Abstract: Soil erosion is one of the most serious ecological problems in the world, and the terrain is an important factor influencing the soil erosion. Some scholars applied the relevant ideas and theories of landscape ecology and the algorithm of partial factors in Universal Soil Loss Equation (USLE), and brought forward the soil loss evaluation index (SL_sw index) for soil erosion measurement at different scales. The SL_sw index can reflect to a certain extent the process of soil loss and erosion, avoid the common problems of physical model, and provide us research ideas and methods in the study of land use and soil loss and erosion on different scales. But it was too simple to study the topography, so this study was to modify the topographic factor (G) in SLsw, and then obtained soil erosion evaluation index with correction factor G (SL_sw_G). This study chose the Loess Hilly Gully Region of Ansai watershed as the research area, and used GIS technology to calculate the SL_sw_G based on the data of Ansai catchment DEM data, daily rainfall data of hydrological observation station, soil sampling data and land use data. The measured sediment transport volume in Ansai watershed was used to verify the modified index. And then the soil erosion intensity and its temporal and spatial variation characteristics of Ansai watershed were analyzed. The results showed that the soil erosion distribution was basically the same before and after the amendment of SL index. The soil erosion was higher in southeast region of Ansai watershed, while it was lower in the northwest region. Due to the influence of human factors, soil conservation measures were relatively weak in the southeastern part of the Ansai watershed. Meanwhile, there were more light soil clay and silt, and the clay and silt had obvious bond when they were wet, so it was easy to be washed away in the course of raining. According to the data of soil sampling analysis, the content of soil clay and silt were higher in the southeast, so it was easy to cause the soil erosion in the southeastern part of the Ansai watershed. According to the T test, the SL_sw and SL_sw_G has the variance homogeneity before and after the amendment, and P=0.037<0.05. Compared with the measured sediment transport volume, the correlation between the modified index and the measured volume was greater than the correlation between the original index and the measured volume. It can be concluded that the modified index can better explain the regional soil erosion of the study area.