No.	EPA Comment	Research Group Response



1	

We have two key concerns that address some of the initial study
objectives as stated in the report. Our first concern is with the
following objective:

"Quantify personal exposures to motor vehicle gasoline evaporative and
combustion emissions in micro environments, (MEs) representing the upper
end of the population exposure. frequency distribution (99th +
percentile) of such exposures," (p. 1-2)

It would be very helpful to evaluate whether the study accomplished this
objective by comparing its findings to other exposure studies. A table
comparing average and "high end" concentrations from this study to those
reported in other studies would be useful in addressing whether this
study captured the "99th+ percentile" exposure concentrations. Studies
that might be appropriate for comparison include population-based
sampling programs such as the National Human Exposure Assessment Survey
(NHEXAS) Phase I studies, the Relationship of Indoor, Outdoor, and
Personal Air (RIOPA) study, and others provided in the attached
reference guide. Studies addressing specific microenvironments addressed
in the 211 (b) report should also be noted (see attached reference
guide).

 In regards to this same concern, it would be useful to define with
specificity the operational definition of "99th+ percentile" for
evaluation of the study objective. For instance, what averaging time
should be considered appropriate for assessing this factor? What metric
(micro-environmental concentration, inhaled dose, etc.)?

	

We are unable to compare “average and high end concentrations from
this study” since no average exposures were sampled during the study. 
 Sampling locations and conditions were chosen to reflect high end
exposures in all cases.  Where average values are mentioned in the
report they refer to averages of replicate high end samples.  The
sampling criteria used to identify high end conditions are described at
page 1-3 and in Appendix A (pp. 3-13 to 3-17) of the report.  

Since we sampled only high end conditions, we’re unable to determine
the precise percentile range of high-end values from the data collected.
 Early discussions with Agency staff during protocol development lead to
the decision to characterize ‘high end’ exposures as ≥ 99th
percentile.  This rationale followed from the compounding effect of
multiple independent selection criteria used to identify high end
conditions. For example, if a single sampling criterion led to exposures
at the 80th or 90th percentile, then samples that met several such
criteria should plausibly exceed the 99th percentile when compounded. 
We believe this to be a conservative estimate since a number of
individual sampling criteria (e.g., low to calm wind speeds, down-wind
locations, cold start situations, congested rush-hour conditions) could
by themselves on occasion change near-source exposures 100-fold from
their opposite extremes.

    

Agency reviewers suggest that direct comparisons to studies of similar
MEs and averaging times would be helpful.  Such preliminary comparisons
were provided earlier to EPA in the Table 4-1 of the 3/14/2003 interim
Atlanta S211b exposure report.  Model ME factor comparisons in Tables
3-2 to 3-4 of the final report also offer additional perspective. 
However, other study values must be adjusted for changes in fuel
composition and vehicle fleet emissions for such comparisons to be
informative and not misleading.  Such an approach is described in the
VCCEP benzene study [  HYPERLINK
"http://www.tera.org/peer/VCCEP/Benzene/BenzeneWelcome.html" 
http://www.tera.org/peer/VCCEP/Benzene/BenzeneWelcome.html  (Appendices
A-2 & B)] that uses a Mobile 6 model-based adjustment procedure. 
However, such supplemental analysis is beyond the scope of the Tier 2
study.

The operational averaging times should be taken in the breathing zone
for periods comparable to those typically spent in such high end MEs. 
For example, refueling ME sampling times were about 20 minutes to
reflect time spent at the service station and in-cabin exposures on
congested freeways were sampled for about 40 minutes to reflect rush
hour commuting trips (see EPA Exposure Factors Handbook:
http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=20563) Reconstructed
1-minute data were provided in the final report to facilitate exposure
model evaluations of alternative behavioral patterns.



2	The other study objective that should be further addressed is:

"Extrapolate study data to other cities and other oxygenated fuels," (p.
1-2)

It is unclear how such extrapolation could be performed for some of the
analyses undertaken. Given the lack of a traditional "balanced"
statistical design, covariance of fuel type and climate is likely to
limit the range of extrapolation. The report should include the
investigators' assessment of the suitability of the various types of
data collected for application.

The presentation of data is in descriptive and graphical form, rather
than statistical. While EPA will employ these data in other contexts for
specific application, some of the tables in which average values are
compared for several variables (e.g., Table 4.3-1 1) would be better
presented using statistical techniques such as ANOVA or mixed models.
Presentations of the average values in the tables are helpful, but the
reader would be better served if significance and statistical power were
explicitly addressed. Furthermore, graphical presentation of statistical
distributions of data (e.g., box plots, cumulative probability
distributions) would be helpful even in interpreting summarization of
data in the absence of statistical comparisons.

	

As noted above, application of this high end ME exposure data to other
times and locations will have to address interim changes in fuel and
fleet characteristics as well as case-by-case similarities and
differences in MEs, fuel component properties, and meteorological
conditions between locations. Extrapolation of the study data to other
cities and other oxygenated fuels is beyond the scope of this Tier 2
study report.

The basic objective of the Tier 2 high end exposure study is to provide
exposure data to EPA for their subsequent analysis.  The descriptive
statistics and data summary presentations were provided as a courtesy to
the Agency.  The supplemental statistical analysis suggested by
reviewers is beyond the scope of the Tier 2 study report.





3	p. ES-5, top:

Background exposure measurements were made during the first 10 minutes
without the test vehicle. What was the rationale for 10 minutes? What
time of day were these made? How was seasonal and other variability
accounted for in the decision to use a single 10-minute measurement as
background? That is, it appears that the study assumes that background
concentrations do not vary in time.

	

The trailing vehicle experiments are described in Section 5 of the final
report. The 10 min period was selected to match other 10 min segments of
the trailing vehicle test protocol. These measurements were used to
establish the initial in-cabin concentrations prior to each run. The
ambient background is not relevant in this context.



4	p. ES-5, line 3:

Change "include" to "included" in "Each trailing vehicle run include
three scenarios. . . ."

	

This typographical error has been corrected.



5	p. ES-5, second paragraph:

We assume that the vehicles parked in a closed garage with the gasoline
powered lawnmower and gasoline storage container all were filled with
the same fuel, but the report should explicitly state whether or not
this is the case.

Also, please further explain the meaning of semi-continuous measurement.

	

The assumption is correct.  A statement to this effect is added at page
ES-5 and is noted in the full description of the attached garage
experiments in Section 4 at page 4-2 of the final report. 

Semi-continuous measurement by solid phase micro-extraction (SPME) is
listed in Table E-1 at page ES-3 and described in detail in Appendix C.
Semi-continuous (10 minute average) is now parenthetically defined at
ES-5.



6	p. ES-5:

The twelve microenvironments appear to be sufficiently diverse to
characterize near-road exposures.

The term "urban canyon" is used here and throughout the report, but
"urban canyon" is never defined or fully characterized. Urban dispersion
is highly variable from one city to the next (e.g., Houston's will be
quite different from Atlanta's) and even within one city (e.g., we
observed significant differences between two nearby urban canyons in NYC
- Rockefeller Center and Madison Square Garden). Thus, the term should
be defined generally and the specific urban canyon (i.e., ME2) in this
study should be fully characterized (e.g., height of buildings, baffling
and obstructions, distance between structures, etc.). Perhaps an
appendix could be added with photographs of the urban canyon monitoring
sites.

	

A general expanded description of the urban canyon microenvironment and
the selection criteria used to identify appropriate locations in the
cities tested are provided at page 3-13 of Appendix A.  Earlier
city-specific reports [May 5, 2004 Data Report – Summer 2003 Chicago &
Atlanta Field Studies – Appendix A; March 5, 2004 Data Report –
Summer 2002 San Antonio & Houston Field Studies] provided to EPA
identified specific street canyon loops selected for testing that are
now footnoted specifically at pages 1-7 to 1-9 for these cities in the
final report.  Although the lengths of the routes for this ME were too
long for any detailed description of the buildings to be meaningful for
pollutant dispersion purposes, a Google Map link at (  HYPERLINK
"http://maps.google.com/maps?hl=en&tab=wl" 
http://maps.google.com/maps?hl=en&tab=wl ) provides satellite
photography and street views that can be used to characterize these
Atlanta, Chicago, and Houston street canyon loops. 



7	p. ES-9, ES- 14, and 4-5, Figure E- 1 :

It might be helpful to make a general statement about the distribution
of the four

species in BTEX. For example, it appears from Figure E-4 on page ES-14
that the

ratios in Houston in the summer and winter are consistent among benzene,
toluene and the xylenes. However, are they evenly distributed across the
twelve MEs, or does one (e.g., benzene) or two dominate in some MEs
while others dominate in other MEs (e.g., xylenes)? Perhaps a discussion
linking the fuel variability (Table 4.3-1 on page 4-5) would be
worthwhile. Similarly, the report may provide some data to answer how
much of the concentration and spread is attributable to fuel effects.

	

In Figure E-4, the Houston ME/monitor BTX ratios appear similar (within
measurement error) across season for the exhaust dominated MEs as the
reviewer suggests.  However, the summer BT ratios do appear larger than
X ratios for ME3 and ME7 which are dominated by refueling evaporative
emissions, although the winter ratios appear less so.  While the fuel
specifications provided in Table 4.3-1 are typical of these cities and
relevant to the garage and trailing vehicle tests discussed in Sections
4 and 5 of the report, they are not directly applicable to the ME
measurements discussed in Section 3 which arise from actual but unknown
mixtures of locally available fuels.  Although seasonal fuel sampling of
local vendors is available in annual surveys for these cities, modeling
of fuel effects in fleet emissions is beyond the scope of the Tier 2
study report.

 



8	p. ES- 15, line 17:

The word "below" is misspelled.

	

This typographical error has been corrected.



9	p. ES- 16:

Can the 2.2 times higher in-cabin exposure levels for the trailing
vehicle relative to the normal emitter mode be attributed to the
differences in ventilation? Or, does vehicle type come into play here
(even though the test vehicles seem to show no significant difference)?

We are unsure that average in-cabin levels were minimally affected by
different vehicles or fuels except for an MTBE fuel effect. Comparing
the values for "truck only" and "car only" in Table E-2 on page ES-17,
it appears that 1,3-butadiene, orthoxylene, carbon monoxide and
formaldehyde differ substantially. Also, fuel type seems to vary
substantially for toluene (Chicago = 1.32, whereas Atlanta = 2.79) and
ethylbenzene (Chicago = 0.120, whereas Atlanta = 0.51), and Chicago fuel
was much lower for the xylenes. It may rest on what is meant by
"minimally affected."

	

The 2-fold higher exposures trailing a malfunctioning lead vehicle are
likely not attributable to cabin ventilation differences given the
efforts made to conduct the experiments under similar meteorological
conditions (e.g., calm or low wind speeds) and identical protocols
regarding ventilation conditions. The relatively higher cabin levels
would be anticipated as a result of higher emissions from a
malfunctioning lead vehicle. 

‘Table E-1’ typo is changed to Table E-2 on page ES-16.

We agree with the reviewer’s comment and modified the statement on
page E16 as follows: “Average in-cabin levels were affected by both
state of maintenance of the vehicle and fuel composition. Use of Atlanta
fuel resulted in higher BTEX concentrations in the trailing vehicle
cabin, which is consistent with higher aromatic contents of Atlanta
fuel, especially in summer. Similarly BTEX concentrations were
consistently higher during the summer, which is also in concert with
higher aromatic contents of summer fuels, especially for Atlanta.
Averaging over the fuels and vehicles, the high emitter mode resulted in
2.2 times higher in-cabin exposure levels for the trailing vehicle than
the normal emitter mode except for HCHO which was relatively unchanged
from background.” Also, the text in Section 5 (page 5-12 and 5-14) was
modified accordingly to reflect this conclusion.

The differences between “car only” and “truck only” seem to be
less consistent and less pronounced than fuel effect.  





10	p. ES-18 and 19:

The conclusions appear to be scientifically sound. The statement that
"although

overall the average population time spent in these high end MEs is
likely the shortest"

is important and should not be left to conjecture. The actual activities
associated with

exposures to evaporative eniissions should be characterized. Activity
patterns often

dictate actual exposures. Concentrations alone tell only part of the
exposure story.

We agree that HAPEMS factors may need to be updated based on these
findings, and

we will provide the final version of this report to the Office of Air
Quality Planning

and Standards who are responsible for the HAMPEM model, which was
recently

upgraded to HAPEM6, http://www.epa.gov/ttn/fera/huma-nh apem.htm1.

	

Other EPA documents provide adequate support for this statement which
need not be repeated in the report.  For example, Table 15-106 of the
EPA Exposure Factors Handbook (EPA/600/P-95/002Fc) reports that 157
(18-64 y) subjects, who had visited a gas station the day before in a
survey population of  9386, averaged 10 minutes (median) outside the
service station with an inter-quartile range of 5-15 minutes ( 
HYPERLINK "http://www.epa.gov/ncea/efh/pdfs/efh-chapter15.pdf" 
http://www.epa.gov/ncea/efh/pdfs/efh-chapter15.pdf ).

The HAPEM5 comment refers to ‘high end’ aspects of the 1998 Rodes et
al. (p. xi) CARB study where the driving protocol highlighted trailing
diesel city buses and HDD trucks when possible, a counter-intuitive
practice for most drivers.  Current HAPEM factors should be lowered to
simulate ‘typical’ behaviors where commuters are more likely to
avoid such vehicles.  Since the current report is also focused on
‘high end’ exposures (see response #1), comparisons with corrected
HAPEM simulations should be restricted to ‘high end’ model
projections.  Report findings should not be used to derive HAPEM factors
thought to characterize average or typical conditions.





11	p. 1-20, final paragraph:

Figure 1.3.2-5 is missing.

	

There is no figure 1.3.2-5. The text on p. 1-20 was corrected
accordingly.



12	p. 1-21: 

P-values should include at least two significant figures.	

The p-values have expanded significant figures at page 1-21 and 1-22.



13	p.2-5:

Note that "OMS" has changed its title to "OTAQ (Office of Transportation
and Air Quality).	

‘OMS’ has been changed to ‘OTAQ’ in Table 2-1 caption at page
2-5.



14	p.2-9:

How were the "error bars" in Figure 2-9 calculated? What components of
uncertainty are accounted for here? Sampling? Instrument precision?
Error propagation?

	

Presumably this question concerns Figure 2-1, as there is only one
figure in this section… A brief discussion of how uncertainties
associated with Figure 2-1 were determined is added to the initial
paragraph at page 2-4.



15	p. 3 - 1, second paragraph:

The 1999 NATA does not assume constant background concentrations of all
pollutants. See http://www.epa.gov/ttn/atw/natal999/background.html for
additional detail.

	

The report does not claim that a uniform background concentration is
assumed for all pollutants.



16	p. 3-3, first paragraph, last sentence:

The italicized text should be evaluated in light of the "general
comment" above.

Notably, this text is appropriate if, in comparison to other studies
looking at similar microenvironments, the concentrations in this study
are higher. The italicized text is likely inappropriate if this
comparison does not suggest that the micro-environmental exposures
presented herein are "extreme."	

We removed the italicization of the text. However,  as noted in
responses #1 & 10, the ‘high end’ study protocol was designed to
sample under conditions leading to more ‘extreme’ exposures that
those thought ‘typical’ or average.  To be informative, the
comparisons suggested with earlier studies, some of which themselves
sampled under ‘high end’ conditions (e.g. Rodes et al., 1998),
should be adjusted for interim changes in fuel composition and fleet
emissions: however, such extended analysis  is beyond the scope of this
report. 



17	p. 4-7 – 8, Table 4.3-6,7:

The tables and text do not seem to identify whether these non-aldehyde
data are from SMPE or canister samples.

	

The report has been edited (p. 4.7) to identify absorption tube,
canister, and SPME values.





18	p. 5 - 1 4, last paragraph:

". . .after the change to low ventilation.. . little change in the
in-cabin concentration.. ."

This text appears to contradict language on p. 5-1 8, second to last
paragraph, where it is noted that the range of concentrations differed
under high ventilation situations.

Please clarify the intended meaning, which appears to suggest that while
average concentrations were unaffected, variability in concentration was
affected by ventilation.

	

The latter reading grasps the meaning intended; language at p. 5-14 has
been clarified to state that although on average the integrated in-cabin
levels were similar for high and low ventilation, the variability (peaks
and valleys) of the in-cabin concentrations was larger during the high
ventilation condition.

