Assessment of alternative forest road routes and landslide susceptibility mapping using machine learning

Background: Forest roads are among the most basic infrastructure used for forestry activities and services. To facilitate increased amount of biomass harvesting adequately, existing road network may require modifications to allow forest transportation within units that not yet accessed by roads. The construction a can trigger landslides, so necessary constraints should be considered when is being planned preclude such problems. Landslide Susceptibility Mapping (LSM) has become an integral part growing process machine learning (ML), providing more effective platform practitioners, planners, decision-makers. This study aims reveal suitable alternative routes road, especially in areas susceptible provide tool Results: For this purpose, two models were developed through ML: Logistic Regression (LR) Random (RF). Elevation, slope, aspect, curvature, Topographic Wetness Index (TWI), Stream Power (SPI), distance from fault, stream, lithology as main landslide susceptibility factors these models. best model was obtained RF approach with Area Under ROC Curve (AUC) value 81.9%, while LR 78.2%. LSM data base, CostPath analysis. Conclusion: It been shown ML methods positively contribute decision-making calculations studies determine network.