Abstract: In view of the large uncertainties embedded in the physical and mechanical parameters of the dynamic numerical model, a quantitative method to assess exceedance probability of landslide run-out distance based on Latin Hypercube Sampling(LHS) is proposed to improve the reliability of landslide slide prediction results.In the proposed method, the input parameters are considered as random variables with probability distribution functions; the LHS is used to generate stratified samplings for the random variables, and the dynamic numerical model is used to compute the landslide run-out distance corresponding to each random sample; and finally, the exceedance probability of landslide run-out distance is calculated by constructing limit state functions for different run-out distance threshold values.The proposed method is applied to conduct a probabilistic assessment of the run-out distance for Dabaozi landslide, and the 95% confidence interval of the run-out distance is computed as196 m, 302 m.According to the run-out distance-exceedance probability curve, the potential affected area of the landslide is classified into five categories(i.e., extremely high, high, medium, low and extremely low) based on different threshold values of exceedance probability(i.e., 50%, 10%, 1% and 0.1%).The calculation results show that the actual run-out distance of landslide is within the 95% confidence interval, and the actual threat range is also within the classified high-extremely high danger zone.The results show that the evaluation results are reasonable and the landslide movement risk evaluation based on the distance exceeding probability is also proved effective.The proposed method may provide new ideas for landslide run-out distance assessment and for hazard zoning, which has important theoretical significance and practical engineering value.