Air Quality Modeling at Neighborhood Scales to Improve Human Exposure Assessment
J. K. S. Ching*. A. Lacser+, D. Byun and W. Benjey
Atmospheric Sciences Modeling Division, Air Resources Laboratory
National Oceanic and Atmospheric Administration, Research Triangle Park, NC 27711, USA
+On leave from the Israel Institute for Biological Research, Ness Ziona, Israel
1. Introduction
Air quality modeling is an integral component of risk assessment and of subsequent development of
effective and efficient management of air quality. Urban areas introduceof fresh sources of pollutants into
regional background producing significant spatial variability of the concentration fields and
corresponding human exposures. Typically, air pollutant concentration data in urban areas used by
exposure models are from central site monitors that provide limited or no information on spatial
variability. This paper describes a methodology for bridging air quality dispersion modeling and
exposure approaches to provide a basis for assessing the impacts of such concentration variation on
human exposures. For this approach the Models-3 Community Multiscale Air Quality Modeling System
(CMAQ) (Byun and Ching, 1999) spatial resolution was refined from 4km to 1.33 km. This preliminary
sensitivity study will illustrate human exposure to several pollutants as a function of these grid sizes. The
approach sets the stage for the modeling of exposure to air toxics.
2. Methodology
2.1 Concentration field
The Eulerian CMAQ model is driven by meteorological and emissions processors. The MM5V2.10
was used as the meteorological model. Runs were made with the nonhydrostatic version and Blackadar
boundary layer parameterization for the 12 and 4-km grid sizes over the NE portion of the US (one way
nesting). The Grell cumulus cloud scheme and analysis nudging were applied only to the 12 km grid. The
12 km run used the surface energy scheme and the 4 km run used the Dudhia's atmospheric radiation
scheme and treated the clouds explicitly. The CMAQ runs for the 1.33-km resolution used interpolated 4-
km grid meteorological values (based on MM5V3 land use). The emission processor (Byun and Ching,
1999) supplied the source intensity as a function of time and space from different sources (area, mobile,
point and biogenic). CMAQ calculates the concentrations of the pollutants (offline from MM5) by taking
into account advection (with piecewise parabolic method), dispersion (with K-theory, PBL similarity),
gas/aqueous chemistry (with CB4 mechanism and the GEAR solver) including aerosols (modal approach)
and deposition. CMAQ was run with 21 vertical levels-10 levels are below 900 m.The model was applied
to the Philadelphia metropolitan area for the period 07/11-07/17/1995 (Byun et. al., 2000).
2.2 Exposure calculations
Exposure is defined as any contact between chemical at a specified concentration and the outer/inner
surface of the human body. The degree of exposure is influenced by the duration (how long one is
exposed), magnitude (concentration) and frequency (how often one is exposed). The basic concept is that
concentrations of pollutants measured at locations (microenvironments) where people are, multiplied by
the time spent in each place will approximate personal exposure. The concentration in the
microenvironment is a function of the ambient concentration, the infiltration rate between the outside and
microenvironment, the air exchange rate and indoor sources. The total exposure is temporally and
spatially dependent. To assess population exposure, one integrates over the population and over
appropriate time duration for significant impact to occur. Currently, human exposure models utilize
stochastic approaches where the parameters that influence exposures are randomly sampled. Population
exposure models can also simulate an individual's exposure using human activity pattern data.
-------
For this initial study, a simple algorithm was used to estimate the average population exposure:
ExpflHZ ij (C jj (t) * {£ k F k P k (t)} * POP y *At)yi ij POP ,j
Where C ij (t) is the CMAQ concentration for time t at grid cell (i j), P k (t) is the percent of population in
microenvironment, k, for time t calculated from activity diary databases (we used 4 microenvironments -
indoors/residential, other indoors, outdoors and in vehicle), F k is the pollutant specific indoor/outdoor
concentration factor for microenvironment k (We used the values of 0.9, 0.8, 1, 1.2 for PM 25 and 0.4,
0.2, 1., 0.4 for 03 respectively (Burke, 2000)). This set assumes that there is no indoor source of
pollution. POP ij is the total population in grid cell (i,j), At is one hour in our case. Model-3/CMAQ
simulations were performed at grid sizes of 12, 4 and 1.33km. We will concentrate on a domain of 120km
x 120 km encompassing Philadelphia, PA. Gridded population density data were derived from US census
data for the census tracts within the model domain (see Figure 1). Figure 2 depicts the average percentage
of population in various microenvironments on an hourly basis. This distribution is applied uniformly
throughout the domain. In this paper we present preliminary predictions of 03 and PM2 5 (consisting of
sulfates, nitrates, ammonium, organic mass and elemental carbon) exposure and their sensitivity to grid
size.
3. Preliminary Results
3.1 Meteorology
The predicted maximum wind speeds in the 4 km grid (first layer) are higher than those in the 12 km
grid. For example, on 07/13 at 1400 UTC the maximum wind speeds were 7.2 m/s and 5.1 m/s in the 4
and 12 km grid, respectively. The domain averaged PBL heights were higher and the growth rate was
steeper in the 4-km grid. The 10-m temperatures were also higher in the 4-km grid (~l-2 ° K). These
results are not shown here.
3.2 Ozone and PM 2.s concentrations
The temporal behavior of the domain averaged ozone concentrations is similar for the three grid
resolutions, whereas the PM2.5 displays dependence on grid size (not shown here). The maximum value
for 03 concentrations within the simulation domain shows small sensitivity to grid size (4 and 1.33 km),
but the 1.33 km PM2.5 maxima are higher than the 4 km and the 12 km grid size predictions (see Figure
3). Note also the much less fluctuating daily behavior of the PM2.5 maxima predicted for the 12 km grid.
Predicting maximum 03 values does not need finer grids than about 4 km, whereas PM2 5 shows
sensitivity to finer grid sizes than 4 km. Part of the reason for that is the fact that 03 is a secondary
pollutant and PM consist also of primary sources and local meteorology might affect their fate more than
the 03 maximum values (personal exposure might be additionally influenced by local sources
contributing to sub-grid variability). Figure 4 shows the domain average concentrations and standard
deviation for 03 and PM25 - the sensitivity to the grid size is much smaller. The number of aerosol
particles per unit volume is also very sensitive to the grid size (not shown here).
3.3 Exposure assessment
The intra-day 03 exposure time series display a typical eastern US summer diurnal behavior (i.e.,
nocturnal minima and daytime maxima). In contrast, the PM2 5 exposure time series differ for each day
and do not show a regular diurnal pattern of behavior during that week (not shown here). Figure 5 shows
that the domain averaged exposure for 03 and PM2.5 are similar for 4 and 1.33 km grid size, but different
from the 12 km grid as knowledge of detailed meteorology is more important for primary sources than for
secondary/regional pollutants.
4. Discussion and Future Study
The exposure modeling approach employed in this initial study is simplified. It shows that one might
need fine grid sizes to predict exposure to primary pollutants, but might be satisfied with coarser grid
-------
sizes for secondary pollutants. The approach does not consider a more realistic activity pattern in which
individuals are exposed to conditions elsewhere from where they reside. Future efforts should model
activity patterns that reflect more realistic spatial distribution of microenvironments of actual exposures
and also express exposure values in a statistical way rather than in a deterministic way as we did here
(e.g., using the SHEDS model - Burke et. al., 2000). From photochemical modeling and PM2.s modeling
'the effort will extend to speciated PM2.5 including number of particles and toxic compounds at
neighborhood scales. Modeling compounds such as semi volatile organic compounds e.g., PAH's and
pesticides can logically be an extension of PM2.5 modeling using gas -to-particle partitioning models.
Additionally, studies will include a) better parameterizations of chemical processes based on laboratory
and theoretical research including new chemical mechanisms that can handle interactions between the
toxic pollutants and short-lived radical species in the atmosphere, b) adding proper urban canopy
parameterizations to MM5 to predict meteorology at fine grids, c) methodologies to transport pollutants
through urban canopies, to link transport between sub-grid to grid scales, and describe exchanges of
pollutants between outdoor and indoor environments, d) data fusion methods that combine monitoring
data with model results, and e) methods to extrapolate episodic model results applicable to longer term
exposure time scale.
5. References
Burke, J.M., (2000) Personal communication.
Burke, J.M., Zufall, M. J. and Ozkaynak H. A. (2000) The contribution of ambient PM 2.5 to total
personal exposures: Results from a population exposure model for Philadelphia, PA. 10'h Annual
Conference oflSEA, October 24-27, 2000, Monterey CA.
Byun, D.W. and Ching, J.K.S., (1999) Science algorithms of the EPA Models-3 Community Multiscale
Air Quality (CMAQ) modeling system. EPA- 600/R-99/030, U.S. Environmental Protection Agency.
Byun, D. W., A. Lacser, B. Benjey and J.K.S Ching (2000) Effects of grid resolution on the simulation of
urban air quality: Application of Models-3/CMAQ to Philadelphia metropolitan area at 12,4 and 1.33 km
resolutions. Third Symposium on the Urban Environment. 14-18 August 2000, UC Davis, CA, pp 88-89.
6. Acknowledgments
Thanks to L. Truppi who produced the gridded population fields using GIS on Census Tract data. Thanks also
to Dr. J. Burke who provided data on the temporal population distribution in different microenvironments and
indoor/outdoor factors.
13.00 27
12.00
LOO
6.00
3.00
27
July 11.1985 000 AO
Min- 0.00 at(10.n M«X- 12.55 «t(14,15)
population distribution
indoors
oth.
indoors
vehicle
20 24
Figure 1. Philadelphia population density
presented on 4 km grid.
Figure 2. Population distribution as a function of
UTC time for the various microenvironments.
-------
OZONE PEAK 07/13
1.33 km —¦—4 km —*—12 km
150
120
a 90
a 60
30
0
12
UTC
16
20
24
PM PEAK 07/13
1.33 km —¦—4 km -
-12 km
120
100
80
40
20
0) @0 i r -A "A a
1F5hFFW1m
12
UTC
16
20 24
Figure 3. Ozone (left) and PM2.5 (right) maximum concentrations on 07/13
AVERAGE OZONE 07/13
¦ - ~ • -avr 12km —¦—avr4 km
8 12 16 20 24
AVERAGE PM 07/13
¦ - -avr 4km —¦—avr 1.3km
40.00
35.00
1 30.00
1 25.00
20.00
15.00
i fjttttffffftt
Wf
I 1
0 4
8 12 16 20 24
UTC
Figure 4. Domain averaged 03 (left) and PM2 5 (right) for different grid sizes (one +0 is plotted as dotted line for
the larger grid and -o as bold line for the smaller grid).
OZONE EXPOSURE 07/13
60
50
£ 40
2b 30
a 20
10
0
: J&
-1.33 km
—4 km
-k—12 km
0 4 8 12 16 20 24
hour
PM EXPOSURE 07/13
50
40
£
«
30
E
O)
20
10
0
- -1.33 km
—4 km
-h.— 12 km
0 4 8 12 16 20 24
hour
Figure 5. Ozone (left) and PM 2.s (right) exposure on 07/13
Disclaimer
This paper has been reviewed according to USEPA peer and administrative review policies and approved for
presentation and publication. Mention of trade names or commercial products does not constitute endorsement or
recommendation for its use.
-------
NERL—RTP—AMD-00-248 TECHNICAL REPORT DATA
1. REPORT NO.
EPA/600/A-01/001
2 .
3 . R
4. TITLE AND SUBTITLE
Air quality modeling at neighborhood scales to
improve human exposure assessment
5.REPORT DATE
6.PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
J.K.S. Ching1, A. Lacser2, D. Byun1, and W. Benjey1
8.PERFORMING ORGANIZATION REPORT NO.
9. PERFORMING ORGANIZATION NAME AND ADDRESS
'Same as Block 12
2Israel Institute for Biological Research, Ness
Ziona, Israel
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
Extended Abstract, FY-00
14. SPONSORING AGENCY CODE
EPA/600/9
15. SUPPLEMENTARY NOTES
16. ABSTRACT
Air quality modeling is an integral component of risk assessment and of subsequent development of effective and efficient management of
air quality. Urban areas introduce of fresh sources of pollutants into regional background producing significant spatial variability of the
concentration fields and corresponding human exposures. Typically, air pollutant concentration data in urban areas used by exposure
models are from central site monitors that provide limited or no information on spatial variability. This paper describes a methodology for
bridging air quality dispersion modeling and exposure approaches to provide a basis for assessing the impacts of such concentration
variation on human exposures. For this approach the Models-3 Community Multiscale Air Quality Modeling System (CMAQ) (Byun and
Ching, 1999) spatial resolution was refined from 4km to 1.33 km. This preliminary sensitivity study will illustrate human exposure to
several pollutants as a function of these grid sizes. The approach sets the stage for the modeling of exposure to air toxics.
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
20. SECURITY CLASS (This
Page)
UNCLASSIFIED
22. PRICE
------- |