The 2012-2020 surface observation and sounding data from Beijing Meteorological Observatory, Zhangjiakou Station and Laoting Station in winter as well as other data are used to establish a set of discrimination criteria for different precipitation types by statistically analyzing vertical temperature and humidity profiles. From forecast perspective, we focus on two key factors of snowflakes forming and melting, and adopt the cloud-top temperature and the 0℃-layer height as new indicators of discrimination criteria for precipitation types. The results show that cloud-top temperature below or equal to -14℃ is a key threshold to generate adequate snowflakes or other ice particles in clouds. This is a necessary condition for snowfall. At the same time, the 0℃-layer height higher than or equal to 0.5 km and lower than or equal to 0.1 km is a threshold for the snowflakes to hardly melt and completely melt, respectively. Sleet or rain is likely to appear with cloud top-temperature between -14℃ and -4℃. The 0℃-layer height at 0.1 km is a threshold of distinguishing sleet from rain. The TS scores of discrimination criteria based on cloud-top temperature and 0℃-layer height in this paper are significantly improved compared with those by using the combination of temperatures at specific layers. The TS scores for snow, sleet and rain reach (improved) 0.93 (0.11), 0.57 (0.39) and 0.86 (0.43), respectively. The results of this study could provide a new reference for forecast and amendment of winter precipitation type.