United States
Environmental Protection
Agency
Atmospheric Research and
Exposure Assessment Laboratory
Research Triangle Park, NC 27711
Research and Development
EPA/600/S3-90/001 August 1990
&EPA Project Summary
Evaluation of the RADM
Gas-Phase Chemical
Mechanism
William P. L. Carter and F.W. Lurmann
A program was conducted to
evaluate the gas-phase mechanism
developed for the Regional Acid
Deposition Model (RADM). An initial
review of the mechanism was
carried out, resulting in several
modifications being made prior to its
further evaluation. New methods
were developed for processing
emissions input for the RADM model
and for representing aggregated
VOCs in the model. The mechanism
was tested by comparing predictions
against results of over 550
environmental chamber experiments
carried out at two laboratories. A
series of 90 test problems for
assessing gas-phase mechanisms in
regional models was developed and
used to test several condensation
approaches for the RADM
mechanism. As a result of this study,
two modified RADM mechanisms
were developed and evaluated. The
most important of the modifications
concerned the representation of
aromatic hydrocarbons. Based on
the results of the evaluation, it was
concluded that the modified RADM
mechanism represents the current
state of the art and is suitable for use
in a regional acid deposition model.
However, there are still major
uncertainties in our understanding of
atmospheric chemistry and available
chamber data are not sufficient for
evaluating many important aspects of
regional oxidant mechanisms.
This Project Summary was
developed by EPA's Atmospheric
Research and Exposure Assessment
Laboratory, Research 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 Regional Acid Deposition Model
(RADM) plays a central role in the
National Acid Precipitation Assessment
Program's (NAPAP) plan to develop an
understanding of source-receptor
relationships for acidifying pollutants.
This model is needed to formulate
appropriate and cost-effective emission
control policies to mitigate the potentially
damaging effects of acid deposition.
While transport, deposition, and
chemistry all play important roles in the
acid deposition phenomenon, the
chemistry plays a particularly important
role in determining the response of the
acid deposition system to major changes
in NOX and SOX emissions. The
chemistry submodel is the only
component of the overall modeling
system that is capable of predicting
nonlinearity in the relationship between
source emissions and acid deposition.
Thus, it is critically important that any
chemical mechanism incorporated into
the RADM model be fully evaluated using
the best available data and that it be
recognized by the scientific community
as representing the current state-of-the-
art in our understanding of atmospheric
chemistry.
Because of the importance of
chemistry in the Regional Acid
Deposition Model, the EPA contracted
the Statewide Air Pollution Research
Center (SAPRC) at the University of
California and Lurmann Associates to
conduct a comprehensive evaluation of
the gas-phase chemical mechanism
proposed for use in RADM-II. As part of
this effort, an initial review of the
mechanism for consistency with current
-------
kinetic and mechanistic data was carried
out; new methods were developed for
processing VOC emissions input and for
representing aggregated VOCs in the
model; the mechanism was tested
against the results of over 550
environmental chamber experiments; a
series of test problems for assessing gas-
phase mechanisms in regional models
was developed and used to test several
condensation approaches for the
mechanism; and two modified versions of
the mechanisms were developed and
evaluated. SAPRC and Lurmann and
Associates were assisted in this effort by
Dr. William R. Stockwell of the State
University of New York, the developer of
the RADM-II chemical mechanism, Dr.
Paulette Middleton of the National Center
of Atmospheric Research, who is
responsible for preparation of emissions
data used by the RADM mechanism, Dr.
Harvey E. Jeffries at the University of
North Carolina (UNC), who is responsible
for the UNC chamber experiments whose
results were used as part of the
evaluation effort, and by Dr. Roger
Atkinson at SAPRC, who assisted in the
initial review of the mechanism for
consistency with current data.
The research findings from the various
efforts in this program, and
recommendations for future research, are
summarized below.
Evaluation of Kinetic and
Mechanistic Parameters
At the beginning of this study, an initial
review of a preliminary version of the
RADM-II chemical mechanism was
carried out, and a number of changes
were recommended. A procedure was
developed for determining rate constants
and products for the reactions of lumped
organic species based on the detailed
speciation of the national emissions
inventory. Many of these initial
recommendations were adopted and
incorporated into the preliminary version
of the RADM-II mechanism prior to its
evaluation against the environmental
chamber data.
Based on the results of testing the
RADM mechanism against chamber data
and sensitivity testing under a range of
atmospheric conditions, two alternate
versions of the RADM mechanism were
developed and evaluated. The first
alternate mechanism, referred to as the
recommended modified RADM
mechanism (RADM-M), had the following
changes relative to the preliminary
version of the mechanism: (1) the
representation of the unknown aspects of
the aromatic photooxidation mechanism
was changed to improve its performance
in simulating chamber data; (2) a more
condensed representation of the >C3
alkanes was incorporated; (3) several
rate constant and product yields for
lumped species were updated based on
a new analysis of the emissions data; (4)
an error in the methylglyoxal quantum
yields that resulted in overestimation of
its photolysis rate was corrected; (5)
several reactions omitted from the
NO3+ alkene reaction system were
added; and (6) the representation of the
reactions of nitrophenols was revised to
improve its chemical fidelity.
Because implementation of the RADM
mechanism was almost completed by the
time the RADM-M mechanism was
developed and evaluated, a second
alternate mechanism was proposed that
involved primarily changes in parameter
values. This version is referred to as the
recommended modified parameter
version (RADM-P). This incorporated an
improved aromatic oxidation mechanism,
which performs approximately as well as
that in RADM-M, but uses the larger
number of species employed in the
RADM mechanism. It also incorporated
the updated lumped species rate
constant and product yields, the reduced
methylglyoxalquantum yields, and the
addition of the omitted NO3+ alkene
reactions that were incorporated in
RADM-M. The RADM developers have
adopted the RADM-P mechanism and
this version is now being implemented in
the full RADM-II model.
Processing of the VOC
Emissions Inventory for the
RADM Mechanism
At the start of this program, only a very
preliminary emission inventory interface
had been developed for the RADM
mechanism; hence,- a large effort- was-
directed towards developing emissions
aggregation procedures for use with the
RADM model. A new two-step system
was developed for this purpose. The first
step involved aggregating the detailed
VOC emissions data into a 32-class VOC
grouping based on reactivity and relative
contributions to total emissions. The
second step involved further aggregating
these emissions groups into the more
limited number of VOC species used in
the RADM mechanism. This effort was
carried out jointly as part of this program
and the RADM development effort.
The aggregation of the emissions
groups into the RADM species involves
use of a new reactivity weighting scheme
that allows for VOCs reacting at different
rates, but with similar mechanisms, to be
represented by the same model species.
This is based on estimating how much of
the VOCs and the model species
undergoes chemical reaction in the
model simulation and adjusting the
amounts of the latter so they are
approximately equal. Although
approximate, the use of reactivity
weighing is preferable to the alternatives
of either ignoring compounds of low but
non-negligible reactivity, or representing
them by much more reactive model
species.
Another feature of this aggregation
system is the use of detailed emissions
data to derive the rate constants and
product yield coefficients of lumped
species in the mechanism based on
.those_fgr_ the. aggregate of compounds
they represent. A software system
recently developed as part of another
program and the 1985 NAPAP
anthropogenic emissions inventory were
used for this purpose. This is considered
a significant advancement that optimizes
the linkage between the chemical
mechanism and the emissions inventory.
Mechanism Evaluation Against
Chamber Data
The gas-phase chemistry is one of the
few modules in the atmospheric modeling
system that can be independently tested
against experimental data. Although the
present chamber data base is not
suitable for testing all aspects of the
mechanism, particularly those involving
predictions of acidic or peroxy species or
simulations of very low NOX conditions
that prevail in rural regions, it is suitable
for testing predictions of ozone formation
and rates of NOX oxidation under urban
conditions. To a lesser extent, the data
can be used to test the mechanisms'
ability to simulate formation of
peroxyacetyl-nitrate -and- formaldehyde
under urban conditions. Simulation of
urban-like conditions is important
because urban areas represent the major
sources of the pollutants that are
transported to remote areas with sensitive
receptors.
A comprehensive evaluation of the
RADM, RADM-M, and RADM-P
mechanisms against environmental
chamber data was performed in this
study. The approach involved simulating
a large number (~550) of experiments
from four different chambers and
statistically evaluating the mechanisms'
performance. Data from (1) the SAPRC
evacuable indoor chamber (EC); (2) the
SAPRC indoor Teflon chamber (ITC); (3)
the SAPRC outdoor Teflon chamber
(OTC), and (4) the UNC outdoor chamber
-------
were utilized for this purpose. Prior to
conducting the evaluation, procedures for
characterizing chamber-dependent
processes were reviewed and updated.
Most of the updated procedures
concerned the UNC outdoor chamber
where a new light characterization model
developed at UNC was utilized. Organic
compounds in experiments employing
complex mixtures of organics were
aggregated using the procedures that will
be employed when the mechanism is
used in RADM. Thus no run-to-run
adjustment of chamber-dependent or
mechanistic parameters was used in the
evaluation.
The major findings from the evaluation
of the RADM, RADM-M, and RADM-P
mechanisms against environmental
chamber data are as follows:
Testing against experiments with
mixtures of organics designed to
represent current VOC emissions
showed that the RADM mechanisms
perform as well as other chemically up-
to-date mechanisms. The mechanisms
are able to predict the maximum ozone
and NOX oxidation rate with a small bias
and a mean error of 20% on the
average.
Testing of of the RADM mechanism
against single organic-NOx-air
experimental data produced mixed
results. With a few exceptions, the
RADM mechanism performed as well
as can reasonably be expected for
experiments containing species for
which the mechanism was designed.
The predictions for experiments with
formaldehyde and propene were
reasonably good. Predictions for
acetaldehyde and methyl ethyl ketone
systematically underpredicted ozone
and NOX oxidation rates, but there are
too few experiments to draw definitive
conclusions on the adequacy of this
portion of the mechanism. The
mechanism has a bias towards
underpredicting reactivity in ethene
runs due to the omission of the
reaction of O(3P) atoms with ethene.
This reaction, however, is unimportant
under conditions where the model will
be applied. The mean error in the
ozone predictions for the 1-butene runs
is greater than is the case for propene,
but the bias is reasonably small. The
predictions for the trans-2-butene
simulations are poor, but there are too
few of these experiments to test this
portion of the mechanism adequately.
The mechanisms had large errors in
simulating the alkane runs due to the
extreme sensitivity of these runs to
chamber effects that tend to vary
unpredictably from run to run.
Compared to the errors, the biases in
the predictions of ozone and NOX
oxidation rates in the alkane runs were
relatively low. In these regards, the
performance of the RADM mechanisms
is comparable to those of other recent
mechanisms that have been tested
against these data.
The performance of the RADM
mechanism in tests against single
aromatic species was not as good as
that reported for other chemically up-
to-date mechanisms. The testing
showed that the RADM aromatics
mechanism has a bias towards
overpredicting ozone and NOX
oxidation rates for the more slowly
reacting compounds and
underpredicting those same
characteristics for the faster reacting
compounds. This results in fairly large
errors in ozone predictions on the
average for toluene, xylenes, and
mesitylene.
The RADM-M and RADM-P
mechanisms were designed to improve
the performance of the mechanism in
simulating the aromatic hydrocarbon
runs. This design objective was
achieved. The performance of these
mechanisms in simulating the results of
these runs was comparable to other
current mechanisms.
Because of the importance of biogenic
emissions in regional model
applications, the RADM mechanism
has a separate species to represent
isoprene. However, the performance of
this mechanism in simulating isoprene
experiments was not as good as those
for most other alkenes. Improving the
performance of the RADM mechanism
in representing isoprene would
probably require adding additional
species to the mechanism.
The RADM mechanisms performed
suprisingly well in simulating the a-
pinene runs, despite the fact that the
mechanisms do not represent this
compound explicitly. However, there
are only a limited number of such runs,
and the apparent good performance
may be fortuitous.
The RADM mechanisms' performance
for species other than those for which it
was designed, such as propioaldehyde,
1-hexene, benzene, naphthalenes, and
tetralin, was generally poor. However,
the level of performance of the RADM
mechanism in this regard is certainly
comparable to what one would expect
for other condensed mechanisms
designed for use in Eulerian
atmospheric models. Satisfactory
performance in simulating these
compounds would probably require
adding new species to the mechanism.
This is probably not worthwhile given
the relatively small contributions of
these compounds to overall VOC
reactivity in most regional model
applications.
The RADM mechanisms do not predict
maximum formaldehyde and PAN
concentrations as accurately as they
predict maximum ozone concentrations
or NO oxidation rates. This is also the
case with other current mechanisms
that have been evaluated against these
data. In the case of formaldehyde,
some of the poor performance may be
attributed to uncertainties in the quality
of the experimental data. In the case of
PAN a positive bias is expected since
the PAN species in the mechanism is
used to represent a number of related
compounds in addition to PAN.
However, in some cases the
performance of the mechanism in
simulating these species represents an
area of concern.
It is very unlikely that mechanisms
could achieve the high level of
performance on ozone and NOX
oxidation rates found in complex
mixture runs without accurately
simulating the production and
destruction of OH radicals. These
results indirectly suggest that the
mechanisms are able to reasonably
predict OH concentrations and,
therefore, the probable yields of nitric
acid and sulfuric acid in daylight hours.
Sensitivity Testing of Alternate
Mechanistic Assumption
Given the nonlinear couplings in
mechanisms, it is often difficult to predict
a priori the conditions and species for
which changes in mechanistic
assumptions will have large effects.
Therefore, it is important in evaluating
alternative mechanistic assumptions to
make sure the testing is performed over
a wide range of conditions. To address
this, a large number of test problems (90)
were developed to represent a wide
range of chemical conditions for
purposes of evaluating selected
mechanistic assumptions.
The predictions of the RADM
mechanism were compared to those of a
detailed version that included all possible
peroxy-peroxy radical combination
reactions. The results showed that even
for hydrogen peroxide and organic
peroxides, which are very sensitive to
peroxy-peroxy radical reactions, the
RADM mechanism gave predictions that
&U.S. GOVERNMENT PRINTING Off ICE-. 1990/7'A8-012/20Q6Q
-------
were almost identical to those of the
detailed version in all cases. Tests were
also carried out to determine if a more
condensed representation of peroxy-
peroxy radical reactions could be used in
the RADM model. The more condensed
representation did not yield predictions
that agreed as closely with those of the
detailed mechanism, though the
differences were probably not significant.
The results suggest that the treatment of
peroxy radicals in the RADM mechanism
is satisfactory.
Tests were carried out to determine if a
more condensed representation of the
reactions of alkanes could be employed
in the RADM model. The results
indicated there are no significant
differences in model predictions if 4he
two higher alkane classes in RADM are
combined, provided the combined class
has the appropriate mechanistic
parameters. These results served as the
basis for the representation of alkanes in
the recommended RADM-M mechanism.
Test calculations indicated that the
versions of the RADM mechanism
incorporating our recommended
modifications (RADM-M and RADM-P)
can in some cases be quite different from
the original mechanism in predicting
species of interest in acid deposition
models. The differences between the
mechanisms tend to be the greatest in
the simulations of urban conditions. This
is due primarily to the recommended
changes in the mechanisms for the
aromatics. The modified mechanisms
tended to predict slightly lower levels of
ozone and in some cases significantly
lower levels of peroxide or acid species
under some conditions. On the other
hand, the modifications had smaller
effects on predictions of nitric acid or
sulfate. This indicates that implementing
the recommended changes in the
mechanism might in some cases have
non-negligible effects on results of
regional model simulations.
Conclusions ._ __
and Recommendations
The RADM gas-phase chemical
mechanism and its associated emissions
aggregation system, revised as a result
of this program, represents the current
state-of-the-science. Therefore, it is
suitable for use in the current version of
the Regional Acid Deposition Model.
However, there remain significant
uncertainties in our understanding of
atmospheric chemistry and further
research is needed to reduce these
uncertainties. The research needed to
William P. L. Carter is with the Statewide Air Pollution Research Center.University
of California, Riverside.CA 92521. Frederick W. Lurmann, formerly with
Lurmann and Associates. Santa Barbara, CA 93103, is now with Sonoma
Technology, Inc., Santa Rosa, CA 95403.
Marcia C. Dodge is the EPA Project Officer (see below).
The complete report, entitled "Evaluation of the RADM Gas-Phase Chemical
Mechanism." (Order No. PB90-164526/AS;Cost: $45.00, (subject to
change) will be available only from:
National Technical Information Service
5285 Port Royal Road Springfield, V'A 221'61
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
address .the major uncertainties in the
representation of gas-phase chemistry in
regional acid deposition models is as
follows:
Development of methods for more
accurate representation of biogenic
VOC emissions in condensed gas-
phased mechanisms.
Additional environmental chamber
studies to acquire data for testing
mechanisms for the terpenes.
Improvement of VOC emissions
inventories, especially for biogenic
organics.
Development of new environmental
chamber technology capable of testing
predictions of gas-phase mechanisms
under... lpw-_Njpx conditionj, and.Joj:...
testing predictions of acidic and
peroxide species. This includes
research into surface effects that affect
results of environmental chamber
experiments.
Additional mechanistic studies
concerning the gas-phase atmospheric
chemistry of aromatic hydrocarbons,
higher alkanes, isoprene, and terpenes.
Development of new instrumentation
needed for making further advances in
elucidating reaction mechanisms.
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Official Business
Penalty for Private Use $300
EPA/600/S3-90/001
------- |