Neural network-based method for the estimation of the rain rate over oceans by measurements of the satellite radiometer AMSR2

The rain rate (RR) retrieval method for the RR estimation over ice-free areas of the ocean is presented. Measurements of the Japanese Advanced Microwave Scanning Radiometer 2 (AMSR2) on board the satellite GCOM-W1 are used. The method is based on the results of the numerical modeling of brightness temperatures of the outgoing microwave radiation of the ocean–atmosphere system and their subsequent conversion into the RR using neural networks. A simplified form of the transfer equation is used. Its errors for the considered wavelengths do not exceed 1 K at an RR of less than 20 mm/h. The method is verified by comparison with the Tropical Rainfall Measuring Mission’s (TRMM) Microwave Instrument (TMI) RR product. As a result of the comparison, the rain rate retrieval error within the range of 20 mm/h is found to be 1 mm/h.