Abstract:Using the actual snowstorm data of 122 national basic meteorological stations in Shandong Province from January to March and November to December of 2017-2021, with a snowfall volume of ≥ 10.0 mm, four methods were used to evaluate and compare the snowstorm forecast in Shandong, including binary classification, neighborhood method, time offset method, and magnitude fuzzy method. The results show that: 1) The total number of snowstorm stations during the period was 182, and the distribution of snowstorm had obvious spatial and temporal distribution characteristics, mainly appearing in the northern part of the peninsula, and the probability of snowstorm was the least in the southeast of Shandong and the southern part of the peninsula. The frequency of snowstorms varies significantly between years and months, with 98 occurring in the most years and only 5 occurring in the least years. February is the most frequent month, accounting for 38.5% of the annual snowstorms. 2) The accuracy of the most widely used binary classification method in current business is relatively low, with only 12.08% in 24 hours. The main reason is that this method has multiple effects in space, time, and magnitude, which cannot accurately reflect the prediction ability. 3) The accuracy rates of neighborhood method, time offset method, and magnitude fuzzy method for 24-hour snow storm forecasting are 14.4%, 14.69%, and 30.05%, respectively. Compared with binary classification method, the accuracy rates of these three testing methods have significantly improved, while the false and false alarm rates have significantly decreased. 4) A comprehensive evalutation method that integrates neighborhood method, time offset method, and magnitude fuzzy method can distinguish forecast differences from three dimensions: space, time, and magnitude. The evaluate results are more precise and accurate, which is conducive to guiding forecasters to put down the burden of "low evaluate scores" and make scientific and objective predictions, improving the effectiveness of forecasting services.