Under the background of climate change, the warming trend in China is significant, especially in winter, which has changed the degree and the spatio-temporal distributions of heat resources. Based on the meteorological data of 32 meteorological stations over the past 56 years in and around the Qilian Mountains and DEM data, rasterizing accumulated negative temperature data was conducted using the "multiple regression plus residual interpolation" method. Results of cross-validation errors revealed that OK gave the best simulation. The results of spatial and temporal variation showed that: (1) Annual average absolute values of accumulated negative temperature were ranged from 607 ℃ to 3507 ℃ in the study area, and the differentiation of regional distributing was large; The area of accumulated temperature segment Ⅲ, Ⅳ and Ⅱcaptured nearly 91.31% of total area; The accumulated temperature segments presented nearly ring distribution and rised gradually from outside to inside except Ⅵ; On grid scale, the dynamic index of accumulated negative temperature decreased gradually from south to north. (2) By using the methods of linear trend estimates, from 1960 to 2015, the results indicated that the initial date showed an obvious delaying tendency with the linear rate of 1.3 d·(10a)~(-1), the termination date showed an obvious forward tendency with the linear rate of 2.4 d·(10a)~(-1), and the sustainable days showed an obvious declining tendency with the linear rate of 3.7 d·(10a)~(-1). Most of the stations passed the statistical tests of P<0.1. (3) From 1960s to 2000s, the area of accumulated temperature segment Ⅰ and Ⅱ decreased significantly, and accumulated temperature segment Ⅴ and Ⅵ increased significantly, which was the most obvious characteristic in Qilian Mountains. However, the trend was opposite from 2000s to after 2010 (20102015), which may be the result of warming stagnation. The area of accumulated temperature segment Ⅲ and Ⅳ decreased and increased respectively, which showed fluctuation changes. (4) The main types of the change of accumulated temperature segment in the study area were Ⅲ~Ⅳ(25.72%), Ⅳ~Ⅴ(19.66%)andⅡ~Ⅲ(17.90%), which captured 97.73% of the total change. The area which keeped unchanged captured 32.48% of the total change. The transfer matrix of accumulated negative temperature indicated that, from 1960s to after 2010, accumulated negative temperature segment was one-way transfer, which transferred from the higher absolute values of accumulated negative temperature to the lower. However, in adjacent time, it presented two-way transfer. The way from the higher absolute values of accumulated negative temperature to the lower was dominant.