Comparisons of Global Retrieved Emissivity for Version 3.1 and Version 3.5

1. Sea Surface EmissivityRetrieval Product EvaluationDenise HaganAIRS Science Team MeetingGreenbelt, Maryland March 31, 2004

2. Outline of Results:Comparisons of Global Retrieved Emissivity for Version 3.1 (similar version at DAAC) and Version 3.5 (latest version)Comparisons of V3.5 Retrieved Emissivity and sensitivity test Comparisons of SST from Regression and Final Retrievals with test resultsSummary and Retrieval Issues

3. Comparisons of Global Retrieved Emissivity for Version 3.1 and Version 3.5

4. Version 3.1 Daytime: Surface Emissivity at 943 cm-1 Emissivity should be 0.99 (0-40 deg view angle) and 0.98 (>40 deg view angle).

5. Version 3.5: Daytime, September 6, 2002 Surface Emissivity at 943 cm-1 Emissivities closer to expected based on model

6. Version 3.5: Daytime September 6, 2002 Ratio of Model and Retrieved Surface Emissivity at 943 cm-1 Retrieved emissivity is less than model by about 1-2% for scan angles >40o

7. Comparisons of V3.5 Retrieved Emissivity and sensitivity test Masuda/Wu and Smith Emissivity model enters retrieval at Start (with climatology guess parameters)and after first Cloud Clearing and Regressionsteps; emissivity error estimates at two steps

8. Version 3.5: Daytime September 6, 2002 Surface Emissivity at 943 cm-1 No scan angle model dependence Emissivity transitions smoothly from low to high scan angles; bias at large scan angles reduced

9. Ratio of Model and Retrieved Surface Emissivity at 943 cm-1 No scan angle model dependence Ratio of -1 Retreival similar to model between 50 N and 50 S, except for regions affected by dust

10. Comparisons of SST from Regression and Final Retrievals with test results

11. Version 3.5: SST Differences Kelvin (AIRS - NCEP) Regression -1

12. Version 3.5: SST Differences Kelvin (AIRS - NCEP) Final Product 1

13. Version 3.5 SST Differences Kelvin (AIRS - NCEP) No Scan Angle Model Dependence Regression 1

14. Version 3.5 SST Differences Kelvin (AIRS - NCEP) No Scan Angle Model Dependence Final Product

15. Version 3.5 SST Differences Kelvin (AIRS - NCEP) Final Product Night

16. Version 3.5 SST Differences Kelvin (AIRS - NCEP) Final Product Night No Scan Angle Dependence

17. (Summary continued)SST outliers observed at large scan angles, and near cloud boundariesMagnitude of SST outliers at scan angles >40 degrees about 2K (for global average)Emissivity of 0.98 @943 cm-1 is equivalent to 1.2 K tropics and 2K mid-latitude for clear skyDownwelling contribution @943 cm-1 equivalent to 0.1 K (small scan angle, clear) to 0.4 K (large scan angle, cloudy) in tropics Next>>>

18. Summary:Retrieval Version 3.5 produces more realistic ocean surface emissivity than V3.1 (eg, version currently resident at DAAC)Emissivity decrease with increasing view angle is greater than model (eg, 2-3% decrease; according to model should be about 1-2% for view angles >40 deg and global mean wind speed 5 ms-1) Next>>>

19. (Summary continued)SST outliers reduced by removing scan angle dependence of emissivity in retrieval algorithm, for both day and nightRetrieved emissivity (without scan angle model dependence) show effects of dust aerosol that are realistic.More SST outliers in final product versus intermediate regression product at large view angles and near cloud boundaries

20. The accuracy goal for AIRS SST is 0 K bias and 0.5 K rms. What changes to retrieval algorithm should be made to reach this goal?Recent results in literature suggest that commonly accepted emissivity model (in AIRS algorithm) may not be correct at low winds, at large scan angles, for different temperatures, with assumption ofLambertian reflectivity.

21. Literature Survey:Friedman (1969) Refractive Index for fresh and salt waterMasuda et al (1988) Extension of Friedman with Cox and Munk (1952); wind + angle dependence of emissivityWu and Smith (1997) full spectral range; MAERI validationWatts et al (1996) ATSR; dropped wind speed dependenceNalli et al (2001) Surface Reflection from Cox-Munk surface Henderson (2003) Monte Carlo model for emissivity/reflectivityHanafin (TBD) modified wind dependenceNewman (TBD) modified temperature dependence

22. Example of emissivity departure from Masuda/Wu and Smith: MAERI data at different wind speeds (55o view angle) compared to model. Emissivity decreases with decreasing wind speed, opposite to model….

23. Different approaches for obtaining more accurate SST:improve solar scattering componentimprove downwelling IR reflectance componentinvestigate amplification of scan angle error in cloud clearing steps use wavelength-dependent only emissivity for forward radiance calcs remove shortwave channels from daytime retrievals

24. Proposed Session: Talking points1) review current state of knowledge on ocean emissivity models and the AIRS ocean emissivity model, including infrared and microwave2) observations versus models (what compares well; what are the problem areas. Are results from in situ instrument (small fov) directly applicable to satellite (large footprint)3) the reflected terms in the forward model; what improvements can be made4) impact of emissivity/reflectivity errors on the accuracy of retrieved gas quantities (as well as the standard temperature products)5) land emissivity model(s)6) understanding effects of using different emissivity models in forward algorithms to derive land versus ocean retrievals