GIS-based optimization framework for shelter site selection and population allocation under multi-hazard scenarios

Mountainous regions face heightened flood and landslide hazards, necessitating efficient population relocation strategies. This study develops an integrated optimization framework for shelter site selection and population allocation in multi-hazard environments. Flood susceptibility was evaluated using six indicators through GIS spatial analysis, while landslide susceptibility was assessed using nine environmental factors with a Random Forest model, validated through 5-fold cross-validation (mean R2 = 0.9282). The two susceptibility maps were integrated through fuzzy overlay to delineate multi-hazard zones and identify safe candidate shelters. Shelter capacities were estimated from available area data and standardized per-capita requirements. A dual-objective optimization model was then formulated to minimize the number of activated shelters and the maximum evacuation distance. Applied to Luchuan County, China, the model produced two contrasting solutions: prioritizing shelter minimization activated 22 shelters with a maximum evacuation distance of 13.19 km, whereas prioritizing evacuation distance activated 24 shelters and reduced the maximum distance to 12.16 km. These findings demonstrate the framework’s effectiveness in generating balanced and practical relocation strategies. Future improvements for the study area include integrating heterogeneous evacuee characteristics, refining shelter capacity estimates with detailed facility data, and incorporating temporal weighting to better capture evolving climate-driven hazard patterns.