Hybrid AI-Geo-informatics framework for river course change prediction and disaster risk mitigation

Rivers are dynamic geomorphological systems that frequently alter their courses due to erosion, sediment deposition, channel migration, and flooding. Although such changes are normal, a sudden and major change like the diversion of the Kosi River in Bihar in 2008 can cause disastrous flooding, displacement and heavy land loss. The conventional methods are a poor fit because manual interpretation of satellites and hydrological modelling is time-consuming and has limited spatial-temporal resolution and lacks predictability. This study utilizes multi-source databases to present an AI-based Geo-Informatics framework for river course change prediction and disaster risk mitigation. Other than satellite imagery the data also includes hydrology, rain and soil data. The suggested hybrid architecture aims to jointly model the spatial river morphology and the evolution of the spatial pattern over time through the use of machine learning models (e.g. Random Forest, Gradient Boosting) and deep learning components (CNN, U-Net, LSTM/ConvLSTM). The framework has the ability to create predictive geospatial risk maps, forecasts of river migration over time, and interactive visualization products that support disaster preparedness and sustainable usage of water resources. Overall, the results demonstrate the potential of AI-driven Geo-Informatics to transform river monitoring from reactive assessment to proactive prediction, contributing to resilience building in accordance with the UN Sendai Framework (2015-2030) and the Sustainable Development Goals on climate action and water management.