Assessment of a Multidimensional Satellite Rainfall Error Model for Ensemble Generation of Satellite Rainfall Data

This letter presents preliminary insights from the pursuit of the following scientific query: "How realistic is ensemble generation of satellite rainfall data by a multidimensional satellite rainfall error model?" The authors first evaluated the scale-dependent multidimensional error structure for two satellite rainfall algorithms developed at the NASA Goddard Space Flight Center, namely: 1) the infrared (IR) estimates known as the 3B41RT product and 2) the combined passive microwave (PMW) and IR estimates known as the 3B42RT product. Ground radar (WSR-88D) rainfall fields from the Southern Plains of the U.S. were used as reference. Next, by reversing the definition of reference and corrupted rain fields produced by a multidimensional satellite rainfall error model (SREM2D, developed by Hossain and Anagnostou), the authors derived the inverse multidimensional error structure of WSR-88D rainfall fields with respect to the satellite rainfall estimation algorithms. SREM2D was then applied on actual satellite rainfall data with the pertinent inverse error parameters to generate an ensemble of most likely realizations of the reference WSR-88D rainfall fields. The simulated ensemble was then compared with that derived from a simpler (bidimensional) inverse error modeling approach. The accuracy of the SREM2D rainfall ensemble was observed to be higher than the simpler error-modeling scheme for the 3B41RT product. No tangible improvement was observed for the 3B42RT product, which is attributed to the heterogeneous nature of 3B42RT data statistics that was not accounted for in the inverse SREM2D approach. The overall conclusion is that a multidimensional error modeling approach such as SREM2D has the potential to generate realistic ensembles of satellite rainfall fields, which should be considered as an improvement over the more widely used simpler error-modeling scheme. A combined use of the multidimensional error model with a sequential error correction scheme could therefore potentially improve the diagnosis of satellite rainfall-based predictability of the global water and energy cycle