Landslide susceptibility assessment by bivariate methods at large scales: Application to a complex mountainous environment

摘要: Statistical assessment of landslide susceptibility has become a major topic research in the last decade. Most progress been accomplished on producing maps at meso-scales (1:50,000–1:25,000). At 1:10,000 scale, which is scale production most regulatory hazard and risk Europe, few tests performance these methods have performed. This paper presents procedure to identify best variables for through bivariate technique (weights evidence, WOE) discusses way minimize conditional independence (CI) between predictive variables. Indeed, violating CI can severely bias simulated by over- or under-estimating probabilities. The proposed strategy includes four steps: (i) identification response variable (RV) represent events, (ii) combination (PVs) neo-predictive (nPVs) increase statistical model, (iii) evaluation simulations appropriate tests, (iv) model expert judgment. study site north-facing hillslope Barcelonnette Basin (France), affected several types landslides characterized complex morphology. Results indicate that are powerful assess scale. However, method limited from geomorphological viewpoint when RVs PVs poorly informative. It demonstrated knowledge still be introduced models produce reliable maps.