The H SAF EPS-SG MWS Day-1 Machine Learning algorithm for snowfall and rainfall surface precipitation rate retrieval: Intercomparison of Machine Learning Techniques and performance analysis
Daniele Casella presents a new algorithm for the Micro-Wave Sounder (MWS) radiometer on board the EPS Second Generation satellites.
The development of precipitation retrieval techniques can now benefit from the availability of unique cloud and precipitation observations by the two space borne radars currently available: the Dualfrequency Precipitation Radar (DPR) on board the NASA/JAXA GPM Core Observatory, and the NASA CloudSat Cloud Profiling Radar (CPR). These two radars have demonstrated their complementarity in the monitoring of precipitation. While DPR has shown a high accuracy in the estimate of medium and intense precipitation regimes, CPR has proven to be very suitable for the retrieval of light rain and snowfall. Within the H SAF a new algorithm for the Micro-Wave Sounder (MWS) radiometer on board the EPS Second Generation satellites (MetOp-SG) has been developed. Different machine learning approaches were tested in order to select the most suitable for optimizing the performance of the algorithm for the detection (a classification problem) and estimate (a regression problem) of rainfall and snowfall.
The details concerning the coincidence datasets creation, the design of the ML modules and the algorithm input selection procedure will be presented, together with results of the algorithm's performance.