Abstract: Based on the daily snowfall data from 27 meteorological stations in Qinghai Province from 1962 to 2012, the spatiotemporal variation patterns of the beginning and ending dates of snowfall and snowfall days were analyzed by means of inverse distance-weighted interpolation, and Mann-Kendall mutation test. The results show that the beginning date of snowfall in Qinghai Province mainly concentrated in September, October and November, but the ending date mainly concentrated in April and May. The earliest snowfall occurred in the Three-River headwaters region, and the latest was in the Qaidam Basin. In addition, the beginning date of snowfall in most areas of Qinghai Province was gradually delayed, but the ending date of snowfall was brought forward. The snowfall days concentrated on March, April and October (mainly light and heavy snows), and showed a decreasing trend at rates of 1-3 days per decade. An obvious abrupt change existed in the beginning and ending dates of snowfall and snowfall days. Latitude, altitude and annual mean temperature correlated strongly with the beginning and ending dates of snowfall and snowfall days, and the correlations of the beginning and ending dates of snowfall with the average daily temperature (mean, maximum and minimum values) were different in different months. Besides, a positive correlation between snow season length and mean snowfall days existed.

Key words: giant panda., species distribution model, Biomod2, MaxEnt