Assessment of Sentinel-1 and Sentinel-2 Data for Landslides Identification using Google Earth Engine

Quick and accurate landslide mapping is crucial for emergency response, disaster mitigation, and increasing the understanding of landslide events. Satellite remote sensing data, such as Synthetic Aperture Radar (SAR) and optical imagery, have been used for identifying landslides. SAR data possess several benefits, such as the ability to penetrate the clouds, operate day and night, and regular revisit time. However, processing SAR data require large amounts of data downloaded to the local system to be processed on a local computer. This study explored the SAR-based amplitude change detection approach using multi-temporal and multi-orbit (ascending and descending) Sentinel-1 data to identify landslides utilizing the cloud computing platform of Google Earth Engine (GEE), which aimed to assist in rapid response in mapping and inventorying landslide location data. This study took two case studies in the Lebak Regency, Banten Province, and Masamba Regency, South Sulawesi Province. Additional topographic data was used to filter out the flat areas which were unlikely to suffer from landslides. For comparison, this study also identified landslides using Sentinel-2 supported optical images on the GEE platform. In this study, the overall value of landslide mapping accuracy is 97.66 on Sentinel-2 and 68.75 on Sentinel-1 for the case study in the Lebak Regency, Banten Province. In comparison, the case study in the Masamba Regency, South Sulawesi Province, shows the overall value of landslide mapping accuracy of 96.69 in Sentinel-2 and 87.6 on Sentinel-1. Thus, the benefits of Sentinel-2 over Sentinel-1 are associated with the resulting accuracy. Still, optical-based images have a weakness in quick response efforts because optical images require sunlight and cloud/shadow-free conditions to spot landslides accurately. © 2021 IEEE.