Landslide susceptibility zonation using statistical and machine learning approaches in Northern Lecco, Italy

This study deals with landslide susceptibility mapping in the northern part of Lecco Province, Lombardy Region, Italy. In so doing, a valid inventory map and thirteen predisposing factors (including elevation, slope aspect, degree, plan curvature, profile distance to waterway, road, fault, soil type, land use, lithology, stream power index, topographic wetness index) form spatial database within geographic information system. The used predictive models comprise bivariate statistical approach called frequency ratio (FR) two machine learning tools, namely multilayer perceptron neural network (MLPNN) adaptive neuro-fuzzy inference system (ANFIS). These first use non-landslide records for comprehending relationship between occurrence factors. Then, values are predicted whole area. accuracy produced maps is measured using area under curve (AUC) according which, MLPNN (AUC = 0.916) presented most accurate map, followed by ANFIS 0.889) FR 0.888). Visual interpretation maps, FR-based correlation analysis, as well importance assessment factors, all indicated significant contribution road networks crucial landslide. Lastly, an explicit formula extracted from implemented model convenient approximation value.