Abstract: Prediction model for movement distance of rock avalanches has important theoretical significance and practical significance due to the characteristic of difficult detection, complex mechanism and high risk of gully-type rock avalanches. This study selects 38 typical rock avalanches induced by Wenchuan earthquake, and obtains their basic data based on the remote sensing and GIS technology combining with field investigation. Through correlation analysis, the results indicate that the influence factors of the maximum horizontal movement distance of rock avalanches (L) are landslide volume (V), the maximum vertical movement distance H, the elevation difference of slip source area (Hs), and the mean slope angle of the lower channel (β). Then a stepwise regression method is adapted to build optimal multiple regression model for the prediction of L. The validation shows that the multiple regression model is suitable for predicting rock avalanche hazardous zones in Wenchuan earthquake areas. The optimal multiple regression model and other internationally common-used prediction models are compared for the movement distance of landslides and debris flows. The result indicates that the forecast index system is suitable for the rock avalanches risk assessment in mountainous areas. The system considers the volume, drop height, the mean slope of the channel, with better goodness of fit index and physical meaning.