Toward a Radar-Based Surface Probabilistic Rain and Snow Segregation

Improving the snow-rain delineation process at the surface is critical for a diverse array of users, including operational National Weather Service forecasters and Department of Transportation officials who are in charge of monitoring road conditions. The Multi-Radar Multi-Sensor (MRMS) product suite is a research system integrating radar, rain gauge, satellite, and numerical weather prediction data creating automated and seamless national 3D radar mosaic and multisensor quantitative precipitation estimates at high temporal and spatial resolutions. It contains Surface Precipitation Type (SPT) fields generated using a decision tree and segregating between seven types of hydrometeors, including rain, snow and hail. Until now, the operational deterministic rain-snow segregation is being computed using simple wet and dry bulb temperatures thresholds, not reflecting the inherent uncertainties related to microphysical processes. Also, this deterministic algorithm is biased towards producing too much rain where there should be snow. We propose to develop a gridded, probabilistic surface precipitation typology to segregate rain and snow. Unique, high-resolution datasets are used as inputs of the algorithm: the RAPid refresh (RAP) model fields analysis of dry and wet bulb temperatures, beam height fields from the MRMS product suite, and citizen-scientist reports of surface hydrometeor types from the meteorological Phenomenon Identification Near the Ground (mPING) project. These datasets are utilized to develop and evaluate the probabilistic model. Results show that a 3D model built on RAP environmental variables and sampling information from the radar network is the most accurate. This algorithm is applied to several case studies involving winter storms with mixed precipitation types.