1. CALCULATING SEA SURFACE TEMPERATURE, EMISSIVITY AND ATMOSPHERIC STATE USING HYPERSPECTRAL RADIANCES J. Le Marshall, J. Jung, W. L. Smith, E. Maturi, J. Derber, Xu Li, R. Treadon, S. Lord, M. Goldberg and W. Wolf

2. Overview JCSDA – Background/Challenge/SST activity Hyperspectral Data Assimilation Hyperspectral emissivity/SST Plans/Future Prospects Summary

3. JCSDA Partners Pending

4. JCSDA Mission and Vision Mission: Accelerate and improve the quantitative use of research and operational satellite data in weather. ocean, climate and environmental analysis and prediction models Vision: A weather, ocean, climate and environmental analysis and prediction community empowered to effectively assimilate increasing amounts of advanced satellite observations and to effectively use the integrated observations of the GEOSS

5. The Challenge Satellite Systems/Global Measurements Aqua Terra TRMM SORCE SeaWiFS Aura Meteor/ SAGE GRACE ICESat Cloudsat Jason CALIPSO GIFTS TOPEX Landsat NOAA/ POES GOES-R WindSAT NPP COSMIC/GPS SSMIS NPOESS MSG

6. 5-Order Magnitude Increase in s atellite Data Over 10 Years Count (Millions) Daily Upper Air Observation Count Year Satellite Instruments by Platform Count NPOESS METEOP NOAA Windsat GOES DMSP 19902010 Year

7. JCSDA Instrument Database – June 2006

10. Satellite Data used in NWP HIRS sounder radiances AMSU-A sounder radiances AMSU-B sounder radiances GOES sounder radiances GOES, Meteosat, GMS winds GOES precipitation rate SSM/I precipitation rates TRMM precipitation rates SSM/I ocean surface wind speeds ERS-2 ocean surface wind vectors Quikscat ocean surface wind vectors AVHRR SST AVHRR vegetation fraction AVHRR surface type Multi-satellite snow cover Multi-satellite sea ice SBUV/2 ozone profile and total ozone Altimeter sea level observations (ocean data assimilation) AIRS MODIS Winds … >32 instruments

11. Sounding data used operationally within the GMAO/NCEP Global Forecast System AIRS HIRS sounder radiances AMSU-A sounder radiances MSU AMSU-B sounder radiances GOES sounder radiances SBUV/2 ozone profile and total ozone On 14 - on 15 - off 16 - off 17 - on 15 - on 16 - on 17 - off 18 - on AQUA 14 - on 15 - on 16 - on 17 - on 10 - on 12 - on 16 - on 17 - on

12. CURRENT SATELLITE DATA - STATUS AIRS v1.Implemented AIRS v2.Completed Operational Trial - NCO MODIS WindsImplemented NOAA-18 AMSU-AImplemented NOAA-18 MHSCompleted Operational Trial - NCO NOAA-17 SBUV Total OzoneImplemented NOAA-17 SBUV Ozone ProfileImplemented SSM/I RadiancesGSI impl. ( prod. Used in SSI) COSMIC/CHAMPRT Assim. in GSI SSMISRT Assim. in GSI MODIS Winds v2.RT Testing WINDSATRT Assim in GSI AMSR/E – Radiance AssimilationRT Assim IN GSI AIRS/MODIS Sounding Channels Assim.ASSIM. Trial GOES – VIS and SW WindsTo be Tested GOES Hourly WindsTo be Tested GOES 11 and 12 Clear Sky Rad. Assim(6.7µm)To be Tested MTSAT 1R Wind Assim.Assim Testing AURA OMIAssim trial TOPEX,JASON1,ERS-2 ENVISAT ALTIMETERTest and Development, Ops 06 GODAS FY – 2CCDW testing Underway Note: ADM – OSSEs Completed ~ 9 new instruments

13. Major Accomplishments Common assimilation infrastructure at NOAA and NASA Community radiative transfer model Common NOAA/NASA land data assimilation system Interfaces between JCSDA models and external researchers Snow/sea ice emissivity model – permits 300% increase in sounding data usage over high latitudes – improved polar forecasts MODIS winds, polar regions, - improved forecasts - Implemented AIRS radiances assimilated – improved forecasts - Implemented Improved physically based SST analysis - Implemented Preparation for advanced satellite data such as METOP (IASI,AMSU,MHS…),, NPP (CrIS, ATMS….), NPOESS, GOES-R data underway. Advanced satellite data systems such as DMSP (SSMIS), CHAMP GPS, COSMIC GPS, Windsat tested for implementation. Impact studies of POES AMSU, HIRS, EOS AIRS/MODIS, DMSP SSMIS, Windsat, CHAMP GPS on NWP through EMC parallel experiments active Data denial experiments completed for major data base components in support of system optimisation OSSE studies completed Strategic plans of all Partners include 4D-VAR

14. William L. Smith, R.O. Knuteson, H.E. Revercomb, W. Feltz, H. B. Howell, W. P. Menzel, N. R. Nalli, Otis Brown, Peter Minnett and Walter McKeown. 1996: Observations of the Infrared Radiative Properties of the Ocean – Implications for the Measurement of Sea Surface Temperature via Satellite Remote Sensing. Bull. Amer. Meteor. Soc. 77, 41 – 51. Nalli, N.R., 1995. Sea surface skin temperature retrieval using the high resolution interferimeter sounder (HIS). M.S. Thesis, Dept. of Atmospheric and Oceanic Sciences, University of Wisconsin – Madison, 117 pp. …. George Aumann et al. 2006: …....... Hyperspectral/AIRS based SSTs

15. USE OF AIRS HYPERSPECTRAL RADIANCES

16. Development and Implementation Progress of Community Radiative Transfer Model (CRTM) P. van Delst, Q. Liu, F. Weng, Y. Chen, D. Groff, B. Yan, N. Nalli, R. Treadon, J. Derber and Y. Han …..

17. Community Contributions Community Research: Radiative transfer science  AER. Inc: Optimal Spectral Sampling (OSS) Method  NRL – Improving Microwave Emissivity Model (MEM) in deserts  NOAA/ETL – Fully polarmetric surface models and microwave radiative transfer model  UCLA – Delta 4 stream vector radiative transfer model  UMBC – aerosol scattering  UWisc – Successive Order of Iteration  CIRA/CU – SHDOMPPDA  UMBC SARTA  Princeton Univ – snow emissivity model improvement  NESDIS/ORA – Snow, sea ice, microwave land emissivity models, vector discrete ordinate radiative transfer (VDISORT), advanced double/adding (ADA), ocean polarimetric, scattering models for all wavelengths Core team (JCSDA - ORA/EMC): Smooth transition from research to operation  Maintenance of CRTM (OPTRAN/OSS coeff., Emissivity upgrade)  CRTM interface  Benchmark tests for model selection  Integration of new science into CRTM

18. Progress CRTM v.0 used in NCEP SSI CRTM v.1 has been integrated into the GSI at NCEP/EMC (Dec. 2005) Beta version CRTM has been released to the public CRTM with OSS (Optimal Spectral Sampling) has been established and is being evaluated and improved.

19. COMMUNITY RADIATIVE TRANSFER MODEL CRTM Below are some of the instruments for which we currently have transmittance coefficients. abi_gr (gr == GOES-R) airs_aqua amsre_aqua amsua_aqua amsua_n15 amsua_n16 amsua_n17 amsua_n18 amsub_n15 amsub_n16 amsub_n17 avhrr2_n10 avhrr2_n11 avhrr2_n12 avhrr2_n14 avhrr3_n15 avhrr3_n16 avhrr3_n17 avhrr3_n18 hirs2_n10 hirs2_n11 hirs2_n12 hirs2_n14 hirs3_n15 hirs3_n16 hirs3_n17 hirs3_n18 hsb_aqua imgr_g08 imgr_g09 imgr_g10 imgr_g11 imgr_g12 mhs_n18 modisD01_aqua (D01 == detector 1, D02 == detector 2, etc) modisD01_terra modisD02_aqua modisD02_terra modisD03_aqua modisD03_terra modisD04_aqua modisD04_terra modisD05_aqua modisD05_terra modisD06_aqua modisD06_terra modisD07_aqua modisD07_terra modisD08_aqua modisD08_terra modisD09_aqua modisD09_terra modisD10_aqua modisD10_terra modis_aqua (detector average) modis_terra (detector average) msu_n14 sndr_g08 sndr_g09 sndr_g10 sndr_g11 sndr_g12 ssmi_f13 ssmi_f14 ssmi_f15 ssmis_f16 ssmt2_f14 vissrDetA_gms5 windsat_coriolis

20. OPTRAN-V7 vs. OSS for AIRS channels OSS OPTRAN CRTM IMPROVED COMMUNITY RADIATIVE TRANSFER MODEL

21. Hyperspectral Data Assimilation AQUA

22. AIRS Data Assimilation J. Le Marshall, J. Jung, J. Derber, R. Treadon, S.J. Lord, M. Goldberg, W. Wolf and H-S Liu, J. Joiner, and J Woollen…… 1 January 2004 – 31 January 2004 Used operational GFS system as Control Used Operational GFS system Plus AIRS as Experimental System

23. Table 1: Satellite data used operationally within the NCEP Global Forecast System HIRS sounder radiances AMSU-A sounder radiances AMSU-B sounder radiances GOES sounder radiances GOES 9,10,12, Meteosat atmospheric motion vectors GOES precipitation rate SSM/I ocean surface wind speeds SSM/I precipitation rates TRMM precipitation rates ERS-2 ocean surface wind vectors Quikscat ocean surface wind vectors AVHRR SST AVHRR vegetation fraction AVHRR surface type Multi-satellite snow cover Multi-satellite sea ice SBUV/2 ozone profile and total ozone

24. AMS 2006 - Future National Operational Environmental Satellites SymposiumRisk Reduction for NPOESS Using Heritage Sensors 24 Improved NCEP SST Analysis Xu Li, John Derber EMC/NCEP  Progress  SST physical retrieval code has been merged into GSI and provided to NCEP marine branch for operational use  An extensive diagnostic study on the diurnal variation signals in in situ and satellite observations, SST retrievals, SST analysis and associated air-sea fluxes (NCEP GFS product) shows the SST diurnal variation needs to be addressed to improve the SST analysis product.  7-day 6-hourly SST analysis has been produced with GSI, after a new analysis variable, in situ and AVHRR data were introduced into GSI.  Plan  Analyze SST by assimilating satellite radiances directly with GSI  Active ocean in the GFS  Aerosol effects

25. Physical/Variational SST Retrieval Formulation Cost Function: is brightness temperature (radiance), skin temperature, atmospheric temperature vertical profile and atmospheric water vapor vertical profile respectively. is calculated with radiative transfer model. is the sensitivity of to respectively. Initially, the and are assumed not varying with height (z). Therefore, The sum of these sensitivities with height is used in the scheme for AVHRR data. Upper-subscription represents analysis, first guess and observation respectively. Lower-subscription means the channel index. is the error variance of and respectively The solutions of are solved by minimizing cost function J

26. AMS 2006 - Future National Operational Environmental Satellites SymposiumRisk Reduction for NPOESS Using Heritage Sensors 26 Improved NCEP SST Analysis Xu Li, John Derber EMC/NCEP

27. Global Forecast System Background Operational SSI (3DVAR) version used Operational GFS T254L64 with reductions in resolution at 84 (T170L42) and 180 (T126L28) hours. 2.5hr cut off

28. The Trial Used `full AIRS data stream used (JPL)  NESDIS (ORA) generated BUFR files  All FOVs, 324(281) channels  1 Jan – 15 Feb ’04 Similar assimilation methodology to that used for operations Operational data cut-offs used Additional cloud handling added to 3D Var. Data thinning to ensure satisfying operational time constraints

29. The Trial AIRS related weights/noise optimised Used NCEP Operational verification scheme.

30. AIRS Assimilation Used 251 Out of 281 Channels - 73 - 86 Removed (Channels peak too High) - 1937 - 2109 Removed (Non LTE) - 2357 Removed (Large Obs – Background Diff.) Used Shortwave at Night  Wavenumber > 2000 cm -1 Downweighted  Wavenumber > 2400cm -1 Removed

31. AIRS data coverage at 06 UTC on 31 January 2004. (Obs-Calc. Brightness Temperatures at 661.8 cm -1 are shown)

32. Figure 5.Spectral locations for 324 AIRS selected channel data distributed to NWP centers.

33. Table 2: AIRS Data Usage per Six Hourly Analysis Cycle Data Category Number of AIRS Channels Total Data Input to Analysis Data Selected for Possible Use Data Used in 3D VAR Analysis(Clear Radiances) ~200x10 6 radiances (channels) ~2.1x10 6 radiances (channels) ~0.85x10 6 radiances (channels)

34. Figure1(a). 1000hPa Anomaly Correlations for the GFS with (Ops.+AIRS) and without (Ops.) AIRS data, Southern hemisphere, January 2004

35. Figure 1(b). 500hPa Z Anomaly Correlations for the GFS with (Ops.+AIRS) and without (Ops.) AIRS data, Southern hemisphere, January 2004

36. Figure 2. 500hPa Z Anomaly Correlations 5 Day Forecast for the GFS with (Ops.+AIRS) and without (Ops.) AIRS data, Southern hemisphere, (1-27) January 2004

37. Figure3(a). 1000hPa Anomaly Correlations for the GFS with (Ops.+AIRS) and without (Ops.) AIRS data, Northern hemisphere, January 2004

38. Figure 3(b). 500hPa Z Anomaly Correlations for the GFS with (Ops.+AIRS) and without (Ops.) AIRS data, Northern hemisphere, January 2004

39. AIRS Data Assimilation J. Le Marshall, J. Jung, J. Derber, R. Treadon, S.J. Lord, M. Goldberg, W. Wolf and H-S Liu, J. Joiner and J Woollen January 2004 Used operational GFS system as Control Used Operational GFS system Plus AIRS as Experimental System Clear Positive Impact Both Hemispheres.Implemented -2005

40. AIRS Data Assimilation MOISTURE Forecast Impact evaluates which forecast (with or without AIRS) is closer to the analysis valid at the same time. Impact = 100* [Err(Cntl) – Err(AIRS)]/Err(Cntl) Where the first term on the right is the error in the Cntl forecast. The second term is the error in the AIRS forecast. Dividing by the error in the control forecast and multiplying by 100 normalizes the results and provides a percent improvement/degradation. A positive Forecast Impact means the forecast is better with AIRS included. Error in AIRS fcst Error in control fcst Error in AIRS fcst Error in control fcst Error in AIRS fcst Error in control fcst Error in AIRS fcst Error in control fcst

42. AIRS Data Assimilation Impact of Data density... 10 August – 20 September 2004

44. AIRS Data Assimilation Impact of Spectral density... 10 August – 20 September 2004

46. AIRS Data Assimilation AIRS in the GSI... 1 January – 15 February 2004

47. AIRS – GSI, v2, … GSI Contral 1/18 fovs v all fov AIRS

48. AIRS Data Assimilation Application of AIRS Radiances over land,water and ice

49. Surface Emissivity (ε) Estimation Methods IRSSE Model Geographic Look Up Tables (LUTs) - CRTM Regression based on theoretical estimates Minimum Variance, provides T surf and ε * Eigenvector technique Variational Minimisation – goal

50. Regression IR HYPERSPECTRAL EMISSIVITY - ICE and SNOW Sample Max/Min Mean computed from synthetic radiance sample From Lihang Zhou Emissivity Wavenumber

51. JCSDA IR Sea Surface Emissivity Model (IRSSE) Initial NCEP IRSSE Model based on Masuda et al. (1998) Updated to calculate Sea Surface Emissivities via Wu and Smith (1997) Van Delst and Wu (2000) Includes high spectral resolution (for instruments such as AIRS) Includes sea surface reflection for larger angles JCSDA Infrared Sea Surface Emissivity Model – Paul Van Delst Proceedings of the 13th International TOVS Study Conference Ste. Adele, Canada, 29 October - 4 November 2003 Surface Emissivity (ε) Estimation Methods

57. Minimum Variance IR HYPERSPECTRAL EMISSIVITY - Water

59. AIRS SST Determination Use AIRS bias corrected radiances from GSI AIRS channels used are : 119 – 129 (11) 154 – 167 (14) 263 – 281 (19) Method is the minimum (emissivity) variance technique Channels used in Pairs : 119, 120; 120, 121; 121, 122;.. etc

60. where I ν, ε ν, B ν, T S, τ ν (z 1, z 2 ), Z and T(z) are observed spectral radiance, spectral emissivity, spectral Planck function, the surface temperature, spectral transmittance at wavenumber ν from altitude z 1 to z 2, sensor altitude z, and air temperature at altitutide z respectively. For a downward looking infrared sensor:

61. The solution can be written as : Where R OBS is the observed upwelling radiance, N↑ represents the upwelling emission from the atmosphere only and N↓ represents the downwelling flux at the surface. The ^ symbol denotes the “effective” quantities as defined in Knuteson et al. (2003).

62. The SST is the T S that minimises : i=1,43

65. Summary: The introduction of AIRS hyperspectral data into environmental prognosis centers has provided improvements in forecast skill. Here we have noted initial results where AIRS hyperspectral data, used within stringent operational constraints, have shown significant positive impact in forecast skill over both the Northern and Southern Hemisphere for January 2004. We have also noted the improvement gained from using AIRS at a spatial density greater than that used generally for operational NWP.

66. Summary: The modeling of surface emissivity in the CRTM and in a number of related studies have also commenced to improve our use of AIRS data over land, water and ice. Initial estimates of emissivity and skin SST based on hyperspectral satellite observations in the IR indicate significant potential for further improving our current estimate of operational skin temperature.

67. Conclusion Given the opportunities for enhancement of the assimilation system and the resolution of the hyperspectral data base, the results here indicate an opportunity to further improve current analysis and forecast systems through the application of hyperspectral data. i.e. further improvements are expected through use of higher spectral and spatial resolution data. Further improvements may also be anticipated through use of data over land, cloudy data and the use of complementary data such as Moderate Resolution Imaging Spectroradiometer (MODIS) radiances to better characterize the AIRS fovs. (Note- all channel and AIRS/MODIS BUFR )

68. The business of looking down is looking up