Flood inundation management in the Narmada Basin: an AIML application for the upstream area of Sardar Sarovar Dam

Abstract

Recurrent flooding poses a significant threat to various sub-catchments of the Narmada River Basin, one of India's major river systems. This study focuses on the flood-prone sub-catchment area upstream of the Sardar Sarovar Dam, where impacts are particularly severe on tribal communities, forests, and the newly formed reservoir ecosystem. To enhance flood risk management, this research investigates the application of Artificial Intelligence and Machine Learning (AIML) for high-resolution flood inundation mapping. The primary objective is to generate high-resolution flood inundation maps that surpass hydrological modelling in accuracy and spatial detail, enabling precise identification of vulnerable areas within the sub-catchment. A comprehensive dataset, including historical rainfall data (1990-2024) from IMD gridded data and local rain gauges, river discharge records from various gauging stations and a 12.5m resolution Digital Elevation Model (DEM), is used to train and validate AIML models (Artificial Neural Network (ANN), Random Forests (RF), and K-Nearest Neighbor (KNN)). Beyond flood inundation, the models were employed to simulate the effects of various flood control measures, including optimized reservoir operation, embankment construction, and afforestation, to inform optimal implementation strategies. The results are expected to demonstrate the superior performance of AIML in capturing and predicting future flood inundations in the region. Based on error calculation, the performance of combined models is expected to be better than that of individual models. The findings will help develop targeted early warning systems, improved land-use planning, and evidence-based decision-making for sustainable flood risk management in the Narmada Basin and contribute to the broader application of AI for disaster risk reduction globally.