Spatial Flood Forecasting Modeling under Lack of Data Using RS and Optimized Support Vector Machine: A Case Study of the Zahedan Watershed

Abstract

This research was conducted with the aim of producing flood susceptible area maps of the Zahedan catchment in Iran during lack of data using remote sensing. To determine the factors to be included in the model, factors identified in previous studies were prepared using remote sensing and were evaluated by the information gain ratio (IGR) method and the multicollinearity diagnostic test, and ultimately 10 factors with the highest IGR values were chosen as the most effective factors in the region. The flood inventory map was produced by processing the Sentinel-1 satellite data. The gathered data were used to map flood susceptibility maps with the SVM model optimized with IWO and ACO. The prediction accuracy of the models was evaluated in terms of root mean square error (RMSE), mean absolute error (MAE), and the area under the receiver operating characteristic (ROC) curve (AUC). While both optimization algorithms are effective in improving the performance of SVM, the hybrid of SVM with IWO shows the best performance in terms of statistical measures and Friedman test results. The results of the study confirm the good performance of the proposed models in spatial prediction of flood susceptible areas. Since over half of the urban lands of the city of Zahedan are at moderate to very high risk of flooding, this area needs more attention in terms of flood prevention and control measures.