Rapid Detection of Landslides Using Sentinel-2 NDVI Change and DEM Metrics on Google Earth Engine

Landslides pose a persistent threat to infrastructure and communities across the rapidly changing Himalayan landscape. Despite the advances in remote sensing, rapid and accurate landslide mapping remains limited due to complex topography, frequent cloud cover and the need for updated inventories to support forecasting models. In this study, a potential landslide detection method was implemented on Google Earth Engine (GEE) using multi-temporal Sentinel-2 imagery and terrain-based masking. A buffer region in Himachal Pradesh was analysed using satellite observations acquired between July and September 2022. Cloud-filtered image pairs were processed to compute NDVI for each date, and significant vegetation loss was used as a proxy for fresh slope disturbances. Terrain parameters derived from the SRTM DEM, specifically slope and surface roughness, were applied to exclude flat or stable areas and enhance the specificity of detection. Pixels showing a substantial decline in NDVI on steep, rugged terrain were automatically flagged as potential landslides and exported as geolocated point features. The results demonstrate that multispectral NDVI change analysis can rapidly highlight areas of probable slope failure within the monsoon season. The validation against published research and news reports demonstrated strong spatial agreement between detected zones of ground displacement and the NDVI-based outputs. Therefore, this confirms the effectiveness of the proposed method in capturing event-scale landslide patterns across Himalayan landscapes. The study presents a fast, scalable and operationally practical method for preliminary landslide screening. Also, it provides valuable support for the growing need for machine-learning-based susceptibility modelling and early warning systems.