Determining Flood-Prone Areas Using Machine Learning Models in the Shahrestank Watershed Area of Khosef City

Flood is one of the most destructive natural phenomenon that occur worldwide, resulting in damage to properties, infrastructure, and even loss of life. This research aims to create a zoning map of flood-prone areas in the watershed of Khosuf County, using machine learning algorithms. In this study, 8 influential parameters and random forest (RF), boosted regression tree (BRT), and support vector machine (SVM) were used to identify areas at susceptibility of flooding in the study area. A total of 81 flood-prone zones were identified. Randomly, 70% of the data were allocated for the modeling process, and the remaining 30% were used for evaluating the generated models. The results of parameter importance analysis indicated that the drainage density parameter had the most significant impact on flood susceptibility. Furthermore, in determining flood-prone areas, the random forest model, boosted regression tree model, and support vector machine showed higher accuracy with receiver operating characteristics (ROC) curves of 0.99, 0.98, and 0.95, respectively. The study area was divided into four categories of flood susceptibility: very high, high, moderate, and low. Overall, the watershed is considered to have a moderate to high flood susceptibility level.