Machine Learning Based Prediction of Urban Flood Susceptibility from Selected Rivers in a Tropical Catchment Area
Benjamin Nnamdi Ekwueme
Abstract
Unexpected flood due to climate change has caused tremendous damage to both lives and properties, especially in tropical areas. Nigeria Southeastern region has been devastated by flood from extreme weather conditions. Flood mitigation involves accurate forecasting, precise prediction, evaluation, and intervention strategy. This study aims at using machine learning solutions to investigate and predict flood susceptibility from selected rivers in the south-eastern region of Nigeria. The regional hydrogeological data from 1981–2019 was collected and analysed. The remote sensing datasets from the National Aeronautics and Space Administration (NASA), Modern-Era Retrospective Analysis for Research and Applications (MERRA) version 2 & 3 platforms from five selected rivers were processed. With the data output of the hydrology, streamline flows, and exposed geology, the ARIMA model was built and used to forecast the flood. The result shows that the flooding pattern would increase by 15-150% within 2020-2024. The forecast indicated that within five years, the river discharge for Adada, Ajali, Imo, Ivo, and Otanmiri will increase within ranges 200-702 m3s-1, 16-26 m3s-1, 508-1280 m3s-1, 43-68.5 m3s-1, and 22-35.1 m3s-1 respectively. Climate change has impacted severely on flood in the region. This knowledge will help the regional agencies and authorities in adapting to flood innuendoes and assessment of hydrologic extremes. Doi: 10.28991/CEJ-2022-08-09-08 Full Text: PDF