The accuracy evaluation of various methods for predicting winter precipitation type during ICE-POP 2018 field campaign

Abstract

The snowfall prediction of winter precipitation types such as rain (RA), snow (SN), sleet (RASN), ice pellet (IP) and freezing RA (FZRA) is challenging because of its variability depending on the atmospheric conditions. There have been several efforts to predict the precipitation types from meteorological parameters: using one parameter (e.g. thickness or wet-bulb temperature) or two parameters (e.g. ground temperature and relative humidity, wet-bulb temperature and low-level lapse rate). Recent attempt is to construct bin microphysics model (e.g. Spectral Bin Model; SBM) to predict precipitation types. The SBM simulates phase change of precipitation particles from cloud top to ground at specific atmospheric condition. The SBM decides precipitation type at ground level from considering simultaneously melting, freezing, evaporation, and sublimation.

In this study, we evaluate the accuracy of existing five methods using the quality-controlled precipitation type from PARSIVEL distrometer. PARSIVEL data is classified into three precipitation types (RA, SN, and RASN) by the frequency of precipitation type based on classification mask from Yuter et al. (2006). The classification criteria are decided from cross-check with other observation data such as sounding and VertiX data at same site. We collect about 140 precipitation cases which the temperature is close to 0 ° during ICE-POP 2018 period. The predictions of different methods using sounding data as input are performed. The probability of detection (POD) of five methods as the observed precipitation type (RA, RASN, SN and total) is calculated under three conditions, respectively: all case (ALL), higher relative humidity (HRH), lower RH (LRH). SBM shows highest POD among five methods at total-ALL, RASN-ALL, total-LRH and RASN-LRH whereas lowest POD at RA-ALL, RA-HRH. In summary, SBM is generally great performance for predict the winter precipitation type, also have advantage at prediction of RASN whereas lower performance at prediction of RA. Accuracy of other methods will be shown.

Acknowledgment

This work was funded by the Korea Meteorological Administration Research and Development Program under Grant KMI2018-06810.