This paper estimated HJ-1 land surface albedos in the Heihe region using the backup algorithm of Moderate Resolution Imaging Spectroradiometer（MODIS） Bidirectional Reflectance Distribution Function（BRDF）/Albedo product（method Ⅰ）

the Bayesian inference-based algorithm（method Ⅱ） and the Lambertian surface algorithm（method Ⅲ）.Compared inversed albedos with surface observations

statistical analysis results showed that:（1） The high resolution albedos from HJ-1 CCD data can provide spatial distribution of underlying surface as well as surface details.Different land cover types’ statistic values indicates that method Ⅰ and method Ⅱ capture similar results

with the absolute error of 0.01 and the relative error of 4% as compared with albedos from MODIS

while method Ⅲ has the absolute error of 0.03 and relative error of 13%.（2） The improvement in the albedo by method Ⅰ and method Ⅱ is almost independent to land cover types

capturing relative error between 2% and 8%;However

the temporal reliance of estimated albedo is more significant

and the improvement is more obvious in the maturity than in the dormancy.（3） Surface albedos estimated by method Ⅰ and method Ⅱ have better consistency with field observations.The root mean square errors are less than 0.05

and relative errors are less than 23%

while results of method Ⅲ are 0.069 and 36.3%

respectively.（4） The retrieval of albedos based on the prior knowledge may depend on the geometry of the sun and the observation

and thus depend on the season and the latitude

as well as sensor specifications.This study will provide significant understanding for space-borne albedo retrieval which lacks of sufficient multi-angular observations.