Flash-flood hazard susceptibility mapping in Kangsabati River Basin, India. Issue 23 (2nd December 2022)

• Record Type: Journal Article
• Authors: Chakrabortty, Rabin; Chandra Pal, Subodh; Rezaie, Fatemeh; Arabameri, Alireza; Lee, Saro; Roy, Paramita; Saha, Asish; Chowdhuri, Indrajit; Moayedi, Hossein
• Is Part Of: Geocarto international. Volume 37:Issue 23(2023)
• Page Start: 6713
• Page End: 6735
• Publication Date: 2022-12-02
• Subjects: flood risk management -- machine learning model -- flash flooding -- subtropical region; Remote sensing -- Periodicals; Geographic information systems -- Periodicals; Geology -- Periodicals; Cartography -- Periodicals; 621.3678
• DOI: 10.1080/10106049.2021.1953618
• Languages: English
• ISSNs: 1010-6049

Abstract

Abstract: Flood-susceptibility mapping is an important component of flood risk management to control the effects of natural hazards and prevention of injury. We used a remote-sensing and geographic information system (GIS) platform and a machine-learning model to develop a flood susceptibility map of Kangsabati River Basin, India where flash flood is common due to monsoon precipitation with short duration and high intensity. And in this subtropical region, climate change's impact helps to influence the distribution of rainfall and temperature variation. We tested three models-particle swarm optimization (PSO), an artificial neural network (ANN), and a deep-leaning neural network (DLNN)-and prepared a final flood susceptibility map to classify flood-prone regions in the study area. Environmental, topographical, hydrological, and geological conditions were included in the models, and the final model was selected based on the relations between potentiality of causative factors and flood risk based on multi-collinearity analysis. The model results were validated and evaluated using the area under receiver operating characteristic (ROC) curve (AUC), which is an indicator of the current state of the environment and a value >0.95 implies a greater risk of flash floods. The AUC values for ANN, DLNN, and PSO for training datasets were 0.914, 0.920, and 0.942, respectively. Among these three models, PSO showed the best performance with an AUC value of 0.942. The PSO approach is applicable for flood susceptibility mapping of the eastern part of India, a subtropical region, to allow flood mitigation and help to improve risk management in this region.