Abstract: The different land surface features and land cover conditions cause anisotropy surface albedo distribution.It is a powerful tool to derive large area of regional surface albedo using remote sensing technique.Land cover of study area is classified into five types, snow covered, bare soil, vegetation, desert and water, in first step.And then, based on the land cover classification over Northwest China, the different surface albedo calculation methods are presented accounting differential land cover condition.Integrated of the different surface albedo calculation methods, a model for regional surface albedo estimate over the heterogeneous landscape is derived.This approach provided a reliable estimate of surface albedo over Northwest China.The accuracy of the model for regional surface albedo estimation is checked by observed surface albedo data.The results of comparison show that the maximal relative error is 18.75 percent, and the minimal error is 8.82 percent, and the average relative error is about 15 percent.The result shows a high accuracy of surface albedo calculation.The main feature of surface albedo over Northwest China is that the value of regional surface albedo synthetically reflects distribution of conditions of soil moisture and vegetation coverage, and the variability of surface albedo is huge in this region.At last, the relationships between surface albedo and vegetation index and surface temperature over Northwest China are discussed in this paper, and these features can be make the conclusions as fellows:1) The relationship between regional surface albedo and NDVI can be given as a power function:Y=18.74^*(x-0.2)^-0.456.2) The relationship between regional surface albedo and surface temperature(T_s)can be written as,y=-72.3+6.45x-0.086x², over Northwest China in July,1995.There are some errors in deriving the surface temperature form NOAA/AVHRR data, and scale transformations are not solved completely in regional scale surface albedo calculating.In addition, the vegetation-soil interaction needs research further.All of those are factors to restrict the precision of the model for regional surface albedo estimation and the objects to be study in the future.