Investigating Limits in Exploiting Assembled Landslide Inventories for Calibrating Regional Susceptibility Models: A Test in Volcanic Areas of El Salvador

This research is focused on the evaluation of reliability regional landslide susceptibility models obtained by exploiting inhomogeneous (for quality, resolution and/or triggering related type and intensity) collected inventories for calibration. At a large-scale glance, merging more can result in well-performing hiding potential strong predictive deficiencies. An example limits that such kinds display given study, which was carried out large sector coastal area El Salvador, where an apparently model (AUC = 0.87) regressing dataset through multivariate adaptive regression splines (MARS), including five from volcanic areas (Ilopango Coatepeque caldera; San Miguel, Vicente Volcanoes). A multiscale validation strategy applied to verify its actual skill local base, bringing light loss power model, with lowering AUC (20% average) effects terms sensitivity specificity.