Dual-frequency radar retrievals of snowfall using Random Forest

The microphysical parameters of snowfall directly impact the hydrological and atmospheric models. Dual-frequency radar retrievals of particle size distribution (PSD) parameters are developed and evaluated over complex terrain during the International Collaborative Experiment held during the Pyeongchang 2018 Olympics and Paralympic winter games (ICE-POP 2018). The observations used to develop retrievals were included the NASA Dual-frequency Dualpolarized Doppler Radar (D3R) and a collection of second-generation Particle Size and Velocity (PARSIVEL²) disdrometer. Conventional look-up table method (LUT) and random forest method are applied to the disdrometer data to develop retrievals for volume-weighted mean diameter Dm, the shape factor mu, snowfall rate S, and ice water content IWC. Evaluations are performed between D3R radar and disdrometer observations using these two methods. The results show that the random forest method performs better in retrieving microphysical parameters because the mean errors of the retrievals relative to disdrometer observations are small compared with the LUT method.