Based on GIMMS-NDVI synthesizing half month maximum every year and precipitation and temperature data from 1982 to 2006, changes of vegetation cover and spatial and temporal characteristics of its correlation with precipitation and temperature were studied using linear regression, Sen+Mann-Kendall analysis and correlation analysis methods. Results are as follows: (1) From 1982 to 2006, change of vegetation cover was obvious, with a changing rate of about 3%/10 a (p=0.19). Changes of vegetation showed a rising trend in most of regions with more surface water and a declining trend mainly in few arid desert regions. (2) The fluctuation of precipitation was obvious in Northwest China during the past 25 a and the overall trend showed a little decreasing. Interannual fluctuation of temperature was relatively small and its overall trend significantly increased. (3) From September to February next year, that is, from autumn to winter, the correlation between vegetation changes and precipitation was strong. From March to August, that is, from spring to summer, the response of vegetation to precipitation was a delay-dependent. The response of vegetation to temperature had one month lag in January, March and December and a significant correlation in other months. (4) Spatially, in the arid regions lacking surface water resources, the relationship between vegetation changes and precipitation was positive and between vegetation changes and temperature, negative. In the regions rich in surface water, the relationship between vegetation changes and precipitation was negative, and between vegetation changes and temperature, positive.