Slope length and slope steepness are critical topographic factors (L and S) in the Universal Soil Loss Equation (USLE) and Chinese Soil Loss Equation (CSLE) for soil erosion modelling.Both slope length and slope gradient are potentially sensitive to spatial resolution when calculated in a GIS framework.The resolution effect on the LS factor and approaches suitable for improving the LS factor at a coarse resolution have not been well identified.To address this problem,the LS factor at 5-m and 30-m resolution in twenty-four watersheds with various terrains was estimated.And a downscale model based on matching of the lower resolution LS cumulative frequency curves to a higher resolution (“Histogram Matching” method) was tested for its potential to improve LS factor estimation accuracy.In the larger relief mountainous area,compared to 5-m resolution,the 30-m resolution generated LS was generally overestimated by more than 20％ and in lower relief areas underestimated by more than 15％.This bias is less than 10％ in medium relief areas.The downscale model improved LS factor estimates compared to the 30-m resolution estimate by more than 10％ when comparing frequency distribution curves and more than 20％ in mean values in larger relief areas.The downscale model worked well in all regions except for the low relief areas,which intuitively are the low soil erosion potential areas.The results of this research help quantify the uncertainty in soil erosion estimates and may ultimately help to improve the assessment of soil erosion through its impact on LS factor estimates,especially at regional and global scales.