Detection of 2011 Sikkim earthquake-induced landslides using neuro-fuzzy classifier and digital elevation model

Abstract

In this paper, a semiautomatic method for landslide detection from satellite images and digital terrain information using generalized improved fuzzy Kohonen clustering network (GIFKCN) classifier is presented. The proposed method classifies the pre- and post-landslide images using the GIFKCN classifier which is trained using spectral indices such as normalized difference vegetation index, normalized difference building index and normalized difference water index. The changes in the vegetation class are identified using the pre- and post-classified images. Generally, landslides result in loss of vegetation; thus, using this property, candidate landslides are identified. Finally, false positives are removed using a rule set created from DEM derivatives slope and aspect. The proposed method is applied on Landsat 5 and Advanced Land Imager EO-1 satellite images to detect earthquake-induced landslides that occurred in Sikkim state of India due to the September 18, 2011, earthquake of magnitude M w = 6.9. The terrain information used is ASTER Global Digital Elevation Model of the area. The accuracy assessment of the method is done, and the results show that the landslides are identified and classified efficiently.