[Background]Soil erosion has become a global ecological and environmental problem. It is now being recognized as a severe threat to socio-ecological security and stability,and it is relative with the food security,resilience to climate change and geosocial stability. Soil erosion is particularly acute in the Loess Plateau. In order to control soil erosion,the quantitative study of soil erosion must be strengthened. Soil erosion is affected by many factors such as climate,vegetation,and land use and soil properties. Among those factors,soil erodibility has been qualitatively evaluated as a key indicator for estimating soil loss and usually being measured by K value. The research of soil erodibility is significant to understand the principle of soil erosion,to estimate soil erosion modulus quantitatively and to control soil and water loss reasonable. The estimation method of soil erodibility is numerous,but the regional applicability of different models remains to be discussed. [Methods]We conducted a study to select the optimal estimation method of soil erodibility (K value) based on the basic data of precipitation,soil, remote sensing images and socioeconomic data in Ansai Watershed. We used GIS technology and CSLE model to calculate soil erosion modulus in 2006 - 2014 and compared it with the corresponding monitoring value of sediment. Tthe K value was calculated by EPIC,NOMO,M-NOMO,Torri and Shirazi model. The comparison between the simulated value of soil erosion modulus and the monitoring value of sediments is based on the principle that the mean absolute error (MAE),the mean relative error (MRE) and the root-mean-square error (RMSE) are closer to 0,the accuracy factor (A_f) is closer to 1,the regional applicability of model is higher. [Results]The mean soil erosion modulus of Ansai watershed in 2006-2014 based on the five models of EPIC,NOMO,MNOMO,Torri,and Shirazi was 65.59,106.00, 108.47,76.69 and 47.68 t /hm~2 respectively. The mean monitoring value of sediment in corresponding year was 17.73 t /hm~2. Based on the above evaluation indexes,we knew that the Shirazi model's regional applicability was the highest,the value of MAE,MRE and RMSE was 30.93,3.25 and 43.66 respectively,the A_f value was 4.41,The regional applicability of EPIC model took second place,the value of A_f was 5.80,The regional applicability of Torri model was in the middle level. The regional applicability of NOMO model and M-NOMO model was lowest,it had the biggest difference with the actual situation,the A_f value was 7.99 and 7.88 respectively. [Conculsions] Based on the above analysis,we concluded that the Shirazi model had the best applicability in study area comparing to the other four K value estimation methods. We should be preferred to choose the Shirazi model in the future watershed scale soil erodibility (K value) estimation and soil erosion evaluation.