The implementation of landslide probability analysis was undertaken in order to evaluate the impact of landsliding on closures of major mountain road networks in Guoshin Township in central Taiwan. To achieve this objective, an event-based landslide probability analysis method was adopted to establish a landslide prediction model using a set of training data from the landslides triggered by Typhoon Mindulle in July 2004. Landslide causative factors and triggering factors were selected in a logistic regression scheme so that the criteria for successfully distinguishing landslides from non-landslides were established. Landslide occurrence probability was mapped in the whole study area as well as along the road route. Locations of high potential for landslide occurrence were thus highlighted along the road route and were proposed for road closure during typhoon events. At last, the proposed locations for road closure were validated by historical road closures caused by subsequent typhoons after Typhoon Mindulle. Validation results show that the proposed model could be used in predicting road closures due to storm-induced landslides.