Assimilation of OMI Data Into NCEP’s GFS

1. Craig Long, S. Zhou, T. Beck, A.J. Miller NOAA/NWS/NCEP/Climate Prediction Center L.Flynn NOAA/NESDIS/STAR Assimilation of OMI Data Into NCEP’s GFS

2. Outline • Background • Improvements due to OMI coverage • OMI Issues • Comparisons between SSI and GSI • How OMI data is assimilated • Thinning possibilities • Summary • What’s Next

3. Aspects of Ozone in NWP • Three aspects of dealing with ozone in NWP • Assimilation of ozone observations • Horizontal and vertical • Agreement between multiple sources • Transport of ozone once in the model • Brewer Dobson Circulation • Ozone Chemistry • Homogeneous: Production and Loss • f: Latitude, Pressure, Season • Heterogeneous: 'Ozone Hole' type depletion • Need additional observations

4. Background • Currently NCEP GFS assimilates SBUV/2 total and profile ozone measurements from both NOAA-16 and 17. • SBUV/2 provides about 90 nadir observations per orbit. • Replacement instrument is the OMPS (Ozone Mapping and Profiler Suite) • Combination of scanning mapper and limb profile • On NPP and NPOESS • Will provide higher vertical and horizontal resolution • Current additional sources of ozone data available: • Aura: OMI*, HIRDLS*, MLS, TES *NRT • MetOp: GOME2*

5. Background cont. • Why is ozone assimilated? • LW and SW radiation schemes need realistic ozone. • Used to extract correct temperature component from the ozone sensitive HIRS channels. • Biggest impacts in terms of temperature and dynamics and should occur in the UT/LS. • Won't improve short term skill (days 1-3) • But should improve days >3 • Ozone forecasts used in UV Index forecasts. • Used for boundary conditions in Air Quality forecasts.

6. OMI Comparison with GFS using SBUV/2 OMI shows finer structure than the GFS, e.g., the relatively high ozone off the East coast is captured by OMI but missed by GFS. OMI GFS

7. Adding OMI makes 5 day total ozone forecast agree more with NASA/TOMS November 11, 2005 SBUV/2 only Adding OMI TOMS obs.

8. November 12, 2005 SBUV/2 only Adding OMI TOMS obs.

9. November 13, 2005 SBUV/2 only Adding OMI TOMS obs.

10. November 14, 2005 SBUV/2 only Adding OMI TOMS obs.

11. November 15, 2005 SBUV/2 only Adding OMI TOMS obs. end

12. OMI Issues • Conflicts with SBUV/2 at high SZA • Also SBUV/2 is V6 product • Is V8 much different? Where? When? • Noise in some channels • affects TO3 at high SZA • Cloud climatology may degrade quality of TO3 • Comparisons with DOAS products • DOAS has striping • But, better estimate of cloud top heights • High density of data • 840 points per single SBUV/2 ob • Needs thinning • Comparisons with surface obs

14. Comparison between OMTO3 (NASA/TOMS) and OMDOAO3 (KNMI/DOAS)

16. OMTO3 vs OMDOAO3 Zonal Mean Total Ozone Mean may average out to near zero, but variability is quite high!

17. Striping in DOAS Total ozone makes it unusable

18. TOMS and SBUV/2 V8 Clim Cloud Tops Results in Total Ozone being too High

19. OMTO3 Cloud Top Pressure Climatology Issue 1000 500 0 1000 500 0 OMTO3 Cloud Top Pressures DOAS Cloud Top Pressures

20. If DOAS Cloud Top Pressures are used, OMTO3 Total Ozone usually is lower 208 258 208 258 OMTO3 using cloud own climatology OMTO3 using DOAS cloud top heights

21. GSI vs SSI (12Z) OMI N-17 SBUV/2 N-16 SBUV/2

22. GSI SSI SBUV/2 only (N16,N17) OMI only SBUV/2 and OMI Total Ozone increment (DU)

23. GSI – SSI differences SBUV/2 only OMI only SBUV/2 and OMI

24. Data Thinning • There are many ways to thin massive amounts of sat. obs. • Experimentation is only way to determine best density: • Sometimes “less is more” • OMI vs SBUV # of obs • 60 OMI obs/scan x 14 scans/SBUV retrieval • Or 840 points per SBUV retrieval • ~76,000 points per orbit • Need to restrict OMI to quality data points • Thin by selection • Fewer points in flat regions - more points in dynamic regions • Background errors may be adjusted to be more sensitive in dynamic regions • Thin by averaging • Uniform coverage • Average out noisy data

25. Dynamic regions Flat region

26. Dynamic ozone regions Flat region

27. Dynamic ozone regions Flat region

28. Data thinning method tested • Method: averaging data in 1o x 1o model grid box. • Selection: when there are overlapped data from multiple orbits within a 1o x 1o box, select data only from one major orbit. • Reduction: total number of data is reduced to ~ 6%.

29. (12Z) OMI N-17 SBUV/2 N-16 SBUV/2

30. 1o (lat) x 1o (lon) thinning 12Z From ~ 76,000 obs per orbit to ~ 4000

31. 1o (lat) x 2o (lon) thinning 12Z From ~ 76,000 obs per orbit to ~ 2000

32. Ozone difference of thinning and non-thinning (GSI)SBUV/2 and OMI DU

33. GSI and SSI TOZ difference (1o x 1o thinning) DU

34. GSI and SSI TOZ difference (1o x 2o thinning) DU

35. Summary • OMI adds additional information in horizontal • OMI data have issues to be rectified • Are ways to improve it! • GSI assimilation of OMI data not significantly different from SSI

36. What’s Next • Move to Aqua computer when available. • Continue experimenting with thinning options. • Quality assessment of data • Assess impacts in forecast mode. • Determine resolution dependence • Impacts to temperatures and dynamics • Strive for improvement in multi-day forecasts. • Begin looking at OMI profile products • Profile total ozone may be better than ‘best’ ozone • Additional profiles • Use March 2006 as test month • Compare profiles with ozonesonde and Lidar data. • HIRDLS data

37. fini

38. EOS AURA was launched in July 2004, which has 4 ozone measuring instruments. MLS OMI HIRDLS TES

39. Aura instruments • OMI(ozone Monitoring Instrument) • total ozone and ozone profile, high horizontal resolution • HIRDLS (High Resolution Dynamics Limb Sounder) • ozone profile, high vertical resolution (1.25 km, 10-80 km) • MLS(Microwave Limb Sounder) • ozone profile (3 km, 8-50 km) • TES (Tropospheric Emission Spectrometer) • tropospheric ozone (0-34 km)