United States
Environmental Protection
Agency
Atmospheric Research and
Exposure Assessment Laboratory
Research Triangle Park, NO 27711
Research and Development
EPA/600/S3-91/050 Dec. 1991
EPA Project Summary
Development of Ozone
Reactivity Scales for Volatile
Organic Compounds
William P. L. Carter
Methods for developing a numerical
scale ranking reactivities of volatile or-
ganic compounds (VOCs) towards ozone
formation were investigated. Effects of
small VOC additions on ozone formation
(incremental reactivities)were calculated
for 140 types of VOCs in model sce-
narios representing a variety of single-
day pollution episodes. Relative
reactivities determined from effects of
the VOCs on maximum ozone concen-
trations (ozone yields) varied widely
among the scenarios, but relative
reactivities determined from effects on
integrated ozone levels were less vari-
able. A "maximum reactivity" scale was
derived from ozone yield reactivities in
scenarios where NOx inputs were ad-
justed so the VOCs had the greatest
effect on ozone, and a "maximum ozone"
scale was derived from scenarios where
NOS inputs gave maximum ozone con-
centrations. These scales gave different
relative reactivities for many VOCs, par-
ticularly aromatics. Several "multi-sce-
nario" scales were derived from the
ozone yield and integrated ozone
reactivities in the unadjusted scenarios.
The maximum ozone scale was more
consistent with averages of ratios of
ozone yield reactivities. The maximum
reactivity scale was more consistent with
ratios of integrated ozone reactivities
and also corresponded best with multi-
scenario scales developed to minimize
the total error in ozone predictions. In-
formation concerning effects of NO lev-
els, of the composition of base case VoC
emissions, and of other scenario condi-
tions on reactivities was also obtained.
Although case-by-case analysis is the
best method to determine effects of VOC
emission changes on ozone formation,
the maximum reactivity scale may be
appropriate when a single VOC ranking
must be used.
This Project Summary was developed
by EPA's Atmospheric Research and
Exposure Assessment Laboratory, Re-
search Triangle Park, NC, to announce
key findings of the research project that
Is fully documented In a separate report
of the same title (see Project Report
ordering information at back).
Introduction
The formation of photochemical ozone
continues to be a complex problem in many
urban areas. Ozone is formed by the in-
teractions of volatile organic compounds
(VOCs) with oxides of nitrogen (NOX), and
control of both is necessary to solve the
ozone problem. One of the ozone control
strategies being considered is to change
the chemical nature of VOC emissions so
that they have less of a tendency to promote
ozone formation, i.e., so that they are less
"reactive." Two example strategies are the
conversion of motor vehicles to alternative
fuels and the substitution of currently used
solvents with less reactive compounds. The
development and assessment of such strat-
egies require a means to quantify and com-
pare the ozone formation reactivities of the
many types of VOCs that can be emitted.
There are,a number of ways to quantify
the reactivities of a VOC. The most direct is
to measure the change in ozone levels
caused by changing the emissions of the
VOC in an actual air pollution episode. This
can be expressed as the "incremental reac-
tivity" of the VOC, which is defined as the
change on ozone caused by adding an
arbitrarily small amount of the VOC to the
Printed on Recycled Paper
-------
emissions, divided by the amount of VOC
added. This takes into accountthe effects of
all aspects of the organic's reaction mecha-
nism on ozone formation. The incremental
reactivities of VOCs in airpollution episodes
can be estimated by using computer airshed
models, given a representation of the con-
ditions of the episode and of the atmo-
spheric chemical reactions of the VOC.
Previous modeling studies have shown
that incremental reactivities of VOCs de-
pond on environmental conditions, particu-
larly on the ratio of total emissions of reac-
tiveorganicgases (ROG)to NOx.This means
that no single scale can predict incremental
reactivities under all conditions. This com-
plicates the development and use of VOC
reactivity scales for regulatory and control
strategy assessment purposes. The most
scientifically justifiable option is not to use a
single reactivity scale but to examine the
effects of all proposed VOC changes on a
case-by-case basis. But this is not practical
for screening purposes or when a large
number of alternative options need to be
considered. Forsuch cases, regulators have
the options of either ignoring reactivity alto-
gether or of using a generalized scale.
Though not applicable to all conditions, the
scale at least provides a practical method
for using reactivity considerations in devel-
oping ozone control strategy.
If use of an appropriate reactivity scale
has the practical effect of achieving better
ozonecontrol strategies than ignoring reac-
tivity, then the second option is to be pre-
ferred. In this case, the problem becomes
developing an optimum scale for such ap-
plications. In this work, we investigate and
compare alternative approaches for deriv-
ing such generalized VOC reactivity scales.
Methods
Representative Pollution
Episodes
A set of 62 idealized single-cell, single-
day ozone pollution scenarios, represent-
ing 12 different urban areas throughout the
United States, was taken as a representa-
tive distribution of ozone pollution episodes.
The "all-city average" mixture of ROG spe-
cies, taken from ground-level air measure-
ments of VOC species in a variety of urban
areas in the United States, was used to
represent organic emissions from all
sources in these scenarios. (This mixture of
VOCs is called the "base ROG mixture" in
the subsequent discussion, and its incre-
mental reactivity is used as the standard
against which the reactivities of the indi-
vidual VOCs are compared.) These 62 sce-
narios may not accurately represent the
conditions of any specific episodes, but as
a set they represent a wide distribution of
theairshed conditions that might affect VOC
reactivity. Therefore, they should be suit-
able for assessing methods to develop re-
activity scales encompassing a wide range
of conditions.
A representative subset of 21 of these 62
scenarios, designated the "base case" sce-
narios, were used to derive several different
"base case, multi-scenario" reactivity scales.
Because of the wide variation of ROG/NOX
ratios in these scenarios, VOC reactivities,
both absolute and relative, varied widely
among these scenarios. To derive reactivity
scales for more consistent sets of chemical
conditions, two sets of modified scenarios
were developed where NOX inputs were
adjusted to yield standard conditions of NO
availability. For the "maximum reactivity
scenarios, the NOX inputs for 29 scenarios
were adjusted such that the base ROG
mixture had the maximum reactivity. Forthe
"maximum ozone" scenarios, the NO^ in-
puts for 21 scenarios were adjusted to yield
the maximum peak ozone concentration.
The NOX inputs yielding maximum reactivity
conditions are always higher than those
yielding maximum ozone. For example, for
a Los Angeles scenario, the maximum reac-
tivity ROG/NOx ratio was 6, and the maxi-
mum ozone ratio was 8. However, these
ratios varied from scenario to scenario.
In addition, two "averaged conditions"
scenarios were developed for studying the
effects on reactivity of changing the base
ROG composition, one for maximum reac-
tivity and one for maximum ozone ROG/
NOK conditions. The scenario conditions
other than NOX inputs were derived roughly
from the average of those for the set of city-
specific scenarios discussed above.
Chemical Mechanism for VOCs
The chemical mechanism used in the
reactivity calculations was recently devel-
oped and is documented in a previously
published journal article. It includes sepa-
rate representations of over 140 types of
VOCs. The representation of approximately
20 of these was tested against results of
environmental chamber experiments. The
representations of the mechanisms for the
others were estimated or derived by anal-
ogy with mechanisms for the species with
better studied mechanisms and by using
measured or estimated rate constants. The
tabulations of the reactivity results for these
species in this work include codes giving the
level of uncertainty of the mechanism used
for each type of VOC and the extent to which
the mechanism for each VOC was tested
against environmental chamber data. Al-
though the atmospheric reaction mecha-
nisms of many, if not most, of these VOCs
are highly uncertain, this mechanism incor-
porates our best present estimates for a
wide variety of VOCs that are emitted into
the atmosphere.
Quantification of Reactivity in
an Individual Scenario
AS indicated above, the incremental re-
activity of a VOC is the change in ozone
formation caused by adding a small amount
of the VOC to the emission in a scenario,
divided by the amount added. Reactivities
are calculated for small amounts of added
VOC to remove the dependence of reactiv-
ity on the amount added and also so that the
incremental reactivities of mixtures can be
calculated by linear summations of those of
theircomponents. The "ozone formed"quan-
tity can be either the maximum ozone con-
centration calculated in the scenario or the
ozone concentration integrated overtime.
In the first case, the incrementa|.reactivities
are referred to as "ozone yield" reactivities,
while in the latter case, they are referred to
as "integrated ozone" reactivities. Both types
of reactivity were calculated for the base
case and the averaged conditions scenarios,
while only ozone yield reactivities were cal-
culated for the maximum reactivity and the
maximum ozone scenarios.
The ozone yield reactivities of VOCs can
be broken down into two components: the
"kinetic reactivity," which is the fraction of
the emitted VOC that reacts in the scenario,
and the "mechanistic reactivity," which is
the amount of ozone formed per molecule of
reacting VOC. Separate estimates of these
two components were made for the maxi-
mum reactivity scenarios and the maximum
ozone scenarios.
Reactivity Scales Derived
Sixdifferentgeneralizedormulti-scenario
reactivity scales were developed from the
incremental reactivities calculated for the
various types of scenarios. The maximum
reactivity (MaxRct) and the maximum ozone
(MaxO3) scales were developed by averag-
ing the kinetic and mechanistic reactivities
in the maximum reactivity or the maximum
ozone scenarios. These can be thought of
as generalized scales for fixed chemical
conditions with respect to NOX but with aver-
aged conditions With respect to the other
scenario inputs. Examples of incremental
reactivities and reactivities relative to the
base ROG mixture are shown for selected
VOCs in Table 1.
The incremental reactivities in the maxi-
mum ozone scale are always lower than
those in the maximum reactivity scale. This
is because the efficiency of ozone formation
from the reacting VOCs (i.e., the mechanis-
tic reactivities) is higher under maximum
-------
Table 1. Summary of Relative Reactivities for Selected VOC Species in Various Reactivity
Scales
Relative Reactivity >
Compound
Carbon Monoxide
Methane
Ethane
n-Butane
n-Octane
Isooctane
n-Pentadecane
Ethene
Propane
trans-2-Butene
1-hexene
Benzene
Toluene
m-Xylene
Naphthalene
Methanol
Ethanol
Formaldehyde
Acetone
Benzaldehyde
Adj. Nox Scales
(Ozone Yield)
MaxRct MaxO3
0.034
0.0052
0.065
0.27
0.17
0.29
0.078
2.2
2.8
3.0
1.25
0.108
0.71
2.3
0.32
0.37
0.53
5.6
0.22
-0.24
0.070
0.0090
0.108
0.45
0.29
0.42
0.128
2.8
3.0
2.9
1.20
0.084
0.38
1.8
0.039
0.53
0.67
4.5
0.20
-1.06
Base Case, Multi-Scenario Scales
(Ozone Yield) (Int'd Ozone)
Avg.R LSFit Avg.R LSFit
0.076
0.0094
0.116
0.45
0.18
0.35
0.0077
3.0
3.3
3.1
1.10
0.080
0.21
1.7
-0.129
0.53
0.67
4.9
0.19
-2.0
0.043
0.0060
0.078
0.32
0.18
0.30
0.074
2.4
2.9
3.1
1.26 .
0.098
0.58
2.2
0.21
0.40
0.57
5.2
0.21
-0.64
0.050
0.0067
0.072
0.29
0.120
0.24
-0.022
2.4
2.9
3.7
0.92
0.086
0.45
2.1
0.122
0.38
0.44
7.3
0.17
-1.27
0.036
0.0048
0.058
0.24
0.140
0.24
0.043
2.2
2.7
3.3
1.13
0.090
0.59
2.3
0.25
0.35
0.44
6.7
0.19
-0.54
•Incremental reactivities of the VOCs relative to the incremental reactivity of the base ROG
mixture. Reactivity given on a carbon-mole basis. Abbreviations used:
Ozone Yield - Scale derived from effects of VOCs on ozone yields
Int'd Ozone - Scale derived from effects on integrated ozone
MaxRct - Maximum reactivity scale
MaxCX - Maximum ozone reactivity scale
Avg. R - Scale derived by the average ratio method
LS Fit - Scale derived by the least-squares fit method
reactivity conditions. More significantly in
terms of VOC control strategies, for many
types of VOCs the relative reactivities are
also different in the two scales. For example,
aromatic hydrocarbons have relatively high
MaxRct reactivities but have relatively low
MaxO3 reactivities. This is because they
have relatively large NOX sinks in their mecha-
nism, which significantly reduces their
reactivities in scenarios where ozone is more
NOX limited. On the other hand, slowly react-
ing compounds such as CO and methane
have higher relative reactivities in the maxi-
mum ozone scale. This is because OH radi-
cal levels tend to be higher under the lower
NO,, maximum ozone scenarios than they
are In the higher NOX maximum reactivity
scenarios. The higher radical levels mean
that larger fractions of the slower reacting
compounds will react. Aldehydes and alk-
enes have high relative reactivities in both
scales.
Four "multi-scenario" relative reactivity
scales were derived from the reactivities in
the unadjusted or "base case" scenarios.
These can be thought of as representing
scales that might be appropriate if these
base case scenarios represented a realis-
tic distribution of airshed conditions. Two
different methods were used to derive the
scales from the distribution of reactivities in
the individual scenarios, and separate
scales were developed for ozone yield
reactivities and for integrated ozone
reactivities. In the "average ratio" method
the scale was derived by averaging the
ratios of incremental reactivities of the VOCs
to the incremental reactivity of the base
ROG mixture for all the scenarios. Each
scenario was weighed equally, including
the highly NO-limited scenarios, where
VOC changes had only a small effect on
reactivity. In the "least-squares error"
method, the relative reactivities were de-
rived to minimize the least-squares error in
absolute ozone predictions throughout the
entire set of scenarios that would resuftfrom
using the single scale. This gives greater
weight to those scenarios where VOC
changes have greater effects on ozone.
Examples of the four base case relative
reactivity scales are also shown in Table 1.
For most VOCs, the relative reactivities in
the multi-scenario scale derived from ozone
yield reactivities and the average ratio
method corresponded roughly to those in
the maximum ozone scale. However, this
scale had a high degree of uncertainty be-
cause the reactivity ratios being averaged
had a high degree of scatter. This is be-
cause ozone yield reactivities depend sig-
nificantly on the ROG/NOX ratio, which var-
ied widely among the base case scenarios.
The multi-scenarioscalederivedfrom ozone
yield reactivities by the least-squares fit
method had less scatter because there was
less variability in reactivity ratios in the sce-
narios that had the largest effect on the
scale. This scale corresponded reasonably
well to the maximum reactivity scale. This is
because the least-squares fit derivation
method gives greater weight to scenarios
that are closer to maximum reactivity condi-
tions.
Integrated ozone relative reactivities
tended to be less variable among the base
case scenarios, and the multi-scenario
scales derived using them tended to be less
dependent on the derivation method. They
also corresponded better to the maximum
reactivity scale than the maximum ozone
scale. This is because integrated ozone
concentrations are always affected by the
ozone formation rate, which is also the
major factor affecting the ozone yield reac-
tivity under maximum reactivity conditions.
The ozone formation rate isaless important
factor in determining ozone yield reactivities
under maximum reactivity conditions. Un-
der maximum ozone conditions, the ulti-
mate ozone formation potential is the most
important factor. -.
Dependence of Reactivities on
Scenario Conditions
The level of uncertainty of any reactivity
scale and its potential use as an ozone
control assessment tool are influenced to a
large extent by the sensitivities of reactivities
to scenario conditions. The dependence on
NOX is known to be a very important factor.
To investigate this further, reactivities rela-
tive to the base ROG surrogate were calcu-
lated as a function of the NOX inputs for a
series of representative VOCs forthe "aver-
aged conditions" scenario. The results indi-
cated that ozone yield reactivities varied
continually as the NOX levels changed and
-------
that there is no single characteristic relative
reactivity of a VOC for NOx-limited condi-
tions. They also indicated that integrated
ozone relative reactivities were much less
dependent on NOX levels than ozone yield
relative reactivities. Integrated ozone and
ozoneyield relative reactivities tended to be
similar at maximum reactivity NOX levels
and to then diverge from each other as NOX
was reduced. This is consistent with the fact
that the multi-scenario, integrated ozone
reactivity scales tended to agree better with
the maximum ozone scale than with the
maximum reactivity scale.
To investigate the effects of non-NO^-
related scenario conditions on reactivity, it
is necessary to factor out the effects of
these conditions on the effective availability
of NOX. A given NOX level or ROG/Npx ratio
may correspond to maximum reactivity con-
ditionsforonescenarioandmaximumozone
or even NOx-limited conditions for another,
depending on how the scenario conditions
affect ozone formation and NOX removal
rates. Therefore, holding NOX inputs con-
stant while varying other scenario condi-
tions does not necessarily mean that condi-
tions of NO availability are also being held
constant. However, if the scenarios being
compared have NOX inputs adjusted to yield
either maximum reactivity or maximum
ozone conditions, then they have consis-
tent conditions of Nonavailability. Thus, any
differences in calculated reactivities for such
scenarios would indicate the effects of other
non-NO^-related scenario conditions on re-
activity.
The scenarios employed in this study had
considerable variation in amounts of dilu-
tion, total VOC inputs, integrated light inten-
sity, amounts of pollutants entrained from
aloft, and relative amounts of pollutants
present initially or emitted during the day.
Although asystematicstudy of the effects of
these factors on reactivity was not carried
out, the variations in reactivities among the
NOX adjusted (maximum reactivity or maxi-
mum ozone) scenarios give an indication of
the sensitivities of reactivities to these fac-
tors. For most VOCs, this variability caused
variations in kinetic and mechanistic
reactivities yielding standard deviations of
averages on the order of ±30%. VOCs with
tow mechanistic reactivity had much higher
relative variability in mechanistic reactivity.
These include high-molecular-weight al-
kanes under all conditions and aromatics
under maximum ozone conditions. Except
for aromatics, the extent of variability of
reactivities tended to be similar for maxi-
mum reactivity and maximum ozone condi-
tions.
These results do not indicate the effects
of uncertainties or variabilities in the base
ROG mixture on reactivity, because the
same mixture was used for all scenarios.
Separate calculations were carried out to
investigate the effects of uncertainties in the
base ROG composition on VOC reactivities.
The aldehyde fraction was varied from 0 to
15% of the mixture, and the amounts of
alkenes and aromatics, relative to alkanes,
were varied by factors of 2 relative to the
standard mixture. The largest effect was
found to be the effect of reducing the alde-
hyde fraction on aldehyde reactivity: de-
creasing it from 5 to 0% increased the
formaldehyde reactivity by almost a factor
of 2. Other than this, it is hard to make any
generalizations concerning the effects of
changing the ROD mixture on VOC reactiv-
ity—the qualitative effects seemed to vary
considerably from case to case. Except for
the large aldehyde-on-aldehyde effect, the
effects of these rather large variations in the
base ROG composition onthe NOx-adjusted
incremental reactivities of these represen-
tative VOCs can be considered to be rela-
tively small. Factorof 2 changes in alkene or
aromatic fractions caused less than 25%
changes in reactivities in most cases.
Conclusions
Developing VOC reactivity scales for
ozone control strategies results in three
types of difficulties: (1) gaps in our knowl-
edge of the gas-phase reaction mecha-
nisms cause major uncertainties in calcu-
lated reactivities for many VOCs, (2) uncer-
tainties in airshed conditions also cause
uncertainties in reactivity calculations, and
(3) the dependence of VOC reactivities on
airshed conditions means that there would
be uncertainties in any reactivity scale that
was developed even if the reactivities of the
VOCs in individual airsheds were known.
The focus of this work has primarily been on
the third problem.
The difference in relative levels of NO is
the most important reason that VOC
reactivities vary from scenario to scenario.
NOX conditions significantly affect ratios of
reactivities of VOCs in different chemical
classes. It can be argued that the maximum
reactivity and the maximum ozone reactiv-
ity scales developed in this work represent
respectively the high and low limits for con-
ditions of Nonavailability that are appropri-
ate for defining a VOC reactivity scale.
Significant concentrations of ozone are not
formed if NOX levels are much higher than
those giving maximum reactivity. If NOX
levels are lower than those yielding maxi-
mum ozone, then ozone formation is NOX
limited, and NOX control is much more effec-
tive than VOC control in reducing ozone.
Therefore, comparison of MaxRct and the
MaxO. scales, gives an appropriate indica-
tion of the effect on reactivity scales of
uncertainties in NOx conditions that are rel-
evant to ozone control strategies.
Since these two reactivity scales can in
some cases give significantly different pre-
dictions of benefits of proposed VOC sub-
stitutions, it can be questioned whether it is
appropriate ever to use reactivity scales in
control strategy applications. Obviously, if
major substitution strategies are being con-
sidered, they should be evaluated as com-
prehensively as possible, and reliance on a
single reactivity scale alone would not be
justifiable. However, when this is not prac-
tical, it is reasonable to expect that the use
of an appropriate reactivity scale would
more likely lead to better ozone control
strategies than if reactivity were ignored
altogether. The problem then is to deter-
mine what is the best type of scale for this
purpose. While not conclusive (because of
the uncertainties in the set of scenarios
employed), the results of this study suggest
a maximum reactivity scale — such as the
MaxRct scale — may give a good approxi-
mation to such an "optimum" scale. It gives
reasonably good predictions of relative
reactivitiesfor airsheds where VOCchanges
have the largest effect on ozone, and it
gives reasonably good predictions of rela-
tive integrated ozone reactivities for a wide
variety of airsheds. Although these conclu-
sions are based on reactivities calculated
for highly simplified, single;day, and per-
haps in some cases inaccurate scenarios,
they are sufficiently varied that it is not
unreasonable to expect that similar results
would be obtained if more detailed and
accurate scenarios were employed.
However, it is clear that further research
is needed to reduce the uncertainties in the
derivation of VOC reactivity scales. The
uncertainties in airshed conditions obviously
need to be reduced. More work is needed
on the effects of scenario conditions on
reactivity. The chemical mechanisms for
many VOCs are highly uncertain, and ex-
perimental data are needed to reduce these
uncertainties, or at least to test their predic-
tions of maximum reactivity.
•fru.S. GOVERNMENT PRINTING OFFICE: 1991 - 648-080/40106
-------
-------
-------
-------
William P. L Carter Is with the University of California, Riverside, CA 92521.
Joseph J. Bufallnl is the EPA Project Officer (see below).
The complete report, entitled "Development of Ozone Reactivity Scales for Volatile
Organic Compounds" (Order No. PB91-243 386/AS; Cost: $26.00, subjectto
change) will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Atmospheric Research and Exposure Assessment Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
United States
Environmental Protection
Agency
Center for Environmental
Research Information
Cincinnati, OH 45268
BULK RATE
POSTAGE & FEES PAID
EPA
PERMIT No. G-35
Official Business
Penalty for Private Use $300
EPA/600/S3-91/050
------- |