NEW DELHI: The Central Water Commission (CWC) has commissioned an internal study to forecast four weeks’ advance floods in strategically important areas of the country, utilising Artificial Intelligence and Machine Learning (AI-ML) in the monsoon season, this newspaper has learnt.
The study is designed to identify flood hotspots in advance, facilitating better preparedness and mitigation efforts.
Till now, CWC mostly depended on IMD’s forecast, which is not effective in long-range forecasting. The internal study, referred to as Extended Range Prediction (ERP), will be used for internal purposes, such as enhancing defence readiness and disaster mitigation in vulnerable regions like the Himalayan area. The ERP will operate for 4 weeks to precisely analyse and predict flood or glacial lake outbreak forecasts.
So far, the AI-ML model has been tested at 94 stations within the Ganga, Brahmaputra, and Godavari river basins in 2025. Results indicate that 31 stations (23 in the Ganga basin, 9 in the Brahmaputra basin, and 1 in the Godavari basin) achieved an accuracy rate greater than 75% across all four weeks.
The study is expected to predict areas that may be affected by flash floods. Last year, the CWC model was applied to the transboundary Jaldhaka River basin from October 1-5, employing the Rainfall Runoff Inundation Model to forecast flash floods.
The study will utilise flash flood data from the India Meteorological Department (IMD) to identify regions likely to experience sudden flooding. It will incorporate computer models and basic elevation maps to simulate how floodwaters may spread, improving understanding of potential impacts and damage under various scenarios.
Currently, the CWC provides short-range flood forecasts for daily and weekly periods based on IMD rainfall forecasts, which are often insufficient for long-range predictions.
While the IMD forecasts rainfall for all four weeks, their accuracy diminishes after the second week, affecting the CWC’s flood forecasts. The CWC provide short-range flood forecasts(24 hrs) and one week medium range forecast by calculating the IMD’s net probable rainfall to make predictions.
The ERP study aims to address the gaps in the existing IMD-based forecasts for managing extreme weather events. “The ERP study will be based on an indigenous model for the purposes of vigilance, particularly for CWC operations and other internal agencies,” stated an officer familiar with the developments.
The name of the indigenously developed model has not been disclosed. The discussion surrounding the ERP was first initiated in February during meetings with state representatives. The IMD has been using the Radar Precipitation Model in the ‘Now Cast’ category to accurately predict extreme weather events across the country. In line with the goals of the ‘Now Cast’, the ERP study will assist in identifying flood hotspots with greater precision and advance notice.
