Integrated GIS-Based MCDA and Machine Learning Techniques in Flood Susceptibility Mapping in Ala River Basin, Nigeria

Abstract Flooding is a recognized form of natural disaster that can lead to loss of life, destruction of critical infrastructure with consequences impacting critical sectors including agriculture and health. This study aims to map out flood susceptible areas within the Ala River basin of Ondo State, Nigeria by integrating the Analytical Hierarchy Process (AHP) Multi-Criteria Decision Analysis (MCDA) technique and Support Vector Machines (SVM) Machine Learning (ML) model. Nineteen factors including elevation, slope, aspect, curvature (profile and plan), roughness, flow direction, flow accumulation, drainage density, distance from the river, TWI, STI, SPI, soil, geology, NDVI, NDMI, LULC, and rainfall were considered as input parameters. Flood susceptibility maps generated from each of these approaches were combined to create a more comprehensive flood susceptibility map of the study area. The AHP analysis has a consistency ratio of 1.8%. Precision, recall, f1-score, accuracy score, and ROC-AUC curve were used in evaluating the AHP-MCDA and SVM-ML model. Based on the evaluation, the combined flood susceptibility map result showed the best performance with the AUC score 0.74, SVM-ML with a score 0.73, and the AHP-MCDA having the least score of 0.59. As these results demonstrate, multiple approaches are required to mitigate flooding.