A PRELIMINARY EVALUATION OF
MODELS-3 CMAQ
Using particulate matter data from the IMPROVE Network
Brian K. Eder, Michelle R. Mebust and Sharon K. LeDuc*
I. INTRODUCTION
The Clean Air Act and its Amendments require the United States Environmental
Protection Agency (EPA) to establish National Ambient Air Quality Standards for
Particulate Matter (PM) and to assess current and future air quality regulations designed
to protect human health and welfare. Air quality models, such as EPA's Models-3
Community Multiscale Air Quality model (CMAQ) [Dyun unci Ching, 1999], provide one
of the most reliable tools for performing such assessments. CMAQ simulates air
concentrations and deposition of various pollutants including PM. These simulations,
which can be conducted on a myriad of spatial and temporal scales, support both
regulatory assessment as well as scientific studies by research institutions. Within
CMAQ is an aerosol component, or module, designed to simulate the complex processes
involving PM, which is commonly separated into PM25 (particles with aerodynamic
diameters < 2.5 |jm) and PMK) (aerodynamic diameters < 10 |.im).
In order to determine its value to the air quality regulatory communities, CMAQ
needs to be evaluated using observational data. One such evaluation, which compared
visibility parameters derived from CMAQ to visibility parameters obtained from National
Weather Service observations, revealed that CMAQ was able to replicate general spatial
and temporal patterns [Eder et al., 2000], The current evaluation compares PM simulated
by CMAQ with PM data collected by the Interagency Monitoring of PROtected Visual
Environments (IMPROVE) network.
* I5rian K. tiler**. Michelle R Mehust. Sharon k LeDuc** NIZR1.. U.S. Lnvironmental Protection Agency.
RT/' NC. 27711,USA. ** On assignment from Air Resources Laboratory. National Oceanic and Atmospheric
Administration, RTP. NC. 27711.USA
I

-------
2
IS.K. EDEIi ETAL
2. CMAQ
2.1.	General description
CMAQ, which is an Eulerian model, simulates the atmospheric and land processes
affecting the transport, transformation and deposition of air pollutants and their
precursors [Byun and Ching, 1999], CMAQ follows first principles and employs a "one
atmosphere" philosophy that tackles the complex interactions among multiple
atmospheric pollutants and between regional and urban scales. Pollutants considered
within CMAQ include tropospheric ozone, PM, airborne toxics, and acidic and nutrient
species. The model also calculates visibility parameters.
2.2.	CMAQ aerosol component description
The aerosol component within CMAQ, described in Binkowski [1999], was derived
from the Regional Particulate Model (RPM) [Binkowski and Shankar, 1995], itself an
extension of the Regional Acid Deposition Model (RADM) [Chang at al.y 1990], Particle
size distributions are represented as the superposition of three lognormnl modes. PM;>
panicles (also called fine particles) are represented by two modes, the Aitken and
accumulation modes, each having variable standard deviations. Aitken mode particles
are those with diameters smaller than about 0.1 nm. Accumulation mode particle
diameters range between 0.1 and 2.5(.im. Each mode receives primary emitted material,
is subject to wet and dry deposition, and may form through condensation of gaseous
precursors. The two modes interact through coagulation, and the Aitken mode may grow
into the accumulation mode and partially merge with it. The fine panicle species
considered within the CMAQ aerosol component include sulfate, nitrate, ammonium,
water, primary organic aerosols, secondary organic aerosols from anthropogenic and
biogenic origin, elemental carbon, and primary aerosol material not otherwise specified.
The coarse particle mode within CMAQ, representing particles having aerodynamic
diameters between 2,5 and 10 jim, consists of wind-blown dust and other large particles
of unspecified origin. Coarse mode particles in the model also undergo wet and dry
deposition.
2.3.	CMAQ model simulation characteristics
The 2000 release of Models-3 CMAQ was used in this evaluation The modeling
domain covers the eastern U.S. (Figure 1.), with each grid cell covering 36 km by 36 km.
The entire month of June 1995 was simulated. The domain's vertical profile contains 21
layers of varying thickness. The simulation used Version 2 of the Regional Acid
Deposition Model chemical mechanism (RADM2), which includes 57 species and 158
reactions, 21 of which are photolytic. The meteorological fields were derived from
MM5, the Fifth-Generation Pennsylvania State University/National Center for
Atmospheric Research (NCAR) Mesoscale Model [Grafl tsi a/., 1994], Emissions were
processed with the Models-3 Emission Processing and Projection System (MEPPS)
[Benjay el a!., 1999].

-------
A PRELIMINARY EVALUATION OF MODELS-3 CMAQ
3. IMPROVE speciated data
IMPROVE is a collaborative monitoring effort governed by a steering committee
composed of representatives from Federal, regional and State organizations [Pitchford
and Scruggs, 2000], The network was designed to (1) establish current visibility and
aerosol conditions, (2) identify the chemical species and emission sources responsible for
visibility degradation, and (3) document long-term visibility trends at over 100 locations
nationwide. The IMPROVE monitors collected two, 24-hour integrated samples
(midnight to midnight, local time) per week in 1995, Additional information concerning
IMPROVE can be obtained from a web site (vista.cira.colostate.edu/improve) maintained
by Colorado State University.
Unfortunately, the majority of IMPROVE monitors are located in western states; as a
result, only 18 sites fell within the model evaluation domain (Figure t). All the sites are
rural, except the urban WASH (Washington D.C.) site. Information concerning these
sites can be found in Table 1.
pf-\
SWi A m ~W J*	!
77—\
¦'MOOS*"' •
\ bowab ,>«** Hi	_
\ (
1
'ftCAP
V
*"\	MACAA.
		—77
• "i	CRSM H--» SI IRQ
DPBll/	'• — -
r^T |f
JU^WASlrtvf BRIG
doso'V'^IvU
y--—y' /•' ¦suiyv,'
K maca>~J ¦jF.rr^
*Sr>
STROMA
lOKEF. {
0v=w^3^r'\
f	1A

V j' - «
Figure I. IMPROVE stations within the CMAQ domain used in (his evaluation.

-------
I!.K. EDKU ETAl.
Table I. IMPROVE sire information
Code
Location
Long (°W)
Lat (°N)
Elev. (111
ACAD
Acadia NP, ME
68.308
44.415
129
OWA
Boundary Waters, MN
91.950
47.950
524
BRIG
Brigantine NWR, NJ
74.472
39.475
9
CHAS
Chassahowitzka NWR, FL
82.567
28.750
2
DOSO
Dolly Sods, WV
79.205
39.143
1158
GRGU
Great Gulf Wilderness, NH
71.217
44.300
439
GRSM
Great Smoky Mountains NP, TN
83.987
35.710
815
JEFF
Jefferson, VA
79.433
37.667
299
LYBR
Lye Brook Wilderness, VT
73.123
43.243
1010
MACA
Mammoth Cave NP, KY
86.075
37.277
248
MOOS
Moosehorn NWR, ME
67.283
45.1 17
76
OKEF
Okefenokee NWR, GA
82.117
30.765
49
ROMA
Cape Romain NWR, SC
79.583
33.033
3
SHEN
Shenandoah NP, VA
78.450
38.543
1098
S1-1RO
Shining Rock Wilderness, NC
83.283
35.650
1621
SIPS
Sipsey Wilderness, AL
87.382
34.358
279
IJPBU
Upper Buffalo Wilderness, AR
93.245
35.880
723
WASH
Washington, D.C.
77.063
38.932
16
4. Evaluation
The scope of this evaluation was somewhat hindered by the IMPROVE network's
twice-a-week sampling schedule, which when incorporated into the one-month
simulation period, limited the number of comparisons. (144 possible observations (18
stations, 8 days: June 3, 7, 10, 14, 17, 21, 24 and 28)). Summary statistics for each of the
5 species included in the evaluation are found below. Scatterplots are found in Figure 2.
PM,„
Organic
Carbon
Table 2. Summary statistics for speciated aerosol data
Species N
Sulfate 129
Nitrate 129
PM2.5 129
Source
CMAQ
IMPROVE
CMAQ
IMPROVE
CMAQ
IMPROVE
Mean
O'g/'n3)
4.98
4.83
0.21
0.31
9.06
12.96
129
CMAQ
IMPROVE
13.74
19.40
112
CMAQ
IMPROVE
1.53
2.32
CV*
Wl6
71.63
62.06
55.39
73.74
50.26
44.23
47.34
M;i\.
255.77
64.28
Mean
(fi«/m )
4.82
26.09
55.64
51.48
0.005
0.25 -0.26
0.32
0.34
4.10
5.59
CoclVicienl ol Variation
Bias ilclined as (IMI'ROVIi - CMAO)/C'MAO

-------
A PRELIMINARY EVALUATION OF MODELS-3 CMAQ
5
n
(D
1
3
£
05
20
15
V
5
0
Sulfate
y
O
O C" O
D
a
~ O
s£ a
3 XP X U
O oS-^
DO!
nS30^ oftr'ofb^oo

~
agfe«°oDo D
fe
10
Observed
20
Nitrate
£ 3 0 rF
C 13$
0.5 10 15
Observed
03 20

ID 20 30
Observed
30 40
Observed
¦D
% 3
"3
E °
CO
11
0:
Organic Carbon
0
nO
D 3D	I
„	c a>J
\s	-
„D ° o „
#3 D «L	3 "
0 r
Observed
Figure 2. Sc.itferplots ol'observed (IMPROVE) versus simulated (CMAQ) for each species Regression line
shown Units arc in MS

-------
e>
U.K. ICDEK F.TAL
4.1.	Sulfate
Examination of Figure 2 and Table 2 reveals a fairly good level of agreement
between simulated and observed sulfate concentrations. The simulated mean
(4.98|jg/m3), and coefficient of variation (79.16%) closely match those observed
(4.83|jg/m3, 71.63%). The overall r2 is 0.63 and the regression equation is
CMAQ=0.60+0.91 "IMPROVE. Given that the CMAQ aerosol module descends from
models designed to address the acid rain problem, it is not surprising that CMAQ
simulated sulfate concentrations agree well with observed sulfate measurements.
There is, however, a tendency for CMAQ to over predict concentrations (mean bias
of 0.37). Examination of the bias across space and time (not shown) reveals that it is
positive at 13 of the sites and on all but two days. An inflated positive bias (2.91) was
observed at the DOSO (Dolly Sods/Otter Creek Wilderness in WV) site/grid cell. This
inflated bias can be attributed to very small concentrations of sulfate observed at this
location (in particular on two days, 24 June and 28 June).
4.2.	Nitrate
Unlike sulfate, examination of Figure 2 and Table 2 reveals a very poor level of
agreement between simulated and observed concentrations of nitrate. While the
simulated mean (0.21 ng/nv') was relatively close to the observed mean (0.3 I uu/nv'). the
coefficient of variation was not (255.77% versus 64.28%). This poor agreement is also
reflected in the scatter plot, overall /•" (0.005) and the regression equation of
CM AQ=0.40+0.32* IMPROVE.
The model consistently underpredicts nitrate concentrations (mean bias = -0.20).
Examination of this negative bias across time and space reveals that it is negative at all
sites except two, GRSM (Great Smoky Mountains NP) and MACA (Mammoth Cave
NP), and on half of the days. Subsequent investigation has determined that, in the CMAQ
simulations used here, the ammonia emissions were too low. These ammonia under
predictions no doubt contributed to the nitrate under prediction seen in the model output.
Efforts are currently underway to determine more realistic ammonia emission levels,
eventually allowing a more accurate model evaluation with respect to nitrate.
4.3.	PMjs
A reasonable level of agreement can be seen between simulated and observed PM25
concentrations (Figure 2., Table2.). The simulated mean (9.06 ng/nv1) and coefficient of
variation (62.06%) reasonably match those observed (12.96 pg/m \ 55.39%). as do the
various percentiles. The overall r~ is 0.55 and the regression equation is
CMAQ= 1.59+0.58* IMPROVE.
Because a large component of PM25 is sulfate, the good agreement seen in the
sulfate evaluation lends itself to reasonable PM3 5 results. The model consistently under
predicts PM2 5 concentrations (mean bias = -0.21) across time (all 8 days) and space (16
of 18 sites).

-------
A PRELIMINARY EVALUATION OF MODELS-3 CMAQ	7
4.4.	PMI0
Examination of Figure 2 and Table 2 reveals a poor level of agreement between
simulated and observed PM|0 concentrations. Although the simulated mean (13.74
|ig/nr), and coefficient of variation (73.74%) are reasonably close to those observed
(19.40 |ig/m3, 50.26%), the overall r7 is only 0.13. The regression equation of
CMAQ=6.52+0.37* IMPROVE and mean bias (-0.26) further reveal the model's
tendency to underpredict PM concentrations. Examination of the bias reveals that it is
negative at 16 of the sites and on all but one day (24 June). The only two sites with
positive bias were SHEN (0.16) and WASH (1.29).
Processes in the CMAQ aerosol module that involve PM|0 need better representation.
Efforts are underway to more accurately model wind-blown dust, as well as to include
sea salt in aerosol dynamics.
4.5.	Organic carbon
CMAQ simulates organic carbon with a modes! level of agreement. The r is 0.25
and the regression equation is CMAQ=0.76+0.34*IMPROVE. As with most of the other
species, CMAQ generally underpredicts organic carbon concentrations (mean bias of -
0.26). This bias is negative at 15 sites and across every simulation day; however, it is not
as large as the bias seen for the other species, as 8 of the 18 -sites and 5 of the 8 days are
within 25%.
This modest level of agreement results partially from the crude physical
representation (currently undergoing improvement) of organics within the CMAQ aerosol
component. Further difficulties arise from incomplete knowledge regarding organic
aerosol constituents, making it difficult not only to adequately model organic species, but
to compare model results with observations.
5. Summary
This evaluation compared speciated aerosol data collected during the month of June
1995 against CMAQ simulations for five species: sulfate, nitrate, organic carbon, PM:5
and PM|0. With the exception of sulfate (36%), the model simulations generally
produced negative biases (model predictions too low) of between -21 and -26%. This
negative bias was generally consistent across the domain and throughout the simulation
period. Agreement between model simulations and observations varied considerably
across species, with r's of 0.63 (sulfate), 0.55 (PM2 5), 0.25 (organic carbon), 0.13 (PM|„)
and 0.005 (nitrate)
Several likely sources of error in the model simulation were identified and include
inadequate emissions inventories and an incomplete understanding of aerosol dynamics,
especially for PM,0. Inadequacies in evaluation data sets have also been identified.
Fortunately, the EPA has recently implemented the National PM2 5 Monitoring Network,
consisting of mass monitoring (1100 sites), routine chemical speciation (300 sites) and

-------
%
b.k. i:ui:k fj al
supersite characterization. These network measurements will eventually provide much
more adequate data, thus allowing for a more thorough evaluation of CMAQ.
6. References
Benjey, W.G., J.M, Godowitch and G.L. Gibson, Emission subsystem, in Science algorithms of the El'A
Models-3 Community Multiscale Air Quality (CMAQ) modeling system, edited by D.W. Byim and J K.S.
Ching, pp. 4-1 - 4-107, EPA-600/R-99/030, U.S. Environmental Protection Agency. U.S. Government
Printing Office, Washington D.C., 1999.
Binkowski, F.S., Tlie aerosol portion of Models-3 CMAQ. in Science algorithms of the EPA Models-3
Community Multiscale Air Quality (CMAQ) modeling system, edited by D.W. Byim ;nul J.K.S Ching pp
10-1 -10-23, EPA-600/R-99/030, U.S. Environmental Protection Agency. US. Government Priming
Office, Washington DC., 1999
Binkowski, F.S and U. Shankar. The regional paniculate matter model. I. Model description and preliminaiy
results. Journal of Geophysical Research, 100, 26.191-26,209, 1995
Byun, D.W. and J.K.S. China, Science algorithms of the El'A Models-3 Community Multiscale Air Quality
(CMAQ) modeling system. EPA-600/R-99/030, U S. Environmental Protection Agency. U.S Government
Printing Office. Washington DC., 1999.
Chang, J.S., F.S. Binkowski, N I.. Seaman. D.W. Byun. J.N Mcllenry. P.J. Samson. W.R, Stockwcll, C.J.
Walcek, S. Madronich. P.B. Middleton. J.E. Pieim and ll.L. I.andsl'ord, Tlie regional acid deposition
model and engineering model, NAPAP SOS/T Report 4, in National Acid Precipitation Assessment
Program Acidic Deposition: State of science und technology. Volume /. Washington DC., 199(1
Eder. B.K... MR. Mebust, F.S Binkowski and S J. Roselle. A preliminary evaluation of Models-3 CMAQ using
visibility parameters. International Symposium on the Measurement of Tosie and Related Air Pollutants.
September 12-14, 2000. Research Triangle Park, NC.
Grell, G.A., i. Dudhia and DR. Stauffer, A description of the fifth-generation I'enn State/NCAR mesosealc
model (MM5), NCAK Technical Note. NCAR/TN-39X->-STlC 1994
Pitchf'ord, M L. and M. Scruggs. IMPROVE network - Current and future configurations, in; Spatial ant!
seasonal patterns and temporal variability of haze and its constituents in the United Slates. \V ('. Malm,
principal author, pp 1-1 - I-IK, Cooperative Institute for Research in the Atmosphere. Colorado State
University. ISSN 0737-352-47. 2000
Disclaimer. This document has been reviewed and approved by the U.S. Environmental
Protection Agency for publication. Mention of trade names or commercial products does
not constitute endorsement or recommendation for use.

-------
k
NERL—RTP-AMD-01—116 TECHNICAL REPORT DATA
1. REPORT NO.
EPA/600/A-01/07 6
2.
3.1
4. TITLE AND SUBTITLE
A Preliminary Evaluation of Models-3 CMAQ using
particulate natter data from the IMPROVE network
5.REPORT DATE
6.PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
Brian K. Eder, Michelle R. Mebust, and Sharon K.
LeDuc
8.PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Same as Block 12
10.PROGRAM ELEMENT NO.
11, CONTRACT/GRANT NO.
12. SPONSORING AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency
Office of Research and Development
National Exposure Research Laboratory
Research Triangle Park, NC 27711
13.TYPE OF REPORT AND PERIOD COVERED
Preprint. FY-01
14. SPONSORING AGENCY CODE
EPA/600/9
15. SUPPLEMENTARY NOTES
16. ABSTRACT
The Clean Air Act and its Amendments require the United States Environmental Protection Agency (EPA) to establish National Ambient Air Quality Standards (or Particulate
Matter (PM) and to assess current and future air quality regulations designed to protect human health and welfare Air quality models, such as EPA's Models-3 Community
Multiscale Air Quality model (CMAQ) [Byun and Ching, 1999], provide one of the most reliable tools for performing such assessments. CMAQ simulates air concentrations and
deposition of various pollutants including PM. These simulations, which can be conducted on a myriad of spatial and temporal scales, support both regulatory assessment as
well as scientific studies by research institutions. Within CMAQ is an aerosol component, or module, designed to simulate the complex processes involving PM, which is
commonly separated into PM,, (partides with aerodynamic diameters £ 2.5 nm) and PM,0 (aerodynamic diameters £10 nm). In order to determine Its value to the air quality
regulatory communities, CMAQ needs to be evaluated using observational data. One such evaluation, which compared visibility parameters derived from CMAQ to visibility
parameters obtained from National Weather Service observations, revealed that CMAQ was able to replicate general spatial and temporal patterns [Ederet al., 2000]. The
current evaluation compares PM simulated by CMAQ with PM data collected by the Interagency Monitoring of PROtected Visual Environments (IMPROVE) network.
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
b.IDENTIFIERS/ OPEN ENDED
TERMS
c.COSATI



18. DISTRIBUTION STATEMENT
RELEASE TO PUBLIC
19. SECURITY CLASS (This
Report)
UNCLASSIFIED
21.NO. OF PAGES
8
20. SECURITY CLASS (This
Page)
UNCLASSIFIED
22. PRICE

-------