Results from GPM GPROF V 4 and Improvements

Results from GPM GPROF V 4 and Improvements Planned for V 5 Christian Kummerow, David Randel and Veljko Petkovic Colorado State University, Fort Collins, USA e-mail: kummerow@atmos. colostate. edu GPROF V 4 Validation Version 4 of the GPM radiometer algorithms were updated and delivered in April 2016. Whereas Version 3 of the algorithm used a-priori databases constructed from a number of proxies such as TRMM, Cloud. Sat, ground based radars and models, Version 4 constitutes the first version constructed from the GPM core satellite itself. GMI – Blue CMB MS – Green Ku – Red DPR MS - Orange Global GMI – Blue CMB MS – Green Ku – Red DPR MS - Orange GPROF V 5 Preparation Land The artifacts in the correlation plot have been resolved by fixing the error in the code. For land pixel, bias against MRMS is +6. 8% with correlation of 0. 58. Snow is included in this plot. GPROF 2014 Unified Algorithm Structure Model Preparation JMA forecast - NRT GANAL - Standard ECMWF - Climatology L 1 C Sensor Data Spacecraft position Pixel Location, Tbs Pixel Time, EIA Chan Freqs & Errors Ancillary Info / Datasets Surface & Emissivity Classes ECMWF / GANAL T 2 m, TPW Autosnow Snow Cover Reynolds Sea-Ice Pre. Processor (sensor specific) Standard input file Sensor Profile Database A-Priori Matched Profiles - GMI/ DPR Denotes Processes running at the SIPS GPROF Precipitation Algorithm Post-processor (Binary to HDF 5) GPROF GMI V 4 follows the Combined algorithm (CMB MS) over oceans and DPR-Ku (Ku) over land quite well. Complete HDF 5 Output file Early Implementation The first internal version of GPROF used the Combined DPR/GMI algorithm (CMB) as a basis for the a–priori database everywhere. The profiles from the combined algorithm were modified slightly to lead to better agreement with the GMI high frequency channels which were not part of the combined algorithm optimization process. Results from GMI can be seen to follow the Combined algorithm quite well but differences were notices against validation from gauges and MRMS GMI – Blue CMB MS – Green Ku – Red DPR MS - Orange GPCC Gauges MRMS Global GMI – Blue CMB MS – Green Ku – Red DPR MS - Orange Compared to the pre-launch algorithm (Version 3), Version 4 shows significantly better agreement with GPCC rain gauge accumulations. While V 3 was significantly low biased against gauges, Version 4 tends to be within 10 -20%. The missing snow not seen by Ku radar has been addressed by using ground based radar (MRMS) over the United States together with coincident satellite overpasses for 2 years to construct a new empirical database as a function of temperature and humidity. It is applied only to snow covered surfaces. Biases are reduced although overall skill of GPROF is still limited. GPROF V 5; Temp < 270 K GPROF V 4; Temp < 270 K Land GPROF/CMB GPROF/DPR-Ku Over land, a database using the DPR-Ku radar for the a-priori database showed significantly better agreement with surface observations. Based on these early result the Combined Radar/Radiometer product is used as the basis for a-priori rainfall products over oceans, while DPR-Ku is used over land. Correlations againstantaneous rain rates from MRMS show artifacts with the Version 4 product that were not there previously. The artifact is due to an error in a biasadjustment intended to eliminate the bias between computed and observed Tb in the a-priori database. Comparisons against MRMS for temperatures less than 270 K (assumed to be snow) also reveal substantial underestimation by GPROF V 4. This can be traced back to missing snow in the Ku radar product which is not sensitive to light snow. An underestimation is also seen at high latitude oceans when compared to GPCP or MERRA. The underestimation is again due to the radar’s inability to detect light drizzle known to occur in these locations MRMS Precipitation Temp < 270 K GPROF V 5 with MRMS Snow database Temp < 270 K Over oceans, Cloud. Sat probabilities of precipitation have been used to determine a cloud water threshold in a non-raining GMI only retrieval. Precipitation will be added to these pixels to minimize differences between observed and simulated Tb. Rainfall is constrained to be less than that observed by the GPM. This step has not been implemented. Conclusions Changes in GPROF V 5 correct implementation errors in the Tb bias correction scheme, as well as dealing with precipitation known to be below the threshold of the GPM radars. Over the remaining regimes, GPROF V 5 will continue to rely on V 4 of the GPM core satellite products for its a-priori database. Three relatively simple but important changes highlighted above are being implemented for Version 5. More difficult changes related to convective structures over land; Convective/Stratiform separation and improved snowfall retrievals are being deferred to future versions.