Abstract
A new approach for distinguishing precipitation types at the surface, the spectral bin classifier (SBC), is presented. This algorithm diagnoses six categories of precipitation: rain (RA), snow (SN), a rain–snow mix (RASN), freezing rain (FZRA), ice pellets (PL), and a freezing rain–ice pellet mix (FZRAPL). It works by calculating the liquid-water fraction fw for a spectrum of falling hydrometeors given a prescribed temperature T and relative humidity profile. Demonstrations of the SBC output show that it provides reasonable estimates of fw of various-sized hydrometeors for the different categories of precipitation. The SBC also faithfully represents the horizontal distribution of precipitation type inasmuch as the model analyses and surface observations are consistent with each other. When applied to a collection of observed soundings associated with RA, SN, FZRA, and PL, the classifier has probabilities of detection (PODs) that range from 62.4% to 98.3%. The PODs do decrease when the effects of model uncertainty are accounted for. This decrease is modest for RA, SN, and PL but is large for FZRA as a result of the fact that this form of precipitation is very sensitive to small changes in the thermal profile. The effects of the choice of the degree of riming above the melting layer, the drop size distribution, and the assumed temperature at which ice nucleates are also examined. Recommendations on how to mitigate all forms of uncertainty are discussed. These include the use of dual-polarized radar observations, incorporating output from the microphysical parameterization scheme, and the use of ensemble model forecasts.
