Just how bad is the air we breathe? The Air Quality Forecasting System (AQF)

1
Just how bad is the air we breathe?The Air
Quality Forecasting System (AQF)

• AQF developed by NOAA and the EPA,
• based on NCEPs WRF model and
• EPAs CMAQ modeling system
• 5x experimental (CONUS),
• 3x operational (eastern U.S.)
• 48-h forecasts, 06Z, 12Z, 1-h and 8-h ave,
• daily max, 1-h gt 125 ppb, 8-h gt 85 ppb

2
(No Transcript)
3
MDLs Evaluation of CMAQ Ozone and Aerosol
Forecasts
Jerry Gorline, Wilson Shaffer, Michael Schenk,
Valery Dagostaro, Arthur Taylor
MDL Air Quality Verification web site
http//slosh.nws.noaa.gov/aqverifUsername
RestrictedPassword Restricted

• 3x Eta vs. 3x WRF comparison, end Eta June 20,
  2006
• Verify predictions over Houston area in Early
  June 06

November 29, 2006
4
Goals Air Quality Verification

• Compare 5x model performance on 3x domain to 3x
  model.
• Performance metrics
• H (Fraction Correct), TS, POD, FAR,
• MAE, ME/Bias.
• Use spatial maps to complement performance
  measures.
• Preliminary 1-h aerosol spatial maps and
  statistics.
• Caveat 5x Isoprene correction Aug. 4, 2006.

5
Definitions of Statistical Terms
6
Timing for 8-h 12Z Ozone Concentrations 2006
Season
Observations from Tanked BUFR files
Model Begins 12Z Day 1
Midnight Day 3
Midnight Day 2
04Z 00 16
11Z 07 47
12Z 08 24
05Z 01 17
04Z 00 40
03Z 23 39
UTC EDT EDT CMAQ Projection
24-h Obs Data 03Z 11Z Day3
24-h Obs Data 04Z 12Z Day2
The 11Z observed 8 hour average spans 03Z -
11Z. 12Z model output at projection 47 spans
projection 40 - 47.
The 12Z observed 8 hour average spans 04Z -
12Z. 12Z model output at projection 24 spans
projection 17 - 24.
24 Hr Model Data Proj 24 47
7
Timing for 8-h 06Z Ozone Concentrations 2006
Season
Observations from Tanked BUFR files
Model Begins 06Z Day 1
Midnight Day 3
Midnight Day 2

UTC EDT EDT CMAQ Projection
21Z 17 39
04Z 00 46
05Z 01 23
22Z 18 16
24-h Obs Data 20Z 04Z Day2
24-h Obs Data 21Z Day1 - 05Z Day2
The 04Z observed 8 hour average spans 20Z - 04Z,
end Day 2 06Z model output at projection 46
spans projection 39 - 46.
The 05Z observed 8 hour average spans 21Z -
05Z. (First 8-h proj with Day 2 data) 06Z model
output at projection 23 spans projection 16 - 23.
24 Hr Model Data Proj 23 46
8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32
Preliminary Aerosol Verification

• Bar charts
• Scatter plots
• MAE/bias/rmse plots
• Spatial maps, daily 1-h max
• Daily statistics

33

34

35

36

37

38

39

40

41

42

43
Comments, 2006

• Lower POD and FAR for 5x compared to 3x, lower
  5x TS during high ozone episodes.
• 5x MAE/bias lower than for 3x.
• 5x model on CONUS grid Consistent
  under-prediction, especially in California.
• 1-h aerosol results Under-prediction,
  forecasted diurnal drop not observed in summer
  but starting to show up in recent statistics.

44
Case Study Surface ozone reduction associated
with thunderstorms, 2005
45
8-h observations for four stations in band of
predicted exceedances
46
Surface observations for Tuesday, July 12, 2005,
2300 UTC