United States
Environmental Protection
Atmospheric Sciences
Research Laboratory
Research Triangle Park NC 27711
Research and Development
EPA/600/S9-87'024  Jan. 1988
Project  Summary
Workshop on Evaluation/
Documentation of Chemical

Roger Atkinson, Harvey Jeffries, Gary Whitten and Fred Lurmann
  Atmospheric photochemists and
developers and users of air quality
models  need  to  discuss the
evaluation  and  documentation  of
chemical mechanisms used  in air
quality  simulation models.  A
workshop, therefore,  was  organized
and conducted  on December 1-3,
1986 by EPA  to discuss the  latest
evidence  and viewpoints on the
subject and to solicit from experts
recommendations  on  optimum
approaches to mechanistic model
evaluation, documentation, and
further  development.  Previous
practices and underlying  issues  in
the subject areas were reviewed and
discussed in background documents
prepared and  distributed in advance
of the  workshop. Participants agreed
that smog chamber data provide the
most  unambiguous  test  of  urban
atmospheric  photochemistry
mechanisms. They  also  agreed,
however, that  there are uncertainties
associated with the representation of
chamber radical  sources and  of
photolytic  rates   in  outdoor
chambers, with  smog  chamber
measurement  errors,  and with the
representation of  as yet unknown
reaction pathways. The participants
recommended that task forces and/or
review groups be established  to
discuss and resolve  existing  smog
chamber methodology issues,  to
assemble a required  smog chamber
data base  for mechanism testing,
and to review/evaluate  relevant
kinetic and smog chamber data and
mechanism  testing  results.
Recommendations were   also
 developed on future research needs
 and  on mechanism documentation
   This Project  Summary  was
 developed  by  EPA's  Atmospheric
 Sciences  Research 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


   The  EPA conducted  this  workshop to
 discuss   the   evaluation  and
 documentation of chemical  mechanisms
 used in air quality simulation models. (A
 chemical  mechanism  is  the set  of
 chemical reactions  and associated  rate
 constants  which describes  the
 transformation of emitted chemicals into
 intermediate  and final products. In  the
 context of this  workshop,  the initially
 emitted chemicals are hydrocarbons and
 oxides  of nitrogen,  and ozone is  the
 product species of major interest.) Goals
 of this workshop were:

 • to  assess  present  practice  in
   photochemical reaction  mechanism
   development  and testing  for those
   mechanisms intended for use in urban
   air quality control calculations;

 • to determine  if there might be a
   commonly agreed upon  mechanism
   evaluation procedure; and

 • to determine  if there might be a
   standard  data base  that  would  be
   useful  in distinguishing among
   different mechanisms.

  A review of previous practice and a
discussion of the underlying  issues was
presented in  background  documents
"The Science of Photochemical Reaction
Mechanism   Development  and
Evaluation"  by Jeffries and  Arnold and
"Need  for Chemical  Mechanism
Documentation"  by  Sexton and  Jeffries,
which were distributed  to  workshop
participants  prior to the meeting.
  Four scientists and  an  EPA  user of
models were  asked to respond  to the
background documents and to offer their
viewpoints on evaluation procedures and
testing data bases.  These were Kenneth
Demerjian,  Roger  Atkinson,  Michael
Gery, Allan  Dunker, and Joseph Tikvart.
These four scientists were in agreement
that there is, and has been,  a generally
accepted procedure for testing the extent
of "reasonable  agreement" between
model  predictions  and  experimental
measurements. This procedure  involves
the  use   of   laboratory  kinetic,
mechanistic, and product data, the use of
smog chamber data, and the use of other
test data, such as captive air irradiation
measurements  and  ambient  air
measurements. Many  participants
believed, however, that the latter type of
comparisons  (e.g.  model  predictions
compared to ambient measurements
and, for some, even captive  air  studies)
require so  many  approximations and
suffer from  such instrumental limitations
that  the extent  of  agreement expected
would be quite  limited  and  thus such
efforts would  not be clear tests of  our
understanding   of  the  chemical
transformation   processes. All  four
scientists agreed  that,  although  there
certainly were problems with their data,
environmental or smog chambers still
provided the most unambiguous data for
the  testing   of  urban   chemical
transformation mechanisms.  Subsequent
discussion confirmed that this approach
or method  was generally  the accepted
approach  used   by  the  workshop
   The EPA  model user (J.  Tikvart)
strongly supported  the draft  proposal for
mechanism  documentation.  Other
workshop attendees,  including  William
Carter, Fred Lurmann, Gary Whitten, and
Gregory McRae described their current
practice  and  recent model  testing
strategies and results.
   While various mechanisms have been
developed  and  tested  against  limited
numbers  of smog chamber experiments,
only two  mechanisms, the  SAPRC/ERT
mechanism and the latest Carbon Bond
Mechanism, have been tested against  a
large number of  chamber experiments
(ca. 500) from different chambers.
   The  Steering Committee  concluded
that, without further work, it  was not
possible  to  choose one  of  these
mechanisms over the other on the basis
of  scientific evidence,  and that  the EPA
should  be  encouraged  to  use both
mechanisms as a method  to estimate the
present  uncertainties in  control
requirement  predictions.  The  Steering
Committee  also concluded that several
review  groups  should  be assembled  to
review  the  kinetic  and  chamber data
bases  and  to review  the  extent   of
agreement among the  models and these
data bases. These recommendations will
be described below.
   It is  clear from the  data presented  at
this  workshop,  together with  work
published over  the past five years, that a
vast amount of progress has  been made
during  the  past  decade,  both with
respect  to  urban-area  chemical
mechanism development  and  the data
base upon which  these mechanisms are
based  and  tested.  Nevertheless,
additional  development  is  clearly
needed.  Recommendations  for  future
chemical  mechanism   development,
documentation, and testing intended  to
extend  the  present urban oxidant-only
mechanisms to the new  areas that will
confront EPA in  the coming years are
also given below.

Guidelines for Mechanism
Development  and Testing

    Demerjian described what most
workshop participants accepted as  a
general  approach  to mechanism
development and  testing.  The  major
components of this  approach are shown
in   Figure  1.  Within this  generally
accepted approach, however, there are
differences in the  practice  among
different modeling groups. With  respect
to  mechanism evaluation,  there was
general agreement that:

•  mechanism  testing should   be
   performed according to a "hierarchy
   of species,"

•  data from at least two, and preferably
   more, chambers should be used in the

•  testing  should include at least two

     -testing   and   refining  the
     representations   of  chamber-
     dependent phenomena, and
    -testing  and refining for  comple,.
    organic  species  and mixtures  of

• as much data as is available should
  used m the testing,  and at a minimum.
  15 experiments should be used in the
  first phase and 50  experiments in the
  second phase for each chamber.
_-- '""^ i
Testing against
Amb/ent Air Data
for Consistency
 Figure 1.    Schematic  Method  for
            Development  and Eval-
            uation   of   Chemical
            Mechanisms.  After  Ken
            neth Demerjian.
   Workshop  discussions  indicated
further  that assessing goodness-to-fit
between the experimental data and the
model predictions is a topic that requires
further  work  and  one for which the
Steering Committee recommends the
creation of a Task Group.  One approach
to showing the"goodness-to-fit," the
developers  of  the  SAPRC/ERT
mechanism used  simple statistical
measures  to describe the goodness-
to-fit for their model  As an example  of
another approach, the  developers of the
Carbon  Bond  Four mechanism  used
several hundred plots  of  model
predictions-vs-smog   chamber
observations to illustrate the adequacy  of
their  fit. Their  reports  also  generally
contain one or  two tables giving the
maximum 03 and time to maximum for
the model predictions  and the observed
values,  but no  scatter plots  or error
distribution plots  are given.  So-
members of the  Steering Committee .
that both of these approaches to showing
 "goodness-to-fit" are  incomplete.

    The detailed  chemical  mechanism,
  developed and  evaluated as discussed
  above, must be  condensed  (see Figure
  1) for use in air quality simulation models
    -  control strategy assessment
 purposes. Procedures  for  condensing
  mechanisms and testing the condensed
  mechanism against  simulations of the
  expanded  or  explicit mechanisms  were
  discussed in three  reports:  Whitten,
  Johnson, and Killus  ("Development of a
  Chemical Kinetic Mechanism for the U.S.
  EPA  Regional  Oxidant Model,"  EPA-
  600/3-85-026,   1985); Whitten  and
  Gery  ("Development  of   CBM-X
  Mechanisms  for  Urban and  Regional
  AQSMs," EPA-600/3-86-012,  1986),
  and  Lurmann, Carter, and   Coyner  ("A
  Surrogate  Species Chemical Mechanism
  for Urban-Scale  Air Quality Simulation
  Models, Adaptation of the Mechanism,"
  EPA-600/3-86-031, 1986).  Specific
  methods  for  mechanism condensation
  include eliminations  of  species that are
  unreactive   or  do  not  change  in
  concentration during  the reactions and
  thus should  not change the numerical
  results of  simulations   using the
  intermediate condensed   mechanism
  •«hen  compared  to   the detailed
   echanism. Other steps include the use
~'6f fractional  stoichiometric  coefficients
  and  intermediate species   which  have
  unimolecular branched pathways  of
  reaction.  As long as the unimolecular
   ;cay lifetimes are reasonably  short, the
**"rTumerical  simulation  results will not  be
  significantly affected. Tests are required
  for lifetimes  longer than a few  minutes.
  Other techniques include the  use of a
  "mass balance" or  "counter  species"
  which are given arbitrarily   high decay
  rates  to  form"intentional"   steady-
  state-like species. Condensation  steps
  beyond those above  involve assumptions
  about typical atmospheric situations, and
  the  condensed  versions   of the
  mechanism must be tested  to determine
  the  bounds of  such  assumptions.
  Techniques  have been developed  to
  identify  unimportant  reactions  and
  species.  Elimination  of unimportant
  reactions  provides  little   benefit  to
  simulation costs, because the costs are
  most sensitive to the number of species.
  Of course, fewer reactions  do  make a
  mechanism easier to understand.
    In   regards   to   mechanism
  documentation,  it was  stressed by  the
  Steering  Committee  that  clear and
   omprehensive documentation  of
  Chemical  mechanisms  by  their
  developers is needed  to ensure  their
  proper use. In his presentation, Joseph
Tikvart reviewed and strongly supported
the approach proposed in  the workshop
background  document  "Need  for
Chemical Mechanism Documentation,"
by Sexton and Jeffries. In this document
an  example  outline of  a  guidance
document for the application of chemical
mechanisms was proposed. Computed
solutions to mechanism test problems
are essential to proper implementation of
chemical mechanisms by  nondeveloper
users. The Steering Committee believes
that a minimum of four test problems are
needed.  These initial  test  problems
should not involve dilution, entrainment,
emissions injection, or  deposition,  but
rather be examples of the pure chemical
kinetics. Ideally, the solutions should be
computed using a high quality algorithm
such  as  the Gear algorithm  with tight
error  control.  In these test cases,  the
algorithm used, the method of computing
the Jocobian, the use  of  absolute  or
relative  error  tolerances,  and  the
minimum, initial, and maximum integrator
step sizes should be documented.  All
species should be integrated rather than
determined from  the  steady-state
assumptions.  Another  documentation
item  discussed  by  the  Steering
Committee was the  need for a standard
set  of  conditions  for  comparing
mechanism  prediction  of  VOC control
requirements  when  using  the  OZIPM
program. It  was suggested that  the
regulatory groups  in EPA should  be
involved in developing these test cases
to insure that they cover typical cases of
interest for regulatory purposes.

Differences  Among   Well-
Tested Mechanisms

   When the guidelines given here  for
chemical mechanism  development  are
followed by multiple research groups, it
is expected that large sections of  the
mechanisms developed will be  very
similar, if not  identical. The portions of
the  detailed  chemical  mechanism,
however, which  are either  unknown  or
only poorly  understood  will have to be
assembled using  estimations  or
arguments  by  analogy,   or perhaps
simply be parameterized. Thus, different
methods of  representing these unknown
or  poorly  known  sections  of  the
chemistry  will arise  in  different
mechanism developments.  The  latter
process will most likely  lead to detailed
(and subsequently condensed) overall
mechanisms  which,  based on  past
experience,  may differ in  their control
strategy predictions, despite the fact that
each  chemical  mechanism  may be
consistent  within  the  bounds  of
reasonable  agreement  with  the
elementary   reaction   kinetics,
mechanisms, and products data bases
and with environmental chamber data.
Other than  further  efforts to  refine the
measure of uncertainty in  both  the
kinetics and chamber  data  and  new
formulations   and  testing   of  the
mechanisms,  there may  well  be no
scientific reason to  accept or  reject one
of these chemical mechanisms over the
other. Given that EPA must proceed with
applications of  the  mechanisms,  it is
recommended that the  EPA not select
only a  single  mechanism for making
control strategy predictions.

Task or Review Groups  Needed

   The   workshop   participants
recommended  that  several interacting
sets of review   processes or  review
groups be established, each with a goal
of assessing and documenting  the  state
of knowledge and degree of reasonable
agreement to  be expected in a given
domain. At  least four different domains
were identified for such activities:

1) kinetic and mechanistic data needed
   in  constructing  photochemical
   transformation mechanisms;

2) environmental chamber data needed
   for  comparison  with mechanism

3) mechanism  intercomparisons tests;
4) user or application tests.

Short-Term Recommendations

•  The Steering  Committee believes that
   it may be possible to further refine
   existing mechanisms  on a  short-time
   scale and without the need  for any
   new  experimental data. While  this
   advancement   needs  no   new
   experimental  data, recommendations
   for  further well-defined research  may
   result from this effort. The first priority
   is to use the presently  available
   environmental  chamber data  base  to
   refine the testing of present  chemical
   mechanisms.  The Steering Committee
   therefore   recommends   the
   establishment of  a task force of  non-
   EPA scientists to evaluate and resolve
   chamber   effects   and  light
   intensity/spectral distribution issues.

•  One of the goals of this workshop was
   to identify a standard  data  base that

  would  be useful  in  distinguishing
  among  different mechanisms.  The
  Steering Committee concluded that it
  was  possible  to  assemble  a
  "necessary"  environmental  chamber
  data set, that  is, one  containing
  chamber experiments  which all
  mechanisms  for use  in urban areas
  would have to simulate. The data  base
  is described  as  "necessary" because
  if a  mechanism could not simulate
  these experiments, it mostly likely
  would  be considered  unsatisfactory
  for  EPA  applications,  but  the  data
  base  would  not be  "sufficient"  to
  resolve all questions associated with
  oxidant  prediction.  The Steering
  Committee  recognized  that  to
  assemble  such  a  data base would
  require significant input from chamber
  operators at both UNC and UCR and
  from model  developers at  SAI and
  UCR, as  well  as  other interested
  parties who  might  wish  to use the
  data. It is envisioned  that  the  data
  base would contain  about  100
  experiments  from  several  chambers.
  Each experiment would contain
  detailed  recommendations  and
  supporting information  on photolytic
  rates, chamber  wall processes, and
  initial and temporal conditions. These
  would  be the result  of  review,
  discussion, and consensus  among
  task force members.


• Establishment of Review  Group for
  Evaluation of Fundamental Kinetic and
  Mechanistic  Data  for  Use in Model

• Establishment of Review  Group for
  Evaluation of Environmental Chamber

• Establishment of Review  Group for
  Mechanism Intercomparison

• Establishment of Review  Group for
  Mechanism   Predictions    in

• Obtain   Additional  Data  on
  Atmospheric  Chemistry of  Organics.
  Specifically obtain  data   (a) on
  absorption  cross-section  and
  photodissociation quantum yields and
  product  data  for  the  carbonyl
  compounds  formed  as intermediate
  products in the  degradation schemes
  or organics, (b) to determine the fates
  of  HO-aromatic  adducts  under
  atmospheric  conditions,  (c)  to
  determine the  reactions of the  >Ce
  alkoxy and alkylperoxy radicals under
  atmospheric  conditions  and  the
  subsequent  reactions   of  their
  products, and  (d) to  determine the
  radicals  formed, and their yields, from
  ozone-alkene  reactions  under
  atmospheric conditions.

• Obtain  additional  Environmental
  Chamber Data of higher quality and
  also  for the  purposes of  (a)
  discriminating  between present
  approaches to representing chamber
  effects,  and  (b) testing of  acid
  deposition  and regional  and
  tropospheric models.

• Conduct  New  Studies of  Chamber-
  Dependent Effects

• Investigate  the Applicability  of the
  present EKiVIA  Method

  Roger Atkinson is with the University of California, Riverside,  CA 92521; Harvey
  J Jeffries is with the University of North Carolina, Chapel Hill, NC 27514; Gary
    Whitten is with Systems Applications, Inc., San Rafael, CA 94903; and Fred
    Lurmann  is with Environmental Research and Technology, Inc., Newbury Park,
    CA 91320.
  Basil Dimitriades  is the EPA Project Officer (see below).
  The complete  report, entitled  "Workshop  on  Evaluation/Documentation  of
  Chemical Mechanisms," (Order No. PB 88-134  358/AS; Cost: $32.95, subject
  to change)  will be available only from:
           National Technical Information Service
           5285 Port Royal Road
           Springfield, VA22161
            Telephone: 703-487-4650
  The EPA Project Officer can be contacted at:
           Atmospheric Sciences Research Laboratory
           U.S. Environmental Protection Agency
           Research Triangle Park, NC 27711
United States
Environmental Protection
Center for Environmental Research
Cincinnati OH 45268
  PERMIT No. G-35
Official Business
Penalty for Private Use $300