Improvement of flood susceptibility mapping by introducing hybrid ensemble learning algorithms and high-resolution satellite imageries

Flood, a dangerous hydro-geomorphic hazard, is one of the most critically applied science research issue. The restoration and recovery are costly and can interrupt communities’ sustainable growth after the extensive flood. Flash floods (FF) are a frequent natural disaster that causes significant casualties and disrupts economic growth in the Brahmaputra River Basin (BRB). Hence, the flood susceptibility modeling of BRB is imperative. The study uses six machine learning (ML) techniques (three stand-alone such as artificial neural network (ANN), fuzzy logic (FL), and random forest (RF), and three hybrid ensemble models (HEMs) including ANN-FL, FL-RF, and RF-ANN) to appraise flash flood Susceptibility (FFS) prediction in BRB considering 16 flash flood susceptibility factors. Area under the curve (AUC), ROC curve, confusion matrix (CM), and Friedman test are applied to assess the performance of the models. Results for the training and testing datasets showed that all HEMs models for FFS prediction in the BRB outperformed the stand-alone models. The RF-ANN has the best prediction ability of all models because the RF meta-classifier improves the ANN model’s base-classifier precision. The RF-ANN model delineated 2908.46 km2 and 874.73 km2 areas as very high and high flood susceptible zones, whereas 995.99 km2, 702.48 km2, and 10,127.57 km2 areas were predicted as moderate, low, and very low flood susceptible zones. Slope, water, vegetation, PrC, aspect, and rainfall all make the BRB sensitive to FF, as per the analysis of InGR and PCM. This work’s accuracy of the ML HEMs used for FFS mapping is promising. Furthermore, the findings of this study may be valuable for flood prevention and management to deal with the current uncertainties and more precisely identify numerous characteristics that impact FFS. This research is helpful for policymakers because it provides information that could be utilized to develop measures to lessen the adverse effects of FF.