A NDVI Based Approach To Detect The Landslides By Using Google Earth Engine

Abstract

Detection of landslide-prone areas plays an important role in planning urban connectivity like roads, bridges, etc. Landslides are generally caused by a variety of factors, the most important of which is rainfall. In this paper, the detection is carried out in four taluks of Chikkamagaluru district, namely Koppa, Sringeri, Mudigere, and Narashimarajpur; these four taluks are located in the Western Ghat region. Landslides are primarily caused by heavy rainfall during the monsoon season. For the detection of landslides, Sentinel optical and SAR data are used because of their 10metre resolution and revisiting period of two to five days. The entire methodology for detecting landslides is carried out in Google Earth Engine due to its large collection of data, which aids in multi-temporal studies. This paper attempts to investigate the capabilities of remote sensing and GIS techniques in the detection of landslides. For the detection of landslides, Normalized Difference Vegetation Index (NDVI) is used for Sentinel-2 data and the SAR backscatter change approach is used for Sentinel-1 images, and I thresholding is applied to both methods to detect areas where landslides had occurred. The main thing is that no previous landslide inventory data is used for detection. The previous landslide inventory is used for validation purposes only. Finally, the performance of both approaches was compared using accuracy assessment properties such as overall accuracy and kappa coefficient to determine which approach is superior. © 2023 IEEE.