[Background] As we all know, soil erosion is considered to be the most common environmental problem in the world influenced by multi-factors in multi-levels and multi-scales. Conducting regular survey on regional soil erosion condition plays an important role in evaluating the effectiveness of control measures. In China, the stratified sampling with unequal probability and the model method were firstly utilized to investigate national soil erosion conditions from 2010 to 2012. The CSLE model and the extrapolation method based on investigation units were mainly utilized in the water erosion regions. The sampling densities of investigation units were 1% for general and 0.25% for plain areas and deep mountains. It is important to understand how different sampling densities and estimation methods based on CSLE model impact on the soil erosion survey and evaluation at the county scale. [Methods] In consideration of the study area, stratified sampling method, the unique situation of soil erosion, and the workload in field investigation of soil erosion, Mengyin County located in Yimeng mountain area was taken as an example. Based on SPOT 6 image and the topographic map at 1:10 000 scale, two sampling densities with 1% and 4%, and three extrapolation methods of direct xtrapolation on sampling units, Kriging extrapolation on sampling units and grid estimation were adopted to calculate the amount of soil erosion at the county scale. Then the influence of different sampling densities and extrapolation methods were explored by contrasting the results of three methods on different sampling densities respectively. [Results] 1) Direct extrapolation and Kriging extrapolation methods were more affected by the sampling densities. The discrepancy percentage of soil erosion area calculated by two methods were 8.82% under 1% sample density and 7.96% under 4% sample density, and the relative discrepancy were 19.05% and 17.43% respectively. But grid estimation method was less affected by the sampling density. The discrepancy percentage of soil erosion area was only 3.13% and the relative discrepancy was only 9.27% comparing 1% with 4% sample density. 2) Under the same sampling density, the results of direct extrapolation and Kriging extrapolation were similar, but these were quite different from the grid estimation result, especially in spatial distribution of soil erosion. Compared to the result of grid estimation, the discrepancy percentage of soil erosion area by direct extrapolation and Kriging extrapolation were between 11.77% and 18.12%, and the relative discrepancy were between 34.72% and 48.93%. [Conclusions] As a consequence,in view of precision and workload in soil erosion survey at the county scale in Yimeng mountain area, the first ecommendation is to adopt grid calculation based on 1% sampling density with high resolution remote sensing images. If without high resolution remote sensing images, it is advisable to adopt extrapolation method based on 4% sampling density. From the study, the discrepancy between different densities and diverse extrapolation methods was known, and suitable sampling density and extrapolation method were selected, thus the field work can be reduced and the precision can be improved, which is important in dynamic monitoring of water and soil loss at county scale.