In this study, landslide potential models for Alishan Mountain Road in southwestern Taiwan were developed and assessed. We collected data on severe typhoon events that occurred in Taiwan between 2007 and 2011 and considered seven intrinsic and extrinsic factors. The dangerous value method and logistic regression were used to build the mountain road landslide potential models, which were calibrated using typhoon event data. The results indicated
that cumulative precipitation and the geological characteristics of the stratum were the most important factors in the models. The logistic regression model generated estimations of historical disaster locations with 96.76% accuracy.The calibrated landslide potential model was used to analyze and predict the distribution of areas with higher landslide potential on Alishan Mountain Road under heavy rain and extremely heavy rain conditions.
Key Words: Landslide potential model, Dangerous value method, Logistic regression model.