Landslide detection using satellite remote sensing imagery

Landslide detection using satellite remote sensing images has been widely studied. This type of applications often involves either change detection or multi-spectral image classification methodologies. If there is only one set of satellite image, the change detection method has limited use. Collecting and analyzing training area data for image classification are costly and time consuming. This study, therefore, utilize only one SPOT satellite image data for estimating the normalized difference vegetation index (NDVI), and to segregate vegetated and non-vegetated areas of the Ta-An River Basin in Central Taiwan. Slope factor and textural feature are then used to identify the landslide area. Results indicate that the accuracy of landslide detection using NDVI alone is about 88%. Using NDVI with slope factor and textural feature increases overall accuracy to 97%. This study successfully demonstrates the capability of using one set of remote sensing image to map landslide area in a large river basin.