Detecting rainfall-induced landslides with Sentinel-1 SAR: Evidence from the Andes

Abstract

Rainfall-induced landslides are among the main geodynamic processes reshaping the Earth’s surface and causing human and material losses worldwide. In Chile, their spatial analysis remains constrained by the scarcity of systematic inventories, particularly in the Andean Cordillera. In this context, this study presents a novel inventory of rainfall-induced landslides in the Biobío Region during 2023 and 2024, based on the analysis of changes in SAR backscatter from Sentinel-1 imagery processed in Google Earth Engine (GEE), complemented by visual interpretation of Sentinel-2 optical images, GIS-based mapping, and selective field validation. Four intense rainfall events were identified; however, only two of them (June and August 2023) triggered detectable landslides, totaling 55 processes dominated by debris flows and debris avalanches. In contrast, no landslides were detected in association with the 2024 event, which was characterized by lower rainfall intensity in the study area. The strong contrast in landslide occurrence between events provides clear evidence of differences in rainfall magnitude and impact, highlighting the role of rainfall intensity and temporal structure in controlling slope instability. In this sense, landslide occurrence can be interpreted as a direct geomorphic response to rainfall forcing, reflecting the effectiveness of precipitation in triggering slope failure. The results demonstrate the high capability of a multitemporal SAR-based approach to detect rainfall-induced landslides in mountainous environments characterized by dense vegetation cover and complex climatic conditions, validating its applicability in Andean territories. The resulting inventory provides a fundamental baseline for calibrating precipitation thresholds, improving susceptibility, hazard, and risk models, and strengthening monitoring systems under scenarios of intensifying hydroclimatic extremes associated with climate change. It also offers a valuable dataset for the development and training of machine learning models aimed at automated landslide detection and regional-scale hazard assessment.