Title: Application of GIS and Machine Learning to zoning flood risk from historical data in Vietnam
Authors: Nguyen Quynh Anh
Volume: 9
Issue: 9
Pages: 54-64
Publication Date: 2025/09/28
Abstract:
Viet Nam as the world's 3rd vertex of the pronounced Southeast Asian Water Triangle that feeds the mighty Mekong River experiences major debilitating flooding in the Central, Northwest, and Mekong Delta regions in the winter-spring months. They decided this as one of the first severe climatic hazards of the region. This study focuses on devising geo-hydrometeorological driven flood risk zones as geo-spatial constructs using records of historical flood events. This has been done using Random Forests for Machine Learning. 11 spatial and climatic variables (elevation, slope, TWI, curvature, distance from the river/drainage, river Curve Number, rainfall extremes, etc.) were processed geospatially through a GIS interface. Validation on Berlin's dataset achieved ~80% accuracy (AUC = 0.80) which is within the accepted range for informative filters, thus, confirming the model's reliability. The decentralized trained model has been implemented in an application that creates binary representations of flood events and scores local defense systems corresponding to the level of vulnerability to flooding. Results demonstrate the combined use of GIS and machine learning as a powerful conceptual and practical tool in systematic spatial planning, proactive alert systems, and climatological adaptations for the region. These models will next be exposed to the real-time flux of geo-spatial data from web-observations across the country to minimize disaster risk.