Landslides pose significant global hazards, especially in regions with heavy rainfall, active tectonic processes, and rugged terrain, impacting lives and infrastructure. Remote sensing, particularly Synthetic Aperture Radar (SAR), plays a crucial role in landslide detection, offering data such as SAR amplitude, SAR Interferometry (InSAR), and SAR Polarimetry (PolSAR) for displacement and land cover change analysis. While SAR provides wide-area detection capabilities, challenges exist in tropical mountainous regions where SAR application is hindered by geometric distortions (e.g., layover, foreshortening). High-resolution satellite data, though commonly used, can be limited operationally due to costs and interpretative demands, making SAR-based approaches a valuable alternative. This study explores landslide detection using Sentinel‑1 amplitude, polarimetric decomposition, and coherence within a Deep Learning (DL) environment. Specifically, it evaluates the application of the U‑Net model in the context of the Colombian Andes, focused on the “La Liboriana” river basin in Salgar, a region affected by intense rainfall and landslides in 2015. The U‑Net model was trained and validated on post-and-pre event data from “La Liboriana” and tested for transferability in the “La Argelia” basin, which shares similar climatic and geomorphologic conditions. Three datasets were developed: Dataset 1 for post-event training/validation, Dataset 2 incorporating pre- and post-event differences, and Dataset 3 for inference in the secondary basin. The U‑Net model achieved high performance (F1-score >0.75) with VV polarization and entropy, demonstrating SAR’s capability to capture geomorphological anomalies and landslide concentration patterns. Integrating DL models with high-resolution satellite imagery and SAR data enables detailed space-time mapping of landslides in complex tropical mountain terrains, supporting rapid emergency response, infrastructure assessment, and improved prediction models for vulnerable communities. © Deutsche Gesellschaft für Photogrammetrie, Fernerkundung und Geoinformation (DGPF) e.V. 2025.