Abstract: Using datasets of routine upper-air and surface observations, wind profile radar, ground-based microwave radiometer and NCEP/NCAR reanalysis with 1°×1° resolution, we have investigated the causes of forecast failure about the first snow event in Shijiazhuang in 2016 as well as the evolution of precipitation types in the event. The results show that the snowfall threshold used in the forecast cannot reflect the conditions under that the snow crystals form and grow in cloud, and the data used to extract the temperature threshold (whether melt or not) in the process of snowfall does not have high enough accuracy, which leads to the forecast failure. Vertical radial velocity from wind profile radar can estimate the start and end of precipitation and its types. There is a consistent downward radial velocity below 3 km when rain falls; the maximum velocity exceeds 1.5 m·s^-1. The precipitation probably changes from rain to snow when the radial velocity is reduced to below 1 m· s^-1, and the height falls below 1 km. The temperature profile, which is retrieved from the microwave radiometer, can be used in the nowcasting of the transition of snow and rain because it can estimate conditions under which the snow crystals form and grow, as well as whether they melt or not.