Multi-hazard spatial modeling via ensembles of machine learning and meta-heuristic techniques

Area of Science:

• Geosciences and Environmental Science
• Disaster Risk Management
• Computational Intelligence

Background:

• Natural disasters pose significant global risks, necessitating integrated hazard assessment.
• Kermanshah Province faces substantial threats from earthquakes, floods, and landslides.
• Effective disaster mitigation requires accurate susceptibility mapping for multiple hazards.

Purpose of the Study:

• To design and implement a multi-hazard map (MHM) for Kermanshah Province, Iran.
• To evaluate the performance of ensemble SWARA-ANFIS-PSO and SWARA-ANFIS-GWO models for flood and landslide susceptibility mapping.
• To integrate earthquake hazard data (Peak Ground Acceleration - PGA) with flood and landslide susceptibility for a comprehensive MHM.

Main Methods:

• Generated flood and landslide inventory maps and identified influencing factors (altitude, slope, rainfall, lithology, land use, etc.).
• Employed the SWARA (Stepwise Weight Assessment Ratio Analysis) method for factor weighting.
• Utilized the Adaptive Neuro-Fuzzy Inference System (ANFIS) machine learning algorithm, optimized with Particle Swarm Optimization (PSO) and Grey Wolf Optimizer (GWO) meta-heuristic algorithms.
• Validated model performance using Receiver Operating Characteristics (ROC), Root Mean Square Error (RMSE), and Mean Square Error (MSE).

Main Results:

• The SWARA-ANFIS-PSO model achieved the highest accuracy for flood susceptibility (ROC=0.936, RMSE=0.346, MSE=0.120).
• The SWARA-ANFIS-PSO model also showed excellent performance for landslide susceptibility (ROC=0.894, RMSE=0.410, MSE=0.168).
• A comprehensive multi-hazard map was created by combining the best susceptibility maps and the PGA map for earthquakes.

Conclusions:

• The SWARA-ANFIS-PSO ensemble model is highly effective for generating accurate flood and landslide susceptibility maps.
• The developed multi-hazard map provides a crucial tool for urban planning and disaster risk reduction in Kermanshah Province.
• Integrated multi-hazard mapping is essential for enhancing resilience and promoting sustainable development in hazard-prone regions.