Global Precipitation Measurement GPM mission Precipitation Processing System
Global Precipitation Measurement (GPM) mission Precipitation Processing System (PPS) Synthetic GMI Level 1 Data Yimin Ji Yimin. ji-1@nasa. gov NASA/GSFC, Code 610. 2, Room S 048 A Presentation Date: October 5, 2010 Yimin Ji - Page 1 October 5, 2010
Synthetic GMI Level 1 Data Outline: • Objective and Requirements • Observations and Data Match-up • CRTM and Tb Transformation • Data Format • Conversion from Tb to Ta and Count • Synthetic Data Processing • Summary Yimin Ji - Page 2 October 5, 2010
Objective Synthetic Data Processing is a part of assurance process for GPM mission success. Overall goal is to guarantee that algorithms and processing systems are correctly implemented at satellite launch. • Generate global database of synthetic coincident brightness temperatures for 13 GMI channels. • Derive antenna temperature, earth view count, science packet from synthetic GMI Tb and calibration counts. • Generate synthetic GMI Level 1 products (geo-located and calibrated Ta, Tb, Count) following file specifications. • Adopt mission simulation to generate GMI orbital Tb file. • Test GMI L 1 to L 3 algorithms before satellite launch and monitor GMI Tb trend after satellite launch. Yimin Ji - Page 3 October 5, 2010
Requirements 1. Data format and data structure meet the requirements of GMI Base and L 1 B file specifications. 2. Realistic GMI brightness temperatures (Tb) and geo-locations for all GMI channels. 3. Tb fields reflect various cloud and precipitation types. 4. Relatively large database to generate monthly data products. 5. Using both existing observation and radiative transfer modeling. Yimin Ji - Page 4 October 5, 2010
Observations (GMI vs. Existing Satellite Sensors) The first step is to match-up instantaneous low frequency pixels (TMI/AMSRE) with high frequency pixels (SSMIS/AMSUB) in time and space. GMI TMI AMSRE 10. 65 V 10. 65 H 18. 7 V 18. 7 H 23. 8 V 36. 5 H 89. 0 V 89. 0 H 165. 5 V 165. 5 H 183 ± 3 183 ± 7 low low high high 10. 65 V 10. 65 H 19. 35 V 19. 35 H 21. 3 V 37. 0 H 85. 5 V 85. 5 H low low high 10. 65 V 10. 65 H 18. 7 V 18. 7 H 23. 8 V 36. 5 H 89. 0 V 89. 0 H low low high 19. 35 V 19. 35 H 22. 23 37. 0 91. 66 150 low V H H 183 ± 7 low V 183 ± 3 low low high high SSMIS AMSUB GMI Ch # Yimin Ji - Page 5 1 2 3 4 5 6 7 89. 0 V 150. 0 V 183 ± 3 183 ± 7 low low 12 13 8 9 10 11 October 5, 2010
Observations (Data Match-up) 1. Matching-up instantaneous TMI and AMSUB(SSMIS) measurements at lower latitude (35 o S-35 o N ). 2. Matching-up instantaneous AMSRE and AMSUB/SSMIS measurements at higher latitude. 3. Spatial resolution: 0. 1 degree in latitude and longitude. Spatial coverage: 0 -360 o longitude, 90 o S-90 o N latitude. 4. Time window: 10 minutes for TMI-AMSUB/SSMIS match-up at lower latitude 35 o S-35 o N. 30 minutes for AMSRE-AMSUB/SSMIS match-up at high latitude > 45 o S/N 1. 90 minutes for AMSRE-AMSUB/SSMIS match-up at other areas. Yimin Ji - Page 6 October 5, 2010
Tb Transformation 1 . Radiative transfer model is needed to transform the observed Tb into synthetic GMI Tb due to the offsets of center frequency and band width. 2. The NOAA Community Radiative Transfer Model (CRTM) was used to generate Tb for all channels of above instruments. 3. Look-up tables were established to transform Tb from one instrument to another for similar center frequencies. 4. For 10 GHz to 89 GHz channels, look-up tables based on AMSRE and TMI match-ups were generated and compared to the simulations. 5. Level 2 products were used to categorize the status of pixels (rain/no rain) Yimin Ji - Page 7 October 5, 2010
Tb Transformation (CRTM Simulation) A large number of atmospheric profiles were used to simulate Tb of existing TMI, AMSRE, AMSUB, SSMIS as well as the future GMI under clear sky and cloud sky conditions. The following figures show TMI Tb offset (GMI-TMI) against the GMI Tb as a function of GMI Tb for 37 GHz and 89 GHz channels under cloud and clear sky conditions. In both conditions, the GMI 36. 5 GHz Tb is normally lower than TMI 37 GHz Tb. The GMI 89 GHz Tb is normally higher than TMI 85 GHz Tb for normal range of Tb. Yimin Ji - Page 8 October 5, 2010
Tb Transformation (CRTM Simulation) TMI Tb offset against the GMI Tb as a function of GMI Tb for 19 GHz and 23 GHz channels under cloud and clear sky conditions. The GMI 18. 7 GHz Tb is normally lower than 19. 35 GHz Tb. The GMI 23. 8 GHz Tb is normally higher than TMI 21. 3 GHz Tb. Yimin Ji - Page 9 October 5, 2010
Tb Transformation (CRTM Simulation) TMI Tb offset against the GMI Tb as a function of GMI Tb for 183 3 GHz and 183 7 GHz channels under cloud and clear sky conditions. For 183 GHz channels, AMSUB and GMI have similar center frequency, but different band width. The differences are small under clear sky conditions. The GMI Tb is higher than the TMI Tb under cloud conditions for both channels. Yimin Ji - Page 10 October 5, 2010
Tb Transformation (AMSRE/TMI Match-up) The CRTM simulation for rain pixels is under developing. Since GMI low frequency channels have exact center frequencies and similar band widths and IFOVs as compared to some of the AMSRE channels, the TMI-GMI Tb transformation can use coincident AMSRE and TMI Tb data. Following figures show offsets of TMI Tb against AMSRE Tb (AMSRE Tb – TMI Tb) for 37 GHz and 89 GHz channels. Yimin Ji - Page 11 October 5, 2010
Tb Transformation (AMSRE/TMI Match-up) Offsets of TMI Tb against AMSRE Tb (AMSRE Tb – TMI Tb) for 19 GHz and 23 GHz channels. Yimin Ji - Page 12 October 5, 2010
GMI L 1 B Synthetic Data Flow Diagram Yimin Ji - Page 13 October 5, 2010
Database of Synthetic GMI Tb of 13 channels and Incidence Angles for June 15, 2008 Yimin Ji - Page 14 October 5, 2010
Database of Synthetic GMI Rain Retrieval for June 15 2008 Yimin Ji - Page 15 October 5, 2010
GMIBASE File Specification (Metadata and S 1 Swath) John Stout at PPS generated file specifications for GMI Base file, GMI L 1 B, and GMI L 1 C files. Here only shows format of GMI Base file. The data are in HDF 5 format. S 1 Swath. Header Metadata S 1 Scan. Time 19 bytes Group: nscan S 1 Latitude 4 bytes Array: npix 1 x nscan S 1 Longitude 4 bytes Array: npix 1 x nscan S 1 scan. Status 17 bytes Group: nscan S 1 navigation 84 bytes Group: nscan File. Header Metadata S 1 calibration 264 bytes Group: nscan Input. Record Metadata S 1 cal. Counts 552 bytes Group: nscan Navigation. Record Metadata S 1 sun. Data 40 bytes Group: nscan File. Info Metadata S 1 incidence. Angle 4 bytes Array: npix 1 x nscan Gprof. Info Metadata S 1 sat. Azimuth. Angle 4 bytes Array: npix 1 x nscan S 1 solar. Zen. Angle 4 bytes Array: npix 1 x nscan S 2 S 1 solar. Azimuth. Angle 4 bytes Array: npix 1 x nscan S 1 sun. Glint. Angle 1 byte Array: npix 1 x nscan S 1 swath represents 9 low frequency channels (10 -90 GHz) Yimin Ji - Page 16 S 1 earth. View. Counts 2 bytes Array: nchan 1 x npix 1 x nscan S 1 Ta 4 bytes Array: nchan 1 x npix 1 x nscan October 5, 2010
GMIBASE File Specification ( S 2 Swath) S 2 swath represents 4 high frequency channels (166 - 183 GHz). Channels of S 2 group have different incidence angle and calibration line than those of S 1 group. Latitude/Longitude as well as calibration are slightly different from those of S 1 group. S 2 Swath. Header Metadata S 2 Scan. Time 19 bytes Group: nscan S 2 Latitude 4 bytes Array: npix 2 x nscan S 2 Longitude 4 bytes Array: npix 2 x nscan S 2 scan. Status 17 bytes Group: nscan S 2 navigation 84 bytes Group: nscan S 2 calibration 124 bytes Group: nscan S 2 cal. Counts 312 bytes Group: nscan S 2 sun. Data 40 bytes Group: nscan S 2 incidence. Angle 4 bytes Array: npix 2 x nscan S 2 sat. Azimuth. Angle 4 bytes Array: npix 2 x nscan S 2 solar. Zen. Angle 4 bytes Array: npix 2 x nscan S 2 solar. Azimuth. Angle 4 bytes Array: npix 2 x nscan S 2 sun. Glint. Angle 1 byte Array: npix 2 x nscan S 2 earth. View. Counts 2 bytes Array: nchan 2 x npix 2 x nscan S 2 Ta 4 bytes Array: nchan 2 x npix 2 x nscan Yimin Ji - Page 17 October 5, 2010
Synthetic orbital GMI L 1 B of 12 Channels for June 15, 2008 Yimin Ji - Page 18 October 5, 2010
Conversion from Tb to Ta and Earth Count Tb -> Ta Ta - > Earth View Count (Vev) Tb=C х Ta + D х Ta* + E Ta = Th + A (Vev – Vh) Tb*=C* х Ta* + D* х Ta + E* A=(Th-Tc)/(Vh-Vc) C, C*, D, D*, E, E* are known coefficient for sensors. Th: Hot load temperature (240 - 320 K). Ta*, and Tb* are Ta and Tb of cross polarization channels or zero if there is no cross polarization channels. Substitute Ta* as a function of Tb* and Ta Ta=C’ х Tb + D’ х Tb* + E’ Where C’, D’, E’ are derived from known coefficients C, C*, D, D*, E, E*. C’=C*/(CC*-DD*), D’=D/(CC*-DD*) E’=(C*E-DE*)/(CC*-DD*) Yimin Ji - Page 19 Tc: Cold Sky temperature (2. 7 K or 3. 2 K). Vh: hot load count of each channel at Th. Vc: cold count of each channel at Tc. Vev = Vh – (Th – Ta)/A = Vh – (Th – Ta)*(Vh – Vc)/(Th – Tc) Vev, Vh, Vc, Th, Tc can be used to generate synthetic L 1 A data once the packet formats are finalized. We will test linear L 1 B algorithm first and then go to the required non-linear algorithm. October 5, 2010
Synthetic Data Processing Yimin Ji - Page 20 October 5, 2010
Summary To ensure all Level 1 algorithms and software running properly, PPS will test them using synthetic data and mission simulation before satellite launch. The software to generate synthetic GMI Tb, Ta, and Count are ready. Samples of global GMI synthetic data are available now. The future plans include: 1. Generate synthetic orbital data set for GMIBASE, GMIL 1 B, GMIL 1 C following the new file specifications for at least a month. Higher level algorithms may be tested with synthetic L 1 data sets. 2. Generate synthetic L 0 data and test GMIBASE, GMIL 1 B, GMIL 1 C algorithm codes. Test all higher level algorithms using both synthetic data and mission simulations. 3. Develop software to monitor the short term and long term trends and test the software using synthetic data. Yimin Ji - Page 21 October 5, 2010
- Slides: 21