Development of Multi-Temporal-Scales Precipitation-Type Separation Method in the Qinghai–Tibetan Plateau

The accurate identification of precipitation types is very important for understanding the hydrological processes in cold regions. Existing identification methods have been established based on daily precipitation and meteorological data, which cannot match the high temporal resolution (such as hourly) simulations of hydrological processes. Based on the minutely surface meteorological data in the QTP from 2012 to 2021, we established three sub-models of the dynamic threshold method with wet-bulb temperature (Tw) and three sub-models of the frequency threshold method with air temperature (Ta) for distinguishing among precipitation types. The results revealed that the mean accuracy (ACC) of the three precipitation types was 0.86, and that these models provided a refined and accurate precipitation identification performance for the Qinghai–Tibet Plateau (QTP). However, these models performed well in the identification of rain and snowfall but performed poorly in the identification of sleet. In addition, the smaller the time scale and regional scales, the better the identification rate. In particular, snowfall is overestimated when daily precipitation-type separation thresholds are input into hourly or minute hydrological models. Therefore, to improve simulation performance, it is important to develop multi-temporal scale precipitation-type partitioning models, take regional variations into account when setting temperature thresholds, and conduct analyses at the finest possible time resolutions to minimize scale-related uncertainties.