A Machine-Learning-Based Multi-Hazard GIS-AHP Framework for Wind Turbine Siting under Earthquake-Landslide Coupling

Abstract

This study presents a machine-learning-based multi-hazard geographical information system (GIS)-analytical hierarchy process (AHP) framework for wind turbine siting that explicitly accounts for the coupled effects of earthquake and landslide hazards. The primary innovation lies in the development of a conditional weighting algorithm that integrates machine-learning-derived hazard assessments with structural engineering logic. Landslide susceptibility is first modeled using a random forest classifier trained on a comprehensive inventory of historical landslide data and 12 geo-environmental conditioning factors, producing a high-resolution susceptibility map with excellent predictive performance (AUC = 0.86). Feature importance analysis indicates that slope, hydrological indices, and geological conditions are the dominant controls on landslide occurrence. This data-driven map is then integrated with earthquake hazard zones and additional environmental and technical constraints within a GIS-AHP framework to generate a comprehensive wind turbine suitability assessment. Results show that explicitly accounting for earthquake-landslide coupling leads to a nearly 20% reduction in high and very high suitability areas, accompanied by an expansion of low and moderate suitability zones, highlighting the limitations of single-hazard planning approaches. The main contribution of this study lies in advancing renewable energy planning through the explicit integration of interdependent natural hazards, demonstrating how earthquake-resistant foundation strategies can simultaneously mitigate landslide risks.