Validation of landslides detection capability using multi-temporal SAR images : A case study on the heavy rainfall in northern Kyushu in July, 2017

Kumiko YAMASHITA, Joko KAMIYAMA, Yamato SUZUKI, Tomoyuki NORO, Jun SUGIMOTO, Takashi SHIBAYAMA and Toshiaki UDONO

Abstract

We verified the detection capability of landslides using SAR images taken before and after the heavy rainfall in northern Kyushu in July, 2017. We have successfully detected about 70% of the landslides that occurred on the slopes facing away from the radar. The landslide detection accuracy tends to be higher where the local incidence angle is large. If these conditions are met, we are able to detect a relatively small landslide with the area of about 2,500m2. This is a quarter area of a landslide that was detected in previous studies. By using multi‐temporal SAR images, the changes on the ground surface conditions can be detected with 90% accuracy.

Key words

synthetic aperture radar (SAR), sediment disaster, image interpretation, detection capability, slope characteristic