Hurricane GPROF: An improved hurricane rain rate retrieval for TRMM and GMI

The current version of the Goddard PROFiling (GPROF 2014) algorithm generally performs well in producing rainfall estimates from TRMM Microwave Imager (TMI) brightness temperatures. However, this algorithm relies upon an a-priori database constructed from coincident Precipitation Radar (PR) and TMI observations of precipitation that occurred during a one year period. Due to the infrequent nature of hurricanes, and the narrow swath of PR rain rates included, the database contains little hurricane data. Consequently, the high rain rates associated with hurricanes are not captured in a statistically representative way. To address this insufficiency, a new empirical database for hurricanes was created. The hurricane database HURDAT2 was used to parameterize the location and spatial extent of Atlantic and Eastern Pacific hurricanes occurring between 2004 and 2012. An empirical database was then compiled from TRMM overpasses that were coincident with HURDAT2 hurricane locations, over ocean surfaces only. Hurricane GPROF incorporates this new empirical hurricane database into the GPROF 2014 retrieval to provide rain rates for TMI that are specific to hurricanes. Initial comparisons between Hurricane GPROF rain rate estimates from TMI and PR rain rates indicate that the main features of the raining scenes are generally captured very well. The very highest rain rates remain difficult to estimate due to their infrequent occurrence in the empirical database. Hurricane GPROF improves upon GPROF 2014 rain rate estimates over hurricanes for TMI, producing higher rain rates and lower biases when compared to PR rain rates. A subset of hurricane overpasses shows that Hurricane GPROF decreases rain rate biases with PR by over 30%, compared to GPROF 2014. Hurricane GPROF produces the largest decrease in bias for rain rates over 7.5 mm.hr-1. It is anticipated that Hurricane GPROF would also be suitable for use with GMI.