Comments on

Revised Guidance for Ozone and Fine Particulate Matter

Permit Modeling
(Issued by EPA on September 20, 2021)

Comments provided by Robert J. Paine, CCM, QEP and David W. Heinold, CCM

AECOM, Chelmsford MA
November 19, 2021

1.0 Introduction.

The authors provide the following comments regarding the United States Environmental Protection
Agency's (EPA's) revised Guidance for Ozone and Fine Particulate Matter Permit Modeling ("Revised
Guidance" or "Guidance"), issued on September 20, 2021. This guidance was referenced in the latest
revisions to the Guideline on Air Quality Models (40 CFR Part 51 Appendix W) promulgated on January
17, 2017 (82 FR 5182) as Reference 59. The Revised Guidance relies upon and references various EPA
guidance documents that have been issued since January 2017, but for which EPA did not solicit
comments. Due to the reliance of the Revised Guidance on these previously issued documents, our
comments address these referenced guidance documents as well.

2.0	EPA has changed its approach yet again regarding the modeling requirements involving emission
increases of applicable pollutants.

The revised policy indicates that if any of the precursor pollutants (or the primary pollutant in the case
of fine particulate matter, or PM2.5) is emitted in amounts greater than the Significant Emission Rate
(SER), then all precursor pollutants (plus the direct pollutant in the case of PM2.5) must be included in
the modeling. This is the case even if some of the remaining precursor or direct pollutants are emitted
in insignificant amounts.

2.1	EPA has been inconsistent in this approach, especially for PM2.5.

In the original 2014 Draft Guidance, EPA's policy was to not require modeling of direct PM2.5 emissions
if they were not emitted in significant amounts by the project, whether or not the precursor emissions
of S02 or NOx were more than their SER of 40 tons per year (tpy). Then, at the 2018
Regional/State/Local Workshop in Boston, EPA presented1 a major update without any formal guidance
document to this policy consistent with the current Revised Guidance. Although this update was only a

1 Presentation available at

ftp://nevyftp.epa.gov/Air/aqmg/SCRAM/workshops/2018 RSL Modelers Workshop/Presentations/1-

20 2018 RSL-Q3 PM25 Modeling Guidance.pdf.

AECOM comments on the Revised Guidance for Permit Modeling for Ozone and PM2.5 November 2021
Page 1


-------
presentation and not official guidance, many regulatory reviews of permit applications submitted
afterward relied upon the 2018 workshop presentation. Then, in February 2020, EPA changed its mind
again and in its Draft Guidance, it decided that only those pollutants emitted in significant amounts
would need to be modeled. Now, EPA has revised the policy once again. This inconsistency conveys
doubt and uncertainty in the durability of EPA's guidance. There are also several technical issues that
complicate this revised policy, as noted in comment areas below.

The increased likelihood of cumulative modeling analyses brings additional focus upon current
conservative features of Appendix W procedures as well as AERMOD features (more details provided
below) that EPA needs to address in the next Appendix W update. These conservative features will
become even more critical to establishing compliance if the PM2.5 NAAQS is further tightened in the
current NAAQS review. EPA needs to review and work to correct several overly conservative features of
cumulative modeling analyses based on current guidance. These features tend to overestimate impacts
of the source seeking a permit and also tend to double-count impacts of all other sources. The result of
these approaches is an overstatement of net air quality impacts that could severely constrain the
permitting process, in some instances blocking environmentally beneficial projects. In this context, we
provide several comments below that address areas that EPA should address in the next Appendix W
revision, especially in light of possible tightening of the PM2.5 NAAQS.

2.2	A review of the MERPs indicates that the SER levels for precursor pollutants would always result
in insignificant impacts.

For precursor pollutant emissions just at the SER levels (40 tons per year for S02, NOx, and VOCs), it is
evident that the combined precursor pollutants Tier 1 Modeled Emission Rates for Precursors (MERP)
concentration would be well under the SILfor both ozone and PM2.5. For ozone, the lowest MERPs
anywhere in the country are 125 TPY for NOx and 1,049 TPY for VOCs. For PM2.5, the lowest MERPs
between the daily and annual averaging times are 188 TPY for S02 and 1,073 TPY for NOx. It is apparent
that the SER value of 40 TPY is a small fraction of the MERPs, and therefore the policy of not requiring
modeling for proposed emissions below the SERs is on sound technical ground. The exclusion of the
contributions in a Tier 1 analysis for small emission increases of a precursor pollutant for a proposed
source in the 2020 Draft Guidance was appropriate because 1) the concentration impact is a low fraction
of the SIL, and 2) the MERP procedures are inherently conservative. Combined, these factors add a
further safety buffer that makes the 2020 Draft Guidance approach suitably protective of air quality.

2.3	The guidance does not provide credit for decreases in precursor pollutant emission rates, which
should be applied to offset increases in other precursor emission rates.

If one of the precursor pollutants for a proposed project will have an emissions decrease, then the
resulting decrease in PM2.5 production should be considered as offsetting the effects of an increase in
the other precursor pollutant. An example of how this offset could be considered is provided below.

AECOM comments on the Revised Guidance for Permit Modeling for Ozone and PM2.5 November 2021
Page 2


-------
The offsets to be provided by the emissions decrease for one precursor pollutant to the emissions
decrease for the other precursor pollutant can be computed using the MERP relationships. For a PM2.5
example, suppose that the applicable MERP for S02 is 200 tons per year and it is 400 tons per year for
NOx. In this hypothetical case, assume an increase in S02 emissions of 50 tons per year and a decrease
in NOx emissions of 50 tons per year. Since in this case NOx emissions are only 50% as effective in
producing secondary PM2.5, a reduction in NOx emissions would offset the S02 production of PM2.5 at
a 1:2 ratio. Therefore, a reduction in NOx emissions of 50 tons per year would be equivalent to an
offset of 25 tons of S02. This would lead to an effective increase in S02 emissions (for the purpose of
secondary formation of PM2.5) of 25 tons per year. This results in a net insignificant increase in S02 and
no increase in NOx. Therefore, the net increase in precursor emissions is insignificant in this case.

2.4 The Tier 1 modeling approach is conservative and leads to overestimates in the modeled
concentration for any given emission rate.

The MERPs are based upon a peak concentration over the entire modeling domain due to a single
precursor pollutant, unpaired in time and space with impacts from other precursor pollutants and (in
the case of PM2.5) with direct emissions of the pollutant. If these peak impacts from multiple
precursors are added, they are implicitly assumed to occur simultaneously and at the same geographical
location. This scenario is so unlikely that it probably never occurs, especially considering that maximum
impacts from directly emitted pollutants occur close to a source and secondary pollutants resulting from
precursor emissions necessarily occur downwind after sufficient transport time. Adding maximum
modeled impacts from multiple precursor and direct pollutants is inherently conservative. Therefore,
the 2020 Draft Guidance was sufficiently conservative in focusing upon the impact of individual large
emission components that may not coincide with the impacts of other minor emission components. The
Revised Guidance increases even more the conservatism of the Tier 1 approach. We recommend that
EPA return to the 2020 guidance for a Tier 1 analysis because it was sufficiently conservative.

If EPA decides not to return to the 2020 guidance approach, then we provide recommendations to
reduce the conservatism of the Tier 1 analyses. Although EPA has provided Tier 1 MERP results as a
function of distance (with 10-km distance intervals), there is no information at distances under 10 km
because many of the photochemical grid modeling runs were conducted with a resolution of 12 km.

This is not a serious issue for ozone modeling, but it is for PM2.5 modeling because peak impacts for
primary PM2.5 could be very close to the facility fenceline, well within 10 km. Due to the conversion of
gas to secondary particle formation is a function of travel time and there is zero conversion at the stack,
it is very likely that for small travel times, the 10-km estimate for the secondary formation is far too
high. EPA should create a distance relationship formula between 0 and 10 km to fill in this gap that
produces realistic results between zero secondary conversion at the stack location and the stated
secondary production at 10 km.

AECOM comments on the Revised Guidance for Permit Modeling for Ozone and PM2.5 November 2021
Page 3


-------
3.0 Tier 2 modeling procedures are likely to be required more frequently if the Revised Guidance is
not changed to reduce over-conservatism, and pre-approved modeling platforms should be
readily accessible to permit applicants.

If the Revised Guidance procedures are retained and the recommended enhancement in the Tier 1
approach noted above is not provided, then permit applicants may need to apply Tier 2 approaches.
The current guidance on the more realistic and refined Tier 2 approaches raises a very high bar and
attendant permitting delays for protocol approval. We encourage EPA to prepare to provide pre-
approved approaches and modeling platforms for Tier 2 photochemical modeling exercises to mitigate
potential delays and protocol complications. The Windows-based CAMx approach noted by Tim Allen2
(of the Fish and Wildlife Service) as presented at the 2021 Air & Waste Management Association is an
example of an emerging method to make such modeling easier for permit applicants as well as agency
reviewers.

4.0 Several aspects of the current AERMOD modeling system need to be refined to avoid excessively
conservative concentration estimates for the proposed new project emissions of direct PM2.5.

It is our understanding that EPA is considering or presently working on several areas for improving
AERMOD performance designated as "white paper" issues. We encourage EPA to implement those
features in a future regulatory version of AERMOD that implements such features as noted below.

Low wind speed improvements. EPA started this process with the ADJ_U* option, but the work has not
been completed. EPA needs to seriously consider adding minimum sigma-v, sigma-w, and/or other
features to the guideline version of AERMOD. The American Iron & Steel Institute provided comments
on this issue in their written submittal to the docket of the 12th EPA Modeling Conference (EPA-HQ-OAR-
2019-0454), and these comments are incorporated here by reference.

Building downwash improvements. EPA is familiar with ongoing developments by the "PRIME2"
workgroup and EPA's Office of Research and Development. Cases of overpredictions in low wind, stable
conditions have been documented, and we are hopeful that updates to AERMOD's treatment of
downwash will improve model performance in these conditions.

Source-caused effects on plume rise. We have documented conditions where the source itself enhances
plume rise: large industrial areas with large fugitive heat releases, heat releases from individual
buildings, stacks in a line for which plumes partially merge, and stacks with moist plumes. With
appropriate documentation, modeling applicants should be permitted to apply these advanced
approaches to obtain improved model performance.

2 Allen, Tim, 2021. Panelist for session # 983742 entitled, "AERMOD Modeling System Updates with U.S. EPA and
Federal Land Manager Initiatives". Air & Waste Management Association's 114th Annual Conference and Exhibition
(Virtual).

AECOM comments on the Revised Guidance for Permit Modeling for Ozone and PM2.5 November 2021
Page 4


-------
Partial plume penetration modeling. We have documented conditions where AERMOD mixes the
penetrated plume to the ground too early in the day, well before the convective mixing layer rises to
intercept it. The "Highly-Buoyant Plume" (HBP) enhancement to AERMOD should be incorporated into
the model to address this deficiency in AERMOD.

Particulate emissions from facility roadways (including "haul roads"). EPA has not adequately resolved
overpredictions from mobile source emissions, especially on facility roadways for which barriers act to
reduce the effective emission rate. Comments by the American Iron & Steel Institute submitted in
association with EPA's 12th modeling conference have more details about model improvements to
consider.

Further progress on BUOYLINE implementation. EPA's implementation of the BLP modeling into
AERMOD starting with version 16216 was a first step in an incomplete integration of buoyant line
sources into AERMOD. Remaining steps should include: a) enabling AERMOD's complex terrain
treatment (using the dividing streamline concept) to be applicable to buoyant line sources, b)
consideration of plume dispersion using AERMOD's formulation rather than the Pasquill-Gifford stability
class-based approach that AERMOD replaced upon promulgation. These advancements could be
accomplished by using BLP's plume rise approach and then utilizing an hourly volume source approach
to have AERMOD implement the advancements noted above.

Updated AERSURFACE roughness lengths need to be reviewed. Another white paper issue involves the
manner in which AERSURFACE is computing surface roughness lengths. These values are lower than
expected due to the geometric mean computational approach. A better approach would focus upon
determining the effects of different surfaces or surface cover types to obtain the effective roughness
height (zOef) for a heterogeneous surface. In this approach, the "tiling method" for determining the
relevant surface energy flux of a surface grid in a numerical weather prediction model can be used (e.g.,
see Manrique-Sunen et al., 2013)3.

5.0 Appendix W should allow modeling approaches that account for emissions variability and
intermittent utilization of a source.

Permit applicants are challenged with representing variable emissions with a modeling system that is
not designed to accommodate the random nature of some source emissions. Although AERMOD can
accommodate planned emission changes by time of day, day of the week, monthly, etc., it currently
does not have the capability to handle the more commonplace random high emission periods. AECOM
has developed an approach4 that can handle this issue with the use of a large number (e.g., 100) sets of

3	Manrique-Sunen, A., A. Nordbo, G. Balsamo, A. Beljaars, and I. Mammarella, 2013: Rep-
resenting land-surface heterogeneity: Offline analysis of the tiling method. J. Hydrometeor., 14, 850-867.

4	Warren, C. and R. Paine, 2017. Modeling of Infrequent and Random Peak Emission Events for Permit Applications.
Paper MO07, presented at the Air & Waste Management Association's Specialty Conference, Guideline on Air
Quality Models: the Changes, Chapel Hill, NC. November 2017.

AECOM comments on the Revised Guidance for Permit Modeling for Ozone and PM2.5 November 2021
Page 5


-------
"Randomly Reassigned Emission" input streams that rely upon a defined emissions distribution. The
placement of the high emission hours for this approach is tailored to the individual situation. The result
is a rigorous modeling test governed by a defined emissions distribution and rules for sequences of high
emission periods.

6.0 EPA's treatment of nearby source emissions in Appendix W Table 8-1 needs to be updated.

For a cumulative modeling exercise, EPA requires the characterization of concentration impacts from
both nearby sources and distant sources. The difference between these two categories is that distant
sources have less spatially-varying impact and can thus be represented by monitoring data. The values
used from the monitoring data are conservatively high, representing a high percentile statistic
comparable to the form of the standard (e.g., 99th percentile value for S02). A refined tier for this
approach is to use a 99th percentile value as a function of hour and season.

Modeling is required for nearby sources (e.g., those sources whose emissions are not affected by the
activity represented by the permit application) because the resulting concentrations can vary in space as
well as in time. In its Appendix W changes, EPA attempted to create more realism for this
characterization by changing one of the factors (temporally representative level) used in the emission
rate calculation.

The only factor that was changed for short-term emission rates for nearby sources was the operating
level (MMBtu/hr), rather than also including the two other factors (emissions limit in Ib/MMBtu and
operating factor (hr/year). This change was a step toward more realism, but additional changes are
needed to make this adequately realistic and workable, for the reasons provided below. In general, the
three factors should be combined into a short-term emission rate that might vary by hour and season,
much like the regional background monitoring data used for the distant sources.

The three factors do not make sense for emissions from processes that do not burn fossil fuels.

Examples of this are S02 emissions from smelting activities; the emissions come from a process that
liberates S02 via a chemical or thermodynamic reaction, rather than fuel consumption. Even for fuel
consumption, there are many cases in which the highest operating levels (e.g., full load) does not
correspond to the periods of the highest emission rate in Ib/MMBtu. Therefore, the product of these
parameters is more relevant than treating them separately. In other cases, sources in a group do not act
independently in that when one unit is down, the one next to it is operated at a very high level.
Therefore, assuming that all units are operated simultaneously at peak levels is unrealistic.

The definition of "nearby sources" should be clarified as including any source, even at the same facility
as that involved in the proposed change, whose emissions are not affected by the proposed activity.

The use of Continuous Emissions Monitoring System data can inform the process for determining the
temporal nature of emission changes, as noted in the current Table 8-1. However, the temporal
AECOM comments on the Revised Guidance for Permit Modeling for Ozone and PM2.5 November 2021
Page 6


-------
changes need to be applied to the product of the three factors (e.g., emission rate in Ib/hr) rather than
just one factor because these factors are not independent of each other. The result of the analysis could
be a varying emission rate for a nearby source by hour and month/season.

7.0 EPA's approach for estimating a conservatively high regional background to add to the nearby
background source impacts can significantly overstate the total concentration.

It is important for EPA to provide a more realistic estimate of background concentrations for a
cumulative analysis. The background concentration has two components: the regional background
representing distant sources and minor local sources not being modeled, and the concentrations due to
local sources that could have a varying impact across the modeling domain (mentioned above in Section
6).

The modeling procedures should be designed such that as the emissions from the proposed project
approach zero, the total modeled concentration should approach the current design concentration
without the contribution from the proposed project emissions. The combination of modeling
conservatively high emission rates for nearby sources plus adding a background design concentration
that may include impacts from sources being modeled will clearly overstate current design
concentrations unless there are no nearby sources to model. Therefore, the selection of the background
monitor should be tailored such that if there are nearby sources being modeled, the background
monitor should be selected from a "clean" area which is not influenced by any modeled sources or
sources local to the selected monitor.

AECOM comments on the Revised Guidance for Permit Modeling for Ozone and PM2.5 November 2021
Page 7


-------