Snow cover depth (D) is one of the important parameters for hydrological model and meteorological forecast. Based on snow cover depth and meteorological data during 1961-2013 at Urumqi Station, the initial date, the ending date and the duration of snow cover and the temporal variation of D within a snow cover year were analyzed. The fitting effects of quadratic function, Gaussian model, Lorentzian model and lognormal model for D process were compared. In addition, the trend test of the maximal snow depth (D_(max)) was conducted; the influence of climate factors on D and D_(max) were also discussed briefly. The results showed that the duration of snow cover in Urumqi lasts for 73 to 207 days, with an average of 141 days. Both the initial date and ending date are postponing and the duration of snow cover is increasing, but not significantly. The time-curves of D within a snow cover year can be classified as four types, i.e., multi-hump and flat type (a), right-skewed type (b), pinnacle type (c) and flat hump type (d). After the D series within a snow cover year being fitted for four curve types, it was found that the Gaussian function was generally good. Thus, the annual variations of the three parameters in Gaussian function over 1961-2013 were obtained. There is no auto-correlation structure in D_(max) series, but there is a significant upward trend in it. Among the climate factors, air temperature, especially the minimal temperature, has negative correlation to D; while precipitation within cold season has positive correlation to D_(max). Fast increasing of air temperature in spring and enough snow cover depth may result in snow-melt type flood, which usually corresponds to the time-curves of D of pinnacle type and flat hump type.