Abstract: Outburst floods caused by the breaching of landslide dams pose severe threats to lives and properties downstream.Rapid and accurate prediction of peak breach flow is vital for scientific assessment of the potential disaster consequences after landslide dam breaching and emergency response.Based on the investigation data of landslide dam breach cases,scholars worldwide have proposed a series of parametric models to predict the peak breach utilizing regression analysis.However,the existing parametric models cannot consider the effect of the grain composition of landslide dam deposits on the peak breach flow.Based on 86 worldwide landslide dam cases with grain-size distribution curves,the grain composition characteristics of three types of landslide dams(soil type,mixed type,and block stone type) are analyzed,and quantitative classification indexes of landslide dam types are proposed.Using experimental test results and field breach cases with measured data,the relationships between landslide dam type,morphology,dammed lake volume,and peak breach flow,dam height,dam width,dammed lake volume,and self-defined erosion factor considering the grain composition are selected as input parameters.A rapid prediction model for the peak breach flow of landslide dams is developed using multiple regression analysis.Five recent landslide dam breach cases are adopted to verify the proposed model in this study,and a comparison study with the existing typical parametric models is conducted.The calculation results show that the proposed model’s relative errors of peak breach flow are all within ±20%.Overall,the proposed model has several advantages,such as quick computation time and better accuracy compared to previous models.