United States
Environmental Protection
Agency
Atmospheric Research and
Exposure Assessment Laboratory
Research Triangle Park, NC 27711
Research and Development
EPA/600/SR-93/075 June 1993
Project Summary
Relationships Between Ozone
Precursor Levels and
Response to Emissions
Reductions:
Analysis of Regional Oxidant
Model Results for the
Northeastern United States
Jana B. Milford, Dongfen Gao, and Antigoni Zafirakou
A detailed analysis of results from
the Regional Ozone Modeling for North-
east Transport (ROMNET) study has
been conducted to examine the condi-
tions under which alternative control
strategies were predicted to be effec-
tive in improving air quality. The
ROMNET study had predicted that for
most of the northeastern United States,
reducing nitrogen oxides (NOx) emis-
sions by about 60% would be more
effective in reducing ozone (O3) than
reducing anthropogenic reactive or-
ganic gas (ROG) emissions by approxi-
mately the same proportion. However,
for the New York City and Baltimore-
Washington areas, ROG controls were
predicted to be highly effective and NOx
controls to be counterproductive!
ROMNET results for cases in which the
reactivity of ROG emissions was re-
duced were similar to those for cases
in which the mass of ROG emissions
was reduced. Plots of O3 versus the
concurrent NO concentration in each
model grid cefl indicated that O3 in-
creased with NOy concentrations up to
10-15 ppb; and either increased or
decreased with higher NO depending
on the associated ROG levels. The
analysis also showed that reducing NOx
emissions by about 60% was uniformly
beneficial for grid cells with NO con-
centrations less than about 25 ppb, but
counterproductive for some grid cells
with NOy above 25 ppb. Ozone was
relatively insensitive to reductions in
ROG emissions in grid cells with NO
concentrations below 5-10 ppb. We
recommend further investigation of the
idea that NOy concentrations could
serve as an indicator of the likely sen-
sitivity of O3 to ROG or NOx controls, if
NOy was monitored along with peak
ozone during photochemical air pollu-
tion episodes.
This Project Summary was developed
by EPA's Atmospheric Research and
Exposure Assessment Laboratory, Re-
search Triangle Park, NC, to announce
key findings of the research report that
is fully documented in a separate re-
port of the same title (see Project Re-
port ordering information at back).
Introduction
The success of efforts to reduce sum-
mertime concentrations of ozone through-
out the northeastern United States de-
pends on our understanding of the chem-
istry involved in ozone formation in this
region. Three questions that are currently
the subject of controversy illustrate the
importance of our ability to describe the
chemistry that is occurring:
(1) What is the relative contribution to
ozone formation of biogenic versus
Printed on Recycled Paper
-------
anthropogenic emissions of reactive
organic gases (ROG)?
(2) Where would reductions in emissions
of nitrogen oxides (NOx) help reduce
ozone, and where would they be
counterproductive?
(3) What benefits could be gained through
control strategies such as substitut-
ing methanol-based fuel for gasoline,
which lower the reactivity of ROG
emissions, rather than cutting the
overall mass of emissions?
For the Northeast, the questions listed
above have been addressed in the Re-
gional Oxidant Modeling for Northeast
Transport (ROMNET) study conducted by
the U.S. Environmental Protection Agency
and state and local agencies within the
region. The study used EPA's three-di-
mensional, regional scale photochemical
grid model, the Regional Oxidant Model
(ROM), to simulate the air quality impacts
of a variety of control strategies for ROG
and/or NOx emissions. This study extends
the analysis of the ROMNET results, fo-
cussing on understanding the chemical
interactions (as predicted by ROM) that
underlie the impact of alternative control
measures on air quality.
Used in predicting how air quality will
respond to emissions reductions, the role
of a photochemical air quality model is to
integrate descriptions of the fundamental
physical and chemical processes that gov-
ern ozone formation, destruction, and
transport. The physicochemical system that
produces ozone on regional and urban
scales is extremely complex, and many
features of the system are difficult, if not
impossible, to elucidate empirically. De-
spite well recognized limitations (notably,
inadequate evaluation by comparison with
observations, but also including the ex-
pense and time required to apply them),
photochemical air quality models repre-
sent the best means currently available to
predict how air quality might respond to
altered emissions.
Because of the limitations of the mod-
els, however, considerable interest exists
in finding observable "indicators" of how
real air masses are likely to respond to
emissions controls (e.g., of NOx versus
ROG-based strategies). Examining model
predictions of the associations between
chemical species, and the chemical con-
ditions under which controls are predicted
to be more or less effective, can help to
identify possible indicators. Once key as-
sociations have been ferreted out from
model results, the next step is to look for
them in field observations.
The study reported here has examined
the predicted distributions and interactions
of chemical species in the ROM simula-
tions, which underlie predictions of how
concentrations of ozone and other sec-
ondary species might respond to controls.
The analysis has explored relationships
between chemical species, and how pre-
dicted concentrations of ROG classes,
radicals, NOx, NO , etc., differ from one
location to another,vand change from simu-
lations with base case emissions to simu-
lations of controlled emissions levels. The
associations between chemical species
seen in the model outputs result from the
modeled interactions of chemical reactions,
transport processes, emissions distribu-
tions, etc., and therefore are not neces-
sarily apparent in the input data or in
theoretical descriptions of individual pro-
cesses. Some of the associations pre-
dicted by the model should be observ-
able, and some suggest new strategies
for predicting the effect of controls, evalu-
ating models, and monitoring the progress
of control efforts.
Discussion and Conclusions
Summary
A key assumption underlying this study
is that photochemical air quality models
such as ROM, despite their limitations,
are useful tools for integrating existing un-
derstanding of the complex array of physi-
cal and chemical processes that act to-
gether to form, transport and destroy sec-
ondary pollutants. In particular, models are
a primary tool for investigating how sec-
ondary pollutants might respond to altered
precursor emissions rates. We thus hoped
that detailed examination of ROM results
showing the chemical conditions under
which alternative control strategies were
predicted to be effective might suggest
ways in which air quality monitoring pro-
grams could be improved, to better assist
in evaluating models and ultimately in pre-
dicting the effects of control strategies.
Within the ROMNET simulation period
of July 2-17, 1988, July 8 was singled
out as the focus of this analysis, as repre-
sentative of a high ozone day. With me-
teorological conditions for the July 1988
episode but with emissions projected to
the year 2005, predicted ozone concen-
trations in the ROMNET domain on July 8
peaked at around 250 ppb, in New York
City. To reduce ozone concentrations, the
ROMNET study predicted that reducing
NOx emissions by about 60% would be
more effective, for most of the model do-
main, than reducing anthropogenic ROG
emissions by approximately the same pro-
portion. Combining the ROG and NOx re-
ductions was predicted to yield similar re-
ductions in ozone to those achieved with
NOx controls alone. However, countervailing
results were predicted for portions of the
New York City and Baltimore-Washington
metropolitan areas, where peak ozone con-
centrations on July 8 were higher under
the NOx control strategy than in the base
case.
The response of PAN concentrations to
the ROG or NOx control strategies was
predicted to be similar to that for ozone.
NOx controls were predicted to be uni-
formly most effective in reducing HNO3;
ROG controls in reducing H2O2 and form-
aldehyde.
Two additional ROMNET scenarios that
we examined were designed to predict
the effect of measures that reduce the
reactivity of volatile organic compound
emissions without substantial reductions
in the mass of emissions. Results for
ozone in the CS20 scenario, in which the
reactivity adjustment was the only change
from the 2005 base case, were similar to
those of CS12, with substantial reductions
limited to the New York City area. Al-
though unweighted ROG concentrations
in urban areas were 30 to 40% higher in
CS20 than in CS12, reactivity weighted
ROG concentrations in CS20 generally
fell within 10 to 20% of CS12 levels.
Because predicted responses in New
York City were so different from those
occurring elsewhere, time series of pre-
dicted concentrations were examined along
an air mass trajectory ending in the core
of the city at 4 p.m. on July 8. The New
York City core was characterized by ex-
tremely high ROG and NO concentrations,
relative to the rest of they model domain.
Based on the time series of precursor and
radical concentrations along the New York
City trajectory, one factor in the sharp
drop in ozone concentrations in the area
that results from ROG controls appears to
be an increase in the lifetime of NOx. With
reduced ROG emissions, OH concentra-
tions are also reduced, and consequently
the conversion of NOx to inactive forms of
NOy is slowed. ROG controls thus appear
to extend the period during which ozone
levels are suppressed by high NOx emis-
sions.
Preliminary analysis of the ROMNET
results looked at associations of ozone
levels with concentrations of NOx, NO ,
ROG and ROG/NOx in the same grid cell.
Consistent with the findings of Sillman et
al. (1990), the association of ozone with
NOy showed the clearest trends with the
least scatter, supporting its use as a refer-
ence variable. As discussed by Sillman et
al., NO has a similar lifetime to that of
ozone, vand reflects time-integrated NOx
emissions received into an air mass.
-------
Scatter plots of 4 p.m. ozone versus 4
p.m. NO concentrations in each grid cell
for the ROMNET 2005 base case and
for the ROG (CS12) and reduced reactiv-
ity (CS20) control cases are consistent in
indicating that ozone increases with NO
concentrations up to 10 - 15 ppb; then
increases or decreases with NO depend-
ing on the associated ROG levels. Scatter
plots of the change in ozone from the
base case to the NOx control case (CS11)
and from CS12 to the combined control
case (CS10) indicate that NOx controls
are uniformly beneficial for grid cells with
4 p.m. NO concentrations less than about
25 ppb. The NOx control scenarios were
predicted to be counterproductive for some
grid cells with NO concentrations above
that level. Ozone concentrations were rela-
tively insensitive to reductions in ROG
emissions in grid cells with NO concen-
trations below 5 to 10 ppb. The associa-
tions of PAN concentrations and of the
response to NOx controls with NO were
qualitatively similar to those found for
ozone.
Discussion
Results from the ROMNET study for
the association of ozone with NOy, and
specifically the threshold NO concventra-
tions below which NOx controls are uni-
formly beneficial, appear to be consistent
with results from several other modeling
studies (Sillman et al., 1990; Sadeghi et
al., 1992; McKeen et ai, 1991; Milford et
al., 1992). The fact that this association is
so consistent across modeling studies sug-
gests that NOy measurements may be a
useful empirical indicator for the sensitiv-
ity of ozone to emissions.
It is well known that the balance be-
tween ROG and NOx levels controls the
sensitivity of ozone to precursor emissions
reductions, with ROG/NOx ratios used to
characterize this balance. Problems with
the use of ROG/NOx ratios as empirical
indicators of ozone sensitivity have been
noted in Section 2.2 of the full report.
Reflecting those problems, preliminary
analysis of the ROMNET results showed
that the response of ozone to ROG or
NOx controls was more consistently re-
lated to simultaneous NO concentrations
than to ROG/NOx ratios. v
As a measured cumulative NOx emis-
sions, NO may serve as an adequate indi-
cator of whether an air mass is ROG or
NOx-limited in part because effective ROG
levels are relatively uniform (Chameides
et al., 1992). Figure 34 in the full report
was notable in supporting this argument,
showing the association of propylene-
equivalent ROG concentrations with
unweighted ROG in the range that en-
compasses the concentrations in most of
the urban grid cells in the ROMNET do-
main, except those in the New York City
area. No correlation is apparent between
these two measures of ROG levels. In
most urban grid cells outside of New York
City, even though unweighted ROG con-
centrations were relatively high, Propy-
Equiv ROG concentrations were close to
the domain average. Gradients of NO
concentrations are generally much sharper
than those of Propy-Equiv ROG.
Our analysis suggests the possibility that
measured NOy concentrations could be
used as indicators of the photochemical
sensitivity of air at times of high ozone
and as a critical test of model perfor-
mance. Establishing a network of NO
monitors would permit evaluation of mod-
els such as ROM against a variable that
is directly related to the simulated effec-
tiveness of ROG versus NO control
strategies, thus increasing confidence in
the use of the models. However, before
such a network could be created, several
issues need to be addressed.
Although measurement techniques for
total NO have performed well in research
applications (Fehsenfeld et al., 1987; NRC,
1991), consideration should be given to
how readily these techniques could be
adapted for more widespread use. An-
other critical issue is monitor siting and
sample timing. Use of NO measurements
as indicators of ozone sensitivity would
require that monitoring locations be re-
moved from direct emissions of NOx, and
that sampling coincide with peak ozone
concentrations.
A final question is the correspondence
between modeled and measured NO . The
CBIV mechanism (Geryetal., 1988; V1989)
included in ROM uses a condensed treat-
ment of organic nitrate species, and ne-
glects gas to aerosol conversions, such
as formation of ammonium nitrate from
nitric acid. The fraction of aerosol nitrate
recovered in NO measurements is un-
known. Moreover, there is some "lost" ni-
trogen in the CBIV mechanism, in that
nitrogen-containing products of some re-
actions are not tracked. According to
Trainer et al. (1991), during the summer
of 1986, NO, NO2, HNO3 and PAN ac-
counted for 85% or more of the NO mea-
sured at their field site in rural Pennsylva-
nia. Thus, for the rural to urban conditions
found in the northeastern U.S., discrepan-
cies between measured and modeled NO
are probably not large. Nevertheless, the
sources of potential discrepancies war-
rant further study.
Recommendations
Further investigation is warranted of the
idea that NO concentrations could serve
as an empirical indicator of the likely sen-
sitivity of ozone to ROG or NOx con-
trols, if monitored along with peak ozone
during photochemical air pollution epi-
sodes. Assessment of the adaptability of
NOy measurement techniques for routine
use in urban areas is needed, along with
detailed evaluation of errors introduced by
the simplified treatment of reactive NO
species in ROM and other photochemical
air pollution models.
To recommend that the Regional Oxi-
dant Model should be evaluated more ex-
tensively, especially with regard to its per-
formance for ozone precursors, is ap-
proaching a cliche. However, the depen-
dence of predicted ROG or NOx control
effectiveness on NO levels displayed by
the ROM results underscores the impor-
tance of this recommendation. To empha-
size the point, the ability of the model to
predict responses to alternative control
strategies is tied to its ability to simulate
grid cell averaged NO levels. With re-
spect to ROG levels, the lack of spatial
correlation between ROG and Propy-Equiv
ROG concentrations suggests the need
for speciated ROG measurements, rather
than measurements of total ROG. The
model results also point to the importance
of measuring concentrations of carbonyl
species.
Finally, a targetted field experiment
would be of interest to verify the ROMNET
result that in the New York City area, in
association with peak ozone concentra-
tions, very high ratios of NOx to NO persist
throughout the day. If observed, unusually
large NOx fractions would support the pre-
diction that ROG emissions reductions will
be highly effective for the New York City
area.
References
Chameides, W.L., F. Fehsenfeld, M.O.
Rogers, C. Cardelino, J. Martinez, D.
Parrish, W. Lonneman, D.R. Lawson,
R.A. Rasmussen, P. Zimmerman, J.
Greenberg, P. Middleton, and T.
Wang. Ozone precursor relationships
in the ambient atmosphere, J.
Geophys. Res. 97, 6037-6055, 1992.
Fehsenfeld, F.C., R.R. Dickerson, G. Hubler,
W.T. Luke, L.J. Nunnermacker, E.J. Wil-
liams, J.M. Roberts, J.G. Calvert, C.M.
Curran, A.C. Delany, C.S. Eubank, D.W.
Fahey, A. Fried, B.W. Gandrud, A.O.
Langford, P.C. Murphy, R.B. Norton, K.E.
Pickering, and B.A. Ridley. A ground-
*U.S. Government Printing Office: 1993 — 750-071/60257
-------
based intercomparison of NO, NOx,
and NO measurement techniques, J.
Geophys. Res., 92: 14710-14722,
1987.
Gery, M.W., G.Z. Whitten, and J.P. Killus.
Development and Testing of the CBM-
IV for Urban and Regional Modeling,
EPA-600/3-88/012, U.S. Environmen-
tal Protection Agency, Research Tri-
angle Park, NC, 1988.
Gery, M.W., G.Z. Whitten, J.P. Killus, and
M.C. Dodge. A photochemical kinet-
ics mechanism for urban and regional
scale computer models, J. Geophys.
Res. 94. 12925-12956, 1989.
McKeen, S.A., E.-Y. Hsie, and S.C. Liu. A
study of the dependence of rural
ozone on ozone precursors in the
eastern United States, J. Geophys.
Res. 96: 15377-15394, 1991.
Milford, J.B., D. Gao, S. Sillman, P.
Blossey, and A.G. Russell. NO as an
indicator of the sensitivity of ozone to
ROG and NOx emissions, paper in
preparation, 1992.
National Research Council. Rethinking the
Ozone Problem in Urban and Re-
gional Air Pollution, National Acad-
emy Press, Washington DC, 1991.
Sadeghi, V.M., L. Bruckman, H.W.
Balentine, B.J. Morrison, S. Coerr, and
I.H. Billick. Evaluation of the efficacy
of NOx controls in ozone
nonattainment areas under section
182(f) of the 1990 Clean Air Act
Amendments, Paper No. 92-89.03,
presented at the 85th Annual Meet-
ing of the Air and Waste Manage-
ment Association, Kansas City, MO,
June 21-26, 1992.
Sillman, S., J.A. Logan, and S.C. Wofsy.
The sensitivity of ozone to nitrogen
oxides and hydrocarbons in regional
ozone episodes, J. Geophys. Res.
95: 1837-1851, 1990.
Trainer, M., M.P. Buhr, C.M. Curran, F.C.
Fehsenfeld, E.Y. Hsie, S.C. Liu, R.B.
Norton, D.D. Parrish, E.J. Williams,
B.W. Gandrud, B.A. Ridley, J.D.
Shetter, E.J. Allwine, and H.H.
Westberg. Observations and model-
ing of the reactive nitrogen photo-
chemistry at a rural site, J. Geophys.
Res. 96: 3045-3063, 1991.
Jana B. Milford, Dongfen Gao, and Antigoni Zafirakou are with the University of
Connecticut, Storrs, CT 06269.
Thomas E. Pierce is the EPA Project Officer (see below).
The complete report, entitled "Relationships Between Ozone Precursor Levels and
Response to Emissions Reductions: Analysis of Regional Oxidant Model Results
for the Northeastern United States," (Order No. PB93-186 294/AS; Cost: $27.00,
subject to change) will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Atmospheric Research and Exposure Assessment Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
United States
Environmental Protection Agency
Center for Environmental Research Information
Cincinnati, OH 45268
BULK RATE
POSTAGE & FEES PAID
EPA
PERMIT No. G-35
Official Business
Penalty for Private Use
$300
EPA/600/SR-93/075
------- |