Automatic detection of landslides from SAR images: Application to the 2011 Kii landslides

We analyzed TerraSAR-X images and found a good combination of filters and enhancement technique for the automatic landslide detection. We were successfully able to suppress noise in the image, identified landslide signals, and classify landslides from other surface changes. With a parameter set focusing on the completeness of detection, we achieved 98 % of detection from landslide catalog, but we have 50 times more false detection. With another parameter set focusing on the accuracy of detection, we achieved 80 % of detection from the catalog, and the signal-to-noise ratio was 27 %. Users can easily change the parameters depending on user's requirement. Quick detection of a landslide occurrence is very important for prompt emergency information, rescue efforts, and mitigation of further damage such as collapse of a landslide dam. Our approach proposed here will contribute to the rapid assessment of landslide hazards.