In this study, logistic regression, a multivariate analysis method, was used for landslide susceptibility
analysis of the Chi-Sun watershed. After extracting and selecting environmental factors, elevation, slope, terrain
roughness, slope roughness, normalized difference vegetation index (NDVI), wetness index, and accumulated
rainfall were selected as causal factors of landslides. An inspection of the distribution of the landslides caused by
Typhoon Morakot revealed that the landslides were located in medium-high and high susceptibility areas of the upstream
watershed. The predicted susceptibility values show a good agreement with the observed landslides, indicating
that both the model and the factors used in the model are valid and that the simulation results match the Typhoon
Morakot event. In addition, past NDVI values obtained by analyzing SPOT-5 satellite images suggest favorable vegetation
recovery in the watershed landslide areas, except in the upstream areas. This study focused on providing suitable
recovery and planning strategies around the riparian areas of the upstream watershed. The effects of vegetation
recovery on landslide susceptibility values were evaluated using the developed model. Landslide hazard areas were
analyzed and mapped on the basis of vegetation recovery for various buffer ranges.
Key Words : Landslide susceptibility, logistic regression, vegetation recovery