[Objective] The main objective of this paper is to quantitatively study soil erosion status of waste residue (such as spoil and slag) from production and construction project under natural rainfall conditions, and provide reference for soil erosion prediction in Jiangxi Province.[Methods] Based on K-means method, the local rainfall was classified into four types:A(maximum rainfall, long rainfall duration, moderate rain intensity, low frequency), B(medium rainfall, medium rainfall duration, light rain intensity, high frequency), C(large rainfall, long rainfall duration, heavy rain intensity, low frequency), and D(medium rainfall, short rainfall duration, extremely heavy rain intensity, high frequency medium rainfall). In addition, the 8 kinds of spoil ground runoff test plots were chosen to carry out quantitative research on soil erosion of spoil in the National Soil and Water Conservation Ecological Science and Technology Park of De'an County, Jiangxi Province.[Results] Through the observation of 43 erosive rainfall in 1 year, the results show that:① The rainfall pattern of A and D are the main types, B has the least influence, and C falls in between A and D. ② There is significant correlation between runoff and sediment for red sandstone(HSY),stencil red clay(SJHT), purple soil (ZSYY), coal gangue(MGS), slag soil(FZT). ③ The rain pattern above large rainfall will lead to the highest sediment yield from the single abandoned waste residue.[Conclusion] When the bare abandoned soil and slag are predicted or monitored in production practice, different types of rainfall should be selected to carry out according to the actual demand, and the largest amount of large-scale rainfall over 50 mm is predicted or observed so as to improve the prediction accuracy and reduce the monitoring cost and improve the monitoring efficiency.