<p><strong>Abstract.</strong> In a high-rainfall, landslide-prone region in this tropical mountain region, a landslide database was constructed from high resolution satellite imagery (HRSI), local reports and field observations. The landslide data was divided into training (80%) and validation sets (20%). From the digital elevation model (DEM), scanned maps and HRSI, twelve landslide conditioning factors were derived and analysed in a GIS environment: elevation, slope angle, slope aspect, plan

more »

... urvature, profile curvature, distance to drainage, soil type, lithology, distance to fault/lineament, land use/land cover, distance to road and normalized difference vegetation index (NDVI). Landslide susceptibility was then estimated using the frequency ratio method as applied on the training data. The detailed procedure is explained herein. The landslide model generated was then evaluated using the validation data set. Results demonstrate that the very high, high, moderate, low and very low susceptibility classes included an average of 86%, 7%, 4%, 3% and 1% of the training cells, and 84%, 7%, 5%, 3% and 1% of the validation cells, respectively. Success and prediction rates obtained were 90% and 89%, respectively. The sound output has discriminated well the landslide prone areas and thus may be used in landslide hazard mitigation for local planning.</p>