United States Environmental Protection Agency	Office of Research and Development

National Exposure Research Laboratory
Research Abstract

Government Performance Results Act (GPRA) Goal 1
Annual Performance Measure #337

Significant Research Findings:

Evaluate Models-3/CMAQ
for Particulate Matter using Episodic Data

Ambient air concentrations of particulate matter (PM) continue to be a major
concern for the Agency. High concentrations have been linked to detrimental
health effects, acid precipitation, regional climate change and visibility
degradation. Accordingly, the Clean Air Act and Amendments call for an
assessment of the impacts of current and future regulations designed to protect
human health and welfare. The most reliable tools for carrying out such
assessments are air quality models such as Models-3 Community Multiscale Air
Quality (CMAQ) model. In order to establish credibility and build confidence
within the user and scientific community, CMAQ like all deterministic models
needs to be evaluated through comparisons with observations. Accordingly, the
purpose of this research is to thoroughly evaluate and characterize the performance
of the CMAQ system and its emissions, meteorological and chemical/transport
modeling components, with an emphasis on air concentrations of fine particles,
which is central to the 2005-2008 SIP implementations.

Characterization of the model's performance involved two types of evaluation.
The first type was an operational evaluation, which utilized data from routine,
nationwide monitoring networks (e.g., the Speciated Trends Network (STN), the
Clean Air Status Trends Network (CASTNet)) to determine if the PM
concentrations of regulatory interest are adequately being simulated (Is the model
providing the right answer?). These are fairly dense networks that collect data on
a long-term continuous basis, but the data are typically daily or weekly average
values. The second type was a diagnostic evaluation, which utilized intensive,
regional field studies (e.g., the Nashville Southern Oxidant Study, Atlanta
Supersite Experiment, EPA Supersite Program) that collect a broader range of
chemical species on an hourly basis to examine the ability of the model to
accurately simulate all of the physical and chemical processes that lead to the
simulated concentration (Is the model providing the right answer for right
reason?). In addition to direct evaluations, model sensitivity analyses were also
conducted as part of this task to characterize model response to uncertainties.

Two simulation periods (4 January - 19 February 2002 and 15 June - 16 July,
1999) were selected for evaluation of the CMAQ model performance for fine
particulate matter. Observational data from four networks were used: the Clean Air
Status and Trends Network (CASTNet), the Interagency Monitoring of Protected
Visual Environments (IMPROVE) network, the Speciated Trends Network (STN),

Scientific
Problem and
Policy Issues

Research
Approach

Results and
Impact


-------
and the Southeastern Aerosol Research and Characterisation (SEARCH) network.
A series of evaluation tests were performed using the FY02 public release version
and beta-versions of the eventual FY03 public release version of the Models-
3/CMAQ. With the exception of nitrate, the FY02 CMAQ simulation results
indicated generally good performance, especially for sulfate and ammonium
where, depending on network and episode, correlations were generally quite good
(e.g., correlation coefficients > 0.70). Nitrate, on the other hand, was greatly over-
predicted by the model. The FY02 CMAQ's simulations of nitrate were deemed
unacceptable, and the correction to this "nitrate problem" was set as a top priority
for the FY03 public release of the model.

In order to identify the cause(s) for the nitrate overprediction, several diagnostic
evaluations were performed using highly time resolved (hourly) data obtained at
the Atlanta and Pittsburgh Supersites. These evaluations showed the FY02
CMAQ had very large over-predictions of nitric acid and/or total nitrate that
peaked at night, suggesting an issue with the nighttime heterogeneous production
of nitric acid on wetted aerosol particles. Reaction rates used in the FY02 CMAQ
for this nighttime pathway for nitric acid production were two to three orders of
magnitude higher than the probability estimates from the most recent literature. A
test of these new literature values for the heterogeneous reaction probabilities
showed a dramatic improvement in the predictions of CMAQ for nitric acid and
aerosol nitrate, although CMAQ was still over-predicting nitric acid at night.
Sensitivity studies showed that eliminating altogether the nighttime heterogeneous
production of nitric acid in CMAQ brought its predictions in line with the
nighttime levels of nitric acid at both special sites. The comparisons of these
CMAQ sensitivity runs now showed a modest over-prediction of nitric acid
occurred during the day, implying that the photochemical mechanisms for ozone
production are also creating part of the nitrate over-prediction problem. The
episodic evaluations after the modifications to the heterogeneous production of
nitric acid in CMAQ indicate that the heterogeneous reaction probability is still
too high; thus, the "nitrate problem" has been greatly ameliorated, but not
eliminated. There are suggestions in the laboratory research community that a
variety of factors exist that further inhibit these nighttime reactions, but none are
published and available for use by the CMAQ developers at this time.

A serious deficiency in the CMAQ simulation was, to an acceptable degree, fixed
with the help of these episodic evaluations and sensitivity tests. Changes to the
nighttime heterogeneous reaction rate for production of nitric acid were
incorporated into the FY03 public release version of CMAQ, and episodic
comparisons against observational networks and special hourly sites indicated
improvements to nitrate predications as a result of these changes. However,

CMAQ still shows a moderate, systematic over-prediction of total nitrate and,
hence, particulate nitrate, based on comparisons against the CASTNet data. It is
important that continued improvements are pursued because CMAQ is expected to
predict ammonia limitation more often than it should in certain areas of the
country because of these total nitrate over-predictions. Thus, the FY03 and FY04
release versions of CMAQ could over-emphasize the nitrate replacement that can
offset part of the reduction in sulfate that will accompany reductions of S02


-------
emissions. As the state of science for nitrate aerosol predictions continues to
improve, these potential impacts on regulatory decision making will be further
tested against episodic data to better understand these interactions and impacts.

Division staff conducted the model evaluation studies. Network data (e.g.,
CASTNet, STN, etc.) were obtained through public access web sites. Episodic
data such as the Nashville SOS intensive were obtained directly from the
Principal Investigators who maintain the data. Key, recent presentations and
publications are provided below:

Dennis, R. L. "CMAQ winter predictions of nitrate: the importance of N205 reactions to HN03

production". Presented at NOAA's Aeronomy Laboratory, June, 2003.

Dennis, R. L. "Time resolved and in-depth evaluation of PM and PM precursors using CMAQ".

Presented at EPA's PM Model Performance Workshop, Chapel Hill, NC. February 2004.
Arnold, J.R., R.L. Dennis, and G.S. Tonnesen. Diagnostic evaluation of numerical air quality

models with specialized ambient observations: Testing the Community Multiscale Air
Quality modeling system (CMAQ) at selected SOS 95 ground sites. Atmospheric
Environmental'. 1185-1198 (2003).

Yu, S, Dennis, R. L., Roselle, S. J. Nenes, A., Walker, J., Eder, B. , Schere, K.and Swall, J. "An
assessment of the ability of 3_D air quality models with current thermodynamics
equilibrium models to predict aerosol nitrate." Journal of Geophysical Research.
American Geophysical Union, Washington, DC, 110(D7):l-22, (2005).

Future Research Future episodic model evaluation research will focus on both winter and summer
months and will use data from the EPA Supersite Program (1) to examine the
diurnal performance of CMAQ on several inorganic and primary species and (2) to
further examine the implications of the nighttime over-prediction of nitric acid on
the control strategy response of CMAQ to reductions in S02 emissions. In the
latter work, indicators of the inorganic system state (whether it is ammonia limited
or nitric acid limited) will be constructed from CMAQ predictions and compared
against indicators derived independently from measurements, in order to assess the
expected reliability of the control strategy predictions of CMAQ for inorganic
species and to evaluate the indicators as diagnostic metrics.

Questions and inquiries can be directed to:

Brian K. Eder, Ph.D.

U.S. EPA, Office of Research and Development
National Exposure Research Laboratory (MD-E243-01)

Research Triangle Park, NC 27711
Phone: 919/541-3994
E-mail: eder.brian@epa.gov

Robin L. Dennis, Ph.D.

U.S. EPA, Office of Research and Development
National Exposure Research Laboratory (MD-E243-01)

Research Triangle Park, NC 27711
Phone: 919/541-2870
E-mail: dennis.robin@epa.gov

Research
Collaboration and
Research
Products

Contacts for

Additional

Information


-------