This research investigates the impact of climate change on flood susceptibility assessment using four advanced deep learning models; Deep Learning Neural Network (DLNN), Artificial Neural Network (ANN), Deepboost, and XGBoost; across different climate projections for the year 2100. The study incorporates climate scenarios under three Shared Socioeconomic Pathways (SSPs); SSP 245 (moderate emissions), SSP 370 (high emissions), and SSP 585 (extreme emissions). Each model demonstrates distinctive strengths in flood risk prediction, with XGBoost offering a balanced and precise classification of flood-prone areas, while DLNN and ANN tend to highlight more extensive high-risk zones. Deepboost adopts a conservative approach, minimizing false positives but potentially underestimating the extent of flood susceptibility. Variables importance analysis shows that rainfall, slope, and land use/land cover (LULC) are critical factors influencing flood risk. The climate projections from the four models—ACCESS, CMCC-ESM2, MIROC6, and MRI-ESM2 show a clear trend: as emissions increase from SSP 245 to SSP 585, flood risks escalate significantly. Under SSP 585, regions considered moderate risk may face severe flood susceptibility by 2100. Under the SSP 370 scenario, flood susceptibility zones expand significantly, with many areas shifting from Moderate or Low to High or Very High risk, highlighting increased flood threats under intermediate climate change. In the more extreme SSP 585 scenario, widespread regions face elevated flood risks, indicating severe future impacts without strong emission reductions. The findings highlight the need for robust flood adaptation strategies, with XGBoost offering a balanced approach for urban planning and DLNN and ANN providing detailed high-risk zone identification for targeted mitigation efforts. This underscores the urgency of global emission reduction efforts to mitigate the worst effects of climate change.
Paramanik et al. (Fri,) studied this question.