Year: 2024
Type: article
Abstract: Among various natural hazards, those induced by hydrological and meteorological factors have emerged as the most frequent and damaging. India is at the second from the top, after Bangladesh, of flood hazard-affected countries of the world, causing immense damage to life and property every year. The Middle Ganga Plain in the country is one of the most severely affected regions due to the convergence of Himalayan rivers in the Ganga (Ganges) plain. This study focuses on optimizing novel ensemble models to enhance the mapping of flood hazard susceptibility in specific topoclimatic setting, characterized by a low-altitude-range, sub-tropical monsoonal climate, and a riverine floodplain environment of the tectonically active Middle Ganga Plain. This region forms part of the Ganga Foreland Basin prone to frequent floods. We employ a comprehensive flood inventory and twelve selected flood conditioning factors and is the first study that has developed these ensemble models in low relief riverine floodplain environmental setting with high resolution geomorphology as one of the conditioning factors. Four artificial intelligence ensemble models were employed in this work: Logistic Regression-Evidential Belief Function (LR-EBF), Logistic Regression-Frequency Ratio (LR-FR), Multi-Layer Perceptron-Evidential Belief Function (MLP-EBF), and Multi-Layer Perceptron-Frequency Ratio (MLP-FR). Results showed that LR-based ensembles outperformed MLP-based ensembles in the studied topoclimatic setting, with LR-EBF model achieving 87.2% and 84.7% success and prediction rates respectively. Our results reveal consistent performance differences between LR-based ensembles with FR and EBF models across different environmental settings. MLP-based ensembles, particularly with Frequency Ratio, displayed more significant performance variations. The findings underscore the importance of considering specific topoclimatic and altitudinal range environments when applying flood hazard susceptibility models. Our study also provides valuable insights for planning and policymaking in flood management practices in similar topoclimatic settings worldwide.
Source: Discover Geoscience
Institutions: Marwadi University, University of Bucharest, Transylvania University of Brașov, National Institute of Hydrology and Water Management, Danube Delta National Institute for Research and Development, Vidyasagar University, Indian Institute of Technology Bombay, Jamia Millia Islamia, China University of Geosciences (Beijing), The University of Adelaide, University of Hawaiʻi at Mānoa, Université Gustave Eiffel
Language: English
Subfield: Global and Planetary Change
Field: Environmental Science
Domain: Physical Sciences
SDG: Climate action
Open Access status: diamond
OpenAlex ID: W4402270129