PB2004-100994
COMMUNITY SCALE AIR TOXICS MODELING WITH CMAQ
Jason Ching*, Sylvain Dupont(1), Jerry Herwehe(2) and Ruen Tang(3)
*Atmospheric Modeling Division, NERL, USEPA, RTP, NC 27711
Email address: china.iason@epa.gov
Voice: (919) 541-4801; FAX: (919) 541-1379
(1)	Postdoctoral Fellow with UCAR at USEPA in RTP, NC, dupont@hpcc.epa.gov
(2)	Atmospheric Turbulence and Diffusion Division, Air Resources Laboratory, Oak Ridge, TN,
herwehe@atdd.noaa.gov
^ Computer Sciences Corp, RTP, NC tang.ruen@epa.gov
1.0 INTRODUCTION
Consideration and movement for an urban air
toxics control strategy is toward a community, exposure
and risk-based modeling approach, with emphasis on
assessments of areas that experience high air toxic
concentration levels, the so-called "hot spots". This
strategy will require information that accurately maps
and characterizes the spatial and temporal variability of
such pollutants. Many air toxic pollutants are active in
photochemistry and ambient concentration levels will,
therefore, depend on both the magnitude of the
secondary products from the inflow regional background
as well as from fresh emissions. In principle, the
Community Multi-scale Air Quality (CMAQ) modeling
system, using multi-scale modeling attributes can
provide the ambient concentrations of air toxics from
both regional and local sources and through advanced
treatment of chemical, transport and deposition
pathways. This paper explores the CMAQ capability to
model air toxics at fine scale to meet the desired air
toxics assessments objectives.
2.0 METHODOLOGY:
We start by setting the nesting of CMAQ for
modeling from regional to fine scales. Modeling results
for various nests will be displayed and discussed. Given
that exposure and risk assessments are typically
focused on populations in urban and industrial areas,
we review the requirements for modeling meteorological
and air pollution fields in urban areas at grid resolutions
of order 1 km. We subsequently utilize the 1.3 km grid
simulations in CMAQ, as a basis for examining the
inherent within-grid spatial variability unresolved at
native coarser scales. We do note that there is
additional sub-spatial grid variability at less than 1.3km,
but their treatment and contribution to sub-grid variability
The corresponding author is on assignment from the Air
Resources Laboratory of NOAA to the National
Exposure Research Laboratory of the United States
Environmental Protection Agency, MD:E243-04, RTP,
NC 27711.
are not discussed here. Rather, the methodology to
attain information at grid scales smaller than 1.3 km will
require utilizing dispersion and transport, and finer scale
modeling, and that their outputs will be in the form of
distribution functions to compliment the 1.3 km CMAQ
simulations.
Simulations for this study were made on episodic
bases and were focused on the Philadelphia area. MM5
and CMAQ simulations were performed using nests at
36, 12, 4 and 1.3 km resolutions and results are shown
for July 12 and 14, 1995. At 1.3 km, urban canopy
parameterizations, UCPs, were introduced into MM5 to
account for the impact of urban building structures on
the meteorological fields (Lacser and Otte, 2002,
Dupont, et al., 2003, and Ching et al., 2003), based on
Brown, 2000, and Martelli et al., 2002. Sensitivity
studies (not shown here) have shown pronounced
affects of the UCP on both the outputs of the MM5 and
the subsequent CMAQ simulations. The emissions were
also spatially resolved at 1.3 km grids. Ten (10)
additional vertical layers were introduced into both MM5
and CMAQ to provide vertical resolution for
implementing the UCP methodology. Sensitivity studies
showed some, albeit relatively small sensitivities to the
layer or layers nearest the surface in which small point,
area and mobile sources were introduced.
3.0 RESULTS
Figure 1 shows an example output of the
simulations for the four nested grids for CO. The results
clearly showed continued enhancements of the spatial
structure (gradients) and the concentration magnitudes
with decreasing grid size. These features are even
more pronounced (not shown here) in the case of
photochemically active pollutants such as NO* and O3.
Also, while not shown here, hot spot features do appear
at the 1.3 km grid resolution for several toxic species
such as formaldehyde and acetaldehyde. During this
study, it was clear that for these latter two toxic pollutant
species, the resultant concentrations consisted of a
relatively large regional component.
We now investigate the relative sensitivity of the
simulations to grid resolution. For this purpose, we

-------
assume the 1.3 km grid simulations as the base and by
aggregating of these values to grids for the coarser grid
modeling domain, we can examine the characteristics of
their within-grid variability. Figure 2 shows results using
the 1.3 km grid simulation for ozone at 4pm EDT
aggregated to 12 km. The results from aggregating
1.3km simulations differ significantly from that of the
native 12 km grid as seen in the right side of the figure.
Our neighborhood scale modeling paradigm for air
toxics assumes that when significant within-grid
concentration variability is known to exist, additional
information on the characteristics of such distributions
will be supplied to complement the grid resolved
simulations for supporting risk-based population
exposure assessments. The next series of figures
provides illustrative statistics based on aggregating the
1.3 km grid results to 4 and 12 km to demonstrate the
qualitative aspects of such distributions. For example,
Figure 3 shows the standard deviation of the within-grid
variability at 4 and 12 km (normalized by its respective
grid value). Moreover, as shown in Figure 4, the
distributions for each of the pollutants do exhibit a wide
range in the value and sign of its skewness. No
apparent form or structure emerge from these patterns;
further, these distributions evolve with time.
Since exposure estimates depend on
concentration and dosage, the magnitude of the range
of the within-grid variability becomes an important
measure of risk. Figure 5 shows such range computed
from the difference in the peak and minimum values of
the 1.3 km results for each cell of the 12 and 4 km
simulations (normalized by their respective coarse scale
aggregated grid mean). In this case, we see in the
central Philadelphia area, that the range can exceed the
mean by up to a factor of 2. No characteristic pattern of
variability of the features on range emerges within this
domain.
Figure 6 shows concentration distribution
histograms from CMAQ simulations for a 12 km grid in
central Philadelphia July 14, 1995 for the time sequence
17-20 GMT. Here, we can see that the histograms can
change rapidly in time, and their characteristics also
differ between the different pollutants. Several of the
distributions exhibit multimode character and such
shapes changes in time.
4.0 DISCUSSION AND SUMMARY
From this limited set of model runs, a few
noteworthy and general points emerge:
(1)	The introduction of UCPs impacts the resulting
modeled flow and air quality fields.
(2)	Resolving the flow and air quality at fine scales will
significantly increase the level of detail in the spatial
features, in the magnitudes of the concentration
gradients and their extreme values.
(3)	Compositing neighborhood scale simulations to
coarser scales yields different results when
compared to coarse grid native simulations; further
the fine scale grid simulations provide indications of
variability in coarser grid solutions. The character
of these results differs depending on the scale of
the coarse grid mesh.
(4)	The degree of within-grid variability is a function of
the grid resolution and pollutant species and of
course the characteristics of such variability are
dependent on many factors, including complexity of
the urban area, its source distribution etc.
(5)	While not presented, within-grid variability will
generally be present even at the 1 km mesh
resolution (These will arise as a combination of
variability due to within grid source configurations
and distribution as well as inherently due to
chemistry and turbulent interactions (Ching et al.,
2003). Investigations of methods to derive such
distributions are underway.
(6)	There is also an important implication arising from
the results of fine scale modeling to model
evaluation. This study suggests that in areas for
which within-grid air quality has an inherently high
degree of spatial variability, a comparison of model
results should factor-in such variability. Since
monitors will not, in general, be adequately sited to
represent the grid resolved value, it follows that
model comparison and evaluation should introduce
some measure of this variability.
Disclaimer: This paper has been reviewed in
accordance with United States Environmental Protection
Agency's peer and administrative review policies and
approved for presentation and publication.
5.0 References
Brown, M.J., 2000: Urban parameterizations for
mesoscale meteorological models, in: Mesoscale
Atmospheric Dispersion. Ed Z Boybeyi, Wessex Press,
448 pp.
Ching, J., A. Lacser, T. Otte, and S. Dupont, 2003: Air
quality simulations at neighborhood scales with CMAQ.
Proceedings of the Fourth Int'l Conference on Urban Air
Quality, Measurements, Modeling and Management,
Prague, CZ.
Dupont, S., T. Otte, and J. Ching, 2003: Simulation of
Meteorological fields within and above urban and rural
canopies with a mesoscale model (MM5) Bound. Layer
Meteorology, (submitted)
Lacser, A. and T. Otte, 2002: Implementation of an
urban canopy parameterization in MM5. Preprints,
Fourth Symposium of Urban Environment. Norfolk, VA.
American Meteorological Society, 153-154.
Martelli, A., A. Clappier, and M.W. Rotach, 2002: An
urban surface exchange parameterizations for
mesoscale models. Bound. Layer Meteor., 261-304.

-------
D [CO
CO j-ji 14,95,6pm (local)
Figure 1. CMAQ simulation of CO: Top (left: 36km,
right: 12 km). Bottom (left: 4 km, right: 1.3km)
Ozone @ 4 PM EDT (12 Km)
Top Itfl • Mejn Trfrn 1.3i BoAOn LrtlFjrcff §11 fcir> RH5 Miai
Prinl
J
Figure 2. CMAQ simulation of ozone, July 12, 1995.
CO @ 07 EDT
Tip: II Jill ijrii
irii mm
t
£-jnc11l 4 fjn aris
3d C#a' Mian
SkF^nesE (12 km|
Tep UfiOS	RHS Can*
iWlsn L'ts ftiWMvA IKS
a? EDI
X
Figure 3. CMAQ simulations of CO for July 12, 1995.
Figure 4. Skewness at 12 km grid resolution derived
from 1.3 km simulations for July 12, 1995

-------
Formaldehyde® 15 EDT
Top (S2 km tjitj!i Qcflum 4 km grid
Range-to-meanrs
{3rd means smn 1 jji
i 4CLi
iOt

UD


>
! IM
m
/
u !¦
»i
Figure 5. Grid and range-to-mean derived from 1.3 km simulations for July 12, 1995
CO at 1700 GMT
05 at 1700 GMT
NO* at 1700 GMT
ALD2 at 1700 GMT
F0RW at 1700 GMT
0 Ci.i &.« 0.6 O.I I
Mixing Ratio (ppmv)
CO at 1800 GMl
Mixing Ratio {pprn
CO at 1900 GMT
Mixing Ratio (ppmv)
CO at 2000 GMT
Mixing Ratio (ppmv)
0.15	0.7
Mixing Ratio (ppmv)
0, at 1800 GMT
0.15	0.7
Mixing Ratio (ppmv)
03 at 1900 GMT
Mixing Ratio (ppmv)
03 at 2000 GMT
0 0.01 0.(12 0.03 O.d 0.05 O.Ot 0.0? O.QI
Mixing Ratio [pprnv)
NO, at 1800 GMT
0.01 0.02 0.03 0.04 0.05 ('.OS 0.07 O.OS
Mixing Ratio [ppmv)
NO* at 1900 GMT
0.01 0.02 O.Q3 O.tu 0.05 O.Ot 0.07 0.09
Mixing Ratio [ppmv)
NO* at 2000 GMT
ALD2 at 1 800 GMT
Mixing Ratio (ppmv)
ALD2 at 1900 GMT
.002 0.003 O.t
Mixing Ratio (ppmv)
ALD2 at 2000 GMT
.002 0.003 0.00' 0.0(15 O.tOS 0.00/ 0.000	O.COO 0,007 O.OOS 0.00» O.0< O.Q11 0,012
Mixing Ratio (ppmv)	Mixing Ratio (ppmv)
FORM at 1800 GMT
.002 0.003 0.004 0.04S O.tOS 0.007 O.OD8	O.COS 0,007 0.004 0.0D9 9.0< 0.011 0.012 O.O
Mixing Ratio (ppmv)
FORM at 1900 GMT
0,007 O.OOS 0.00* O.OI 0.011 0.01
Mixing Ratio (ppmv)
FORM at 2000 GMT
Mixing Ratio (ppmv)
Mixing Ratio [ppmv)
Mixing Ratio (ppmv)
Mixing Ratio (ppmv)
Figure 6. Concentration distribution histogram for 12 km cell in Central Philadelphia. From left: CO, O3, NOx
Acetaldehyde and Formaldehyde. From top, 1700, 1800, 1900, and 2000 GMT, July 14, 1995

-------