Distinguishing compound and cumulative hazards using machine learning and fuzzy logic in multi-hazard susceptibility...

Area of Science:

• Geosciences
• Environmental Science
• Data Science

Background:

• Mountainous regions face diverse natural hazards like floods, avalanches, rockfalls, and landslides.
• Accurate susceptibility mapping is crucial for effective spatial planning and risk reduction.

Purpose of the Study:

• To conduct an integrated multi-hazard susceptibility assessment in a mountainous region of northern Iran.
• To compare the performance of machine learning models for single-hazard susceptibility.
• To develop and evaluate a fuzzy logic-based framework for multi-hazard integration.

Main Methods:

• Applied Artificial Neural Network (ANN), Random Forest (RF), and Support Vector Machine (SVM) for single-hazard modeling.
• Utilized 21 topographic, climatic, geological, land-cover, and proximity variables at 30m resolution.
• Implemented a Fuzzy Logic integration framework with AND, OR, and GAMMA operators for multi-hazard assessment.

Main Results:

• RF models showed highest performance for flood and rockfall; SVM performed well for avalanche and landslide.
• Fuzzy logic integration enhanced multi-hazard prediction accuracy.
• The AND operator identified high-confidence compound hazard hotspots, while GAMMA provided broader cumulative susceptibility zones.

Conclusions:

• The study demonstrates the effectiveness of machine learning and fuzzy logic for integrated multi-hazard susceptibility assessment.
• Fuzzy integration, particularly with the AND operator, offers improved predictive reliability for complex mountainous environments.
• The framework provides a robust tool for regional hazard assessment and spatial planning.