Abstract: Benggang is one of the most serious failures in soil erosion, particularly in the red loam hilly areas in southern China. In this study, the risk assessment was carried out on Benggang development using GIS. A case study was taken from the granite area of Jiangxi Province, China. 17 049 sample points of Benggang were obtained using remote sensing interpretation and field survey. 10 potential influence factors of Benggang were also selected, such as the slope, aspect, elevation, topographic relief, rainfall erosion, temperature, lithological, soil type, land use type, and vegetation coverage. A systematic investigation was then implemented to explore the relationship among parameters and avalanche occurrence; Gray correlation and Pearson correlation analysis were applied to calculate the importance and correlation matrices of the influence factors. As such, the index system of Benggang risk assessment was obtained to screen the influence factors; Furthermore, quantitative analysis was also performed on the relationship among environmental impact factors and the development of Benggang; The weighted information and frequency ratio models were utilized to calculate the risk index of Benggang, according to the natural breakpoint grading. A series of tests were conducted to quantitatively evaluate the accuracy of the risk assessment. The best performance was selected to optimize the risk assessment model. The results show that: 1) The temperature, elevation and land use type were the main influencing factors on the development of Benggang in Jiangxi Province. Among them, 29 of the 67 graded indicators were used to promote the development of Benggang, while 38 indicators were to inhibit. 2) The number and percentage of Benggang in the zoning of the three models were positively correlated with the risk level. A higher percentage was found with the number of Benggang in the extreme and normal high-risk zones. Three models were selected to effectively assess the susceptibility of Benggang. The model accuracy was evaluated on the working characteristic curve of the subjects. The test results show that the AUC values of the information value model, the weighted information value model and the frequency ratio model were 0.860, 0.852 and 0.833, respectively, indicating the better assessment accuracy of three models. All of them were suitable for the assessment of the risk of Benggang. There was the slightly higher accuracy of the information value model. A more objective and accurate assessment was achieved on the risk of avalanche occurrence. 3) The risk level of Benggang shared a trend of gradual increase from the north to the south. The extreme high-risk area was concentrated mainly in the southern, while the high-risk area was distributed in the south and central. The finding can also provide a strong reference and data support to the prevention and control of Benggang erosion in Jiangxi Province.