On 23 February 2026, the South Asia Hydromet Forum (SAHF) Working Group (WG) on Hydrology convened its second meeting, bringing together delegates from Bangladesh, Bhutan, India, Maldives, Myanmar, Nepal, Pakistan, and Sri Lanka.
The session included experts from the World Meteorological Organization (WMO), the Regional Integrated Multi-Hazard Early Warning Systems (RIMES), Indonesia's Badan Meteorologi, Klimatologi, dan Geofisika (BMKG), India's Central Water Commission (CWC), and various academic institutions.
Following up on the initial virtual meeting in July 2025, and the approval of the Climate Risk and Early Warning Systems (CREWS) South Asia project , the meeting discussions centered around the sustainability of the Flash Flood Guidance System and the potential transition towards open-source, interoperable, and scalable flood forecasting frameworks.
The meeting was chaired by Dr. K.J. Ramesh, senior advisor at RIMES, and opened by Dr. Mrutyunjay Mohapatra, Chair of the SAHF Executive Council and Director General of the India Meteorological Department. In his remarks, Dr Mohapatra highlighted the necessity of multi-model ensemble approaches to increase the accuracy of flash flood forecasts. He also noted the expanding role of Artificial Intelligence (AI) and Machine Learning (ML) in refining forecast systems and transforming unstructured, crowdsourced data into actionable datasets. Dr Hwirin Kim, Chief of WMO Hydrological Modelling and Forecasting (HMF) Section, advocated for a framework that would prioritize open-source, modular, and member-driven solutions to ensure better integration of existing tools and platforms.
During the meeting, participants presented updates on national and regional efforts to strengthen flash flood forecasting. WMO introduced the Flood Forecasting Framework (FFF), which was presented at the Advisory Group meeting of the Flood Forecasting Initiative (FFI-AG) in 2025. The framework envisions a modular, distributed architecture incorporating multiple data sources and delivering tailored outputs to the end-users. Designed to scale from sub-basin to national levels, the framework is also adaptable to different types of flooding. A concept note is being developed for formal approval by WMO Congress by the end of 2026. The FFF is centered around five principles: (i) Interoperability, (ii) Open-Source, (iii) Sustainability, (iv) Member-driven, and (v) User-centric. These principles apply not just to the FFF but should be advocated by Members for all projects and partnerships. SAHF Members were encouraged to apply these principles in future initiatives.
Country presentations highlighted practical implementations of flash flood forecasting approaches. In Bangladesh, RIMES has developed an operational flash flood forecasting system using a multi-model framework comprising the Tank Hydrological Model with up to 15 days lead time, WRF-Hydrology with up to 3 days lead time, and Rainfall Threshold-Based Flood Forecasting. The system is being hosted at the Bangladesh Water Development Board 's (BWDB) Flood Forecasting and Warning Center (FFWC). The layered approach presented a pragmatic way to align tools with varying forecast windows and operational needs.
In Pakistan and Afghanistan, BMKG is piloting the Ensemble Framework for Flash Flood Forecasting (EF5), which is a distributed, physics-based, open-source hydrological modelling framework developed by the National Oceanic and Atmospheric Administration (NOAA) and the University of Oklahoma. BMKG has implemented a customized version of EF5, which is configured with precipitation estimates from the Self-Calibrating Multivariate Precipitation Retrieval (SCaMPR) algorithm and from the European Centre for Medium Range Weather Forecasting (ECMWF). The system calculates water balance and runs kinematic wave routing modules to generate stream flow, runoff volume, and landslide potential outputs. The key features include the ability to place virtual rain gauges along river networks and to run ensemble or deterministic forecasts for up to 10 days lead time.
In India, several research groups are exploring various approaches to better forecast flash floods. The Indian Institute of Technology (IIT) Delhi is using a novel flood framework that couples EF5 and Triton (a GPU-based hydrodynamic model). The system is still under testing and makes use of IMD's High Resolution Rapid Refresh (HRRR) product.
The meeting outlined the need to continue the operations for the existing FFGS until an alternative solution is found. It was noted that EF5 or any other similar open-source framework is a promising but maturing candidate for transition, requiring significant efforts to ensure operationalization. It was agreed that a multi-model approach, rather than reliance on a single model, was more suitable for the region. The need to ensure robust data sharing and integration was also critical to ensure the success of the solution.
A set of key recommendations was drafted for the SAHF Executive Council, centering around the way forward with respect to flood forecasting and warnings in the region, including the endorsement of the Flood Forecasting Framework. It was also requested that additional resources be mobilized to support pilot initiatives for flood forecasting in the region, to help find the most optimal solutions.
