Spatial modeling of flood susceptibility using machine learning algorithms

Floods constitute one of the most devastating and destructive natural forces in the world. They have a considerable impact on the economy and can result in significant loss of life. Several strategies, including studies by advanced data analysis methods, have been adopted to curb this phenomenon and ultimately limit the accompanying damage. In this study, four supervised models based on machine learning (ML) algorithms were used to map flood vulnerability in the Souss watershed located in southern Morocco. They include random forest, x-gradient boost, k-nearest neighbors and artificial neural network. Thirteen predisposing factors including aspect, curvature, digital elevation model (DEM), distance to rivers, drainage density, flow accumulation, flow direction, geology, land use, rainfall, slope, soil type, and topographic wetness index (TWI) were selected as inputs to achieve this. Four different models were developed for each ML algorithm based on variable selection and one-hot encoding. Overall, all the models of the four algorithms have an AUC score above 80% for the testing data, which means that they all performed very well. The ranking of the ML algorithms used by considering only the most efficient model of each algorithm is as follows: KNN (98.6%), RF (98.1%), XGB (97.2%), and NNET (95.9%). Finally, all the models were overlaid to identify points of agreement or disagreement. This study provides evidence of ML models being successful in mapping flood vulnerability. These findings can be beneficial, serving as an important resource in mitigating the impacts of floods in the highlighted vulnerable areas presented in the flood vulnerability maps.