On the Verification of Particulate Matter Simulated by the NOAA-EPA Air Quality Forecast System

1
On the Verification of Particulate Matter
Simulated by the NOAA-EPA Air Quality Forecast
System

• Ho-Chun Huang1, Pius Lee1, Binbin Zhou1, Jian
  Zeng6,
• Marina Tsidulko1, You-Hua Tang1, Jeff McQueen3,
  Qiang Zhao7,
• Shobha Kondragunta2, Rohit Mathur4, Jon Pleim4,
  George Pouliot4,
• Geoff DiMego3, Ken Schere4, and Paula Davidson5
• 1 Scientific Applications International
  Corporation, Camp Springs, Maryland.
• 2 NOAA/NESDIS Center for Satellite Applications
  and Research, Camp Springs, Maryland.
• 3 National Centers for Environmental Prediction,
  Camp Springs, Maryland.
• 4 National Oceanic and Atmospheric
  Administration, Research Triangle Park, NC. (On
  assignment to the National Exposure Research
  Laboratory, US EPA)
• 5 Office of Science and Technology, National
  Weather Service, Silver Spring, MD.
• 6 Earth Resources Technology Inc., Annapolis
  Junction, MD.
• 7I.M. Systems Group, Inc., Rockville, MD.

2
Outline

• NOAA-EPA Air Quality Forecast System
• GOES and AQF atmospheric optical depth (AOD)
• NCEP verification results
• Summary

3
Air Quality Forecast System

• CONUS (ozone) became operational model on
  September 18, 2007
• Developmental model operational PM Chemistry
• CMAQ v4.5 driven by the WRF/NMM at 12 km
• NEI (2001), BEIS3, Mobile 6
• AERO3 Aerosol Module with SOA (no sea salt)
• Updated ISORROPIA for numerical stability at low
  relative humidity
• Euler Backward Iterative (EBI) Solver for CB4
• Minimum Kz to mimic urban island

4
AOD Comparisons

• In-site measurement (AERONET, AIRS) (Marina
  Tsidulko)
• Satellite measurement GOES product comparisons
  with AERONET and MODIS (Prados et al, 2007)
• (AERO) good for AOD gt 0.15, Negative bias for AOD
  gt 0.35
• (MODIS) good agreement and correlation of high
  AOD
• CMAQ AOD comparison with IMPROVE, MODIS, and
  AERONET in the eastern US (Roy et al, 2007)
• good spatial and temporal patterns
• CMAQ AOD is often less than MODIS AOD for the
  same concentration

5
The NCEP/EMC Real-timeAOD Verification

• AQF AOD The column integration of extinction (s)
  due to particulate scattering and absorption and
  layer thicknesses (?Zi)
• AQF AOD vs. GOES AOD
• Frequency Daily (April to September 2007)
• Data hourly from 1215 2115 UTC
• Domains CONUS, East US, and West US

6
The GOES Derived AOD (Prados et al, 2007)
Visible
Infrared
7
AQF modeling and verification domains
8
Null GOES AOD
mean over the period Total 66.6 Cloud
41.8 White Noise 24.8
9
mean over the period Total 55.0 Cloud
46.4 White Noise 8.6
mean over the period Total 78.3 Cloud
34.8 White Noise 43.5
10
The NCEP/EMC Real-timeAOD Verification

• Thresholds
• lt0.1, gt0.1, gt0.2, gt0.3, gt0.4, gt0.5, gt0.6, gt0.8,
  gt1.0, gt1.5, and gt 2.0
• Skill Scores
• Critical Success Index (Threat Score CSI)
• Probability of Detection (POD)
• False Alarm Ratio (FAR)
• _of_Fcst / _of_Obsv (BIAS)
• Equitable Threat Score (ETS)
• Accuracy rate
• Type of figures
• Daily average time series (per month)
• Daily average by threshold
• Monthly average by threshold

11
http//www.emc.ncep.noaa.gov/mmb/aq/
12
http//www.emc.ncep.noaa.gov/mmb/hchuang/web/html/
score_mon.html
13
(No Transcript)
14
Bias F/O (ab)/(ac) CSI H/(FO-H)
a/(abc) POD H/O a/(ac) False Alarm ratio
1-H/F b/(ab) Accuracy rate (N-F-O2H)/N
(ad)/(abcd)
15
Bias number of points
lt 0.1
gt 0.4
gt 0.1
gt 0.5
gt 0.2
gt 0.6
gt 0.3
gt 0.8
16
Probability of Detection
lt 0.1
gt 0.4
gt 0.1
gt 0.5
gt 0.2
gt 0.6
gt 0.3
gt 0.8
17
AQF does not account foradditional particulate
sources?

• Inventory wild fire emissions, not real-time data
• Sea Salt
• Long range transport of dust, aerosol, and
  chemical species across modeling boundary

18
(No Transcript)
19
X
X
20
Pearson Correlation Coefficientbetween the AQF
skill score (CSI) and the number of Null GOES
data due to Cloud

• TOTAL DAYS 183
• CSI CLUD CONUS gt 0.1 NUM 167 r -0.3165 r2
  0.1002 t -4.2852
• CSI CLUD CONUS gt 0.2 NUM 167 r -0.2978 r2
  0.0887 t -4.0075
• CSI CLUD CONUS gt 0.3 NUM 167 r -0.2595 r2
  0.0673 t -3.4512
• CSI CLUD E US gt 0.1 NUM 167 r -0.3580 r2
  0.1282 t -4.9254
• CSI CLUD E US gt 0.2 NUM 167 r -0.2774 r2
  0.0769 t -3.7087
• CSI CLUD E US gt 0.3 NUM 167 r -0.2462 r2
  0.0606 t -3.2630
• CSI CLUD W US gt 0.1 NUM 167 r -0.0520 r2
  0.0027 t -0.6686
• CSI CLUD W US gt 0.2 NUM 167 r 0.1309 r2
  0.0171 t 1.6958
• CSI CLUD W US gt 0.3 NUM 167 r 0.1286 r2
  0.0165 t 1.6661

21
August 2-5, 2007
22
August 15, 2007
23
SUMMARY

• Good spatial AQF PM distribution with low bias on
  the AOD ? unresolved PM sources or processes
• It is difficult to access the AQF PM skill in the
  western US due to strong surface reflectivity
• Negative correlation (in the eastern US) between
  the AQF PM skill score and null satellite AOD
  because of cloud (clear sky ? better skill score)
  was observed
• Further investigation is needed to understand the
  (non-linear) relationship between cloudiness
  and AQF PM skill, as well as the processes that
  impact AQF PM skill

24
http//www.emc.ncep.noaa.gov/mmb/aq/