line provides the best estimate of hourly varying stack temperatures over the
3-year simulation period.
Figure 5.2-16. Colver Power Linear Fitted Equation for Stack Temperature
Colver Power Heat Input versus Stack Temperature
2013-15 Round 3 DRR Modeling Information
y= 370+ 0.04 x
R2 = 0.28
200-
Heat Input (mmBtu/hr)
Stack velocities were configured using the same approach though hours with invalid flow
MODC were excluded from the analysis. Figure 5.2-17 shows Colver Power's heat input versus
stack velocity from CAMD and the DRR Round 3 modeling model hourly emission file. R-
squared values for the linear fit are poor. While the R squared value indicates a poor fit, EPA
believes the developed linear fit line provides the best estimate of hourly varying stack velocities
over the 3-year simulation period. EPA utilized this equation to construct Colver Power's hourly
varying stack velocity by plugging in the hourly unit heat input values over the 3-year model
simulation period into the equation. This information along with the corresponding hourly SO2
emission rate and hourly stack temperature was then put into the AERMOD emission input file
for the final 3-year simulation period.
49
-------
w 20-
£
Figure 5.2-17. Colver Power Linear Fitted Equation for Stack Velocity
Colver Power Heat Input versus Stack Velocity
2013-15 Round 3 DRR Modeling Information
o-
500 1,000 1,500 2,000
Heat Input (mmBtu/hr)
5.2.5.3.3. Model Input Parameters for Ebensburg Power
Ebensburg Power Modeled Hourly Emission Rates: EPA downloaded Ebensburg Power's
(actual) hourly emissions for its waste-coal fired unit over the 3-year model simulation period (1
July 2017 through 20 June 2020). The CAMD hourly emission rate was used in the 3-year
modeling analysis.
Ebensburg Power is one of the 3 waste-coal or GOB fired facilities in Cambria County. Waste
coal is consumed in the CFB unit and controlled via lime injection to reduce SO2 emissions.
Ebensburg Power's listed Title V boiler rating is 820 mmBtu/hr making it Cambria County's
smallest waste coal unit. As far as boiler rating size, Ebensburg Power is less than one fifth the
size of Seward's combined listed boiler rating.
Figure 5.2-18 shows Ebensburg Power's hourly SO2 emission rate over the 3-year model
simulation period (highlighted in green). Similar to other sources, EPA identified which hours
were "measured" (with valid flow and concentration MODC) and which hours were "calculated"
(either invalid flow and/or concentration MODC).
50
-------
Figure 5.2-18. Ebensburg Power's Hourly CA.MD SO2 Emissions for 3-vr Simulation
Period
Ebensburg Power Hourly Emissions
CAMD Part 75 2017-20
2,500-
Measured
Calculated
2,000-
(/)
w 1,500-
Q)
CD
K.
c
o
'(/)
-------
Figure 5.2-19 shows Ebensburg Power's heat input versus stack temperature from CAMD and
the DRR Round 3 modeling's hourly emission file. Unlike the other sources, R-squared values
for the linear fit are relatively good. EPA utilized this equation to construct Ebensburg Power's
hourly varying stack temperatures by plugging in its hourly unit heat input values over the 3-year
model simulation period into the equation.
Stack velocities were configured using the same approach though hours with invalid flow
MODC were excluded from the analysis. Figure 5.2-20 shows Ebensburg Power's heat input
versus stack velocity from CAMD and the DRR Round 3 modeling model hourly emission file.
R-squared values for the linear fit are excellent. EPA utilized this equation to construct
Ebensburg Power's hourly varying stack velocity by plugging in the hourly unit heat input values
over the 3-year model simulation period into the equation. This information along with the
corresponding hourly SO2 emission rate and hourly stack temperature was then put into the
AERMOD hourly emission file for the final 3-year simulation period.
Figure 5.2-19. Ebensburg Power Linear Fitted Equation for Stack Temperature
Ebensburg Power Heat Input versus Stack Temperature
2013-15 Round 3 DRR Modeling Information
y = 380 + 0.074;
R2 = 0.76
<
-400-
CD
=J
CTJ
1—
Q)
Q_
E
d)
o
2 300-
O)
200-
Heat Input (mmBtu/hr)
52
-------
Figure 5.2-20. Ebensburg Power Linear Fitted Equation for Stack Velocity
Ebensburg Power Heat Input versus Stack Velocity
2013-15 Round 3 DRR Modeling Information
y = 3.9 + 0.024 x
R2 = 0.94
if
I
0 250 500 750 1,000
Heat Input (mmBtu/hr)
5.2.6. Modeling Parameter: Meteorology and Surface Characteristics
As noted in the Modeling TAD, the most recent 3 years of meteorological data (concurrent with
the most recent 3 years of emissions data) should be used in designations efforts. The selection
of data should be based on spatial and climatological (temporal) representativeness. The
representativeness of the data is determined based on: 1) the proximity of the meteorological
monitoring site to the area under consideration, 2) the complexity of terrain, 3) the exposure of
the meteorological site, and 4) the period of time during which data are collected. Sources of
meteorological data include National Weather Service (NWS) stations, site-specific or onsite
data, and other sources such as universities, Federal Aviation Administration (FAA), and
military stations.
Site Specific Meteorological Data (Ash Site #1): Sensitivity and evaluation studies (Paine,
2001 and Paine et al, 2013) have shown that AERMOD has the potential to over-estimate the
downwind plume impacts from sources located in or near hilly terrain like Conemaugh and
Seward. This can be especially true if the analysis is performed using meteorological data that
are not site-specific and consists of only a single low level (e.g., 10-m), wind measurements such
as at National Weather Service stations. Significant improvement in AERMOD performance for
impacts in complex terrain from tall-stack emissions would be expected with the use of site-
specific multiple-level tower and SOnic Detection And Ranging (SODAR) wind profiler system.
53
-------
It was for this reason that a plan for site-specific meteorological measurements was formulated to
address dispersion characteristics near the Conemaugh and Seward power plants. This led to an
EPA-approved meteorological monitoring protocol in the spring of 2015, and the installation of a
100-meter height meteorological tower equipped with multiple levels of meteorological sensors
(at 2, 10, 50, 75, and 100 m) along with a SODAR wind profiler system (with measurements
starting at 50 m and extending upwards in 50-m increments to 500 m).
A meteorological measurement site was located on the Ash Site #1 located between the
Conemaugh and Seward power plants (see figures 5.2-21 and 5.2-22). AERMOD was
specifically designed to accommodate multiple levels of meteorological data to more accurately
estimate vertical profiles of meteorological variables used in the modeling. For the monitoring
program, the EPA Guidelines for Air Quality Modeling (40 CFR Part 51, Appendix W) and
EPA's meteorological monitoring guidance (EPA, 2000) provided the general guidance for
sensor and parameter selection and siting of the tower and SODAR. A more detailed description
of the monitoring equipment, collection site and data gathering procedures is described in
AECOM's Meteorological Monitoring Station Design and Quality Assurance Project Plan for
the Conemaugh and Seward Generating Stations - Indiana County, PA dated March 2015.
Figure 5.2-21. Site Specific Meteorological Data Collection Locations
Site Specific Meteorological Data Collection • Ash #1 Site
Legend
Sources
^^Con emau gh
Seward
Ash #1 Site
0 Site-Specific (Met Tower)
©Site-Specific (SODAR)
\! ys '
.Indiana, PA Nonattainment
Area
% . U » ^
2 Kilometers
«EPA
54
-------
Figure 5.2-22. From AECOM-2, Figure 1-4, View of Meteorological Site, Conemaugh and
Seward. Ligonier Valley, Looking Southwest with Chestnut Ridge in Background
Soil Moisture and Snow Cover Analysis/AERSURFACE Processing: AERMET, the
meteorological preprocessor for AERMOD, has advanced boundary layer algorithms that require
user-specified surface characteristics for albedo, Bowen ratio, and surface roughness length. To
aid the user community with an objective method for determining these AERMET-required
surface characteristics, EPA developed the AERSURFACE tool, which was first released in
2008. AERSURFACE generates estimates of realistic and reproducible surface characteristic
values using LULC data from the National Land Cover Database (NLCD).
AERSURFACE is not a regulatory component of the AERMOD Modeling System as listed in
Appendix A to the Guideline on Air Quality Models (or Appendix W to 40 CFR Part 51).
Section 8.4.2(b) of Appendix W recommends the use of the latest version of AERSURFACE for
determining surface characteristics when processing measured meteorological data through
AERMET (i.e., representative site-specific data or data from a nearby National Weather Service
or comparable station).
AERMET-ready surface characteristics including surface roughness length, albedo and Bowen
ratio were based on the location of the meteorological site-specific collection site (Ash Site #1).
Pennsylvania SIP submittals utilized a previous version of AERSURFACE. Surface
characteristics need to be determined using the most recent version AERSURFACE. The newer
version allows for the use of more recent LULC data that would better align with the site-specific
meteorological data collection period. Site-specific meteorological data was collected over a 13-
month period from 1 August 2015 through 31 August 2016. The final 1-year collection period
used in PA's Supplemental Analysis spanned from 1 September 2015 through 31 August 2016.
This was due to better SODAR data capture percentages over this period (AECOM, 2019).
I
55
-------
EPA reran AERSURFACE to generate surface characteristics using the 100-m meteorological
tower location. The release of AERSURFACE version 20060 replaces version 13016 and
finalizes many of the updates and enhancements implemented in the 19039 DRFT version. EPA
used processing steps outlined in its AERMOD Implementation Guide, AERSURFACE users
guide (EPA, 2020) and AERSURFACE transmission memo18. The AERMOD Implementation
Guide recommends the use of a circular 1-km radius centered at the meteorological station site
for surface roughness calculations. Bowen ratios and albedo values were determined in
accordance with guidance using a 10-km by 10-km region centered on the measurement site.
AERSURFACE links various land cover categories to a set of seasonal surface characteristics
and requires specification of the seasonal category for each month of the year that are assigned
based on local conditions. Bowen ratios are dependent on surface wetness characteristics and
were assigned "wet", "average" and "dry" categories in AERSURFACE using a 30-year
precipitation data set for Pennsylvania Climate Division 9. Albedo values also need adjustments
to account for wintertime (monthly) snow cover. A month is considered to have "continuous"
snow cover if over half the days have at least 1 inch snow depths. Snow cover was retrieved
from the Community Collaborative Rain, Hail, and Snow Network for the "Belmont 0.1 NE" site
(PA-CM-4 on the CoCoRaHS website) located approximately 18 km southeast of the
Conemaugh and Seward. Figure 5.2-23 shows the Pennsylvania Climate Division 9, the Ash Site
#1 tower location and the Belmont 0.1 NE site used to assess snow cover.
EPA reexamined the surface wetness characteristics for its modeling analysis. Pennsylvania's
supplemental modeling analysis utilized monthly precipitation from 1981 to 2010 as the 30-year
period. We updated the time period and used PA Climate Division 9 monthly precipitation data
from 1991 through 2020 for surface moisture determination. Actual AERSURFACE soil
moisture categories were based off of monthly precipitation totals collected at the Ash Site #1.
Soil moisture categories were determined by taking the Ash Site #1 monthly precipitation and
comparing them to the 30-year precipitation data from Pennsylvania Climate Division 9 based on
the divisions explained in section 2.3.3 of the AERSURFACE users guide.
Snow-cover from Pennsylvania's modeling analysis was also reexamined. Only the month of
January (2016) had more than 50% of the days with at least 1 inch of snow cover during the site-
specific meteorological data collection period. Other winter designated months (Nov, Dec, Feb
and Mar) were not adjusted to account for continuous monthly snow cover.
Table 5.2-11 shows the soil moisture category breakdown during the site-specific meteorological
data collection period. It includes the corresponding monthly Climate Division 9 precipitation
and the cut off values for "wet" and "dry" months. Final soil categories were based on where the
Ash Site #l's monthly precipitation fell within the Climate Division 9 thirty-year survey period.
56
-------
Figure 5.2-23. Location of Met Tower, Belmont 0.1 NE and PA Climate Division 9
Legend
Met Sites
0 Ash Site #1 Met Tower
Q Belmont 0.1 NE
~ Indiana, PA
Nonattainment Area
PA Climate Division 9
Elevation
rence
f^rmlst'rohgi
Meters
Indiana
|?N'eghenyj
¦MM
kwmmm
Pennsylvania Climate Division 9 / Ash Site #11 Belmont 0.1 NE
100 Kilometers
1
SERA;
18 See EPA's Support Center for Regulatory Atmospheric Modeling (SCRAM) website:
https://vvww.eDa.gOY/scram/air-qualitv-disDersion-inodeling-related-model-suppoit-prograins
57
-------
Table 5.2-11. Monthly AERSURFACE Soil Moisture Category Breakdown
Ash Site #1 AERSURFACE Monthly Soil Moisture Category/Snow Cover
Year
Month
Dry Cutoff
Wet Cutoff
PA Climate 09
Ash Site #1
Category
Snow Cover
2015
Sep
2.45
4.60
4.94
2.80
Average
2015
Oct
2.36
4.29
3.49
2,89
Average
2015
Nov
2.52
4.13
2.04
1.65
Dry
No Snow
2015
Dec
2.69
4.18
4.11
3.81
Average
No Snow
2016
Jan
2.33
3.84
2.25
1.40
Dry
Snow
2016
Feb
2.08
3.30
3,17
1,62
Dry
No Snow
2016
Mar
2.78
4 54
2.53
1.94
Dry
No Snow
2016
Apr
2.96
4.47
2.76
1.52
Dry
2016
May
3.34
5.24
4.32
4.89
Average
2016
Jun
3.72
5.59
4.24
4.26
Average
2016
Jul
3.76
4.90
3.94
3.07
Dry
2016
Aug
3.03
4.62
4,37
5,78
Wet
EPA updated surface characteristics using AERSURFACE (20060) centered on the Ash Site #l's
100-m tower location. Standard settings were used to determine surface roughness values
surrounding the met tower; ZORADIUS was set to 1.0 km. Albedo and Bowen ratios were
determined by a 10 km by 10 km survey area centered on the tower location. Land use/land
cover (LULC), impervious surface and tree canopy data for 2016 were downloaded from the
Multi-Resolution Land Characteristics (MRLC) Consortium website and used as
AERSURFACE's source data. Figure 5.2-24 shows the location of the Ash Site #1 met tower
along with the surface roughness and albedo/Bowen ratio survey areas.
AESURFACE sectors can range from 1 to 12 for input to AERMET, though sectors must be a
minimum of 30°. Surface roughness values were assigned to 8 sectors surrounding the Ash Site
#1 met tower. Sector width spacing varied and was based on the visual presentation of the 2016
land use categories within 1-km of the Ash Site #1. All sectors were defined as non-airport (see
section 2.3.2 of the AERSURFACE users guide for additional information). Figure 5.2-25 shows
the chosen sectors and the 2016 LULC. The 2016 LULC contains much more developed LULC
categories than the 1992 data used in the Indiana, PA SIP. The 1992 data was the only data
available at the time of SIP preparation.
The bulk of the developed LULC categories in Figure 5.2-25 are generally confined to the
Conemaugh and Seward power plants. The Ash Site #1 is also categorized as developed. This
may be because the LULC category assignment is made using spectral analysis (possibly in the
infrared) and the material in the ash landfill resembles properties of concrete or macadam
surfaces. While this is a significant change from the 1992 LULC, EPA believes the impact of this
change would be minimal since the Ash Site #1 makes up a small fraction of the 1 km survey
area and its impacts are further reduced by dividing the area among the 8 defined sectors.
58
-------
EPA also notes the 2016 LULC includes a significant area of woody wetlands along the
southeastern side of the Conemaugh River but not on the power plant side of the river. We
surmise that any low-lying areas near the Conemaugh and Seward power plants were probably
filled in and raised to prevent either facility from being flooded, a common occurrence along this
river over the last century or so. The last major flooding event on the Conemaugh River occurred
in 1977 and permanently displaced residents from the hamlet of Robindale.
Impervious surface and tree canopy data (for 2016) were also processed in AERSURFACE to
supplement the surface roughness calculations. These are displayed in Figure 5.2-26 and 5.2-27.
Roads and railroads are clearly delineated in the impervious surface files. The Conemaugh and
Seward power plant structures also show up in the impervious surface files, as does the Ash Site
#1 landfill. EPA does not think the Ash Site #1 will make a significant impact on final surface
roughness calculations since including the impervious surface information appears to only make
small differences in the AERSURFACE derived values. Tree canopy data appears to be accurate.
Roads and rail roads are clearly visible along with power line cuts. The Ash Site #1 does not
appear to have any tree cover, which makes sense since disposal sites are generally prohibited
from allowing any woody vegetation19 to take hold (to protect any capping structures).
AERSURFACE was run multiple times to construct the monthly varying surface characteristics
for the Ash Site #1 met tower location. This accounts for the different monthly soil
characteristics and snow cover information (as listed in Table 5.2-11). Seasonal settings for the
Ash Site #1 were identical to the ones assigned in Pennsylvania's original and supplemental SIP
analysis. Winter included the months of November, December, January, February and March.
Spring included April and May. Summer included the months of June, July and August. Autumn
included the months of September and October. Given the variability in monthly soil moisture
and continuous snow cover, 4 AERSURFACE runs were needed. These included Average (soil
moisture) no snow, dry-snow, dry-no snow and wet-no snow. The results of these
AERSURFACE runs were used as stage 2 input during the AERMET processing described in the
next section.
19 EPA is not certain if the Ash Site #1 is subject to state disposal site regulations but 25 PA Code § 288.237 (b)
describing standards for successful revegetation of disposal areas prohibits trees, woody shrubs or deep-rooted
plants that would allow the penetration of the disposal area's cap or drainage layer.
59
-------
Figure 5.2-24. AERSURFACE Survey Areas for Surface Roughness, Albedo and Bowen
Ratio
AERSURFACE Albedo & Bowen Ratio Survey - Ash Site #1 / 2016 LULC
Legend
It Met Tower
Sources
^^^Conemaugh
— Bowen Ratio & Albedo Survey
E3 Surface Roughness (1-km)
^—Indiana, PA Nonattainment
"""Area
2016 LULC Categories
| Open Walei
Developed. Open Space
| Developed, low Intensity
| Developed. Medium Intensity
| Developed. High intensity
| Barron Land
] Deciduous Forest
J Scrub Brush
] Grass ai
| Pasture/Hay
I Cultivated Crops
| Woody Wetlands
| emergent Herbaceous Wetlands
10 Kilometers
v=,ERA
Figure 5.2-25. AERSURFACE Sector Assignment for Surface Roughness Calculations
AERSURFACE Surface Roughness Sectors - Ash Site #1 / 2016 LULC
Legend
A! Met Tower
Sources
Conemaugh
_ Defined Surface Roughness
Sectors
——Indiana, PA Nonattainment
aArea
SERA
60
-------
Figure 5.2-26. AERSURFACE Impervious Surface Input
AERSURFACE Surface Roughness Sectors - Ash Site #1 / 2016 LULC
Legend
Met Tower
Sources
Conemaugh
Defined Surface Roughness
Sectors
.Indiana, PA Nonattainment
'Area
°-5 | 1 2 Kilometers oEFA£s™'"~™
Figure 5.2-27. AERSURFACE Tree Canopy Input
61
-------
AI RM IT Processing: Once the AERSURFACE files were generated using the 2016 LULC
data, the Ash Site #1 met tower and SODAR collected data had to be run through EPA's
AERMET preprocessor. EPA followed the same processing steps completed in Pennsylvania's
SIP submission. This essentially updates the SIP modeling meteorological input files to use the
most current versions and produced the final meteorological input files needed to run AERMOD.
Meteorological data from several sources were used in EPA's AERMET processor. Site-specific
data was processed for the Ash Site #1 100-m tower and SODAR. Additional surface
measurements from the nearby Johnstown-Cambria County airport were also included. Upper air
data in the form of morning soundings came from a site near the Pittsburgh International Airport
in western Pennsylvania.
The locations of these sites are shown on Figure 5.2-28. The Ash Site #1 is located between the
Conemaugh and Seward power plants. This site is located within the Ligonier Valley at about the
same elevation as Conemaugh and Seward and is within 2 km of either facility. The Johnstown-
Cambria County airport is located approximately 20 km east of the Ash Site #1, Conemaugh and
Seward. The airport's ASOS measurements are taken at nearly 700 m in elevation. This is nearly
360 m in elevation higher than the Ash Site #1. Upper air morning sounds taken near the
Pittsburgh International Airport approximately 100 km west of Conemaugh and Seward (but at
similar elevations to Conemaugh and Seward).
As described in Pennsylvania's supplemental modeling submittal, there were some problems
with the SODAR capture20. Figure 5.2-29 shows the 150-m SODAR wind rose. There appears to
be a "gap" in the southwest quadrant of the wind rose. The 100-m tower wind rose does not
show the same feature (see Figure 5.2-30). The wind direction count suppression extended
through all layers of the SODAR measurements. Figure 5.2-31 shows the 100-m met tower and
500-m SODAR wind roses.
After an analysis of the SODAR data and other research, this wind direction data gap in the
southwest quadrant was attributed to moisture plumes from Conemaugh's 2 hyperbolic cooling
towers. Moisture plumes interfered with the SODAR signals when winds were coming from this
direction. As shown, this interference did not impact wind direction measurements taken on the
100-m met tower. To compensate for this wind direction interference, final processed wind
directions were set to missing when SODAR level wind directions were between 235° and 290°
as per AECOM's analysis. SODAR wind measurements outside of this range were retained so
that wind information collected by the SODAR could still be utilized. Missing wind directions
(in the SODAR measurements) would be filled by extending the 100-m tower wind direction
data upwards for any given hour removed. SODAR wind speed and sigma w measurements
taken during these hours were available for the final AERMET processed profile file.
20 See Appendix A of AECOM's Supplemental SO 2 NAAQS Compliance Modeling Report for the Indiana, PA SO 2
Non-Attainment Area - Focus on Areas Near the Conemaugh and Seward Generating Stations (Revision No. 1),
December 2019
62
-------
As a check, EPA also created wind roses for each SODAR. level using the 100-m tower wind
direction substitution described previously. Figure 5.2-32 shows wind roses for the 150-m level
with the 100-ni tower wind direction substitution and wind roses showing all the actual 150-m
SODAR collected data. Some "filling" is evident in the SODAR wind direction gap. Note that
AERMET does not do this substitution explicitly within the AERMET profile file. Wind
directions will be "extended" upwards from the next available level from the tower wind
measurements.
Figure 5.2-28. Meteorological Collection Sites Used to Develop the AERMET Files
Met Sites
H Johnstown -Cambria
County Airport
a Pitts burgh International
Airport
® Ash Site #1 Met Tower
~ Indiana, PA
N on att ai n me nt A rea
Elevation
Meters
High : 2035
[Keystone^
JSfjSeward
•nemaugh^
Meteorological Data Sites for EPA Modeling Analysis of Indiana, PA Nonattainment Area
60 Kilometers
—I
>EFA
63
-------
Figure 5.2-29. Ash Site #1 150-meter SODAR Wind Rose
Indiana, PA S02 SIR - On-Site Wind Rose: SODAR (150-m)
Indiana, PA S02 SIP - On-Site Wind Rose: Met Tower (100 m)
Figure 5.2-30. Ash Site #1 100-ineter Tower Wind Rose
1.6 Kilometers
Legend
On-Site SODAR (150 m}
Wind Speeds
[ 0.5-2.10 m/s
| 2.10 - 3 .60 mfe
| 3 90 - 5.70 m£
¦ 5.70-8.80 m&
¦ 8.80 -11.10 rrVs
] >11.10 mA
Indiana, PA Nonattainrrtent/£vee
Legend
On-Site Tower (100-m)
Wind Speeds
0.5-2.10 mft
2.10 - 3 60 mfts
| 360 - 5 70 m/s
| 5.70-8 00 m/s
| 8 80 -11 10 m/a
I »11.10 m/s
Indiana. PA Nonattainment Area
¦R-JEW [«
mmm
64
-------
Figure 5.2-31. 100-meter Met Tower and 500-meter SODAR Wind Roses
Ash site #1 (100-meter Tower) Wind Rose
10%
5%
Z5%
~
~
*
,*
~
4
Ash site #1 (500-meter SODAR) Wind Rose
1Z8%
10%
7.S%
s
0 to 2 2 to 4 4 to 0 6 to 8 8 to 10 10 to 16.9
{m s~1)
Frequency of counts by wind direction {%)
0 to2 2 to 4 4 to 6 6 to 8 8 to 1010 to 19.30
{m s"')
Frequency of counts by wind direction (%)
Figure 5.2-32. 150-meter SODAR Wind Roses With and Without 100-m Tower
Substitution
Ash site #1 (150-meter SODAR with Tower Adjustment) Wind Rose Ash site #1 (150-meter SODAR) Wind Rose
10%
7.5%
S%
N
I 1
S
0 to 2 2 to 4 4 to 6 ©to 8 8 to 1010 to 16.07
(m s"1)
Frequency of counts by wind direction (%)
0 to 2 2 to 4 4 to 0 0 to 8 8 to 1010 to 10.07
(m s"1)
Frequency of counts by wind direction (%)
65
-------
EPA reprocessed the Ash Site #1 met tower and SODAR through AERMET using the previously
described AERSURFACE settings. These impact the monthly surface roughness, albedo and
Bowen ratio values (in the sfc output file). This was done within the AERMET stage 2
processing step. This generated multiple AERMET surface files for each of the AERSURFACE
varying categories (as listed in Table 5.2-11). Files were generated for average-no snow, dry-no
snow, dry-snow and wet-no snow settings within AERSURFACE. The final 1-year site specific
surface met file had to be concatenated together based on each month's setting. This was done
using R to produce the final surface file. The AERMET profile file did not need similar
adjustments.
Tower and SODAR turbulence measurements were collected as part of the site-specific
meteorological survey period. As noted in the AERMET users guide (section 4.7.6.5) and
Appendix W, for site-specific data sets, such as the Ash Site #1 used in this modeling analysis,
turbulence measurements should not be used in tandem with the adjusted u-star option. EPA's
own analysis of its field study data (see FR 82, 5187, January 17, 2017; Appendix W) showed
results with site-specific turbulence data did not show a bias toward underprediction without the
adjusted u-star option (ADJU*) but did show a bias toward underprediction using turbulence
data with the ADJ U* option. Two sets of meteorological data were therefore available for final
processing. One using the turbulence measurements and one excluding the turbulence
measurements but utilizing the adjusted u-star (ADJ U*) option in the final stage 2 AERMET
processing step.
AERMET processing also included surface data from the National Weather Service (NWS)
Johnstown-Cambria County airport ASOS site for the Ash Site #1 collection period. The stage 2
processing utilized AERMET's cloud cover and temperature substitutions. AERMET includes
substitutions for missing cloud cover and temperature data based on linear interpolation across
gaps of 1 or 2 hours. Linear interpolation across short gaps is a reasonable approach for these
variables since ambient temperatures tend to follow a diurnal cycle and do not vary significantly
from hour to hour. Additionally, AERMOD is relatively insensitive to hourly fluctuations in
cloud cover, especially during convective hours since the heat flux is integrated across the day.
Gaps of 1 or 2 hours for these parameters near the early morning transition to a convective
boundary layer may result in all convective hours for that day being missing. A more complete
description of cloud cover and temperature substitution procedures when using site-specific and
NSW data in AERMET is available in section 4.7.6.6 of EPA's (2021) AERMET users guide.
AERMOD needs a morning temperature profile to characterize dispersion during the day. These
can be provided from available NWS sites scattered across the United States. Upper air morning
and evening (12 GMT and 24 or 0 GMT) soundings are scheduled to be taken on a daily basis
near Pittsburgh, PA. These measurements are the closest site to the Conemaugh and Seward
power plants and are representative of conditions near these facilities.
66
-------
EPA reviewed the AERMET stage 1 reports after it processed the Ash Site #1 site-specific data
and the corresponding upper air rawinsonde data from Pittsburgh. AERMET was run with the
option to capture the morning sounding within ±3 hours of the normal 12 GMT sounding
collection time. EPA noted that the AERMET stage 1 report contained several warning messages
related to the Pittsburgh upper air measurement processed in AERMET. AERMET generated the
following warning message:
UPPERAIR W31 READFSL SKIP SOUNDING; 1 ST LEVEL TYPE 4 NOT TYPE 9,
FOR SOUNDING # 1059 DATE: 20160606 HR 07
This warning message was attached to 3 dates across the 1-year Ash Site #1 collection period.
They were 6 June 2016, 19 June 2016 and 26 August 2016. EPA reviewed the final concatenated
AERMET surface file using R and determined that these 3 dates did not contain any convective
mixing height calculations; all convective mixing heights in the sfc file were coded as missing.
This was due to the morning soundings on these dates not being processed in AERMET. A
review of the Pittsburgh, PA upper sounding file showed that vertical profile measurements
existed on these 3 dates, but the surface code line (coded 9) was missing from these particular
morning soundings.
To allow AERMET to process these 3 morning soundings, the upper-air input file was edited to
include a line 9 code filled in with missing parameters to reflect the line of missing surface
information21. This allowed the morning upper air soundings to be fully processed in AERMET
to allow for AERMOD to generate model concentrations during the daytime hours for these 3
"missing" dates.
To the best of our ability, AERMET processing was conducted in accordance with current EPA
guidance. AERMET produced the surface and profile files needed as input into AERMOD with
some editing via R for the surface and upper air profile file described in this section.
Final Processing to Produce AERMOD-Ready Meteorological Data: EPA's Modeling TAD
recommends modeling 3 years of emissions to assess source impacts. The most representative
meteorological data available to assess the impacts of Conemaugh and Seward is the site-specific
meteorological data collected at the Ash Site #1. This data set represents 1-year of data (1 Sep
2015 through 31 Aug 2016) whereas the emissions window is 3 years in length (1 Jul 2017
through 30 Jun 2020). The meteorological data, therefore, needed to be adjusted to match the
dates of 3-year hourly emissions data.
EPA used R to change the years of the Ash Site #1 meteorological data to match the 3-year
emission period. Following section 7.4 of EPA's Modeling TAD, months, days, and hours
remain unchanged. Both the Ash Site #1 and emission period contained one leap year so the
meteorological data for that date was only used once over the 3-year modeling interval. All other
dates and hours were repeated 3 times over the simulation period, so the met data matched the
21 A more thorough explanation of the steps EPA took to ensure these missing line 9 codes were included for
AERMET processing can be found in the Region 3 rundown presented during the 2022 Regional/State/Local
Modelers workshop. See slides 5-7 available on EPA SCRAM website.
67
-------
emissions.
As noted in EPA's Modeling TAD, the use of older site-specific data should be used with
caution if source emissions are somewhat dependent on meteorological data. Electric Generating
Units (EGUs) like Conemaugh, Seward and the Cambria County waste-coal sources are subject
to load demands from the electric grid in which they are connected. Extremes in weather
conditions, such as cold snaps or heat waves can increase electric demand and therefore
influence emissions for sources providing power to the electric grid.
Conemaugh and Seward's maximum combined power generation is equal to approximately 2.4
gigawatts of electricity, enough power to supply approximately 1.5 million homes. These plants
far surpass local power needs; the combined population of Indiana and Armstrong counties is
less than 150,000 and both counties have experienced long-term population declines. The vast
majority of the power generated by these plants is probably exported via the PJM electric grid.
Based on this information, EPA feels it would be difficult to assess the impact of having older
site-specific meteorological data matched with the more recent 3-year simulation period. We
expect there to be some impact, but no analysis was undertaken to assess differences in grid
demand and its impact on Conemaugh and Seward's SO2 emissions between the 2 data sets.
5.2.7. Modeling Parameter: Background SO2 Concentrations
The Modeling TAD offers 2 mechanisms for characterizing background concentrations of SO2
that are ultimately added to the modeled design values: 1) a "tier 1" approach, based on a
monitored design value, or 2) a temporally varying "tier 2" approach, based on the 99th
percentile monitored concentrations by hour of day and season or month. Section 8.3 of EPA's
Guideline on Air Quality Models provides additional discussion on background monitoring
concentrations for air quality analyses. Additional guidance points regarding the determination of
background concentrations for the 1-hr SO2 NAAQS are also outlined in EPA's March 1, 2011
1-hour NO2 clarification memo including using temporally varying background concentrations.
Background concentrations are essential in constructing the design concentration, or total air
quality concentration, as part of any NAAQS analysis. In selecting an appropriate background
concentration, it is important to not include the ambient impacts of the project source under
consideration. Typically, state or local air monitoring stations (SLAMS) provide background
concentrations for air quality analyses.
To avoid source influence on background monitor concentrations from the primary Indiana, PA
nonattainment sources, Pennsylvania constructed background concentrations from the South
Fayette monitor in western Allegheny County in its original SIP and Supplementary Analysis.
The South Fayette monitor is roughly 77 km southwest of Keystone, 85 km south-southwest
from Homer City and 95 km west of Conemaugh and Seward.
Two other possible background monitors in the area include the Strongstown monitor in Indiana
County and the Johnstown monitor located in neighboring Cambria County. Figure 5.2-33 shows
68
-------
the location of the Indiana, PA SIP sources, the 3 Cambria County waste-coal sources and the 3
SO2 monitoring sites considered as possible background monitoring sites.
Figure 5.2-33. Background SO2 Monitoring Sites
learfidd
Krm strong!
diana
Strongs town]
Allegheny*
[SoutntF-ayettel
ambria
flohnstownj
^estmorelancfl
mmi
lOtrfei-Yet]
Possible 1-Hour SO2 Background Monitoring Sites for Westmreiand/Cambria County Area
Jefferson
Legend
Primary SOjSources
Conemaugh
Keystone
Homer City
Seward Power Plant
Cambria Cogen (Shut Down)
0 Colver Power
F*1 Ebensburg Power
Indiana, PA N onattainment
Area
Elevation
Meters,
^igh:
20 35
0 20 40 80 Kilometers
1 1 1 1 1 1 1 1 1
AEFA
69
-------
From a geographical perspective, the South Fayette monitor is the most distant monitor of the 3
considered. This monitor was used in the previous modeling analyses that supported
Pennsylvania's original and Supplemental Analysis SIP submittals. It was chosen to represent a
true background for the nonattainment area. South Fayette was considered far enough upwind of
the nonattainment area to not be impacted by the 4 primary SIP sources in the Indiana County,
PA nonattainment area. There are other SO2 monitors that are closer to the Indiana, PA
nonattainment area but they were found to be impacted by other nearby sources and therefore not
representative of true background.
The next 2 background sites are the Johnstown and Strongstown sites. Of the 2, the Johnstown
monitor is closest to the sources of interest (Conemaugh and Seward). The Johnstown monitor is
located in the City of Johnstown roughly 15 km southeast of the Conemaugh and Seward power
plants included in EPA's modeling analysis. Located along Stony creek River, the Johnstown
monitor sits at a significantly lower elevation, around 435 meters, than the surrounding terrain,
which rises to over 800 meters in places.
The Strongstown monitor is located in elevated terrain in eastern Indiana County near its
boundary with Cambria County. Base elevation at this monitor is approximately 580 meters.
Unlike the Johnstown monitor, there are no real imposing terrain features between it and any of
the 4 primary SIP sources in the Indiana, PA nonattainment area. The Chestnut and Laurel
ridges, by contrast, present several physical impediments to plumes originating from the Indiana,
PA SIP sources.
Historical Concentrations
EPA downloaded hourly SO2 concentrations for the South Fayette, Johnstown and Strongstown
monitors using R's RAQSAPI's library. This information was used to graphically display
monitor hourly concentrations along with yearly exceedances (hours above the 1-hour SO2
NAAQS of 75 ppb), yearly high 4th-high concentrations (or 99th% values) and design value
concentration between 2009 and 2021. EPA's original Round 1 SO2 designations were based on
2009-11 monitor design values though 2010-12 were also considered since designation were not
completed until October of 2013. This was done to examine any trends at the individual monitors
and determine and overall trends with the 3 monitoring sites to support the selection of the
background monitoring site.
Hourly SO2 Concentrations: South Fayette, Johnstown and Strongstowns' hourly SO2
concentrations were downloaded using R. Hourly values from 2009 through 2021 were
examined for each monitor. Plots showing hourly SO2 concentrations, Figure 5.2-24a-c, along
with the 1-hr SO2 NAAQS (75 ppb) were developed from each monitor and show general trends
over time.
70
-------
Strongstown's hourly SO2 concentrations appear to have more spikes than the other 2 monitors.
South Fayette's hourly SO2 concentrations appear to be the lowest of the 3 monitors. Spikes in
monitor SO2 concentrations have generally decreased over time at all 3 sites. Exceedances at
Strongstown were much more common than the other 2 monitors. These instances also appear to
have declined in recent years. The last exceedance at Strongstown appears to have occurred in
early 2017.
71
-------
Figure 5.2-24a. 2009 through 2021 Hourly SO2 Concentrations
South Fayette
South Fayette, PA Monitor (42-003-0067) S02 Concentrations in Parts Per Billion (ppb)
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
Date
Figure 5.2-341) (Continued). 2009 through 2021 Hourly S02 Concentrations
Johnstown
Johnstown, PA Monitor (42-021-0011) S02 Concentrations in Parts Per Billion (ppb)
2022
72
-------
Figure 5.2-34c (Continued). 2009 through 2021 Hourly S02 Concentrations
Strongstown
Strongstown, PA Monitor (42-063-0004) S02 Concentrations in Parts Per Billion (ppb)
150-
2009 2010 2011
2012
2013
2014
2015 2016
Date
2017
2018
2019
2020
2021
2022
Table 5.2-12 summarizes the 3 monitors' 99 Percentile (%) (high 4 high), design value and
days with 1 hr SO2 concentrations above 75 ppb (exceedance) from 2009 through 2021.
As previously shown in the hourly monitor plots, Strongstown has the highest SO2
concentrations of the 3 monitors considered for developing the background monitor
concentrations. Strongstown's design values and 99th% 1-hr SO2 concentrations are about 2
times higher than either South Fayette or Johnstown. Recent values at Strongstown are about 5
times lower than they were a decade ago. There is also a marked decline at Strongstown
beginning in 2016. Prior to 2016, Strongstown's 99th% values were in the mid-60 to mid-70 ppb
range. Afterwards, they fell into the mid-20 ppb range.
EPA believes changes in SO2 emissions at the Keystone and Homer City power plants account
for the decline in Strongstown's measured 1-hr SO2 concentrations. Homer City installed Novel
Integrated Desulfurization or NIDs on units 1 and 2 in 2016. These units are sometimes called
dry scrubbers since they use significantly less water than flue gas desulfurization or FGD units,
such as the ones operating on Homer City unit 3 and at the Conemaugh and Keystone power
plants.
73
-------
Table 5.2-12. Summary of 99th%, Design Values and Exceedances: SO2 in parts per billion
South Fayette
Johnstown
Strongstown
Year
H4H
Design
Exceedances
Valid
H4H
Design
Exceedances
Valid
H4H
Design
Exceedances
Valid
2009
63
2
Y
68
3
Y
82
4
Y
2010
39
1
Y
56
1
Y
95
7
Y
2011
28
43
0
Y
39
54
0
Y
68
82
2
Y
2012
20
29
0
Y
52
49*"
0
N
69
77
3
Y
2013
19
22
0
Y
32
41 *"
0
N
66
68
2
Y
2014
21
20
0
Y
54
46*"*
0
Y
66
67***
3
N
2015
18
19
0
Y
34
40*"
0
Y
73
68*"
3
Y
2016
9
16
0
Y
24
37
0
Y
39
59***
0
Y
2017
8
12
0
Y
19
26
0
Y
24
45
1
Y
2018
10
9
0
Y
17
20
0
Y
26
30
0
Y
2019
15
11
0
Y
13
16
0
Y
27
26
0
Y
2020
7
11
0
Y
8
13
0
Y
20
24
0
Y
2021
8
10
0
Y
10
10
0
Y
20
22
0
Y
'*" Data is Incomplete; Quarter(s) with < 75% Valid Days
EPA pulled SO2 emissions for both Keystone and Homer City that are reported to EPA's Clean
Air Markets Division (CAMD) as part of the Part 75 emissions reporting program. Each
facility's hourly SO2 emissions from 2009 through 2021 are depicted in Figure 5.2-35 and 5.2-
36. Hourly emissions are either measured or calculated based on method of determination codes
(MODC) for the flow and SO2 concentration instruments. This distinction, however, isn't
important for this analysis.
Keystone's CAMD SO2 emissions in 2009 are much higher because this period preceded the
installation of the facility's wet Flue-Gas Desulfurization (FGD) units. Additionally, the
facility's SO2 emissions show much less spiking after 2016. The red line on the figure represents
Keystone's modeled critical emission value or CEV, which is 9,711.1 lbs/hr. The impacts of the
installation of the NIDs controls on Homer City units 1 and 2 has a more dramatic impact on
total hourly SO2 emissions. Prior to 2016, Homer City's total SO2 emissions ranged from 10,000
to 50,000 lbs/hr. After the NIDs were installed, Homer City's hourly emissions fell to under
5,000 lbs/hr. Homer City's combined unit CEV is 6,360 lbs/hr. Anecdotally, it appears that
emission reductions at Homer City and possibly Keystone coincide with the dramatic drop in
Strongstown's monitored SO2 values.
74
-------
Figure 5.2-35. Keystone 2010-2020 Hourly SO2 Emissions
Keystone Hourly Emissions
CAMD Part 75 2009-21
O
w
Measured
Calculated
CEV
Figure 5.2-36. Homer City 2010-20 Hourly SO2 Emissions
Homer City Hourly Emissions
CAMD Part 75 2009-21
| 20,000-
O
CO
Measured
Calculated
CEV
ttiETite
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
Date
75
-------
Model Background Values
Model background values can be determined using monitor concentrations. EPA's modeling
analysis used a background concentration formulated following its March 1, 2011 1-hr NO2
clarification memo.22 Background concentrations were formulated using a season by hour of day
method described in pages 18-20 of the previously referenced clarification memo. Model
background values were constructed using R from 2019-21 hourly monitoring data for the South
Fayette, Johnstown and Strongstown monitors.
Each monitor's calculated season by hour of day background is shown in Tables 5.2-13a-c. The
seasonal breakdown is as follows; winter includes the months of December, January and
February, spring includes the months of March, April and May, summer includes the months of
June, July and August and fall includes the months of September, October and November. Hour
of day values represent the (seasonal) average of the 3-year sample period (2019-21). The table
also includes the number of mission hours over the 3-year period and the total number of
available hours for each hour of day.
Note that the Johnstown and Strongstown monitors have no valid measurements for hour 2. This
is because the Commonwealth of Pennsylvania, who maintains these monitors, consistently
performs maintenance checks (span checks for example) at this time preventing any sampling
during these hours.
All 3 monitors tend to have higher overall concentrations during the Winter and Fall seasons.
There is also a tendency for daytime (background) concentrations to be slightly higher than
overnight concentrations. Background peaks also tend to occur between the late morning and
early afternoon hours. This may be the result of downward mixing of overnight plumes as the
boundary layer expands due to daytime surface heating.
22 https://www.epa.gov/sites/default/files/2020-10/documents/additional clarifications appendixw faonilv~no2~
naaas final 03-01-2011.pdf
76
-------
Table 5.2-13a. South Fayette Monitor Season by Hour of Day Background Concentrations
South Fayette, PA (42-003-0067): 2019-21 Background S02 Concentrations (ppb) by Season/Hour of Day
Winter
Spring
Summer
Fall
Hour
Average
Missing
Total
Hours
Average
Missing
Total
Hours
Average
Missing
Total
Hours
Average
Missing
Total
Hours
1
2.3
0
271
2.3
0
276
2.0
0
276
2.7
0
273
2
3.3
0
271
2.3
0
276
2,0
0
276
2.7
0
273
3
3.0
0
271
2.7
0
276
1.7
0
276
2.0
0
273
4
3.7
0
271
3.7
0
276
2.0
0
276
2.0
0
273
5
3.0
0
271
2.0
0
276
1.7
0
276
2.3
0
273
6
3.7
0
271
2.7
0
276
2.0
0
276
2.0
0
273
7
2.7
0
271
3.0
0
276
2.3
0
276
2.0
0
273
8
3.0
0
271
3.3
6
276
3.3
8
276
2.7
6
273
9
3.7
2
271
3.3
3
276
3.3
1
276
2.3
0
273
10
4.0
1
271
3.0
12
276
3.0
23
276
3.0
10
273
11
5.7
9
271
3.0
25
276
3.0
21
276
3.0
24
273
12
4.7
27
271
3.0
14
276
3.0
6
276
3.0
14
273
13
4.7
13
271
2.3
7
276
3.0
5
276
4.3
8
273
14
4.7
6
271
2.0
1
276
3.3
4
276
4.0
2
273
15
4.0
1
271
2.3
1
276
3.3
4
276
2.7
0
273
16
4.7
0
271
2.0
0
276
4.0
2
276
2.7
0
273
17
3.3
1
271
2.3
0
276
5.0
1
276
3.7
0
273
18
3.7
0
271
2.3
0
276
5.0
0
276
4.0
0
273
19
3.7
0
271
3.3
0
276
4.3
0
276
2.7
0
273
20
4.0
0
271
3.3
0
276
4.0
0
276
2.3
0
273
21
2.7
0
271
3.0
0
276
3.7
0
276
2.0
0
273
22
4.3
0
271
2.0
0
276
2.7
0
276
2.7
0
273
23
3.3
0
271
2.7
0
276
3.0
0
276
2.7
0
273
24
3.3
0
271
2.3
0
276
2.0
0
276
2.7
0
273
77
-------
Table 5.2-13b. Johnstown Monitor Season by Hour of Day Background Concentrations
Johnstown, PA (42-021-0004): 2019-21 Background S02 Concentrations (ppb) by Season/Hour of Day
Winter
Spring
Summer
Fall
Hour
Average
Missing
Total
Hours
Average
Missing
Total
Hours
Average
Missing
Total
Hours
Average
Missing
Total
Hours
1
3.3
3
271
3.3
13
276
0.7
21
276
1.0
6
273
2
271
271
276
276
276
276
273
273
3
3.3
2
271
2.3
15
276
0.3
21
276
1.0
4
273
4
3.7
2
271
2.3
14
276
0.3
21
276
0.7
3
273
5
3.0
3
271
1.7
13
276
0.3
21
276
0.3
3
273
6
3.7
3
271
3.3
13
276
0.7
21
276
0.0
3
273
7
3.0
3
271
3.3
14
276
0.7
21
276
0.7
3
273
8
3.0
3
271
3.3
14
276
1.7
21
276
1.3
2
273
9
3.7
2
271
3.7
15
276
2.3
21
276
1.7
4
273
10
4.0
2
271
4.7
15
276
4.0
24
276
2.7
3
273
11
6.0
2
271
6.3
14
276
5.7
26
276
3.0
5
273
12
6.0
5
271
5.3
15
276
6.3
23
276
3.3
4
273
13
7.0
4
271
3.7
14
276
4.3
26
276
4.0
2
273
14
10.0
7
271
3.7
16
276
4.0
23
276
5.0
5
273
15
9.0
6
271
3.0
17
276
4.7
22
276
3.3
4
273
16
8.0
4
271
3.3
14
276
4.7
22
276
4.7
3
273
17
8.0
1
271
3.3
13
276
5.3
21
276
4.0
2
273
18
4.3
2
271
4.0
13
276
5.3
21
276
3.0
3
273
19
4.3
2
271
3.3
13
276
3.7
21
276
2.0
4
273
20
4.0
2
271
3.0
13
276
2.7
21
276
2.0
4
273
21
3.0
3
271
3.7
13
276
2.7
21
276
1.3
5
273
22
4.7
3
271
3.7
13
276
1.3
21
276
1.3
4
273
23
4.7
3
271
3.0
13
276
1.3
21
276
1.3
4
273
24
3.7
3
271
2.7
13
276
0.7
21
276
1.0
4
273
78
-------
Table 5.2-13c. Strongstown Monitor Season by Hour of Day Background Concentrations
Strongstown, PA (42-063-0004): 2019-21 Background S02 Concentrations (ppb) by Season/Hour of Day
Winter
Spring
Summer
Fall
Hour
Average
Missing
Total
Hours
Average
Missing
Total
Hours
Average
Missing
Total
Hours
Average
Missing
Total
Hours
1
3.3
6
271
2.0
19
276
1.3
4
276
3.0
7
273
2
271
271
276
276
276
276
273
273
3
3.0
6
271
3.3
18
276
2.7
6
276
2.3
7
273
4
3.3
6
271
2.7
19
276
2.0
5
276
2.0
8
273
5
2.7
6
271
1.7
18
276
1.0
5
276
2.3
8
273
6
3.0
7
271
2.3
18
276
1.3
5
276
2.7
8
273
7
4.3
6
271
2.7
19
276
3.7
5
276
1.7
7
273
8
3.3
5
271
3.3
18
276
8.3
6
276
3.0
8
273
9
3.7
5
271
5.7
20
276
6.7
7
276
5.3
8
273
10
5.0
9
271
7.0
24
276
7.0
9
276
9.0
12
273
11
9.0
9
271
5.0
23
276
8.7
9
276
9.7
10
273
12
8.3
9
271
4.7
20
276
5.0
5
276
5.3
12
273
13
9.3
7
271
4.3
19
276
4.7
5
276
6.0
12
273
14
11.7
5
271
3.3
17
276
4.3
3
276
5.7
13
273
15
7.7
7
271
4.0
17
276
5.3
4
276
4.0
11
273
16
6.0
6
271
3.3
17
276
5.3
4
276
4.3
9
273
17
6.3
6
271
4.3
18
276
4.3
3
276
2.7
8
273
18
7.7
6
271
2.0
17
276
5.7
3
276
2.0
8
273
19
6.7
6
271
2.3
17
276
4.7
4
276
2.0
8
273
20
4.0
6
271
2.3
17
276
4.7
4
276
3.3
8
273
21
3.7
5
271
3.3
17
276
3.3
4
276
5.3
8
273
22
3.0
6
271
2.7
17
276
4.0
4
276
3.0
8
273
23
3.3
5
271
2.3
17
276
4.3
4
276
2.0
8
273
24
4.0
5
271
3.7
18
276
2.3
4
276
2.0
8
273
Background Monitor Selection
The Strongstown monitoring site was selected over the Johnstown and South Fayette sites
because it probably captures a reasonable background concentration from the Keystone and
Homer City sources that are not explicitly modelled. Distance wise, Strongstown is nearly as
distant from Keystone and Homer City as these sources are from the Laurel Ridge in
Westmoreland County. Strongstown's elevation also would allow it to be exposed to tall-stack
emissions from both plants. The Johnstown monitor is located along Stoneycreek in the
Sandyvale Memorial Gardens and Conservancy just south of the Cambria War Memorial arena.
Johnstown sits at a much lower elevation and is probably relatively unaffected by the tall stack
emissions from the Indiana, PA SIP sources except during the day when there is good vertical
mixing. Vertical mixing in the (Johnstown) valley may be impeded by local inversions given the
terrain.
79
-------
Once the background site (Strongstown) was selected, monitoring data was processed to
determine the model background concentration for the analysis. EPA followed a method
described in its March 1, 2011 1-hr NO2 clarification memo to generate seasonal by hour of day
background concentration that will be added to hourly generated AERMOD concentrations.
These values were entered into AERMOD using the BACKGRND keyword. Strongstown's SO2
concentrations for hour 2 were consistently missing due to this hour being used for monitor span
checks and other maintenance activities. An interpolated value using the hour 1 and 3 values was
used as a background concentration for the AERMOD simulation.
5.2.8. EPA Site-Specific Adjusted U-star Modeling Summary and Results
EPA ran AERMOD using actual hourly SO2 emissions for the Conemaugh and Seward power
plants in Indiana County and the Cambria County waste coal sources over a 3-year period as
described previously. Hourly emissions for Conemaugh and Seward were provided as part of the
September 2020 materials supplied by AECOM and shared with EPA and Pennsylvania. EPA
developed hourly emissions for the Cambria County sources using CAMD emissions and unit
heat input information to generate hourly varying stack parameters based on Round 3 DRR
modeling done previously by Pennsylvania.
As previously described, AERMOD (version 22112) was used with adjustments made in BPIP to
correct Seward's stack height and building layouts, seasonal by hour of day background
concentrations from Strongstown, and reprocessed meteorological data utilizing AERSURFACE
output and adjusted SODAR data from the Ash Site #1. Final processed meteorological data
from the Ash Site #1 was reformatted to match the hourly emissions files for Conemaugh and
Seward along with the Cambria County sources in accordance with EPA's Modeling TAD. We
note that EPA utilized the VECTORWS option within AERMOD to account for vector (not
scalar) measurements via the SODAR inputs in the AERMET preprocessor stage. Pennsylvania's
Supplemental Analysis used both vector and scalar processing with AERMOD and determined
the scalar version produced slightly higher model results. EPA did not make this comparison but
chose to use the VECTORWS as the more appropriate option within AERMOD for its analysis
based on the rationale provided above.
EPA produced 2 sets of final AERMOD-ready meteorological files. One set using the adjusted u-
star option in AERMET (stage 2) without the Ash Site #1 tower and SODAR turbulence
measurements (both horizontal and vertical, SA and SW) and another set using the Ash Site #1
tower and SODAR turbulence measurements (but not the adjusted u-star processing option).
Results for the EPA Site-Specific Adjusted U-Star modeling are provided in this section, and the
following section provides the EPA Site-Specific Turbulence results. Both modeling analyses
conducted by EPA produced design values over the NAAQS. An analysis to determine if either
meteorological data set is more appropriate was not conducted as part of EPA's modeling
analyses.
80
-------
Pennsylvania's Attainment Plan for the Indiana, PA SO2 NAA utilized the adjusted u-star
meteorological data set for areas near the Conemaugh and Seward power plants in the extreme
southeast portions of the Indiana, PA 1-hr SO2 nonattainment area.
Final design value concentrations for the EPA Site-Specific Adjusted U-star modeling are shown
in Figure 5.2-37. The model peak concentration (117.9 ppb) occurred along the Laurel Ridge
east of the Conemaugh and Seward power plants. Model concentrations exceeded the 1-hr SO2
NAAQS along portions of the ridge. Figure 5.2-38 shows a close up focused on the areas where
AERMOD shows receptors with modeled design values that exceed 75 ppb. Model violations of
the 1-hr SO2 NAAQS occur along the Laurel Ridge south of the Conemaugh River Gorge.
Modeled violations occur in Westmoreland along the portion of the Laurel Ridge facing
Conemaugh and Seward and also on the backside of the ridge in Cambria County. Violations
extend about 7 km southwest along the ridgeline. The demarcation line dividing Cambria and
Westmoreland counties roughly follows the top of the Laurel Ridge.
81
-------
Figure 5.2-37. AERMOD Adjusted U-Star Results for All Sources Plus Background
EPA Modeling Analysis of Portions of the Laurel Ridge Near Conemaugh and Seward
Legend
S^Peak Model Receptor
Adjusted U-star Run
SO2 1-Hr Design Value
~ <25.0 ppb
E3 25.0 - 50.0 ppb
¦ 50.0-75.0 ppb
Q75.0 - 100.0 ppb
¦1> 100.0 ppb
—Indiana, PA Nonattainment
Area
Elevation
Meters
High : 2035
(olver
Indiana
Ca mbria"uogen"( S hut)Uo>vn)]
, EbensburgBr-
parrTpnai
Gonemaugni
Westmoreland
20 Kilometers
vyEPA
82
-------
Figure 5.2-38. Close-up of Violating Receptors Along the Laurel Ridge
Indiana
Cambria
Westmoreland
EPA Modeling Analysis of Portions of the Laurel Ridge Near Conemaugh and Seward
035ZDGS}*
Legend
t Peak Model Receptor
—Indiana, PA Nonattainment
Area
Adjusted U-star Run
SO21-Hr Design Value
Zj<25.0 ppb
3 25.0 - 50.0 ppb
~ 50.0-75.0 ppb
I75.0 - 100.0 ppb
i~!> 100.0 ppb
Elevation
Meters
^ High : 2035
Low : -99 n
0 2.25 4.5 9 Kilometers O CR/V E^ironm?ntal Protection
| 1 1 1 1 1 1 1 1 Agency
Table 5.2-14 summarizes results for the peak model receptor, which had a simulated design
value of 117.9 ppb. EPA converted the AERMOD concentration from Lig/nr to parts per billion
or ppb by using a value of 196.4 jig/m3 as representing the 1-hr SO2 NAAQS value (75 ppb).
Universal Transverse Mercator or UTM (zone 17) location, elevation and AERMOD hill-height
scales are also included along with each year's 99th% value. Each 99th% value's corresponding
date and hour of day during the 3-yr model simulation period are also included in the table. The
simulated peak model receptor's design value is the average of each year's high 4th high (or 99th
%) daily 1-hr maximums over the 3-year simulation period. Note there is some year-to-year
variability in the 99,h% values that contribute to the 3-year modeled design value. Model output
indicates values exceed the 1-hr SO2 NAAQS through the 12th rank.
83
-------
Table 5.2-14. Final EPA Site-Specific Adjusted U-star Model Peak Receptor Summary
EPA AERMOD 22112 Adjusted U-star Run Summary for Peak Model Receptor
Item
EPA DRR Simulation Value
UTM Easting (m)
669816
UTM Northing (m)
4471618
Elevation (m)
743.69
Hill Height Scale (m)
762.01
S02 Design Value (ppb)
117.87
S02 Design Value (ug/m3)
308.67
Year 1 S02 High 4th High (ppb)
143.2
Year 1 Date
06-28-2018
Year 1 Hour
23
Year 2 S02 High 4th High (ppb)
122.59
Year 2 Date
01-05-2019
Year 2 Hour
17
Year 3 S02 High 4th High (ppb)
87.83
Year 3 Date
07-01-2019
Year 3 Hour
22
The peak model receptor's 99th% values generally occurred late in the day during the evening
and overnight hours. To examine this possible trend, EPA pulled all 2,580 model receptors
(including the peak model receptor) that violated the 1-hr SO2 NAAQS and examined the hour of
day for each of the 99th% values that made up the receptor's design value (3 hour of day values
for each violating receptor). Figure 5.2-39 shows the hour of day occurances for the violating
model receptors. The 99th% values for the violating model receptors appear to occur more
frequently in the early overnight hours (hours 20 through 23) and hour 6. This suggest model
concentrations, to some extent, tend to be higher under more stable atmospheric settings. There
may also be an emission/operating trend that accounts for this observation since actual hourly
varying emission rates were modeled.
84
-------
Figure 5.2-39. Hour Of Day for 99th% Values for Violating Model Receptors
High 4th-High Hour of Day Occurrences
for Model Receptors Above 75 ppb
1000-
800-
c
o
O
600-
400-
200-
254 254
101
~
111
~
115
88
~
~
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour of Day
Beginning with version 11059, AERMOD has incorporated 3 output options to support the 1-
hour NO2 and SO2 standards, especially the analyses that may be required to determine a
source's (or group of sources) contributions to modeled violations of the NAAQS and for
comparison to the Significant Impact Level (SIL). The form of these standards, based on
averages of ranked values across multiple years, complicates this analysis, especially for the 1-
hour NO2 and SO2 standards, which are based on ranked values from the distribution of daily
maximum 1-hour averages. The MAXDCONT option within AEMOD, applicable to 1-hour SO2
standards, can be used to determine the contribution of each user-defined source group to the
high ranked values for a target source group, paired in time and space. This is accomplished as
an internal post-processing routine after the main model run is completed. Section 3.7.2.8 of
EPA's AERMOD user guide (2022) has a more detailed description of this processing option.
EPA utilized the MAXDCONT options within AERMOD to ascertain the contributions of
emissions from Conemaugh, Seward and the Cambria County sources along with the background
concentration to the final modeled design values. These are shown in Table 5.2-15 for the
adjusted u-star run's peak model receptor. Seward is the largest contributor to the peak model
receptor's AERMOD concentration, contributing to almost 96% of the peak receptor's modeled
design value. The next largest contributor is the season by hour of day background concentration
at almost 4%. Emissions from Conemaugh and the 3 Cambria County sources are minor
contributors with a combined impact of less than 0.03%.
85
-------
Table 5.2-15. MAXDCONT Output Source/Background Contribution to Peak Receptor
Contributions in jig/m3, Total Concentration Converted to ppb using 196.4 jig/m3= 75 ppb
EPA AERMOD 22112 Adjusted U-Star Run Peak Receptor Source Contribution in ug/m3
Conemaugh
Background
Total
0.08658
295.85021 0.00196
12.72997
308.66872
EPA was able to pull each source's hourly emission rates at for each of the 99th% occurances that
made up the peak model receptor's modeled 1-hr SO2 design value. Based on the peak model
receptor's MAXDCON results, the only contributing source for the model simulation is Seward.
Table 5.2-16 shows each modeled source's hourly emission rate at the time the 99th% modeled
value occurred. Under the right meteorological conditions, it appears Seward can cause
exceedances along the Laurel Ridge with SO2 emission rates in the 1,500 to 1,600 lbs/hr range.
Even lower rates may cause model exceedances since the simulation produced values above 75
ppb at ranks below the high 4th high according to the MAXDCON file output.
Table 5.2-16. Hourly Emission Rates for Modeled 99th% or High 4th-High Occurances
EPA AERMOD 22112 Adjusted U-star Run: Peak Receptor High-4th High & Corresponding S02 Emission Rates
Date
Hour
S02 (ug/m3)
S02 (ppb)
Conemaugh
(lbs/hr)
Seward
(lbs/hr)
Cambria Cogen
(lbs/hr)
Colver
(lbs/hr)
Ebensburg
(lbs/hr)
06-28-2018
23
375.00234
143.20
634.1
3,245.2
711.92
711.92
380.17
01-05-2019
17
321.01417
122.59
370.5
1,553.1
807.16
711.12
239.69
07-01-2019
22
229.98964
87.83
1,467.2
1,514.6
0.00
706.36
294.45
86
-------
5.2.9. EPA Site-Specific Turbulence Summary and Results
EPA also modeled using the site-specific (Ash Site #1) processed turbulence data as a
designation analysis for portions of Cambria and Westmoreland counties near the Conemaugh
and Seward power plants. Preprocessing steps for this analysis are identical to the processing
steps contained in section 5.2.1- 5.2.7 of this document. The only difference is the
meteorological processing for this modeling analysis utilized the site-specific turbulence
measurements in the AERMET preprocessor (without using the adjusted u-star option in the
AERMET stage 3 processing step). At this time, EPA is not endorsing either data set (adjusted u-
star or turbulence) as the most suitable for the circumstances. As noted, the adjusted u-star data
set is consistent with Pennsylvania's Indiana attainment plan SIP submission. EPA believes
consistency in this modeling parameter is important because the area in Indiana County, PA and
Cambria and Westmoreland Counties that are the focuses of these analyses are right next to each
other, spanning only a small radius from the two key facilities of Seward and Conemaugh.
Final model results for the site-specific turbulence processed meteorological data showed
violations of the 1-hr SO2 NAAQS along the Laurel Ridge. The extent and magnitude of the
modeled 1-hr SO2 NAAQS violations were significantly reduced compared to the adjusted u-star
values. The peak modeled concentration using the site-specific turbulence processed
meteorological data was 77.3 ppb. A total of 24 model receptors violated the 1-hr SO2 NAAQS.
This is substantially fewer violating receptors than the adjusted u-star simulation.
Figure 5.2-40 shows model results using the site-specific turbulence measurements over the EPA
model grid. The extent of the areas violating the 1-hr SO2 NAAQS is much smaller than the
adjusted u-star run.
A close up of portions of the Laurel Ridge containing the peak modeled values is shown in
Figure 5.2-41. The area where modeled concentrations exceed the 1-hr SO2 NAAQS is much
smaller than the simulation using the adjusted u-star processed meteorological data. Modeled
violations are confined to portions of the Laurel Ridge facing Conemaugh and Seward and do
not extend past the ridgeline into Cambria County.
87
-------
Figure 5.2-40. EPA Site-Specific Turbulence Modeling Results for All Sources Plus
Background
O'oneriiaugj^Hfl
Westmoreland
EPA Modeling Analysis of Portions of the Laurel Ridge Near Conemaugh and Seward
Indiana
I ( ambria"uogen"( Miut'UwHl )1
lliTnslnTr 75.0 ppb
Elevation
Meters
High : 2035
0 5 10 20 Kilometers O CRA Eni'ranmf03•
| 1 1 1 1 1 1 1 1 Er2T\ Agency
If one compares modeled design values within the lower terrain of the Ligonier Valley from both
simulations, it appears the simulation using the site-specific turbulence measurements produces
higher model design values than the adjusted u-star simulation. Modeled design values for the
site-specific turbulence simulation are about twice as high as the adjusted u-star in the Ligonier
Valley though they are still well below the NAAQS. This result coupled with the higher adjusted
u-star simulation design values versus the site-specific turbulence design values along the Laurel
Ridge suggests there is an elevation sensitivity between the 2 meteorological data sets. The
adjusted u-star processed meteorological data produces higher model concentration in elevated
terrain compared to the meteorological data processed with the site-specific turbulence
measurements. In lower terrain, it appears the opposite occurs, the adjusted u-star meteorological
data produces lower model design values than the site-specific turbulence measurements.
88
-------
As a clarifying point, EPA is not, at this time, endorsing the use of either the adjusted u-star or
the use of the site-specific turbulence measurements as the appropriate meteorological input into
AERMOD. The merits of each meteorological data set were not fully investigated for our
analysis. As we have shown, the selection of either data set shows modeled design values along
the Laurel Ridge in Westmoreland County exceed the 1-hour SO2 design value (75 ppb) thus
supporting EPA's contention that the area needs to be redesignated to nonattainment
Figure 5.2-41. Close-up of Violating Receptors Along the Laurel Ridge
EPA Modeling Analysis of Portions of the Laurel Ridge Near Conemaugh and Seward
Legend
^Peak Model Receptor
0 Receptors Above 75ppb
Turbulence Run
SO21-Hr Design Value
E=]<25.0 ppb
~ 25.0 - 50.0 ppb
¦ 50.0-75.0 ppb
B> 75.0 ppb
-Indiana, PA Nonattainment
Area
Elevation
Meters
^ High : 2035
Low : -99
0 2.25 4.5 9 Kilometers
1 1 1 1 1 1 1 1 1
oEPA
Table 5.2-17 summarizes results for the peak model receptor, which had a simulated design
value of 77.3 ppb. EPA converted the AERMOD concentration from |ig/m3 to parts per billion or
ppb by using a value of 196.4 ng/m3 as representing the 1-hr SO2 NAAQS value (75 ppb).
Universal Transverse Mercator or UTM (zone 17) location, elevation and AERMOD hill-height
scales are also included along with each year's 99th% value. Each 99,h% value's corresponding
date and hour of day during the 3-yr model simulation period are also included in the table. The
simulated peak model receptor's design value is the average of each year's high 4th high (or 99th
%) daily 1-hr maximums over the 3-year simulation period.
89
-------
There is some year-to-year variability in the 99th% values that make up the 3-year modeled
design value. The first year's 99th% value is just below 75 ppb and the third-year value is
actually below 75 ppb. Second year modeled 99th% values is well above 75 ppb and brings the
calculated modeled concentration above the 1-hr SO2 NAAQS.
Table 5.2-17. Final EPA Model Peak Receptor Summary
EPA AERMOD 22112 Turbulence Run: Peak Receptor High-4th High &
Corresponding S02 Emission Rates
Item
EPA DRR Simulation Value
UTM Easting (m)
670244
UTM Northing (m)
4471888
Elevation (m)
751.43
Hill Height Scale (m)
765.23
S02 Design Value (ppb)
77.28
S02 Design Value (ug/m3)
202.38
Year 1 S02 High 4th High (ppb)
74.7
Year 1 Date
09-15-2017
Year 1 Hour
08
Year 2 S02 High 4th High (ppb)
101.92
Year 2 Date
11-08-2018
Year 2 Hour
07
Year 3 S02 High 4th High (ppb)
55.23
Year 3 Date
06-08-2020
Year 3 Hour
23
The modeled hour for each of the peak receptor's 99th% value appears to be more variable than
the adjusted u-star simulation. EPA examined the hour of day for each of the violating receptors'
99th% values and plotted them. Figure 5.2-42 shows the distribution of hour of day for the
violating receptors; 24 receptors with 3 99th% values. Given there are far fewer receptors to draw
from, it's difficult to extrapolate any definitive patterns. It appears that higher values are still
occuring in the early overnight hours and close to sunrise. This result indicates peak model
concentrations may be occuring under stable atmospheric conditions.
90
-------
Figure 5.2-42. Hour Of Day for 99th% Values for Violating Model Receptors
High 4th-High Hour of Day Occurrences
for Model Receptors Above 75 ppb
C
D
o
O
2
Q
~.
~
o-
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour of Day
Utilizing AERMOD's MAXDCONT option, EPA examined all 24 violating model receptor's
source contributions. Table 5.5-10 shows the source contributions for each model receptor that
exceeded the 1-hr SO2 NAAQS. Similar to the adjusted u-star peak model receptor results,
Seward appears to be the most significant contributor to modeled violations. Background
contributions are the next largest contributor but generally less than 10%. The Cambria County
sources are contributing less than 1 ppb to modeled design values at the violating model
receptors.
There are only 3 instances where Conemaugh is a significant contributor to modeled violations.
They are highlighted in beige on the table. In these 3 instances, Seward's contribution is higher
than Conemaugh's with Conemaugh contributing between 24 to 38% towards the receptor's
modeled concentration.
91
-------
Table 5.2-18. MAXDCONT Output Source/Background Contribution to Peak Receptor
EPA AERMOD 22112 Turbulence Run Peak Receptor Source Contribution in ug/m3
Rank
Conemaugh
Seward
Cambria
County
Background
Total
PPb
Rank
Conemaugh
Seward
Cambria
County
Background
Total
ppb
1
0.03626
194.19771
0.29712
7.84724
202.37834
77.28
17
46.68970
134.71839
0.99311
15.08414
197.48534
75.41
2
0.04318
187.94874
1.12893
12.81716
201.93802
77.11
18
0.03791
188.98841
0.29480
7.84724
197.16836
75.29
3
0,03791
193.70608
0.29493
7.84724
201.88615
77.10
19
0,03679
181.52484
1.12327
14.21222
196.89712
75.19
4
0.03441
193.08195
0.29913
7.84724
201.26273
76,86
20
75.36091
111.66587
0.18139
9.59107
196.79925
75.15
5
0.03717
185.60357
1.12025
14.21222
200.97322
76.75
21
0.03480
181.35577
1.16082
14.21222
196.76362
75.14
6
0.03683
192.39006
0.29643
7.84724
200.57056
76,59
22
74.52908
113.97227
0.36118
7.84724
196.70977
75.12
7
0.03505
192.20603
0.29854
7.84724
200.38686
76.52
23
0.01868
186.91490
0.17191
9.59107
196.69656
75.11
8
0.04301
185.78020
1.11869
12.81716
199.75905
76.28
24
0.03005
189.14738
0.14893
7.32409
196.65045
75.10
9
0.03580
184.09130
1.13897
14.21222
199.47829
76.18
25
10
0.03844
191.15933
0.29413
7.84724
199.33914
76.12
26
11
0.03859
190.67668
0.29390
7.84724
198.85642
75.94
27
12
0.02278
190.43648
0.12070
7.84724
198.42720
75.77
28
13
0.04298
184.35491
1.13010
12.81716
198.34515
75.74
29
14
0.02125
189.91990
0.11599
7.84724
197.90438
75.57
30
15
0.03726
182.41378
1.08792
14.21222
197.75119
75.52
31
16
0.02366
189.70684
0.12334
7.84724
197.70108
75.50
32
EPA pulled each source's hourly emission rates for each of the 99th% occurances that made up
the peak model receptor's modeled 1-hr SO2 design value. Based on the peak model receptor's
MAXDCON results, the only contributing source for the model simulation is Seward. Table 5.2-
19 shows each modeled source's hourly emission rate at the time the 99th% modeled value
occurred. Under the right meteorological conditions, it appears Seward can cause exceedances
along the ridge with SO2 emission rates slightly above the 1,800 lbs/hr range. Note that the 1st
and 3rd year 99th% occurrence resulted in model concentrations under 75 ppb.
Table 5.2-19. Hourly Emission Rates for Modeled 99th% or High 4th-High Occurances
EPA AERWIOD 22112 Turbulence Run: Peak Receptor High-4th High & Corresponding S02 Emission Rates
Date
Hour
S02
(ug/m3)
S02
(PPb)
Conemaugh
(lbs/hr)
Seward
(lbs/hr)
Cambria Cogen
(lbs/hr)
Colver
(lbs/hr)
Ebensburg
(lbs/hr)
09-15-2017
8
195.60960
74.70
515.1
1,799.5
450.01
734.93
303.18
11-08-2018
7
266.90361
101.92
1,370.3
2,158.0
667.47
677.00
379.37
06-08-2020
23
144.62180
55.23
114.0
1,223.1
0.00
0.00
253.18
5.3. Air Quality Monitoring Data for the Westmoreland and Cambria Counties,
PA
92
-------
The Johnstown monitor is located in the City of Johnstown roughly 15 km southeast of the
Conemaugh and Seward power plants. Located along Stonycreek River, the Johnstown monitor
sits at a significantly lower elevation, around 435 meters, than the surrounding terrain, which
rises to over 800 meters in places. The location of this monitor is not in the area of maximum
modeled concentration, therefore, the monitoring data from this site isn't sufficient to
characterize air quality concentration for designation purposes.
EPA considered design values for the air quality monitor located in Cambria County by
assessing the most recent 3 consecutive years (i.e., 2019-2021) of quality-assured, certified
ambient air quality data in the EPA Air Quality System (AQS) using data from Federal
Reference Method and Federal Equivalent Method monitors that are sited and operated in
accordance with 40 CFR parts 50 and 58.23 Procedures for using monitored air quality data to
determine whether a violation has occurred are given in 40 CFR part 50 Appendix T, as revised
in conjunction with the 2010 SO2 NAAQS. The 2010 1-hour SO2 NAAQS is met when the
design value is 75 ppb or less. Table 5.3-1 contains the 2019-2021 design values for the area of
analysis.
Table 5.3-1. 2010 SO2 NAAQS Design Values in Cambria County
AQS Site II)
Monitor Location
2019-2021
Design
Value (ppb)
2019 99"'
Percentile
(ppb)
2020 99"'
Percentile
(ppb)
2021 99"'
Percentile
(ppb)
42-011-0021
Miller Auto Shop, 1
Messenger St.,
Johnstown, PA
40.309944, -78.915444
10
13
8
10
Based on available ambient air quality data collected between 2019-2021, Cambria County does
not show a violation of the 2010 SO2 NAAQS at its monitor. However, the absence of a
violating monitor when considering the distance from Conemaugh and Seward Stations coupled
with the terrain of the Laurel Ridge is not a sufficient technical justification to rule out that an
exceedance of the 2010 SO2 NAAQS may occur in the immediate vicinity of the Facilities.
Therefore, EPA is considering air quality modeling to determine whether the areas in
Westmoreland and Cambria Counties are in attainment.
5.4. Intended Designation Boundary Determination
Under CAA section 107(d)(3)(A), the Administrator may at any time inform the governor of a
state that available information indicates that the designation of any area should be revised.
Based on the air quality modeling information summarized above, the EPA is informing the
governor of Pennsylvania that portions of Westmoreland and Cambria counties should be
redesignated to nonattainment. In this section, we consider the appropriate geographical extent of
the nonattainment area.
23 SO; air quality data are available from EPA's website at https://www.epa. gov/oiitdoor-air-onatitv-data. SO; air
quality design values are available at https://www.epa.gov/air-trends/air-analitv-design-values.
93
-------
In notifying the governor of the boundaries of our intended redesignation, the EPA is relying on
the same technical bases used for its initial designations process for the 2010 SO2 NAAQS. For
those initial designations, EPA designated as nonattainment those areas containing the area
violating the NAAQS (e.g., the area around a violating monitor or encompassing modeled
violations), as well as any nearby areas (e.g., counties or portions thereof) that contain emissions
sources contributing to ambient air quality in the violating area. (See CAA section
107(d)(l)(A)(i)). Accordingly, although EPA considers county boundaries as the analytical
starting point for determining SO2 nonattainment areas, an evaluation of five factors for each
area may be considered in determining the geographic scope of a nonattainment boundary.
Thus, boundaries area evaluated on five factors: 1) ambient air quality data or dispersion
modeling results; 2) emissions-related data; 3) meteorology; 4) geography and topography; and
5) jurisdictional boundaries, as well as other relevant available information. While the factors are
presented individually, they are not independent. Instead, the five-factor analysis process
carefully considers their interconnections and the dependence of each factor on one or more of
the others.
5.4.1. Factor 1: Ambient Air Quality Data and Dispersion Modeling Results
As described above in section 5.2, EPA modeled actual emissions for Seward and Conemaugh,
Colver Power and Ebensburg Power, and results are depicted in Figure 5.2-37. The model
receptors with concentrations over the standard are noted in blue. The violating model receptors
are located in Lower Yoder Township in Cambria County, and St. Clair Township in
Westmoreland County. EPA also considered other modeling analyses which are discussed in
detail in section 5.5. which showed peak modeled receptors generally in the same location.
There are no monitors located in St. Clair Township or in Lower Yoder Township. As noted
above, the Johnstown monitor (42-011-0021), which has a 2021 design value of 10 ppb, is
located at significantly lower terrain, and the Laurel Ridge hinders emissions from the two
facilities in Indiana County from significantly impacting that monitor. Thus, lack of a monitored
violation at that location, does not rule out a violation of the standard closer to the facilities
(which EPA modeling indicates).
5.4.2. Factor 2: Emissions-Related Data
EPA believes that it is reasonable to evaluate SO2 emissions data from EPA's Emission
Inventory System (EIS) and CEMS data. Table 5.4-2 shows the most recent four years of
emissions data for the facilities that are being characterized by the modeling analyses described
previously.
94
-------
Table 5.4-2. SO2 Emissions of Sources in t
le Indiana/Westmoreland/ Cambria Area
2018 SO2
2019 SO2
2020 SO2
2021 SO2
County
Kacililv Name
K missions
Kmissions
Kmissions
Kmissions
(Ions)
(Ions)
(Ions)
(Ions)
Indiana
Conemaugh
4831
4299
2758
2888
Indiana
Seward
7963
5782
6314
7569
Cambria
Colver Power
2728
2259
1565
2334
Cambria
Ebensburg
Power
1855
1319
1359
1022
Cambria
Cambria
Cogen
2520
491
0
0
The EPA has not received any additional information on emissions reductions resulting from
controls put into place after the date of the emissions inventory data provided in the table above.
5.4.3. Factor 3: Meteorology
EPA evaluated meteorological data to determine how weather conditions, including wind speed
and direction, affect the plume of sources contributing to the ambient SO2 concentrations. A
detailed description of the meteorology of the area is included in section 5.2.6. EPA conducted
two modeling analyses, which included site specific meteorology data (collected between
Conemaugh and Seward plants).
5.4.4. Factor 4: Geography and Topography
EPA examined the physical features of the land that may affect the distribution of emissions and
may help define nonattainment area boundaries. A detailed description of the land use data used
in the modeling analysis was provided earlier.
Both Conemaugh and Seward are located along the Conemaugh River in Indiana County and are
contained within the Ligonier Valley. The Chestnut Ridge lies to the west of these facilities and
the Laurel Ridge lies to the east. Both terrain features largely pinch out to the north but extend
many miles to the south. Water drainage is to the west, eventually becoming part of the Ohio
River Basin. The Conemaugh River bisects both ridges creating the Conemaugh River Gorge as
it passes through the Laurel Ridge. The Chestnut and Laurel ridges present several physical
impediments to plumes originating from Conemaugh and Seward.
5.4.5. Factor 5: Jurisdictional Boundaries
EPA considers existing jurisdictional boundaries for the purposes of providing a clearly defined
legal boundary for carrying out the air quality planning and enforcement functions for the area.
Our goal is to base designations on clearly defined legal boundaries that align with existing
administrative boundaries when reasonable. Existing jurisdictional boundaries used to define a
95
-------
nonattainment area must encompass the area that has been identified as meeting the
nonattainment definition.
Modeled violations are constrained to St. Clair Township in Westmoreland County and Lower
Yoder Township in Cambria County. The main impacting sources are located in East and West
Wheatfield Townships in Indiana County (a nonattainment area), across the Conemaugh River
from the modeled violations.
5.4.6. Intended Nonattainment Area Boundary
In consideration of the five factors, EPA intends to designate the Township of St. Clair in
Westmoreland County, and the Township of Lower Yoder in Cambria County as nonattainment
for the 2010 SO2 NAAQS (see Figure 6-1). This new nonattainment area would not include the
main contributing sources of Conemaugh and Seward power plants, which reside in neighboring
Indiana County, an already designated nonattainment area for the 2010 SO2NAAQS. EPA
considered options for a nonattainment area boundary that included Conemaugh and Seward
plants, which are discussed below. Ultimately, EPA believes that an attainment plan for the
intended Westmoreland/Cambria nonattainment area would require the same stringency in terms
of emission limits as one for a nonattainment area whose boundary included the townships where
Conemaugh and Seward plants are located. Specifically, EPA recognizes that the state's
obligation under section 110(a) of the CAA in developing an attainment plan for a nonattainment
area is to submit "... a plan which provides for implementation, maintenance, and enforcement
of such primary standard in each air quality control region (or portion thereof) within such
State." CAA section 110(a)(1). Section 110 further provides that "[i]n the case of a plan or plan
revision for an area designated as a nonattainment area, meet the applicable requirements of part
D of this subchapter (relating to nonattainment areas)." CAA section 110(a)(2)(I). Section
172(c)(6) then requires the SIP for a nonattainment area to include enforceable emission
limitations and control measures as necessary or appropriate to provide for NAAQS attainment
"in such area." CAA section 172(c)(6). Therefore, EPA maintains that a nonattainment area
without the contributing sources does not preclude the state from requiring emission limits on
sources contributing to the air quality violations in the nonattainment area.
EPA considered whether the townships where Seward and Conemaugh power plants are located
should be included in the Westmoreland and Cambria nonattainment area. While it is EPA's
general policy to include sources that cause violations within the nonattainment area boundary,
EPA recognizes the uniqueness of this situation in that Seward and Conemaugh power plants are
already included in the Indiana, PA SO2 nonattainment area, and thus those townships are
already subject to CAA Part D nonattainment planning requirements. Additionally, the
attainment plan for the Indiana, PA area was recently partially disapproved and partially
approved, which initiated a sanctions clock under CAA section 179, providing for emission
offset sanctions for new sources if EPA has not fully approved a revised SIP attainment plan
within 18 months after final partial disapproval, and providing for highway funding sanctions if
EPA has not fully approved a revised plan within 6 months thereafter. The sanctions clock can
be stopped only if the conditions of EPA's regulations at 40 CFR 52.31 are met. This action also
initiated an obligation for EPA to promulgate a Federal implementation plan within 24 months
unless Pennsylvania has submitted, and EPA has fully approved, a plan addressing these
96
-------
attainment planning requirements. In order to avoid SIP planning uncertainty for the existing
Indiana, PA nonattainment area, EPA believes the intended boundary proposed for the new
Westmoreland/Cambria nonattainment area, which does not include the culprit sources, is
reasonable, as it will result in Pennsylvania ultimately needing to demonstrate that the emissions
from Seward and Conemaugh are sufficient to provide for NAAQS attainment in both areas
without disrupting the pre-existing requirement that Pennsylvania demonstrate, on the already
established schedule for that duty, that the emissions from those sources and other sources in the
Indiana County area are sufficient to provide for Indiana County's attainment.
5.5. Modeling Analyses Provided by Other Parties
The EPA received additional information relevant to the designation of this area. This section
will outline several different modeling analyses that were submitted during the public comment
period for EPA's partial disapproval and partial approval of the Indiana, PA attainment plan, and
an additional analysis EPA conducted using an alternatively processed meteorological data set.
EPA's designation modeling used site-specific meteorological data processed with the adjusted
u-star option (without the site-specific turbulence measurements in accordance with EPA
guidance). For completeness considerations, EPA included the site-specific turbulence AERMET
processed data in this section.
Table 5.5-1 summarizes the results from the 3 additional modeling analyses covering areas of
Cambria and Westmoreland counties near the Conemaugh and Seward power plants. Additional
details regarding for each of these additional analyses are provided in sections following the
additional modeling summary tables.
97
-------
Table 5.5-1: Summary of AERMOD Modeling Input Parameters for the Area of Analysis
for the Westmoreland and Cambria Area
1 iiput Parameter
Sierra Club Yalue
KEY-CON
EIW-Turhulenee
AERMOD Version
21112
21112
22112
Dispersion Characteristics
Rural
Rural
Rural
Indiana, PA SIP
Sources
Conemaugh,
Seward, Cambria
Modeled Sources
Conemaugh and
Seward (Conemaugh
single stack)
(Conemaugh, Homer
City, Keystone,
Seward)
Cogen, Colver
Power, Ebensburg
Power
6 (Merged Stack for
Conemaugh)
7 (Conemaugh 3
stacks; 2 for
individual unit flue
Modeled Stacks
2
and merged)
Modeled Structures
33
33
Modeled Fencelines
None
None
Total receptors
34,040
10,705
Emissions Type
Actual (CAMD)
Actual
Actual
2015-17, 2016-18,
2019-21
1 July 2019 through
2017-19, 2018-20 and
30 June 2020
Emissions Years
2019-21
1 September 2015
through 31 August
2016
1 September 2015
through 31 August
2016
Meteorology Years
Same as modeled
emissions periods
Met Data transposed
to fit emission
period as per
Modeling TAD
Met Data
transposed to fit
emission period as
per Modeling TAD
Source
Source
Specific/ASOS
Ash Site #1 &
Specific/ASOS
Ash Site #1 &
NWS Station for Surface
Johnstown-Cambria
Johnstown/Cambria
Johnstown/Cambria
Meteorology
County ASOS
County ASOS
County ASOS
NWS Station Upper Air
Meteorology
Pittsburgh, PA
Pittsburgh, PA
Pittsburgh, PA
NWS Station for
Johnstown-Cambria
Ash Site #1 (uniform
Ash Site #1 (user
Calculating Surface
Characteristics
County ASOS
(uniform 30° sectors)
30° sectors)
defined sectors)
Methodology for
Season by Hour of
Season by Hour of
Season by Hour of
Calculating Background
SO2 Concentration
Day, South Fayette
(2016-18)
Day, South Fayette,
PA
Day, Strongstown,
PA
Calculated Background
SO2 Concentration
Variable
Variable
Variable
98
-------
The results presented below in Table 5.5-2 and Figure in the following sections show the
geographic extent of the predicted modeled violations based on the input parameters.
Table 5.5-2. Predicted 99th Percentile Daily Maximum 1-Hour SO2 Concentration
Averaged Over Three Years for the Area of Analysis for the Westmoreland and Cambria
Area
99,h percentile daily
Ueceplor Location
maximum
1-hour SO2
|IT.M /one I7|
Concent rat ion (ug/nr*)
Modeled
concentration
Averaging
Data
U'M.
I IM.
(including
NAAQS
Modeler
Period
Period
Lasting
northing
background)
1 .cvcl
Sierra
99th Percentile
Club
1-Hour Average
2019-20
669597.38
4471747.00
244.64084
196.4*
KEY-
99th Percentile
CON
1-Hour Average
2019-21
670337.39
4471875.04
193.23181
196.4
EPA-
1 July
turbulence
99th Percentile
2019 to
30 June
1-Hour Average
2020
670244
4471888
220.18960
196.4
* Equivalent to the 2010 SO: NAAQS of 75 ppb using a 2.619 (.ig/m3 conversion factor
99
-------
5.5.7. KEY-CON
A protocol for conducting modeling to evaluate the impacts outside of the Indiana, PA NAA was
submitted by KEY-CON and Seward Stations to the Pennsylvania Department of Environmental
Protection (PA DEP) in September 2020. EPA and PA DEP provided comments to this protocol
in a 17 February 2021 email sent to John Shimshock (KEY-CON) from PA DEP's Andrew
Fleck. Actions to pursue this modeling, however, were suspended while the review of the
modeling for the area inside the nonattainment area was ongoing. KEY-CON performed the
modeling study based on the September 2020 modeling protocol with minor updates. KEY-CON
submitted its results during the public comment period for the proposed partial disapproval and
partial approval of the Indiana, PA attainment plan (87 FR 15166). A brief overview of the
modeling analysis and the results will be discussed in the following sections.
5.5.1.1. Modeled Emissions/Stack Parameters
KEY-CON modeled emissions over a slightly different 3-year period than it included in its
modeling protocol. The time period from the September 2020 modeling protocol was 1 July
2017 through 30 June 2020. The (updated) modeling period for the KEY-CON submitted
modeling was 1 Jan 2019 through 31 Dec 2021.
All sources that were included in the Indiana, PA SIP modeling were also included in the KEY-
CON modeling analysis. These included Conemaugh (one stack), Homer City (3 stacks),
Keystone (1 stack) and Seward (1 stack); Keystone and Homer City were not sources included in
the September 2020 modeling protocol. Stack locations, base elevations and stack diameters for
each of the 5 KEY-CON modeled stacks in the AERMOD input files were compared to
Pennsylvania's original SIP and Supplementary Analysis documentation. EPA verified that stack
base elevations, stack heights and stack diameters used by KEY-CON matched values in the
original Indiana, PA SIP modeling input files. We also note that modeled stack diameters for
Conemaugh and Keystone represent merged diameters representing an equivalent area of the
source's individual flues within their FGD stack structures. As noted previously, merged stack
diameters, while appropriate when both coal-fired units are operating, may not be appropriate for
times when only one unit is operating over the 3-year model simulation period.
EPA checked KEY-CON's modeled hourly SO2 emission rates versus hourly emission rates
from EPA's CAMD database. Hourly CAMD emissions, MODC flow and concentration
information was downloaded using EPA's FACT software. These were then processed using R
for direct comparison with the KEY-CON hourly input file. EPA had no real ability to check
KEY-CON's model inputs for each stack's stack temperature and stack velocity. These were
based on each facility's CEMS measured data. While stack flow rates are part of CAMD
reported emissions, they are generally not usable since the flow information is adjusted to
standard cubic feet per hour, not actual cubic feet per hour. Flow MODC could, however, be
used to identify hours with potentially invalid stack velocities.
100
-------
The following sections examine the hourly CAMD emissions for each source included in the
KEY-CON modeling analysis over the 3-year simulation period. Comparisons of the modeled
hourly emission rates and CAjMD emissions were also conducted.
Coneniaugh Modeled Hourly Emission Rates: Figure 5.5-1 shows Conemaugh's combined
CAMD hourly emissions over the KEY-CON model simulation period. EPA combined hourly
emissions for both of Conemaugh's coal-fired units. As noted previously, CAMD hourly data
include information on the validity of the flow rate and concentration instruments used to
measure hourly SO2 emissions. If flow and concentration codes are valid, a "measured" value is
generated and if either or both flow and concentration instruments malfunction a "calculated"
hourly SO2 emission value is generated. The modeled critical emission value, or CEV is also
included on the figure. Pennsylvania's Supplemental Analysis modeling indicated that
Conemaugh's CEV was 3,381 lbs/hr. This represents the level where modeled emissions just
meet the 1-hr SO2NAAQS. As the figure shows, Conemaugh's hourly SO2 emissions rarely
exceed its modeled CEV.
Figure 5.5-1. Coneniaugh Combined Unit CAMD SO2 Emissions for 2019 through 2021
Conemaugh Hourly Emissions
CAMD Part 75 1 Jan 2019 to 31 Dec 2021
20,000-
Measured
• Calculated
— CEV
_ 15,000-
Date
Figure 5.5-2 shows a comparison of hourly emissions over the 3-year simulation period. A 1 -to-1
(red) tend line is also included on the figure. If the CAMD and KEY-CON's modeled hourly SO2
emission rates are identical, then they will graph as a point along the 1-to-l (red) trend line.
Hourly CAMD emissions and KEY-CON's modeled emission rates are overall well matched.
There only 18 hours across the simulation period where CAMD hourly SO2 emissions are more
than 50 lbs/hr higher than KEY-CON's modeled emission rates.
101
-------
Figure 5.5-2. Conemaugh's CAMD versus KEY-CON Hourly SO2 Emissions
Conemaugh Merged Stack Emission Rate Comparison
CAMD vs. KEY-CON
1 Jan 2019 through 31 Dec 2021
4,000-
1,500-
1,000-
0
0
0
& 0
c**>
1 1 1 1 1 1 1
500 1,000 1,500 2,000 2,500 3,000 3,500 4,000
CAMD S02 Emission Rate (Ibs/hr)
Homer City Unit 1: Homer City is comprised of 3 coal-fired units. Each unit has a dedicated
stack meaning emissions from each coal unit are vented to a single stack structure. CAMD
emissions over the 3-year simulation period were therefore plotted for each unit separately. The
KEY-CON model simulation contained separate stacks for each Homer City coal-fired unit.
Figure 5.5-3 shows CAMD hourly emissions over the 3-year KEY-CON simulation period. The
figure also includes the modeled CEV for unit 1, which is 1,550 lbs/hr. As described previously,
hourly SO2 emissions are differentiated between measured and calculated values on the figure.
Over the 3-year simulation period, CAMD information indicates Homer City unit 1 was off
13,336 hours out of 26,304 total hours. Unit 1 emissions appear to exceed the model CEV
periodically over the 3-year simulation period.
Figure 5.5-4 shows a comparison of CAMD and modeled hourly SO2 emissions over the 3-year
simulation period. A 1-to-l (red) tend line is also included on the figure. If the CAMD and KEY-
CON's modeled hourly SO2 emission rates are identical, then they will graph as a point along the
1-to-l (red) trend line.
102
-------
Figure 5.5-3. Homer City Unit 1 CAMD SO2 Emissions for 2019 through 2021
Homer City Unit 1 Hourly Emissions
CAMD Part 75 1 Jan 2019 to 31 Dec 2021
Measured
—¦ Calculated
— CEV
1 L
Mil lillL
Jj.i
1 ilnt ill
I Li 1
it wiii
ilM K
1 «M
A k IwM
Ik 11
$ $ $ $ $ $
*V 'V *V *V 'V 'V
#' £ # d* / # # ^ o* #'
Date
Figure 5.5-4. Homer City Unit 1 CAMD versus KEY-CON Hourly SO2 Emissions
(/)
_Q
Homer City Unit 1 Emission Rate Comparison
CAMD vs. KEY-CON
1 Jan 2019 through 31 Dec 2021
10,000-
CD
-+—>
05
a:
c:
o
'-
LU
7,500-
5,000-
2,500-
2,500 5,000 7,500
CAMD S02 Emission Rate (Ibs/hr)
10,000
103
-------
Hourly CAMD emissions and KEY-CON's modeled emission rates are overall well matched.
There only 31 hours across the simulation period where CAMD hourly SO2 emissions are more
than 50 lbs/hr higher than KEY-CON's modeled emission rates.
Homer City Unit 2: Figure 5.5-5 shows CAMD hourly emissions over the 3-year KEY-CON
simulation period. The figure also includes the modeled CEV for unit 2, which is 1,550 lbs/hr.
Hourly SO2 emissions are differentiated between measured and calculated values on the figure
Over the 3-year simulation period, CAMD information indicates Homer City unit 2 was off
14,487 hours out of 26,304 total hours. Unit 2 emissions appear to exceed the model CEV
periodically over the 3-year simulation period.
Figure 5.5-5. Homer City Unit 2 CAMD SO2 Emissions for 2019 through 2021
Homer City Unit 2 Hourly Emissions
CAMD Part 75 1 Jan 2019 to 31 Dec 2021
Measured
• Calculated
— CEV
Hk
11.iii Li
pri
1 kilib
1
uL-l
pjp 1
. ik
(Hf
Q) o> O) o> c> e> o> c> *> k *•»
Q> Q> c>
-------
Figure 5.5-6. Homer City Unit 2 CAMD versus KEY-CON Hourly SO2 Emissions
Homer City Unit 2 Emission Rate Comparison
CAMD vs. KEY-CON
1 Jan 2019 through 31 Dec 2021
CAMD S02 Emission Rate (Ibs/hr)
Homer City Unit 3: Figure 5.5-7 shows CAMD hourly emissions over the 3-year KEY-CON
simulation period. The figure also includes the modeled CEV for unit 3, which is 3,260 lbs/hr.
Hourly SO2 emissions are divided between measured and calculated values on the figure. Over
the 3-year simulation period, CAMD information indicates Homer City unit 3 was off 11,464
hours out of 26,304 total hours. Unit 3 emissions appear to occasionally exceed the model CEV
over the 3-year simulation period.
Figure 5.5-8 shows a comparison of CAMD and modeled hourly SO2 emissions over the 3-year
simulation period. A 1-to-l (red) tend line is also included on the figure. If the CAMD and KEY-
CON's modeled hourly SO2 emission rates are identical, then they will graph as a point along the
1-to-l (red) trend line.
105
-------
Figure 5.5-7. Homer City Unit 3 CAMD SO2 Emissions for 2019 through 2021
Homer City Unit 3 Hourly Emissions
CAMD Part 75 1 Jan 2019 to 31 Dec 2021
Measured
— Calculated
— CEV
lli 1
Jk 1
hi
II Www 4
1
1
ii ifl
$ $ $ $ $ $
*V 'V *V *V 'V 'V
#' £ # d* / # # ^ o* #'
Date
Figure 5.5-8. Homer City Unit 3 CAMD versus KEY-CON Hourly SO2 Emissions
Homer City Unit 3 Emission Rate Comparison
CAMD vs. KEY-CON
1 Jan 2019 through 31 Dec 2021
10,000-
(/)
_Q
CD
-+—>
05
a:
c:
o
'-
LU
7,500-
5,000-
2,500-
2,500 5,000 7,500
CAMD S02 Emission Rate (Ibs/hr)
10,000
106
-------
Hourly CAMD emissions and KEY-CON's modeled emission rates are overall well matched.
There only 19 hours across the simulation period where CAMD hourly SO2 emissions are more
than 50 lbs/hr higher than KEY-CON's modeled emission rates.
Keystone Modeled Hourly Emission Rates: Figure 5.5-9 shows Keystone's combined CAMD
hourly emissions over the KEY-CON model simulation period. Keystone and Conemaugh are
sister plants sharing a similar configuration. EPA combined hourly emissions for both of
Keystone's coal-fired units. Emission lines are labeled "measured" or "calculated" following the
same conventions described previously. The modeled critical emission value, or CEV is also
included on the figure. Pennsylvania's SIP modeling indicated Keystone's CEV was 9,711
lbs/hr. This represents the level where modeled emissions just meet the 1-hr SO2 NAAQS. As
the figure shows, Keystone's hourly SO2 emissions do exceed its modeled CEV. Overall,
Keystone's SO2 emissions are consistently higher than the other KEY-CON modeled sources.
Figure 5.5-9. Keystone's Combined Unit CAMD SO2 Emissions for 2019 through 2021
Keystone Hourly Emissions
CAMD Part 75 1 Jan 2019 to 31 Dec 2021
20,000-
Measured
— Calculated
— CEV
15.000-
Date
Figure 5.5-10 shows a comparison of hourly emissions over the 3-year simulation period. A 1-to-
1 (red) tend line is also included on the figure. If the CAMD and KEY-CON's modeled hourly
SO2 emission rates are identical, then they will graph as a point along the 1-to-l (red) trend line.
107
-------
Hourly CAMD emissions and KEY-CON's modeled emission rates show a lot more spread
around the 1-to-l trend line than the other modeled sources. There is a total of 1,655 hours across
the simulation period where differences between the CAMD reported hourly and KEY-CON
modeled hourly SO2 emissions are more than ±50 lbs/hr. The bulk of these differences occurred
over the 2020 portion of the 3-year modeling period. The impact of these emission differences
may be minor considering Keystone is over 40 km northwest of the Laurel Ridge.
Figure 5.5-10. Keystone's Combined CAMD versus KEY-CON Hourly SO2 Emissions
Keystone Combined Emission Rate Comparison
CAMD vs. KEY-CON
1 Jan 2019 through 31 Dec 2021
20,000
£ 17,500
CD
-•—»
03
01
c
o
'(/)
(/)
'£
LU
CM
o
w
o
o
I
>-
LU
15,000
12,500
10,000
7,500
2,500
5,000 7,500 10,000 12,500 15,000
CAMD S02 Emission Rate (lbs/hr)
17,500
20,000
108
-------
Seward Modeled Hourly Emission Rates: Figure 5.5-11 shows Seward's CAMD hourly
emissions over the KEY-CON model simulation period. Emission lines are labeled "measured"
or "calculated" following the same conventions described previously. The modeled critical
emission value, or CEV is also included on the figure. Pennsylvania's Supplemental Analysis
modeling established Seward's CEV at 4,500 lbs/hr. This represents the level where modeled
emissions just meet the 1-hr SO2 NAAQS. As the figure shows, Seward's hourly SO2 emissions
occasionally exceed its modeled CEV.
Figure 5.5-11. Seward's CAMD SO2 Emissions for 2019 through 2021
Seward Hourly Emissions
CAMD Part 75 1 Jan 2019 to 31 Dec 2021
20,000-
Measured
Calculated
— CEV
15,000-
Date
Figure 5.5-12 shows a comparison of hourly emissions over the 3-year simulation period. A 1-to-
1 (red) tend line is also included on the figure. If the CAMD and KEY-CON's modeled hourly
SO2 emission rates are identical, then they will graph as a point along the 1-to-l (red) trend line.
CAMD emissions and KEY-CON's modeled emission rates largely fall along the 1-to-l trend
line indicating good agreement. There only 9 hours across the simulation period where CAMD
hourly SO2 emissions were more than 50 lbs/hr higher than the KEY-CON modeled emission
rates.
109
-------
Figure 5.5-12. Seward's CAMD versus KEY-CON Hourly SO2 Emissions
-
LU
10,000-
9,000-
8,000-
7,000-
6,000-
5,000-
4,000-
3,000-
2,000-
1,000 -
0-
0
Seward Stack Emission Rate Comparison
CAMD vs. KEY-CON
1 Jan 2019 through 31 Dec 2021
1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000 10,000
CAMD S02 Emission Rate (Ibs/hr)
110
-------
5.5.1.2. Meteorology and Surface Characteristics
KEY-CON used the site-specific meteorological data that EPA utilized for its modeling analysis.
This data was collected and used for Pennsylvania's Supplemental Analysis for the southeast
portion of the Indiana, PA nonattainment area near Conemaugh and Seward. It consists of 1 year
(September 2015 - August 2016) of hourly surface observations from the on-site meteorological
tower and SODAR along with 1 year of concurrent cloud cover data from the Johnstown-
Cambria County airport (JST) and upper air data from Pittsburgh International Airport.
AERSURFACE (version 20060) and AERMET (version 21112) were used to produce the final
processed meteorological data
In general, the processing steps performed by KEY-CON largely resemble the processing steps
completed for EPA's analysis as described in section 5.2.6. EPA notes the following differences
between what EPA used in its analysis and KEY-CON's final processed meteorological data:
• an additional month of continuous snow cover (Feb 2016) was processed in
AERSURFACE
• AERSURFACE was run using 12 equal (30°) sectors versus EPA's use of 8 surface
varying sectors
• AERSURFACE utilized the USGS 1992 land use-land cover (LULC)
In regard to using the 1992 LULC data in AERSURFCE, KEY-CON's September 2020 protocol
offers the following support for its decision:
[0]ver the periodfrom 1992 through 2016, the land cover has remained relatively
unchanged, .... However, after visual inspection of the NLCD 1992 and 2016 datasets within
1 km of the on-site tower, there are significant differences noted. The most obvious difference
is in the 2016 data the land cover being classified within approximately 150 m of the on-site
tower is developed/high intensity. This apparent misclassification is typically reservedfor
highly industrialized areas and is not representative of the actual land cover. Accordingly,
we are proposing to use the 1992 land cover data in this modeling for the AERSURFACE
application, which the most representative and is consistent with previous modeling
demonstrations for this area.
Figure 5.5-13 shows both the 1992 and 2016 LULC classifications within 1 km of the site-
specific met tower collection site (Ash Site #1). EPA generally recommends using LULC data
sets that represent conditions at the time of the meteorological collection period. In this case, the
2016 LULC data would more closely match land use at the time of the site-specific
meteorological collection period (1 Sep 2015 through 31 Aug 2016).
While we note KEY-CON's odd LULC categories near the Ash Site #1 met tower, we'd also
counter that the ash landfill should not be covered with deciduous forest as shown in the 1992
LULC data (assuming woody vegetation is prohibited on a capped disposal area). Additionally,
the Seward power plant (in sector II) was rebuilt in the early 2000s and therefore not correctly
captured using the 1992 LULC data. We would also point out that the Conemaugh River area
and associated wetlands are poorly defined in the 1992 LULC dataset.
Ill
-------
Figure 5.5-13. Land Use Land Cover Within 1-kin of the Ash Site #1 Meteorological Tower
Westmoreland-Cambria, PA: KEY-CON AERSURFACE
Legend
Ash Site #1 Met Tower
-AERSURFACE Sectors
2 Kilometers
-------
EPA reprocessed the AERSURFACE processing using the 2016 LULC (and impervious surface
and tree canopy information) following the sector definitions chosen by KEY-CON. Surface
moisture conditions and snow cover were entered into AERSURFACE's CLIMATE keyword
following KEY-CON's September 2020 modeling protocol. Table 5.5-3 summarizes the
AERSURFACE CLIMATE settings over the site-specific simulation period.
Table 5.5-3. KEY-CON AERSURFACE CLIMATE Keyword Settings
KEY-CON AERSURFACE CLIMATE Keyword Settings
Date
Surface Moisture
Snow Cover
Sep 2015
AVG
Oct 2015
AVG
Nov 2015
DRY
Dec 2015
WET
Jan 2016
DRY
SNOW
Feb 2016
DRY
SNOW
Mar 2016
DRY
Apr 2016
DRY
May 2016
AVG
Jun 2016
AVG
Jul 2016
DRY
Aug 2016
WET
AERSURFACE (version 20060) was rerun for each of the CLIMATE keyword categories
covering the site-specific collection period. Results were then processed in R so that comparisons
between the 1992 LULC and 2016 LULC data sets could be made for AERSURFACE
determined albedo, Bowen ratio and surface roughness lengths.
Table 5.5-4 shows the AERSURFACE generated albedo values using the 1992 LULC and 2016
LULC data sets. Albedo values are determined by a simple geometric mean of the values of the
individual grid cells that make up the 10 km x 10 km area centered on the measurement site. The
same value is used for all sectors so only the monthly values are displayed on the table.
There are only small variations in the Albedo values between the 1992 and 2016 LULC data sets.
These are not expected to have much impact on AERMOD concentrations once they are
processed in AERMET.
113
-------
Table 5.5-4. AERSURFACE Generated Albedo Values for the 1992 and 2016 LULC Data
Sets
KEY-CON AERSURFACE Albedo Comparison
Month
Albedo 1992 LULC
Albedo 2016 LULC
1
0.50
0.49
2
0.50
0.49
3
0.17
0.17
4
0.16
0.15
5
0.16
0.15
6
0.16
0.16
7
0.16
0.16
8
0.16
0.16
9
0.16
0.16
10
0.16
0.16
11
0.17
0.17
12
0.17
0.17
Table 5.5-5 shows the AERSURFACE generated Bowen ratio values using the 1992 LULC and
2016 LULC data sets. Bowen ratio values are determined using the same methodology as the
albedo values described previously; a simple geometric mean of the values of the individual grid
cells that make up the 10 km x 10 km area centered on the measurement site. As with albedo, the
same value is used for all sectors so only the monthly values are displayed on the table.
Table 5.5-5. AERSURFACE Generated Bowen Ratio Values for the 1992 and 2016 LULC
Data Sets
KEY-CON AERSURFACE Bowen Ratio Comparison
Month
Bowen Ratio 1992 LULC
Bowen Ratio 2016 LULC
1
0.50
0.49
2
0.50
0.49
3
1.98
1.87
4
1.47
1.37
5
0.66
0.61
6
0.33
0.34
7
0.68
0.72
8
0.22
0.22
9
0.96
0.90
10
0.96
0.90
11
1.98
1.87
12
0.41
0.39
114
-------
Bowen ratio values also do not appear to vary significantly between the 1992 and 2016 LULC
data sets. Differences of this magnitude should not impact final AERMOD concentrations once
they are processed in AERMET.
Surface roughness lengths are based on inverse distance-weighted geometric means. The mean is
calculated from the roughness values associated with the land cover category that defines each
land cover grid cell within the area or individual sectors out to a fixed radial distance from the
meteorological tower. KEY-CON used the 1 km recommended and default radial in its
AERSURFACE processing.
AESURFACE sector and monthly varying surface roughness lengths (zO) for both the 1992 and
2016 LULC data sets are summarized in Table 5.5-6a-c. Unlike the albedo and Bowen ratio
values, there appears to be significant differences between values extracted from the 1992 LULC
data set and values from the 2016 LULC data sets. In general, the 1992 LULC data set yields
higher surface roughness lengths than the 2016 LULC data set (complimented with impervious
surface and tree canopy data). Sectors 10 and 11 are highlighted since these cover the wind
directions that are most likely to impact AERMOD concentrations along the Laurel Ridge.
Table 5.5-6a. AERSURFACE Generated Surface Roughness Lengths (zO) for the 1992 and
2016 LULC Data Sets
Months 1-4 (Jan - Apr)
KEY-CON AERSURFACE Surface Roughness Lengths
Month Sector
zO
1992
LULC
zO
2016
LULC
Month
Sector
zO
1992
LULC
zO
2016
LULC
Month
Sector
zO
1992
LULC
zO
2016
LULC
Month
Sector
zO
1992
LULC
zO
2016
LULC
1 1
0.476
0.250
2
1
0.476
0.250
3
1
0.573
0.317
4
1
0.949
0.489
1 2
0.422
0.039
2
2
0.422
0.039
3
2
0.493
0.060
4
2
0.724
0.073
1 3
0.456
0.156
2
3
0.456
0.156
3
3
0.546
0.197
4
3
0.844
0.283
1 4
0.501
0.221
2
4
0.501
0.221
3
4
0.597
0.277
4
4
0.945
0.415
1 5
0.474
0.267
2
5
0.474
0.267
3
5
0.572
0.329
4
5
0.928
0.488
1 6
0.507
0.197
2
6
0.507
0.197
3
6
0.606
0.247
4
6
0.975
0.367
1 7
0.455
0.177
2
7
0.455
0.177
3
7
0.541
0.219
4
7
0.847
0.327
1 8
0.494
0.114
2
8
0.494
0.114
3
8
0.579
0.133
4
8
0.891
0.189
1 9
0.467
0.078
2
9
0.467
0.078
3
9
0.561
0.115
4
9
0.921
0.183
1 10
0.453
0.140
2
10
0.453
0.140
3
10
0.538
0.194
4
10
0.849
0.320
1 11
0.456
0.212
2
11
0.456
0.212
3
11
0.543
0.275
4
11
0.865
0.401
1 12
0.462
0.183
2
12
0.462
0.183
3
12
0.557
0.241
4
12
0.920
0.385
115
-------
Table 5.5-6b. AERSURFACE Generated Surface Roughness Lengths (zO) for the 1992 and
2016 LULC Data Sets
Months 5-8 (May - Aug)
KEY-CON AERSURFACE Surface Roughness Lengths
Month
Sector
zO
1992
LULC
zO
2016
LULC
Month
Sector
zO
1992
LULC
zO
2016
LULC
Month
Sector
zO
1992
LULC
zO
2016
LULC
Month Sector
zO
1992
LULC
zO
2016
LULC
5
1
0.949
0.489
6
1
1.248
0.643
7
1
1.248
0.643
8 1
1.248
0.643
5
2
0,724
0.073
6
2
0.932
0.084
7
2
0.932
0.084
8 2
0.932
0.084
5
3
0.844
0.283
6
3
1.119
0.346
7
3
1.119
0.346
8 3
1.119
0.346
5
4
0.945
0.415
6
4
1.232
0.531
7
4
1.232
0.531
8 4
1.232
0.531
5
5
0.928
0.488
6
5
1.233
0.630
7
5
1.233
0.630
8 5
1.233
0.630
5
6
0.975
0.367
6
6
1.268
0.474
7
6
1.268
0.474
8 6
1.268
0.474
5
7
0.847
0.327
6
7
1.100
0.411
7
7
1.100
0.411
8 7
1.100
0.411
5
8
0.891
0.189
6
8
1.122
0.227
7
8
1.122
0.227
8 8
1.122
0.227
5
9
0.921
0.183
6
9
1.210
0.369
7
9
1.210
0.369
8 9
1.210
0.369
5
10
0.849
0.320
6
10
1.100
0.547
7
10
1.100
0.547
8 10
1.100
0.547
5
11
0.865
0.401
6
11
1.119
0.541
7
11
1.119
0.541
8 11
1.119
0.541
5
12
0.920
0.385
6
12
1.215
0.549
7
12
1.215
0.549
8 12
1.215
0.549
Table 5.5-6c. AERSURFACE Generated Surface Roughness Lengths (zO) for the 1992 and
2016 LULC Data Sets
Months 9-12 (Sep - Dec)
zO zO zO zO zO zO zO zO
Month Sector 1992 2016 Month Sector 1992 2016 Month Sector 1992 2016 Month Sector 1992 2016
LULC LULC LULC LULC LULC LULC LULC LULC
9
1
1.248
0.622
10
1
1.248
0.622 1
1
0.573
0.317
12
1
0.573
0.317
9
2
0.932
0.081
10
2
0.932
0.081 1
2
0.493
0.060
12
2
0.493
0.060
9
3
1.119
0.336
10
3
1.119
0.336 1
3
0.546
0.197
12
3
0.546
0.197
9
4
1.232
0.518
10
4
1.232
0.518 1
4
0.597
0.277
12
4
0.597
0.277
9
5
1.233
0.617
10
5
1.233
0.617 1
5
0.572
0.329
12
5
0.572
0.329
9
6
1.268
0.461
10
6
1.268
0.461 1
6
0.606
0.247
12
6
0.606
0.247
9
7
1.100
0.403
10
7
1.100
0.403 1
7
0.541
0.219
12
7
0.541
0.219
9
8
1.122
0.227
10
8
1.122
0.227 1
8
0.579
0.133
12
8
0.579
0.133
9
9
1.210
0.362
10
9
1.210
0.362 1
9
0.561
0.115
12
9
0.561
0.115
9
10
1.100
0.538
10
10
1.100
0.538 1
10
0.538
0.194
12
10
0.538
0.194
9
11
1.119
0.520
10
11
1.119
0.520 1
11
0.543
0.275
12
11
0.543
0.275
9
12
1.215
0.533
10
12
1.215
0.533 1
12
0.557
0.241
12
12
0.557
0.241
5.5.1.3. Aera of Analysis/Receptor Grid
KEY-CON's model receptor grid was detailed in it September 2020 modeling protocol provided
to Pennsylvania and EPA Region 3. It describes the model receptor grid construction as follows:
116
-------
[T]he receptor grid is centered at approximately the center point between Seward and
Conemaugh Stations and extends outward approximately 10 km to areas outside Indiana
County. Receptors throughout the modeling domain are spaced no more than 100 m apart.
Receptors in areas expected to be associated with peak modeled impacts have been spaced at
25-m intervals, as shown in the figure.
Elevations and receptor height scales (used in AERMOD) are developed by AERMAP, the
terrain preprocessor for AERMOD, which requires processing of terrain data files. The
height scale is the terrain elevation in the vicinity of a receptor that is used in the critical
dividing streamline height calculation for interaction of the plume with terrain.
The current version of AERMAP has the ability to process USGS National Elevation Dataset
(NED) data in place of Digital Elevation Model files. The appropriate file for 1/3-arc-
second, or 10-m, NED data was obtainedfrom the Multi-Resolution Land Characteristics
Consortium (MRLC) link at http://www.mrlc.gov/viewerjs/.
Figure 5.5-14 shows an overview of the KEY-CON modeling grid along with the primary SO2
sources included in the modeling analysis. Overall, the grid has 34,040 receptors. A close up
view around the Conemaugh and Seward power plants is shown in Figure 5.5-15. This shows the
model receptor grid density along the Laurel Ridge to the southeast of the power plants. As noted
previously, Model receptor spacing is 100 m throughout the modeling domain with a 25-m
spaced Cartesian grid placed on portions of the Laurel Ridge (with the highest model
concentrations). This finer grid should ensure peak model concentrations are captured in KEY-
CON's modeling analysis.
117
-------
Figure 5.5-14. KEY-CON Model Receptor Grid and Primary Modeled SO2 Sources
KEY-CON Model Receptor Grid Overview
118
-------
Figure 5.5-15. KEY-CON Model Receptor Grid and Primary Modeled SO2 Sources
KEY-CON Model Receptor Grid Near Conemaugh & Seward
5.5.1.4. Background Concentration
KEY-CON's modeling analysis included a season by hour of day varying background
concentration. This follows the background concentration construction method outlined in EPA's
March 1, 2011 1-hour NO2 clarification memo for use of a temporally varying background
concentrations. Season by hour of day 1-hr SO2 background concentrations were taken from the
South Fayette monitor located in Allegheny County. KEY-CON's AERMOD input file identifies
the monitor and period (2019-21) used to develop the model background concentration.
EPA downloaded hourly SO2 concentrations for the South Fayette monitor using R's
RAQSAPFs library for the 2019-21 time period. We used R to configure the season by hour of
day 1-hr S02 background concentrations in accordance with our March L 2011 guidance. Table
5.5-7 summarizes EPA's constructed season by hour of day background concentrations from
South Fayette's 2019-21 monitor values. The values shown in EPA's table generally match the
values KEY-CON used in their AERMOD input file. We note that KEY-CON's season by hour
of day background concentrations (in parts per billion or ppb) were entered to 2 decimal places.
Our values, also in ppb, preserve 1 decimal place. EPA's table also includes information on the
number of missing and total hours avail able.
119
-------
Table 5.5-7. EPA Constructed 2019-21 Season by Hour of Day Background
Concentrations for the South Fayette, PA Monitor Located in Allegheny County.
Winter
Spring
Summer
Fall
South Fayette, PA (42-003-0067): 2019-21 Background S02 Concentrations (ppb) by Season/Hour of Day
Hour
Average
Missing
Total
Hours
Average
Missing
Total
Hours
Average
Missing
Total
Hours
Average
Missing
Total
Hours
1
2.3
0
271
2.3
0
276
2.0
0
276
2.7
0
273
2
3.3
0
271
2.3
0
276
2.0
0
276
2.7
0
273
3
3.0
0
271
2.7
0
276
1.7
0
276
2.0
0
273
4
3.7
0
271
3.7
0
276
2.0
0
276
2.0
0
273
5
3.0
0
271
2.0
0
276
1.7
0
276
2.3
0
273
6
3.7
0
271
2.7
0
276
2.0
0
276
2.0
0
273
7
2.7
0
271
3.0
0
276
2.3
0
276
2.0
0
273
8
3.0
0
271
3.3
6
276
3.3
8
276
2.7
6
273
9
3.7
2
271
3.3
3
276
3.3
1
276
2.3
0
273
10
4.0
1
271
3.0
12
276
3.0
23
276
3.0
10
273
11
5.7
9
271
3.0
25
276
3.0
21
276
3.0
24
273
12
4.7
27
271
3.0
14
276
3.0
6
276
3.0
14
273
13
4.7
13
271
2.3
7
276
3.0
5
276
4.3
8
273
14
4.7
6
271
2.0
1
276
3.3
4
276
4.0
2
273
15
4.0
1
271
2.3
1
276
3.3
4
276
2.7
0
273
16
4.7
0
271
2.0
0
276
4.0
2
276
2.7
0
273
17
3.3
1
271
2.3
0
276
5.0
1
276
3.7
0
273
18
3.7
0
271
2.3
0
276
5.0
0
276
4.0
0
273
19
3.7
0
271
3.3
0
276
4.3
0
276
2.7
0
273
20
4.0
0
271
3.3
0
276
4.0
0
276
2.3
0
273
21
2.7
0
271
3.0
0
276
3.7
0
276
2.0
0
273
22
4.3
0
271
2.0
0
276
2.7
0
276
2.7
0
273
23
3.3
0
271
2.7
0
276
3.0
0
276
2.7
0
273
24
3.3
0
271
2.3
0
276
2.0
0
276
2.7
0
273
5.5.1.5. KEY-CON Model Results
KEY-CON's modeling generally followed EPA's Modeling TAD. EPA reviewed the modeling
with actual 2019-21 SO2 emissions from all 4 Indiana, PA nonattainment area sources. KEY-
CON's modeling analysis was confined to portions of Westmoreland and Cambria counties
adjacent to the Indiana, PA nonattainment area.
Figure 5.5-16 shows KEY-CON's 2019-21 modeled 1-hr SO2 design values over its entire
modeling domain. KEY-CON's AERMOD concentrations were converted to parts per billion or
ppb by multiplying the model concentrations by a conversion factor; 75 ppb over 196.4 |ig/m3.
Model concentrations are overlain over the local topographic elevations.
Model 1-hr SO2 design values are highest along the Laurel Ridge southeast of Conemaugh and
Seward as noted in Figure 5.5-16. The peak receptor had a modeled design value of 73.8 ppb.
KEY-CON's modeled design values are just below the 1-hr SO2 NAAQS (75 ppb) along the
120
-------
Laurel Ridge facing Conemaugh and Seward. We note that KEY-CON's 25-m or fine Cartesian
grid doesn't extend along the entire ridge. It may be possible that there are other model peaks
occuring along the Laurel Ridge that may exceed KEY-CON's peak model values.
Figure 5.5-16. KEY-CON AERMOD Results for All Sources Plus Background
KEY-CON Model Results Overview
As noted previously, EPA believes KEY-CON used an outdated LULC (1992) data set in its
meteorological data processing. Section 5.5.1.2 outlines the differences between this data set and
the more up to date 2016 LULC data set that, in EPA's opinion, is more reflective of conditions
that were present during the site-specific meteorological data collection period.
EPA reran KEY-CON's analysis using the 2016 LULC data in the meteorological preprocessing
to see what difference using the more up to date land used information had on peak model design
values. The final peak model design value (not shown) along the Laurel Ridge was 76.4 ppb or
just slightly above the 1-hr SO2 NAAQS. It appears the 2016 LULC (smaller) surface roughness
lengths, shown in Table 5.5-6a-c, increased final model concentrations by about 3.6%. This
value is roughly in line with EPA's modeling analysis using the site-specific turbulence data (to
be discussed in section 5.5.3). EPA's analysis differs slightly from KEY-CON using slightly
different meteorological processing steps, a different model receptor grid, a different 3-year
emission period and different background concentrations. Despite these differences EPA's final
121
-------
model design value (using the site specific turbulence data), 77.3 ppb, was very close to the
value derived using KEY-CON's analysis with the 2016 LULC data.
Figure 5.5-17. KEY-CON AERMOD Results Along the Laurel Ridge
KEY-CON Model Results Along the Laurel Ridge
5.5.2. Sierra Club
Sierra Club conducted air modeling impact analysis to determine if large emission sources are
causing exceedances of the 1-hour SO2 NAAQS. This section provides a brief summary of the
modeling analysis, results and procedures for evaluating emissions from the Seward Generating
Station (Seward) in Seward, Pennsylvania and Conemaugh Generating Station (Conemaugh) in
New Florence, Pennsylvania. Both plants are located in Indiana County. This analysis
determined if the plants contribute to exceedances of the NAAQS in and around the Indiana
County nonattainment area.
EPA summarized Sierra Club's modeling analysis based on the report summary provided in the
comment period (Exhibit 4) from Wingra Engineering, dated 13 April 2022. EPA's summary
assessment is also based on a review of the modeling files submitted during the public comment
period.
122
-------
5.5.2.1. Modeled Emissions/Stack Parameters
Actual hourly emission rates for the Conemaugh and Seward power plants were used for Sierra
Club's modeling analysis. Because emission rates from either of the facilities' continuous
emissions monitoring systems (CEM) were not publicly available, the Sierra Club modeling
analysis relied on hourly emissions data from EPA's CAMD database. Source emissions were
modeled for 5 distinct 3-year periods: 2015-17, 2016-18, 2017-2019, 2018-20, and 2019-21.
EPA's summary focuses on the last 3-year (2019-21) period for its assessment. The other 3-year
simulation periods had higher final modeled concentrations (along the Laurel Ridge) and may
indicate over time that there were some reductions in Seward's actual hourly SO2 emissions as
noted by EPA in its Remand documentation.
Physical stack parameters, such as stack base elevation, stack heights and stack diameter were
obtained from the December 2019 report prepared by AECOM for the Conemaugh and Seward
power plants. Pennsylvania's Supplemental Analysis was focused on the areas in the southeast
portion of the Indiana, PA nonattainment area where Sierra Club's previous modeling had
identified 1 area along the Laurel Ridge inside the Indiana, PA nonattainment area that exceeded
the 1-hr SO2NAAQS.
The Sierra Club model input files did not explicitly name the 2 sources that were modeled; they
were referred to in the input model input files as source "S01" and "S09". EPA matched the
stack location and elevation parameters with the Supplemental Modeling input files to identify
source "S01" as Conemaugh's merged FGD stack and source "S09" as Seward's combine CFB
units' stack. EPA verified that Sierra Club's location, stack base elevation matched what was in
Pennsylvania's Supplemental Analysis modeling input file.
Modeled stack diameters for Seward were identical for the Sierra Club and Supplement
Modeling input files. Sierra Club used Conemaugh's Supplemental Modeling analysis' merged
(FGD) stack diameter. Conemaugh's FGD stack has 2 individual flues (each with 7.32 meter
stack diameters) that service each individual coal-fired boiler units. Using the merged stack
diameter may impact final model concentrations since their combined flow rates were used to
estimate hourly stack emissions. This is especially important for hours when only one of
Conemaugh's coal-fired units is operating. Passing the CAMD flow rates through a merged stack
diameter, which is intended to be used when both units are on simultaneously, probably
underestimates stack velocity (and increases modeled concentrations).
EPA examined the Sierra Club's hourly emission file, which included hourly varying emission
rates, stack temperature and stack velocity inputs into AERMOD. Hourly emissions rates were
compared with the corresponding measurements for Conemaugh and Seward that EPA
downloaded using its CAMD FACT software for the 2019 through 2021 time period. Each
source will be examined separately in the following sections.
Conemaugh Modeled Hourly Emission Rates: Hourly emissions from Sierra Club's
AERMOD input file and CAMD were processed using R so that they could be directly compared
over the 3-year simulation period (1 Jan 2019 through 31 Dec 2021). Note that AERMOD inputs
utilized metric units while CAMD reports in Imperial units. Conversions were made to the
AERMOD input file to convert all hourly emission rates to pounds per hour (versus grams per
123
-------
second) using National Institute of Standards and Technology or NIST conversion factors. EPA
utilized information on instrument validity (MODC discussed previously) to identify hours with
valid measurements, referred to as "measured" versus invalid hours which are referred to as
"calculated". Unless otherwise noted, most comparisons were limited to hours CAMD identified
as "measured". Total "calculated" hours are generally limited to less than 100 hours over the 3-
year simulation period. Therefore, excluding these hours is not important as far as identifying
any serious potential differences between Sierra Club's modeled emission rates and ones from
the CAMD database.
Conemaugh's hourly CAMD SO2 emission rates over the 2019 to 2021 simulation period were
shown in the KEY-CON section and for brevity are omitted here. Figure 5.5-17 shows a
comparison of hourly emissions over the 3-year simulation period. A 1-to-l (red) tend line is also
included on the figure. If the CAMD and Sierra Club's modeled hourly SO2 emission rates are
identical, then they will graph as a point along the 1-to-l (red) trend line.
Figure 5.5-17. Conemaugh's CAMD versus Sierra Club Hourly SO2 Emissions
Conemaugh Merged Stack Emission Rate Comparison
CAMD vs. Sierra Club
1 Jan 2019 through 31 Dec 2021
£ 4,000-
"c/T
—I 3,500-
(D
-I—»
03
a:
o 2,500-
-------
emission rates for Conemaugh are nearly identical to the hourly SO2 emission rates used by
KEY-CON.
Conemaugh Modeled Hourly Stack Temperatures: In general, hourly stack temperatures and
stack velocities are only available from stack CEMS units, which are not usually available for
public examination. Sierra Club' modeling utilized a constant stack temperature for
Conemaugh's FGD (merged) stack. A stack temperature of 325 K was selected based on Table
2-2 from Pennsylvania's Supplemental Analysis' model documentation. EPA confirmed this
temperature by reviewing the Supplemental Analysis documentation and Sierra Club's model
input file. For comparison, KEY-CON's modeled stack temperatures (based on CEMS data) for
Conemaugh ranged from 291.074 K to 327.928 K with a mean of 322.6 K. Based on this
comparison, Sierra Club's modeled stack temperature for Conemaugh seems reasonable.
Conemaugh Modeled Hourly Stack Velocities: Sierra Club's construction of Conemaugh's
(merged FGD) modeled hourly varying stack velocities is described on page 4 of its modeling
documentation (Exhibit 4). Supporting spreadsheets were provided by the Sierra Club to show
how it constructed a flow to heat input ratio to determine modeled exit (stack) velocities.
EPA compared Sierra Club's modeled hourly stack velocities (in meters per second or m/s) to
the corresponding hours from the KEY-CON simulation. KEY-CON's hourly stack velocities for
Conemaugh's (merged FGD) stack were based on CEMS data. EPA combined the CAMD and
KEY-CON hourly emissions to eliminate hours with invalid flow MODC to ensure only hours
with valid flow data (according to CAMD) were compared. There are 23,783 hours identified
with "measured" values. The simulation period included a total of 26,304 hours, the difference
being hours with both of Conemaugh's units off or having "calculated" values.
Figure 5.5-18 shows a comparison of the Sierra Club's stack velocity for Conemaugh versus the
corresponding hour value from the KEY-CON input files. A 1-to-l (red) trend line is added onto
the graph. Hourly stack velocities on or close to the 1-to-l (red) trend line indicate Sierra Club's
modeled stack velocity is or is nearly identical to the CEMS based stack velocity (from KEY-
CON).
Overall, the Sierra Club's stack velocities appear to exceed ones based on Conemaugh's CEMS
data. There is a slight overestimation bias of about 0.398 m/s. We also note that the Sierra Club's
stack velocities on the upper end of the distribution appear to be above the corresponding CEMS
values. This could mean Sierra Club's stack velocities are biased high when the units are near
maximum operations (and correspondingly at their maximum emission rates). Both of these
observations may lead to model underpredictions since higher stack velocities are generally
associated with lower dispersion model concentrations.
125
-------
Figure 5.5-18. Conemaugh's KEY-CON versus Sierra Club Hourly Stack Velocity
Comparison
Conemaugh Merged Stack Velocity Comparison
KEY-CON vs. Sierra Club
1 Jan 2019 through 31 Dec 2021
KEY-CON Stack Velocity (m/s)
Seward Modeled Hourly Emission Rates: EPA used the same processing steps described
earlier to compare the Sierra Club's hourly modeled emission rates with corresponding hourly
emissions from the EPA's CAMD database. Again, our comparisons were limited to hours
CAMD identified as "measured". Total "calculated" hours are generally limited to 129 hours
over the 3-year simulation period. Therefore, excluding these hours is not important as far as
identifying any serious potential differences between Sierra Club's modeled emission rates and
ones from the CAMD database.
Seward's hourly CAMD SO2 emission rates over the 2019 to 2021 simulation period were shown
in the KEY-CON section and for brevity are omitted here. Figure 5.5-19 shows a comparison of
hourly emissions over the 3-year simulation period. A 1-to-l (red) trend line is also included on
the figure. If the CAMD and modeled SO2 hourly emission rates are identical, then they will
graph as a point along the 1-to-l (red) tend line.
Hourly CAMD emissions versus the Sierra Club's modeled emission rate shows, for the most
part, the modeled and CAMD SO2 emissions are well matched. There only 9 hours across the
simulation period where CAMD hourly SO2 emissions were more than 50 lbs/hr higher than the
126
-------
Sierra Club's modeled emission rates. We note that the Sierra Club's hourly SO2 emission rates
for Seward are nearly identical to the hourly SO2 emission rates used by KEY-CON.
Figure 5.5-19. Seward's CAMD versus Sierra Club Hourly SO2 Emissions
a)
-1—»
03
£Z
o
'
-------
EPA compared Sierra Club's modeled hourly stack velocities (in meters per second or m/s) to
the corresponding hours from the KEY-CON simulation. KEY-CON's hourly stack velocities for
Seward's stack were based on their CEMS data. EPA combined the CAMD and KEY-CON
hourly emissions to eliminate hours with invalid flow MODC to ensure only hours with valid
flow data (according to CAMD) were compared. There are 21,657 hours identified with
"measured" values. The simulation period included a total of 26,304 hours, the difference being
hours with Seward's units not operating or having "calculated" values.
Figure 5.5-20 shows a comparison of the Sierra Club's stack velocity for Seward versus the
corresponding hour value from the KEY-CON model input files. A 1-to-l (red) trend line is
added to the graph. Hourly stack velocities on or close to the 1-to-l (red) trend line indicate
Sierra Club's modeled stack velocity is or is nearly identical to the CEMS based stack velocity
(from KEY-CON).
Overall, the Sierra Club's stack velocities appear to exceed ones based on Seward's CEMS data.
There is an overall underestimation bias of about -1.533 m/s. There doesn't appear to be a good
match between the stack velocities Sierra Club used versus stack velocities KEY-CON
constructed from its CEM data. There is a large grouping of stack velocities where Sierra Club's
stack velocities are about half of the corresponding KEY-CON values. Underestimating stack
velocities on this magnitude will probably contribute to higher model concentrations and
significantly impact where peak model concentrations are simulated by the model.
Figure 5.5-20. Seward's KEY-CON versus Sierra Club Hourly Stack Velocity Comparison
o
03
-i—•
CO
_Q
o
CO
s_
CD
Seward Stack Velocity Comparison
KEY-CON vs. Sierra Club
1 Jan 2019 through 31 Dec 2021
0 i ii 'i i i i i i i
0 5 10 15 20 25 30 35 40 45 50
KEY-CON Stack Velocity (m/s)
128
-------
5.5.2.2. Meteorology and Surface Characteristics
Sierra Club processed surface meteorological data from the Johnstown-Cambria County airport
(JST) ASOS tower with concurrent upper air data collected near the Pittsburgh International
airport in Allegheny County, PA. Multiple 3-year periods from 2015 through 2021 were
processed with AERMET (version 21112). One and five-minute data was processed using
AERMINUTE (version 15272) to supplement the hourly collected Integrated Surface Database
(ISD) data.
Surface characteristics were included in the AERMET processing for JST. This was done by
running AERSURFACE (version 20060) for the JST ASOS tower location. Sierra Club's
modeling utilized default values for determining surface roughness, zo radius set to 1 km, with
Bowen and albedo values determined for a 10 km area surrounding the ASOS tower.
AERSURFACE sector widths were set to 12 equal 30° sectors from the input tower location with
all sectors set to airport settings. Monthly values were exported for each sector with average
(moisture) and snow cover assumed for all winter month according to Sierra Club's
AERSURFACE input file. The month season settings from the AERSURFACE input file were
set accordingly; winter (Jan, Feb, Dec with snow cover), spring (Mar, Apr, May), summer (Jun,
Jul, Aug) and fall (Sep, Oct, Nov).
Figure 5.5-21 shows the AERSURFACE sectors used by Sierra Club. The aerial photo shows
how the AERSURFACE sectors are aligned. EPA checked the AERSURFACE configuration
Sierra Club used and found the following possible flaws in Sierra Club's AERSURFACE
processing:
• The JST ASOS tower location Sierra Club used is approximately 300 m west-northwest
of where the ASOS tower is actually located. This could introduce significant errors in
the surface roughness calculations made by AERESURFACE.
• No adjustments were made to the FREQ SECT's airportflag values in the
AERSURFACE input file (see section 3.2.9 of the AERSURFACE User's Guide for
additional discussion). A visual inspection of the sectors shows some sectors are probably
not confined to the formal airport footprint (sector IV for example). AERSURFACE uses
the airport flag value to adjust the sector surface roughness values. Non-airport defined
sectors, especially if they include significant tree cover in the land use/land cover
settings, may have significantly different surface roughness values without the non-
airport distinction.
• Snow cover was assumed for all winter months without supporting evidence (a survey of
local snow cover information). Assumed snow cover over the winter months (Jan, Feb,
Dec) significantly changes the albedo values calculated by AERSURFACE and may
impact final AERMOD concentrations during these months.
• Average soil moisture values were assumed over the entire modeling period.
AERSURFACE modifies Bowen ratio values based on monthly rainfall totals compared
to long-term 30-year climate averages (see section 2.3.3 of AERSURFACE User's
Guide). Not properly accounting for seasonal variability in precipitation/soil moisture
values will impact the final AERMOD simulation values.
129
-------
Figure 5.5-21 Sierra Club Modeling AERSURFACE Sector Definition
Westmoreland-Cambria, PA: Sierra Club AERSURFACE
V
/"v. '•5'/
Legend
/"
^ Sierra Club JST ASOS
™ Location
• A *
Actual J ST A SOS Tower
jPjESPip'
~*| Sierra Club Sectors
N
S
r wm •• " * * ^
0 0.4 0.8 1.6 Kilometers rn/i j-imii st>i»
1 1 1 1 1 1 1 1 1 sy ERr^oVrcr"'"
Sierra Club chose to use the Johnstown-Cambria County airport in its modeling analysis due to
the site's close proximity to the Conemaugh and Seward power plants; the airport is nearly 20
km east of Conemaugh and Seward. As noted previously, there are significant elevation (and
terrain) differences between the location of the Johnstown-Cambria County airport and
Conemaugh and Seward. Base elevations at the airport are almost 360 m higher than the
Conemaugh and Seward power plants. Additionally, the power plants are located in the Ligonier
Valley with significant terrain features to the west (Chestnut Ridge) and east (Laurel Ridge).
Johnstown-Cambria County airport is located on very high terrain (base elevations around 700
m) making it very exposed to the elements. By contrast, the Conemaugh and Seward power
plants reside in a broad valley that is probably subject to topographically induced flow patterns.
Given the unique features impacting local meteorological patterns, a site-specific meteorological
collection program was undertaken with collocated meteorological tower instruments and
SODAR instruments deployed to collect 1 year of data at the Ash Site #1 to represent conditions
near the Conemaugh and Seward power plants. This data was used in EPA's analysis since it
best represents local conditions that impact the emissions from these plants. Utilizing surface
meteorological data from the Johnstown-Cambria County airport will probably not entirely
capture transport characteristics in the vicinity of the Conemaugh and Seward power plants.
Figure 5.5-22 shows a wind rose for the 3-year Johnstown-Cambria County airport reviewed by
EPA. It is taken from the AERSURFACE processed AERMET file included in Sierra Club's
public comment submittal and was produced using R's openair package. Predominant winds
130
-------
were from the west. Winds from the northeast quadrant occur very infrequently. Average wind
speeds were about 4.3 m/s or a little over 8 nautical miles per hour or knots. As noted, the
Johnstown-Cambria County airport is located in exposed elevated terrain and is therefore subject
to higher wind speeds.
Figure 5.5-22. Johnstown-Cambria County Airport Wind Rose for 2019-21
Sierra Club AERMET SFC File: 2019-21
10%
5%
w
N
^ E
mean = 4.3448
s calm = 0.7%
10 to 14.66
8 to 10
6 to 8
4 to 6
2 to 4
0 to 2
(ms"')
Frequency of counts by wind direction (%)
EPA would also like to document the following potential issues with Sierra Club's AERMET
processing:
• Sierra Club set Johnstown-Cambria County's anemometer height at 7.82 m (in the stage 3
AERMET (version 21112) input file. The actual anemometer height is 26 feet or 7.92 m.
• There are several warning messages in the AERMET stage 1 and 3 output files noting
missing morning upper air soundings. EPA described the issue previously (section 5.2.6)
and how they can be filled to ensure the data is processed into the final AERMET ready
meteorological output files.
131
-------
5.5.2.3. Area of Analysis and Receptor Grid
Sierra Club's model receptor grid was described in its 13 April 2022 Evaluation of Compliance
Report from Wingra Engineering as follows:
[T]wo receptor grids were employed:
• A 100-meter Cartesian receptor grid centered on the two plants and extending out 10
kilometers.
• A 500-meter Cartesian receptor grid centered on the two plants and extending out 20
kilometers.
To reflect a representative inhalation level, a flagpole height of 1.5 meters was not usedfor
all modeled receptors. The use of a flagpole height is not expected to significantly affect the
predicted impacts. This is similar to the approach usedfor the December 2019 modeling
report.
Elevations for receptors were obtainedfrom National Elevation Dataset (NED) GeoTiffdata.
GeoTiffis a binary file that includes data descriptors and geo-referencing information
necessary for extracting terrain elevations. These elevations were extractedfrom 1 arc-
second (30 meter) resolution NED files. The USEPA software program AERMAP v. 18081 is
usedfor these tasks.
EPA confirmed Sierra Club's model receptor spacing description; we did not reprocess the
receptor grid through AERMAP to confirm receptor elevations and hill-height scales. Figure 5.5-
23 shows an overview of the Sierra Club model receptor grid, which cover parts of the Indiana,
PA nonattainment area and extends into portions of Cambria, Indiana, Somerset and
Westmoreland counties near the Conemaugh and Seward power plants.
Figure 5.5-24 shows model receptor spacing near the power plants and along the Laurel Ridge.
Note that Conemaugh and Seward's ambient air boundaries are not defined in the Sierra Club's
model receptor grid. The 100-m grid spacing may not be fine enough to capture the maximum
modeled concentration for the Sierra Club simulation. A more refined grid along the Laurel
Ridge would probably yield higher model concentrations.
132
-------
Figure 5.5-23. Sierra Club Model Receptor Grid and Primary Modeled SO2 Sources
Sierra Club Model Receptor Grid Overview
-&EPA
Figure 5.5-24. Sierra Club Model Receptor Grid and Primary Modeled SO2 Sources
Sierra Club Model Receptor Grid Near Conemaugh & Seward
SERA
133
-------
5.5.2.4. Background Concentration
Sierra Club's modeling analysis included a season by hour of day varying background
concentration. This follows the background concentration construction method outlined in EPA's
March 1, 2011 1-hour NO2 clarification memo allowing for the use of a temporally varying
background concentrations. Season by hour of day 1-hr SO2 background concentrations were
taken from the South Fayette monitor located in Allegheny County. Sierra Club's AERMOD
input file identifies the monitor and period (2016-18) used to develop the model background
concentration. This is the same monitor and period used in Pennsylvania's Supplemental
Analysis. EPA compared Sierra Club's and Pennsylvania's Supplemental Analysis and verified
that the season by hour of day background values matched.
EPA downloaded hourly SO2 concentrations for the South Fayette monitor using R's
RAQSAPI's library for the 2016-18 time period. We used R to configure the season by hour of
day 1-hr S02 background concentrations in accordance with our March 1, 2011 guidance. Table
5.5-8 summarizes EPA's constructed season by hour of day background concentrations from
South Fayette's 2016-18 monitor values. This allows a comparison of the Sierra Club's modeled
background concentrations versus EPA's and KEY-CON's. We note that KEY-CON's season by
hour of day background concentrations (in parts per billion or ppb) were entered to 2 decimal
places. Our values, also in ppb, preserve 1 decimal place.
Generally, the model background concentrations used in Sierra Club's modeling analysis are
slightly higher than the ones used by EPA and KEY-CON. This may reflect slightly more
regional coal usage over the background period Sierra Club used versus more recently.
134
-------
Table 5.5-8. EPA Constructed 2019-21 Season by Hour of Day Background
Concentrations for the South Fayette, PA Monitor Located in Allegheny County.
Winter
Spring
Summer
Fall
South Fayette, PA (42-003-0067): 2016-18 Background S02 Concentrations (ppb) by Season/Hour of Day
Hour
Average
Missing
Total
Hours
Average
Missing
Total
Hours
Average
Missing
Total
Hours
Average
Missing
Total
Hours
1
4.0
0
271
3.3
1
276
2.7
2
276
3.3
1
273
2
4.0
0
271
2.0
1
276
3.3
3
276
3.3
0
273
3
4.3
0
271
2.0
1
276
1.7
2
276
3.0
0
273
4
3.3
0
271
2.0
1
276
1.7
2
276
3.0
0
273
5
3.7
0
271
2.3
1
276
1.7
2
276
2.3
0
273
6
4.0
0
271
3.0
1
276
2.0
2
276
2.3
0
273
7
5.0
0
271
5.0
1
276
2.3
3
276
2.3
0
273
8
4.3
0
271
4.7
1
276
3.3
3
276
3.0
0
273
9
5.3
0
271
3.7
3
276
3.3
3
276
4.3
0
273
10
5.3
0
271
4.7
7
276
4.7
21
276
3.3
8
273
11
5.0
3
271
3.7
32
276
3.0
32
276
3.3
23
273
12
4.3
27
271
2.7
9
276
3.3
4
276
4.0
16
273
13
3.3
7
271
2.7
2
276
3.3
3
276
2.7
2
273
14
3.3
1
271
3.3
1
276
3.0
4
276
2.7
2
273
15
4.0
0
271
3.0
1
276
4.0
4
276
4.0
1
273
16
3.3
0
271
3.0
1
276
4.0
1
276
3.7
0
273
17
2.7
0
271
3.3
1
276
3.7
1
276
4.3
0
273
18
3.3
0
271
4.7
1
276
3.7
1
276
5.0
0
273
19
3.3
0
271
6.0
1
276
4.0
2
276
4.0
0
273
20
4.0
0
271
3.7
1
276
4.0
2
276
3.7
0
273
21
3.3
0
271
2.7
1
276
2.7
2
276
4.0
0
273
22
4.0
0
271
2.7
1
276
2.3
2
276
3.3
0
273
23
4.3
0
271
2.7
1
276
2.7
2
276
2.7
0
273
24
4.0
0
271
2.3
1
276
2.3
2
276
3.3
0
273
5.5.2.5. Sierra Club Model Results
Sierra Club's modeling roughly followed EPA's Modeling TAD. EPA reviewed the modeling
processing for the 2019-21 time period, which is the most recent of the 3-year periods Sierra
Club submitted during the Remand comment period.
Figure 5.5-25 shows Sierra Club's 2019-21 modeled 1-hr SO2 design values over its entire
modeling domain. Sierra Club's AERMOD concentrations were converted to parts per billion or
ppb by multiplying the model concentrations by a conversion factor; 75 ppb over 196.4 |ig/m3.
Model concentrations are overlain over the local topographic elevations.
Model 1-hr SO2 design values are elevated along the Laurel Ridge that resides southeast of
Conemaugh and Seward as noted in Figure 5.5-26. The peak receptor had a modeled design
value of 93.4 ppb. Sierra Club's modeling shows patches of areas with 1-hr SO2 design values
above 75 ppb along the Laurel Ridge. All areas above the 1-hr SO2 NAAQS are confined to
135
-------
portions of the Laurel Ridge that face Conemaugh and Seward. There are also minor peaks in
model 1-hr SO2 design values along the Chestnut Ridge west of the plants and also peaks to the
northeast near Robindale Heights in East Wheatfield Township, Indiana County.
Figure 5.5-25. Sierra Club AERMOD Results for All Sources Plus Background
Sierra Club Model Results
136
-------
Figure 5.5-26. Sierra Club AERMOD Results Along the Laurel Ridge
6. Summary of EPA's Intended Redesignation for the Westmoreland
and Cambria Area
After careful evaluation of all modeling analyses and other technical information, as well as all
available relevant information, the EPA is notifying Pennsylvania that the designation for
portions of Westmoreland and Cambria Counties should be revised to nonattainment for the
2010 SO2 NA AQS. Specifically, the boundaries of the revised area should be comprised of
Lower Yoder Township in Cambria County and St. Clair Township (including the boroughs of
New Franklin and Seward) in Westmoreland County. Figure 6-1 shows the boundary of this
intended revised designated area.
Additionally, the EPA does not intend to change the designations of the remainder of
Westmoreland and Cambria counties.
137
-------
Figure 6-1. Boundary of the Intended Revised Westmoreland and Cambria Nonattainment
Area
Seward
Indiana
r r-i
.SEWARD
NEW»FfLORENCE
Cambria
ST CLAIR
Westmoreland
LOWER YODER
PSMi
Proposed Westmoreland-Cambria, PA 1-Hour SO2 Nonattainment Area
Legend
SO; Sources
Conemaugh
Q Indian a, PA Nonattainment Area
f__Westmoreland-Cambria, PA
^ Nonattainment Area
0 3 6 12 Kilometers C CPA wImmw
1 1 1 1 1 ! 1 1 1 0£irV.rs™,:r''1
138
-------
7. References
AECOM, 2015. Meteorological Monitoring Station Design and Quality Assurance Project Plan
for the Conemaugh and Seward Generating Stations - Indiana County, PA
AECOM, 2019. Supplemental SO2 NAAQS Compliance Modeling Report for the Indiana, PA
Non-Attainment Area (Revision No. 1).
Air & Waste Management Association's Specialty Conference, Guideline on Air Quality
Models: the Path Forward, Raleigh, NC. March 2013.
EPA, 2021, AERMOD Implementation Guide, EPA-454/B-21-006.
EPA, 2021, User's Guide for the AMS/EPA Regulatory Model (AERMOD)EPA,-454fB-2\-00\
EPA, 2021, User's Guide for the AERMOD Meteorological Preprocessor (AERME1), EPA-
454/B-21-004.
EPA, 2020, User's Guide for AERSURFACE Tool. EPA-454/B-20-008, U.S. Environmental
Protection Agency, Research Triangle Park, North Carolina.
EPA, 2016. SO 2 NAAQS Designations Modeling Technical Assistance Document (DRAFT),
August 2016 (https://www.epa.eov/so2-polliition/technical-assistance-dociiments-implementine-
2010- sulfur-di oxi de~ stan dard).
EPA, 2000. Meteorological Monitoring Guidance for Regulatory Modeling Applications, EPA-
454/R-99-005.
"GEP Stack Height Evaluation for Pennsylvania Electric Company's Seward Plant", CPP Project
86-0336. January 26, 1989.
Paine, R.J. 2001. Meteorological Input Data for AERMOD Applications. Presented at the Air &
Waste Management Association Specialty Conference on Guideline on Air Quality Models: A
New Beginning. Rhode Island. April, 2001
Paine, R., F. Tringale, and S. Gossett, 2013. Resolution of 1-hour SO 2 Non-attainment Area in
Kingsport, IN: Advanced Meteorological and Monitoring Study. Control #7, 139resented at the
Shaffner, M. N, Geology and mineral resources of the New Florence quadrangle, Pennsylvania
(Bolivar, New Florence, Wilpen, and Rachelwood 7.5-minute quadrangles, Indiana,
Westmoreland, Cambria, and Somerset Counties), 1958, PA DC NR. publication, available at:
http://maps.dcnr.pa. eov/publications/Default.aspx?id=9
139
------- |