Snow cover information is important for climate change research,water resource utilization and snow disaster monitoring.Estimating sub-pixel snow cover fraction can reduce the error caused by the mixed information within apixel,thus improve the accuracy of snow cover mapping.Compared with the MODIS(moderate-resolution imaging spectroradiometer)snow cover fraction(FRA)product,aper-pixel estimating model was developed and validated to achieve a more precise estimation in typical area of Bayanbulak,Xinjiang Province.Using HJ-1B30-m observations asground truth,the percentage of snow cover was calculated for 500-m cells.Then a regression relationship between 500-m normalized difference snow index(NDSI)observations and FRA was developed for each pixel.The results showed that compared with the MOD10A1FRA product,the standard error and the mean absolute error of the propose model were reduced.But the per-pixel estimating model tended to overestimate FRA overall.The problem was resulted from solar radiation difference between sunny and shady slope,the fragmented snow in fringe area and the complications of snow in forested area.Overall,the per-pixel estimating model can estimate sub-pixel snow cover fraction more accurately,thus provides support for reasonable water resource utilization and snow disaster mitigation in pastoral areas.