EPA Response to External Peer Review
Comments on the EPA Report
Model-extrapolated Estimates of
Airborne Lead Concentrations
at U.S. Airports
(formerly titled 'Methods for Estimating Airborne
Lead Concentrations at Airports Nationwide')
oEPA
United States
Environmental Protection
Agency

-------
EPA Response to External Peer Review
Comments on the EPA Report
Model-extrapolated Estimates of
Airborne Lead Concentrations
at U.S. Airports
(formerly titled 'Methods for Estimating Airborne
Lead Concentrations at Airports Nationwide')
Assessment and Standards Division
Office of Transportation and Air Quality
U.S. Environmental Protection Agency
NOTICE
This technical report does not necessarily represent final EPA decisions or
positions. It is intended to present technical analysis of issues using data
that are currently available. The purpose in the release of such reports is to
facilitate the exchange of technical information and to inform the public of
technical developments.
EPA
Environmental Protection	EPA-420-R-20-004
Agency	February 2020

-------
Table of Contents
1.	Introduction	2
1.1 Peer Reviewers	2
2.	Charge to Reviewers	2
3.	Reponses to Peer Reviewer Comments	4
3.1	Response to Comments Received from Reviewer 1: Steven Barrett	4
3.2	Response to Comments Received from Reviewer 2: Michael Kleeman	18
3.3	Response to Comments Received from Reviewer 3: Barbara Morin	30
3.4	Response to Comments Received from Reviewer 4: John R. Pehrson	47
3.5	Response to Comments Received from Reviewer 5: Sandy Webb	55

-------
1. Introduction
The United States (US) Environmental Protection Agency (EPA) Office of Transportation
and Air Quality (OTAQ) contracted with RTI International to conduct a scientific peer
review of a draft technical analysis report that describes the methods for estimating
airborne lead concentrations at airports nationwide. RTI International, an independent
contractor, facilitated the peer review in compliance with EPA Science Policy Council
Peer Review Handbook, 4th Edition. RTI selected five peer reviewers with expertise in air
quality monitoring and modeling, piston-engine aircraft operations, and potential
impacts of piston-engine aircraft sources. Reviewers were charged with evaluating the
methodology, assumptions, and supporting data used to estimate concentrations of
lead in air from piston-engine aircraft activity at and around airports in the US.
Reviewers were also asked to identify any alternative data/approaches that may
improve EPA's understanding of the potential impacts of piston-engine aircraft activity
on concentrations of lead in air at and near airports. A full description of the peer
review process can be found in in Appendix A, which includes the Contractor's report.
1.1 Peer Reviewers
RTI International selected the following individuals to review the report provided by the
EPA. Reviewers are referred to by reviewer number throughout the response document,
as assigned here alphabetically.
	Reviewer 1: Steven Barrett, Massachusetts Institute of Technology Department
of Aeronautics & Astronautics
	Reviewer 2: Michael Kleeman, University of California-Davis Department of Civil
and Environmental Engineering
	Reviewer 3: Barbara Morin, Rhode Island Department of Environmental
Management
	Reviewer 4: John Pehrson, CDM Smith
	Reviewer 5: Sandy Webb, Environmental Consulting Group LLC
2. Charge to Reviewers
The following charge questions were provided to the reviewers to guide their review
and highlight specific areas for input and comment.
1. Sections 1 and 2 describe the nature of how piston-engine aircraft operate for
safety and logistical reasons, along with the previous work that EPA and others
have conducted to characterize concentrations of lead in air at individual
2

-------
airports. As stated in the report, conducting detailed air quality modeling or
monitoring at all US airports is not feasible due to resource constraints. Please
comment on the extent to which this information is clearly described and
provide your perspective on the approach selected to utilize modeling from an
individual airport in order to characterize concentrations of lead in air at and
downwind of maximum impact areas of airports nationwide.
2.	Section 3.1 presents the methods to calculate Air Quality Factors (AQFs) at the
model airport. Please comment on the approach used to calculate AQFs at the
model airport specifically for the purposes of using these factors to estimate
concentrations at and downwind of maximum impact areas of airports
nationwide.
3.	Table 2 and accompanying text in Section 3.2 describe the methods used to
estimate piston-engine aircraft landing and take-offs (LTOs) at individual runway
ends on a rolling 3-month basis (e.g., apportioning out piston-engine-specific
LTOs from total LTOs at each airport, allocating annual activity to daily and then
hourly periods). Are these methods clearly described and do you have
recommended changes to the steps taken? Please explain any alternative
options and provide the location of data sources that would support such
alternative options.
4.	Section 3.3 presents an analysis to refine estimates of piston-engine aircraft
activity using airport-specific data for a subset of airports. Please comment on
whether there are alternative airport-specific data, or analysis approaches, that
could improve estimates of piston-engine aircraft activity at a subset of airports.
In addition, please comment on whether parameters other than piston-engine
aircraft activity should be included in analyses to potentially improve model-
extrapolated concentrations at a subset of airports, noting that additional
parameters are evaluated at all airports in the uncertainty and variability
analyses presented in Section 4.
5.	EPA provides coarse comparisons of monitored lead concentrations to model-
extrapolation results from the national and airport-specific analyses, in Sections
4.1 and 4.2, respectively. In Section 4.3, EPA provides a more detailed
comparison of data from lead monitors placed in close proximity to the locations
of model-extrapolated concentrations. Please comment on the appropriateness
of the approaches to compare model-extrapolated results to monitored
concentrations of lead given available monitoring data. Based on your
understanding of the methods presented in the report and the comparisons of
monitor and model-extrapolated concentrations, please provide your
3

-------
perspective on the performance of the methods in characterizing the ranges of
lead concentrations from piston-engine aircraft at and downwind of US airports.
6. Section 4.3 presents quantitative and qualitative uncertainty analyses of the
model-extrapolated results provided in previous sections. Please provide your
perspective on the methods used to conduct these uncertainty and variability
analyses, as well as the key parameters EPA included in the analyses (based on
previous work discussed in Section 2). Please provide your perspective on the
application of this analysis to further characterize the range of lead
concentrations attributable to piston aircraft activity at airports nationwide.
3. Reponses to Peer Reviewer Comments
The following sections provide the full comments as received from each reviewer along
with EPA's response where warranted. Small editorial errors present in reviewer's
comments (e.g., misspellings, duplicated words) are corrected in this section; the full,
uncorrected comments from reviewers are provided in the contractor's report which is
an appendix to this document. References to report section numbers in this document
refer to the final report. Full citations to works cited in both this document and the
report are available in the References section of the report.
3.1 Response to Comments Received from Reviewer 1: Steven Barrett
Massachusetts Institute of Technology, Department of Aeronautics &
Astronautics
Summary Assessment
1.	The report represents and in places goes beyond best practice in estimating the
range of maximum likely lead concentrations in air due to piston engine aircraft
considering over 13,000 US airports.
2.	The uncertainty analysis is particularly commendable and provides useful insight
on the likely range of concentrations when accounting for biases and
uncertainties.
3.	The data used is the best available and is appropriately treated in the context of
a data-challenged analysis.
4

-------
4.	The model extrapolation method is a rigorously derived approach that is likely
to yield very reasonable estimates. While further refinements could be made to
account for various factors, these are unlikely to make a material difference and
are likely small compared to the uncertainties captured in the uncertainty
analysis.
5.	Overall this is a comprehensive and high-quality analysis that has been
conducted to the highest standard given the limitations of the data available.
6.	Notwithstanding this, there are uncertainties in the results which are
transparently described and explored both quantitatively and qualitatively.
7.	These uncertainties may be significant at any one specific airport of the 13,000,
but as a national analysis they are likely to be small. As such the overall
conclusions of the analysis are in my judgement robust.
Comments on Sections 1 and 2 (Charge Question 1)
Excerpt from charge question 1: "comment on the extent to which this information is
clearly described and provide your perspective on the approach selected to utilize
modeling from an individual airport in order to characterize concentrations of lead in air
at and downwind of maximum impact areas of airports nationwide."
Reviewer 1 Responses:
8.	Section 1 contains a high-quality summary of the nature of aircraft lead
emissions, including the quantity and technical purpose of leaded aviation
gasoline. Importantly the report also notes that this issue does not apply to jet
fuel - which is used in vastly higher quantities.
9.	The judgement that the run-up location dominates lead concentration maxima is
well justified in my view. This is because aircraft run at high power for an
extended period while not moving, cf. takeoff operations where emissions are
spread out.
10.	In my judgement, EPA are correct in asserting that conducting detailed
monitoring and/or modeling at every one of the 13,000 US airports is not
reasonable, feasible, or necessary given the aims of the work.
5

-------
11.	The overall approach of a detailed assessment of individual representative model
airport is a cogent and well justified method. The notion of an air quality factor
to relate emissions to concentrations is rational and transparent, and is a
reasonable approach to make assessing airports nationwide tractable problem.
12.	EPA make clear that uncertainty and variability characterization is considered in
the work, consistent with best scientific practice.
13.	Section 1 describes the structure of the report, which is logical and clear.
14.	AERMOD is in my judgement a robust and scientifically justified tool for
dispersion modeling of the type conducted by EPA. It contains a detailed and
practically applicable representation of atmospheric dispersion, and is suitable
for application to an airport environment. While it has limitations when applied
to jet aircraft sources, these limitations are not an issue when applied to GA
sources.
15.	The report includes a model evaluation (i.e. relative to data), which is consistent
with the very best practice. The R2 achieved is excellent in the context of
atmospheric dispersion modeling, providing confidence in both the methodology
used to estimate activity and emissions, and dispersion modeling.
16.	The EPA report transparently notes limitations in reproducing modeled values in
section 2.2. In my judgement, however, this level of model performance is
consistent with the best available approaches and demonstrates a level of model
skill that is beyond what I would consider acceptable.
17.	My view is that a 7-day monitoring period is more than sufficient to provide
confidence in the modeling results and a range of meteorological conditions
occurs.
18.	The assertion that the 3-month averaging time limits the importance of day-to-
day variability is correct and further supports the AQF approach in my
assessment.
19.	The approaches used for modeling GA aircraft sources are appropriate and
consistent with best practice.
6

-------
20.	The approach for estimating aircraft activity and emissions is well described and
more than sufficiently detailed.
21.	The altitude cut-off and altitude emissions approach is appropriate. While at-
altitude emissions do impact surface air quality in my estimation, these are not
at all relevant to calculating maximum concentrations as is relevant in this
report.
22.	The surface and upper air meteorological data stations are close enough to be
usefully representative of the meteorology at the airport being modeled. The
distance of the surface air station may introduce some uncertainty, but this is
likely to be small relative to overall atmospheric dispersion and other modeling
uncertainties (except potentially at specific airports).
EPA Response: We have added this point to the qualitative discussion of
meteorological uncertainty in Section 4.4.1.
23.	The receptor placements are logical and more than sufficiently resolved.
Comments on Section 3.1 (Charge Question 2)
Excerpt from charge question 2: "comment on the approach used to calculate AQFs at
the model airport specifically for the purposes of using these factors to estimate
concentrations at and downwind of maximum impact areas of airports nationwide."
Responses:
24.	The approach of calculating different AQFs for different types of operation and
for single vs. multi-engine aircraft is robust and appropriately accounts for the
variability in emissions that is to be expected.
25.	The AQFs are logically and clearly defined (e.g. Eq 1). This provides a transparent
and practically usable way of characterizing maximum 3-month average lead
concentrations given aircraft activity.
26.	The use of 14 months is more than sufficient to capture variability in conditions
and impacts. It is unlikely this approach leads to any over- or under-estimate that
7

-------
is of significance relative to the uncertainties inherent in atmospheric dispersion
modeling.
27.	The specific steps used to calculate AQFs are clearly described and logical.
28.	The report correctly notes that the remaining question is the extent to which the
results apply to airports throughout the country - addressed in section 4.
29.	It would aid clarity if Table 1 used scientific notation (i.e. 1.5xl0~5 rather than
1.5E-5 etc.).
EPA Response: Numbers in Table 1 have been changed to scientific
notation.
Comments on Section 3.2 and Table 2 (Charge Question 3)
Excerpt from charge question 3: "Table 2 and accompanying text in Section 3.2 describe
the methods used to estimate piston- engine aircraft landing and take-offs (LTOs) at
individual runway ends... Are these methods clearly described and do you have
recommended changes to the steps taken? Please explain any alternative options..."
Responses:
30.	The overall approach used to estimate the number of LTOs by piston-engine
aircraft is rational and it is hard to see how it could be improved given the
available data.
31.	The approach is also well-established in NEI use. There will be uncertainties given
the data limitations, but I am not aware of alternative reasonably usable data.
The additional airport-specific data used serves to understand uncertainties
associated with these data limitations.
32.	The approach used is clearly described. In particular, Table 2 clearly and in a well-
structured way describes the approach along with supporting rationale. (Step l's
title should delete the word "Determine".)
EPA Response: "Determine" has been removed from the title to Step 1.
8

-------
33. The 2011 NEI data source is still relevant and appropriate. Although this could
potentially be updated and there may be advantages to that, I do not expect this
would materially affect results.
EPA Response: We agree that the choice of the analysis year is useful to
evaluate. Year-on-year changes in the production ofavgas (a reflection
of national piston-engine aircraft activity) have ranged from a 3%
increase to a 13% decrease during the period from 2011 through 2016
(https://www.eia.gov/dnav/pet/pet pnp refp2 a eppv ypy mbbl a.ht
m). This information supports the conclusion that the choice of a more
recent analysis year would not materially affect results of the estimates of
ranges of lead concentrations at airports nationwide. We have further
addressed this comment in response to Reviewer 1, Comment 50 below.
34.	The use of a national average percentage of GA/AT aircraft that are piston-
engined is appropriate given the limited data available, and enables the use of
per airport LTO data. This would introduce no uncertainty on average, and some
uncertainty per airport. It is unlikely that this uncertainty is significant given the
uncertainties inherent in atmospheric dispersion modeling.
35.	The division into single and multi-engine aircraft is sufficient to capture the
important variability in emissions, along with the division into full LTO and touch
and go operations.
36.	Using daily activity counts from towered airports to extrapolate activity profiles
to other airports is a rational and likely appropriately accurate approach. This
includes the use of the closest towered airport for the untowered airports.
37.	The wind direction runway assignment approach is appropriate and, on average,
is unlikely to introduce significant uncertainty.
38.	The assumption in step 12 that the period of maximum activity is assumed to be
the period of maximum concentration is robust given that we are considering
local passive dispersion modeling. (This assumption could not be transferred to
regional chemistry-transport modeling, for example.)
39.	The avgas Pb scaling approach is robust and will introduce no uncertainty.
9

-------
40. Overall these methods are robust and clearly described. There do not appear to
me to be viable improvements that should or could be made.
Comments on Sections 3.3 (Charge Question 4)
Excerpt from charge question 4: "comment on whether there are alternative airport-
specific data, or analysis approaches, that could improve estimates of piston-engine
aircraft activity at a subset of airports. In addition, please comment on whether
parameters other than piston-engine aircraft activity should be included in analyses to
potentially improve model-extrapolated concentrations at a subset of airports..."
Responses:
41.	The criteria used to select airports for further detailed study appears logical and
clearly described. In particular, the use of 100% instead of the national fraction
of piston-engine AT and GA aircraft is sensible given the potential variability in
these numbers across the nation.
42.	One possible refinement on a per airport basis that could be applied when
evaluating if Pb concentrations come to within 10% of the limit, is to correct for
local or nearest available average wind speed where that is known.
EPA Response: See response to Reviewer 1 comment immediately below.
43. Specifically, the concentration of a passive tracer scales with , where u is
wind speed, and angled brackets imply a time average [e.g. Barrett and Britter
(2008), Development of algorithms and approximations for rapid operational air
quality modelling. Atmospheric Environment 42 (34), pp. 8105-8111. DOI:
10.1016/j.atmosenv.2008.06.020.] If the wind speed at the model airport is v
and at a specific airport is u, then the wind-speed corrected concentration would
be the concentration estimated by the AQF approach multiplied by /.
This would mean that if the wind speed at a specific airport is lower on average
[and so  would be higher] resulting in a higher concentration, this would be
captured.
EPA Response: We have conducted this wind speed correction at all
airports as noted in Step 15 of Table 2; results of this wind speed
correction and the impact on lead concentrations at the maximum impact
site are provided in Section 4.1.
10

-------
44.	This wind speed correction is likely unnecessary in general, but may provide
additional robustness in avoiding missing airports that may breach the NAAQS
limit. One possibility could be to apply a larger (e.g. 50%) margin instead of 10%,
and where the larger margin is reached apply the wind speed correction to
determine if the concentrations approach the NAAQS limit.
EPA Response: See response to Reviewer 1, Comment 47.
45.	Such wind speed corrections could be based on nearby or closest ground
measurements, or from pre-existing WRF modeling output possibly with
appropriate correction for low wind speed conditions.
EPA Response: We used the same ASOS station wind data used to assign
aircraft to specific runway ends. The provenance of the data and the
methodology are described in Section 3.2 and Appendix B.
46.	I would note that I view this refinement as optional as the approach taken by
EPA is scientifically robust. The decision on if to do this is in part a practicality
and resource issue. It would, however, provide additional assurance in the result.
EPA Response: We appreciate the comment and agree that it provides
additional assurance to the estimates of lead concentrations provided in
this report.
47.	Other alternatives to a wind speed correction may also be possible to provide
additional assurance that airports breaching the Pb limit are not being missed.
For example, further rationale and/or discussion of the 10% as used, or a larger
margin.
EPA Response: We have taken additional steps to identify airports with
the potential for lead concentrations at the maximum impact area to be
above the level of the NAAQS for lead. See Table 4 for the full description
of these steps.
48.	The overall approach is in my judgement robust and consistent with best
practice, with one optional potential refinement in terms of wind speed
correction to account partially for location effects.
11

-------
Comments on Sections 4.1 and 4.2 (Charge Question 5)
Excerpt from charge question 5: "comment on the appropriateness of the approaches to
compare model-extrapolated results to monitored concentrations of lead given
available monitoring data. Based on your understanding of the methods presented in
the report and the comparisons of monitor and model-extrapolated concentrations,
please provide your perspective on the performance of the methods in characterizing
the ranges of lead concentrations from piston-engine aircraft at and downwind of US
airports."
Responses:
49.	While there are inconsistencies (as noted by EPA) both in time and space
between monitored data and the modeled data, comparisons are still valuable
for assessing confidence in results.
50.	The difference in the particular year (2011 vs. other years) is likely to introduce
minimal additional error compared to other uncertainties. It may be helpful for
EPA to note the change in GA activity (or some equivalent measure, such as
avgas sales) nationally over a period of time to give general readers assurance
that this is not a significant source of error.
EPA Response: As noted earlier, EPA agrees that the choice of analysis
year is useful to evaluate. Year-on-year changes in the production of
avgas (a reflection of national piston-engine aircraft activity) have ranged
from a 3% increase to a 13% decrease from 2011 through 2016
(https://www.eia.gov/dnav/pet/pet pnp refp2 a eppv ypy mbbl a.ht
m).
Because individual airports may show greater variability than national
totals, we evaluated historic operational data from AT ADS for the top 50
most active GA airports. We described this information regarding year-
on-year variability in Appendix B, and we have referenced the comparison
in Section 3.2. The data shows that, while decadal operational trends may
be significant, the median year-on-year change in operations at the top
50 GA airports ranges from -4% to 4% for a given year, and the
interquartile of airport year-on-year operational changes is between +/-
10% for all years. However, individual airport year-on-year changes range
from -25% to +44%.
12

-------
51.	The specific locations being based on judgement will not, in my assessment,
introduce any significant error.
52.	The approach of providing the monitor location and several model locations
graphically, thus showing the complexity of the environments, is a useful way of
presenting the data. I believe this to be an appropriate approach given the data
available, and is of a very high degree of transparency.
53.	My assessment of the coarse monitor/extrapolated model comparisons is that
the results are close enough to support the approach taken. It should be noted
that the decay rate with distance is significant, and the results are consistent
with this.
54.	The only outlier in terms of monitored vs. measured results is Airport MM.
However, given the mean wind direction it is difficult to compare the monitored
data to extrapolated modeled data. Overall the conclusion (5) that the
extrapolated modeled data reproduces to a very acceptable degree the
monitoring results stands.
EPA Response: We have updated the figure captions of the monitor-to-
model comparison figures and the text of Section 4.1 and 4.2 to reflect
the reviewer's comment that it is difficult to directly compare the
monitored data to extrapolated model data for reasons including wind
direction, monitor location, and differences in data years for the modeled
and monitored data. Further, we acknowledge the reviewer's point that
the extrapolated modeled data generally captures expected
concentrations evident in monitored results at airports nationwide in
accordance with the report's aims; the performance of the model
extrapolation in reproducing monitored results may vary from airport to
airport depending on local considerations (operations, pilot behavior,
active fleet, meteorological conditions, etc.). We expanded our qualitative
uncertainty assessment in Section 4.4 to capture this comment.
Additionally, we learned after the draft report was provided to peer
reviewers that the airport in panel MM conducts the majority of landing
and take-off and therefore run-up checks at a different runway and
therefore the monitor was not sited to capture the maximum impact site;
this figure has been removed from the report.
13

-------
55.	Section 4.2 contains a thorough and useful discussion of the key factors that
result in the variability found. This suggests the reasons are well understood, and
that the alternate criteria for selection (100% of aircraft being piston engined) is
appropriate.
56.	Figure 8 provides a clear depiction of the national and airport-specific results. It
may be helpful to also show on this chart the upper bound used in the criteria to
select airports for further study.
EPA Response: We have taken additional steps to identify airports with
the potential for lead concentrations in the maximum impact area to be
above the level of the NAAQSfor lead (see Table 4). These steps are
described in Section 3.3 and the results included in a chart (Table 7) that
explains which airports met which criteria. Given these changes and the
absence of a single decision metric, we chose not to demarcate a single
upper bound value on the figure.
57.	Overall my assessment is that the performance is fit for purpose and, within the
limitations of the data currently available, represent a best-practice approach in
regulatory modeling.
58.	Additional uncertainties are considered in my response to charge question 6.
14

-------
Comments on Section 4.3 (Charge Question 6)
Excerpt from charge question 6: "Section 4.3 presents quantitative and qualitative
uncertainty analyses of the model- extrapolated results provided in previous sections.
Please provide your perspective on the methods used to conduct these uncertainty and
variability analyses, as well as the key parameters EPA included in the analyses... Provide
your perspective on the application of this analysis to further characterize the range of
lead concentrations attributable to piston aircraft activity at airports nationwide."
Responses:
59.	The inclusion of an uncertainty and variability analysis of this degree of
comprehensiveness and quality is to be commended. It is rare to see such an
analysis done to this level of quality in regulatory (and academic) practice.
60.	Atmospheric stability conditions could also be listed as an uncertainty, along
with local roughness conditions, on page 46 end of second paragraph as part of
parameter 5.
EPA Response: This comment has been addressed by mentioning
atmospheric stability and surface roughness as additional sources of
uncertainty in Section 4.4.
61.	In my view the correct key uncertainty parameter groups have been identified
and treated appropriately. The choices made have been well justified and
explained.
62.	The data used to justify the distribution for run-up times appears robust and to
materially add value to the work. It is unlikely that assuming the airports for
which data is available represent airports more broadly introduces significant
error.
63.	The same comment (4) applies to avgas lead concentrations.
64.	The Monte Carlo approach applied is rigorous and appropriate.
65.	Table 5 provides a clear and transparent description of the uncertain parameters
and their rationale.
15

-------
66.	In the first row of Table 5, Assumptions column - it is likely more accurate to say
that concentrations vary as a power law with distance (as in ~x~s). This does not
materially affect the work.
EPA Response: We revised the text to note that the lead concentration
attributable to run-up decreases as a negative power law with distance
from the maximum impact site.
67.	The work supports the finding that the uncertainty in run-up time is key. Given
the nature of run-up times (being at the discretion of the pilot in command and
being safety critical in nature), it is unlikely more could reasonably be done to
estimate and characterize uncertainty in this factor.
68.	It may be helpful to give an aggregate expected mean (as a percentage) bias due
to the run-up time and avgas lead concentration combined. This could be
compared with the mean under-estimate of measured values.
EPA Response: We have added aggregate statistics on the median and
97.5th percentiles of the analysis for run-up time and avgas lead
concentration to Section 4.3.1 in accordance with the reviewer's
suggestion. However, it is difficult to directly compare the results of the
Monte Carlo assessment in Section 4.3.1 to the model-to-monitor
comparison presented in Section 2.2. The time-in-mode data underlying
the model comparison in Section 2.2 was developed from airport-specific
recorded time-in-mode survey data; the Monte Carlo assessment time-in-
mode distribution is developed from a meta-analysis of time-in-mode
data across different airports and different studies. Thus, the uncertainty
in the national extrapolation is not necessarily representative of the
possible uncertainty or bias in the model airport results.
69.	The comparison in Figure 12 between extrapolated model results (with
uncertainty quantified) and monitored data suggests that the baseline modeling
and uncertainty analysis captures real-world variability.
70.	Section 4.3.2 contains a comprehensive discussion of qualitative factors affecting
uncertainty and variability. As well as mixing height, it should also mention
atmospheric stability.
EPA Response: This comment has been addressed; atmospheric stability is
now mentioned along with mixing height in Section 4.4.
16

-------
71.	It may be helpful to note that rather than the average wind speed, it is the
average of one over wind speed, that drives average concentrations.
EPA Response: We have revised this text to describe the impact of one
over wind speed on the average lead concentrations.
72.	The sensitivity analysis using varying meteorological factors is useful and
provides a clear indication of the potential size of this uncertainty.
73.	Overall the application of the uncertainty approach described in the report is
robust and provides useful additional information about the uncertainty in
model extrapolated values.
74.	The results also makes clear that my suggested wind speed correction is indeed
optional as this is likely not of great significance relative to the run-up time
uncertainty.
17

-------
3.2 Response to Comments Received from Reviewer 2: Michael Kleeman
University of California-Davis, Department of Civil and Environmental
Engineering
Summary:
The purpose of this report is to assess Pb concentrations from aircraft across the United
States. A detailed analysis was carried out for a single representative airport using
measurements and site specific modeling. Dispersion fields from the representative
airport were then extrapolated to the 13,000 US airports using activity data from each
specific airport. Based on this generic exercise, a subset of airports were identified
where predicted concentrations were close to the Pb NAAQS (or exceeded the NAAQS).
Further site specific modeling was conducted at these target airports to more accurately
represent activity data and meteorological conditions.
Comment 1:
The approach summarized above is logical given finite resources, but it would be
relatively simple to make improvements at only minor additional cost. For example, it
seems possible (likely?) that airports with lower average wind speed than the single
model airport used in the current report could have under-predicted Pb concentrations
which might mean that some of these airports were not identified for additional
analysis. A more-accurate pre-screening could have been performed by pre-sorting the
13,000 airports into approximately 5 categories based on average wind speeds and/or
mixing height measured at each location. Detailed modeling could then be conducted
for a representative airport within each category using site specific information. The
dispersion fields developed for each of these 5 categories could then be extrapolated
out to the 13,000 remaining airports by choosing the representative model airport that
most closely represents the actual airport. This suggested improvement would more
accurately capture the approximate wind speed and mixing height at each target
airport. The additional computational expense of this modification would be minor, and
it would lead to an improved screening to identify airports that merit more detailed
modeling.
EPA Response: The approach suggested here for refining the
extrapolated lead concentrations is appropriate and logical, yet
conducting onsite modeling at additional airports is resource
18

-------
intensive (with regard to extrapolating within categories of
facilities; we agree with the reviewer that the computational
expense of the extrapolation itself would be relatively minor).
Based on comments from multiple reviewers, we elected to refine
the methodology for extrapolating lead concentrations to account
for local wind conditions using a scaler approach. We scaled
model-extrapolated concentrations using the relationship between
local pollutant concentration and concurrent average inverse
wind-speed, as suggested by Reviewer 1 Comment 43, and noted
this change in Step 15 of Table 2. The results of this wind speed
correction and the impact on concentrations are provided in
Section 4.1.
Comment 2:
AERMOD is essentially a steady state plume model that estimates pollutant dispersion
based on regional atmospheric conditions measured (or predicted) for the site.
AERMOD does not account for complex air flow around buildings or complex air flow
generated in the propeller wash region of the aircraft. The report uses AERMOD to
predict Pb concentrations at the "maximum impact location immediately adjacent to
the run-up area at a runway end". Predicted concentrations at this location are most
likely not accurate because the effects of propeller wash on atmospheric mixing have
not been accounted for. More complex modeling would be required to accurately
predict concentrations at the maximum impact location. This complex modeling should
either be performed, or more realistically, the concentrations at the maximum impact
location should be removed from the report. This latter option may be preferred since
the maximum impact location is generally not accessible to the public and
concentrations at this site are therefore not a public health issue. Concentrations should
be reported at locations further downwind from the aircraft (25m? or 50m?) where the
assumptions inherent in the model are valid.
EPA Response: As noted in Section 2, EPA conducted novel, proof of
concept modeling at the Santa Monica airport that has since undergone
peer review (Carr et a!., 2011). This modeling specifically incorporates
initial conditions for aircraft exhaust in parameterizing this source in
AERMOD so that maximum impact site concentrations could be
evaluated. This work included the incorporation of propeller wash, which
creates turbulent mixing over the wings of the aircraft and utilized initial
vertical and horizontal mixing using the exhaust temperature and height
19

-------
relevant for a piston-engine aircraft. Additional information on these
parameters is now included in Appendix A of the report. The approach for
parameterizing piston-engine aircraft emissions was evaluated using
model-to-monitor comparisons at the Santa Monica airport. This
modeling framework was then applied to the model airport used in this
report and, as described in Section 2.2 of the report, the model performed
well at a second airport with regard to estimates of lead concentrations
at the maximum impact location.
We have further clarified in the report that the maximum impact
site at the model airport was 15 meters behind the aircraft. At
airports, including those with high traffic volumes, these locations
proximate to piston aircraft exhaust may be in very close proximity
(e.g., within 50 meters) to areas accessible by the general public,
and are therefore relevant for evaluation in this report.
Carr, E., et al. (2011). "Development and evaluation of an air
quality modeling approach to assess near-field impacts of lead
emissions from piston-engine aircraft operating on leaded
aviation gasoline." Atmospheric Environment 45(32): 5795-5804.
Comment 3:
Table 1 - remove the column for 0 m based on the issue raised in Comment 2.
EPA Response: See response to Reviewer 2, Comment 2 above. We
have changed the text describing what we previously labeled as "0
m" in Table 1 to consistently refer to this location as the maximum
impact site. This location was 15 meters behind the run-up
location at the model airport as described in Footnote 5 in the
report.
Table 4 - remove the column for max site based on the issue raised in Comment 2.
EPA Response: See response to Reviewer 2, Comment 2 on Page
19.
20

-------
Figure 5 illustrates LTOs associated with Pb concentrations at the max impact site. This
should be revised to illustrate LTOs associated with Pb concentrations at some other
location where the predicted concentrations are valid. See Comment 2.
EPA Response: See response to Reviewer 2, Comment 2 on Page
19.
Figure 6 illustrates predicted Pb concentrations at the max impact site. This should be
revised to illustrate predicted concentrations at some other location where the model is
valid. See Comment 2.
EPA Response: See response to Reviewer 2, Comment 2 on Page
19.
Figure 10 should remove predicted concentrations at max impact site as discussed in
comment 2.
EPA Response: See response to Reviewer 2, Comment 2 on Page
19.
Figure 11 should plot concentrations at some location other than the max impact site as
discussed in comment 2. Suggest choosing location that is well outside zone where
propeller wash enhances mixing. See comment 2.
EPA Response: See response to Reviewer 2, Comment 2 on Page
19.
Comment 4:
Table 1 - it is somewhat surprising that the ME concentrations are ~4 times greater than
the SE concentrations. Do most ME aircraft have 4 engines? Perhaps this is discussed
elsewhere in the report but a note should be made in this table caption to make it
obvious to the reader.
EPA Response: ME aircraft typically have two engines, although
they can have more than two; the engines used in ME aircraft
have greater fuel consumption due to their larger displacement
(providing greater horsepower than engines typically used in a
single engine aircraft) and typically conduct longer run-up
21

-------
durations. We have included additional information relevant to
this point in Footnote 15.
Comment 5:
Figure 10 caption references shaded blue area but this is not present in the actual
figure. Revise caption to match figure.
EPA Response: This comment has been addressed by revising the
figure caption (now Figure 12).
Comment 6:
Page 40 discusses the comparison of the predicted vs. measured Pb concentrations at
airports where monitoring was performed. The study limits comparisons to locations
where the monitor was proximate to the maximum impact area or downwind of that
area. This seems to be overly restrictive. The model predicts a continuous field that can
be compared to any monitor within a few km of the airport. Model receptor points were
arranged in a regular grid and concentrations at sites between those points can be easily
interpolated. A comparison should be made to all available measurements at all
airports.
EPA Response: The rationale for limiting the comparisons
presented to those where monitors were proximate to the
maximum impact areas is directly related to the analysis and goal
of the estimates being presented in this report. Namely, we are
providing estimates of lead concentrations in the maximum
impact areas, and therefore present available comparisons of the
estimated concentrations with monitored concentrations relevant
to this general location. EPA and others have identified the run-up
location and downwind areas as the maximum impact locations,
due in part to the common attribute of piston-engine aircraft
activity conducting run-up in a designated location at each airport.
Therefore, maximum concentrations in and downwind of the run-
up area is an important commonality among general aviation
airports. Aircraft activity outside this common area differs among
airports depending on hangar locations, startup and idle locations,
taxi-ways, refueling stations and other areas where piston aircraft
operate. Because of this, we do not find it instructive for the
intended purpose here, to compare modeled estimates of lead
22

-------
from the model airport to monitored airports where the monitor
was located distant from the area of maximum impact.
Comment 7:
The uncertainty in the model predictions should be more fully explored through a
comparison between predictions and all available measurements (see Comment 6). All
available monitors should be used for this analysis. The uncertainty derived from these
calculations more accurately represents the uncertainty of the overall modeling
approach than the results of the Monte-Carlo analysis. This uncertainty should be
incorporated into the error bars for Pb concentrations at all reported airports.
EPA Response: Please see response to Reviewer 2, Comment 6
above. We disagree that the model-to-monitor comparisons
provide a more accurate representation of the uncertainty in the
overall modeling approach compared with the Monte-Carlo
analysis. First, there has been very limited monitoring at or near
maximum impact locations at airports, which prevents the type of
analysis suggested. As shown and discussed in the coarse
concentration comparisons in Sections 4.1 and 4.2, monitor
locations vary both in axial and lateral downwind distance from
the maximum impact area. Monitored concentration data and
model-extrapolated concentrations also do not necessarily reflect
the same operational years. Given that the focus of our analysis is
on the maximum impact area and areas downwind (e.g., EPA is
not attempting to estimate lead concentrations at all locations on
or near airports in the US), and these model-to-monitor
comparisons do not account for non-aeronautical sources of lead
and variation in background concentration, we are limiting our
comparisons of uncertainty to the model parameters that have
been demonstrated as being influential at the maximum impact
area. Expanding the number of comparisons by evaluating
monitors distant from the maximum impact location is not
instructive for quantifying concentrations at the maximum impact
location or their associated uncertainty.
Further, as noted in the title of Section 4.3, the Monte Carlo
analysis is solely addressing the quantitative influence of those
parameters that have been demonstrated as being influential in
23

-------
the maximum impact and downwind areas; it is not meant to be
interpreted as capturing all sources of variability and uncertainty.
We have text to Section 4.3, to address the reviewer's comment by
noting our focus in this section on the key parameters that have
been demonstrated in previous studies to impact lead
concentrations at and downwind from the maximum impact area
at airports while recognizing that additional variables and local
considerations may contribute to uncertainty at individual
airports. We point the reader to our analysis in Section 4.4 in
which we have also expanded the qualitative uncertainty
discussion to capture this limitation in response to the reviewer's
comment.
Comment 8:
The Monte-Carlo analysis for the effect of run-up time and Pb concentration in fuel
seems unnecessarily complicated. Each of these parameters is assumed to be linearly
related to ambient concentrations at downwind locations. This has simple and
predictable impact on the concentration variable as described below.
Linear relationships between run-up duration and concentrations at various downwind
distances are summarized in Table C-l with the form concentration=a+b*run-up-time
where a and b are constants. The linear properties of the expectation operator predict
that the variance in the concentration will simply be the variance in the run-up-time
multiplied by b2.
A linear relationship is assumed between fuel Pb concentrations and ambient
concentrations with the form concentration = concentration * Fuel-Pb /Fuel-Pbo. If
concentratono is influenced by the variation in run-up time, then we simply substitute
this into the equation yielding concentration = (a+b*run-uptime) * Fuel-Pb/Fuel-Pbo.
The linear properties of expectation should yield the result that the variance in the
predicted concentration is simply b2 * variance of run-up-time * variance of Fuel-
Pb/Fuel-Pbo.
The simple analysis presented above suggests that the variance of the ambient Pb
concentration in response to variance in run-up time and Fuel-Pb can be easily
calculated without the need for 10,000 iterations of a Monte-Carlo analysis.
24

-------
If the actual distribution of Pb concentrations is needed in addition to the calculated
variance, then the full Monte-Carlo analysis may be warranted. Figure 11 displays the
2.5% and 97.5% concentrations and so perhaps this motivates the analysis. If these
thresholds have some regulatory significance then that information should be described
to the reader in the caption for Figure 11 and/or the associated text discussion. If these
are arbitrary thresholds designed to show the range of concentrations, then perhaps the
standard error (square root of the variance) can be quoted instead at a greatly reduced
computational cost.
Suggest consulting with a statistician to confirm the most efficient approach that is still
accurate.
EPA Response: Monte Carlo analysis is a useful technique for
performing local and global uncertainty analysis, and it can readily
be expanded to accommodate additional uncertain parameters.
Thus, while the identified parameters could potentially more easily
be assessed by techniques like summing uncertainty in
quadrature, the focus of this report is in developing a robust
methodology for understanding lead concentrations at airports
nationwide and associated uncertainty. For that reason, Monte
Carlo is an appropriate technique and consistent with uncertainty
assessment of other aircraft emissions on impact assessment tools
in the literature. Examples from literature include:
Lee, Joosung J., et al. "System for assessing aviation's global
emissions (SAGE), part 2: uncertainty
assessment." Transportation Research Part D: Transport and
Environment 12.6 (2007): 381-395.
Allaire, D., and K. Willcox. "Surrogate Modeling for Uncertainty
Assessment with Application to Aviation Environmental System
Models." Al A A journal 48.8 (2010): 1791-1803.
Simone, Nicholas W., Marc EJ Stettler, and Steven RH Barrett.
"Rapid estimation of global civil aviation emissions with
uncertainty quantification." Transportation Research Part D:
Transport and Environment 25 (2013): 33-41.
25

-------
Comment 9:
Page 49 states that "Available data had an average lead concentration of 1.79 g/gal and
were normally distributed within the range specified for 100LL (i.e., 1.70 to 2.12 g/gal)
(see Appendix C for details on avgas lead data and their distribution)".
Close inspection of Figure C-l shows that the measured Pb-fuel concentration is *not*
normally distributed but rather is bi-modal with a first peak at 1.55 g Pb / gallon and a
second peak at 2.05 g Pb / gallon. This may be an artifact of poorly chosen number of
histogram bins. Recommended number of bins would be approximately the square root
of the sample size N (~10 in this case). The histogram should be replotted to confirm
that it is bimodal.
If the report retains the full Monte-Carlo analysis, the correct distribution for fuel-Pb
should be used. If the report drops the Monte-Carlo analysis in favor of just calculating
the variance of the ambient Pb concentration as described in Comment 10, then no
further action is required.
EPA Response: We acknowledge the reviewer's point that the
binning of the histogram was not optimally chosen, and we have
replotted the histogram of lead concentrations in fuel in
accordance with the reviewer's suggestion. The replotted
histogram, shown below, indeed does not follow a bimodal
distribution, and the text of Appendix C has been updated to
reflect the improved analysis of fuel lead concentrations.
The question of what form and range to select for a "correct
distribution" for 3-month average lead concentration at a given
facility is a difficult one. Because an airport may be serviced by
multiple fuel deliveries over a three-month period; because
aircraft landing and taking off at a given airport may have been
fueled at a different facility; and because the avgas lead
concentration sample data contained noticeable outliers and
lacked temporal and spatial resolution, we determined that using
the full distribution of avgas lead concentration samples was not
appropriate for quantifying uncertainty in three-month average
lead concentrations at airports. We use both the central limit
theorem and the ASTM fuel specifications to guide the choice of
avgas lead concentration distributions used in the Monte Carlo
analysis.
26

-------
Distribution of Avgas Lead Concentrations
35
cy cy O" n" "V n* rv v "v rv*
V V 
-------
intended to constitute a determination by EPA. As noted in the
Federal Lead Action Plan to Reduce Childhood Lead Exposures and
Associated Health Impacts (https://www.epa.gov/lead/federal-
action-plan-reduce-childhood-lead-exposure), EPA is evaluating
aircraft lead emissions and their impact on air quality because this
source is currently the dominant contributor to air-related lead
emissions in the US. EPA's activities regarding aircraft lead
emissions can be found at the following website:
https://www. epa. aov/reaulations-emissions-vehicles-and-
enaines/reaulations-lead-emissions-aircraft, and the Federal
Aviation Administration activities regarding the evaluation of
unleaded fuel options can be found here:
https://www.faa.aov/about/initiatives/avaas/.
The model-extrapolated estimates of lead concentrations provided
in this report cannot be used in making determinations regarding
violations of the NAAQSfor lead; EPA relies on the lead
surveillance monitoring network for such determinations with
regard to lead. EPA's guidance on this matter is provided in the
National Ambient Air Quality Standards for Lead Final Rule
(http://www.apo.gov/fdsvs/pka/FR-2008-ll-12/pdf/E8-
25654.pdf).
Comment 11:
Suggest adding the following paragraph (or similar) to the Summary section of the
report. A similar statement defining the reasonable scope of the report and proper
interpretation of the results should also be included in the introduction.
"The model predictions described in the current report should be viewed as a
screening tool to assess the need for additional measurements of ambient Pb
concentrations at airports. The calculated results provide a ranking of the
locations where measurements may be most useful to determine if ambient Pb
concentrations violate NAAQS levels. The actual maximum concentration values
described in this report represent reasonable estimates for ambient Pb
concentrations, but they should be verified with measurements before any
determination of a NAAQS violation is made."
28

-------
EPA Response: We agree that additional clarity on this point is
needed; we revised the Summary and the Introduction clarifying
the purpose of the report and noting that our model-extrapolated
values cannot be used to determine compliance with the lead
NAAQS.
Minor Comments
Page 34: Sentence reading "Data from previous EPA studies at six airports showed
agreement (within 10%) between the number of SE and ME aircraft based at an airport
and onsite observations of piston engine aircraft activity the airport (see Appendix B for
study details)." should have "at" inserted to read
"Data from previous EPA studies at six airports showed agreement (within 10%)
between the number of SE and ME aircraft based at an airport and onsite
observations of piston-engine aircraft activity at the airport (see Appendix B for
study details)."
EPA Response: This comment has been addressed.
Page 47: Sentence reading "Run-up emissions accounted for 82% of the 3-month
average lead concentration attributable to piston-engine aircraft in EPA air quality
modeling at a model facility, and was a primary contributor to emissions in modeling
conducted by Feinberg et al. (Section 2, Appendix A){Feinberg, 2016 #11}." Appears to
have a reference that was not properly formatted.
EPA Response: This comment has been addressed.
Page 56: Sentence reading "As noted in Section 3.1, the mean of the twelve 3-month
average AQFs from the model airport was used to calculate model-extrapolated
concentrations; this average was used in order capture the influence on lead
concentrations from the full range of wind speeds, mixing heights, and other
meteorological parameters that occurred at the model airport." should have "to"
inserted to read
"As noted in Section 3.1, the mean of the twelve 3-month average AQFs from
the model airport was used to calculate model-extrapolated concentrations; this
average was used in order to capture the influence on lead concentrations from
the full range of wind speeds, mixing heights, and other meteorological
parameters that occurred at the model airport."
EPA Response: This comment has been addressed.
29

-------
3.3 Response to Comments Received from Reviewer 3: Barbara Morin
Rhode Island Department of Environmental Management
General Comments-
In general, I am comfortable with the methods used by [EPA] to evaluate impacts at
the model airport. I am, however, concerned about the uncertainties associated with
extrapolating those results to other airports, as reflected in my comments below. I
believe that this analysis appropriately addresses EPA's objective of providing "an
understanding of the potential range in lead concentrations in air at the approximately
13,000 airports with piston-engine aircraft activity in the US." However, if modeling
results are to be used, alone or in conjunction with monitoring results, to demonstrate
compliance with the NAAQS or to evaluate the potential for site-specific exposures, it
is my hope that those analyses will utilize airport-specific information rather than
relying on the AQFs derived in this study.
EPA Response: We appreciate the Reviewer's feedback that the analysis
included in the report appropriately addresses the objective described in
the report. We incorporated additional text in the Introduction to the
Report to further emphasize that in making determinations regarding
violations of the NAAQS for lead, EPA relies solely on the lead surveillance
monitoring network. EPA's guidance on this matter is provided in the
National Ambient Air Quality Standards for Lead Final Rule
(http://www. qpo. aov/fds vs/oka/FR-2008-11 -12/pdf/E8-25654.pdf).
Charge Questions and Responses-
1. Sections 1 and 2 describe the nature of how piston-engine aircraft operate for safety
and logistical reasons, along with the previous work that EPA and others have
conducted to characterize concentrations of lead in air at individual airports. As stated
in the report, conducting detailed air quality modeling or monitoring at all US
airports is not feasible due to resource constraints. Please comment on the extent to
which this information is clearly described and provide your perspective on the
approach selected to utilize modeling from an individual airport in order to
characterize concentrations of lead in air at and downwind of maximum impact areas
of airports nationwide.
While the information presented in Section 1 and 2 is clear and useful, short (one
or two sentence) explanations addressing the following topics would further aid in
the understandability of this material:
30

-------
The document explains the function of lead in aviation gasoline. Since lead
previously served a similar function in automobile gasoline and has been
banned from that fuel for more than 30 years, a sentence about why removing
lead from aviation gasoline has not yet been considered feasible would be
helpful.
EPA Response: We have added a footnote that unleaded motor vehicle
fuel cannot generally be used in piston-engine aircraft because of the
minimum octane requirements as well as other carefully controlled fuel
parameters in avgas.
It is relevant to note here that the general aviation industry and fuel
providers, together with the Federal Aviation Administration are currently
engaged in a multi-year program to identify and evaluate unleaded fuels
for use in piston-engine aircraft (information available at the following
link: http://www.faa.eov/about/initiatives/aveas/).
It took me a little while to reconcile the statement that "Run-up activity is
estimated to contribute over 80% of the lead concentrations at and downwind of the
area where the runup mode of operation occurs" with the emissions breakdown by
operation type in Table A-l, which associates 36% of ME LTO emissions and 15%
of SE LTO emissions with run-up operations and Figure A-9, which shows fuel
consumption rates during run-up to be similar to those during approach and less
than those during take-off and climb modes. I assume that this is because the
aircraft is at ground level and stationary during run-up, unlike during taxiing and
take-off operations, as well as the fact that run-up operations take place near the
end of the runway, but further explanation would be helpful.
EPA Response: We have added text to this section to clarify the
contribution of run-up emissions to lead concentrations at and
immediately downwind of the run-up location. The text we added
communicates that run-up operations are typically conducted adjacent to
the runway end from which aircraft take-off and the brakes are engaged
so the aircraft is stationary. As a result of the stationary aircraft, duration
of run-up, and high fuel consumption rate, emissions from run-up activity
are the largest contributor to local maximum atmospheric lead
concentrations; run-up emissions are estimated to contribute over 80% of
the lead concentrations at and immediately downwind of the area where
the run-up mode of operation occurs, even though this mode of operation
does not have the highest fuel consumption rate. We refer the reader to
Appendix A of the report for more intormation.
31

-------
A sentence about why RHV was chosen as the model airport would be useful.
EPA Response: EPA selected RHV as the model airport because it was
representative of general aviation airports where piston-engine aircraft
operate. Additionally, this airport allowed EPA to evaluate the use of
AERMOD in a more complex airport setting (e.g., parallel runways)
compared with the earlier proof of concept modeling EPA conducted at
SMO. We have added a sentence to the report noting this information.
Is longer-term monitoring being conducted at RHV to evaluate annual impact
predictions?
EPA Response: Lead monitoring at the RHV airport is required to continue
per requirements stipulated in the NAAQSfor lead (lead concentrations
measured above half the NAAQS necessitate continued monitoring).
These data could be used to understand year-to-year changes in lead
concentrations at this facility.
Although I understand that the short-term model-monitor comparisons were
considered to be within the acceptable bounds, the fact that the model under-
predicted the monitored values on 6 of the 7 days at the maximum impact site
and on all 7 days at the downwind site is not reassuring. Was any consideration
given to adjusting model results to account for this under-prediction? An
explanation of this decision would be helpful.
EPA Response: We acknowledge that our model airport evaluation
suggests the modeled concentrations are somewhat lower than the
monitored values at the model airport. As stated in the report and noted
in this comment, the difference we observed between modeled and
monitored concentrations is within the commonly held acceptable
bounds; moreover, when the modeling approach was applied elsewhere,
results showed under- and overestimates (Carr et a!., 2011; Fein berg et
a!., 2016). One could conduct a sensitivity analysis to evaluate the impact
of adjusting model results for under-prediction (e.g., apply corrections
based on the seven days of model-to-monitor comparison at the model
airport). We did not conduct such a sensitivity analysis, in part because it
would presume that the difference between modeled and monitored
concentrations quantitatively captures all of the relevant uncertainties
and is consistently, directionally correct. In addition, this type of
sensitivity analysis does not account for any uncertainty in monitoring
results, which may contribute, along with model uncertainty or bias, to
32

-------
differences modeled versus monitored concentrations. Rather than adjust
modeling results at the model airport a priori, we utilized a series of
uncertainty analyses to evaluate how key parameters may impact model-
extrapolated concentrations at airports nationwide. The specific
uncertainty analyses were selected based on observations at several
airports which identified key parameters that impact modeled
concentrations of lead from piston-engine aircraft activity. Each
uncertainty or sensitivity analysis is described in detail in Sections 3 and 4,
with supporting information available in Appendix C.
Carr, E., et al. (2011). "Development and evaluation of an air
quality modeling approach to assess near-field impacts of lead
emissions from piston-engine aircraft operating on leaded
aviation gasoline." Atmospheric Environment 45(32): 5795-5804.
The concept of aircraft emissions as volume sources (page 10) is
counterintuitive, since emissions are from the tailpipe. This topic is explained
further in A. 1.5, but it would be helpful to either add a sentence to the
introduction to address that issue or to add a reference to A. 1.5 in the
introduction.
EPA Response: We added a reference to Appendix A, Section A.1.5.
 Page 10 states that "wind direction data were used to identify the runway end
from which piston-engine aircraft took off during each hour of each day in the
year of modeling." To facilitate analysis of monitoring data around TF Green
Airport, RIAC, the operator of that airport, provides RIDEM/RIDOH with data
on the time, aircraft and runway for each LTO. Are similar data available for
RHV or for any of the other airports evaluated? If so, was there any attempt to
use such data, even if incomplete, to verify the runway assumptions based on
wind direction?
EPA Response: EPA or EPA contractors visited ten general aviation
airports (RHV, SMO, DAB, ACK, CRQ, SQL, DAB, MRI, PTK, and VNY) to
visually verify the use of wind direction to evaluate the active runway
used. We do not have runway-specific identification of activity for 2011
from these facilities (other than RHV) to compare with the runway activity
estimated in this analysis. We conducted several sensitivity analyses to
evaluate areas of uncertainty such as runway assignment during the peak
33

-------
period of piston-engine aircraft activity. These additional analyses are
described in Section 3.3 with results presented in Section 4.2.
2. Section 3.1 presents the methods to calculate Air Quality Factors (AQFs) at the
model airport. Please comment on the approach used to calculate AQFs at the model
airport specifically for the purposes of using these factors to estimate concentrations
at and downwind of maximum impact areas of airports nationwide.
I would appreciate a sentence of phrase explaining the purpose of touch and go
operations. I assume that these are training exercises.
EPA Response: We have added additional text to Footnote 3 on touch-
and-go operations to explain that they are part of pilot training.
 Page 12 characterizes the maximum impact site as "the runway end at which
LTOs most frequently occurred at the model airport facility." The document
says earlier that 80% of the lead concentration at the maximum impact site is
from run-up operations. I understand that run-up operations are generally
conducted at the runway end that is being used for take-offs. However, since
run-ups are not associated with landing operations, I assume that the number of
take-off operations, rather than the number of landing and take-off operations,
would be the relevant factor for determining the maximum impact location.
This may make a difference at some airports, if the diurnal patterns for landing
and take-off operations differ.
EPA Response: While run-up is the main contributor to lead
concentrations at the maximum impact site, both landings and takeoffs
contribute to lead concentrations at and downwind of the maximum
impact location. Since, over a 3-month period, each takeoff must also be
associated with a landing, and airports do not identify operations as take-
off vs. landing, but just'operations', it is appropriate to model each
landing and takeoff as a linked pair. While for each particular landing and
takeoff pair, the takeoff must precede the landing, because general
aviation operations are typically of short duration, we do not expect there
to be a significant difference between the landing profile and takeoff
profile across the day other than at the margins. Given that it is
significantly less computationally intensive to run the extrapolation model
for each LTO cycle as a pair and because we have limited evidence of any
significant difference between the landing and takeoff profiles across the
day, we model the two as a single operational unit (an LTO or a T&G). We
have added to Appendix B a short discussion of the landing and takeoff
profiles, showing that the landing and takeoff profiles at six airports are
34

-------
not appreciably and consistently different across days where aircraft
takeoffs and landings were surveyed. Further, we discuss that given
evidence from modeling, literature, and surveys of aircraft operations, the
3-month average lead concentrations are not expected to be sensitive to
diurnal profile uncertainty. We show that using a generic diurnal profile is
expected to contribute only 2% uncertainty to lead concentrations, based
on a comparison of profiles at airports where both landing and takeoff
survey data exist. However, we acknowledge that diurnal profile may
contribute additional uncertainty at individual airports if there are specific
local operational patterns that would significantly separate take-off
operations from landing operations diurnally. A discussion of diurnal
profile uncertainty has also been added to Section 4.4.3 with additional
details in Appendix B.
I am also a little confused that there is no distinction between landing and take-
off in calculation of the AQF ratio (i.e. that the equation used to calculate AQF
uses LTOs), since emissions during landing and take-off operations are very
different (and take-offs also involve run-up operations). Is the assumption that
every landing is associated with a take-off? Does the fact that those activities
may take place on different runways (e.g. because of diurnal wind variations)
impact the accuracy of this calculation?
EPA Response: As noted above, EPA does make the logical assumption
that each takeoff is associated with a landing at a given airport. While
some airports may have a significant fraction of itinerant flying (i.e., from
one airport to another), we make the simplifying assumption that over a
three-month period, modeling each take-off as being associated with a
landing is appropriate.
It is not expected that the diurnal profile in take-offs and landings is a
significant parameter that would impact rolling three-month average
lead concentrations at most airports. Previous modeling of aircraft lead
concentrations found that monthly atmospheric lead concentration
estimates showed relatively small variations based on a wide range of
input diurnal profiles because any hourly differences (from, for example,
wind speed) average out over longer periods [Feinberg and Turner, 2013].
Extending the averaging period from one month to three months should
further reduce modeled concentration sensitivity to diurnal profile.
To further understand sensitivity to diurnal profile, EPA examined the
impact of using a generic "operational" diurnal profile vs. a "landing" or
35

-------
"take-off" specific diurnal profile as suggested by the reviewer's
comment. For this choice of diurnal profile to be impactful on three-
month averaged concentrations, two factors would need to occur: the
difference between the diurnal profile of landings and the diurnal profile
of take-off would need to be significantly different, and the average wind
direction at the time of a potential overestimate of take-offs would need
to be significantly different than the average wind direction at the time of
a potential under-estimate of take-offs.
EPA examined the landing and take-off patterns at RHV (the model
airport) and at five airports for which airport survey data was available.
Given that piston aircraft do not typically operate at night and that an
aircraft must first take-off for it to land, there is an expectation that take-
offs will (on average) occur earlier than landings. However, piston-engine
aircraft typically perform short operational missions. Thus, while at the
margins landings should occur later than takeoffs, we do not expect the
profile of landings and takeoffs to differ significantly.
Airport surveys at each airport reported counts of landing and take-offs
during operating hours or a subset of hours for between three and six
days of operation. Operational survey data were excluded for any day
that did not have survey data covering at least 80% of operational hours
or for any days where both landing and take-off data were not available.
The figure below shows the difference in percentage points of the landing
and take-off diurnal profiles at each of these airports as reported in
survey data. The data confirms the expectation that take-offs were
relatively more prevalent than landings over the first hour of monitored
operations (and, consequently, in the last hour of operations, take-offs
were relatively less prevalent than landings), but that profiles were
otherwise similar over the day. The figure shows that, on average the
difference between a landing diurnal profile and a takeoff diurnal profile
is 2.6 percentage points. Further, 95% of examined hours show a
difference of less than 6 percentage points between the landing diurnal
profile and the take-off diurnal profile. Thus, using a generic "operation
(LTO)" profile will, on average, over or under predict takeoffs by 1.3% in
any given hour.
36

-------
25
20
15
10
5
0
-5
-10
-15
-20
-25
RHV
MRI
CRQ
DCU
VNY
DVT
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour
Figure 1: Difference between landing diurnal profile and takeoff diurnal
profile from 3 to 6 days of survey data at each of 6 airports.
In our analysis of individual airports, we consider +/-10% maximum 3-
month concentrations (in addition to considering variation in other more
sensitive parameters such as the expected split between multi- and single-
engine aircraft), which far exceeds the +/-1.3% uncertainty from
differences in landing and take-off diurnal profile.
Further, while the average difference between a generic operation diurnal
profile and a take-off only diurnal profile is 1.3%, the actual uncertainty in
resulting 3-month average concentration may be even smaller. Since
aircraft are assigned to runway primarily based on wind direction, a
modeled difference in operations would require the wind direction to
change significantly from the time in which takeoffs may potentially be
overestimated to the time in which they may potentially be
underestimated.
The figure below shows the average wind direction at 938 ASOS stations
nationwide for each hour of the day. Wind direction is normalized such
that the wind direction at 00:00 is 0 at all stations. Each ASOS station is
plotted along a circle of different unit radius, and each hour is modeled by
a dot where angle represents difference in wind direction from 00:00 and
color represents time of day. The figure shows that across all ASOS
stations, 86% of all hours have average wind direction that fall within 90
of the initial recorded wind direction. Therefore, for example, at a single
runway airport, even if the diurnal profile were moving 2% of operations
37

-------
from the morning to the afternoon, there is an expectation that, averaged
over a three-month period, those operations would still generally he
assigned to the same runway end.
Figure 2: Normalized average wind direction by hour of the day at 938
ASOS stations.
Given the evidence that 3-month average concentrations are not
highly sensitive to diurnal profile, the selection of the RHV diurnal
profile is appropriate for the national analysis of lead concentrations.
To better explain and document the modeling of diurnal profile, we
have added the above discussion to Appendix B to support the existing
discussion of diurnal profile. We have referenced this discussion and
the associated uncertainty from using a single diurnal profile in an
expanded section on Operational Parameter uncertainty (Section
4.4.3).
Feinberg, Stephen, arid Jay Turner. "Dispersion Modeling of Lead
Emissions from Piston Engine Aircraft at General Aviation
38

-------
Facilities." Transportation Research Record: Journal of the
Transportation Research Board 2325 (2013): 34-42.
The document states that "AQFs are calculated as the ratio of the average lead
concentration over rolling 3-month time periods to piston-engine aircraft LTOs
at the most frequently used runway end over the same 3-month period." Are
there cases where a receptor is impacted by LTOs from more than one runway?
This is the case at TF Green. While this may not have a significant impact at the
maximum receptor, it can be a factor in calculating three-month average
concentrations at downwind receptors, so focusing only on one runway may
underestimate those impacts.
EPA Response: EPA agrees that the approach described in this report
would underestimate impacts for cases where the maximum impact site
is impacted by piston-engine aircraft activity in addition to that
characterized at the model airport (i.e., multiple runways, additional
taxi and idle locations). An airport-specific assessment would be needed
to comprehensively evaluate lead emissions and concentrations for
complex airports in which a range of aircraft activities could be
influencing lead concentrations downwind from the maximum impact
area.
The document states that "In order to average across the largest range in
meteorology inputs to AQFs (e.g., wind speed), the resulting 12 AQFs were
averaged to provide a single 3-month AQF for each aircraft class, operation-
type, and location combination." Since the NAAQS is a not-to exceed,
maximum 3-month concentration, why would you average the 3-month AQFs
calculated? Due to seasonal variations in meteorology and emissions, the 3-
month AQFs calculated for some 3-month time periods would legitimately be
higher than for other times of the year. By averaging the 12 values, you are
losing that range, and potentially underestimating 3-month averages that may
exceed the NAAQS.
EPA Response: We acknowledge that the use of the maximum AQF
from the model airport would provide higher estimates of extrapolated
lead concentrations; as noted in the report, the maximum weighted
AQF is 23% higher than the average weighted AQF. To achieve our goal
of using the model airport to best characterize lead concentrations at
airports nationwide and in acknowledgement of the importance of
meteorology on lead concentrations, we used the average AQF and
then wind-speed corrected the estimated concentrations as a means to
39

-------
further refine and make comparisons with the lead NAAQS (the wind-
speed correction is a new analysis added in response to peer review
comments we received).
3. Table 2 and accompanying text in Section 3.2 describe the methods used to estimate
piston- engine aircraft landing and take-offs (LTOs) at individual runway ends on a
rolling 3-month basis (e.g., apportioning out piston-engine-specific LTOs from total
LTOs at each airport, allocating annual activity to daily and then hourly periods). Are
these methods clearly described and do you have recommended changes to the steps
taken? Please explain any alternative options and provide the location of data sources
that would support such alternative options.
The methods are clearly described. However, while I understand the need to
generalize when extrapolating to a large number of other airports, I question whether
there is so much uncertainty associated with those extrapolations that the predictions
are not meaningful. For instance:
EPA Response: We address each of the Reviewer's specific points below,
but address the general concern regarding uncertainty in the
extrapolated concentrations here. As noted in the introduction statement,
the methods used are intended to provide results that are informative for
understanding ranges of lead concentrations in air at airports nationwide.
As such, we provide a detailed analysis of uncertainty and variability in
key parameters in Section 4 of the report and where individual airports
are evaluated, we utilize airport-specific data to the fullest extent
possible, while maintaining a quantitative uncertainty analysis of lead
concentrations at these facilities as well. We recognize the reviewers
concern that use of these model-extrapolated concentrations outside the
aims if this report may be less meaningful. We have explicitly added a
statement that these results are not to be used to determine NAAQS
attainment status in the Introduction and have revised the report to be
more specific as to the interpretation of the results.
Appendix B says "A comparison of the diurnal profiles across these four
facilities (airports) shows the same basic features: a ramp-up of activity in early
morning, peaks in activity in late morning and early afternoon, and decreasing
operations in the evening." However, the diurnal patterns for RHV shown in
Figure B-2 appear to be distinctively different from those for the other airports.
At RHV, the volume of operations remains high through most of the day,
peaking at 4:00 PM on weekdays, while, at the other airports, activity peaks in
the morning and drops off in the afternoon. This is significant because
atmospheric dispersion characteristics tend to be much less favorable in the
40

-------
early morning than at midday or in the afternoon, so emissions from morning
flights could have significantly higher impacts than emissions that occur later in
the day. As an example of this phenomenon, measured pollutant levels near
highways tend to be significantly higher during morning rush hour than during
evening rush hour, despite similar traffic volumes and congestion levels. We see
similar effects at monitors near TF Green Airport.
EPA Response: We agree with the reviewer that increased atmospheric
stability (lower mixing height) in the morning hours generally leads to
greater concentrations from ground-based emissions compared with
afternoon hours. Wind speed is also strongly related to mixing height, as
shown in Figure 14, with a correlation coefficient of 0.98 at the model
airport; we have conducted a new analysis of the impact of wind speed on
lead concentrations which refines the extrapolated estimates of lead
concentration (See Section 3.2 and Step 15 of Table 2). With regard to
the impact of the diurnal profile in aircraft activity on the extrapolated
concentrations, it is useful to point out that while one can observe
potential differences between airport diurnal profiles presented in Figure
B-2, the diurnal profiles across each of the airports are broadly consistent.
For example, while surveys at RHV show 4.5% more operations at 16:00
relative to RVS (normalizing for total operations), the surveys also show
RVS having 4.5% more operations at 18:00 relative to RHV. Further, as
discussed in the response to comments above, while hourly
concentrations may be sensitive to operational profiles, long-term
average concentrations (such as the 3-month average concentrations
developed here) are not expected to be sensitive to diurnal profile. We
added discussion to Appendix B and Section 4.4.3 to further discuss
uncertainty contributions of diurnal profile modeling choices. Lastly, it is
relevant to note that differences in the diurnal profile across the four
airports for which extended operational survey data are available may be
driven by local or regional operational patterns, seasonal differences, or
an artifact of survey length and methodology, which, when taken
together, further suggest that the assumption that these activity profiles
are broadly similar is a reasonable one given available evidence.
As mentioned above, impact calculations were done only for the most used
runway end. There are several reasons why this may underestimate impacts.
First, some downwind sites may be impacted by LTOs from more than one
runway end, so discounting all flights except those that on the most-used
runway may underestimate 3-month average exposures at those locations. In
addition, the number of flights isn't the only determinant of impact. If wind
41

-------
speeds associated with one wind direction are stronger than those associated
with another direction (not a far-fetched idea), impacts associated with a
runway end with lighter average wind speeds may be higher than impacts from
a somewhat more well-used runway end that is associated with higher wind
speeds. For evaluating impacts on ambient air, as would be done to determine
compliance with the NAAQS, the proximity of the runway end to the property
line would also be an important factor.
EPA Response: We acknowledge that the methods used in this
assessment do not include potential impacts at specific airport facilities
where activity at multiple runways or taxiways may increase lead
concentrations at downwind receptors. We also acknowledge the
important impact of wind speed on ambient lead concentrations, and
have incorporated a new analysis to refine the extrapolated lead
concentrations by correcting for wind speed at each individual airport
during the period of maximum activity. For the analysis of specific
airports where estimated lead concentrations were above the level of
the lead NAAQS, we used visual inspection of Google Earth images and
on-site inspections to evaluate whether the area within 50 meters of the
maximum impact site has unrestricted access. As noted in responses to
comments above, and in the report, EPA relies solely on the lead
surveillance monitoring network for lead NAAQS attainment
determinations. EPA's guidance on this matter is provided in the
National Ambient Air Quality Standards for Lead Final Rule
(http://www.qpo.gov/fdsys/Dkg/FR-2008-ll-12/pdf/E8-25654.Ddf).
 Did the modeling take into account diurnal air traffic patterns? Due to the
differences in dispersion characteristics at different times of day discussed
above, diurnal patterns may make a significant difference in impacts.
EPA Response: Air quality modeling at the model airport did incorporate
the diurnal profile in air traffic, as well as hourly meteorology.
According to Figure B-l, the ME full LTO fraction at RHV peaks in the late
afternoon (4:00 PM on weekdays and 5:00 PM on weekends), but that peak is
not seen with SE full LTOs. This distinction may be important because of the
approximately 10X greater impacts of ME than SE planes, as shown in Table 1.
Was this diurnal difference used in the modeling, or just to select the runway
end with the highest total piston engine traffic?
42

-------
EPA Response: EPA incorporated this diurnal profile difference between
ME and SE activity in the air quality modeling (see Appendix A for details
on air quality modeling at the model airport).
Again, AQFs were generated by averaging the 12 values calculated. Since the
NAAQS is a not-to -exceed value and the activity for the most active runway
for the most active 3-month period is used in the analysis, shouldn't the analysis
also use the AQFs that correspond to that period, in order to take into account
seasonal variations in meteorological conditions?
EPA Response: We have responded to this comment above.
4.	Section 3.3 presents an analysis to refine estimates of piston-engine aircraft activity
using airport-specific data for a subset of airports. Please comment on whether there
are alternative airport-specific data, or analysis approaches, that could improve
estimates of piston-engine aircraft activity at a subset of airports. In addition, please
comment on whether parameters other than piston-engine aircraft activity should be
included in analyses to potentially improve model-extrapolated concentrations at a
subset of airports, noting that additional parameters are evaluated at all airports in the
uncertainty and variability analyses presented in Section 4.
The refined analyses used more airport-specific factors for fraction of piston planes
and for breakdown between SE and ME. I would like to see modeling done at a
subset of the airports to look at the impacts of some of the other factors discussed
above (e.g. diurnal variation).
EPA Response: EPA acknowledges that to take steps beyond extrapolated
estimates, conducting airport-specific air quality modeling is an approach
that would provide a robust assessment of lead concentrations
attributable to aircraft emissions. Additional modeling is outside the
scope of this effort; however, we have identified the key airport-specific
parameters that would need to be incorporated in such modeling if
further assessment of specific airports is conducted.
5.	EPA provides coarse comparisons of monitored lead concentrations to model-
extrapolation results from the national and airport-specific analyses, in Sections 4.1
and 4.2, respectively. In Section 4.3, EPA provides a more detailed comparison of
data from lead monitors placed in close proximity to the locations of model-
extrapolated concentrations. Please comment on the appropriateness of the
approaches to compare model-extrapolated results to monitored concentrations of
lead given available monitoring data. Based on your understanding of the methods
presented in the report and the comparisons of monitor and model-extrapolated
concentrations, please provide your perspective on the performance of the methods in
43

-------
characterizing the ranges of lead concentrations from piston-engine aircraft at and
downwind of US airports.
I'm not clear what point is being made by the discussion in 4.1 about the model
showing decreased concentration with distance and with lower levels of piston-
engine activity. Wouldn't that always be the case for a ground-level volume source?
If the modeling included multiple sources or an elevated stationary source, maximum
impacts may occur at some distance from the highest source. However, in this case,
only aircraft using the busiest runway end were modeled and both the run-up and
take-off operations take place at that end. Similarly, it also seems obvious that the
per plane modeled impacts from ME planes would be higher than for SE planes,
since the emissions are substantially higher.
EPA Response: As the Reviewer points out, the model-extrapolation
results conform to expectations regarding pollutant gradients and the
relative contribution of SE versus ME emissions rates. We point to these
characteristics of the results simply to confirm for readers that the data
are in line with these basic measures, particularly for readers who are
less familiar aircraft emissions sources and modeling.
I'm not sure what conclusion to draw from the monitor to model comparison in
Figure 7. In most cases, the comparisons look reasonable, but, from the limited data
presented, it is impossible to see the shape of the variability of concentration. Would
it be possible to model with a receptor at the location of the monitors?
EPA Response: EPA understands the interest in seeing the gradient in lead
concentrations from extrapolated values at the exact monitoring location.
Resource limitations prohibit our ability to run AERMOD at the eight
airports shown in these figures (now Figures 9 and 11). The goal in
comparing the model-extrapolated lead concentrations with monitored
values is to provide information on which the reader can draw a general
understanding of the reasonableness of the extrapolated lead
concentrations estimates at, and downwind from the area of maximum
impact (also, as noted in Section 4.3.2, there are many inherent
differences in time, space and methods between these concentrations
that need to be taken into account). Also see responses to Comment 6
from Reviewer 2.
The monitor-model comparability using airport-specific data in Figure 9 appears to be
considerably better than the comparability using the national estimates in Figure 7.
Again, it would be interesting to model a receptor at the location of the monitor. The
fact that the extrapolated modeled results in Panel A of Figure 12 are considerably
44

-------
lower than, and do not appear to be well correlated with, the monitored values is not
reassuring.
EPA Response: We acknowledge the Reviewer's point regarding the
relationship between model-extrapolated and monitored concentrations
in Panel A of Figure 12; we note in the report that these model-
extrapolated lead concentrations were 12% to 52% lower than the
primary monitor concentrations. We have conducted extensive analysis
to describe the variability and uncertainty in the extrapolated estimates
and it is instructive to note that the lead concentrations at the primary
monitor all fall within the quantitative uncertainty bounds provided by
evaluating variability in run-up time and avgas lead concentration. EPA is
providing the data in Figure 12 to illustrate the ability of the extrapolation
method to appropriately identify airports with the potential for lead
concentrations to be above the level of the lead NAAQS. To that purpose,
the figure presents several rolling 3-month average lead concentration
values at specific airports where monitored lead concentrations have
consistently violated the NAAQS (panel A) and been consistently below
the NAAQS (panel B).
6. Section 4.3 presents quantitative and qualitative uncertainty analyses of the model-
extrapolated results provided in previous sections. Please provide your perspective on
the methods used to conduct these uncertainty and variability analyses, as well as the
key parameters EPA included in the analyses (based on previous work discussed in
Section 2). Please provide your perspective on the application of this analysis to
further characterize the range of lead concentrations attributable to piston aircraft
activity at airports nationwide.
Figure 10 is confusing - the caption doesn't match the figure.
EPA Response: This comment has been addressed.
While the methods used to conduct the uncertainty analysis appear to be
reasonable, the results are not convincing. In particular, it appears that run-up time
durations have such a dramatic effect on impacts that calculations of impacts
using the AQFs would not provide useful estimates of impacts at other airports
unless they are adjusted to account for differences in that parameter. Note that the
document states that "run-up emissions accounted for 82% of the 3-month average
lead concentration attributable to piston-engine aircraft in EPA air quality
45

-------
modeling", that the "variation between the 5th and 95th percentiles of average
run-up times observed in EPA modeling resulted in an almost 8-fold variation in
concentration attributable to only run-up emissions" and that "the model-
extrapolated concentrations from the airport-specific activity analysis are
consistently at or near the 2.5th percentile of the Monte Carlo bounds while the
97.5th percentile of the Monte Carlo analysis is up to 2-fold higher than the
model-extrapolated concentrations from the airport-specific activity analysis" due
to the use of "a much shorter run-up time in developing the model-extrapolated
lead concentrations in the national analysis compared with run-up times that have
been observed at other airports." Other factors, including wind speed variations,
further add to this uncertainty.
Therefore, although the analysis presented in this paper provides information
about the range of impacts, I am skeptical about the advisability of using the
AQFs for calculating impacts and, potentially, compliance with the NAAQS, at
individual airports.
EPA Response: As the Reviewer points out, and, in alignment with the
goals of this assessment, the report provides information about the
range of estimated concentrations of lead in air at and downwind of
the maximum impact area at airports nationwide. We further refine
theses estimates at a subset of airports where there may be the
potential for lead concentrations to be above the level of the lead
NAAQS. EPA is not using these estimated lead concentrations to
evaluate attainment of the lead NAAQS.
We incorporated additional text in the Introduction to the Report to
further emphasize that in making determinations regarding violations
of the NAAQS for lead, EPA relies solely on the lead surveillance
monitoring network. EPA's guidance on this matter is provided in the
National Ambient Air Quality Standards for Lead Final Rule
fhttp://www, gpo.gov/fdsYs/pke/FR-2008-ll-12/pdf/E8-25654.pda
46

-------
3.4 Response to Comments Received from Reviewer 4: John R. Pehrson
CDM Smith
The general description of piston-engine aircraft operations, use of lead in AvGas,
and the lead impact locations at airports is adequate for this evaluation. The
discussion clearly states the reason for the study, the general approaches used to
develop the lead inventories and air quality factors, and comparison to monitored
concentrations at the model airport.
However, an item of potential concern is the fairly consistent model under-
prediction of monitored lead concentrations at the maximum impact location and at
locations 60 meters downwind (shown in Figure 1 on page 9) at the model airport.
The report identifies three likely sources of uncertainty: (i) exact location of run-up
activity relative to monitoring station locations, (ii) duration of run-up activity for
each flight, and (iii) whether single-engine or multi-engine aircraft were being used.
A recent study also highlights the variability of piston-engine aircraft emissions for
the same engine type and pilot (Yacovitch et al. 2016):
"In contrast, piston engines, which drive small propeller planes, operate in a much
more fexible manner. Piston engines are rugged and imprecise and pilots can operate
them in various ways with simple levers (e.g., the throttle and mixer) in the cockpit
Power and emissions are weakly linked, particularly in low-power states like idle and
taxi. The nature of piston engines means that there is also a great deal of variability in
their emissions, even for the same pilot operating the same airplane." [Chapter 3, page
11]
This is worth noting more clearly in the early sections of the USEPA report. The
findings presented in Yacovitch et al. indicate that measurements at the same
location, with the same aircraft, pilot, and wind conditions could still produce
noticeably different results.
EPA Response: We appreciate the information provided in Yacovitch et
a I., and we now cite these findings as a contribution to variability and
uncertainty in Section 4.4. The variability in individual pilot behavior and
aircraft emissions that the Reviewer points to could influence
concentrations both higher or lower than modeled concentrations. To the
extent that the modeling at RHV is under-predicting lead concentrations,
we acknowledge the Reviewer's point that this is not necessarily tied only
to the three parameters previously highlighted. We have added text to
Section 4.3 to highlight additional sources of variability; we also added
text to Section 2 to further clarify that the under-prediction in modeling
47

-------
results at the model airport is within the common model evaluation
criterion of values falling within 2:1 of monitored concentrations.
Notably, these are daily values, which inherently include more variability
than the 3-month rolling averages used to calculate model-extrapolated
concentrations at airports nationwide. Comparisons between model-
extrapolated concentrations and monitor data, presented in Sections 4.1,
4.2, and 4.3 also suggest that any under-prediction at the model airport
still yields general agreement between model-extrapolated and
monitored concentrations.
Returning to other causes of uncertainty, Appendix A discusses sensitivity analyses
conducted to determine the potential cause of the modeled discrepancy with
monitored results. On pages 22 through 25 of Appendix A, the three likely sources
of uncertainty were noted and sensitivity analyses were conducted. Run-up
duration is one of the parameters considered in the sensitivity analyses. It appears
from the data presented in Appendix A, ten days of activity data including engine
run-up durations were collected - all in the Summer season. Would weather
conditions encountered in other seasons prompt pilots to spend more time in
engine run-up before taking off?
EPA Response: There may be a short increase in run-up time for
carbureted engines to conduct de-icing during winter conditions. We
account for variation in run-up duration, which could result from
differences in seasonal conditions, individual airport characteristics, or
other attributes, by conducting a Monte-Carlo analysis that draws from a
distribution of run-up times across five airport studies. Further, the
maximum activity period for a given airport is likely to be in the summer,
as GA operations nationwide peak in May and reach a minimum in
January (Wang and Horn. 1985).
Wang, G. H., & Horn, R. J. (1985). Temporal patterns of aircraft
operations at US Airports: A statistical analysis. Transportation Research
Part A: General, 19(4), 325-335.
Following this discussion, two other sources of modeling uncertainty were noted
beginning on page 25 of Appendix A: selection of the initial sigma-y and sigma-z
values, and source exclusion zones for volume sources. It is not clear if any
sensitivity evaluations were conducted on these parameters. The model exclusion
zone issue can be avoided if area sources were used to model engine run-ups
instead of volume sources. The size of the area would require some thought,
48

-------
although selecting parameters that mimic the initial sigma-y value could be used.
With regard to choosing appropriate sigma-y and sigma-z values, one study
measured those parameters (Wayson, et al. 2003) and presented values for
turboprop commuter aircraft as well as large commercial aircraft.
EPA Response: We appreciate the Reviewer's comments regarding the
choice of volume vs. area source for aircraft and the selection of initial
sigma-y and sigma-z values. These are parameters for which we did not
conduct sensitivity analysis. We selected volume sources to treat piston-
engine aircraft emissions given the similarity between exhaust from these
engines and motor vehicles for which volume sources are typically
applied when evaluating the near field environment. This approach is
consistent with other piston-engine aircraft emission studies available in
the literature (Carr et al. 2011, Heiken et al. 2014, Feinberg et al. 2016).
We are unaware of any published work evaluating the treatment of
piston-engine aircraft sources as area vs. volume sources. The treatment
of aircraft sources as area or volume sources is a modeling topic noted
for future research, particularly when evaluating concentrations in the
near field environment from jet engine operations (Arunachalam, et al.,
2017).
Arunachalm, S., Valencia, A. et al., (2017). Dispersion Modeling Guidance
for Airports Addressing Local Air Quality Health Concerns. National
Academy of Sciences Airport Cooperative Research Program. Research
Report 179.
Carr, E., et al. (2011). "Development and evaluation of an air
quality modeling approach to assess near-field impacts of lead
emissions from piston-engine aircraft operating on leaded
aviation gasoline." Atmospheric Environment 45(32): 5795-5804.
Feinberg, Stephen, and Jay Turner. "Dispersion Modeling of Lead
Emissions from Piston Engine Aircraft at General Aviation
Facilities." Transportation Research Record: Journal of the
Transportation Research Board 2325 (2013): 34-42.
Heiken, J., et al. (2014). Quantifying Aircraft Lead Emissions At
Airports. ACRP Report 133.
49

-------
http://www.nap.edu/catalog/22142/quantifying-aircraft-lead-
emissions-at-airports.
The estimation of initial sigma-y and sigma-z values for volume source
representation of aircraft lead emission dispersion, the values could be compared to
the values for commuter aircraft (mostly turboprops) developed from a LIDAR
study of aircraft exhaust plumes presented in Wayson et al. Typical turboprop
initial sigma-y values during takeoff were 10.3 meters, and initial sigma-z values
were 4.1 meters during the takeoff roll (from 52 measurement events for commuter
aircraft). The measured values from Wayson, et al., 2003 implies that the sigma-y
values for takeoff in the report under review may be high by a factor of
approximately 2, and the taxi sigma-y value may also be high in the current report.
The chart below compares the values from the two reports.
Volume Source Initial Sigma Values for Aircraft

Takeoff, Wayson et al.
A







Run-Up, USEPA







Ta
xi, USEPA 2017 O



Takec
ff, USEPA 2017 O




















0	5	10	15	20	25
Sigma-Yo, meters
Given that commuter turboprops operating from a large commercial airport are
usually multiengine aircraft with seating capacity for 10 to 20 passengers, the sigma
values presented in Wayson et al., for takeoff would likely over-estimate initial
sigma values for typical general aviation aircraft. The USEPA study under review
indicates a substantial difference in takeoff/taxi initial sigma-y values to those for
run-up operations. If such a relationship holds across different aircraft sizes, and if
the takeoff sigma-y value for turboprops from Wayson et al. is more appropriate for
50

-------
general aviation aircraft, then the run-up initial sigma values presented in this
USEPA report might also be over-estimated. As noted on page 25 of Appendix A,
such an overestimation of initial sigma values would likely result in
underprediction of concentrations at nearby receptors.
EPA Response: We agree with the Reviewer's evaluation that, in
comparison to Wayson's measurements of initial sigma values during
take-off for turboprops, the initial sigma values we selected for modeling
piston aircraft taxi and takeoff might be over-estimated, and could
additionally contribute to the underestimate of lead concentrations at
our model airport. We note this as a source of uncertainty in Section
4.4.2.
While the text in the last paragraph on page 8 implies that seven (7) days of
modeled and monitored data were compared, Appendix A, page 3 (middle
paragraph) states that only three (3) days had both surveyed hourly operational
data and lead monitoring data, and the other four (4) days of monitored
concentration data used an average activity profile developed from 10 days of
activity surveys to predict the monitored concentrations. This may be a point worth
noting (footnoting?) in Section 2. In addition, it is not clear which three days of
simultaneous measurements and activity data were used since the list of days with
collected activity data included four (4) days of simultaneous activity data and lead
measurements (8/20, 8/23, 8/26, and 8/28) - per Appendix A, pages 1 and 2, and
footnote 1.
EPA Response: The three days of simultaneous air monitoring and activity
data collection were 8/23, 8/26, and 8/28; southerly winds on 8/20
resulted in operations occurring predominantly on Runway 13L where we
did not have monitors placed to evaluate the maximum impact and
downwind gradient in lead from piston aircraft; therefore, model-to-
monitor comparisons were not used for this day. Clarification on this
point was added to Footnote 1 in Appendix A.
Potential typographical error: Section 1, Page 4, 2nd Paragraph. In the center of the
paragraph it is stated that: "Piston-engine aircraft conduct approximately 62 million
landing and takeoff operations (LTOs) annually (USEPA 2011)." In reviewing the
data referenced (the 2011 NEI data site), it appears that the piston-engine LTOs
included in Table 5 of EPA-420-B-13-040 is approximately 32 million LTOs, not 62
51

-------
million. Not sure if this is a typo, or if the 62 million figure is meant to also include
touch-and-go operations and/or helicopter operations.
EPA Response: The original sentence mistakenly referred to the 62 million
individual operations (landing and take-off events separately) as LTOs.
We have corrected the sentence to state that 32 million LTOs were
conducted by piston-engine aircraft.
The description of the approach to develop AQFs was clear. The AQFs were
developed from modeled results for each operation type (LTO vs T&G) and aircraft
size (SE vs ME). I'm assuming that there isn't sufficient monitoring data to conduct a
multivariate regression analysis on daily monitored concentrations to tease out
AQFs.
EPA Response: This assumption is correct.
I found Table 2 to be quite understandable. The steps to disaggregate total
operations down to daily/hourly operations, assignment of operations to specific
runway ends, and aggregating the results for the most commonly used runway was
clearly described in Table 2. Given that the detailed hourly data is difficult to obtain,
or is non-existent, the approach used in this study is well considered. Figures 2 and
3 also clarified the approach.
The methods used to estimate airport-specific activities and SE to ME ratio appears
reasonable.
The findings presented in Figures 7 and 9 are rather compelling, given the range of
potential uncertainty in the parameters used to estimate lead concentrations. These
figures clearly indicate that the overall approach to estimating lead concentrations
at general aviation airports produces results that are fairly consistent (same order
of magnitude and often within the expected range) with measured values at a
handful of airports with lead monitoring stations.
My primary concern with the results presented in Section 4 is the apparent
assumption that the maximum impact site is considered ambient air, especially
since the concentrations estimated at a distance of only 50 meters from the
maximum impact site already fall below the lead NAAQS. Footnote 40 at the bottom
of page 32 provides the official definition of ambient air by USEPA. I believe that
USEPA generally accepts that ambient air begins outside of access-controlled areas
of an air emissions source, at least for stationary source permitting. If the airport
were considered the source, and if a fence is installed around the airport, then
52

-------
ambient air would occur along the fence line. Given that the airport operator does
not generally own the aircraft using the airport, this typical definition of ambient air
may not apply. However, if this definition is considered to apply, then most of the
"maximum impact sites" for lead concentrations noted in this study would not be
ambient air if the general public does not have unrestricted access to these sites.
EPA Response: Ambient air is defined by EPA regulations as that portion
of the atmosphere, external to buildings, to which the general public has
access. At airports, the general public includes recreational pilots and
their passengers, members of the public who visit the airport for special
events, and may include other populations (e.g., people who rent
hangars). Locations at airports to which this population has access
include parking lots, observation decks, hangars, and access roads to
hangars.
For purposes of this report, instead of evaluating individual airports for
areas of potential ambient air, we have elected to apply the simple
criterion of unrestricted access. The final report identifies only those
facilities for which there is unrestricted access within 50 meters of the
maximum impact site at airports with model-extrapolated lead
concentration estimates above the level of the lead NAAQS.
The document should provide a clear description of what is meant on Page 6, 2nd
paragraph: "Following EPA practice, this analysis focuses on the maximum impact
area at airports nationwide..." While it is technically easier to calculate lead
concentration near aircraft run-up areas, it may not be the appropriate point to
estimate impacts relative to ambient air quality standards.
EPA Response: We revised this section of the report to be more clear. It is
common practice, particularly in a screening analysis, to evaluate the
maximum impact locations. And, with regard to the NAAQS for lead, EPA
specifically noted the need to evaluate maximum impact locations (see
source-oriented monitoring in https://www.gpo.qov/fdsvs/pka/FR-2008-
ll-12ZpdfZE8-25654.pdf).
Editorial comment: Figures A-l and A-2 are not very clear. These figures may each
need to be full page to see all of the detail discussed in the figure captions.
EPA Response: We have increased the size of the figures.
53

-------
Potential typo: Appendix A, Page 20, Equation A-4: Is the term "Eq. 4" in Equation A-
4 really meant to be "Equation A-3"l
EPA Response: We have corrected this error in the report.
Format consistency: The text on page 5 of Appendix A uses dots (.) in the table
numbers: "... in Table A.l..." while the actual table titles use dashed (-): Table A-l.
Might check this throughout the document.
EPA Response: The references to the table have been corrected for
consistency.
Potential typo: Appendix C, Page 4, 2nd Paragraph: A figure is suggested, but not
actually shown at: "The concentration from only run-up emissions at the maximum
impact site receptor was 0.034 |ig/m3 for the 5th percentile, 0.257 |ig/m3 for the
95th percentile, and 0.092 |ig/m3 for the default run-up duration, as shown in
Figure C-. Lead concentrations..."
EPA Response: We have corrected this typo in the report.
References
Wayson R.L., Fleming, G.G., Kim, B., Eberhard, W.L., Brewer, W.A., Draper, J., Pehrson,
J., and Johnson, R., 2003. The Use of LIDAR to Charaterize Aircraft Exhaust Plumes.
Air & Waste Management 2003 Annual Conference and Exhibition. Paper No.
69965.
Yacovitch, T.I., Zhenhong, Y., Herndon, S.C., Miake-Lye, R., Liscensky, D., Knighton,
W.B., Kenney, M., Schoonard, C., and Pringle, P., 2016. ACRP Report 164 - Exhaust
Emissions from In-Use General
54

-------
3.5 Response to Comments Received from Reviewer 5: Sandy Webb
Environmental Consulting Group LLC
Overall the report was well written with straight forward, systematic, and detailed
methodology descriptions. The "way finding" (section introductions, how sections
were organized, descriptions of section contents, etc.) throughout the report was
very helpful.
1. Sections 1 and 2 describe the nature of how piston-engine aircraft
operate for safety and logistical reasons, along with the previous work
that EPA and others have conducted to characterize concentrations of
lead in air at individual airports. As stated in the report, conducting
detailed air quality modeling or monitoring at all US airports is not
feasible due to resource constraints. Please comment on the extent to
which this information is clearly described and provide your
perspective on the approach selected to utilize modeling from an
individual airport in order to characterize concentrations of lead in air
at and downwind of maximum impact areas of airports nationwide.
The methodology described is very systematic, rational, and clearly written. Each
section was clearly introduced and the sequence of information was logical.
Page 4, paragraph 4 describes clearly the source and purpose of lead in avgas.
However, it is perhaps an over simplification. Tetraethyl lead is "splash blended"
into unleaded gasoline resulting in highly variable lead concentration. This can be
seen in Appendix C. While the spec is for 2.12 grams/gallon, it can be quite a bit
higher and lower. Also this was the only mention of 100 octane avgas. More
justification for not modeling the higher lead-content gasoline should be provided.
EPA Response: We have added explanation to Appendix C regarding the
rationale for not modeling the higher lead-content avgas in this
assessment. Appendix C also includes discussion of variability in avgas
lead concentrations.
It would be very useful in Section 1 to include at least a qualitative material balance
for lead into and out of a piston engine - describing TEL in gasoline into the engine,
combustion transforming the lead, lead captured in engine oil, and lead in exhaust.
Include a physical description of the lead emissions (Are there any aerosol
emissions containing lead? Is there information on particle size distribution? Are all
lead emissions as elemental lead?) A more quantitative explanation of some of this
55

-------
appears in Appendix A, page A3, Aircraft Emission Rates. This will help explain the
importance of fuel consumption rates and times-in-mode.
EPA Response: We have included information requested by the Reviewer
in Appendix A.
In Section 1 paragraph 2 and footnote 5 as well as Section 2.1 in the last paragraph
on page 7, the run-up area is declared the "maximum impact area." While this is
likely true, a more complete explanation is warranted. From these descriptions it
was not clear whether ground-level lead concentrations in the maximum impact
area were based on modeling, which would include emissions dispersed into the
run-up area from other parts of the airport, or if these concentrations were solely
the result of run-up emissions. I believe it was the modeled emissions but was
confused by the write up in Section 2.1 Appendix A, Table A-l shows much higher
monthly and annual emissions from taxi-out than run-up. Also, taxiways, parallel to
runways, will largely be upwind of the run-up area for whichever runway end is in
use. Were these emissions captured in the model airport modeling that was the
basis for the AQF? Could taxi-out emissions be a significant addition to run-up
emissions in the maximum impact area? Whether or not there is a significant
contribution this should be explained. It was difficult to tell from reading the AQF
methodology (Equation 1) if it was a top-down computation from airport-wide
emissions calculations or a bottoms-up computation based on run-up activity. I may
well have missed something in the explanation but it was not clear to me.
EPA Response: To quantitatively evaluate the relative contribution of
aircraft lead emissions during different modes of operation on lead
concentrations in air, EPA modeled the emissions, including locations and
relative emission rates of lead at the model airport. The taxi-way
emissions were included in the development of the AQFs at the model
airport. The Reviewer makes an important distinction between the
relative contribution of run-up and taxi-out lead emissions versus the
contribution of these emissions to concentrations in the defined area of
maximum impact near the run-up location. We added clarification to the
Introduction of the report to note that while taxi-out emissions of lead
can be higher than lead emissions during run-up, the taxi-out emissions
occur while the aircraft is underway, which causes much greater
dispersion of the emissions compared with the ground-based emissions
during run-up which occur while the aircraft is stationary. Thus, the
quantitative contribution of run-up emissions to lead concentrations in
air at the maximum impact site is much larger than the contribution of
56

-------
taxi-out emissions at this location. However, in order to holistically
evaluate aircraft emissions on air quality, EPA has included all aircraft
emissions when developing AQFs from the study at the model airport
(i.e., taxi and other modes are included in AQFs along with run-up) (we
added a Footnote in Section 3.1 to clarify this point).
Section 2.2 says "...the model tended to under predict monitored concentrations ..."
Systematically under predicting monitored concentrations raises questions about
the methodology. See also Figure 8. This was practically dismissed as a concern.
Perhaps a more complete explanation is warranted. I was unable to discern the
reason.
EPA Response: EPA has added additional text to the description and
discussion of the model-to-monitor comparisons conducted at the model
airport. See response to Reviewer 3 on Page 32. We note that this
underestimation may be due to day-to-day operational variability,
changes in the location of run-up procedures, and variability in the
monitored concentration, among other reasons addressed in the
response to Reviewer 3; however, this under prediction does not appear
to be systematic considering modeling performed at other airports (see
references in Section 2.2) using the same or similar approaches.
Additional discussion has been added to Sections 2.2 and 4.4 to address
sources of uncertainty and their implication on the analysis.
The Reviewer also points to the airport-specific analysis later in the report
with the reference to Figure 8 (now Figure 10). There, we expanded our
discussion of the evaluation of individual airports in Section 3.3. This
expanded discussion now includes a series of sensitivity analyses to
capture potential sources of uncertainty and variability. We examine
airports where concentrations may be underestimated in the national
analysis by re-evaluating these airports with the assumption that all GA
activity and 50% of AT activity would be performed by piston-engine
aircraft. This sensitivity analysis, in effect, increases the maximum
modeled concentration by a minimum of 28% for each airport. The seven
days of model-to-monitor comparison at the monitor airport showed a
model underestimation of, on average, of 16.8%. Thus, we believe our
sensitivity analyses are, to first order, sufficient for capturing uncertainty
from potential model underestimation at the model airport.
57

-------
Regardless of these comments, I do think the overall methodology is a good one for
calculating lead emissions for GA airports nationwide.
2. Section 3.1 presents the methods to calculate Air Quality Factors (AQFs)
at the model airport. Please comment on the approach used to calculate
AQFs at the model airport specifically for the purposes of using these
factors to estimate concentrations at and downwind of maximum
impact areas of airports nationwide.
Developing the AQFs by aircraft class and applying those to other airports is a
legitimate approach and should provide consistent results from airport to airport.
However, as noted above I couldn't tell whether the AQFs were computed based on
total airport emissions or only those from the run-up area. If the AQFs are
appropriately developed this procedure for estimating concentrations makes sense
and works well for a national analysis.
EPA Response: We appreciate the Reviewer's perspectives on the
approach and confirm here that all aircraft emissions (i.e., during all
modes of operation) were included in developing the AQFs.
3. Table 2 and accompanying text in Section 3.2 describe the methods
used to estimate piston- engine aircraft landing and take-offs (LTOs) at
individual runway ends on a rolling 3-month basis (e.g., apportioning
out piston-engine-specific LTOs from total LTOs at each airport,
allocating annual activity to daily and then hourly periods). Are these
methods clearly described and do you have recommended changes to
the steps taken? Please explain any alternative options and provide the
location of data sources that would support such alternative options.
Table 1 in Section 3.1 would be easier to read if the data was presented as (10-6 mg
Pb/m3/LTO).
EPA Response: This comment has been addressed.
Table 2 in Section 3.2 was very systematic and clear. Including the rationale for each
step was very helpful. Overall the table is so detailed and long it is difficult to keep
all steps in mind. The calculations for number of operations by aircraft type by hour,
month, and averaging period all seemed reasonable and appropriate for developing
a national analysis. This process appears to get the proper result so no suggestions
for alternative options or data sources.
58

-------
I found Step 13avi in Table 2 confusing. Does this imply the need for and availability
of avgas lead concentration at each airport or is this only for the model airport?
EPA Response: Step 13avi in Table 2 describes the step introduced to
provide extrapolated lead concentrations at airports where lead
concentration in avgas has not been measured. The avgas lead
concentration at the model airport was used in the air quality modeling
results that provided the AQFs, which were in turn used to calculate
model-extrapolated concentrations at each airport nationwide. In order
to use a standardized lead avgas concentration instead of the
concentration measured at the model airport, we scaled model-
extrapolated concentrations at each airport by the ratio of the ASTM
maximum standard to the concentration at the model airport. The result
provides concentrations of lead in air at each US airport that include an
avgas lead concentration equal to the ASTM maximum.
4. Section 3.3 presents an analysis to refine estimates of piston-engine
aircraft activity using airport-specific data for the subset of airports.
Please comment on whether there are alternative airport-specific data,
or analysis approaches, that could improve estimates of piston-engine
aircraft activity at a subset of airports. In addition, please comment on
whether parameters other than piston-engine aircraft activity should
be included in analyses to potentially improve model-extrapolated
concentrations at a subset of airports, noting that additional
parameters are evaluated at all airports in the uncertainty and
variability analyses presented in Section 4.
National data sources of data on GA activity are notoriously unreliable because of
the nature of much of the GA activity. There is significant training activity at many
GA airports but often no compelling reason to track T&Gs. Many aircraft are flown
infrequently. Data reports to FAA frequently are not prepared rigorously. All
emissions analysis projects regarding general aviation have to deal with these
limitations. The most accurate analyses are based on data collected on-site. This is
not feasible for preparing a national estimate of lead concentrations. As far as my
experience with GA data sources, EPA has selected the most reliable and the
uncertainty analysis suggests that data limitations will not have a significant effect
on the overall estimation results.
On page 29, the discussion of Figure A-9 references concentrations "...labeled as
Heiken Fuel Consumption..." There are no references to Heiken in the table.
59

-------
EPA Response: This comment has been address and the reference has
been corrected.
Figure A-9 Run-Up and the discussion in paragraph 2 on page 31 - Based on this
data, the run-up fuel consumption (time-in-mode and consumption rate) used in
this analysis looks low compared to Heiken et al's findings. Also for the discussion of
run-up times-in-mode in Section C.2, especially the last sentence in paragraph 1 on
page 4 in C.2.2, "Their modeling found that changing the emissions attributable to
run-up from 3% of modeled emissions to 5% of modeled emissions resulted in a
34% increase in annual atmospheric lead concentrations." Perhaps the run-up fuel
consumption should be reconsidered and the concentrations recomputed using a
higher TIM and consumption rate. This could account for the lower modeled
concentrations compared to the measured concentrations.
EPA Response: The time in mode (TIM) data at the model airport are
based on observations at that facility, and while we agree that the fuel
consumption rate for run-up emissions used in the air quality modeling at
the model airport is low compared with data from Heiken et a I., any
increase in fuel consumption during the run-up mode of operation would
improve the model-monitor comparisons for some of the days of under-
prediction, but would result in an over-prediction for some days. Given
the sparse dataset for fuel consumption during run-up (i.e., only four
unique fuel consumption rates across engines types compared with 18
unique fuel consumption rates for other modes of operation), we did not
include this parameter in a quantitative uncertainty analysis. As noted by
a separate Reviewer, fuel consumption rates and times in mode vary
from pilot to pilot and within events for the same pilot. We note
parameters other than TIM and fuel consumption rates that may
contribute to the model-to-monitor differences and quantify their
impacts in Appendix A. When extrapolating the model airport results to
other airports, we account for differences in TIM during run-up by
conducting a quantitative uncertainty analysis to characterize the impact
of this important parameter on lead concentrations in maximum impact
areas of airports nationwide. While a sufficient sample size was available
to conduct such a quantitative analysis for run-up TIM, data were
insufficient to conduct a similar analysis on fuel consumption rate.
The second paragraph on page 4 in Section C.2.2 references a figure with no figure
number ("Figure C-"). It appears there is a missing figure showing sensitivity
analysis.
60

-------
EPA Response: This comment has been addressed.
5. EPA provides coarse comparisons of monitored lead concentrations to
model-extrapolation results from the national and airport t-specific
analyses, in Sections 4.1 and 4.2, respectively. In Section 4.3, EPA
provides a more detailed comparison of data from lead monitors placed
in close proximity to the locations of model-extrapolated
concentrations. Please comment on the appropriateness of the
approaches to compare mo del-extrapolated results to monitored
concentrations of lead given available monitoring data. Based on your
understanding of the methods presented in the report and the
comparisons of monitor and model-extrapolated concentrations, please
provide your perspective on the performance of the methods in
characterizing the ranges of lead concentrations from piston-engine
aircraft at and downwind of US airports.
The model-extrapolated approach is probably the best approach for estimating lead
concentrations at airports nation-wide. The modeled vs. monitored and model
airport analysis confirm this. As noted in Section 4.2, only about 27 out of 13,000
airports are likely to exceed the NAAQS. These are all high activity airports and will
likely have the best data on operations, based aircraft, mix of aircraft type, etc.
As noted in an earlier comment, having the model-extrapolated estimates falling
consistently below airport-specific analysis raises a concern, however, the reason
for this is not evident.
EPA Response: We believe the Reviewer is raising two potentially inter-
related points. The first, in referring to a previous comment, notes the
potential systematic under-estimation in modeled concentrations at the
model airport. As discussed in the response to the comment about
Section 2.2 above, modeling conducted by EPA using similar methods
indicates that the approach used does not systematically underestimate
monitor concentrations. We have added text to that section and Section
4.4 to better indicate that we recognize model performance as a source
of uncertainty and its implications on interpreting these results. This
comment may also pertain to the data we provide regarding the
uncertainty analysis around the national analysis results which indicate
that the model-extrapolated concentration estimates are consistently
lower that alternative values that use the range of data available for run-
61

-------
up duration. This is because the run-up duration at the model airport
was at the low end of the range of values measured at other airports.
The second point refers more specifically to the evaluation of individual
airports for potential to have lead levels above the lead NAAQS. In the
airport-specific analysis, we compare two model-extrapolated lead
concentrations for each of the airports with the potential to have lead
levels above the lead NAAQS. The national analysis estimates typically
(but not for all airports) fall below the airport-specific estimates as a
result of replacing national average piston-engine activity fractions with
airport-specific data. A potential explanation for this is that many
airports with a significant amount of GA activity may have a higher
percentage of GA activity performed by piston-engine aircraft than at
other airports. This occurrence would result in the model-extrapolated
concentrations in the airport-specific analysis exceeding the estimate for
the model-extrapolated concentrations using national default piston-
engine percentages. Thus, the airports that are identified in the airport-
specific analysis may represent airports that both have high activity and a
high percentage of piston operations. We recognize that activity and
operational data, both using national default and airport specific values,
is a source of potential uncertainty, and we have added text to better
describe activity uncertainty to Section 4.4 of the report.
Page 47, last paragraph, second sentence says "As a conservative assumption, the
ASTM standard for the maximum lead concentration in 100LL was used in the
national analysis ..." As shown in Figure C-l, the lead concentration is often higher
than the standard so I would not consider this a "conservative" assumption,
however, I think it is a reasonable assumption.
EPA Response: We agree and have deleted the use of this word in the
report in reference to concentration estimates.
6. Section 4.3 presents quantitative and qualitative uncertainty analyses
of the model-extrapolated results provided in previous sections. Please
provide your perspective on the methods used to conduct these
uncertainty and variability analyses, as well as the key parameters EPA
included in the analyses (based on previous work discussed in Section
2). Please provide your perspective on the application of this analysis to
further characterize the range of lead concentrations attributable to
piston aircraft activity at airports nationwide.
62

-------
The analysis methods used to assess the analytical results are appropriate and seem
to be well carried out. My experience and knowledge of statistical techniques is
rather limited and I will rely on the other reviewers to assess this section. I know
some of them are experts with these techniques.
The description of the results of Monte Carlo analysis gets a bit tedious and could
perhaps be moved to an appendix.
EPA Response: We endeavored to keep the text describing the Monte
Carlo analysis as streamlined as possible while allowing the reader to
understand the basic steps undertaken.
The caption for Figure 10 is not very clear, especially the explanation of the black
and blue concentration values.
EPA Response: This comment has been addressed and the figure caption
has been improved for clarity.
The text should point out the different scales used in Figures 10 and 11.
EPA Response: This comment has been addressed.
It is very difficult to read the concentration scales on Figure 12, both panel A and B.
EPA Response: We have made changes to the text to make it easier to
read.
Section 4.3.2 - the first half of the first paragraph is a very clear explanation of the
approach taken with this analysis. In the second half of the paragraph, I find the
discussion of mixing height confusing. It was not clear to me that mixing height
would have much of an effect on concentrations in the run-up area/maximum
impact location. In the 3rd paragraph, it would be helpful to give range and not just
the difference for the 3-month AQF.
EPA Response: We have rewritten Section 4.4.1 to respond to reviewer
comments on meteorological parameters and other sources of
uncertainty. We have placed the discussion of mixing height and its
contribution to concentration uncertainty in the context of other
meteorological parameters. We agree with the Reviewer that, given that
the largest contributor to concentrations at the maximum impact
location is nearby run-up emissions, mixing height would not be the
63

-------
biggest contributor to uncertainty. However, we do note that even near-
source concentrations of primary pollutants have been shown to be
dependent on meteorological variables and these variables may
contribute to greater uncertainty at downwind locations.
Section 5, paragraph 2, sentence 1 - provide the max value and put "greater than
NAASQ" into the parenthetical statement.
EPA Response: This comment has been addressed.
7. Editorial Comments
Page 4, paragraph 3, delete unneeded words, "... aviation gasoline for several
reasons, namely to help..."
Page 4, footnote 1 - second sentence should read "Facility types other than airports
tt
Page 8, paragraph 2 - beginning of second line has inconsistent use of a dash
"aircraft- and meteorological data," - should be a dash to follow meteorological or
not one after aircraft.
Page 13, paragraph after numbered list - line 3 should read "... were used to
evaluate ..."
Page 18, line 10 should read "... runway is multiplied ..." rather than "... runway as
multiplied ..."
Page 20, Table 2, Step 1, step description - delete the first word "Determine"
Page 23, Table 2, footnote 29, last line - the wording after "airports" is very
awkward where it says airports "likely have a distinct activity profile from GA
airports." This should be explained more clearly/simply.
Page 48, first paragraph, Table 5 on page 50 is too far from the initial table callout
on page 48, paragraph 1, line 9
Page 51, first paragraph, figure 10 on page 52 is too far from the initial figure callout
on page 51, paragraph 1, line 1
Page 56, paragraph 3, line 6 should read "... order to capture ..."
64

-------
Page 62, references, for the Heiken reference, it should read ACRP Report 133
rather than ACRP 02-34. Once published, ACRP drops all references to project
numbers in deference to the report number. This report is also referenced on pages
A32, Appendix B references, and C9
Page B1 through end of Appendix B; page numbers are missing
Figure B-l is split across two pages (B3 and B4) making it look like a figure number
is missing on page B3
Figure C-6 (a), (b), (c), and (d) - add Run-Up Time units of measure to y-axis
EPA Response: The editorial comments above have been addressed.
65

-------