Flood susceptibility zoning using machine learning improved by genetic algorithm

Due to the increase in the risk of floods, especially in the cities, and the emergence of human, financial, and environmental risks due to its increase, the flood zoning areas are of great importance. Therefore, in this study, flood susceptible areas in Birjand plain were tried to be zoned with the help of effective criteria. In this regard, the data-driven methods of support vector machine (SVM) and random forest (RF) were used in combination with genetic algorithm to zoning flood susceptible areas. Therefore, in order to implement and validate the mentioned models, 42 flood prone locations in the study area were extracted. In addition, 19 hydrogeological, topographical, geological and environmental criteria affecting flood susceptibility in the study area were extracted to be used to predict flood susceptibility map. Area under the curve (AUC) and a variety of other statistical indicators including coefficient of determination (R2) and Root mean square error (RMSE) were used to evaluate the performances of the models. The values of R2, RMSE and AUC obtained from the SVM-GA method were 0.9032, 0.2751 and 0.931, respectively, and the RF-GA method were 0.9823, 0.2321 and 0.914, respectively, which indicate the compatibility and The RF model is more accurate than the SVM model. The results also showed that the susceptibility of flooding in the central areas of the study area, due to lower altitude and slope angle, is higher than other areas.