Improving Landslide Detection on SAR Data Through Deep Learning

In this letter, we use deep learning convolutional neural networks (CNNs) to compare the landslide mapping and classification performances of optical images (from Sentinel-2) and synthetic aperture radar (SAR) images (from Sentinel-1). The training, validation, and test zones used to independently evaluate the performance of the CNN on different datasets are located in the eastern Iburi subprefecture in Hokkaido, where, at 03.08 local time (JST) on September 6, 2018, an Mw 6.6 earthquake triggered about 8000 coseismic landslides. We analyzed the conditions before and after the earthquake exploiting multipolarization SAR as well as optical data by means of a CNN implemented in TensorFlow that points out the locations where the landslide class is predicted as more likely. As expected, the CNN runs on optical images proved itself excellent for the landslide detection task, achieving an overall accuracy of 98.96%, while CNNs based on the combination of ground range detected (GRD) SAR data reached overall accuracies beyond 95%. Our findings show that the integrated use of SAR data may also allow for rapid detection even during storms and under dense cloud cover and provides comparable accuracy to classical optical change detection in landslide recognition and detection.

Automated summary

In this study, deep learning convolutional neural networks (CNNs) were used to compare the mapping and classification performances of optical images and synthetic aperture radar (SAR) images in landslide detection. The results showed that CNNs based on optical images achieved an overall accuracy of 98.96% in landslide detection, while CNNs based on SAR data reached accuracies beyond 95% in ground range detection.