A Performance Analysis of the CMAQ Model and Its
Sensitive to Ozone Precursors in East Tennessee
Control/Tracking Number: 05-A-623-AWMA
Karthikeyan Ramaswamy, Terry Miller
University of Tennessee, Department of Civil and Environmental Engineering, 223
Perkins Hall, Knoxville, TN 37996
ABSTRACT
This paper presents the results of performance and sensitivity analyses of the CMAQ
Model for ozone concentrations in East Tennessee. The objective of the research was to
conduct a sensitivity analysis on ozone precursor emissions and to determine the
optimum level in reducing the ozone precursor’s for attaining the federal 8-hr ozone
standard. In this research, Models-3/Community Multi-scale Air Quality (CMAQ) and
Sparse Matrix Operator Kernel Emissions (SMOKE) were utilized to predict base case
ozone concentrations based on 1999 National Emission Inventory (NEI) in East
Tennessee from August 27th to September 9th, 1999. The modeling domain consisted of a
three-tiered nested domain with grid resolutions of 36 km, 12 km and 4km. The 4 km
domain covered the entire State of Tennessee. Performance analysis of the CMAQ
model, comparing predicted results to actual measurement data, was followed by
sensitivity analyses for ozone precursors. The goal was to determine the possibility of
attaining the 8-hr ozone National Ambient Air Quality Standard (NAAQS) in East
Tennessee. The results suggest that 20 to 30 percent NOx reductions are needed in mobile
and/or point sources to attain the ozone NAAQS. The results also revealed that NOx
emission reductions from mobile sources are more effective in reducing ozone
concentrations than reductions from point sources, especially at high elevation rural sites.
The findings also address the feasibility of attaining the 8-hr ozone standard throughout
the modeling domain, with recommendations for improving the model.
INTRODUCTION
Community Multiscale Air Quality (CMAQ)[1], a Third Generation Photochemical Model
developed by USEPA was applied to the summer of East Tennessee, episode starting
from 27th August 1999 to 9th September 1999. The main purpose of this study was to
explore the sensitivity of the CMAQ model to various control scenarios and to analyze
the performance of the model compared to measured ozone concentrations in East
Tennessee air quality region, shown in Figure 1.
1
Figure.1 Counties in East Tennessee Air Quality Region [2]
To determine the technical feasibility of attaining the 8-hr ozone standard, and to identify
the best precursor reduction strategies for moving towards attainment, it is necessary to
understand the precursor’s sensitivity in that particular region since ozone formation rate
is highly depends upon ozone precursor emissions rates. The response of ozone
concentrations to changes in local VOC and NOx emission rates is therefore an essential
factor in designing an effective ozone abatement strategy [1].
The domain-used in this study is shown in Figure 2. Eleven layers in the vertical direction
and three levels of nested domains with grid resolutions of 36km for the outmost domain,
12km for the intermediate domain and 4km for the innermost domain were used in this
study. The outmost domain encompasses most of the eastern United States. The
intermediate domain was centered on Tennessee and other southeastern states. The
innermost 4km grid domain covers the entire state of Tennessee.
2
Figure.2 Nested Domains [3]
Emission Inventory
The emissions inventory used in this study was the 1999 TDEC [4] inventory and was
used as the base case emissions to support the performance evaluation modeling as well
as for sensitivity modeling. The base year inventory consisted of emissions from point,
area, biogenic and mobile sources. This emission inventory includes emissions of NOx,
VOC, CO, NH3, SO2, PM10 and PM2.5 in tons/day and in tons/year.
Monitors
Eight monitoring stations were chosen to represent East Tennessee. Table 1 shows the
locations, AIRS identification number, elevation, and the latitude - longitude of each
monitoring station in East Tennessee [5]. Real time values from these monitoring stations
were used to compare how well the CMAQ model simulated the ozone concentrations
that were actually measured at eight ozone-monitoring stations during the episode
considered.
3
Table.1 Monitoring Station Information [5]
Latitude:
(Degree)
Longitud
e:
(Degree)
Station Name
AIRS
Identification
Number:
Anderson
47-001-0101
780 ft
237.7 m
35.96508
-84.22324
Blount-Cades Cove
47-009-0102
1850 ft
563.8 m
35.60304
-83.78362
Blount-Look Rook
47-009-0101
2700 ft
822.9 m
35.6315
-83.94368
Jefferson
47-089-0002
1017 ft
309.9 m
36.11444
-83.60111
Knox-Rutledge Pike
47-093-0021
980 ft
298.7 m
36.08505
-83.76466
Knox-Mildred Drive
47-093-1020
1056 ft
321.8 m
36.01834
-83.87614
Sevier-Cove Mountain
47-155-0101
4150 ft
1265 m
35.69653
-83.60986
Sevier-Clingmans Dome
47-155-0102
6610 ft
2015 m
35.56279
-83.49807
Elevation
Nitrogen Oxides and VOCs
In Tennessee, based upon 1999 emission inventories, 41% of the annual NOx emissions
originated from the mobile sources (transportation sector), 43% of the annual NOx
emissions from the point sources (industrial sector) and 16% from the Area sources
(Figure 3). VOCs are the other precursor involved in ozone formation. Area sources and
mobile sources produced about 51% and 30% of the total VOC emissions, respectively
(Figure 4) [3].
NOx - Tennessee
16.00%
43.00%
Point
Mobile
Area
41.00%
Figure.3 1999 Annual NOx Emissions from Major Sources in Tennessee
4
VOCs - Tennessee
19.00%
Point
Mobile
51.00%
Area
30.00%
Figure.4 1999 Annual VOC emissions from Major sources in Tennessee
Base Case
A rolling 8-hr average ozone concentration was computed from hourly concentrations
and the maximum predicted concentration each day was compared to the maximum 8hour average monitored concentrations. The predicted maximum 8-hour ozone
concentration was chosen from the hourly maximum 8-hour average ozone
concentrations predicted in a grid matrix of 7 cells by 7 cells, which encompasses the
monitoring grid at the center (Refer Figure 5).
x
Center Grid
(Monitor Location)
Figure.5 7 x 7 grid matrix around the Center grid.
In order for the CMAQ to utilize emission data, the emission inventory was spatially
gridded, temporary resolved and chemically speciated by MCNC's Sparse Matrix
Operator Kernel Emissions (SMOKE) modeling system [6]. It generates hourly
precursor’s emissions needed by CMAQ for 36km, 12km and 4km grid and the emissions
were speciated for the Carbon Bond-IV (CB-IV) chemical mechanism [7]. Once the
individual emissions were processed, all the emissions were merged to a single output file
called merged output. The merged outputs were used for analyzing the distribution of
NOx and VOCs emission and used as an input to CMAQ to predict the ozone
concentration [6]. The Carbon Bond-IV Mechanism is intended to simulate the formation
of ozone from its precursor’s (NOx and VOC).
5
Scenarios
Based on each scenario, emissions from point and mobile sources were reduced for the
whole modeling domain. In each case mentioned below, the inventory was modified to
reflect the corresponding changes. Biogenic source emissions and area source emissions
were held fixed for all modeling runs while point and mobile emissions were changed to
reflect the changes that could possibly arise for various emission reduction scenarios.
The nine control strategies that were considered in this study is given below:









Run1: Base case using average emission in 11 layers (Base case)
Run2: Both NOx and VOC emissions from mobile sources were reduced by 100%
Run3: Both NOx and VOC emissions from point sources were reduced by 100%
Run4: NOx emissions from mobile sources were reduced by 15%
Run5: NOx emissions from mobile sources were reduced by 30%
Run6: NOx emissions from mobile sources were reduced by 100%
Run7: NOx emissions from point sources were reduced by 15%
Run8: NOx emissions from point sources were reduced by 30%
Run9: VOC emissions from mobile sources were reduced by 100%
Review of Current Studies on CMAQ Performance and Sensitivity Analyses
P. Georgopoulos et al. (1999) presented studies on evaluation of the performances of a
CMAQ in predicting ambient ozone concentrations over the Northeastern U.S. The study
revealed that the correlation between the simulated and observed values is slightly better
for fine grid compared to the coarse grid simulations. The study also summarized that the
performance of CMAQ in predicting ozone with both fine and coarse grids is generally
consistent with the EPA’s recommendation of MNGE (Mean Normalized Gross Error)
and MNB (Mean Normalized Bias) for ozone predictions of 35% and 15%, respectively.
The study stated that the model predictions appeared to be generally consistent with
observations, even though the predicted values for the lowest percentile seem to have a
minimum of about 20ppb while the observed values were close to zero. The author
attributes this to be caused by the fact that CMAQ simulations represent averages over
expanded areas and cannot capture localized maxima [8].
C. Hogrefea, et al presented a similar paper based on the evaluation results of the
modeling system used to simulate ozone air quality over the eastern United States. It was
mentioned that the comparison of observed and predicted spatial patterns of daily
maximum ozone concentrations showed the best performance in predicting patterns for
average and above-average ozone concentrations. He concluded that the MM5/CMAQ
system is a suitable tool for the simulation of summertime surface temperature and ozone
air quality conditions over the eastern United States [9].
Jinyou Liang et al conducted a study based on the comprehensive field monitoring
campaign of the 2000 Central California Ozone Study (CCOS). In which CMAQ and
CAMx (Comprehensive Air-quality Model with extensions) models were employed to
simulate the ozone concentration for a July 31- August 2, 2000 episode. Jinyou Liang et
al determined that the domain-averaged surface ozone was higher in CAMx than CMAQ
6
at all hours and surface NOX values in CMAQ were less during the daytime and higher
during the nighttime than CAMx. The results suggest that CMAQ predicted higher peak
ozone in areas influenced by forest fires and under predicted ozone in the Bay Area but
performed better than CAMx in Sacramento and the San Joaquin Valley. Studies showed
that ozone precursors (NOX, NMHCs, and HCHO) were under predicted overall but still
CMAQ met the U.S. EPA model performance criteria for ozone in the Bay Area [10].
Results and Analysis for East Tennessee
To compare the model's output to the actual measured value at monitors, both these
values were plotted for each day, on the same graph for each location as shown in Figures
6 - 13. From the first impression of these graphs, the predictions appeared to be in
reasonably good agreement with observation data for all locations. The modeled ozone
concentrations were significantly lower than observations for the first part of episode but
captured the basic trend of real time data during the 11 day episode at Rutledge Pike
(Knox County), Knoxville Mildred Drive (Knox County), Clingmans Dome (Sevier
County), Cove Mountain (Sevier County) and Look Rock (Blount County). Anderson
County and Jefferson County (Figure 6 and 9) didn’t capture the same trend as the
monitoring station but the overall mean ozone concentrations predicted by the model was
very similar to the mean observed ozone concentrations (Refer Table 2). For the stations
at Anderson and Jefferson County the average bias (-2.8% & 5.3% respectively) was
much lower than the other counties. Knox County results were better than for Blount and
Sevier County (Refer Table 2). For Anderson, Jefferson and Knox County, the second
part of the episode from 6th to 8th of September the CMAQ model captured the rising
trend with observed data and over predicted the ozone concentration (Refer Figure 6, 9 to
11).
Table 2 Overall mean concentrations – Observed vs. Predicted
Predicted Max 8- Observed Max 8hr average
hr average at
ppm
Average
Station
nearby grids
monitor site
difference
Bias, %
(ppm)
(ppm)
Anderson
0.069
0.073
-0.003
-2.78
Blount Cades Cove
0.071
0.078
-0.007
-7.11
Blount Look Rock
0.071
0.092
-0.021
-22.01
Jefferson
Knox- Rutledge
Pike
Knox- Middle
Drive
Sevier- Cove
Mountain
Sevier- Clingmans
Dome
0.066
0.065
0.001
5.33
0.065
0.074
-0.009
-10.57
0.068
0.076
-0.008
-9.67
0.070
0.092
-0.022
-23.42
0.066
0.087
-0.021
-23.30
7
Predicted
Observed
Daily Max 8-hr Average - Observed and Predicted
Anderson County
Freels Bend Area Melton Lake, Site ID: 47-001-0101
Ozone Concentration (ppm)
0.120
0.100
0.080
0.060
0.040
0.020
0.000
8/28/1999
8/30/1999
9/1/1999
9/3/1999
9/5/1999
9/7/1999
9/9/1999
Date
Figure.6 Anderson County – Max 8-hr Average – Observed vs. Predicted
Predicted
Observed
Daily Max 8-hr Average - Observed and Predicted
Blount County
GSMNP Cades Cove, Site ID: 47-009-0102
0.100
Ozone Concentration (ppm)
0.090
0.080
0.070
0.060
0.050
0.040
0.030
0.020
0.010
0.000
8/28/1999
8/30/1999
9/1/1999
9/3/1999
9/5/1999
9/7/1999
9/9/1999
Date
Figure.7 Blount County (Cades Cove) – Max 8-hr Average – Observed vs. Predicted
8
Predicted
Observed
Daily Max 8-hr Average - Observed and Predicted
Blount County
GSMNP Look Rock, Site ID: 47-009-0101
Ozone Concentration (ppm)
0.120
0.100
0.080
0.060
0.040
0.020
0.000
8/28/1999
8/30/1999
9/1/1999
9/3/1999
9/5/1999
9/7/1999
9/9/1999
Date
Figure.8 Blount County (Look Rook) – Max 8-hr Average – Observed vs. Predicted
Predicted
Observed
Daily Max 8-hr Average - Observed and Predicted
Jefferson County
Lost Creek Road, Site ID: 47-089-0002
Ozone Concentration (ppm)
0.120
0.100
0.080
0.060
0.040
0.020
0.000
8/28/1999
8/30/1999
9/1/1999
9/3/1999
9/5/1999
9/7/1999
9/9/1999
Date
Figure.9 Jefferson County – Max 8-hr Average – Observed vs. Predicted
9
Predicted
Observed
Daily Max 8-hr Average - Observed and Predicted
Knox County
Rutledge Pike, Mascot TN, Site ID: 47-093-0021
Ozone Concentration (ppm)
0.120
0.100
0.080
0.060
0.040
0.020
0.000
8/28/1999
8/30/1999
9/1/1999
9/3/1999
9/5/1999
9/7/1999
9/9/1999
Date
Figure.10 Knox County (Rutledge Pike) – Max 8-hr Average – Observed vs. Predicted
Predicted
Observed
Daily Max 8-hr Average - Observed and Predicted
Knox County
Knoxville Mildred Drive, Site ID: 47-093-1020
Ozone Concentration (ppm)
0.120
0.100
0.080
0.060
0.040
0.020
0.000
8/28/1999
8/30/1999
9/1/1999
9/3/1999
9/5/1999
9/7/1999
9/9/1999
Date
Figure.11 Knox County (Middle Drive) – Max 8-hr Average – Observed vs. Predicted
10
Predicted
Observed
Daily Max 8-hr Average - Observed and Predicted
Sevier County
GSMNP Cove Mountain, Site ID: 47-155-0101
Ozone Concentration (ppm)
0.120
0.100
0.080
0.060
0.040
0.020
0.000
8/28/1999
8/30/1999
9/1/1999
9/3/1999
9/5/1999
9/7/1999
9/9/1999
Date
Figure.12 Sevier County (Cove Mountain) – Max 8-hr Average – Observed vs. Predicted
Predicted
Observed
Daily Max 8-hr Average - Observed and Predicted
Sevier County
GSMNP Clingmans Dome, Site ID: 47-155-0102
Ozone Concentration (ppm)
0.120
0.100
0.080
0.060
0.040
0.020
0.000
8/28/1999
8/30/1999
9/1/1999
9/3/1999
9/5/1999
9/7/1999
9/9/1999
Date
Figure.13 Sevier County (Clingmans Dome) – Max 8-hr Average – Observed vs.
Predicted
The CMAQ model over predicted the average 8-hour ozone concentration only at Lost
Creek Road (Jefferson County) by 1 ppb (5.3%). For the monitoring stations at Knox and
Anderson County, the model under predicted the average 8-hour ozone from 3 to 9 ppb
11
(Refer Table 2 and Figure 14). The model under predicted the average 8-hour maximum
ozone concentration close to 10% at Knox County and by 3.44% at Anderson County.
Generally CMAQ performed better at Anderson, Jefferson and Knox County compared to
the monitors in Blount and Sevier Counties. There Blount and Sevier County monitors
were located within the Great Smoky Mountains National Park at high elevations of
2700ft to 6610 ft above sea level. Anderson, Knox and Jefferson County monitors were at
relatively low elevations (<1000 ft) on the valley floor. High difference were found
between the average observed and modeled concentrations for both the stations in Sevier
County: the maximum difference of 22 ppb was observed at Sevier County when the
average concentration predicted by the model was only 70 ppb while the average
observed concentration was 92 ppb. Similar differences were observed at Look Rock
(Blount County). The above result suggests that the modeled performed poorer at rural,
high elevation locations.
For the episode considered, Figure 14 shows the average bias of each monitor. Modeled
values underpredict the observed values by 7.11% on average for the station at Cades
Cove, which is at 1850ft above mean sea level. For the elevated locations (Sevier and
Blount County) the model under predicted the observed values by nearly 20% on average
though the predictions appeared to follow the same day-to-day trend as the observations.
Averaged Bias for Monitoring Station
10.00%
5.00%
661
0ft
Sev
ie r
-01
01
@
415
0ft
ie r
-01
02
@
Sev
ox102
0@
Kn
Kn
ox002
1@
980
105
6ft
ft
ft
101
7
@
Jef
fer
s on
Rk
Lo
ok
Blo
unt
Cv
@
Ca
de s
unt
Blo
-10.00%
@
185
0f t
780
ft
@
der
son
An
Percentage
-5.00%
270
0ft
0.00%
-15.00%
-20.00%
-25.00%
Stations
Figure.14 Average Biases for predicted values at monitors
Comparisons to EPA’s Bias Limit
A Bias test was conducted to determine if model predictions were within a desired
<±20% performance goal. To pass the performance test, modeled daily maximum 8-hr
ozone concentrations should be less than ±20% of observed values at the monitors.
Biases were calculated for each day at all locations. The results of Anderson and
Jefferson County are shown in Figure 15 and 16. These graphs illustrate that for most of
12
the days in episode, base case predictions fell within the EPA’s prescribed bias limit of
<±20%. Most of the days the bias falls out of the bias limit in Blount and Sevier County
and again showing that the model under predicts at high elevation ozone monitoring sites.
Bias of Daily Maximum 8-hr Average - Site Specific
Anderson County
Freels Bend Area Melton Lake, Site ID: 47-001-0101
40
30
20
Bias %
10
0
8/28/1999
8/30/1999
9/1/1999
9/3/1999
9/5/1999
9/7/1999
9/9/1999
-10
-20
-30
-40
Date
Figure.15 Anderson County – Bias of Daily Max 8-hr Average
Bias of Daily Maximum 8-hr Average - Site Specific
Jefferson County
Lost Creek Road, Site ID: 47-089-0002
60
40
Bias %
20
0
8/28/1999
8/30/1999
9/1/1999
9/3/1999
-20
-40
-60
-80
Date
13
9/5/1999
9/7/1999
9/9/1999
Figure.16 Jefferson County – Bias of Daily Max 8-hr Average
Results of Various Emissions Reduction Scenarios
For the various sensitivity analyses conducted on mobile and point sources at the
monitoring locations, the following results were obtained as outlined below:



Is an ozone formation in the region “NOx Limited or VOCs Limited”? If ozone
formation is limited by NOx emissions, reducing the emissions of VOCs may
have little or no effect in reducing the level of ozone. Greater reduction in ozone
concentrations were observed in Table 3 and 4 for 100% reduction in mobile NOx
emissions and modest reductions in ozone were observed for 100% mobile VOCs
reduction. For the same percent reduction in ozone precursors, NOx yielded more
ozone reduction than the VOCs. Furthermore 100% reduction in mobile VOCs
emissions yields almost the same reduction in ozone concentration as 30%
reduction in mobile NOx emissions. Therefore mobile source NOx emissions
would remain the target pollutant to be controlled in East Tennessee based on a
percent basis reduction.
Reducing mobile source NOx and VOC by one ton yields 0.14ppb and 0.12ppb of
ozone reduction for per ton of reduction respectively. This result indicates that
both NOx and VOCs are effective in controlling the ozone formation in East
Tennessee.
Table 4 shows for Knox County, the average percent reduction of ozone
concentration for a 100% reduction in point source NOx & VOC emissions
(12.31% and 13.24%) was closer to the 100% reduction in mobile source (NOx &
VOC) emissions (12.31% and 10.29%). This indicates that point source reduction
was as effective as mobile source reduction in Knox County.

Table 3 shows that in Jefferson County the average percent reduction of ozone
concentration for a 100% reduction in point source NOx emissions (12.12%) was
close to the 100% reduction in mobile source NOx emissions (13.64%). This
indicates that point source emission reduction was as effective as mobile source
emission reduction for Jefferson County.

For rest of the counties (especially rural, high elevation sites), mobile source NOx
emission reductions were more effective than point source emission reductions in
lowering ozone levels. As shown in Table 4, a 100% reduction in mobile source
NOx emissions yielded an 18% reduction in ozone at Blount and Sevier County
sites, while a 100% reduction in point source NOx yielded less than 13%
reduction in ozone.
The reduction in ozone with reduction in mobile source NOx emissions was not
necessary linear. For the episode considered, Figure 17 and 18 show the overall
response of ozone reduction for variable NOx emission reduction from mobile
source at Anderson and Knox County. For the Knox County site, a 15% and 30 %
reduction in mobile NOx emissions gave roughly the same result.

14
Location
Anderson
Blount Cades Cove
Blount Look Rook
Jefferson
Knox -0021
Knox -1020
Sevier-0101
Sevier- 0102
Average
Tons/day in Tennessee
(36km domain)
100%
100%
30%
15%
reduction in
reduction in reduction in reduction in
Mobile NOx
Mobile NOx Mobile NOx Mobile NOx
and VOCs
emissions
emissions
emissions
emission
12
11
4
3
14
13
6
5
13
12
6
5
10
9
3
3
8
7
3
3
9
7
3
3
14
13
6
5
12
12
5
4
12
11
5
4
100%
reduction in
Mobile VOCs
emissions
792.12
4
6
6
4
4
4
6
5
5
100%
reduction in
Point NOx
and VOCs
emissions
9
9
9
8
8
7
9
8
8
427.5
Reduction of 792.12 tons of mobile NOx/day
Reduction of 1 ton of mobile NOx/day
Reduction of 427.5 tons of mobile VOC/day
Reduction of 1 ton of mobile VOC/day
Yields
11
0.014
5
0.012
ppb of ozone reduction
ppb of ozone reduction
ppb of ozone reduction
ppb of ozone reduction
Table.3 Summary of ozone reduction (in ppb) from base for various scenarios
15
30%
reduction in
Point NOx
emissions
15%
reduction in
Point NOx
emissions
1
3
3
3
3
2
3
2
3
1
2
2
2
2
2
2
1
2
Location
Anderson
Blount Cades Cove
Blount Look Rook
Jefferson
Knox -0021
Knox -1020
Sevier-0101
Sevier- 0102
100%
reduction in
Mobile NOx
and VOCs
emission
17.39%
19.72%
18.31%
15.15%
12.31%
13.24%
20.00%
18.18%
100%
30%
15%
reduction in reduction in reduction in
Mobile NOx Mobile NOx Mobile NOx
emissions
emissions
emissions
15.94%
18.31%
16.90%
13.64%
10.77%
10.29%
18.57%
18.18%
5.80%
8.45%
8.45%
4.55%
4.62%
4.41%
8.57%
7.58%
4.35%
7.04%
7.04%
4.55%
4.62%
4.41%
7.14%
6.06%
100%
reduction in
Mobile VOCs
emissions
5.80%
8.45%
8.45%
6.06%
6.15%
5.88%
8.57%
7.58%
100%
reduction in
Point NOx
and VOCs
emissions
13.04%
12.68%
12.68%
12.12%
12.31%
10.29%
12.86%
12.12%
30%
reduction in
Point NOx
emissions
15%
reduction in
Point NOx
emissions
1.45%
4.23%
4.23%
4.55%
4.62%
2.94%
4.29%
3.03%
1.45%
2.82%
2.82%
3.03%
3.08%
2.94%
2.86%
1.52%
Table.4 Summary of percent reduction in ozone concentration from base case for various scenarios
16
Overall Response of Ozone for various percent reduction in Mobile NOx Emissions at Anderson
18
16
15.94
% reduction in Ozone
14
12
10
8
6
5.8
4.35
4
2
0
0
0
20
40
60
80
100
120
% reduction in Mobile NOx
Figure.17 Overall Response of Ozone for various percent reductions in Mobile NOx
Emission at Anderson
Overall Response of Ozone for various percent reduction in Mobile NOx Emissions at Knox-0021
12
10.77
% of Ozone Reduction
10
8
6
4.62
4.62
4
2
0
0
0
20
40
60
80
100
120
% of Mobile NOx Reduction
Figure.18 Overall Response of Ozone for various percent reductions in Mobile NOx
Emission at Knox – 0021
17
Conclusion:
 The CMAQ model captured the general trend for the maximum 8 -hour average
ozone concentrations in East Tennessee.
 Based on the ppb difference between average predicted and average observed
concentrations and average bias values, the model performed overall better at the
Anderson, Jefferson and Knox County monitors where the model predictions
were only from 3 to 9 ppb lower than the measured values.
 Based on the ppb difference between average predicted and average observed
concentrations and average bias values the model performance was poor at
elevated locations like the Sevier County and Blount County monitors where the
model under predicted measured concentrations by 20 to 22 ppb.
 For both Jefferson and Knox County, a 15% reduction in mobile NOx emissions
yields the same reduction as a 30% reduction in mobile NOx (i.e. 3 ppb ozone
reduced) See Table 3 and 4. A 100% reduction in mobile source NOx emissions
yields a 7 to 9 ppb reduction in ozone. Reducing both NOx and VOC’s by 100%
yields an ozone reduction of 8 to 10 ppb. Reducing only mobile source VOC’s by
100% reduces ozone by only 4 ppb.
 For most locations model is more sensitive to reductions in mobile source NOx
emissions than point source NOx emissions. A 15% reduction in mobile NOx
emissions reduced ozone by an average of 4 ppb, while a 15% reduction in point
source NOx lowered ozone only 2 ppb.
 The current ozone design value for East Tennessee is 89 ppb. It will require a 5%
reduction in maximum 8-hour ozone concentrations to achieve the NAAQS of 85
ppb. The CMAQ modeling indicates that a 20% to 30% reduction in NOx
emissions may allow the NAAQS to be achieved.
References
1. Carolina Environmental Program, (CEP), Community Multiscale Air Quality,
http://www.cmascenter.org, (accessed January 18, 2005).
2. The East Tennessee Economic Development Agency, http://www.eteda.org
(accessed January 1, 2005).
3. Prakash Doraiswamy, “Modeling and Source Apportionment of Primary and
Secondary PM2.5 in the Atmosphere”, PhD Dissertation, July 2004, The
University of Tennessee, Knoxville.
4. Tennessee Department of Environment and Conservation (TDEC),
http://www.state.tn.us/environment/ (accessed January 18, 2005).
5. Tennessee Department of Environmental and Conservation (TDEC), Ozone
Monitors in Tennessee, http://www.state.tn.us/environment/apc/ozone/monitors/
(accessed October 4,2004).
6. Carolina Environmental Program, (CEP), Sparse Matrix Operator Kernel
Emissions (SMOKE) Modeling System, Emission Processing Tool,
http://www.cep.unc.edu/empd/products/smoke/index.shtml (accessed October 4,
2004).
18
7. Atmospheric Chemistry International Research Site for Information and
Technology Exchange (AIRSITE), http://airsite.unc.edu/soft/cb4/CB4docs.html /
(accessed August 18, 2005).
8. P. Georgopoulos, Q. Sun and A. Chandrasekar, “Evaluation of Grid-Based Ozone
and PM Modeling for a 1999 Summer Episode”, Journal of the Air & Waste
Management Association, (accessed October 20, 2004).
9. C.Hogrefe, J. Biswas, K.Civerolo and B.Lynn, “Climate Change and Ozone Air
Quality: Applications of a Coupled GCM/MM5/CMAQ Modeling System”, 2003
Models-3 Users’ Workshop, (accessed October 20, 2004).
10. Jinyou Liang *, Philip T. Martien, Su-Tzai Soong, and Saffet Tanrikulu, “A
Photochemical model comparison study: CAMx and CMAQ performance in
Central California”, 13th Conference on the Applications of Air Pollution
Meteorolgy with the Air and Waste Management Association.
19