Post-landslide vegetation recovery assessment in Dumbarawatta, Sri Lanka using S2DR3 model - based downscaling of sentinel-2 imagery

Abstract

Landslides are among the most destructive natural hazards, causing significant damage to infrastructure, ecosystems, and human settlements. Monitoring vegetation recovery after such events is critical for ecological restoration and effective hazard management. Given the challenges of conducting field surveys and UAV-based monitoring in inaccessible terrain, remote sensing approaches can be an efficient alternative. Sentinel-2 optical imagery, with a native spatial resolution of 10 meters and minimal cloud interference, provides a valuable resource for monitoring vegetation dynamics. However, to obtain more detailed and accurate spatial information, alternative approaches such as downscaling are required. In this study, the Sentinel-2 Deep Resolution 3.0 model (S2DR3) is used to downscale imagery from 10m to 1m resolution. Leveraging this enhanced resolution, the Normalized Difference Vegetation Index (NDVI) is applied to monitor vegetation disturbance and regrowth over time following a rainfall-induced landslide that occurred on 4 June 2021 in Dumbarawatta, Sri Lanka. The analysis is conducted using the Google Earth Engine platform, offering a scalable and cost-effective methodology for post-landslide environmental monitoring. This approach supports informed decision-making in landscape recovery and land management, particularly in complex and mountainous terrains. The S2DR3 downscaled product detected up to 5% more vegetation cover than Sentinel-2, enhancing accuracy in post-landslide recovery assessments.