Enhancing flood mapping through ensemble machine learning in the Gamasyab watershed, Western Iran

Area of Science:

• Hydrology and Environmental Science
• Geospatial Analysis
• Machine Learning Applications

Background:

• Floods are increasing in frequency and severity, posing significant threats to human life and financial stability.
• Accurate flood susceptibility mapping (FSM) is essential for effective disaster risk reduction and management.
• Traditional methods often lack the precision needed for complex hydrological assessments.

Purpose of the Study:

• To evaluate machine learning (ML) techniques for flood susceptibility mapping (FSM) in the Gamasyab watershed, Iran.
• To compare the performance of Random Forest (RF) and Support Vector Machine (SVM) models, including ensemble approaches.
• To identify key factors influencing flood occurrence within the study area.

Main Methods:

• Utilized Random Forest (RF), Support Vector Machine (SVM), and ensemble models integrated with a Geographic Information System (GIS).
• Incorporated 10 effective flood-influencing factors and 82 historical flood locations.
• Applied resampling techniques (bootstrap, subsampling) for robust model training and testing.

Main Results:

• Elevation, slope, and precipitation were identified as primary drivers of flood susceptibility.
• The ensemble model significantly outperformed individual RF and SVM models, achieving an AUC of 0.9.
• The ensemble model demonstrated high accuracy with COR of 0.79 and TSS of 0.83.

Conclusions:

• Integrating ensemble ML models with GIS provides a powerful and effective tool for accurate flood susceptibility mapping.
• The study highlights the potential of advanced computational techniques for enhancing flood risk assessment and mitigation strategies.
• Findings offer valuable insights for watershed management and disaster preparedness in flood-prone regions.