Neural network retrieval of precipitation using NPOESS microwave sensors

This paper will present efforts for the development and validation of passive microwave precipitation retrieval algorithms for the NPOESS (national polar-orbiting operational environmental satellite system) satellite program and the NPOESS preparatory project (NPP) prior to the launch of the first satellite in 2009. The advanced technology microwave sounder (ATMS) offers improvements including finer sampling and spatial resolution over heritage instruments such as the advanced microwave sounding unit instruments AMSU-A/B aboard the NOAA-15, NOAA-16, and NOAA-17, and similar instruments. The conical scanning microwave sounder (CSMS) is planned for the second and subsequent NPOESS satellites. A system for simulating ATMS and CSMS microwave observations from atmospheric data has been developed. This system has shown encouraging results when validated with observations from AMSU-B on NOAA-16. This system is flexible and can be used not only with cross-track scanning instruments but also with conically scanning instruments. A neural network was trained to estimate 5.2deg MM5 rain rates from simulated ATMS observations. Encouraging agreement was observed. However, this algorithm is only preliminary and many improvements are in progress.