Xiaolin, a small village in southern Taiwan, was completely buried by a deep-seated landslide during the extreme rainfall of Typhoon Morakot in 2009, and 500 people were killed. This disaster caused great public concern. This study examines the rainfall-induced landslide characteristics in the area and their correlations with various susceptibility factors. We emphasize the importance of producing an accurate landslide inventory using visual interpretation and propose a new systematized set of subcategories of susceptibility factors, especially stream factors. We used satellite images, digital elevation models, and landslide inventories from before and after the disaster to perform spatial analysis using geographic information system (GIS) tools. Of the topographical factors considered, only slope aspect was important, and the landslides bore little relation to tectonic lines. The stream factors had the strongest influence on these landslides. Almost 100 % of the landslide area was related to various types of erosion caused by the stream network. Most of the large-scale landslides were induced by headward erosion, and all occurred on dip or oblique slopes. Thus, a combination of dip or oblique slope types and stream factors resulted in more landslides, which suggests that landslides are often induced by a combination of multiple factors.