Abstract: Abstract: It is valuable for soil and water loss control and its benefits evaluation to understand the relationship between rainfall and soil erosion process. In order to make clear what rainfall type accelerates soil erosion on slopes of red soil in south part of China, and to ascertain the response of soil erosion intensity to rainfall characteristic under various rainfall types, data of meteorology and hydrology of 134 individual rainfalls in experimental plots on bare slopes of red soil in north part of Jiangxi Province were collected and analyzed. The amount, duration and intensity of rainfall were selected as the characteristic indices for erosive rainfall. Two methods including quick clustering and discriminant clustering were introduced to classify the rainfall types resulting in soil erosion. Pearson correlation analysis was then used to find the main indices that contributed the most to the slope soil erosion. Further, Mann-Kendall test and the other statistics method were used to determine the relationship between soil erosion intensity and the rainfall characteristic indices under different rainfall types. The results showed that the erosive rainfall in north part of Jiangxi Province could be classified into three types: type A (high frequency, short duration, minor amount of rainfall, and strong rainfall intensity), type B (medium frequency, medium duration, medium amount of rainfall, and medium rainfall intensity) and type C (low frequency, long duration, great amount of rainfall, and low rainfall intensity). Soil erosion on bare slopes of red soil in south part of China was mainly caused by type A. During the process of individual rainfall, soil erosion on bare slopes of red soil was mainly affected by the joint influences of rainfall intensity and amount, the former resulted in erosion change directly through the closest index - maximum 30 min rainfall intensity, the latter had indirectly effects by changing the former. With the increase of maximum 30 min rainfall intensity, soil erosion occurred remarkably different properties at rainfall intensity of 15 mm/h. The soil erosion increased relatively slowly before that critical point but increased rapidly after the point. Generally, soil erosion intensity increased with the amount of rainfall, whereas the erosion ability of different rainfall types varied, which followed the pattern: type A > type B > type C. Excluding the influence of rainfall happened in earlier time, the erosion intensity on bare slopes of red soil had a significant power function relationship with both the maximum 30 min rainfall intensity and the rainfall amount, however, the exponents and coefficients were different when different rainfall types occurred. It has been proven that this model is not only best for intensive rainfall with short duration, but also runs very well while applying to erosion prediction for all three rainfall types - the values of multiple correlation coefficient and the Nash-Suttclife efficiency coefficient were both above 0.9, and the average relative error was less than 25%. In conclusion, this study is very helpful both in exploring further red soil erosion laws and in process-based prediction model establishment.