Landslide disasters are affected by different kinds of factors. Choosing appropriate environmental indicators to investigate landslide causes and predicting potential locations would contribute to disaster management and prevention in the future. In this study, Si-de Bridge, which is located at the Qi-Shan Creek of Kaohsiung County, was used as the outlet of a watershed which suffered heavy damage during Typhoon Morakot. The landslide area was extracted from two-date SPOT-5 images with the image subtraction method. DEM satellite images were applied to calculate some environmental factors. Then, through principal component analysis (PCA), we analyzed the
important factors and built a discriminant analysis (DA) model to explore the relationship between environmental indices and landslides and generated a potential landslide map. The results showed that the extracted three principal component axes were mainly vegetation index, slope and difference from mean elevation; elevation, slope, NDVI, slope roughness, topographic wetness index and deviation
from mean elevation were used to build the DA model, with classification accuracy up to 70%. According to the landslide susceptibility map, the areas with higher topographic wetness index and lower slope roughness had higher landslide susceptibility.
Key Words : Landslide, environmental indices, susceptibility classification, remote sensing.