Predictive Modeling of Flood Susceptibility in Tetouan, Morocco Using Machine Learning Algorithms

Floods represent a significant natural hazard causing extensive damages. The research aims to demonstrate the robustness of employing Machine Learning (ML) models, namely Random Forest (RF), Support Vector Machine (SVM), Logistic Regression (LR), K-nearest neighbor (KNN), and Decision Tree (DT) to generate flood susceptibility maps for Tetouan city in Morocco. The methodology relies on a spatial dataset comprising 1000 samples, including eight conditioning factors: elevation, slope, distance to the river (DR), drainage density (DD), Land Use (LU), Stream Power Index (SPI), Topographic Witness Index (TWI), and Normalized Difference Vegetation Index (NDVI). These factors were extracted using remote sensing techniques. Performance comparisons of ML algorithms reveal that RF exhibited the highest accuracy and area under the curve (AUC) values, reaching 95%, thereby outperforming other models. The key findings of this study can serve as guidelines for authorities and hydrologists to proactively predict flood-prone areas and implement necessary measures to mitigate risks.