Develop an Approach for Mapping an Accurate and Appropriate Flood Susceptibility for Quang Binh Province, Vietnam, Using Machine Learning Algorithms and Remote Sensing

Abstract Flooding is one of the natural catastrophes that causes significant damage to Vietnam year over year, particularly in low-lying areas along rivers and coastal regions. This study aimed to develop an approach for mapping an accurate and appropriate flood susceptibility for Vietnam’s condition, thereby identifying areas prone to high potential flooding, preventing and minimizing flooding damage. Quang Binh province (Vietnam), which experienced a historical flood in 2020, was the study area. Nine modern machine learning algorithm models (ANN, KNN, SVM, CART, Naive Bayes, GBM, RF, XGBoost, and LightGBM) were applied and evaluated. The models were trained on not only a dataset of storm-induced water level rise and historical flood, but also a set of 11 topographic and hydrological variables (Slope, Aspect, Curvature, Flow direction, Flow accumulation, Topographic wetness index, Land cover, Soil type, River density, Road density, and Distance to river). The results demonstrated that the RF, CART, LightGBM, and XGBoost models achieved superior performance with accuracy (above 98%, AUC > 0.9, and F1 > 0.7) and produced consistent flood interpretation findings from satellite images. Besides, despite a high accuracy, the GBM model witnessed a biased spatial distribution, emphasizing the significance of integrating quantitative evaluation with spatial validation. Additionally, decisive factors (Slope, Topographic wetness index, Distance to river, and River density) were revealed. This study successfully affirmed the potential of machine learning algorithms not only in analysing current flood conditions but also in simulating flood susceptibility under future sea level rise scenarios. These findings were expected to offer an essential approach for risk assessment, disaster management, and future planning.