A verification study over europe of amsu-a/mhs and ssmis passive microwave precipitation retrievals

Global monitoring of the precipitation requires the full exploitation of all overpasses of present and future satellites carrying cross-track and conically scanning passive microwave (PMW) radiometers. Therefore, it is essential to achieve consistency and accuracy of passive microwave precipitation retrievals from the different sensors orbiting around the globe. Within the EUMETSAT H-SAF program (Satellite Application Facility on Support to Operational Hydrology and Water Management, http://hsaf.meteoam.it) we have developed two different passive microwave precipitation retrieval algorithms, one based on a physically-based Bayesian approach for conically scanning radiometers (i.e., SSMIS), and the other one based on Neural Network approach for cross-track scanning radiometers (i.e., AMSU-A/MHS). The two algorithms are based on the same physical foundation, i.e., same cloud-radiation model simulations to be used as a priori information in the Bayesian solver and as training dataset in the neural network approach. They also use similar procedures for screening of non-precipitating pixels, identification of frozen background surface, presence of snowfall, and determination of a pixel based quality index of the surface precipitation retrievals. These procedures are calibrated according to the different characteristics of the radiometers used. The two algorithms use dynamical/meteorological/environmental variables as ancillary information to characterize the observed event, and mitigate the ambiguity of the rainfall rates at the ground associated to any given set of measured multichannel brightness temperatures. A verification study of the latest versions of the two algorithms has been carried out, where the PMW rainfall estimates are compared against radar observations and raingauge network measurements for several precipitation events in Europe, characterized by different environmental, meteorological, dynamical conditions, and by different precipitation regimes. In this paper we present the main results of this study, discussing strengths and potentials of the two algorithms in relation to the different characteristics of the observed events. In addition we describe the efforts made to achieve consistency of the retrievals from close in time overpasses of the cross-track and conically scanning radiometers for the same event, foreseeing potential improvements in nowcasting and/or hydrological applications.