The aim was to further research soil erosion characteristics and accurately predict soil erosion amount in karst areas. Based on field surveys and research achievements available, yellow soils, which are widely distributed, were chosen as test soil samples and slope, rain intensity, vegetation coverage and bare-rock ratio were taken as soil erosion factors. Artificial rain simulation instruments （needle-type） were made use of to simulate correlation of rain intensity, vegetation coverage, and bare-rock ratio with soil erosion quantity. Furthermore, multiple-factor linear regression analysis, stepwise regression analysis and multiple-factor non-linear regression analy- sis were made to establish a multiple-factor formula of soil erosion modulus with dif- ferent slopes and select regression models with high correlation coefficients. The re- sults show that a non-linear regression model reached extremely significant level or significant level （0.692〈FF〈0.988） and linear regression model achieved significant lev- el （0.523〈FF〈0.634）. The effects of erosion modulus changed from decreasing to in- creasing and the erosion factors from high to low were rain intensity, vegetation cov- erage and bare-rock ratio when slope gradient was at 6~, 16~, 26~ and 36~. The mod- el is of high accuracy for predicting gentle slope and abtupt slope, which reveals correlation of erosion modulus with erosion factors in karst areas.
