Environmental Protection
Agency
Office of
Development
DC
vvEPA
Evaluating the Relation
Between Ozone, NOX and
Hydrocarbons: The
Method of Photochemical
Indicators
-------
EVALUATING THE RELATION BETWEEN OZONE, NOX AND
HYDROCARBONS: THE METHOD OF PHOTOCHEMICAL INDICATORS
by
Sanford Sillman
Department of Atmospheric, Oceanic and Space Sciences
University of Michigan
Ann Arbor, Michigan 481Q9-2143
Cooperative Agreement No. CR822083-01
Project Officer
James M. Godowitch
Atmospheric Modeling Division
National Exposure Research Laboratory
Research Triangle Park, NC 27711
NATIONAL EXPOSURE RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
RESEARCH TRIANGLE PARK, NC 27711
-------
NOTICE
The information in this document has been funded wholly or in part by the United
States Environmental Protection Agency under Cooperative Agreement No. CR822083-01
to the University of Michigan. It has been subjected to the Agency's peer and
administrative review, and it has been approved for publication as an EPA document.
Mention of trade names or commercial products does not constitute an endorcement or
recommendation for use.
ACKNOWLEDGMENTS
The work represented here was funded by Cooperative Agreement No. CR822083-
01 with the U. S'. Environmental Protection Agency. Results for the UAM simulations for
New York and Los Angeles were provided by Jim Godowitch and Neil Wheeler, each of
whom made many helpful suggestions. I also wish to thank Neil Wheeler and the
California Air Resources Board for providing me with measurments in the Los Angeles
basin, Bob Imhoff and his colleagues in the Southern Oxidant Study for providing me with
measurements from Atlanta, and Pete Daum and his colleagues for sharing their NARE
results with me. Discussions with Jana Milford and Tom Ryerson were also especially
helpful.
-------
ABSTRACT
The method of photochemical indicators is a way to evaluate the sensitivity of O3 to
its two main precursors, nitrogen oxides (NOX) and reactive organic gases (ROG), directly
from ambient measurements. The method is based on identifying measureable species or
species ratios that are closely associated with O3-NOX-ROG predictions in photochemical
models. Several species of this type have been identified: Os/NOy (where NOy is total
reactive nitrogen), Os/NOz (where NOz=NOy-NOx), O3/HNO3, H2O2/HNO3,
H2O2/NOZ, and H2O2/NOy. In each case, high values of the proposed indicator are
associated with NOx-sensitive chemistry in models and low values are associated with
ROG-sensitive chemistry. This report presents a summary of O3-NOX-ROG sensitivity in
models and the chemistry that motivate the choice of species as photochemical indicators.
It shows the correlation between model NOX-ROG predictions and indicator values for a
variety of models, including simulations for Lake Michigan, the northeast corridor, Atlanta
and Los Angeles. The indicator-NOx-ROG correlation remains consistent in model
scenarios with radically different assumptions about anthropogenic and biogenic emission
rates and meteorology and in models with different chemical mechanisms. The report
shows correlations between 63, NOZ and H2O2 in models and in ambient measurements,
which provide a basis for evaluating the accuracy of critical assumptions associated with
the indicator method. Case studies are described for Atlanta and Los Angeles in which
measured values of indicator ratios were used as a basis for evaluating model scenarios. It
is shown in which model scenarios with different NOX-ROG predictions often give similar
values for peak 03 but different values for indicator ratios. The case studies illustrate how
comparisons between model results and measured indicator values can be used as a basis
for model evaluation. Uncertainties associated with the indicator method (including
measurement uncertainties, dry deposition and surface effects) are discussed.
in
-------
CONTENTS
Notice [[[ ii
Acknowledgments [[[ ii
Abstract [[[ iii
Figures [[[ v
Tables [[[ vi
1. Overview [[[ 1
2. Chemistry of 03, NOX and ROG ........................... . .......................... 9
2.1 Factors affecting O3-NOX-ROG sensitivity in urban locations ............. 10
2.2 Chemistry of 03, NOX and ROG and its relation to indicators ............. 15
3. Results from photochemical simulations.... ...................................... ,.24
3.1 Description of photochemical simulations .............. ................. .....25
(a) Lake Michigan regional simulations . . . . . ................ 25
(b) Northeast corridor regional simulations ................. 26
(c) New York urban simulations... ........ ....... . ..... ,...,27
(d) Atlanta urban simulations. ........... ..... ................ 28
(e) Los Angeles urban simulations ...... ..... ............ ....29
(f) Nashville simulations (preliminary) ................... .; . . 30
3.2 Results [[[ 31
03 03 .O3
31
H202 H202 . H202
N0y~
(c) Other indicators ............................................. 53
4. Species cross-correlations as a basis for evaluating the
-------
FIGURES
Number
2-1 Isopleths showing net rates of ozone production ............................. 21
2-2 Predicted ozone response at locations in the Los Angeles basin ............ 22
2-3 O3-NOX-ROG response vs. NOy .............................................. 23
3-1 O3-NOX-ROG response vs. O3/NOZ .......................................... 38
3-2 O3-NOX-ROG response vsO3/NOy ........................................... 41
3-3 Cumulative percentile distribution of O3/NOZ ................................. 41
03 03 03
3-4 95th and 5th percentile values for MQ~. NO"3™* HNOT ................ ^
3-5 O3-NOX-ROG response for ............................................ 47
3-6 O3-NOX-ROG response for -j=p ............................................ 50
.3-7 95th and 5th percentile values for TTXTQT , -J^Q— and ^^j-= ............. 51
3-8 O3-NOX-ROG response for (O3-40ppb)/NOy ................ . ............. 56
3-9 O3-NOX-ROG response for O3/NOX ........ .......... ........ . .............. 56
3- 1 0 O3-NOX-ROG response for ROG/NOX and ROG/NOy .............. ...... 57
3-11 95th and 5th percentile values for other indicators ..................... . ...... 58
4-1 Measured 03 vs. NOZ at four rural sites ............ ..... ...................... 64
4-2 Model 03 vs. NOZ for NOX- and ROG-sensitive locations ..... ........... ; '64
4-3 Model correlation between 03, NOZ and H2O2 .......... ..... ........ ...... 65
4-4 03 vs. NOZ from aircraft measurements over the Atlantic Ocean . .......... 66
4-5 NOZ+2H2O2 vs. 03 from aircraft measurements over the Atlantic. ....... 66
5-1 Measured 03 vs. NOy in Atlanta ...... ................................. ... ..... 72
5-2 Peak 03 and concurrent NOy in Atlanta models and measurements ........ 73
5-3 Measured 03 vs. NOy and 03 vs. NOZ in Los Angeles ..................... 78
5-4 03 vs. NOZ in model scenarios for Los Angeles ............................. 79
-------
TABLES
Number
03 03 03
3-1 NOX- and ROG-sensitive values for — , -- and T ............. 43
H7O2 H?O? H7O?
3-2 NOX- and ROG-sensitive values for ^Q , NQ and NQ ......... 52
3-3 NOX- and ROG-sensitive values for for other indicators .................... 59
5-1 03, NOy and NOX-ROG sensitivity in Atlanta ................................ 74
VI
-------
SECTION 1
OVERVIEW
For many years there has been uncertainty about the relation between low-level
ozone and its two major anthropogenic precursors: reactive organic gases (ROG) and
nitrogen oxides (NOX). It is generally known that for some conditions ozone
concentrations increase with increasing NOx emissions and are largely independent of
ROG, while for other conditions ozone increases with ROG and does not increase (or may
even decrease) with NOX. However, it is difficult to determine whether individual air
pollution events are dominated by NOx-sensitive or ROG-sensitive chemistry.
The uncertain relationship between ozone, ROG and NOX has made it especially
difficult to develop effective control policies for ozone. Ozone as an air pollutant is
responsible for widespread violations of air quality standards, affecting most major
metropolitan areas in the U.S. A large effort has been made to reduce emission of ozone
precursors, but it is unclear whether these efforts have been effective in controling ozone
levels (NRC, 1991). Because of the complex relationship between ozone and its
precursors, there is far greater uncertainty about the effectiveness of control strategies for
ozone than for other pollutants associated with air quality violations. A control strategy that
relies on reductions in emissions of ROG will not be effective in a region where ozone
concentrations are driven by NOx-sensitive chemistry. Similarly, a control strategy that
relies on NOx reductions will not be effective during air pollution events with ROG-
sensitive chemistry.
The task of analyzing the ozone-NOx-ROG relationship is commonly based on
predictions from photochemical models. These models use estimates for emission rates of
ozone precursors in combination with meteorological information (wind speeds, vertical
diffusion, temperatures, etc.) and a representation of ozone chemistry in order to simulate
the ozone formation process for a specific air pollution event. Predictions for the
effectiveness of ROG vs. NOX controls are derived by repeating the simulations with
reduced emission rates for ROG or NOX. The accuracy of these control strategy
predictions has frequently been called into question, largely because of uncertainties
-------
associated with emission inventories. Inventories for anthropogenic ROG emissions may
be underestimated by up to a factor of two (Fujita et al., 1992), while recent research has
shown that the emission rate for isoprene, the most important biogenic ROG, is
approximately four times higher than previous estimates (Geron et al., 1994, 1995). Both
of these uncertainties have a large impact on model predictions of the effectiveness of ROG
vs. NOX controls. It is frequently possible to generate model scenarios with either NOX-
and ROG-sensitive chemistry while remaining within the bounds of uncertainty associated
with emissions and other model input.
A more fundamental problem with the model-based approach for O3-NOX-ROG
sensitivity is the difficulty in evaluating model predictions against real-world
measurements. There is no direct way to evaluate a control strategy prediction, because it
is never possible to repeat an air pollution event with reduced ROG or NOX emissions and
compare its outcome with model results. Model evaluations can only be done by
comparing predictions from the model base case with ambient measurements. The problem
with most model evaluations is that the accuracy of model predictions for the base case
does not guarantee that model predictions for control strategies are also accurate. In
particular, the most common test of model performance - comparison with ambient ozone -
gives no basis for confidence in model predictions for control strategies. This is because
different model scenarios often give similar values for ozone in the base case, while giving
very different predictions for the response of ozone to reductions in ROG or NOX.
This report presents a new approach to the problem of evaluating Qs-NOx-ROG
sensitivity: the method of photochemical indicators. The goal of the method is to identify
measureable species or species ratios that are closely linked to model control strategy
predictions, so that NOx-sensitive or ROG-sensitive chemistry can be associated with
specific values of these species ratios. If successful, the method of .photochemical
indicators would provide a way to evaluate NOX-ROG sensitivity based solely on ambient
measurements, without using models. A more modest application of the method would be
to evaluate the performance of model scenarios. The nature of the indicator ratios
associated with this method is that they will always assume different values in models with
NOx-sensitive chemistry as opposed to models with ROG-sensitive chemistry. Thus, the
indicator ratios (unlike ozone) provide a basis for choosing between NOx-sensitive and
ROG-sensitive model scenarios.
The method of photochemical indicators is part of a broader attempt to develop
observation-based methods (OEMs) that seek to analyze O3-NOX-ROG chemistry based on
extensive field measurements rather than model predictions (e.g. Cardelino et al., 1995).
-------
The method of photochemical indicators was developed based on a series of photochemical
model applications designed to represent a wide range of physical conditions and
situations. These have included simulations for four metropolitan regions: the Lake
Michigan airshed, the New York-Boston urban corridor, Atlanta and Los Angeles. They
have included simulations with anthropogenic ROG emissions doubled from inventory
estimates, reflecting uncertainty in inventories identified by Fujita et al. (1992). They have
included scenarios with anthropogenic ROG emissions reduced by half from inventory
estimates, reflecting conditions that may occur in Europe today or in the U.S. following the
application of an effective ROG control program. They have included scenarios with
biogenic ROG emissions based on the current BEIS2 inventory (Geron et al., 1994) and
also on the older BEIS1 inventory with much lower emission rates (Pierce et al., 1990),
and scenarios with zero biogenic emissions. They have included simulations with changed
wind speeds and vertical mixing rates in order to represent conditions that might be either
more or less stagnant than the events used as test cases. They have included results from
two very different model types: a regional-scale model developed by Sillman et al. at the
University of Michigan (Sillman et al., 1993) and the Urban Airshed Model (UAM-
rV)(Morris and Myers, 1990), familiar to air pollution researchers. These two models
include different representations of photochemistry: a mechanism based on Lurmann et al.
(1986) (similar to the more familiar SAPRC mechanism) with various updates and
isoprene chemistry based on work by Paulson and Seinfeld (1992) in the Michigan model;
and the Carbon Bond IV mechanism (Gery et al., 1989) in UAM-IV.
Species and species ratios that can be used as indicators for NOX-ROG sensitivity
were sought based on an analysis of model results. In order to be useful as a
photochemical indicator a species or species ratio must satisfy the following four criteria:
(i) The indicator must involve measureable species.
(ii) The indicator must consistently show different values in model scenarios with
NOx-sensitive chemistry as opposed to model scenarios with ROG-sensitive chemistry.
The difference between NOx-sensitive and ROG-sensitive indicator values must also be
large enough so that the NOX-ROG indication is less likely to be obscured by
uncertainties in measurement techniques.
(iii) The correlation between indicator values and NOX-ROG sensitivity must remain
reasonably constant in models with widely different assumptions. The purpose of the
method would be undermined if the indicator correlation were limited to models with a
fixed range of assumptions, e.g. about biogenic ROG.
-------
(iv) The indicator should be closely linked with the chemical factors that govern NOX-
ROG sensitivity. It should not just represent an empirical curve-fitting exercise or be
an artifact of an individual model.
Two classes of species have been identified that appear to meet these criteria. These
species are as follows.
A. Ratios involving ozone and total reactive nitrogen: J^Q~, where NOy represents total
reactive nitrogen (NOX, PAN, HNOs, alkyl nitrates and other nitrogen-containing species
03
produced from NOX); MQ~. where NOZ represents NOX reaction products, NOy-NOx;
O
B. Ratios involving hydrogen peroxide and total reactive nitrogen: \f(H2O2,HNC>3),
H202 .H202
NOZ ' ana NOy '
In each case, a high value for the indicator ratio corresponds to a situation in which
the ozone concentration is primarily sensitive to NOX, while a low value corresponds to a
situation in which 03 is primarily sensitive to ROG. This interpretation applies only for
indicator values during the afternoon (concurrent with peak 03) and only for NOX-ROG
sensitivity at the same location as the indicator value.
Among these alternatives, the ratios involving hydrogen peroxide (especially
HNQ.J provide more consistent results in photochemical models and are more closely
linked.to the photochemical processes that drive NOX-ROG chemistry, but these are also
more difficult to measure. Among the ratios involving ozone, T- has a stronger link with
03
NOX-ROG chemistry than -, but the latter has certain advantages associated with
situations involving power plants.
An important component of the indicator method concerns the need to evaluate the
accuracy of the indicator predictions. The correlation between NOX-ROG sensitivity and
the indicator ratios is based entirely on the results of photochemical models. As such, the
indicator method faces the same type of challenge as the conventional photochemical
models: how can the predicted NOX-ROG sensitivity be proven?
Model results have suggested an important test for the validity of the method. The
chemistry associated with NOX-RQG sensitivity and the indicator ratios also suggests that a
linear correlation should be found between 03 and the sum: NOz+2H2O2, evaluated
during the afternoon following a period of photochemical activity. This linear correlation is
predicted in three-dimensional models that include both chemistry and transport, and is
-------
shown in Section 4. The correlation is closely associated with the role of 03, NOZ and
H2O2 as indicators for NOX-ROG sensitivity. It is well-known that 03 increases with
NOZ but that the rate of increase (dO3/dNOz) decreases at higher NOZ and is lower in some
urban areas (e.g. Los Angeles) relative to rural sites (Trainer et al., 1993). The indicator
method provides an additional interpretation: high O3/NOZ ratios in rural areas represent
NOx-sensitive chemistry while low O3/NOZ ratios in central Los Angeles represent ROG-
sensitive chemistry. If this interpretation is valid, then the correlation between 03 and the
sum NOZ+2H2O2 will remain linear, even while the 03-NOZ slope varies. Thus, an
examination of measured 03, NOZ and H2O2 provides both a general evaluation of the
indicator method and a test for problems (e.g. measurement errors or erroneous model
assumptions) in each individual application of the method.
In summary, the indicator method provides several advantages over conventional
methods of analysis of O3-NOx-ROG sensitivity. Unlike conventional photochemical
models, the indicator predictions are not sensitive to assumptions about emission rates for
anthropogenic or biogenic ROG. When used in combination with photochemical models
the indicator method provides a meaningful evaluation of the accuracy of model NOX-ROG
predictions. Finally, the indicator method includes a diagnostic test for its own accuracy,
based on the predicted linear correlation between 03, NOZ and H2O2- The indicator
method can be applied based on measurements for just three species, O3,.NOX and NOy,
although the inclusion of measurements of H2O2, HNO3 and CO would improve the
reliability of the method.
The indicator method, like any new approach, involves problems and uncertainties
that are not present in other methods. The following is a list of caveats for researchers who
seek to use the indicator method.
(1) The indicator method provides information about NOX-ROG sensitivity only at the time
and place of the measurements. It is well known that NOX-ROG sensitivity varies with
location within a metropolitan area, with ROG-sensitive chemistry more likely in the urban
center and NOx-sensitive chemistry more likely downwind (Milford et al., 1989). Thus, it
is inaccurate to infer NOX-ROG sensitivity for an entire metropolitan area from
measurements at a small number of locations. NOX-ROG sensitivity can also vary from
event to event and changes with time of day. Illustrations of this will be shown in the
report below.
(2) The indicator method as presented here is designed to estimate NOX-ROG sensitivity
associated with ozone concentrations, not the instantaneous rate of ozone production. The
NOX-ROG sensitivity of ozone concentrations depends on photochemistry over an
extended time period (typically 1-2 days) and transport from a significant distance upwind.
-------
This impact of upwind transport and photochemistry is included in the models that were
used to derive the indicator correlations.
(3) The indicator method is critically dependent on the accuracy of measurements,
especially NOy. Luke and Dickerson (1987) have warned that certain types of
measurement systems characteristically underestimate NOy, because HNOs (a major NOy
component) is deposited on the inlet tube. This type of measurement error would
invalidate the use of NOy as a NOX-ROG indicator.
Model results suggest that values for Os/NOy and Os/NOz associated with NOX-
and ROG-sensitive chemistry differ by a factor of two, and values for H2O2/HNO3 and
other ratios with peroxides show an even larger difference between NOX- and ROG-
sensitive locations. Errors in measurements are likely to be important only if they are
significant in comparison with this range of variation, i.e. 25% or higher. However errors
in measurements are especially troublesome if they might lead to a systematic overestimate
or underestimate in measured indicator ratios, since these would cause bias in the indicator
NOx-ROG interpretation.
Because of potential problems with NOy measurements it is important to carefully
evaluate measurement accuracy. One possible method for evaluating the accuracy of
measured NOy is to include simultaneous measurements of CO and SO2 and examine
correlations of these species with NOy. Previous studies have identified a correlation
between these species, reflecting their common anthropogenic origin (Panish et al., 1991,
Buhr et al., 1995). A deviation from the expected correlation between these species might
be used to identify erroneous measurements.
Because these three species have a common anthropogenic origin, it should be
possible to identify a consistent correlation between them, which could then be used to
evaluate the accuracy of individual NOy measurements.
(4) Species ratios that use NOX, PAN or NOX+PAN in place of NOy, NOZ or HNO3 do
not correlate well with NOX-ROG sensitivity and should not be used as photochemical
indicators. This is a problem because NOy measurements that rely on older techniques
have been found to represent the sum NOX+PAN rather than true NOX or NOy (Logan,
1989). Recent reports have also erroneously used the ratio O3/NO2 as a NOX-ROG
indicator (LADCO, 1994). Also, species ratios must be based on simultaneous
measurements (rather than 03 in the afternoon versus NOy in the morning or at a different
location).
(5) Many of the species associated with the indicator method (NOy, NOZ, H2O2 and
HNO3) can be rapidly removed from the atmosphere (and removed at varying rates) by
surface deposition or deposition on aerosols. This may represent a major uncertainty for
-------
the method. Correlations between NOX-ROG senstivity and indicator values may change
significantly if changes are made to model surface deposition rates. In addition, surface-
level measurements at night cannot be used as NOX-ROG indicators because the indicator
concentrations at night are dominated by surface deposition processes.
(6) The indicator method was derived from models that did not include aerosol nitrates
(NO3~) which are formed from gas-phase HNO3. It is therefore appropriate to include
aerosol nitrate (converted to gas-phase equivalent units) in the sum for NOz, and interpret
HNO3 in the indicator ratios as representing the sum of HNO3 and NO3-. Failure to
include aqueous chemistry represents an additional source of uncertainty in the method.
(7) Surface measurements of NOy are sometimes dominated by NOX sources close to the
measurement site. This would interfere with the direct use of NOy in indicator ratios (e.g.
O3/NOy or H2O2/NOy). This problem can be avoided by using NOZ (based on
simultaneous measurement of NOX and NOy) instead of (or in addition to) NOy.
This report presents a summary of research on the indicator method to date.
Section 2 provides background information on the chemistry of 03, NOX and ROG that
serves as a motivating factor for the selection of the various indicator ratios. This section
also presents a concise summary of the current understanding of O3-NOX-ROG sensitivity
that has emerged in recent years. The state of science associated with O3-NOX-ROG
chemistry has seen great changes in the past ten years. The summary in Section 2 may
serve both as an introduction to the field and a review for experienced researchers and
regulators.
Section 3 presents the correlation between O3-NOX-ROG predictions and the
various indicator ratios in photochemical models. It includes a description of the models
used, general results for O3-NOX-ROG sensitivity, the extent of variation of the indicator-
NOX-ROG correlation in different model scenarios, and methods for quantitatively
identifying the indicator values corresponding to the transition between NOX- and ROG-
sensitive chemistry. It also shows results for alternative choices (NOy, AIRTRAK,
O3/NOX, ROG/NOX) which do not perform as well as the recommended indicator ratios.
A complete presentation of model results is also included in the Appendix.
Section 4 examines the correlation between 03, NOZ and H2O2 as predicted by
photochemical models and as observed during field measurement campaigns. As described
above, this correlation represents a critical test for the accuracy of the indicator method.
Section 5 shows results from applications of the indicator method for specific events in
Atlanta and Los Angeles. These results include comparisons between predicted indicator
values from a series of model scenarios, each designed to give different predictions for
NOX-ROG chemistry. These predictions are compared with measured indicator species
7
-------
and ratios, which are used as a basis for accepting the results of some model scenarios and
rejecting others. The comparison between model and measured values of photochemical
indicators associated with predicted and observed (non-paired) peak 03 is proposed as a
criterion for evaluating model performance. This criterion provides a much stronger basis
for model evaluation then criteria based solely on model vs. measured 03. It is hoped that
these case studies can be used as examples for future applications.
The contents of this report are based models and measurements for the northeast
corridor and Lake Michigan (Sillman, 1995a), Atlanta (Sillman et al., 1995b, 1997a), New
York and Los Angeles (1997a). More recent results from the Middle Tennesee Ozone
Study (Sillman et al., 1997b) have been included only in summary form.
8
-------
SECTION 2
CHEMISTRY OF O3, NOX AND ROG
This section includes two parts. The first part provides a summary of the various
factors that can affect O3-NOX-ROG sensitivity, based on results from photochemical
models. The original understanding that NOX-ROG sensitivity is determined by ROG/NOX
ratios has been greatly modified in recent years. Additional factors include the impact of
biogenic ROG, the geographical variation in NOX-ROG chemistry as an air mass ages and
moves downwind from an urban center, and the influence of meteorological stagnation in
causing day-to-day variations in NOX-ROG chemistry. An understanding of these factors
can be of great help to researchers and to regulators who need to interpret the results of
models or other NOX-ROG analyses.
The second part analyses the specific chemical reactions and reaction sequences that
create the division into NOx-sensitive and ROG-sensitive regimes; This section also
provides the theoretical basis for the link between NOX-ROG sensitivity and the identified
indicator ratios. As discussed below, Os-NOx-ROG chemistry is derived from reaction
cycles involving odd hydrogen radicals (OH, HO2 and RO2, where R represents a carbon-
hydrogen chain). NOx-sensitive chemistry occurs when radical-radical reactions
(HO2+HO2 and HO2+RO2, making peroxides) are the dominant sink for odd hydrogen,
while ROG-sensitive chemistry occurs when nitrate-forming reactions (OH+NO2, making
nitric acid) are the dominant sink. From these chemical reaction sequences it is possible to
derive a theoretical relationship between NOX-ROG sensitivity and the species ratios that
have been identified as photochemical indicators: O3/NOZ and H2O2/HNO3. Section 3
will show how the correlation between NOX-ROG sensitivity and photochemical indicators
also appears in more complete photochemical simulations.
A more complete review of the state of science associated with 03, NOX and ROG
is given in Sillman (1997c).
-------
2.1. Factors affecting O3-NOX-ROG sensitivity in urban locations
For purposes of this study the terms "NOx-sensitive" and "ROG-sensitive" will be
defined based on the predicted response of ozone concentrations to a moderate (usually
35%) reduction in anthropogenic emissions of NOx or ROG. Locations will be defined as
NOx-sensitive if the model results with reduced NOx show a greater decrease of 03 than
the same percent reduction of ROG. However it should be recognized that the NOx-
sensitive and ROG-sensitive labels also depend on the size of the reductions. In general,
NOx reductions are more likely to result in reduced 03 if a large percent reduction is
applied. Roselle et al. (1995) have reported cases where a 25% reduction in ROG is
predicted to result in lower 03 than a 25% reduction in NOX, but where a 75% reduction in
NOX results in lower 03 than a 75% reduction in ROG.
The following discussion describes individual factors that have a large impact on
NOX-ROG predictions in models. The discussion refers to the the impact of each
individual factor in isolation; i.e. all other things being equal, a change in the individual
factor will tend to shift model responses in the direction of NOx-sensitive or ROG-sensitive
chemistry. These should not be used to infer NOX-ROG sensitivity for individual
locations, which result from the combination of many factors.
ROG/NOx ratios: It is well known that the division between NOx-sensitive and ROG-
sensitive ozone photochemistry is closely associated with ROG/NOX ratios (e.g. NRC,
1991). The OS'-NOx-ROG relationship is conveiently summarized by isopleth plots, a
form of which is shown in figure 2-1. These plots show that the rate of ozone formation
as a function of NOX and ROG is highly nonlinear. When ROG/NOX ratios are high the
rate of ozone production increases with increasing NOX but shows little sensitivity to ROG
(NOx-sensitive regime). When ROG/NOx ratios are low the rate of ozone formation
increases with increasing ROG and decreases with increasing NOx (ROG-sensitive
regime). Early analyses suggested that ROG-sensitive chemistry could be associated with
ROG/NOx ratio lower than 10 in an urban center during the morning hours while NOx-
sensitive chemistry could be associated with ROG/NOX ratios greater than 20 (Blanchard
etal., 1991).
The view that morning (6-9 a.m.) ROG/NOx ratios can be used to determine NOX-
ROG sensitivity has been largely discredited (NRC, 1991, and references below). The
problems with ROG/NOx concentration ratios as a NOX-ROG indicator include the
following:
10
-------
(i) It fails to account for the impact of biogenic ROG, which are mainly emitted during the
midday and afternoon hours rather than at 6-9 a.m. and which are also disproportionately
reactive (Chameides et al., 1988,1992).
(ii) It fails to account for variations in ROG reactivity, which may occur either due to
variations between urban areas or to changes in estimates of the speciation of emitted ROG
(Chameides et al., 1992, Carter et al., 1994, 1995)
(iii) It fails to account for the geographic variation of NOX-ROG sensitivity within a
metropolitan area (Milford et al., 1989).
(iv) It fails to explain the fact that NOX-ROG sensitivity can vary from event to event, even
for the same metropolitan area. Event-by-event variations in NOX-ROG sensitivity have
been found in photochemical models, even though the ROG7NOX emission ratios and
morning concentrations do not vary (Milford et al., 1994, Roselle et al., 1995).
(v) It fails to account for the impact of multiday transport, which usually has greater
sensitivity to NOX than single-day events (Sillman et al., 1990a).
Despite these flaws, the morning ROG/NOX ratio continues to be used,
inappropriately, to analyze ROG-NOX sensitivity (Hanna et al., 1996). Because of its
repeated misuse, it is important to understand both the rationale for its use (i.e. the isopleth
diagram) and its flaws.
Biogenic ROG and ROG reactivity: Chameides et al. (1988) showed that biogenic
ROG, especially isoprene, have a significant impact on ozone chemistry in urban areas.
More recently, results from the Ozone Transport and Assessment Group (OTAG)
demonstrated that a switch from the older BEIS1 emission inventory to the newer BEIS2
inventory (with four times higher emission rates for isoprene) causes a shift from ROG-
sensitive chemistry to NOx-sensitive chemistry in models for several urban areas (OTAG,
1996). Similar results from Sillman et al. (1995b) for Atlanta, are included in this report.
Two aspects of biogenic ROG are especially important to understand. First,
emission rates for isoprene, the most important biogenic ROG, are zero at night and very
low during the morning. The maximum emission rate occurs during early afternoon
(Geron et al., 1994, 1995). Consequently, analyses based on measured ROG during the
morning hours would seriously underestimate the impact of biogenics. Second, isoprene
and other biogenic ROG are extremely reactive, and their impact on photochemistry is
therefore larger than would be suggested by their atmospheric concentration in comparison
with anthropogenic ROG. When speciated ROG measurements are weighted according to
reactivity or ozone-forming potential (Chameides et al., 1992, Carter et al., 1994, 1995)
the impact of biogenic ROG always appears much larger than it would without reactivity-
weighting.
11
-------
Geographical variation and photochemical aging: A central feature of NOX-ROG
chemistry is the tendency for air parcels to change from ROG-sensitive chemistry to NOX-
sensitive chemistry as they age. Thus, an air parcel frequently has ROG-sensitive
chemistry while it is close to its emission sources and increasingly NOx-sensitive chemistry
as it moves downwind.
Milford et al. (1989) demonstrated the resulting geographical variation in NOX-
ROG chemistry in a model for Los Angeles. As shown in Figure 2-2, model-derived
isopleth plots show strongly ROG-sensitive chemistry for locations near downtown and
NOx-sensitive chemistry at downwind sites. This split between ROG-sensitive downtown
and NOx-sensitive downwind has appeared repeatedly in model applications and is a
central feature of O3-NOX-ROG chemistry.
A similar, better-known pattern appears for power plant plumes. Power plants
typically cause a decrease in 03 in the fresh plume immediately downwind of the power
plant, followed by an increase in 03 relative to the background concentration as the plume
moves further downwind (White et al., 1983). Although the decrease in 03 in the fresh
plume is due to NO titration (via the reaction NO+O3->NO2) rather than NOX-ROG
chemistry, the fresh plume also exhibits ROG-sensitive chemistry with respect to ozone
formation, while the downwind plume with enhanced 03 has NOx-sensitive chemistry.
The evolution towards NOx-sensitive chemistry as air moves downwind is due
largely to the removal of NOX as the air mass ages. NOX has a relatively short
photochemical lifetime (3-6 hours) while many ROG species have lifetimes of 1 day or
longer. Consequently, ROG/NOX ratios always increase as an air mass ages. The
addition of biogenic ROG as the air mass moves downwind also contributes to the
conversion from ROG-sensitive to NOx-sensitive chemistry. The increase in ROG/NOX
ratios with age is illustrated in Figure 2-1, which shows calculated ROG and NOX
concentrations in simulated air parcels, superimposed on isopleths that give the ozone
production rate as a function of ROG and NOX. In these model calculations, air parcels
were initialized with an ROG-NOX mix that would initially exhibit ROG-sensitive
chemistry, but they evolve towards a higher ROG/NOX ratio and NOx-sensitive chemistry
over the simulated 8-hour period.
It is generally accepted that rural locations are characterized by NOx-sensitive
chemistry, except in locations that are directly impacted by urban or power plant plumes
(e.g. Sillman et al., 1990a, McKeen et al., 1991, Roselle et al., 1995). Rural air masses
are typically far downwind from emission sources (except for biogenics) and show the
chemical characteristics of aged air (Trainer et al., 1993), which includes NOx-sensitive
chemistry. Because these air masses frequently have 80-100 ppb ozone they can be
12
-------
associated with regional transport and can have a significant impact on NOX-ROG
chemistry even in downwind urban areas (Sillman et al., 1990a). Similarly, events
involving transport for periods of longer than one day are more likely to be N0x-sensitive.
ROG-sensitive chemistry is more likely in events dominated by local photochemical
production rather than transport.
Meterological stagnation and density of emissions: A little-known feature of
ozone photochemistry is the tendency for extremely stagnant episodes (i.e. with light winds
and low mixing heights) in locations with high emission rates to exhibit ROG-sensitive
chemistry, while episodes with greater meteorological dispersion or lower emission rates
show NOx-sensitive chemistry. This feature needs to be distinguished from the effects of
chemical composition (i.e. ROG/NOx ratios and biogenic ROG) or transport vs. local
photochemistry.
Milford et al. (1994 , see also Sillman et al., 1993) demonstrated the tendency for
model calculations with high ROG and NOX concentrations to exhibit ROG-sensitive
chemistry while events with lower ROG and NOX concentrations but with the same initial
ROG/NOX ratio would show NOx-sensitive chemistry. This is also illustrated in the air
parcel calculations in Figure 2-1. The series of air parcels shown here initially had the
same ROG/NOX ratio but different concentrations. Following an 8-hour simulation the air
parcels with higher .initial concentrations (corresponding to stagnant meteorology) still had
ROG-sensitive chemistry while the parcels with lower initial concentrations (corresponding
to greater meteorological dispersion) had NOx-sensitive chemistry. Milford et al. also
found that ROG- and NOx-sensitive chemistry in models is correlated with high and low
values of NOy respectively (see Figure 2-3). The higher NOy corresponded to stagnant
meteorology while lower NOy corresponded to locations with greater dispersion.
Similar results were found in the ROM simulation by Roselle et al. (1995). Roselle
et al. showed that events with the highest 03 concentrations were more likely to exhibit
ROG-sensitive behavior, while events with lower 03 (corresponding to the 99th and 95th
percentile of 03 in the simulation) were more likely to show NOx-sensitive behavior. The
difference between the higher and lower 03 in these simulations corresponded to stagnant
events in contrast to events with greater dispersion.
The difference in NOX-ROG chemistry between stagnant events and events with
dispersion explains the tendency for models to predict different NOX-ROG chemistry in the
same location for different events. It also explains the tendency for ROG-sensitive
chemistry to be associated with the largest cities while NOx-sensitive chemistry is more
often associated with smaller cities (e.g. Roselle and Schere., 1995).
13
-------
The connection between NOX-ROG chemistry and meteorological stagnation adds
an additional dimension of uncertainty to the NOX-ROG predictions of photochemical
models, especially when model accuracy is evaluated solely in terms of ozone
concentrations. Dispersion rates are especially difficult to estimate when winds are light
and variable, and simulated peak ozone increases sharply as wind speed is decreased. This
dependency of ozone on wind speed increases as wind speeds get lower. Models can
simulate correct ozone through a series of compensating errors, e.g. overly stagnant
meteorology in combination with underestimated biogenic ROG or underestimated regional
transport; or vice versus. These types of compensating errors would cause changes in
model predictions for NOx-ROG sensitivity.
The association between ROG-sensitive chemistry, large cities and stagnant
meteorology is due in part to the influence of biogenic ROG. Stagnant events tend to have
higher concentrations of anthropogenic NOx and ROG, whose sources are concentrated
near urban centers, but not biogenic ROG which have a more ubiquitous source.
However, there is also a factor directly related to photochemistry. Events with higher NOX
tend to have slower, less efficient chemistry (on the basis of ozone production per NOX)
(Liu et al., 1987, Lin et al., 1988) and consequently have a slower rate of evolution
towards NOx-sensitive chemistry.
Summary: The above description creates an image of the type of situations that are more
likely to be associated with ROG-sensitive chemistry as opposed to NOx-sensitive
chemistry. ROG-sensitive chemistry is associated with: low ROG/NOX ratios; relatively
low impact of biogenic hydrocarbons; locations close to emission sources as opposed to
downwind; situations dominated by short-term local photochemistry as opposed to
multiday transport; and large cities with stagnant meteorology. NOx-sensitive chemistry
is associated with the opposite. However, these characterizations should be used as a
general basis for understanding the behavior of photochemical models and not as
predictions for specific locations. NOX-ROG predictions are subject to a great deal of
uncertainty and need to be examined individually for each location. The description given
here is only useful as a basis for understanding the impact of individual factors in isolation
and for understanding differences in NOx-ROG predictions between various model
scenarios. Subsequent sections (Section 3 and the Appendix) will show NOX-ROG
predictions from individual model scenarios with varying ROG/NOX emission ratios,
biogenic ROG, location, vertical mixing rates and wind speeds. The differences between
scenarios are consistent with the general description given here.
14
-------
2.2 Chemistry of 03, NOX and ROG and its relation to photochemical indicators.
The chemistry of 63, NOX and ROG and its division into NOx-sensitive and ROG-
sensitive regimes can be derived directly from the reaction sequences that produce ozone.
The analysis of NOX-ROG chemistry in terms of reaction sequences, presented here, is
important only in terms of theory. However it provides a rationale for the selection of
specific species ratios as NOX-ROG indicators. While the justification for NOX-ROG
indicators comes from results of more complete photochemical simulations, it is useful to
show that the identified species are linked to NOX-ROG chemistry in a more general sense
than can be shown just from model results.
The chemistry of ozone formation ™" **•• '"^'•rstood from the following simplified
reaction sequence.
RH+ >2 + H2O Rl
RO2+NO ______ +HO2 + NO2 R2
HO2 + NO ^ OH .+ NO2 R3
CO + OH HO2 + C02 R4
RH represents a generic hydrocarbon chain, and RO2 represents an organic peroxy radical.
Most hydrocarbon reaction sequences follow this format. The sequence is initiated by a
reaction between a primary hydrocarbon and the OH radical, creating an RO2 chain, and
followed by a rapid reaction of the RO2 with nitric oxide. The RO2+NO reaction leads to
the production of HO2, which also reacts with NO to return the original OH radical. The
RO2+NO reaction also leads to the production of intermediate ROG species (represented
here as R'CHO), which continue to react via a similar sequence until it is oxidized to CO or
C02*.
* For example, here is the reaction sequence for propane
C3Hg + OH [2?] C3H702 + H2O R 1 '
C3H?02 +NO C2H5CHO + HO2 + NO2 R2 '
15
-------
Reactions R2 and R3 are the ozone-producing reactions, because the conversion
from NO to NO2 is followed by photolysis of NO2 (NO2+hn-*NO+O,
O+O2+M-»O3+M). In this simplified version, the rate of ozone production can be
associated with the rate of reaction Rl, or with the summed rates of R2 and R3.
It is obvious that the rate of the above reaction sequence is controlled by the
availability of odd hydrogen (or odd-H) radicals (OH, HO2 and RO2). The sum of odd-H
radicals (OH+HO2+YRO2) is conserved in each of the above reactions. The behavior of
the reaction system can be understood in terms of the following reactions that act as odd-H
sinks or sources (Kleinman et al., 1986, Sillman et al., 1990a, 1995a):
HO2 + HO2 ->H2O2 + O2 R5
RO2 + HO2 ->ROOH + O2 R6
NO2->HNO3 R7
hv 2 OH R8
Photolysis of ozone (R8) is often the largest source of odd-H radicals. Photolysis
of intermediate hydrocarbons (chiefly formaldehyde, HCHO, and other aldehydes) also
provides a significant radical source and may exceed R8 in urban locations. The two major
sinks for radicals are the formation of peroxides (R5 and R6) and the formation of nitric
acid (R7). Net formation of PAN can also be a significant radical sink (Sillman et al.,
1995c). The PAN reactions are included here for completeness:
RO2 + NO2 ->PAN R9
PAN ->RO2 + NO2 RIO
Because the odd hydrogen radicals are short-lived, reactions R5-R8 form a steady
state in which the radical source reaction (R8) is balanced by the radical sinks (R5, R6 and
R7, along with net production of PAN). The resulting equation for odd-H radicals is:
SH = PHNOS + 2PH202 + 2PROOH + PPAN (2- 1 )
16
-------
where SH represents the odd-H source ( equal to jg[O3] with photolysis rate jg for reaction
R8), PHNOS represents the production rate for HNOs (equal to k7[OH][NO2] with rate
constant kv for reaction R7), PR2O2 and PROOH represents the production rate for
peroxides (equal to ks[HO2]2 and k6[HO2][RO2l, respectively) and PPAN represents net
production of PAN and other PAN-like species.
The split into NOx-sensitive and ROG-sensitive regimes for ozone can be derived
directly from the above analysis of odd-H radicals. When peroxides represent the
dominant sink for radicals, then equation (1) reduces to:
SH= PROOH =2k5[H02] + 2k6[H02][R02] (2-2)
In this situation the sum [HO2]+[RO2] will be largely unaffected by ROG and NOX, being
determined entirely by the size of the radical source (SH, dependent on sunlight, H2O and
03). The rate of ozone production, equal to the summed rates of reactions R2 (=RO2+NO)
and R3 (=HO2+NO), will increase in direct proportion with NOX and will show little
dependence on ROG. ROG contributes to ozone production only to the extent that it
increases the radical source (SH), and even this impact is reduced by the quadratic term for
HO2- This corresponds to the NOx-sensitive regime for ozone.
When nitric acid represents the dominant sink for radicals the situation is very
different:
SH= PHN03 = k7[OH][N02] (2-3)
Here the concentration of the OH radical decreases with increasing NOX. OH may also
increase slightly with increasing ROG because ROG adds to the radical source (SH). Since
the rate of ozone production is approximately equal to the rate of reaction Rl (=RH+OH),
ozone production will decrease with increasing NOX. Ozone production will also increase
with increasing ROG, possibly in a more-than-linear fashion. This corresponds to the
ROG-sensitive regime.
A complete solution for the above nine-reaction system (Sillman, 1995a) showed
that the split between the ROG-sensitive and NOx-sensitive regimes occurs when the
radical sinks through the formation of peroxides (R5 and R6) and the radical sink through
the formation of nitric acid (R7) are exactly equal. Since each peroxide represents a sink
for two odd-H radicals the transition from ROG- to NOx-sensitive chemistry corresponds
to
PHNOS = 2(PH202 + PROOH) (2-4)
17
-------
The instantaneous rate of production of 03 will be NOx-sensitive whenever the
instantaneous rate of production of peroxides (PH2O2 + PROOH) is greater than the
instantaneous rate of production of nitric acid (PHNO3) multiplied by 0.5, and that the rate
of production of 03 will be ROG-sensitive whenever the rate of production of nitric acid
multiplied by 0.5 is greater than the rate of production of peroxides. This result suggests
that the NOx-ROG sensitivity of ozone concentrations (as opposed to the instantaneous
production rate) might be correlated with the ratio TT. using measured H2O2 and
HNO3 concentrations rather than production rates. Both of these species are relatively
long-lived (1-2 days in the convective mixed layer, assuming dry deposition velocities of
1.0 cm/sec for H2O2 arid 2.5 cm/sec for HNO3). Deposition represents the largest sink
H?O2
for both species. The 1-2 day lifetime suggests that TTXT. will represent photochemical
conditions representative of ozone formation over a 1-2 day period, which is appropriate
for a diagnosis of NOX-ROG sensitivity for ozone concentrations. The validity of
as an indicator for NOX-ROG chemistry will be examined in Section 3.
The theoretical basis for O3/NOZ and O3/HNO3 as NOX-ROG indicators can also
be derived from the steady-state equation for odd-H radicals. Assuming that the radical
source (Sn) is due to ozone photolysis, R8, the radical equation (2-1) can be rewritten as
J8[O3] = PHN03 + 2PR202 + 2PROOH + PPAN (2- la)
Since the transition between the NOX- and ROG-sensitive regimes occurs when the radical
sink terms for peroxides and nitric acid are equal to each other (equation 2-4), the NOX-
ROG transition also corresponds to:
jg[O3l= 2PRN03 +PPAN (2-5a)
or
[031 _ 1
2PHNO3 +PPAN js
This result suggests the use of either O3/NOZ or O3/HNO3 as a NOX-ROG indicator. A
[03]
low value for O3/NOZ or O3/HNO3 suggests that the ratio 2PHNO3 + PPAN WSS
for much of the air mass history, corresponding to ROG-sensitive chemistry, while a high
value for O3/NOZ or O3/HNO3 suggests that the ratio 2PHNO3 + PPAN WaS high
much of the air mass history, corresponding to NOx-sensitive chemistry As was the case
18
(2-5b)
-------
for TTKTQ . the indicator ratios O3/NOz and Os/HNOs both involves species with a 1-2
day lifetime and therefore reflect the chemistry associated with a cumulative period of ozone
H9O7 03
production. Unlike TTXT,,, the equation for TT- suggests that the NOX-ROG transition
may vary with cloud cover or with water vapor, both of which affect the reaction rate
parameter, jg.
The linear correlation between 03 and the sum 2H2O2+NOZ, proposed as an
evaluation for the indicator method, is also based on the theoretical analysis of the steady-
state radical equation (2-1). The radical equation suggests that there should be a linear
correlation between 03 and (2H2O2+NOZ) because the former represents the odd-H radical
source and the latter represents the sum of radical sinks. Since the indicator method is
linked in theory to the role of these species as radical sources and sinks, it is important that
the measured correlation between these species show reasonable correspondence with
model results. The slope between 03 and (2H2O2+NOZ) is expected to vary with water
vapor concentration because the rate of the photolysis reaction R8 includes dependence on
water vapor. The slope should also vary from day to day based on cloud cover (which also
affects the photolysis reaction) and rainfall (which removes H2O2 and HNO3). However
it should show relatively little variaion between rural and urban locations during the same
time period. This predicted linear correlation should contrast with the relation between 03
and NOz, which often shows adecreasing O3-NOZ slope as NOz increases (Daum et al.,
1996). The slope between 03 and (2H2O2+NOZ) might be expected to increase slightly in
urban locations due to the added radical source associated with hydrocarbons.
The above analysis has identified correlations between NOX-ROG sensitivity and
indicator ratios based solely on an analysis of ozone production. It has not included the
impact of NOX titration, i.e. removal of ozone through the reaction O3+NO which is
commonly associated with power plants and other large NOX emission sources.
Subsequent results are based on models with a more complete representation of
photochemistry and include this effect. However the indicator-NOx-ROG correlations (for
both O3/NOZ and JJVTQ.J show a characteristic failure in locations where O3-NOZ-ROG
chemistry is dominated by ozone destruction via the O3+NO reaction rather than by
photochemical production of ozone. Since NOX-ROG chemistry is primarily of interest in
situations dominated by ozone production this problem should not limit its use. The
indicator correlations shown in the next section appear to function correctly in models that
include large area NOX sources (i.e. large cities). However care must be taken when
19
-------
indicator measurements are made in the immediate vicinity of a large NOX point source,
e.g. a power plant.
20
-------
50
20
10
X
10 20
io -
50
100 200
500
Hydrocarbons (ppbC)
Figure 2-1. Isopleths showing net rates of ozone production in ppb/hour, averaged over
an 8-hour daytime period, as a function of NO\ and ROG concentrations.
Isopleths range from 0 to 20 ppb/hour in intervals of 2.5 ppb/hr. The dashed
lines represent calculated evolution of NOX and ROG concentrations over an
eight-hour period for air parcels with an initial ROG/NOX ratio of 5:1, adapted
from Milfordetal. (1994).
21
-------
'100 80 60 40 20
"100 80 60 40 20
RMOto Ogtnat fit Consul)
BO 60 40 20
React** Oreanics (% ConwO
-100 80 60 40 20
Reactive Orgarics (K Coneof)
BO 60 40 20
React* Oryutct p4 Cortrd)
Figure 2-2. Predicted response of peak ozone concentrations (ppm) at locations across
the Los Angeles basin, to spatially uniform NOx and ROG emissions
reductions. The upper right-hand comer of each response diagram
corresponds to the base case, (a) Downtown Los Angeles; (b) Pasadena; (c)
San Bernadino; (d) Chino; (e) Roubidoux. From Milford et al., 1989.
22
-------
60
S 40^
Q.
& 20 ± ---
I 0
(a) Lake Michigan regional simulation - base case
100
NOy (ppb)
(b) Northeast corridor regional simulation - base case
Figure 2-3. Predicted reduction in peak 03 (in ppb) resulting from a 35% reduction in
the emission rate for anthropogenic ROG (crosses) and from a 35% reduction in
the emission rate for NOx (circles) plotted against NOy (ppb) coincident with
the ozone peak. Simulations are described in Section 3.1. From Sillman
(1995a) based on Milford et al. (1994) and Sillman (1993).
23
-------
SECTION 3
RESULTS FROM PHOTOCHEMICAL SIMULATIONS
This section will present results for O3-NOX-ROG sensitivity from a series of
photochemical simulations and analyze the correlation between NOx-ROG sensitivity and
the various photochemical indicators as predicted by the models.
Results will be based on groups of three simulations: an initial scenario, a
simulation with anthropogenic ROG emissions reduced by a fixed percentage (usually
35%) relative to the initial scenario, and a simulation with anthropogenic NOx emissions
reduced by the same percentage relative to the initial scenario. Results will be shown for a
specific hour (usually in the afternoon, and corresponding to the time of peak 03 occurring
in the initial model scenario). ROG-sensiti vity will be reported as the difference between
03 in the initial scenario and 03 in the simulation with reduced ROG at the same time and
location. NOx-sensitivity will be similarly reported as the difference between 03 in the
initial scenario and 03 in the: simulation with reduced NOX. These results for model ROG-
and NOx-sensitivity will be presented in comparison with values for photochemical
indicators at the same time and location in the initial model scenario. Results will typically
be reported for every location in the model domain or sub-region of interest.
Complete graphical results will be shown for two indicator ratios: 1- and
Results for the other indicator ratios are visually very similar to these two, and will be
presented in abbreviated form. In addition, a concise method for tabulating results of the
NOx-ROG-indicator correlation will be developed, based on a statistically defined
transition between indicator values associated with ROG-sensitive chemistry and indicator
values associated with NOx-sensitive chemistry. This statistically defined transition point
will be used to summarize results from all model scenarios and for each indicator ratio.
Graphical results will also be shown for indicator ratios that did not correlate well with
03
NOx-ROG sensitivity: *TQ- and the ratio of reactivity-weighted hydrocarbons to NOy.
24
-------
The simulation results shown here should not be interpreted as recommendations
for specific ozone abatement strategies or as statements about NOX-ROG sensitivity in
specific locations. Many of the simulations use outdated inventories for anthropogenic
emissions. Most of the simulations use the BEIS1 inventory for biogenics, which
underestimates isoprene by a factor of three or more (Geron et al., 1994) and consequently
causes model results to be biased in favor of ROG controls. In addition, the selection of
model scenarios for the NOx-ROG analysis deliberately favored ROG-sensitive scenarios,
especially for Atlanta. A meaningful evaluation of NOX-ROG indicators is only possible
in model scenarios that include both NOx-sensitive and ROG-sensitive subregions, which
did not occur in some of the NOx-sensitive scenarios. Despite these caveats, the model
NOx-ROG results are offered as meaningful indications of NOX-ROG sensitivity in relation
to each other: i.e., the differences in NOX-ROG sensitivity between model scenarios
represent tendencies that are likely to be reproduced in other NOX-ROG models.
3.1 Description of photochemical simulations
Results are based on five separate model applications: a regional-scale simulation
for the Lake Michigan airshed, a regional-scale simulation for the northeast corridor, an
urban-scale (UAM-IV) simulation for New York, an urban-scale (UAM-IV) simulation for
Atlanta, and an urban-scale (UAM-IV) simulation for Los Angeles. Preliminary results are
also shown for the Middle Tennesee Ozone Study in Nashville. A detailed description of
each simulation is given here.
(a) Lake Michigan regional simulations:
Event: August 1-2, 1988.
Model: Regional-scale model developed at the University of Michigan.
Chemistry: Lurman et al. (1986) with various updates, including reaction rates from
DeMore et al. (1992), added RO2+HO2 reactions from Jacob and Wofsy (1988),
isoprene chemistry from Paulson and Seinfeld (1992), photolysis rates from Madronich
(1987), column ozone equal to 325 DU, aerosol optical depth 0.68, representing
moderately polluted conditions, clear skies. Dry deposition velocities over land were:
03 and NO2, 0.6cm s'1; NO, 0.1 cm s'1; HNO3, 2.5 cm s'1; PAN, 0.25 cm s'1;
and H2O2, 1.0 cm s" 1. Deposition velocities over water were reduced to 0.05 cm s~ 1
for all these species, reflecting a probable surface inversion over cold water during the
simulated events.
25
-------
Grid resolution and domain size: 20x20 km horizontal resolution, 3 vertical layers.
The vertical structure accounts for stable conditions over Lake Michigan during the
daytime and generates a confined sub-layer for urban emissions from Chicago, with
typical heights 200-500 meters. The vertically confined Chicago plume is consistent
with recent aircraft measurements by Hillery (1995). The domain extends from south
of Chicago to the northern end of Lake Michigan, 220x380 km. This was combined
with a coarse-resolution model for regional transport that included most of the eastern
U.S. (Sillman et al., 1990b). NOX-ROG sensitivity was based on reduced emissions
throughout both model domains.
Horizontal advection: Smolarkiewicz (1983) for the local domain, Prather (1986) for
regional transport.
Meteorology: Winds were interpolated from measurements at sites in the Lake Michigan
region and from regional measurements from the National Weather Service. Some
modifications were made in wind speeds to improve model performance vs. measured
ozone. Mixing heights were based on measured vertical temperature profiles at land-
based sites, along with the assumption that growth of the mixed layer ceases as air
travels over Lake Michigan.
Emissions: Anthropogenic emissions are from the NAPAP 1980 inventory (EPA,
1986). Biogenic emissions are derived from data by Lamb et al. ((1985) in
combination with land use data by Matthews (1983), similar to BEIS1.
Alternate scenarios: Modified scenarios include doubled anthropogenic ROG
emissions and anthropogenic emissions reduced by half. NOX-ROG results are
reported for 6 pm, which corresponded to peak or near-peak 03 in all scenarios.
Reference: Sillman et al., 1993.
(b) Northeast corridor regional simulations:
Event: June 14-15, 1988.
Model: Regional-scale model developed at the University of Michigan.
Chemistry: Lurman et al. (1986) with various updates, including reaction rates from
DeMore et al. (1992), added RO2+HO2 reactions from Jacob and Wofsy (1988),
isoprene chemistry from Paulson and Seinfeld (1992), photolysis rates from Madronich
(1987), column ozone equal to 325 DU, aerosol optical depth 0.68, representing
moderately polluted conditions, clear skies. Dry deposition velocities over land were:
O3 and NO2, 0.6 cm s'1; NO, 0.1 cm s'1; HNO3, 2.5 cm s'1; PAN, 0.25 cm s'1;
and H2O2, 1.0 cm s'1. Deposition velocities over water were reduced to 0.05 cm s'l
26
-------
for all these species, reflecting a probable surface inversion over cold water during the
simulated events.
Grid resolution and domain size: 20x20 km horizontal resolution, 3 vertical layers.
The vertical structure accounts for stable conditions over the Atlantic Ocean during the
daytime and generates a confined sub-layer for land-based emissions. The domain
extends from south of Chicago to the northern end of Lake Michigan, 480x560 km.
This was combined with a coarse-resolution model for regional transport that included
most of the eastern U.S. (Sillman et al., 1990b). NOX-ROG sensitivity was based on
reduced emissions throughout both model domains.
Horizontal advection: Smolarkiewicz (1983) in the local domain, Prather (1986) for
regional transport.
Meteorology: Winds were interpolated from measurements at sites along coastal New
England and from regional measurements from the National Weather Service. Some
modifications were made in wind speeds to improve model performance vs. measured
ozone. Mixing heights were based on measured vertical temperature profiles at land-
based sites, along with the assumption that growth of the mixed layer ceases as air
travels over the Atlantic Ocean.
Emissions: Anthropogenic emissions are from the NAPAP 1980 inventory (1986).
Biogenic emissions are derived from data by Lamb et al. ((1985) in combination with
land use data by Matthews (1983), similar to BEIS1.
Alternate scenarios: Modified scenarios include the following: a scenario with zero
biogenic emissions; a scenario with doubled isoprene emissions and mixed layer
heights reduced by half; a scenario with wind speeds reduced by half; a scenario with
increased deposition velocities (4 cm s'l for H2O2 and HNO3 over land) and wind
speeds reduced by half. NOX-ROG results are reported for 6 pm, which corresponded
to peak or near-peak 03 in all scenarios.
Reference: Sillman et al., 1993.
(c) New York urban simulations:
Event: July 21, 1980.
Model: Urban Airshed Model (UAM-IV).
Chemistry: Carbon Bond IV.
Grid resolution and domain size: 8x8 km horizontal grid cell resolution, 5 or 7
vertical layers. The domain extends from central New Jersey to eastern Connecticut
and Long Island (248x200 km). Upwind conditions were derived from the Regional
27
-------
Oxidant Model, with nominal 20x20 km horizontal grid resolution and domain covering
the eastern half of the U.S.
Horizontal advection: Different scenarios use either Smolarkiewicz (1983) or Bott
(1989).
Meteorology: Two scenarios were included with different meteorology. One used the
UAM-IV diagnostic wind processor in combination with measured vertical profiles of
temperature and wind speeds from the National Weather Service. The other scenario
used results from a dynamic mesoscale meteorological model (MMM) by Ulrickson and
Hass (1990), including four-dimensional data assimilation.
Emissions: Anthropogenic emissions are from the Rao and Sistla( 1993). Biogenic
emissions are from BEIS1 (Pierce et al., 1990).
Alternate scenarios: Two scenarios were included: a scenario with five model vertical
layers and meteorology from the UAM diagnostic wind processor ("DWM-5") and a
scenario with seven model vertical layers and meteorology from the mesoscale
meteorological model ("MMM-7"). The latter scenario had lower wind speeds and
resulted in higher ozone. NOx-ROG results are reported for 5 pm. Peak 03 occured at
3 pm.
(d) Atlanta urban simulations:
Event: August 9-11, 1992.
Model: Urban Airshed Model (UAM-IV).
Chemistry: Carbon Bond IV.
Grid resolution and domain size: 4x4 km horizontal resolution, 5 or 7 vertical layers
in different scenarios. The domain includes metropolitan Atlanta (108x140 km) but
does not include regional-scale chemistry and transport. Upwind conditions had
O3=55 ppb, based on measurements upwind of Atlanta, and NOX=1 ppb. In addition,
the scenarios included here have upwind H2O2 (=1 ppb), HNO3 (=2 ppb) and PAN
(=0.75 ppb). These modified upwind concentrations were derived from correlations
between O3 and NOZ (Trainer et al., 1993) and 03 and PAN (Roberts et al., 1995) at
rural sites in eastern North America.
Horizontal advection: Smolarkiewicz (1983).
Meteorology: UAM-IV diagnostic wind processor in combination with measured
vertical profiles of temperature and wind speeds. Mixing heights based on measured
vertical profiles and methods described by Marsik et al. (1995).
28
-------
Emissions: Anthropogenic emissions are from Cardelino et al. (1994). Biogenic
emissions are from BEIS1 (Pierce et al., 1990).
Alternate scenarios: The scenarios shown here all had mixing heights reduced by 20%
relative to the base case, in order to enhance model sensitivity to ROG. They include a
scenario with BEIS1 biogenics and a scenario with tripled emission rates for isoprene,
approximately equal to the BEIS2 inventory (Geron et al., 1995). NOX-ROG results
are reported for 5 pm.
Reference: Sillman et al., 1995b, 1997a.
(e) Los Angeles urban simulations:
Event: August 26-29, 1987.
Model: Urban Airshed Model (UAM-IV).
Chemistry: Carbon Bond IV.
Grid resolution and domain size: 5x5 km horizontal grid cell size, 5 vertical layers.
The domain includes metropolitan Los Angeles (180x325 km). Boundary conditions
were 30 ppb 03.
Horizontal advection: Different scenarios use either Smolarkiewicz (1983) or Bott
(1989).
Meteorology: UAM-IV diagnostic wind processor in combination with measured
vertical profiles of temperature and wind speeds.
Emissions: Anthropogenic emissions are from the Southern California Air Quality Study
(SCAQS) (Lawson et al., 1990, Wheeler et al., 1992). Biogenic emissions are from
BEIS1 (Pierce et al., 1990).
Alternate scenarios: The scenario by Godowitch and Vukovich (1994) uses a version
of UAM-IV with horizontal advection based on Bott (1989). Scenarios by Wagner et
al. (1992) use the standard UAM-IV with horizontal advection based on Smolarkiewicz
(1983). The scenarios by Wagner et al. include a base case, a case with doubled
anthropogenic ROG and a case with tripled anthropogenic ROG. Simulated control
strategies are based on 25% reductions relative to the initial scenarios, rather than 35%
in the other case studies. NOX-ROG results are reported for 3 pm.
Note: Recent, more detailed photochemical models for Los Angeles have been developed
by Harley et al. (1993), Kumar et al. (1994) and Lu and Turco (1995).
Reference: Godowitch and Vukovich (1994), Wagner et al. (1992).
29
-------
(f) Nashville simulations (preliminary):
Event: July 11-13,1995.
Model: Nested urban and regional-scale model developed at the University of Michigan.
Chemistry: Lurrnan et al. (1986) with various updates, including reaction rates from
DeMore et al. (1992), added RO2+HO2 reactions from Jacob and Wofsy (1988),
isoprene chemistry from Paulson and Seinfeld (1992), photolysis rates from Madronich
(1987), column ozone equal to 325 DU, aerosol optical depth 0.68, representing
moderately polluted conditions, clear skies. The chemistry for this simulation also
includes modified RO2-RO2 reactions and rates recommended by Kirchner and
Stockwell (1996), which were not included in previous cases with this model. Dry
deposition velocities over land were: 03 and NO2, 0.6 cm s~l; NO, 0.1 cm s"^;
HNO3,2.5 cm s'1; PAN, 0.25 cm s'1; and H2O2,2.5 cm s'1. Deposition velocities
for H2O2 were increased in comparison with (a) and (b) above, based on recent work
by Hall etal.( 1997).
Grid resolution and domain size: 5x5 km horizontal resolution, 3 vertical layers.
The domain is 180x180 km, centered on downtown Nashville. This was combined
with a coarse-resolution model for regional transport that included most of the eastern
U.S. (Sillman et al., 1990b). NOX-ROG sensitivity was based on reduced emissions
throughout both model domains.
Horizontal advection: Smolarkiewicz (1983) for the local domain, Prather (1986) for
regional transport.
Meteorology: Winds and mixing heights were derived from the Rapid Update Cycle
(RUC) prognostic model (Benjamin, 1994), with modifications based on on-site
measurements in the Nashville area.
Emissions: Anthropogenic emissions are from the NAPAP 1990 inventory (EPA, 1993)
for the regional background. Emissions for the Nashville urban domain are from an
indentory developed as part of the 1988 State Implementation Plan (SIP) for Nashville,
developed by W. Davis (University of Tennesee, Knoxville) and used with permission
from J. Walton (Tennesee Department of Environmental Quality). Biogenic emissions
are from BEIS2 (Geron et al., 1994).
Alternate scenarios: Modified scenarios include cases with dry deposition of H2O2
and HNO3 increased to 4 cm s'1, cases without the RO2-RO2 reactions recommended
by Kirchner and Stockwell, cases with changed wind speeds and with reduced
emission of biogenic hydrocarbons. NOX-ROG results are reported for 4 pm, July 13,
which corresponded to peak or near-peak 03 in all scenarios.
30
-------
Reference: Sillman et al., 1997b. This publication has been recently submitted for
publication and has not yet been through peer review. Only preliminary results are
included here.
3.2 Results
,. 03 03 . 03
(a) iwrO vrT7and
Z' NOy miu HNO3-
Figure 3-1 shows the correlation between model NOX-ROG sensitivity and Tj/y- for
six model scenarios. The figure shows the reduction in 03 associated with a particular
percentage reduction in anthropogenic ROG or NOX emissions (usually 35%), defined as
the difference between 03 in the initial scenario and 03 in the scenario with reduced ROG
or NOX at the same time and location. Negative values represent locations in which
reduced anthropogenic emissions results in an increase in 03. The simulated NOX-ROG
O3
reductions are plotted for all locations in the model domain, in comparison with ^^- in the
initial model scenario at the same time and location.
Each simulation shows a similar pattern. Locations with larger ozone reductions in
response to reduced ROG rather than reduced NOX (i.e. locations with ROG-sensitive
03
chemistry) also have low — (<8) in the initial simulation, while locations with larger
ozone reductions in response to reduced NOX (i.e. locations with NOx-sensitive chemistry)
03 03
also have high — (>10). There is also a well-defined value for — that defines the
transition between NOx-sensitive and ROG-sensitive regions. This transition value
remains the same in simulations for different metropolitan areas and for different model
scenarios.
Figure 3-1 also shows that the correlation between NOX-ROG sensitivity and the
03
indicator ratio TT— remains consistent in model scenarios with very different overall
NOX-ROG results. For example: the Lake Michigan base case (Figure 3- la) has primarily
ROG-sensitive chemistry, especially in the locations with highest ozone in the plume from
Chicago. When the Lake Michigan simulation is repeated with initial doubled ROG
emissions (Figure 3-lb) the predicted NOX-ROG sensitivity changes and the locations with
highest ozone show approximately equal sensitivity to reductions in NOX and ROG. A
03
comparison of these two scenarios shows that »IQ— also changes from low values in the
ROG-sensitive scenario to higher values in the more NOx-sensitive scenario. The change
31
-------
in VT- between the two scenarios is consistent with the overall correlation between NOX-
03 03
ROG chemistry and ™% and suggests that measured — might be used as a basis for
choosing between the NOx-ROG predictions of these two model scenarios.
Results from the two simulations for New York (Figures 3-lc and 3- Id) illustrate
the impact of meteorology on NOx-ROG predictions. The northeast corridor base case
(Figure 3-lc) is characterized by relatively strong winds and vigorous vertical mixing, and
results in a largely NOx-sensitive simulation. The New York UAM scenario (with MMM-
7 meteorology) has lighter winds and results in a largely ROG-sensitive simulation. The
03
ROG-sensitive simulation also shows low — while the NOx-sensitive simulation shows
03 03
high vTi- As with the Lake Michigan case, the change in — between these two
03
scenarios is consistent with the overall correlation between — and NOX-ROG
sensitivity. Although these two results were produced with different photochemical models
the difference between NOx-ROG predictions appears to be due to the difference in
meteorology between the two events rather than to differences between the models. When
the northeast corridor simulation is repeated with lower wind speeds it shows ROG-
sensitive chemistry, and when the New York UAM scenario is repeated with stronger wind
speeds it shows NOx-sensitive chemistry. (Results for these scenarios are shown in
03
Appendix B). The correlation between — and NOX-ROG sensitivity appears consistent
between the two models.
Lastly, the contrast between simulations for Atlanta and Los Angeles (Figures 3-le
and 3- If) illustrates the differences between cities with characteristics that favor NOx-
sensitive chemistry as opposed to ROG-sensitive chemistry. Los Angeles as an urban area
has several characteristics that favor ROG-sensitive chemistry based on the description in
Section 2: a large city with high density ofemissions; limited vertical dilution; relatively
low emission rates of biogenic ROG in comparison with anthropogenic ROG; and
domination by local chemistry rather than regional transport. By contrast, Atlanta is a
relatively smaller city with lower emission density; higher ventillation rates than Los
Angeles, and high rates of biogenic emissions. The simulation results in Figures 3-le and
3- If reflect these differences, showing extensive NOx-sensitive chemistry in Atlanta and
extensive ROG-sensitive chemistry in Los Angeles. The simulations show a similar
03
difference between high c- in Atlanta in association with NOx-sensitive chemistry and
low TTT- in Los Angeles in association with ROG-sensitive chemistry. This difference in
32
-------
O3
between Atlanta and Los Angeles, predicted by photochemical models, is also
consistent with ambient measurements (see Section 5).
Although the correlation between - and NOX-ROG sensitivity appears
consistently in all six simulations, there are a few locations in the Los Angeles simulation
(Figure 3f) that violate the general pattern. These locations are characterized by high
(>12) and ROG-sensitive chemistry. These locations are all characterized by a large
negative response to reduced NOX, i.e. reduced NOx causes a significant increase in
simulated 03 at these locations. These exceptional points occur in locations where NOX-
ROG sensitivity results are dominated by titration of 03 by direct emissions from a local
O3
NOX source. As explained in Section 2, the association between ~r- and NOX-ROG
sensitivity is derived in theory from the chemistry of ozone production and does not include
NOX titration. The same type of exception appears in results with most of the.other
indicator ratios (e.g. see Figure 3-4f). The exceptions do not appear in results for
ilustrated in Figure 3-2, because locations dominated by NOX titration all have
O3
characteristically high NOy and low
Results from other simulations for the indicator ratios ^FT, XT<4 andTTXTl are
NOz NOy HNO3
all visually similar to the results in Figures 3-1 and 3-2. Complete results are shown in the
Appendix. Summary transition values forNOx- vs ROG-sensititive chemistry are: 8-10
03 03 01
for j^-; 6-7.5 for ^-; and 10-15 for J
A useful way to describe the correlation between NOX-ROG sensitivity and
indicator ratios is to compare the statistical distribution of indicator values associated with
NOx-sensitive chemistry with the distribution of indicator values associated with ROG-
sensitive chemistry in individual photochemical simulations. The distribution of ^^- for
NOx-sensitive and ROG-sensitive locations in the simulation for Los Angeles is shown in
Figure 3-3. The following definition is used for this analysis: a location is defined as
NOx-sensitive if 03 in the simulation with reduced NOX emissions is lower than 03 in
both the initial simulation and the simulation with reduced ROG by at least 5 ppb.
Similarly, a location is ROG-sensitive if 03 in the simulation with reduced ROG emissions
is lower than 03 in both the initial simulation and the simulation with reduced NOX by at
least 5 ppb. Results of this analysis (Figure 3-3) show that a large majority of ROG-
03
sensitive locations have much lower XT<^ than the NOx-sensitive locations and there is
i\uz
33
-------
relatively little overlap between the range of - for ROG-sensitive locations and the
range of \f(O3,NOz) for NOx-sensitive locations. The 92nd percentile of the distribution
03
of — for ROG-sensitive locations (8.25) is almost identical with the 8th percentile of the
distribution of — for NOx-sensitive locations (8.16). In other words, 92% of the ROG-
sensitive locations have - below this value while 92% of the NOx-sensitive locations
03
have - above this value. ,
The presentation of model results in terms of percentile distributions of NOX- and
ROG-sensitive indicator values also provides a concise way of tablulating the indicator-
NOX-ROG correlation from individual simulations. Table 3-1 shows the 95th percentile
indicator values for ROG-sensitive locations and the 5th-percentile indicator values for
NOx-sensitive locations for each model scenario. A comparison between the 95th and 5th
percentile values provides an estimate for the extent of overlap between ROG-sensitive and
NOx-sensitive indicator values; i.e. when the 95th percentile ROG-sensitive value is equal
to or lower than the 5th percentile NOx-sensitive value there is little overlap between NOX-
sensitive and ROG-sensitive indicator values. The 95th and 5th percentile values also
define the transition point between NOx-sensitive and ROG-sensitive indicator values in
each scenario and provide a convenient basis for comparison between indicator transition
points in each simulation. This information is also presented graphically in Figure 3-4.
The information in Figure 3-4 can be easily understood if it is kept in mind that 95% of the
locations with ROG-sensitive chemistry in each simulation have indicator values below the
solid line in the figure, while 95% of location with NOx-sensitive chemistry have indicator
values above the dashed line.
As shown in Table 3-1 and Figure 3-4 the correlation between NOX-ROG
sensitivity and the indicator ratios J^Q— , J^Q— and ^Q remains consistent for a wide
range of simulations, including scenarios for different cities, different meteorology, widely
varying anthropogenic emission rates, ROG/NOX ratios and biogenic emission rates. The
transition point between NOX- and ROG-sensitive chemistry shows relatively little variation
among the model scenarios. The values for the NOX-ROG transition for each indicator,
cited above, also appear clearly in Figure 3-4. The subsequent notes focus results that
represent exceptions to this generally consistent pattern. The first three issues below
(multiday transport, deposition rates and paniculate nitrate) relate to uncertainties in
science. The others (Smolarkiewicz advection and upwind boundary conditions) relate to
model artifacts.
34
-------
(1) The simulation for the northeast corridor with zero isoprene also shows NOX-
ROG transition values that are significantly higher than the other scenarios. The apparent
reason for this discrepancy is that the urban plume from the New York metropolitan area
still has ROG-sensitive chemistry even after 24 hours of downwind transport.
Conseqnently the NOX-ROG transition is affected by nighttime chemistry. In the other
scenarios, plumes that are more than 24 hours downwind from emission sources are either
all NOx-sensitive (due to photochemical aging and loss of NOX at night) or else have exited
from the model domain. The NOX-ROG transition in a 2-day-old plume may occur at a
higher indicator ratio due to more rapid removal of HNO3 relative to 03 as the plume ages.
(2) The simulation for the northeast corridor with increased dry deposition rates for
HNOs and H2O2 has a higher NOX-ROG transition point than the other scenarios. This
may represent a significant scientific uncertainty. The uncertain removal rate for HNO3 is
also related to uncertainties in the interpretation of paniculate nitrate measurements,
discussed below. In a sensitivity test shown here, increased deposition causes the NOx-
ROG transition to increase by only 10%, but the change in deposition was relatively small
(4 cm/s for HNO3 vs 2.5 cm/s in the base case). Preliminary results for Nashville,
included in Table 3-1, also has higher NOX-ROG transition points in a scenario with high
deposition.
(3) None of the simulations shown here include paniculate nitrate. It is unclear
whether paniculate nitrate should be included implicitly in HNO3, NOZ or NOy. To the
extent that paniculate nitrate is formed directly from HNQs and is ultimately derived from
NOX emissions, it should be either represented as a loss term for HNO3 in model
calculations or else included implicitly in model gas-phase HNO3. But if paniculate nitrate
is emitted directly into the atmosphere, this directly emitted nitrate should not be included in
the indicator ratios. Parrish et al. (1993) found that paniculate nitrate was small compared
to HNO3 and other components of NOZ at rural sites in eastern North America, but Tuazon
et al (1980, 1981) found that NH4NO3 aerosols accounted for a significant fraction of total
reactive nitrogen (assumed to include NO3" but not NH3 or NH4+) at one site (but not all
sites) in Los Angeles. Interpretation of paniculate nitrate represents a significant
uncertainty.
(4) The UAM-IV simulations for Los Angeles from Wheeler et al. (1992)
generated much greater overlap between ROG-sensitive and NOx-sensitive indicator values
than any of the other simulations. The degree of overlap in the Wheeler et al. simulations
might be due to the Smolarkiewicz advection calculation in combination with the extremely
sharp gradients in 03 and reactive nitrogen in the simulations. These simulations included
gradients in ozone concentrations of 50 ppb or higher between adjacent 5x5 km horizontal
35
-------
grid cells, with corresponding gradients in reactive nitrogen and H2O2. In the
Smolarkiewicz advection calculation the presence of steep concentration gradients can cause
results for individual locations to be affected by the concentrations of neighboring grid
cells. This may be the cause of the imprecise NOx-ROG-indicator correlation in the
simulations by Wheeler et al. The simulation by Godowitch et al. represents the same
event and uses similar meteorology and the same photochemical model, but advection is
calculated based on Bott (1989) instead of Smolarkiewicz. The resulting simulation shows
much less overlap between NOx-sensitive and ROG-sensitive indicator values.
Imprecision associated with the Smolarkiewicz advection calculation has been cited
elsewhere (Chock, 1991-).
Results tend to be somewhat better for — than — , especially for the Los
03
Angeles simulations by Wagner et al. Despite this, it is recommended that — be used as
in indicator rather than (or in addition to) — whenever possible. This recommendation
is partly based on the fact that a useful correlation has been found in measurements between
03 and NOZ (see the Los Angeles case study in Section 5, also Figures 4-3 and 4-4 and
results in Trainer et al.; 1993). In addition, measured — at surface sites may be
dominated by on-site NOx emissions.
In situations where only NO and NOy measurements. are available a useful
approximation for NOX and NOZ might be made by using the steady-state, relation between
03, NOandNO2:
where j 1 1 and kl 2 are the reaction rates for NO2+hv->NO and NO+O3->NO2,
respectively. Chameides et al. (1990), Ridley et al. (1992) and others have found that the
NO
above equation underestimates rr in rural locations by -30% due to the ozone-producing
reactions (HO2+NO->OH+NO2). The underestimate is lower in urban centers. A
plausible approximation for NOX as a function of NO would be:
NOX = NO( 1+0.05 03) (3-2)
where the coefficient 0.05 represents the ratio ji i/k 12 during full sunlight. This
approximation tends to underestimate NOX- This type of approximation has never been
used as a basis for analyzing data sets in the absence of NO2 and therefore must be
36
-------
regarded as speculative. However it might provide a basis for evaluating the accuracy of
measured NOy and for identifying cases where measured NOy may be dominated by on-
site NOx emissions.
37
-------
•5 -20 5P
ROG controls
NOx controls
20
25
(a) Lake Michigan regional simulation - base case (6 pm)
c
JO
"o
3
•o
o
CO
O
X ROG controls
NOx controls
(b) Lake Michigan regional simulation - doubled ROG (6 pm)
Figure 3-1. Predicted reduction in afternoon 03 (in ppb) resulting from a 35% reduction
in the emission rate for anthropogenic ROG (crosses) and from a 35% reduction
in the emission rate for NOx (circles) plotted against O3/NOz at the same time
and location. Simulations are described in Section 3.1. From Sillman (1995a,
1997a).
38
-------
J2
Q.
^
C
_0
'*-
u
3
•o
0)
CO
O
X ROG controls
e NOx controls
0
25
(c) Northeast corridor regional simulation - base case (6 pm)
.o
Q.
u
3
•o
o
CO
O
X ROG controls
NOx controls
0
25
(d) New York urban simulation - "MMM-7" meteorology (5 pm)
Figure 3-1 (continued): Predicted reduction in afternoon 03 (in ppb) resulting from a
35% reduction in the emission rate for anthropogenic ROG (crosses) and from a
35% reduction in the emission rate for NOX (circles) plotted against O3/NOZ at
the same time and location. Simulations are described in Section 3.1. From
Sillman (1995a, 1997a).
39
-------
.0
o.
a
o
3
•o
a
CO
O
60
40
20
0
-20 I
-40-i
-60
0
X ROG controls
• NOx controls
10 15
03/NOz
20 25
(e) Atlanta urban simulation - BEIS1 and low mixing height (5 pm)
X2
a
c
o
•o
o
CO
O
O
e e
-x—x
X ROG controls
• NOx controls
20
25
(f) Los Angeles urban simulation - Godowitch and Vukovich(1994) base case
(3 pm)
Figure 3-1 (continued): Predicted reduction in afternoon 03 (in ppb) resulting from a
35% reduction in the emission rate for anthropogenic ROG (crosses) and from a
35% reduction in the emission rate for NOX (circles) plotted against O3/NOZ at
the same time and location. Simulations are described in Section 3,1. From
Sillman (1995a, 1997a).
40
-------
X ROG controls
NOx controls
15
20
Figure 3-2 : Predicted reduction in afternoon 03 (in ppb) resulting from a 35% reduction
in the emission rate for anthropogenic ROG (crosses) and from a 35% reduction
in the emission rate for NOX (circles) plotted against Os/NOy at the same time
and location. Results are for the Los Angeles urban simulation by Godowitch
and Vukovich( 1994). From Sillmanetal. (1997a).
c
o
*3
3
.Q
tn
b
03/NOz
Figure 3-3 : Cumulative percentile distribution of O3/NOZ for locations with ROG-
sensitive chemistry (solid line) and for locations with NOx-sensitive chemistry
(dashed line). Percentiles represent the fraction of the total array with O3/NOZ
lower than the specified value. Results are for the Los Angeles urban
simulation by Godowitch and Vukovich (1994) and based on analysis reported
in Sillmanetal. (1997 a).
41
-------
16
N
o
CO
O
O
z
co
O
6-E
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
jL .J.-L.J--L-J--L_-I-_L-J-_L_.
i i i — i i i i
1 2 3 4 5 67 8 9 10 11 12 13 14 15 16 17 18
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Simulation
Figure 3-4: 95th-percentile indicator value for ROG-sensitive chemistry (solid line) and
5th percentile indicator value for NOx-sensitive chemistry (dashed line) from 18
model scenarios identified in Table 3-1, for the indicator ratios G>3/NOZ,
O3/NOy and O3/HNO3. In each simulation 95% of the locations with ROG
sensitive chemistry have indicator ratios below the solid line, while 95% of the
locations with NOx-sensitive chemistry have indicator ratios above the dashed
line.
42
-------
Table 3-1
NOv- and
values for
03
and
03
as photochemical
indicators
The table gives the 95th percentile value for the distribution of each indicator ratio over ROG-
sensitive locations and the 5th percentile value for NOx-sensitive locations for various model
scenarios.
Model scenario
Lake Michigan regional simulations:
1. Base case
2. Doubled ROG
3. ROG reduced by half
4. Base case at 12:00 pm.
Northeast corridor regional simulations:
5. Base case
6. Zero isoprene
7. Low mixing
8. Doubled isoprene and low mixing
9. Reduced wind speeds
10. Increased deposition, reduced winds
New York urban simulations:
1 1 . MMM-7 meteorology
12. DWM-5 meteorology
Atlanta urban simulations:
13. BEIS1
14. Tripled isoprene
Los Angeles urban simulaions:
15. Godowitch base case
16. Wagner base case
17. Wagner, doubled ROG
18. Wagner, tripled ROG
Nashville (preliminary)
19. Base case
20. Older chemistry
21. Higher deposition (4 cm s"^)
03
NO^
95th
ROG
8.2
8.6
7.9
9.3
7.6
13.6
8.2
7.0
9.2
10.2
7.4
7.3
9.6
7.7
10.4
20
14
14
5.6
5.8
6.4
5th
NOX
11.5
8.1
7.8
13.0
9.7
.14.0
9.6
7.5
10.2
11.1
9.9
9.4
12
11
9.8
6.3
6.2
6.2
7.1
7.0
8.4
Os
J
NO
95th
ROG
6.7
7.6
6.0
6.9
6.5
10.4
7.1
5.2
6.9
7.5
7.1
4.5
6.7
5.4
6.1
8.0
7.6
8.3
7.3
7.4
8.6 .
n-,
y
5th
NOX
9.4
7.7
6.8
9.4
8.2
11.8
8.4
6.2
8.2
9.0
6.7
7.8
8.1
7.5
7.5
5.6
5.7
5.5
8.4
8.1
10.2
~->
HNO3
95th
ROG
11.8
15.2
10.1
15
15
19
15
15
17
21
12.1
8.4
13
9.8
14
34
38
48
9.1
9.5
11
5th
NOX
18.1
12.4
9.4
21 .
17
16
17
17
18
21.
12.6
11.1
16
16
14
7.3
7.4
7.7
12
12
15
43
-------
H202 H202 . H2O2
(b) HN03' "NOT andT*Oy~ *
The correlation between NOX-ROG sensitivity and TjCrfC is shown in Figure 3-5.
Complete results are shown in the Appendix. The correlation for KTT> and JL is
lNL/2 INUy
visually similar to TTVTQ- but with a different value for the transition between ROG-
sensitive and NOx-sensitive chemistry. The NOX-ROG transition points are identified
through the 95th percentile ROG-sensitive and 5th percentile NOx-sensitive indicator
values, which are shown in Figure 3-7 and Table 2. As was the case for N( , the ratio
H2O2
HNQo is strongly correlated with NOX-ROG sensitivity in model results, and the
correlation remains consistent for different locations, model types and model scenarios.
Changes in model assumptions that cause a shift in NOX-ROG sensitivity also cause an
equivalent change in the ratio TTXJQ-. The same pattern is repeated for the other H2O2
ratios. As described in Section 2.1, the correlation between T and NOX-ROG
sensitivity is derived directly from the chemistry of ozone production and consequently has
a strong theoretical justification. Model results show that JL and JL correlate with
NOX-ROG sensitivity equally well, although the theoretical justification is derived from
H202
HNO3'
The ratios involving H2O2 have two important advantages over N(X and its
equivalents as an indicator for NOX-ROG chemistry. First, the NCbt-ROG correlation is
stronger for the indicators involving H2O2- A comparison of Figures 3-1 and 3-5 shows
that the difference between NOX- and ROG-sensitive values is much larger for the H2O2-
based indicators than for the O3-based indicators, making the possibility of error due to
imprecise measurement less. The transition region between NOX- and ROG-sensitive
chemistry is also smaller in comparison to the difference between NOX- and ROG-sensitive
values for the H2O2-based indicators. The 95th percentile ROG and 5th percentile NOX-
sensitive values in Table 2 demonstrate the small exteent of overlap between NOx-sensitive
and ROG sensitive values for the H2O2-based indicators. Second, species correlations
involving H2O2 provide an important diagnostic test for the validity of the indicator
method, transition points and the accuracy of measurements. This is described below in
44
-------
Section 4. The inclusion of tests for the accuracy of method rather than simply accepting
the indicator NOX-ROG correlation "on faith" is an important advantage for data sets that
include H2O2- However H2O2 measurements are more difficult and less accurate than 03,
and the H2O2-based indicators and are more sensitive to model uncertainties, especially
with regard to deposition rates (see below).
Most of the uncertainties described above in connection with the O3-based
indicators also apply to the H2O2-based indicators. As shown in Figure 3-4 for Los
Angeles, an exception to the indicator NOX-ROG correlation occurs for location where
NOx-ROG sensitivity is due to titration by direct NOX emissions rather than by differing
rates of ozone production. These exceptions do not appear in the NOX-ROG correlation for
theraio TT~~ (Figure 3-5), although TTT has a stronger theoretical justification. The
model results shown here do not include particulate nitrate, and it is uncertain whether
paniculate nitrate should be included in the indicator ratios. In addition, deposition rates
for H2O2 and HNOs represent a greater uncertainty for the H2O2-based indicators than
for the ratios involving 03. Results from Hall and Claiborn (1997) show that the dry
deposition rate for H2O2 should be equal to the rate for HNO3 and may can have very high
values (4 cm s~l or more). Scenarios with increased deposition of H2O2 (including both
the scenario for the northeast corridor with increased deposition and the Nashville base
case) have NOX-ROG transition values lower by 25%. The ratio NQ is especially
sensitive to changes in assumed deposition rates. H2O2 is also affected by wet deposition,
not included here.
An additional uncertainty for the H2O2-based indicators is the interpretation given
to H2O2 in models. Many photochemical mechanisms, including the CB-TV mechanism
used in the UAM, do not include formation of organic peroxides. In these mechanisms
H2O2 should be interpreted as representative of the sum of H2O2 and organic peroxides.
Simulations with a more complete representation of organic peroxides (Sillman et al.,
1995b, 1997b) have predicted that H2O2 will represent 50%-70% of total peroxides.
NOX-ROG transition values for NOX- vs ROG-sensititive chemistry based on the
H2O2 H2O2
simulations used in this study are: 0.3-0.6 for ; 0.2-0.35 for NQ ; and 0.2-
0.3 for V|Q~- These transition values appear clearly in Figure 3-7. However
preliminary results from Nashville (in simulations with updated photochemistry and higher
dry deposition rates) have resulted in a lower estimate for NOX-ROG transition values:
0.2-0.3 for j: ; 0.15-0.20 for ^ ; and 0.13-0.17 for -. The difference
45
-------
between the transition values for Nashville and the earlier estimates represents a major
uncertainty and may lead to future modifications.
46
-------
^ 60 T
ROG controls
0 NOx controls
0
0.5 1 1.5
H202/HN03
(a) Lake Michigan regional simulation - base case (6 pm)
^-K OU
c
o
*3
O
3
•o
0)
CO
O
0
X ROG controls
* NOx controls
:-*_x>xccXooxx
0.5
1.5
H202/HN03
(b) Lake Michigan regional simulation - doubled ROG (6 pm)
Figure 3-5. Predicted reduction in afternoon 63 (in ppb) resulting from a 35% reduction
in the emission rate for anthropogenic ROG (crosses) and from a 35% reduction
in the emission rate for NOx (circles) plotted against H2O2/HNO3 at the same
time and location. Simulations are described in Section 3.1. From Sillman
(1995a, 1997a).
47
-------
Si
a
^
c
£
"o
3
•o
a
CO
O
0
0.5
1.5
H2O2/HNO3
(c) Northeast corridor regional simulation - base case (6 pm)
x ROG controls
® NOx controls
H2O2/HN03
(d) New York urban simulation - "MMM-7" meteorology (5 pm)
Figure 3-5 (continued): Predicted reduction in afternoon 03 (in ppb) resulting from a
35% reduction in the emission rate for anthropogenic ROG (crosses) and from a
35% reduction in the emission rate for NOx (circles) plotted against
H2O2/HN03 at the same time and location. Simulations are described in
Section 3.1. From Sillman (1995a, 1997a).
48
-------
ROG controls
NOx controls
•o -204: -•
o
0.5 1 1.5
H2O2/HNO3
(e) Atlanta urban simulation - modified upwind conditions. (5 pm)
60 ..x
.a
a.
o.
c
o
4-1
o
3
•o
O
CO
O
x ROG controls
* NOx controls
0.5 1 1.5
H2O2/HNO3
(f) Los Angeles urban simulation - Godowitch and Vukovich(1994) base case
(3pm)
Figure 3-5 (continued): Predicted reduction in afternoon 03 (in ppb) resulting from a
35% reduction in the emission rate for anthropogenic ROG (crosses) and from a
35% reduction in the emission rate for NOx (circles) plotted against
H2O2/HNO3 at the same time and location. Simulations are described in
Section 3.1. From Sillman (1995a, 1997a).
49
-------
^ 60 .:x
Si
0.
a
•a
0)
CO
O
ROG controls
NOx controls
0 0.2 0.4 0.6
H2O2/NOy
0.8
1
Figure 3-6 : Predicted reduction in afternoon 03 (in ppb) resulting from a 35% reduction
in the emission rate for anthropogenic ROG (crosses) and from a 35% reduction
in the emission rate for NOX (circles) plotted against H2O2/NOy at the same
time and location. Results are for the Los Angeles urban simulation by
Godowitch and Vukovich (1994). Based on the analysis reported in Sillman et
al. (1997a).
50
-------
CO
O
CM
O
CM
23456789 1011 12131415161718
N
O
Z
CM
O
CM
12345678 9101112131415161718
0.1
12345678 9101112131415161718
Simulation
Figure 3-7: 95th-percentile indicator value for ROG-sensitive chemistry (solid line) and
5th percentile indicator value for NOx-sensitive chemistry (dashed line) from 18
model scenarios identified in Table 3-2, for the indicator ratios H2O2/HNO3,
H2O2/NOZ, H2O2/NOy. In each simulation 95% of the locations with ROG
sensitive chemistry have indicator ratios below the solid line, while 95% of the
locations with NOx-sensitive chemistry have indicator ratios above the dashed
line.
51
-------
Table 3-2
VT~ _ _,.„ •*• , « H2O2 H2O2 . H2O2
NOx- and ROG-sensitive values for , -— and T as
photochemical indicators
The table gives the 95th percentile value for the distribution of each indicator ratio over ROG-
sensitive locations and the 5th percentile value for NOx-sensitive locations for various model
scenarios.
Model scenario
Lake Michigan regional simulations:
1. Base case
2. Doubled ROG
3. ROG reduced by half
4. Base case at 12:00 pm.
Northeast corridor regional simulations:
5. Base case
6. Zero isoprene
7. Low mixing
8. Doubled isoprene, low mixing
9. Reduced wind speeds
10. Increased deposition, reduced winds
New York urban simulations:
1 1 . MMM-7 meteorology
12. DWM-5 meteorology
Atlanta urban simulations:
13. BEIS1
14. Tripled isoprene
Los Angeles urban simulaions:
15. Godowitch base case
16. Wagner base case
17. Wagner, doubled ROG
18. Wagner, tripled ROG
Nashville (preliminary)
19. Base case
20. Older chemistry
21. Higher deposition (4 cm s"^)
H202
HNOs
95th
ROG
.24
.33
.15
.38
:40
.54
.55
.61
.43
.31
.55
.31
.51
.29
.59
.40
.44
1.3
.23
.24
.23
5th
NOX
.43
.28
.27
.68
.69
.68
.68
.68
.49
.42
.55
.59
.56
.67
.39
.43
.34
.45
.30
.38
.34
H202
NOZ
95th
ROG
.20
.22
.12
.24
.21
-36
.34
.28
.23
,15
.42
.27
.42
.23
.45
.25
.19
.37
.19
.19
.18
5th
NOX
.28
.20
.22
.41
.39
.54
.35
.28
.29
.20
.40
.50
.43
.45
.25
.29
.23
.21
.21
.25
.21
H202
NOy
95th
ROG
.16
.21
.11
.20
.17
.28
.32
.17
.17
.11
.30
.16
.18
.14
.24
.14
.11
.16
.12
.14
.11
5th
NOX
.22
.18
.20
.29
.31
.43
.32
.23
.23
.15
.27 .
.41
.29
.31
.22
.24
.20
.18
.17
.20
.16
52
-------
(c) Other indicators:
NOy: Milford et al. (1994) recommended the use of NOy as a NOX-ROG indicator. As
shown above in Figure 2-3, NOx-sensitive chemistry is associated with low NOy and
ROG-sensitive chemistry is associated with high NOy. The 95-th percentile NOx-sensitive
and 5th-percentile ROG-sensitive values for NOy are shown in Table 3-3 and Figure 3-1 1
below. NOy succeeds as a NOX-ROG indicator because ROG-sensitive chemistry is
associated with urban centers and large cities, which have characteristically high NOy,
while NOx-sensitive chemistry is associated with smaller cities and downwind and rural
locations that have low NOy. However the NOX-ROG transition for NOy changes
significantly in models when emission rates of anthropogenic and/or biogenic ROG are
changed. The variation in the NOX-ROG transition for NOy is illustrated in Table 3-3 and
Figure 3-1 1 below and in Sillman (1995a). This report recommends the use of XTQ— rather
than NOy as a NOX-ROG indicator because the former shows less sensitivity to model
assumptions.
3b and AIRTRAK: Graham Johnson (1984, 1990, see also Blanchard et al.,
J
1993) developed a correlation between NOX-ROG sensitivity and measured parameters
based on a series of smog chamber experiments. In the Johnson formulation, NOX-
sensitive chemistry is expected when chemistry has run to completion, i.e. most of the
initially emitted NOX has been converted to NOz species. For NOx-sensitive conditions the
rate of ozone production per emitted NOX is expected to reach a constant, maximum value
(~4 in smog chamber experiments). For ROG-sensitive conditions the rate of ozone
production per emitted NOX is expected to be lower. More sophisticated model
calculations by Liu et al. (1987) and Lin et al. (1988) suggest that the ozone production
efficiency (defined as the rate of ozone production per NOX removal, and associated with
the slope of ozone vs. NOz) shows significant variation.
The AIRTRAK method can be evaluated by using the ratio ^Q as a NOX-ROG
indicator in the photochemical models described in this report. The parameter O3b
represents background ozone, i.e. upwind conditions in the simulations (usually 40 ppb).
03-035
The exact formulation by Johnson is slightly different from ^Q — because it includes
modifications for production of particulate nitrate (assumed to be zero in the models used
53
-------
here). The correlation between model NOx- ROG sensitivity and VTQ is shown in
Figures 3-8, Figure 3-1 1 and Table 3-3.
Results show that — J^Q — does correlate successfully with NOX-ROG sensitivity.
The NOx-ROG transition point (3.5-5.5) is comparable with results from smog chambers
(4.1). The NOx-ROG correlation is less precise than the correlations for T^Q~ shown
above. There is also greater overlap between the NOx-and ROG-sensitive values of \f(O3-
03b, N0y) .
ROG
and .,,. ratios: Results for these ratios show a poor correlation with model
NOX -ROG sensitivity. Results are shown in Figures 3-9 and 3-10.
The use of - as a NOx-ROG indicator has never been formally suggested, but in recent
03
years it has been used as a popular substitute for J^Q— when NOy measurements are not
available (LADCO, 1994). Results (Figure 3-9) show that ^- does not correlate well
with NOx-ROG sensitivity. The ratio (O3, NOX+PAN) also does not correlate well with
NOx-ROG sensitivity.
ROG/NOx ratios have frequently been used to evaluate NOx-ROG sensitivity,
especially in regulatory applications (Blanchard et al., 1991). Correlations between NOx-
ROG sensitivity and various forms of the ROG/NOx ratio, including ROG/NOy and
reactivity-weighted ROG (expressed as propene-equivalent carbon, defined by Chameides
et al. (1992)) were tested in photochemical simulations. Results (Figure 3-10) show that
ROG/NOx ratios correlate poorly with model NOX-ROG sensitivity. It has previously
been shown that the ROG/NOx ratio in the morning is a poor predictor of NOx-ROG
sensitivity at the time of maximum ozone (see Section 2.1). Figure 3-10 suggests that
ROG/NOx ratios during the afternoon also correlate poorly with concurrent NOX-ROG
sensitivity even when the reactivity of the ROG mix is taken into account. The poor
correlation occurs because the various ROG/NOx ratios are associated with instantaneous
O3-NOx-ROG chemistry at the time and place of measurement but not with the process of
ozone production over a full day. It is possible that future research will identify ways of
using this ROG/NOx ratios as indicators for NOx-ROG dependence of instantaneous rates
of ozone production (as opposed to NOx-ROG sensitivity of ozone concentrations, which
has been investigated here.)
Cardelino et al. (1995) developed an observation-based method for evaluating
NOx-ROG chemistry based on measured NOX and ROG, which are readily available
54
-------
through the PAMS network. Their method includes an extensive analysis of instantaneous
NOX-ROG chemistry (derived from NOX and ROG measurements) to determine the
cumulative NOX-ROG sensitivity of ozone that is produced over a multi-hour period. The
poor results for ROG/NOX ratios as photochemical indicators in this study does not have
any implications concerning the accuracy of the method of Cardelino et al. (1995).
Measured ROG (especially speciated ROG and biogenics) and NOX should also be an
important part of model evaluations. However the conventional use of ROG/NOX ratios as
a "rule of thumb" for NOX-ROG chemistry is inaccurate.
HCHO
Sillman (1995a) used the ratio -J^Q— as a NOX-ROG indicator. This gave better
results than the other ROG/NOX ratios because HCHO is related both to ROG reactivity
and the chemistry over a period of several hours. Results have been included in Table 3-3
HC'HO
and Figure 3-11. The correlation between -jr^— and NOX-ROG sensitivity is plausible
but not as strong as the other indicators.
55
-------
JQ
a
a.
•a
a
CO
O
X ROG controls
• NOx controls
246
(O3-40 ppb)/NOy
8
Figure 3-8 : Predicted reduction in afternoon 03 (in ppb) resulting from a 35% reduction
in the emission rate for anthropogenic ROG (crosses) and from a 35% reduction
in the emission rate for NOX (circles) plotted against
(O3-40 ppb)/NOy at the same time and location. Results are for the Lake
Michigan urban simulation. From Sillman (1995a).
.Q
Q.
Q.
O
3
•o
O
CO
O
0
50 100 150
03/NOx
200
250
Figure 3-9 : Predicted reduction in afternoon 03 (in ppb) resulting from a 35% reduction
in the emission rate for anthropogenic ROG (crosses) and from a 35% reduction
in the emission rate for NOX (circles) plotted against O3/NOX at the same time
and location. Results are for the Lake Michigan urban simulation.
56
-------
c
o
°J
o
3
T3
O
O -60
10
20 30
ROG/NOx
40
50
(a)
o
3
TJ
0)
CO
O
ROG controls
* NOx controls
10 15
r-ROG/NOx
(b)
Figure 3-10 : Predicted reduction in afternoon 03 (in ppb) resulting from a 35%
reduction in the emission rate for anthropogenic ROG (crosses) and from a 35%
reduction in the emission rate for NOX (circles) plotted against (a) ROG/NOX
and (b) rROG/NOy at the same time and location, where rROG represents
reactivity-weighted ROG as propene equivalent carbon (Chameides et al. (1992)
in ppbC. Results are for the Lake Michigan urban simulation. Based on results
in Sillman(1995a).
57
-------
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
1 2345 67 8 9 10 11 12 13 14 15 16 17 18
123456789 101112131415161718
Simulation
Figure 3-11: 95th-percentile indicator value for ROG-sensitive chemistry (solid line)
and 5th percentile indicator value for NOx-sensitive chemistry (dashed line)
from 18 model scenarios identified in Table 3-3, for the indicators: NOy, (03-
40ppb)/NOy, HCHO/NOy. Results for NOy show the 95th percentile value for
NOx-sensitive chemistry (solid line) and 5th percentile for ROG-sensitive
chemistry because ROG-sensitive chemistry is associated with higher values for
this indicator.
58
-------
Table 3-3
NOx- and ROG-sensitive values for other photochemical indicators.
The table gives the 95th percentile value for the distribution of each indicator ratio over ROG-
sensitive locations and the 5th percentile value for NOx-sensitive locations for various model
scenarios, the indicator s shown here are not recommended for use.
Model scenario
NOy*
95th
NOX
5th
O3-40 ppb
NOZ
95th 5th
ROG NOX
HCHO
NOV
95th
ROG
5th
NOX
Lake Michigan regional simulations:
1. Base case 11 15
2. Doubled ROG 26 16
3, ROG reduced by half 12 15
4. Base case at 12:00 pm. 7.8 14
Northeast corridor regional simulations:
5. Base case 20 22
6. Zero isoprene 6.9 8.6
7. Low mixing 23 21
8. Doubled isoprene, low mixing 27 34
9. Reduced wind speeds 16 22
10. Increased deposition, reduced winds 15 20
New York urban simulations:
11. MMM-7 meteorology 20 25
12. DWM-5 meteorology 19 38
Atlanta urban simulations:
13. BEIS1 16 20
14. Tripled isoprene 20 31
Los Angeles urban simulaions:
15. Godowitch base case 26 24
16. Wagner base case 22 18
17. Wagner, doubled ROG 29 22
18. Wagner, tripled ROG 47 20
4.5
5.9
3.3
4.3
4.8
6.5
5.4
4.2
5.1
5.5
5.6
3.5
4.8
4.1
4.5
5.9
6.2
7.2
4.6
5.0
3.3
3.8
5.8
5.5
5.9
4.8
5.2
5.6
5.4
5.8
5.3
5.1
5.5
3.6
3.7
3.8
.27
.40
.21
.33
.40
.38
.51
.43-
.52
.56
.50
.20
.35
.38
.36
.44
.52
.67
.34
.38
.21
.42
.33
.24
.35
.45
.36
.41
.38
.29
.39
.55
.34
.28
.35
.41
* Results for NOy give the 95th percentile for NOx-sensitive locations and the 5th percentile for ROG-sensitive
locations because low NOy corresponds to NOx-sensitive chemistry.
59
-------
SECTION 4
SPECIES CROSS-CORRELATIONS AS A BASIS FOR
EVALUATING THE INDICATOR METHOD
Significant insights into the chemistry of ozone production can be gained by
examining cross-correlations between various species associated with ozone chemistry.
The use of species cross-correlations has been developed largely by Trainer et al. (1993)
and Parrish et al. (1993) who analyzed correlations between ambient ozone, NOy and
NOZ. In this section it will be shown that the method of photochemical indicators is closely
related to a predicted correlation between three species: 03, NOz and H2O2- This
predicted correlation can be used to evaluate the viability of the indicator method in general,
and can also be used as a diagnostic tool to evaluate the accuracy of individual field
measurements.
Trainer et al. (1993) has shown that measured 03 consistently increases with NOZ .
in measurements of photochemically aged air (defined by — <0.4) during the afternoon
at rural sites in eastern North America. As shown in Figure 4-1, the slope of 03 vs. NOZ
tends to be highest at low NOZ (<5 ppb) and decreases at higher NOZ, giving the O3-NOZ
correlation a nonlinear or "bent-over" appearance. Trainer et al. speculated that the
difference in the O3-NOZ slope at low and high NOZ was related to the switch between
NOx-sensitive and ROG-sensitive chemistry. They also reported that 03 increased with
NOZ in the Los Angeles basin with a slope that was significantly lower than the slope at
rural eastern sites.
Results from this report and from Sillman ( 1995a) suggest that NOX-ROG
03
chemistry is correlated with the absolute value of the Nr) ratio rather than the slope
between 03 and NOZ for an individual data set. As shown in Figure 4-2, model O3-NOZ
correlations show an ambiguous pattern when compared with model NOx-ROG sensitivity.
The characteristic nonlinear "bend-over" in the O3-NOZ correlation can appear for both
03
NOx-sensitive and ROG-sensitive conditions, depending on the — ratio. As shown in
60
-------
Figures 4-1 and 4-2, there is a characteristic decrease in the slope between 03 and NOZ as
NOZ increases from 3 to 10 ppb, even though these represent likely NOx-sensitive rural
measurements (Figure 4-1, Trainer et al., 1993) and NOx-sensitive simulation results. By
contrast, 03 increases at a near-constant rate as NOZ increases from 15 to 30 ppb in Figure
4-2, even though this region represents ROG-sensitive simulation results. From Figure 4-
2, it might be possible to associate NOx-sensitive locations with a high slope between 03
and NOZ, ROG-sensitive locations with a lower slope, and the region of transition between
NOX- and ROG-sensitive chemistry with a change in slope. However a stronger
03
correlation with NOx-ROG chemistry can be found by using the — ratio rather than the
slope as an indicator (see also Figures 3-la and 3-8).
The theory associated with photochemical indicators suggests that 03 should
increase with NOZ for both NOx-sensitive and ROG-sensitive chemistry but that the rate of
increase should be lower for ROG-sensitive chemistry. Under NOx-sensitive conditions
03 is expected to increase with NOZ at a rate determined by the production efficiency for
ozone (defined by Lin et al. (1988) as the rate of production of ozone divided by the rate of
removal of NOX, T CNO — ^), But for RQG-sensitive conditions the correlation between 03
and NOZ is determined by the supply of odd-H radicals, where the source is proportional to
03 and the NOZ represents the accumulated sink. The sum of odd-H sinks, approximately
equal to the sum NOZ+2H2O2, should increase linearly with 03. The predicted linear
correlation between 03 and the sum N02-+2H2O2 is a central feature of the indicator
theory.
As shown in Figure 4-3, correlations between 03 and NOZ should show a range of
variation between a maximum O3-NOZ slope, representing NOx-sensitive chemistry, and a
minimum O3/NOZ ratio that represents the ROG-sensitive limit. The correlation between
03 and the sum NOZ+2H2O2 shows a linear increase and represents the ROG-sensitive
Os •
limit for the ^7^- ratio (since for the ROG-sensitive limit NOZ»H2O2). The transition
INUZ
between NOx-sensitive and ROG-sensitive chemistry occurs when \f(O3,NOz) is greater
than \f(O3,NOz+2H2O2) by 50%, i.e. the NOx-sensitive region is defined by:
°3
NOZ ^
In simulations with high deposition (4 cm/sec for both H2O2 and HNO3) the ratio
/-.
J^Q- — 2"r Q was found to decrease by 25%. The NOX-ROG transition point for
also decreased by 25% but the NOx-ROG transition point for ^Q— decreased by only 10%
(see Tables 3-1 and 3-2). It is not yet clear whether the criteria given above
61
-------
03
> 1-5 -TTr ) can be used to establish the NOx-ROG transition but it
represents an important test for the theory.
The predicted linear correlation between 03 and the sum NOz+2H2O2 provides the
following tests for the indicator theory:
(1) The predicted linear correlation should be confirmed by ambient measurements. If
confirmed, the linear correlation provides evidence that the interpretation of 03, NOz and
H2O2 as NOX-ROG indicators is consistent with the ambient behavior of these species.
(2) It provides independent confirmation of the NOx-ROG transition point for ^r-.
The NOx-ROG transition points identified in Section 3.0 are sensitive to model
assumptions about sunlight, water vapor and dry deposition rates. These can be evaluated
03
for individual data sets by comparing the ratio MO W^ ra^os *n Pnotochemical
models. The NOX-ROG transition point for an individual data set can be established from
the formula: M7\I=l-5 NQ +2H9O?' ^"s can ^ comParc^ to me transition points
from the photochemical models in Section 3.
(3) It provides a diagnostic test for the accuracy of individual measurements. Deviation
from the expected correlation between 03, NOz and H2O2 may provide evidence that an
individual set of measurements is flawed. Similar diagnostic tests might be performed by
comparing 03 - NOz correlations with previous ambient measurements by Trainer et al.
Recent aircraft measurements by Daum et al. (1996) provide important confirmation
for the method. Daum et al. measured 03, NOx, NOy and H2O2 over the Atlantic Ocean
between Cape Cod and Nova Scotia, corresponding to air exported from the northeastern
U.S. The resulting correlation for 03 vs. NOz (Figure 4-4) shows a nonlinear increase,
comparable to the results from Trainer et al. (1993), with the slope for 03 vs. NOZ
decreasing at high NOZ. The same series of measurements showed a linear correlation
03
between 03 and the sum NOZ+2H2O2 (Figure 4-5). The ratio MQ +2H?O9 var*es ^rom
7 at low 03 to 5 at high 03 (1 10 ppb). These measurements show excellent agreement
03
with the photochemical model from Sillman (1995a), shown in Figure 4-3 (NQ +2H?O2 =
6). Staffelbach et al. (1997a,b) also measured 03, NOX, NOy and H2O2 during an air
pollution event in Switzerland. They found a similar linear correlation between 03 and the
sum NOz+2H2O2 but with a higher ratio CNQ +2H?O7 = ^' '^le n*&ner ratio is
attributed to higher deposition rates.
62
-------
The recent field measurement campaign (1995) in Nashville, TN, associated with
the Southern Oxidant Study, included extensive measurements of 03, NOX, NOy H2O2
and other related species (PAN, HNO3). Preliminary results (Sillman et al., 1997b)
showed a linear correlation between 03 and the sum NOz+2H2O2, similar to the results
reported by Daum et al. (1996) but with a slightly higher ratio (MQ .i/moQo =^' ^s
pattern is consistent with model results for Nashville and with results shown here, though
it supports the use of higher dry deposition estimates. A similar correlation and ratio was
found in aircraft measurements on two separate days (July 11 and 13, 1995). Aircraft
measurements during a third day (July 18) reported significantly higher ratio
=9^' P°ssiblv due to rainout of H2O2-
63
-------
150
100
I
1 1 1 r
1 1 r
J L
J L
o Bondville
o Egbert
a Scotia
A Wh'rtetop
_L
10 20
Products of NO, Oxidation (ppbv)
30
Figure 4-1 : Measured ozone as a function of NO 2 at four rural sites in eastern North
America (Bondville, IL, Egbert, ON, Scotia, PA and Whitetop, NC). For
Whitetop NOZ is represented as the sum of measured PAN, HNOs and NO 3*
For the other sites NOz is the difference between measured NOy and
From Trainer et aL (1993).
Q.
O.
180
160
140
120
100
80
60
40
20
0
0
20
NOy-NOx (ppb)
30
40
Figure 4-2 : Simulated afternoon concentrations of 03 vs. NOz for locations with NO*-
sensitive chemistry (triangles), ROG-sensitive chemistry (points) and
intermediate chemistry (circles) from the Lake Michigan regional simulation
base case (Sillman et al., 1993).
6A
-------
200
0
10 20
NOz, 2H2O2+NOZ (ppb)
30
Figure 4-3. Correlation between 63 and NOZ (crosses) and 03 and me sum
2H2O2+NOZ (dashes), all in ppb, from the northeast corridor regional
simulation base case. The lines correspond to an slope between 03 and NOZ of
9:1 and a ratio of 6:1, representing maximum and minimum values of 63 vs.
NOZ in NOx-sensitive and ROG-sensitive regions respectively.
65
-------
s - 54.0 + 4.6*(NOy-N02)
5 10
(NOy-NOp.ppbv
Figure 4-4 : 03 concentration (ppb) vs (NOy-NO2) from aircraft measurements over the
Atlantic Ocean. The data were collected at 10-second intervals, ordered by
(NOy-NO2) concentrations and divided into 10 intervals. Each point on the
plot represents the average NO y-NO2 corresponding to one of those intervals.
Horizontal bars represent the range of NOyN02 included in each interval;
vertical bars represent the standard deviation of the 03 concentration
measured over the interval. Slope and intercept were calculated from the
individual data points. From Daum et al. (1996).
35
Q.
Q.
CM
30 \r
25
20
15
10
5
0
20 40 60 80 100 120 140
03, ppbv
Figure 4-5 : The sum NOz+2H2O2 vs. 63 (ppb) from aircraft measurements over the
Atlantic Ocean. Sixty second ordered by 03 concentrations and divided into
10 intervals. Each point on the plot represents the average 03 corresponding
to one of those intervals. Horizontal bars represent the range of 03 included
in each interval; vertical bars represent the standard deviation of the
NOz+2H2O2 sums over the interval. Slope and intercept were calculated
from the individual data points. From Daum et aL (1996).
66
-------
SECTION 5
CASE STUDIES OF THE INDICATOR METHOD
FOR ATLANTA AND LOS ANGELES
This section presents two case studies that illustrate the possible use of the method
of photochemical indicators. The case studies are based on field measurement campaigns
during specific air pollution events in Atlanta and Los Angeles. Measured 03, NOX and
NOy have been used in combination with photochemical models to evaluate NOX -ROG
sensitivity in each city. It is hoped that these methods will serve as examples for similar
studies elsewhere!
An important facet of the case studies shown here is the development of a series of
model scenarios with different ROG-NOX chemistry. In the case study for Atlanta a
deliberate effort was made to generate plausible scenarios with both NOx-sensitive and
ROG-sensitive chemistry. This was achieved by varying the assumed rate of
anthropogenic and biogenic emissions (including both the BEIS1 and BEIS2 inventories)
and by varying wind speeds and mixed layer heights by 20%. In the case study for Los
Angeles scenarios were generated with anthropogenic ROG emissions increased by up to a
factor of three, resulting in a significant change in the predicted NOX-ROG response. Each
of these model scenarios generated different values for the indicator ratios XTrr and
NOy miu NOZ
which can then be evaluated against ambient measurements. A central feature of the
method is the development of model scenarios with different NOy-ROG responses, rather
than an attempt to "validate" a single model scenario or to only include scenarios with
identical NO^-ROG responses.
Results of this section also demonstrate that the measured indicator ratio KT/-. has
INUy
different values in a likely NOx-sensitive environment (Atlanta) in comparison with a likely
ROG-sensitive environment (Los Angeles). The difference in measured xT7=r~ between
Atlanta and Los Angeles corresponds closely to the difference between NOX- and ROG-
61
-------
sensitive environments predicted by the photochemical models in Section 3. However, the
NOx-ROG predictions are subject to a number of caveats: (1) NOx-ROG sensitivity varies
greatly from event to event, so that events in Atlanta with more stagnant meteorology might
have ROG-sensitive chemistry. (2) NOX-ROG sensitivity has been shown to vary with
location in Los Angeles (Milford et al., 1989), and downwind locations in the Los Angeles
basin are likely to have NOx-sensitive chemistry. (3) The NOX-ROG sensitivity results
are critically dependent on the accuracy of measured NOy.
The methods shown here would be strengthened if measured H2O2 were also
included. This is because the indicator ratios with H2O2 show a stronger correlation with
NOx-ROG chemistry and are less dependent on model assumptions, and because measured
H2O2 would allow the field measurements to be evaluated in terms of the O3-NOz-H2O2
correlation. The Nashville field measurement campaign associated with the Southern
Oxidant Study should provide an example of an analysis based on H2O2 (Sillman et al.,
1997b). A case study based on H2O2 was also carried out by Jacob et al. (1995) in rural
Virginia. They found that shifts from values associated with NOx-sensitive
chemistry in summer to values associated with ROG-sensitive chemistry in fall, possibly
associated with a seasonal change in photochemistry.
5.1 Atlanta
Model scenarios were developed for the air pollution episode of August 9-1 1 , 1992
in Atlanta. The simulations are based on the Urban Airshed Model (UAM-IV) and are
described in detail in Section 3. Simulations were performed using two different
inventories for anthropogenic emissions: an inventory developed by the Georgia
Department of Natural Resources(1987) based on the NAPAP 1985 national inventory(R)
and an updated inventory developed by Cardelino et al. (1994). Estimates for NOX
emissions from both point and area sources differ by a factor of two between the
inventories, and similar differences are found for ROG/NOx ratios. Smaller differences are
found for individual ROG. Separate model scenarios were also generated with biogenic
emissions from the BEIS 1 inventory and with tripled isoprene emissions, comparable to
the newer BEIS2 inventory (Geron et al., 1995). Additional scenarios were generated with
the height of the daytime mixed layer reduced by 20% from the initial estimate. The 20%
reduction was used to generate a plausible scenario with ROG-sensitive chemistry, and lies
within the range of uncertainty for vertical mixing identified by Marsik et al. (1995). The
resulting six scenarios are:
68
-------
(Al) 1987 anthropogenic emission inventory, and BEIS1 biogenic inventory,
equivalent to the base case in Sillman et al (1995b).
(A2) 1987 anthropogenic emission inventory with tripled isoprene.
(Bl) 1994 anthropogenic emission inventory, BEIS1 biogenics.
(B2) 1994 anthropogenic emission inventory with tripled isoprene.
(Cl) 1994 inventory with mixing heights reduced by 20%.
(C2) 1994 inventory with mixing heights reduced by 20% and tripled isoprene.
Aircraft-based measurements for 03 and NOy were made downwind of Atlanta
between 4:15 and 5:15 pm on both August 10 and August 11 (Imhoff et al., 1995). NOy
was measured using a gold tube converter with CO injection (Bellinger et al., 1983) which
was mounted external to the helicopter in the free air stream. The NO produced was drawn
through ~6 m of Teflon tubing to a Thermo Electron Instruments Model 42 nitrogen oxides
analyzer. Tests indicated that very little 03 (<3 ppb) survived the gold-tube converter and
therefore no adjustment was applied for the possible reduction of NO with 03 in the
tubing. Wind direction was from the north on August 10 and from the southwest on
August 11, and the location of the measurements was correspondingly different. On both
days the measurements covered a region extending from near downtown Atlanta to
approximately 40 km downwind! The spatial pattern of measured 03 and NOy confirmed
that the aircraft regularly intercepted a plume of polluted air, which was believed to be
associated with. Atlanta. The measurements were made at 500-600 m. altitude above
ground. .
Figure 5-1 shows measured values for 03 vs. NOy on August 10, 1992 in
comparison with model results from each of the six model scenarios. The measured 03-
NOy correlations show a broad region in which 03 increases with NOy, generally
associated with NOy between 5 and 15 ppb. This O3-NOy correlation is consistent with
previous rural measurements (Trainer et al., 1993). The measurements also show a range
of locations in which 03 does not increase with NOy, typically associated with higher
NOy. These uncorrelated points may represent locations close to emission sources where
NOx/NOy is large, so that photochemical production of 03 that would otherwise be
associated with the observed NOy has not yet occurred. The use of O3/NOy as a
photochemical indicator suggests that the locations with high NOy and low O3/NOy may
also represent ROG-sensitive chemistry.
The model results in Figure 5-1 represent a subset of the model domain including
downtown Atlanta and the downwind urban plume, at the same time (4-5 pm) and altitude
(600 km) as the measurements but not necessarily at the same horizontal location. Results
for all model scenarios show that 03 increases with NOy in a pattern broadly consistent
69
-------
with the observed Os-NOy correlation. However, there are some differences between the
simulated and observed O3-NOy correlation and differences between individual
simulations. The models show a much smaller region with uncorrelated 03 and NOy,
possibly because the model grid resolution (4x4 km) is too coarse to include NOX plumes
from individual emission sources. Model scenarios with the BEIS1 biogenic inventory all
show a lower rate of increase of 03 with NOy, both in comparison with the model
scenarios with increased isoprene and in comparison with measurements. Peak 03
occurred further downwind in the model (40 km south of the Georgia Tech campus) than in
measurements (20 km south of Georgia Tech), but the region with measured high 03
(>140 ppb) extends south and overlaps with the location of the model peak.
The most important evaluation for the six model scenarios is an examination of
measured 03 and concurrent NOy in the vicinity of the measured peak 03, in comparison
with peak 03 and concurrent NOZ in each model scenario. This comparison is especially
important because the model behavior in the vicinity of peak 03 has a dominant impact on
NOx-ROG control strategies, and the difference in the NOx-ROG responses of the six
model scenarios is most obvious for the peak 03 in each scenario. The near-continuous
sampling of 03 and NOy in the aircraft-based measurements provides an excellent basis
for characterizing the region of the Atlanta urban plume associated with peak 03. Because
photochemical models use 4 km grid resolution, we have used the range of measured 03
and NOy within 4 km of the measured peak 03 as a basis for characterizing the peak-O3
region. As shown in Figure 5-2, ozone within 4 km of the measured peak varies by 20.
ppb (130-150 ppb) while NOy varies between 7 and 15 ppb. This range of 03 and NOy is
used as a basis for evaluating the performance of individual model scenarios, i.e. model
peak 03 and concurrent NOy is expected to fall within the range of measured 03, NOy
and O3/NOy ratios found within 4 km of the measured peak 03.
Results (Figure 5-2) show that several model scenarios (Bl, Cl) generate peak
ozone that is close to the observed peak, but with concurrent NOy that is significantly
higher than the observed value. The poor performance of these scenarios in comparison
with measured 03 and NOy can also be seen in the complete data set (Figure 5-1), but is
highlighted when the evaluation focuses on peak 03 and concurrent NOy . Other model
scenarios (Al, A2) show good agreement with both peak 03 and concurrent NOy.
The indicator ratio i^/=r~ associated with peak 03 shows great variation among the
six model scenarios, reflecting the difference in model NOX-ROG sensitivity. The NOX-
sensitive model scenarios (Al, A2 and B2) all show J^Q— greater than the NOX-ROG
transition ratio (7). The ratios for scenarios Al and A2 (11-13) are significantly higher
70
-------
than the transition ratio. By contrast the ROG-sensitive scenario (Cl) has a ™- (=5.5)
03
lower than the transition ratio. This correlation between TTQ- and NOX-ROG sensitivity is
consistent with the results of Section 3.
Measured — in the vicinity peak O3 shows a range of values (10-20) that is
03
significantly higher than the NOx-ROG transition. This high »— is consistent with NOx-
sensitive chemistry. The measurements show good agreement with the strongly NOX-
sensitive model scenarios (Al and A2) and show poor agreement with the ROG-sensitive
scenario (Cl). It is important to note that the ROG-sensitive scenario shows good
agreement with measured peak 03. If model performance was evaluated solely in terms of
measured peak O^t. then scenarios AL A2. Bl and Cl would all be acceptable even though
they show different NOv_-ROG responses. The model-measurement comparison for the
indicator ratio >j/=r" in connection with peak O3 provides a stronger basis for evaluating
J-
model scenarios.
Results for both August 10 and August 1 1 are summarized in Table 2, are similar to
August 10. The model results and measurements both suggest that August 1 1 is more
likely to have NOx-sensitive chemistry than August 10.
71
-------
ScmrioM
Sc«ttrioA2
N0y(ppb)
10 IS
N0y(ppb)
20
25
Scenario 81
Scenario B2
10 15
NOy (ppb)
20
Sc«nu
-------
Atlanta, GA: August 10, 1992
1 O\J '
170 .
160 -
150 -
*-» 140 -
Q.
o. 130 •
CO
O 120 -
110 -
100 -
90 -
80 .
C
*
3
•
•
:
j
|
-
> -• . f
*
: ' V
i 1
N
l^1 "
:
V
•
/
/
0 1
Oy (ppb
D
/
/
5 2
)
C2
L/
/C1
0 2.
Figure 5-2. Peak 63 and concurrent NOy (ppb) in the Atlanta urban plume on 8-10-92.
The crosses represent measurements at 600 m elevation, 4-5 pm, for locations
within 4 km of the measured peak 63. Bold letters represent domain-wide peak
03 and concurrent NOy at 600 m elevation, 4-5 pm for the model scenarios
identified in the text. The line (O3/NOy=7) corresponds to the transition point
between NOX- and ROG-sensitive chemistry based on O3/NOy as a
photochemical indicator.
73
-------
Table 5-1
03, NOy and NOX-ROG sensitivity in simulations for Atlanta
The table gives simulated peak 03, concurrent NOy, peak 03 in simulations with 35% reductions
in anthropogenic ROG emissions and peak 03 in simulations with 35% reductions in NOx
emissions, for six model scenarios. Scenarios Al and A2 use anthropogenic emissions from
Georgia^; the other scenarios use Cardelinol 1. Scenarios Cl and C2 have daytime mixing heights
reduced by 20% relative to the base case. Scenarios Al, B1 and Cl have biogenic emissions from
(BEIS112; scenarios A2, B2 and C2 have isoprene emissions increased by a factor of three.
Observed values give the range of measured values within 4 km of observed peak 03.
August 10, 1992
Scenario
Al
A2
Bl
B2
Cl
C2
peak
03
132
141
132
157
149
177
NOy
11
11
15-21
22
16-27
25
reduced
ROG
126
137
126
150
143
167
reduced
NOX
118
122
127
144
.147
164
August 11, 1992
peak
03
131
136
147
166
165
184
NOy
9
9
12
17
16
19
reduced
ROG
128
132
141
163
159
180
reduced
NOX
112
113
128
145
153 .
160
Observed 130-150 7-15
123-148 8-12
74
-------
5.2 Los Angeles
Results for Los Angeles are based on simulations for the event of August 26-29,
1987 developed by Godowitch and Vukovich and by Wagner et al. (1992), described in
Section 3. These models used an anthropogenic emission inventory for southern California
developed in association with the Southern California Air Quality Study (SCAQS). In
addition, Wagner et al. (1992) generated alternative scenarios with emissions of
anthropogenic ROG increased by factors of two and three relative to the initial estimate.
NOX-ROG responses were tested relative to each of these cases.
Measurements of 03, NO, NO2, PAN, HNO3 and aerosol nitrate were made at
eight surface sites in the Los Angeles basin (Lawson et al., 1990). The measurement sites
were: downtown Los Angeles, Hawthorne, Long Beach City College, Anaheim,
Burbank, Azuza, Claremont and Rubidoux. HNO3 and paniculate nitrate were measured
using a transition-flow reactor (Ellestad et al., 1989), and 03, NO2, NOX and PAN were
measured using a Luminox detector (R-30). In this report it is assumed that NOy is equal
to the sum of the five measured nitrogen-containing species. Parrish et al.(1993) found
that these five species account for 90% of measured NOy in the eastern U.S. Aerosol
nitrate is not included in model photochemistry, but it is appropriate to assume that model
HNO3 is equal to the sum of measured HNO3 and aerosol nitrate in model-measurement
comparisons.
Figure 5-3 shows measured 03 vs. NOy and 03 vs. NOZ from all eight surface
measurement sites in the Los Angeles basin during the hours of 1-5 pm, August 27,1987.
The combined results from all sites provide enough measurements to identify correlations
among the species. Virtually no correlation appears between 03 and NOy, but there is a
strong correlation between 03 and NOZ. A similar correlation between 03 and NOZ has
appeared in previous measurements in Los Angeles (Tuazon et al., 1981, Trainer et al.,
1993).
As reported by Trainer et al., the correlation between 03 and NOZ in Los Angeles is
similar to the O3-NOZ correlations at rural sites in eastern North America. However
in Los Angeles is much lower than either the measured ratios at rural eastern sites and also
03
much lower than TT- reported above for Atlanta. The 03 and NOZ measurements for Los
03
Angeles from Figure 5-3 have an average O3/NOZ ratio equal to 4, and TT— associated
75
-------
with peak Os is equal to 6. This contrasts with both Atlanta ( j^j- = 14, above) and with
the rural eastern sites from Trainer et al. shown in Figure 4-1 ( T- = 15). If the Os/NOz
ratio is interpreted as a photochemical indicator, the low T- in Los Angeles would
suggest ROG-sensitive chemistry, in contrast with the higher - and NOx-sensitive
03
chemistry in Atlanta and the rural eastern sites. Measured ^— in Los Angeles is also
comparable with measured values for o °ver ^e ^ant^c O°ean fr°m Daum et
al. (1996), presented above (Figure 4-5). The latter ratio represents a likely limiting value
for >T- under ROG-sensitive conditions.
Figure 5-4 shows 63 and NOZ from simulations for Los Angeles by Wagner et al.
(1992) including scenarios with anthropogenic emissions equal to the base case and with
anthropogenic ROG increased by factors of two and three. The simulated values represent
the same time period (1-5 pm) and the same locations as the measurements. A comparison
between Figures 5-3 arid 5-4 is used to evaluate the performance of model scenarios. It is
apparent that the base case represents a significant underestimate of peak 03, although the
03 03
J^Q— ratio associated with peak 03 ( J^Q- =5) is reasonable in comparison with
measurements. The scenario with doubled anthropogenic ROG shows good agreement
• " " • fVi ' " f\^ fV>
both with measured 03 and with j (average - =6, - =5 concurrent with peak
03). The simulation with tripled anthropogenic ROG significantly overestimates
(average O3/NOZ= 7, J^Q- =9 concurrent with peak 03). This scenario shows greater
sensitivity to NOX ( see Wheeler et al., 1992),which is consistent with the higher
In contrast with the Atlanta event, evaluation of the three model scenarios for Los
Angeles are consistent with evaluations based solely on measured 03. The model
evaluation shown in Figures 5-3 and 5-4 suggest that the scenario with doubled ROG is
more accurate than either the base case or the scenario with tripled ROG. This conclusion
is consistent with the original analysis by Wheeler et al. (1992) and also with the analysis
of emission rates in southern California by Fujita et al. (1992). Fujita et al. found that
measured ROG/NOX ratios during the morning lower than ROG/NOX ratios in the UAM
base case by approximately a factor of two.
Results from this case study were constrained by the limited geographical extent of
the measurements. Milford et al. (1989) has shown that model NOX-ROG sensitivity in
76
-------
Los Angeles shows great geographical variation with ROG-sensitive chemistry downtown
and NOx-sensitive chemistry at downwind sites. The measurement sites included in this
analysis were almost all located in the region identified as ROG-sensitive in Milford et al.
A more complete analysis would require measurements that would identify the region of
transition between ROG-sensitive and NOx-sensitive chemistry.
77
-------
CO
O
300
250
200 -I
150. __/—*.-•
100-|-r
X
0
0
-------
^^
.a
n
a.
CO
250-
200-
150-
50-
0-
C
I
ysb-ajp--
J
) 20
NOz
i
-------
SECTION 6
SUMMARY AND CONCLUSIONS
This report has presented a series of results associated with the use of
photochemical indicators as a basis for investigating O3-NO\-ROG sensitivity. The
method of photochemical indicators seeks to identify species or species ratios that are
closely associated with NOX-ROG predictions in models. A successful correlation between
model NOx-ROG predictions and indicator values has been found for six species ratios:
03 03 03 H202 H202 . H2O2 T . . . ,
*" each case' hlSh values m associated
NO/ N0£ ffiTOs-'-iDTOs' ~NOi~'
with NOx-sensitive model predictions for ozone and low values are associated with ROG-
sensitive predictions. The close correlation between NOx-ROG predictions and indicator
values suggests that measured indicator values can be used as a basis for evaluating the
accuracy of model NOx-ROG predictions.
Indicator-NOx-ROG correlations have been examined for a wide variety of model
applications, including two different model types (UAM-IV and a regional model
developed at the University of Michigan) with different photochemical mechanisms. They
have been examined for several locations (Lake Michigan, northeast corridor, Atlanta and
Los Angeles). They have been examined in model scenarios with a range of assumptions
about emission rates, including scenarios with anthropogenic ROG emissions doubled or
reduced by half in comparison with inventory values and scenarios with radically different
emission rates for biogenic ROG. They have been examined in scenarios with changed
wind speeds and vertical mixing heights. They have been examined in scenarios with
strongly ROG-sensitive chemistry and in scenarios in strongly NOx-sensitive chemistry.
In all these cases, the correlation between model NOX-ROG predictions and indicator
values remains largely unchanged, even though the model NOx-ROG predictions vary.
It is especially noteworthy that changes in model assumptions that affect NOX-ROG
predictions also cause a corresponding change in the model values for photochemical
indicators. This feature is especially striking in the Atlanta case study (Table 5-1 and
Figure 5-2). In this case, model predictions for O3/NOy concurrent with peak 03 varied
80
-------
from 13 in the strongly NOx-sensitive model scenarios to 6 in the ROG-sensitive scenario.
The Atlanta case study is especially important because the NOx-sensitive and ROG-
sensitive scenarios each give similar predictions for peak 03, suggesting that an evaluation
vs. measured 03 does not provide a basis for confidence in model NOX-ROG predictions.
A similar contrast between NOx-sensitive and ROG-sensitive scenarios is apparent in the
scenarios for Lake Michigan with different ROG emission rates (Figures 3-la and b) and in
the simulations for New York /northeast corridor for events with different meteorology
(Figures 3-lc and 3-Id). The predicted contrast in indicator values between N0x-sensitive
and ROG-sensitive locations is partially confirmed by measurements in Atlanta and Los
Angeles. Model results (Figures 3-le and 3-lf) predict high values for O3/NOZ in Atlanta
and low values in Los Angeles. This prediction is consistent with measurements (Figures
5-1,5-2 and 5-3).
This study has identified a several factors that might be sources of error in the
indicator method. A list of caveats is given in Section 1. The most important uncertainties
are associated with the removal rate for the indicator species (especially through wet and
dry deposition); the possible role of particulate nitrate as a sink for NOy; and uncertainties
associated with the chemistry of organic peroxides, which are not represented in some of
the chemical mechanisms that are frequently used for air quality analysis (e.g. CB-IV). It
is especially imporant to identify uncertainties that might lead to bias in the NOX-ROG
interpretation of indicator values.
A more general problem associated with the method of photochemical indicators is
that, as with all methods designed to predict the sensitivity of 03 to NOX and ROG, it is
very difficult to find direct confirmation that the predictions are accurate. This report has
recommended investigation of model and measured correlations between 03, NOZ and
H2O2 as a basis for evaluating model assumptions that represent uncertainties in the
indicator method. The models that were used to derive indicator-NOx-ROG correlations
also predict a linear correlation between 03 and the sum: NOZ+2H2O2- This prediction
should be verified vs. measurements. In addition, model results predict that the ratio \f(O35
NOZ+2H2O2) should have similar values in NOx-sensitive and ROG-sensitive locations
(although it may show diurnal variations and day-to-day variation in responses to changing
cloud cover and the effect of wet deposition). Because this ratio is closely associated with
the proposed indicators (especially with O3/NOZ), it might also be used as a basis for
evaluating the accuracy of the proposed NOX-ROG transition values associated with the
indicators.
81
-------
REFERENCES
Benjamin S. G., K. J. Brundage and L. L. Morone, Implementation of the Rapid Update Cycle,
Part I: Analysis/Model Description, NOAA/NWS Technical Procedures Bulletin, Series No.
416, 1994.
Blanchard, C. L., P. M. Roth and G. Z. Whitten. The influence of NOx and VOC emissions on
ozone concentrations in rural environments. Electric Power Research Institute, Palo Alto, CA,
May, 1991.
Blanchard, C. L., P. M. Roth and H. E. Jeffries. Spatial mapping of preferred strategies for
reducing ambient ozone concentrations nationwide. Paper #93-TA-37A.04, presented at the
Air and Waste Management Association 86th Annual Meeting and Exposition, Denver, CO,
June 13-18, 1993.
Bellinger, M. J., R. E. Sievers, D. W. Fahey, and F. C. Fehsenfeld, Conversion of nitrogen
dioxide, nitric acid and n-propyl nitrate to nitric oxide by gold-catalyzed reduction with carbon
monoxide. Ana. Chem., 55, 1980-1986, 1983.
Bott, A., A positive definite advection scheme obtained by nonlinear renormalization of the
advective fluxes. Man. Wea. Rev., 117,1006-1015, 1989.
Buhr, M., D Parrish, J. Elliot, J. Holloway, J. Carpenter, P. Goldan, W. Kuster, M. Trainer, S.
Montzka, S. McKeen, and F. C. Fehsenfeld, Evaluation of ozone precursor source types
using principal component analysis of ambient air measurements in rural Alabama. J.
Geophys. Res., 100, 22853-22860, 1995.
Cardelino, C., W-L. Chang and M. E. Chang, 1994: Comparison of emissions inventory
estimates and ambient concentrations of ozone precursors in Atlanta, Georgia. Presented at the
Air and Waste Management Association International Conference on the Emission Inventory:
Applications and Improvement, Raleigh, N.C., November 1-3, 1994.
Cardelino, C. and W. L. Chameides. An observation-based model for analyzing ozone-precursor
relationships in the urban atmosphere. J. Air Waste Manage. Assoc., 45, 161-180, 1995
Carter, W. P. L. Development of ozone reactivity scales for volatile organic compounds, J. Air
Waste Manage. Assoc., 44, 881-899, 1994.
Carter, W. P. L., Computer modeling of environmental chamber studies of maximum incremental
reactivities of volatile organic compounds, Atmos. Environ., 29-18, p. 2513, 1995.
82
-------
Chameides, W. L., F. Fehsenfeld, M. O. Rodgers, C. Cardellino, 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, 6037-6056, 1992.
Chameides, W. L., D. D. Davis, J. Bradshaw, S. Sandholm, M. Rodgers, B. Baum, B. Ridley,
S. Madronich, M. A. Carroll, G. Gregory, H. I. Schiff, D. R. Hastie, A. Torres, E. Condon,
Observed and model-calculated NO2/NO ratios in tropospheric air sampled during the NASA
GTE/CITE2 field study. /. Geophys. Res., 95, 10235-10247, 1990.
Chameides, W. L., R. W. Lindsay, J. Richardson, C. S. Kiang, The role of biogenic
hydrocarbons in urban photochemical smog: Atlanta as a case study, Science, 241,1473-
1474, 1988.
Chang, M. E., D. Hartley, C. Cardelino, and W.-L. Chang, Temporal and spatial distribution of
biogenic emissions of isoprene based on an inverse method using ambient isoprene
observations from the 1992 Southern Oxidants Study, Atlanta Intensive, Atmos. Environ., in
press, 1996.
Chock, D.P. A comparison of numerical methods for solving the advection equation - HI. Atmos.
Environ., 25A, 853-871, 1991.
Daum, P. H., L. I. Kleinman,L. Newman, W. T. Luke, J. Weinstein-Lloyd, C. M. Berkowitz
and K. M. Busness, Chemical and physical properties of anthropogenic pollutants transported
over the North Atlantic during NARE, /. Geophys. Res., 101, 29029-29042, 1996.
DeMore, W. B., S. P. Sander, D. M. Golden, R. F. Hampson, M. J. Kurylo, C. J. Howard, A.
R. Ravishankara, C. E. Kolb, and M. J. Molina. Chemical kinetics and photochemical data
for use in stratospheric modeling. JPL 92-20, Jet Propulsion Laboratory, NASA, 1992.
Drummond, J., H. Schiff, D-. Karecki, G. Mackay. Measurements of NO2, 03, PAN, HNOs,
H2O2 an H2CO during the Southern California Air Quality Study. Presented at the 82nd
annual meeting of the Air and Waste Management Association, Anaheim, CA, 1989; Paper 89-
139.4.
Ellestad, T. G., L. Stockburger, and K. T. Knapp. Measurements of nitric acid and paniculate
nitrate by transition-flow reactor during the 1987 Southern California Air Quality Study.
Presented at the 82nd annual meeting of the Air and Waste Management Association, Anaheim,
CA, 1989; Paper 89-152.2.
Environmental Protection Agency (EPA) Development of the 1980 NAPAP emissions inventory,
EPA-600/7-86~057a, Environmental Protection Agency, Research Triangle Park, N.C. 1986.
Environmental Protection Agency (EPA). The 1985 NAPAP emissions inventory (version 2):
development of the annual data and modelers' tapes, EPA-600/7-89-012a, Environmental
Protection Agency, Research Triangle Park, NC, 1989.
Environmental Protection Agency (EPA), Regional Interim Emission Inventories (1987-1991),
Volume I and n, EPA-454/R93-021a and b, Environmental Protection Agency, Research
Triangle Park, N.C., 1993.
Fujita, E. M., B. E. Croes, C. L. Bennett, D. R. Lawson, F. W. Lurmann and H. H. Main,
Comparison of emission and ambient concentration ratios of CO, NOx, and NMOG in
California's south coast air basin. J. Air Waste Mgmt. Assoc., 42:264-276 1992.
83
-------
Georgia Department of Natural Resources, 1987: Georgia's State Implementation Plan for Ozone
in the Atlanta Area. Air Protection Branch.
Geron, C. D., A. B. Guenther, and T. E. Pierce, An improved model for estimating emissions of
volatile organic compounds from forests in the eastern United States. /. Geophys. Res., 99,
12773-12791, 1994.
Geron, C. D., T.E. Pierce and A.B. Guenther, Reassessment of biogenic volatile organic
compound emissions in the Atlanta area, Atmos. Environ., 29, 1569-1578, 1995
Gery, M. W. G. Z. Whitten, J. P. Killus and M. C. Dodge. A photochemical kinetics mechanism
for urban and regional computer modeling. J. Geophys. Res., 94,12925-12956, 1989.
Godowitch, J. M. and J. M. Vukovich, 1994: Photochemical urban airshed modeling using
diagnostic and dynamic meteorological fields. Presented at the 87th Air and Waste
Management Association Meeting and Exhibition, Cincinnati, OH, June 19-24, 1994.
Hall, B. D. and C. S. Claibom, Measurements of the dry deposition of peroxides to a Canadian
boreal forest, J. Geophys. Res., in press, 1997.
Hanna, S. R., G. E. Moore and M. E. Fernau. Evaluation of photochemical grid models (UAM-
IV, UAM-V, and the ROM/UAM-FV couple) using data from the Lake Michigan Ozone Study
(LMOS). Atmos. Environ., 30, 3265-3279, 1996.
Harley, R. A., A. G. Russell, G. J. McRae, G. R. Cass, and J.'H. Seinfeld. Photochemical
modeling of the Southern California Air Quality Study. Environ. Sci. Technol., 27, 378-388,
1993.
Hillery, J. F. Spatial and temporal distribution of aloft ozone and oxides of nitrogen in a
mesoscale lakeshore environment In Ranzieri, A. J. and P. A. Solomon, eds., Regional
Photochemical Measurement and Modeling Studies: The Proceedings of an International
•.Specialty Conference, San Diego, CA, November 8-12,1993. Air and Waste Management
Association, 1995.
Imhoff, R. E., R. J. Valente, J. F. Meagher and M. Luria, 1995: The production of 03 in and
urban plume: airborne sampling of the Atlanta urban plume. Atmos. Environ.., 29, 2349-
2358.
Jacob, D. J., B. G. Heikes, R. R. Dickerson, R. S. Artz and W. C. Keene. Evidence for a
seasonal transition from NOX- to hydrocarbon-limited ozone production at Shenandoah
National Park, Virginia. J. Geophys. Res., 100, 9315-9324, 1995.
Johnson, G. M., A simple model for predicting the ozone concentration of ambient air. Proc.
Eighth Inter. Clean Air Confer., Melbourne, Australia, May 2,1984, p. 715-731.
Johnson, G. M., S. M. Quigley, and J. G. Smith. Management of photochemical smog using the
AIRTRAK approach. 10th International Conference of the Clean Air Society of Australia and
New Zealand, Aukland, New Zealand, March, 1990, p. 209-214.
Kirchner, F., and W. R. Stockwell, The effect o peroxy radical reactions on the predicted
concentrations of ozone, nitrogenous compounds and radicals. J. Geophys. Res., 101, 21007-
21023, 1996.
84
-------
Kleinman, L. I., Photochemical formation of peroxides in the boundary layer. J. Geophys. Res.,
97,10889-10904, 1986.
Kleinman, L. I., Seasonal dependence of boundary layer peroxide concentration: the low and high
NOX regimes. /. Geophys. Res., 96, 20721-20734, 1991.
Kleinman, L. L, Low and high NOX tropospheric photochemistry. J. Geophys. Res., 99,
16831-16838, 1994.
Kumar, N., M. T. Odman, and A. G. Russell. Multiscale air quality modeling: Application to
southern California. J. Geophys. Res., 99, 5385-5397, 1994.
Lake Michigan Air Directors Consortium (LADCO). Lake Michigan Ozone Study: Evaluation of
the UAM-V Photochemical Grid Model in the Lake Michigan Region. Version 2.0. Submitted
to U.S. Environmental Protection Agency by Lake Michigan Air Directors Consortium, Des
Plaines, IL, September, 1994.
Lamb, B., Westberg, H., Allwine, G. and Quarles, T. (1985) Biogenic hydrocarbon emissions
from deciduous and coniferous trees in the United States, J. Geophys. Res., 90, 2380.
Lawson, D. R., The Southern California Air Quality Study. J. Air. Waste Manage. Assoc., 40,
156-165, 1990.
Logan, J. A., Ozone in rural areas of the United States, J. Geophys. Res., 94, 8511-8532,
1989.
Lu, R. and R. P. Turco. Air pollution transport in a coastal environment: Part II: three-
dimensional simulations over the Los Angeles basin. Atmos, Environ., 29, 1499-1518, 1995.
Luke, W. T., and R. R. Dickerson, The flux of reactive nitrogen compounds from eastern North
America to the western Atlantic Ocean, Global Biogeochem. Cycles, 1, 329-343,1987.
Lurmann, F. W., Lloyd, A. C, and Atkinson, R. A chemical mechanism for use in long-range
transport/acid deposition computer modeling. J. Geophys. Res. 91, 10905-10936, 1986.
Madronich, S. Photodissociation in the atmosphere: 1. Actinic flux and the effect of ground
reflections and clouds, J. Geophys. Res. 92, 9740-9752, 1987.
Marsik, F. M., K. Fischer, T. D. McDonald and P. J. Samson, 1993: Comparison of methods
for estimating mixing height used during the 1992 Atlanta field intensive./ Appl. Meteo., 34,
1802-1814, 1995.
Matthews, E. Global vegetation and land use: new high-resolution data bases for climate studies.
J. Climate Appl. Meteorol, 22, 474-487, 1983.
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., Gao, D., Sillman, S., Blossey, P., and Russell, A. G., Total reactive nitrogen
(NOy) as an indicator for the sensitivity of ozone to NOX and hydrocarbons. J. Geophys.
Res., 99, 3533-3542, 1994.
85
-------
Milford, J. B., A. G. Russell, and G. J. McRae. A new approach to photochemical pollution
control: Implications of spatial patterns in pollutant responses to reductions in nitrogen oxides
and reactive organic gas emissions. Environ. Sci. Tech., 23, 1290-1301, 1989.
Morris, R. E. and Myers, T. C, 1990: User's guide for the urban airshed model, Vol. I-V. EPA-
450/4-90-007A-E.
National Research Council (NRC), Committee on Tropospheric Ozone Formation and
Measurement. Rethinking the Ozone Problem in Urban and Regional Air Pollution, National
Academy Press, 1991.
Ozone Transport and Assessment Group (OTAG). Summary of Basecase A OTAG UAM-V
modeling: BEIS1 and BEIS2. Presented at the OTAG Regional and Urban Scale Modeling
Workgroup meeting, May 2,1996.
Parrish, D. D., M. Trainer, M. P. Buhr, B. A. Watkins, and F. C. Fehsenfeld. Carbon Monoxide
concentrations and their relation to concentrations of total reactive oxidized nitrogen at two
rural U.S. sites. J. Geophys. Res., 96, 9309-9320, 1991.
Parrish, D. D., M. P. Buhr, M. Trainer, R. B. Norton, J. P. Shimshock, F. C. Fehsenfeld, K.
G. Anlauf, J. W. Bottenheim, Y. Z. Tang, H. A. Wiebe, J. M. Roberts, R. L. Tanner, L.
Newman, V. C. Bowersox, K. J. Olszyna, E. M. Bailey, M. O. Rodgers, T. Want, H.
Berresheim, U. K. Roychowdhury, and K. L. Demerjian, The total reactive oxidized nitrogen
levels and the partitioning between individual species at six rural sites in eastern North
America. /. Geophys. Res., 98, 2927-2939, 1993.
Paulson, S. E. and J. H. Seinfeld. Development and evaluation of a photooxidation mechanism
for isoprene. J. Geophys. Res., 97, 20703-20715, 1992.
Pierce, T. E., B. K. Lamb and A. R. Van Meter, 1990: Development of a biogenic emissions
inventory system for regional scale air pollution models. Presented at the 83rd Air and Waste
Management Association Annual Meeting, Pittsburgh, PA, Paper No. 90-94.3, June 24-29,
1990.
Prather, M. J., Numerical advection by conservation of second-order moments, J. Geophys Res.,
91, 6671-6681, 1986.
Rao, S. T. and G. Sistla, Efficacy of nitrogen oxides and hydrocarbon emissions controls in ozone
attainment strategies as predicted by the Urban Airshed Model. Water Air Soil Pollut., 67,95-
116, 1993.
Ridley, B. A., S. Madronich, R. B. Chatfield, J. G. Walega, R. E. Shelter, M. A. Carroll, and D.
D. Montzka, Measurements and model simulations of the photostationary state during the
Mauna Loa Observatory Photochemistry Experiment: Implications for radical concentrations
and ozone production and loss rates. J. Geophys. Res., 97, 10375-10388, 1992.
Roberts, J. M., R. L. Tanner L.Newman, V. C. Bowersox J. W. Bottenheim, K. G. Anlauf, K.
A. Brice, D. D. Parrish, F. C. Fehsenfeld, M. P. Buhr, J. F. Meagher, and E. M. Bailey,
Relationships between PAN and ozone at sites in eastern North America. J. Geophys. Res.,
100, 22821-22830, 1995. :
Roselle, S. J. and K. L. Schere. Modeled response of photochemical oxidants to systematic
reductions in anthropogenic volatile organic compound and NOx emissions. J. Geophys.
Res., 100, 22929-22941, 1995.
86
-------
Sillman, S., Logan, J. A. and Wofsy, S. C. The sensitivity of ozone to nitrogen oxides and
hydrocarbons in regional ozone episodes. /. Geophys. Res., 95, 1837-1851, 1990a.
Sillman, S., Logan, J. A. and Wofsy, S. C. A regional-scale model for ozone in the United
States with a sub-grid representation of urban and power plant plumes, J. Geophys. Res, 95,
5731-5748, 1990b.
Sillman, S., P. J. Samson and J. M. Masters, Ozone production in urban plumes transported over
water: photochemical model and case studies in the northeastern and midwestern U.S. J.
Geophys. Res., 98, 12687-12699, 1993.
Sillman, S., The use of NOy, H2O2 and HNOs as indicators for ozone-NOx-ROG sensitivity in
urban locations. J. Geophys. Res., 100, 14175-14188, 1995a.
Sillman, S. and Samson, P. J., The impact of temperature on oxidant formation in urban, polluted
rural and remote environments. J. Geophys. Res., 100, 11497-11508, 1995b.
Sillman, S., K. Al-Wali, F. J. Marsik, P. Nowatski, P. J. Samson, M. O. Rodgers, L. J.
Garland, J. E. Martinez, C. Stoneking, R. E. Imhoff, J-H. Lee, J. B. Weinstein-Lloyd, L.
Newman and V. Aneja. Photochemistry of ozone formation in Atlanta, GA: models and
measurements. Atmos. Environ., 29, 3055-3066, 1995c.
Sillman, S., D-Y. He, C. Cardelino and R. E. Imhoff, The use of photochemical indicators to
evaluate ozone-NOx-hydrocarbon sensitivity: case studies from Atlanta, New York and Los
Angeles. J. Air Waste Manage.. Assoc. , 47, 1030-1040, 1997a.
Sillman, S., D. He, M. Pippin, P. Daum, L. Kleinman, J. H. Lee and J. Weinstein-Lloyd.
Model correlations for ozone, reactive nitrogen and peroxides for Nashville in comparison with
measurements: implications for VOC-NOx sensitivity. Submitted to J. Geophys. Res.,
upcoming special section on the Middle Tennessee Ozone Study, 1997b.
Sillman, S., Review Article' The Relation between Ozone, NOX and Hydrocarbons in Urban and
Polluted Rural Environments. Submitted to J. Geophys. Res., upcoming special section on
the Middle Tennessee Ozone Study, 1997c.
Smolarkiewicz, P. K. A simple positive definite advection scheme with small implicit diffusion,
Mon. Wea. /tev.,111, 479-486, 1983.
Staffelbach, T., A. Neftel, A. Blatter, A. Gut, M. Fahrni, J. Stahelin A. Prevot, A. Hering, M.
Lehning, B. Neininger, M. Baumie, G. L. Ko, J. Dommen, M. Hutterli and M. Anklin.
Photochemical oxidant formation over southern Switzerland, part I: Results from summer,
1994. J. Geophys. Res., in press, 1997a.
Staffelbach, T., A. Neftel and L. W. Horowitz. Photochemical oxidant formation over southern
Switzerland, part II: model results. J. Geophys. Res., in press, 1997b.
Trainer, M., D. D. Parrish, M. P. Buhr, R. B. Norton, F. C. Fehsenfeld, K. G. Anlauf, J. W.
Bottenheim, Y.Z. Tang, H.A. Wiebe, J.M. Roberts, R.L. Tanner, L. Newman, V.C.
Bowersox, J.M. Maugher, K.J. Olszyna, M.O. Rodgers, T. Wang, H. Berresheim, and K.
Demerjian, 1993: Correlation of ozone with NOy in photochemically aged air. J. Geophys.
Res., 98, 2917-2926.
87
-------
Tuazon, E. C., A. M. Winer, and J. N. Pitts, Jr. Trace pollutant concentrations in a multiday
smog episode in the California south coast air basin by long path length Fourier Transform
Infrafred Spectroscoy. Environ. Sci. Technol., 15, 1232-1237, 1981.
Wagner, K. W., N. J. M. Wheeler and D. L. McNerny, 1992: The effect of emission inventory
uncertainty on Urban Airshed Model sensitivity to emission reductions. Presented at the Air
and Waste Management Association International Specialty Conference, "Tropospheric Ozone:
Nonattainment and Design Value Issues", Boston, October, 1992.
Wheeler, N. J. M., and K. K. Wagner, SCAQS Modeling Project Documentation. Technical
Support Division Report, California Air Resources Board, Sacramento, 1992.
White, W. H., D. E. Patterson and W. E. Wilson, Jr. Urban exports to the nonurban troposphere:
results from project MISTT. /. Geophys. Res., 88, 10745-10752, 1983.
Ulrickson, B. L. and C. F. Mass, Numerical investigation of mesoscale circulations over the Los
Angeles basin. Parti: a verification study. Man. Wea. Rev., 118, 2138-2161, 1990.
88
-------
APPENDIX:
ADDITIONAL PLOTS OF NOX-ROG SENSITIVITY vs.
PHOTOCHEMICAL INDICATORS
The appendix shows correlations between NOX-ROG sensitivity and photochemical
indicators for the complete set of model results. Each figure shows the predicted reduction
in 03 for the specified hour (usually 3-6 pm) resulting from a 35% reduction in
anthropogenic ROG emissions (crosses) and from a 35% reduction in NOX emissions
(circles) at each location in the model domain. The simulated ozone reductions are plotted
vs. the simulated value for the specified indicator ratio in the initial model scenario (without
emission reductions) at the same time and location. Simulations are described in Section
3.1.
89
-------
0
Lake Michigan, ROG reduced by half - O3/NOz
25
-60
X ROG controls
NOx controls
0
20
25
Lake Michigan, base case at noon - O3/NOz
90
-------
XJ
Q.
C
^0
"o
3
•o
0)
CO
O
-40-
0
X ROG controls
NOx controls
25
Northeast corridor, zero isoprene - O3/NOz
.n
a
^.
c
o
I
-a
0)
CO
O
X reduced ROG
° reduced NOx
03/NOy-NOx
Northeast corridor, doubled isoprene: O3/NOz
91
-------
reduced mixing height
X ROG controls
NOx controls
0
03/NOy-NOx
Northeast corridor, low mixing: O3/NOz
0
25
Northeast corridor, light winds and high deposition: O3/NOz
92
-------
j5
g- 40-
^ 20-
tS 0.
o
•a -20-
0>
" -40-
«
0 -60.
_j
•^*P$^
1
V
e© do
vr vv
X ROG controls
® NOx controls
0
10 15
03/NOz
20
25
New York, DWM-5 meteorology: O3/NOz
c
o
•o
o
CO
X ROG controls
NOx controls
-60
25
Atlanta, increased isoprene: O3/NOz
93
-------
0
Los Angeles, Wagner base case: O3/NOz
a.
o.
c
o
TJ
0)
CO
O
Los Angeles, Wagner with doubled ROG: O3/NOz
94
-------
.fl
a
Q.
C
O
^
u
3
•o
a>
-200 -E
0
10
15
OS/NOz
20
25
Los Angeles, Wagner with tripled ROG: 03/NOz
O3 reduction (ppb
x ROG controls
NOx controls
15
20
Lake Michigan base case O3/NOy
95
-------
^ 60
.Q
£ 40-t
X ROG controls
• NOx controls
o
Lake Michigan, doubled ROG emissions: O3/NOy
20
X ROG controls
® NOx controls
20
Northeast corridor base case O3/NOy
96
-------
y)A^X v J»
-*---—
X ROG controls
NOx controls
0
20
New York with MM7 meteorology: O3/NOy
^ 60
a.
a.
o
•o
0)
CO
O
O3/NOy
Atlanta, BEIS1 biogenics: O3/NOy
97
-------
:'xxxx:;—X-
e
X ROG controls
NOx controls
15
20
Los Angeles (Godowitch): O3/NOy
^ 60 T
Q. 40- ... ..'-
xx<;xx;xx
-------
.a
a.
c
-------
^ 60
.Q
40-1:
c ^u
1 o
•o -20 -E
o>
"- -40-E ........ -.
CO
O -60
>»i^>X;>&XX-X-
0
10
X reduced ROG
° reduced NOx
20
O3/NOy
Northeast corridor, low mixing and doubled isoprene: O3/NOy
60
40. .......jv. ;
•g -20. ....e.®
o>
" -40 J:
eo
O -60
0
10
X ROG controls
• NOx controls
20
03/NOy
Northeast corridor, low mixing: O3/NOy
100
-------
0
10 15
03/NOy
X ROG controls
® NOx controls
20
25
Northeast corridor, light winds and high deposition: O3/NOy
a 40-
^ 20-
0
"•? 0 .
0
•a -20-
V
u -40 -
CO
° -60.
£
;
_.-.-.-..._
"
^<<_X
L ______
0 5 10
O3/NOy
»°
(X-
e & 9
y V"X
X ROG controls
• NOx controls
15 20
New York DWM-5 meteorology: O3/NOy
101
-------
.Q
Q.
a.
c
10
•a
0)
eo
O
0
10
03/NOy
X ROG controls
• NOx controls
15
20
Atlanta, increased biogenis: O3/NOy
o.
D.
C
-------
0
Los Angeles, Wagner with doubled ROG: O3/NOY
a
a.
c
o
o
3
O
CO
O
-200
0
10
O3/NOY
15
20
Los Angeles, Wagner with tripled ROG: O3/NOy
103
-------
ROG controls
NOx controls
20 40
O3/HNO3
60
Lake Michigan base case: O3/HNO3
J3
Q.
£.
C
o
•o
o
CO
X ROG controls
NOx controls
-40-E
0
20 40
03/HN03
60
Lake Michigan, doubled ROG emissions: O3/HNO3
104
-------
.a
a
c
o
o
3
•O
O
CO
O
X ROG controls
NOx controls
20 40
O3/HNO3
60
Northeast corridor base case: O3/HNO3
.0
Q.
C
O
*3
O
3
•o
O
CO
O
X ROG controls
NOx controls
-40 - !-- -,
20 40
O3/HNO3
New York with MM7 meteorology: O3/HNO3
105
-------
.Q
Q.
C
o
3
TJ
X ROG controls
® NOx controls
r»oaBy»a W" ~TQ| V
:x:xx:x-)x-x-x-x—
0
20
40
60
O3/HNO3
Atlanta, BEIS1 biogenics: O3/HNO3
X ROG controls
• NOx controls
Los Angeles (Godowitch): O3/HNO3
106
-------
JD
Q.
5;
c
£
O
3
•o
0)
X ROG controls
® NOx controls
-40 - ! -,
20 40
03/HN03
Northeast corridor, light winds and high deposition: O3/HNQ3
Q.
Q.
i
O
3
•o
0
CO
O
0
40-
20-
0 -
-20-
-40-
-60.
fflffijtMci^
« .^.•XCCJV.w>v>vx
;
;
____________J
X ROG controls
® NOx controls
20 40
O3/HNO3
60
New York, DWM-5 meteorology: O3/HNO3
107
-------
J2
a
o.
c
xixx>xx<
0 0.5 1 1.5
H2O2/HNO3
Lake Michigan, base case at noon: H2O2/HNO3
108
-------
.D
a
a.
c
o
*3
O
3
TJ
0)
i_
CO
O
ROG controls
NOx controls
0
0.5 1 1.5
H2O2/HNO3
Northeast corridor with zero isoprene: H2O2/HNO3
reduced ROG
reduced NOx
0.5 1 1.5
H2O2/HNO3
Northeast corridor, low mixing and doubled isoprene: H2O2/HNO3
109
-------
o.
a.
c
o
•a
a
CO
O
xx:xxx»x>>x
-------
^ 60
£J
Q.
c
3= 0
o
3
•o
0)
CO
o
-20-
-40-
-60
0
i
i
. _i.
X ROG controls
® NOx controls
^-i^?^^X:.y?x: xv
•->-
•
-r-
i
i
0.5 1 1.5
H2O2/HNO3
New York DWM-5 meteorology: H2O2/HNO3
x ROG controls
NOx controls
0.5
H2O2/HNO3
1.5
Atlanta, increased biogenics: H2O2/HNO3
11
-------
* ROG controls
0 NOx controls
0
H2O2/HN03
Los Angeles, Wagner base case: H2O2/HNO3
.a
a.
c
JO
^
o
3
TJ
O
CO
O
0.5 1 1.5
H2O2/HNO3
Los Angeles, Wagner with doubled ROG: H2O2/HNO3
112
-------
.0
a
a.
g
•o
v
CO
O
0
0.5
1.5
H2O2/HNO3
Los Angeles, Wagner with tripled ROG: H2O2/HNO3
^ 60 T
-Q
Q.
C
O
~
o
3
T3
O
CO
O
ROG controls
0 NOx controls
0
0.2
0.4 0.6
H2O2/NO2
0.8
Lake Michigan base case H2O2/NOz
13
-------
^ 60
X ROG controls
9 NOx controls
0.4 0.6
H2O2/NOZ
Lake Michigan, doubled ROG emissions: H2O2/NOz
0
0.2
0.4 0.6
H2O2/NOZ
Northeast corridor base case, H2O2/NOz
114
-------
.a
Q.
c
o
TJ
0)
CO
O
-60
9^-0-
i
i
0
0.2
0.4
0.6
H2O2/NOZ
New York MM7: H2O2/NOz
ROG controls
NOx controls
UTo
J2
Q.
0.
'C
_o
4—
O
3
0)
CO
O
ROG controls
c NOx controls
0 0.2 0.4 0.6
H2O2/NOZ
0.8
Atlanta, BEIS1 biogenics - H2O2/NOz
15
-------
^ 60 -;x:
x<'>x:::xx.xx:x_
eo
x ROG controls
NOx controls
0.4 0.6
H2O2/NOZ
Los Angeles (Godowitch): H2O2/NOz
0.8
1
c
o
*-<
o
: 3
•o
O
CO
O
v** x xx xx <-:x ox x(
0.4 0.6
H2O2/NOZ
Northeast corridor, light winds and high deposition, H2O2/NOz
116
-------
o
0.2
0.4 0.6
H2O2/NOy
0.8
Lake Michigan base case H2O2/NOy
^ 60
x ROG controls
® NOx controls
0.2
0.4 0.6
H2O2/NOy
0.8
Lake Michigan, doubled ROG: H2O2/NOy
117
-------
0
0.2
0.4 0.6
H2O2/NOy
0.8
Northeast corridor base case H2O2/NOy
o
3
•o
0)
w
O
0
0.2
0.4
X ROG controls
® NOx controls
0.6
u.o
H2O2/NOy
New York MM7 simulation: H2O2/NOy
118
-------
^ 60
X ROG controls
0 NOx controls
0 0.2 0.4 0.6
H2O2/NOy
Atlanta, BEIS1 biogenics: H2O2/NOy
0.8
ROG controls
e NOx controls
0.2
0.4 0.6
H2O2/NOy
0.8
Lx)S Angeles (Godowitch): H2O2/NOy
119
-------
.0
Q.
C
O
3
•a
o
CO
O
o
* ROG controls
0 NOx controls
0.2
0.4 0.6
H2O2/NOy
0.8
Northeast corridor, light winds and high deposition, H2O2/NOy
120
------- |