Window-Based Morphometric Indices as Predictive Variables for Landslide Susceptibility Models

The identification of areas that are prone to landslides is essential in mitigating associated risks. This usually achieved using landslide susceptibility models, which estimate likelihood given local terrain conditions and the location known past events. Detailed databases covering different conditioning factors paramount producing reliable maps. However, thematic data from developing countries scarce. As a result, models often rely on morphometric parameters derived widely-available digital elevation models. In most cases, simple parameters, such as slope, aspect, curvature, computed moving window 3 × pixels, used. Recently, use window-based indices an additional input has increased. These user-defined observation size. this contribution, we examine influence size when core predictive variables for assessment. We variety include calculated with sizes, compared capabilities reliability resulting predictions. All based random forest algorithm. results improved significantly each index was meaningful (AUC-ROC 0.89 AUC-PR 0.87). sensitivity analysis highlights both highly-informative windows impact their selection model performance. also stress importance evaluating while adapted metrics performance reliability.