United States
Environmental Protection
Agency
Atmospheric Research and Exposure
Assessment Laboratory
Research Triangle Park NC 27711
Research and Development
EPA/600/S3-89/010 Aug. 1989
Project Summary
Validation Testing of New
Mechanisms With Outdoor
Chamber Data
H. E. Jeffries, K. G. Sexton, J. R. Arnold, and J. L. Li
The UNC smog chamber data base
was used to compare the perform-
ance of two state-of-the-sctence pho-
tochemical mechanisms: the Carbon
Bond Four Mechanism and the Carter,
Atkinson, Lurmann Mechanism. A
number of tasks had to be performed
before the comparisons could be
conducted. These included com-
paring and reconciling the thermal
rate constants used in the two mech-
anisms, re-assessing the photolytic
rates for the UNC chamber and pro-
ducing a new radiation transfer
model, performing new experiments
with new instrumentation for formal-
dehyde and hydrogen peroxide to
confirm older data, and analyzing
three years of ambient hydrocarbon
data to determine a default compos-
ition for use in EKMA comparisons. A
series of chamber experiments was
modeled with both mechanisms to
assess their performance for specific
chemistry. A series of State Imple-
mentation Plan EKMA calculations
was also performed with both
mechanisms for a range of condi-
tions.
Although there were other
differences between the two photo-
chemical methods, both mechanisms
showed good agreement for ozone
and oxides of nitrogen chamber data.
The agreement for other products
such as formaldehyde and peroxy-
acetylnitrate was not as good. Neither
mechanism was considered superior
to the other either in fitting smog
chamber data or in predicting VOC
control requirements.
This Project Summary was devel-
oped 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
four separate reports of the same title
(see Project Report ordering informa-
tion at back).
Introduction
The U.S. Environmental Protection
Agency (EPA) has a long-standing pro-
gram to develop accurate photochemical
mechanisms for incorporation into air-
quality models. The EPA has sponsored
the development of two basic types of
mechanisms: the Carbon Bond approach
with its emphasis on conservation of
carbon mass, and the surrogate approach
with its emphasis on quasi-explicit
chemistry. Neither approach has a clear
theoretical advantage; both require that
choices be make between detail and
accuracy, and each may have specific
advantages in particular applications.
The work described in this summary
compared two specific mechanisms: the
Carbon Bond Four mechanism (CB4)
developed by System Application Inc.
and the Carter, Atkinson, Lurmann mech-
anism (CAL) developed by the University
of California, Riverside (UCR) and Envi-
ronmental Research and Technology
(ERT). During their development, both
mechanisms had been tested using se-
lected data from the UNC and UCR
chambers and were found to agree
reasonably well with these data. Of
concern was whether the two mecha-
nisms would predict similar VOC control
requirements when used in EPA's EKMA
procedure, and, if not, to determine why
the mechanisms disagreed.
-------
This comparison task involved six
activities: (1) comparison of rate con-
stants for the portion of the mechanisms
considered to be "well known", and a
reconciliation of the differences, (2) a
complete re-assessment of the photolytic
rates used to simulate the UNC outdoor
chamber, (3) conducting a two-month
outdoor chamber experimental program
comparing five methods for measuring
formaldehyde and three methods for
hydrogen peroxide - products that the
mechanisms were not predicting well, (4)
simulating a selected set of UNC
chamber experiments using identical in-
puts for both mechanisms, (5) analysis of
ambient NMOC data to determine default
hydrocarbon compositions for use in
EKMA, (6) comparison of the mechan-
isms' performance in a series of State
Implementation Plan scenarios using the
EKMA method. An overview of each of
these tasks is described below.
Rate Constant Comparison
When the inorganic and carbonyl
reaction sets of both mechanisms were
compared in the early part of the pro-
gram, significant differences were found.
The differences were reconciled with the
mechanisms' authors and the resulting
versions of the mechanisms are now
comparable in their inorganic and
carbonyl sections.
Photolytic Rate Calculation
As part of the experimental portion of
this program, a computerized, portable,
spectroradiometer was used to collect a
large body of spectral irradiance meas-
urements outside and inside the UNC
chamber. These data showed that neither
mechanism developer had correctly cal-
culated in-chamber photolytic rates, nor
had the UNC researchers supplied suf-
ficient information to estimate the rates
properly. In addition, analysis of these
spectral data showed that the method
that had been used for more than 12
years, a look-up table based upon
Peterson's calculation of actinic flux,
provided a poor approximation of the
radiation field at the UNC site because of
the effect of aerosols on the radiant flux.
In response to these findings, a new
radiation transfer model was developed.
This model uses the World Radiation
Center extraterrestrial solar flux and a
one-layer approximation of atmospheric
transmission due to Rayleigh scattering,
total column ozone, total column water
vapor, and aerosol absorption and scat-
tering to compute the surface spectral
irradiance and surface spherical spectral
irradiance. It incorporates three aerosol
models that are sensitive to relative
humidity and composition. In addition, the
model includes two broad-band models,
one for the total solar radiation (TSR)
sensor and one for the Eppley ultraviolet
(UV) radiometer.
Studies were conducted to determine
the solar spectrum optical properties of
the TFE Teflon film used to enclose the
UNC chamber. Additional studies were
conducted to determine the spectral
albedo of the reflective chamber floor.
Chamber geometry relative to the sun's
path was re-assessed and a full chamber
transmission model was developed. By
operating radiometers inside and outside
the chamber, sufficient information was
collected to develop an in-chamber
actinic flux model.
The TSR and UV radiometer data for
286 days from 1976 to 1986, were re-
assessed and modeled with the broad-
band radiation models. Cloud effects
were then introduced into the model and
correction factors for the theoretical mo-
del were derived for each day. Finally,
the actinic flux (spherical spectral ir-
radiance) at 16-minute intervals for the
286 days was calculated.
New Formaldehyde and
Hydrogen Peroxide Data
In past tests by the developers, the
mechanisms have generally failed to
predict the observed formaldehyde
(HCHO) product profiles in the UNC
chamber data. The model developers
suggested that the UNC HCHO data
could be too high by a factor of two.
Further, in a previous study conducted at
UNC, a laser system was used to monitor
hydrogen peroxide (H202) and the results
were difficult to understand given today's
knowledge of H202 chemistry. Therefore,
an intercomparison study of five HCHO
methods and three H202 methods was
undertaken. Research groups from EPA,
NCAR, Texas Tech University, Uni-
Search, Inc., and UNC set up their HCHO
methods at the UNC smog chamber
laboratory to collect HCHO data during
chamber experiments over a two-week
period. The NCAR, Texas Tech, and Uni-
Search groups also made H202 measure-
ments.
The results show that HCHO monitor-
ing methods can perform nearly identi-
cally, if the calibration sources are
reconciled. The standard UNC HCHO
method was shown to be in very close
agreement with the other methods and.
not as suggested by the modeler
subject to a positive interference of up
a factor of two.
The hydrogen peroxide data from <
three measurement methods were
good agreement in the range 0-800 pf
H202. The half-life for H202 in the UN
chamber, however, was extremely sho
ranging from 1 hour in the dark to '
minutes in the noon sun. Furthermor
the addition of 350 ppm CO, whic
should have increased the half-life t
recycling hydroxyl radicals back
hydroperoxy radicals, actually accele
ated the lost of H2O2 dramatically. Mo
confusing of all, during the accelerate
decay of the 350 ppb H2O2 there was
loss of 55 ppm CO.
Similarly, during experiments designe
to produce large quantities of H202 fro
chemical reactions of VOC and NOX, on
4-10 ppb was ever observed. Tl-
assumption is that the insitu-produce
H2O2 also decays rapidly, and therefoi
high levels are never reached in tr
chamber.
Speculation as to the mechanism i
H202 loss in the UNC chamber include
direct absorption on the walls, absorptic
into a thin water film on the walls, or
catalytic process involving the absorbe
nitrogen species, CO, and H2O2 th;
rapidly consumes H2O2 and, by a cha
process, CO.
The basic conclusion reached froi
these results is that the chamber walls i
the UNC chamber are sufficiently reactiv
towards H2O2 that it is not possible I
produce data to test mechanism patf
ways for generating H2O2. Based upc
the nitric acid data reported by th
UniSearch laser system, nitric acid also i
probably lost almost as rapidly to th
walls.
Demonstration Modeling with
New Inputs
To demonstrate the effect of differer
mechanism rate constants and reactio
products, 22 dual UNC chamber expei
iments were modeled with both meet
anisms using the same simulation cor
ditions. The 22 case days included 1
single hydrocarbon component expei
iments and 12 mixture experiments. Th
following points were noted:
1. Both mechanisms show exceller
agreement with experimental data ft
HCHO-NOX systems.
2. In ethylene experiments, the mecr
anisms show good fits to NO, NOJ
03, and ethylene decay profile!
-------
Both mechanisms, however, under-
predict the formation of HCHO and
CO.
3. Neither mechanism does well at sim-
ulating 1-butene experiments on cool
days; both mechanisms are much
too reactive in producing ozone.
CAL, however, produces an excellent
fit to t-2-butene. CB4's surrogate of
two ALD2s for t-2-butene shows too
little reactivity throughout the simula-
tion, but does begin to converge
towards the correct ozone concen-
tration near the end of the
experiment.
4. CB4 does better than CAL in
simulating toluene experiments; CAL
is better than CB4 in simulating
xylene experiments. Both mechan-
isms, however, overpredict HCHO in
these systems by nearly a factor of
two or three. Both mechanisms
predict PAN fairly well in the xylene
experiments, but both overpredict
PAN in the toluene experiments with
CAL predicting nearly twice as much
PAN as CB4. Both mechanisms
overestimate the reactivity of o-
xylene compared to m-xylene and
both underpredict CO formation in
these systems.
5. In a HC mixture with no aromatics,
both mechanisms provide excellent
fits to NO, N02 and 03, and correctly
simulate the decay of propylene,
ethylene, n-pentane and n-butane.
HCHO is underpredicted, however
(by 50% in the case of UNCMIX).
6. In the SynAuto HC mixture experi-
ments, both mechanisms provide
excellent fits to NO, N02 and 03, and
to propylene, ethylene and toluene
decays. HCHO predictions tend to be
much closer, suggesting a compen-
sation between an overproduction in
the aromatics chemistry and an
underproduction in the olefin
chemistry.
7. Both mechanisms are uniformly too
reactive in generating ozone for the
SynUrban mixture experiments, even
though the propylene, ethylene, and
toluene decays are well fitted.
Urban Hydrocarbon
Composition
The application of the two mechanisms
in EKMA control strategy calculations
requires the specification of an urban hy-
drocarbon composition. Previously, SAI
had supplied the so-called "default"
^hydrocarbon composition, but new data
rom several EPA projects to monitor
hydrocarbons in ambient air and to
speciate the hydrocarbons using gas
chromatography provided a new data
base. An analysis of this data base was
done by OAQPS and another was done
by ERT as part of the CAL mechanism
development. The carbon fractions ob-
tained in these two efforts were signif-
icantly different. A new analysis of the
data was undertaken for both the CB4
and CAL mechanisms using a single set
of assumptions and speciation guidance
from the model developers. A total of 66
city-years of detailed HC data (compris-
ing 773 individual analyses) was avail-
able. A total of 56 HC samples collected
aloft by Washington State University was
also added to the data base. These data
were speciated into the carbon fractions
for hydrocarbon species used in the two
mechanisms. Fractions for each mechan-
ism are reported for each city-year, for
five clusters of city years with similar
internal composition, and for the total.
The five clusters were determined using
two forms of cluster analysis. Multiple re-
gression was also used to confirm the
overall average carbon fractions. Four
analyses of automobile exhaust from the
UNC autoexhaust smog chamber study
were also speciated to assist modelers in
testing chamber data in the same manner
as atmospheric data. Recommendations
for default carbon fractions for both
mechanisms were developed and the
fractions have now been included in the
computer code for OZIPM4. Selected
simulations using both mechanisms were
also conducted to demonstrate the
effects of HC composition on ozone
predictions.
EKMA Control Calculation
Comparison
A number of EKMA calculations were
carried out with both mechanisms using
the OZIPM4 program. The conditions
selected were derived from cases used
for State Implementation Plan develop-
ment. Fourteen cases were simulated for
Nashville, Tulsa, Puget, Washington DC,
Phoenix, Philadelphia, and New York.
Baseline ozone values range from 0.13 to
0.234 ppm and calculated VOC reduction
estimates ranged from 13% to 57%.
Sometimes CB4 predicted the higher
control requirement and sometimes CAL
predicted the higher control requirement.
The two mechanisms gave control re-
quirements that differed at most by only
7%.
In a series of sensitivity tests, the CB4
mechanism showed more sensitivity to
external radical sources (e.g. hydrocar-
bon aloft) than did CAL. CB4 calculated
controls also exhibited a much larger
sensitivity to temperature than did CAL:
CB4 showed a 10% change in control
requirements for a 5°F change in simu-
lation temperature while CAL showed
only a 5% change for the same temper-
ature variation. This sensitivity is directly
related to the different temperature
dependencies used in the two mechan-
isms to describe the formation of PAN.
CB4 uses a much higher negative value
for the activation energy than does CAL.
The CB4 value is based upon fits needed
to simulate winter experiments in the
UNC chamber, whereas the CAL value is
based upon the limited temperature
range of the kinetic data for this reaction
that are found in the literature. The large
value used in the CB4 must be regarded
as speculation until experimental tests
confirm or refute the value. This may limit
the application of these mechanisms to
ambient temperatures above about 75
degrees F.
Conclusions
In this project, two state-of-the-science
photochemical mechanisms were com-
pared. Many differences occurring during
the development of the mechanisms
were examined and resolved where
possible. Other activities needed to ob-
tain common inputs, such as photolysis
rates, were undertaken.
The mechanisms were compared
against chamber experiments. Both
mechanisms simulate HCHO/NOX sys-
tems nearly perfectly. Both mechanisms
simulate ethylene systems reasonably
well, but require variable "wall radical"
sources to fit ozone, NOX, and ethylene
data. Neither mechanism fits HCHO pro-
duct data in these systems. For organics
more complex than ethylene, the two
mechanisms do diverge and the quality
of the fits varies. For example, CAL is
better for propylene, neither is very good
for 1-butene, but CAL is much better for
trans-2-butene. CB4 is better for toluene
and CAL is better for xylene. Neither is
very good at prediction PAN or HCHO
formation from aromatics. Both are good
for simple urban-tike mixtures, and both
are good for synthetic autoexhaust. Both,
however, are too reactive for synthetic
urban hydrocarbon mixtures. Both mech-
anisms give similar predictions of control
requirements in EKMA applications, but
CB4 is very temperature sensitive.
Finally, neither mechanism is superior
to the other in either fitting smog cham-
ber data or in predicting VOC control
requirements. Thus the choice between
them must be made on other grounds.
-------
H. E. Jeffries, K. G. Sexton, J. R. Arnold, and J. L Li are with the University of
North Carolina, Chapel Hill, NC 27514.
Mania C. Dodge is the EPA Project Officer (see below).
The complete report consists of four volumes, entitled 'Validation Testing of New
Mechanisms with Outdoor Chamber Data:" (4 volume set: Order No. PB 89-159
024/AS; Cost: $81.00)
'Volume 1. Comparison of CB4 and CAL Mechanisms," (Order No. PB 89-159
032/AS; Cost: $28.95)
'Volume 2. Analysis of VOC Data for the CB4 and CAL Photochemical Mech-
anisms," (Order No. PB 89-159 040/AS; Cost: $21.95)
'Volume 3. Calculation of Photochemial Reaction Photolysis Rates in the UNC
Outdoor Chamber." (Order No. PB 89-159 057/AS; Cost: $21.95)
'Volume 4. Appendixes to Photochemical Reaction Photolysis Rates in the UNC
Outdoor Chamber," (Order No. PB 89-159 065/AS; Cost: $21.95)
The above documents will te available only from: (Costs subject to change)
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22 1SJ
The EPA Prefect Officer caw tee
------- |