Operational Air Quality and Source
Contribution Forecasting in Georgia
Yongtao Hu1, M. Talat Odman1, Michael E. Chang2 and
Armistead G. Russell1
1School
2Brook
of Civil & Environmental Engineering,
Byers Institute of Sustainable Systems
Georgia Institute of Technology
10th Annual CMAS Conference, October 24th, 2010
Georgia Institute of Technology
Outline
Local Air Quality Forecasting in Georgia
The Hi-Res air quality forecasting system
Evolution of Hi-Res during 2006-2011
Met performance and O3 and PM2.5 performance for metro Atlanta
New SOA module and its impact on PM2.5 performance
Pilot Source Contribution Forecasting
Georgia Institute of Technology
Local Forecasting: How it works in Georgia
A forecasting panel with 5-7 local expert members
•air quality researchers, modelers, meteorologists, etc.
“Voting” through a decision making online system
•the consensus of the panel will be the “ensemble” forecasts
Base their “opinion” on multiple available resources
•NOAA O3 guidance, Statistical model results, Weather forecasts
•Hi-Res O3 and PM2.5 forecasts
•Experiences
Broadcast to the public
•Website: http://www.gaepd.org/air/smogforecast/ and AirNOW
•Smog alert through highway traffic information system
Georgia Institute of Technology
Hi-Res: forecasting ozone and PM2.5 at a 4-km
resolution for metro areas in Georgia
Hi-Res Modeling Domains
Hi-Res Air Quality Forecasting System
Serving Metro-Atlanta Area since 2006
Emissions
Meteorology
Emission
Inventory
NAM 84-hr
Forecast
SMOKE
WRF
Forecast
Product
CMAQ
4-km (123x123)
12-km (123x138)
Air Quality
Hi-Res Cycle
Cycle 1
00Z
66Z
15Z
ramp up
Cycle 2
36-12- & 4-km forecasts
27Z
12Z
ramp up
Operation
Forecast
Forecast
Forecast Day
Day 22
Forecast Day
Day 11
7pm 0am
R&
R&S
R& 1 F1&R1
36-km (148x112)
S2
78Z
7
Forecast
Time (UTC)
36-12- & 4-km forecasts
Computer
Forecast Day 2
Forecast Day 1
0am
Time (EST)
0am
0am
&S
F2 R2&S
Georgia Institute of Technology
Hi-Res Forecast Products
“Single Value” Report: tomorrow’s AQI, ozone and PM2.5 by metro area in Georgia
Air Quality Forecasts: AQI, ozone and PM2.5, 48-hrs spatial plots and station profiles
Meteorological Forecasts: precipitation, temperature and winds, 48-hrs spatial plots and
station profiles
Performance Evaluation: time series comparison and scatter plots for the previous day
Snapshots from Hi-Res homepage: http://forecast.ce.gatech.edu
Georgia Institute of Technology
Evolution of Hi-Res during 2006-2011
Updated to latest release of WRF each year before the ozone season.
•WRF 2.1, 2.2, 3.0, 3.1, 3.2 and 3.3
CMAQ is typically one version behind.
•CMAQ 4.6 with Georgia Tech extensions
Projected NEI to current year in the very beginning of each year.
Updated forecast products website each year before ozone season.
Switched from single-cycle forecasting to two-cycles in 2008.
Enlarged 4-km domain to cover the entire state of Georgia in 2009.
Introduced Georgia Tech’s new SOA module in 2009.
Enlarged 36-km domain to cover the CONUS and enlarged 12-km
domain to cover the eastern US in 2011.
In 2011 switched landuse data from old USGS data to new MODIS
data in WRF, to reflect recent changes in land cover.
Enlarged Modeling Domains
Current
2006-2010
4-km (99x78)
Current 4-km (123x123)
12-km (72x72)
4-km (123x123)
12-km (123x138)
36-km (148x112)
36-km (72x72)
Georgia Institute of Technology
Ambient Monitoring Sites for Performance
Evaluation in Atlanta Metro
Kennesaw
Gwinnet
NE Atlanta
Yorkville
Atlanta
West Atlanta Confederate Ave
Douglasville
South DeKalb
Atlanta
Conyers
Fayetteville
Newnan
Walton
McDonough
Peachtree City
SLAMS O3
SLAMS PM2.5
NWS Met
Met Performance
Overall 2006-2011 Performance (Ozone Season):
Atlanta Metro
Humidity
Temperature
315
20
18
16
Observed Humidity at 2m (g/kg)
Observed Daily Hi Temp at 2m (K)
310
305
300
295
14
12
10
8
6
4
2
290
290
0
295
300
305
310
315
0
2
4
6
8
10
12
14
Forecast Humidity at 2m (h/kg)
Forecast Daily Hi Temp at 2m (K)
MB
-0.39K
MB
-0.68g/kg
ME
1.55K
ME
1.19g/kg
Georgia Institute of Technology
16
18
20
Forecast vs. Observed Temperature
2007
315
310
310
Daily Hi Temp 2m (K)
Daily Hi Temp at 2m (K)
2006
315
305
300
MB
0.50K
300
295
295
ME
290
5/1/2006
305
5/31/2006
6/30/2006
1.57K
7/30/2006
8/29/2006
290
5/1/2007
9/28/2006
5/21/2007
6/10/2007 6/30/2007
7/20/2007
8/9/2007
8/29/2007 9/18/2007
310
Daily Hi Temp at 2m (K)
310
Daily Hi Temp at 2m(K)
1.90K
315
315
305
300
MB
-0.80K
305
300
MB
295
295
ME
5/21/2008
6/10/2008
6/30/2008
7/20/2008
1.94K
8/9/2008
8/29/2008
9/18/2008
290
5/1/2009
0.37K
ME
5/21/2009
6/10/2009
6/30/2009
Date
1.47K
7/20/2009
8/9/2009
8/29/2009
9/18/2009
Date
2011
2010
315
315
310
Daily Hi Temp at 2m (K)
310
Daily Hi Temp at 2m (K)
ME
2009
2008
305
300
295
290
05/01/10
-1.16K
Date
Date
290
5/1/2008
MB
05/21/10
06/10/10
MB
-0.19K
ME
1.15K
06/30/10
07/20/10
Date
08/09/10
305
300
295
290
MB
-0.05K
ME
1.29K
Georgia Institute of Technology
08/29/10
09/18/10
285
05/01/11
05/21/11
06/10/11
06/30/11
07/20/11
Date
08/09/11
08/29/11
09/18/11
Forecast vs. Observed Humidity
2006
2007
20
20
18
Daily Avg Humidity at 2m (g/kg)
Daily Avg Humidity at 2m (g/kg)
18
16
14
12
10
8
MB
6
4
ME
2
-0.74g/kg
1.39g/kg
5/21/2006 6/10/2006 6/30/2006 7/20/2006
8/9/2006
14
12
10
8
MB
6
4
2
0
5/1/2006
16
0
5/1/2007
8/29/2006 9/18/2006
ME
5/21/2007 6/10/2007
Daily Avg Humidity at 2m (g/kg)
18
16
14
12
10
ME
2
0
5/1/2008
5/21/2008
6/10/2008
6/30/2008
7/20/2008
-0.89g/kg
1.33g/kg
8/9/2008
8/29/2008
14
12
10
8
6
2
0
5/1/2009
9/18/2008
MB
-0.36g/kg
ME
0.96g/kg
4
5/21/2009
6/10/2009
6/30/2009
18
16
16
Daily Avg Humidity at 2m (g/kg)
20
18
14
12
10
8
4
ME
2
0
05/01/10
05/21/10
06/10/10
06/30/10
07/20/10
Date
8/9/2009
8/29/2009
9/18/2009
08/29/11
09/18/11
2011
20
MB
7/20/2009
Date
2010
6
9/18/2007
16
Date
Daily Avg Humidity at 2m (g/kg)
Daily Avg Humidity at 2m (g/kg)
20
18
4
8/29/2007
2009
20
MB
8/9/2007
Date
2008
6
1.22g/kg
6/30/2007 7/20/2007
Date
8
-0.59g/kg
-0.73g/kg
14
12
10
8
6
4
1.03g/kg
Georgia Institute of Technology
2
08/09/10
08/29/10
09/18/10
0
05/01/11
05/21/11
MB
-0.73g/kg
ME
1.22g/kg
06/10/11
06/30/11
07/20/11
Date
08/09/11
Air Quality Performance Metrics
Forecast
False Alarms
Hits
NAAQS
NAAQS
Correct
Nonevents
Missed
Exceedences
Observation
MNB 
1
N
N

k 1
ck  ck
m
o
c
o
k
MNE 
1
N
N
ck  ck
k 1
ck

m
o
o
Overall 2006-2011 Performance (Ozone Season):
Atlanta Metro
PM2.5
Ozone
150
70.0
146
0
0
4-km
4-km
134
75
35.0
792
58
525
52
0
0.0
150
75
0
0
Obs.
35
70
Obs.
MNB
17%
MNB
-17%
MNE
23%
MNE
32%
Georgia Institute of Technology
Ozone Performance
Forecast vs. Observed O3
2006
MNB
11%
MNE
29%
2007
120.0
100.0
120
MNB
8.5%
MNE
19%
100
80
O 3 (p p b )
O3 (ppb)
80.0
60.0
60
40.0
40
20.0
20
0.0
May-01
0
Jun-01
Jul-01
Aug-01
Obs.
M a y-0 7
Sep-01
120
100
MNB
17%
MNE
23%
2009
120
100
O3 (ppb)
O3 (ppb)
S e p -0 7
4 -k m
MNB
28%
MNE
30%
80
60
60
40
40
20
20
Jun-08
Jul-08
Aug-08
Obs
0
May-09
Sep-08
Jun-09
Jul-09
120
Aug-09
Obs
4-km
2010
MNB
14%
MNE
18%
120
100
100
80
80
60
Sep-09
4-km
2011
O3 (ppb)
O3 (ppb)
A ug -0 7
Obs
80
MNB
18%
MNE
21%
60
40
40
20
20
0
May-10
J ul-0 7
4-km
2008
0
May-08
J un-0 7
Georgia Institute of Technology
Jun-10
Jul-10
Aug-10
Obs
4-km
Sep-10
0
May-11
Jun-11
Jul-11
Aug-11
Obs
4-km
Sep-11
2009 O3 Performance: Hi-Res vs. GA EPD’s
Our 4-km Forecast
EPD Ensemble Forecast
150
150
7
75
75
4
108
6
7
EPD
4-km
24
5
125
0
0
0
75
150
0
75
Obs.
Obs.
MNB
28%
MNB
13%
MNE
30%
MNE
21%
150
2010 O3 Performance: Hi-Res vs. GA EPD’s
Our 4-km Forecast
EPD Ensemble Forecast
150
150
EPD
4-km
75
75
11
105
10
16
12
14
12
104
0
0
0
75
150
0
75
150
Obs.
Obs.
MNB
14%
MNB
9%
MNE
18%
MNE
17%
2011 O3 Performance: Hi-Res vs. GA EPD’s
Our 4-km Forecast
EPD Ensemble Forecast
150
150
EPD
4-km
75
75
10
76
32
13
36
26
12
91
0
0
0
75
150
0
75
150
Obs.
Obs.
MNB
18%
MNB
8%
MNE
21%
MNE
16%
PM2.5 Performance
Summer
Forecast vs. Observed PM2.5
2006
45.0
-38%
MNE
43%
2007
-37%
MNE
44%
45
40
35.0
35
30.0
30
O 3 (p p b )
3
PM 2.5 (mg/m )
40.0
MNB
MNB
25.0
20.0
15
10.0
10
5.0
5
May-01
0
Jun-01
Jul-01
Aug-01
Obs.
Sep-01
M a y-0 7
45
40
35
MNB
-38%
MNE
42%
20
45
40
10
5
0
May-09
Sep-08
Jul-09
MNE
35
Aug-09
Obs
MNB
40
4%
25
20
25%
45
40
21%
Sep-09
4-km
2011
35
PM2.5 (ug/m3)
30
MNB
-2%
MNE
25%
30
25
20
15
15
10
10
5
5
0
May-10
Jun-09
4-km
45
MNE
20
5
2010
8%
25
15
Obs
MNB
30
10
Aug-08
S e p -0 7
35
15
Jul-08
A ug -0 7
4 -k m
2009
PM2.5 (ug/m3)
25
Jun-08
J ul-0 7
Obs
30
0
May-08
J un-0 7
4-km
2008
O3 (ppb)
20
15.0
0.0
PM2.5 (ug/m3)
25
Georgia Institute of Technology
Jun-10
Jul-10
Aug-10
Obs
4-km
Sep-10
0
May-11
Jun-11
Jul-11
Aug-11
Obs
4-km
Sep-11
2009 PM2.5 Performance: 4-km vs. GA EPD’s
Our 4-km Forecast
EPD Ensemble Forecast
70
70
EPD
4-km
0
0
0
0
35
35
129
129
3
3
0
0
0
35
70
0
35
70
Obs.
Obs.
MNB
8%
MNB
11%
MNE
25%
MNE
24%
2010 PM2.5 Performance: 4-km vs. GA EPD’s
Our 4-km Forecast
EPD Ensemble Forecast
70
70
0
0
0
EPD
4-km
0
35
138
35
138
0
0
0
0
0
35
70
0
35
Obs.
Obs.
MNB
4%
MNB
14%
MNE
21%
MNE
30%
70
2011 PM2.5 Performance: 4-km vs. GA EPD’s
Our 4-km Forecast
EPD Ensemble Forecast
70
70
0
2
1
EPD
4-km
0
35
141
35
141
3
5
0
0
70
35
0
35
0
Obs.
Obs.
MNB
-2%
MNB
7%
MNE
25%
MNE
24%
70
Winter
Forecasted vs. Observed PM2.5
2008
45
45
40
40
35
35
30
30
O 3 (ppb)
O 3 (p p b )
2007
25
20
25
20
15
15
10
10
5
5
0
O ct-0 7
N o v-0 7
D e c-0 7
Ja n-0 8
Obs
F e b -0 8
0
Oct-08
M a r-0 8
Nov-08
Dec-08
4 -km
45
45
40
40
35
35
30
30
25
20
20
15
10
10
5
5
Dec-09
Mar-09
25
15
Nov-09
Feb-09
4-km
2010
PM2.5 (ug/m3)
PM2.5 (ug/m3)
2009
0
Oct-09
Jan-09
Obs
Jan-10
Obs
Feb-10
4-km
Mar-10
Apr-10
0
Oct-10
Nov-10
Dec-10
Jan-11
Obs
Feb-11
4-km
Mar-11
Apr-11
A New SOA Module (Baek, J., Georgia Tech, 2009)
Included processes:
•SOA partitioned from anthropogenic
VOCs’ oxidations (8 SVOCs)
•From monoterpenes (2 SVOCs)
•From isoprene (2 SVOCs added)
•From sesquiterpenes (1 SVOC added,
gas phase oxidation reactions added for
α-caryphyllene, β-humulene, and other
sesquiterpenes)
•Multigenerational oxidation of all
semi-volatile organic carbons (SVOCs)
added
Multigenerational Oxidation : HSVOC
SOA species in CMAQ:
•AORGAJ and AORGAI
•AORGBJ and AORGBI
•AORGBISJ and AORGBISI
•AORGBSQJ and AORGBSQI
•AORGAGJ and AORGAGI
+OH,+O3
LSVOC
+OH,+O3
Aerosol
Forecast vs. Observed OC at South DeKalb
15
forecast
14
2009
13
Ozone Season May-September
obs
12
11
Organic Carbon PM2.5
10
16
2006
2007
2008
2009
2010
9
8
7
6
5
4
12
3
1
0
5/1/2009
5/21/2009
6/10/2009
6/30/2009
7/20/2009
8/9/2009
8/29/2009
9/18/2009
Date
8
15
forecast
14
obs
2010
13
12
4
11
10
0
0
4
8
Observed (ug m-3)
12
16
Organic Carbon PM2.5
Forecast (ug m-3)
2
9
8
7
6
5
4
3
2
1
0
5/1/2010
5/21/2010
6/10/2010
6/30/2010
7/20/2010
Date
8/9/2010
8/29/2010
9/18/2010
Source Contribution Forecasting in Pilot Operation
Source contribution forecasts for ozone and PM2.5.
•Traffic, Power plants and others such as prescribed fires (needs
additional efforts).
Extra information on top of ozone and PM2.5 concentration forecasts.
•Providing quantitative information on specific source contributions
•Alerting on specific source impacts
•To help public targeting actions to prevent pollution events
Using forward sensitivity tool DDM3D in CMAQ to calculate first
order sensitivity coefficients
•Interpret such sensitivities of ozone and PM2.5 concentrations to
total emissions from specific sources as their contributions
Challenges in operational forecasting
•Computationally expensive, but doable
•Instability in calculation
Preliminary Source Contribution Forecasts:
Traffic and Power Plants Impacts
2009
2010
Summary
• 2006-2011 Temperature and humidity performance in
May-September are good.
– Daily high temperature bias is -0.39K and error is 1.55K
– Daily average humidity bias is and error is -0.68g/kg and 1.19g/kg
• 2006-2011 Ozone forecasts are good.
– Overall bias is +17% and error is 23%
• 2006-2011 PM2.5 forecasts are not very accurate.
– May-September bias is -17% and error is 32%
• The new SOA module helped much better 2009-2011
PM2.5 performances in May-September
– Bias is +8% and error is 25% for 2009
– Bias is +4% and error is 21% for 2010
– Bias is -2% and error is 25% for 2011
• Preliminary source contribution forecasting products.
Georgia Institute of Technology
Acknowledgements
We thank Georgia EPD for funding the Hi-Res forecasts,
Our former group member Dr. Jaemeen Baek for the new
SOA module, and
Dr. Carlos Cardelino of Georgia Tech for team forecasts.