New tools for hydrological and water forecasting in practice

On March 23 in Hanoi, the Viet Nam Meteorological and Hydrological Administration (under the Ministry of Agriculture and Environment), in collaboration with the University of Science (Vietnam National University, Hanoi), organized a scientific workshop titled “New solutions and technologies for hydrological and water resources forecasting and their application in operations.” The event was held in response to World Water Day (March 22) and World Meteorological Day (March 23).

According to Mr. Nguyen Thuong Hien, Director General of the Viet Nam Meteorological and Hydrological Administration, hydrology and water resources are no longer confined to a single technical field but have become strategic issues closely linked to water security, disaster safety, and the country’s sustainable development. In this context, the demands placed on hydrological forecasting are increasing.

“We not only need more accurate forecasts, but also earlier and more detailed ones—most importantly, forecasts that can directly support real-world decision-making and operations. This is a major challenge, requiring the hydrometeorological sector to innovate strongly in both mindset and technology,” emphasized Nguyen Thuong Hien, Director General of the Vietnam Meteorological and Hydrological Administration.

Science and technology have made significant advances in this field, including numerical hydrological–hydraulic models, remote sensing, satellite data, IoT-based sensor systems, and artificial intelligence (AI). These tools have enabled hydrological processes to be monitored, observed, and analyzed with increasing precision and shorter latency, moving toward real-time forecasting. However, Mr. Hien candidly acknowledged that a gap still exists between research and operational application. Many valuable research outcomes have yet to be translated into practical tools, while real-world forecasting operations continue to face challenges that require scientific solutions.

Highlighting these challenges, Dr. Hoang Van Dai, Deputy Director of the National Center for Hydro-Meteorological Forecasting, noted that the center is currently operating several advanced models such as NAM, MIKE 11, MARINE (France), and IFAS (Japan). However, key constraints remain, including dependence on the accuracy of rainfall forecasts, limited real-time reservoir operation data, and insufficient capability for high-resolution flood forecasting.

A major bottleneck lies in transboundary river flows. Ms. Nguyen Y Nhu of the University of Science pointed out that the Red River Basin depends on more than 40% of its water originating from China. The lack of a data-sharing mechanism for upstream reservoirs significantly increases forecasting errors.

Empirical data show that when relying solely on domestic observation data, machine learning models can provide forecasts for less than seven days - too short for local authorities to proactively respond to major floods or prolonged droughts.

To address data gaps, experts have proposed breakthrough technological solutions. Nguyen Y Nhu’s research team introduced a hybrid model combining physics-based approaches (the VIC model) with AI algorithms (such as LSTM and GRU). By enabling AI to “learn” the physical patterns of baseflow, the system can compensate for missing upstream information. Test results at the Pac Ma station in Lai Chau Province show that accurate forecast lead times can be extended up to 100 days - a significant advancement compared to traditional methods.

Associate Professor Dr. Trinh Quang Toan from the Vietnam Academy for Water Resources cited record-breaking storms in 2025, such as Kajiki and Bualoi, to emphasize that hydrological forecasts cannot remain at general numerical levels. He proposed shifting to “impact-oriented forecasting”, integrating multi-source data (radar, satellites) into the WEHY model combined with AI to update ultra-short-term scenarios within 0-6 hours. This approach is critical for making timely reservoir management decisions at major dams like Ban Ve and Son La.

Associate Professor Dr. Ngo Le An from Thuy Loi University also presented an AI-based solution to directly forecast seasonal water flow from raw rainfall data without prior adjustments. He suggested integrating data from irrigation infrastructure into forecasting models to enhance the reliability of agricultural planning.

Professor Tran Thuc, President of the Vietnam Meteorological and Hydrological Society, acknowledged technological progress but warned about fragmentation in research. Current solutions are scattered and have yet to be unified into a comprehensive framework.

The meteorological sector needs a paradigm shift from phenomenon-based to impact-based forecasting. Instead of merely reporting 200mm of rainfall, forecasts should specify which areas will flood, the expected depth of flooding, and the potential effects on residents. Achieving this requires close coordination between meteorological agencies and other relevant bodies to build a sufficiently large cross-sectoral database. Meteorological authorities should serve as the central hub, integrating fragmented scientific and technological solutions into a cohesive system to reduce errors and improve the reliability of early warnings and forecasts.

Experts agreed on prioritizing investment in open-source technologies and deep AI integration for hydrological forecasting. The ultimate goal is to establish an intelligent early warning system that transforms technical data into actionable, understandable information for communities, helping to mitigate losses and damages in the face of increasingly unpredictable disasters.