Machine Learning Enhanced Snowfall Rate Retrievals from Passive Microwave Sensors

The NOAA NESDIS operational Snowfall Rate (SFR) product is generated at near real time from a suite of passive microwave (PMW) sensors onboard several polar orbiting satellites operated by NOAA and its partner agencies. It estimates liquid equivalent snowfall rate over global land and has been in operation since 2012. The SFR algorithm consists of two main components: snowfall detection (SD) and snowfall rate estimation (SFR). The SD component is an eXtreme Gradient Boosting (XGB) model trained on the global ground observations from the NOAA Integrated Surface Database (ISD) and the SFR component is a physical based 1DVAR model enhanced by Neural Networks (NN) to improve the Ice Water Path (IWP) initialization and SFR systematic bias correction. Through the employment of the machine learning (ML) technique, the NESDIS SFR product has also been extended over the ocean, including open ocean, sea ice and coast, and to extremely cold conditions such as those in the polar regions. The ML SD and SFR algorithms have been successfully applied to cross-scanning instrument such as ATMS and MHS sensors and the conical scanning GMI sensors to generate the SFR product operationally. In this presentation, we will introduce our recent development of the ML SD and SFR algorithm for SSMIS sensors aboard the DMSP-F16, F17 and F18 and their global validation results and then present the development to cross calibrate the SFR products from different satellites. Case study compared with Stage IV radar and gauge combined precipitation rate, CPR near surface snowfall rate, NOHRSC hourly snowfall analysis and ERA5 hourly snowfall rate reanalysis will also be provided.