利用1961—2018年10月至次年4月新疆巴音郭楞蒙古自治州（巴州）北部6个国家观测站的天气现象资料，统计分析近58 a降雨、降雪、雨夹雪、雨转雪4种降水相态的气候特征，得出巴州北部10月和4月以降雨为主，12月和1月以降雪为主，雨夹雪和雨转雪主要在11月和3月。利用2003—2018年10月至次年4月库尔勒站探空资料，得到与降水相态转换关系密切的10种物理量，量化了降水相态判识特征及指标。结果表明：（1）地面最低温度、近地层温度、850 hPa温度、500~850 hPa之间的位势厚度、700~850 hPa之间的位势厚度、0 ℃层高度可将4种降水相态完全判别，500~850 hPa之间的温差、700~850 hPa之间的温差能有效判别降雨、降雪、雨夹雪。（2）构建了降水相态预报评分办法，库尉轮平原和焉耆盆地的综合指标准确率达92.06%和94.36%，综合判定后预报评分达93.58%。（3）特性层温度及温差对降雨和降雪的预报效果好于雨夹雪；位势高度及厚度对降雪的预报效果好于降雨和雨夹雪；位势厚度对雨转雪的预报效果要好于特性层温度。研究降水相态综合判据对巴州北部降水相态的判别有很好的参考价值，为提高雨雪相态转换预报准确率提供科学依据。

Based on the weather phenomenon data from six national weather stations in northern Bayingol Prefecture of Xinjiang, China, the climatic characteristics of rain, snow, sleet, and rain to snow from October to April in the last 58 years (1961—2018) are analyzed. The results show that the major precipitation types in northern Bayingol Prefecture are rain in October and April, snow in December and January, and sleet and rain to snow mainly occur in November and March. The discrimination criterion and index of precipitation types are quantified using 10 physical variables closely related to the precipitation types transition discovered using sounding data from the Korla station from October 2003 to April 2018. The results show that: (1) Surface minimum temperature, near-surface air temperature, temperature at 850 hPa, geopotential thickness between 500 hPa and 850 hPa, geopotential thickness between 700 hPa and 850 hPa, and 0 ℃ level height can completely distinguish the four precipitation types, and the temperature difference between 500 hPa and 850 hPa, temperature difference between 700 hPa and 850 hPa can distinguish rain, snow, and sleet better. (2) A phase state scoring method for precipitation forecasts was developed, and after a thorough analysis, the combined index accuracy was 92.06% and 94.36% for Korla-Yuli-Luntai Plain and Yanqi Basin, respectively, and the forecast score was 93.58%. (3) The characteristic layer temperature and temperature difference forecast rain and snow with more accuracy than sleet, the geopotential height and thickness forecast snow with greater accuracy than rain and sleet, and the geopotential thickness forecast rain to snow with greater accuracy than the characteristic layer temperature. These comprehensive precipitation type indices have a high reference value for distinguishing precipitation types in northern Bayingol Prefecture and can provide a scientific foundation for improving rain-snow phase transition forecasting accuracy.