Modeling flood susceptibility on the onset of the Kerala floods of 2018

• Published in: Environmental earth sciences Vol. 83; no. 4; p. 123
• Main Authors: Chithra, K., Binoy, B. V., Bimal, P.
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01-02-2024; Springer Nature B.V
• Subjects: Accuracy; Biogeosciences; Bivariate analysis; Climate and land use; Climate and vegetation; Conditioning; Drainage density; Earth and Environmental Science; Earth Sciences; Environmental Science and Engineering; Flood control; Flood management; Flood mapping; Flooding; Floods; Geochemistry; Geographical information systems; Geology; Hydrology/Water Resources; Independent variables; Information processing; Land use; Mathematical models; Modelling; Normalized difference vegetative index; Original Article; Parameters; Rainfall; Regression analysis; Soil types; Statistical analysis; Statistical models; Susceptibility; Sustainable development; Terrain; Terrestrial Pollution; Vegetation; Vegetation index; Wetlands; Wetness index; Kerala India; India; GIS; Logistic regression; Relative frequency ratio; Flood conditioning parameters; Susceptibility mapping
• ISSN: 1866-6280; 1866-6299
• DOI: 10.1007/s12665-023-11412-1

Floods are the most devastating global hazard which affect the environment and economy of several regions in the world. Flood management requires the identification of areas susceptible to flooding and measuring the impact of flood conditioning parameters. This study examines the application of bivariate relative frequency ratio (RFR) and multivariate logistic regression (L.R.) models to identify flood-susceptible regions in three districts of northern Kerala, India. A comprehensive flood susceptibility study utilizing high-resolution terrain information and past flood inventory is conducted in the study area. The current study generated the flood inventory of August 2018 during the Kerala floods using Sentinel-imagery of 10 m resolution. Thirteen flood conditioning parameters related to the terrain, land usage, climate and vegetation are used as independent variables in the statistical modeling. The terrain-related parameters such as elevation, slope, curvature, flow accumulation, topographic wetness index (TWI), and stream power index (SPI) are derived from CartoDEM of 10 m resolution. The other independent variables used are rainfall, normalized difference vegetation index (NDVI), waterbody distance, drainage density, soil type, and geology. In GIS, the dependent and independent variables are spatially combined, and SPSS and R are used for statistical modeling and validation. The final flood susceptibility map is divided into the risk categories of very low, low, medium, high, and very high. Both RFR and L.R. model results were found reliable, and the low-lying coastal wetlands are highly susceptible to flooding in the study area. The area under curve values show that the L.R. model performs better with 92.7% accuracy than the RFR model with 85.6% accuracy. The generated flood susceptibility map can be a valuable tool for sustainable planning and development in the target region.