Improving Observations Of Precipitation Type At The Surface: A 5-Year Verification Of A Radar-Derived Product From The United Kingdom'S Met Office

This study aims to verify the skill of a radar-based surface precipitation type (SPT) product with observations on the ground. Social and economic impacts can occur from SPT because it is not well forecast or observed. Observations from the Met Office's weather radar network are combined with postprocessed numerical weather prediction (NWP) freezing-level heights in a Boolean logic algorithm to create a 1-km resolution Cartesian-gridded map of SPT. Here 5 years of discrete nonprobabilistic outputs of rain, mixed-phase, and snow are compared against surface observations made by trained observers, automatic weather stations, and laser disdrometers. The novel skill verification method developed as part of this study employs several tolerances of space and time from the SPT product, indicating the precision of the product for a desired accuracy. In general the results indicate that the tolerance verification method works well and produces reasonable statistical score ranges grounded in physical constraints. Using this method, we find that the mixed precipitation class is the least well diagnosed, which is due to a negative bias in the input temperature height field, resulting in rain events frequently being classified as mixed. Snow is captured well by the product, which is entirely reliant upon a postprocessed NWP temperature field, although a single period of anomalously cold temperatures positively skewed snow scores with low-skill events. Furthermore, we conclude that more verification consistency is needed among studies to help identify successful approaches and thus improve SPT forecasts.