evidoria

← Back to browse

Good practice Imported

Vietnam's National Centre for Hydro-Meteorological Forecasting Pairs AI with Google and Weathernews to Sharpen Typhoon Warnings

Vietnam · Hanoi · See the Vietnam profile · See the Hanoi profile

Evidence: Observational / pre–post Top 62% 53/100 · Ask Evidence Copilot about this practice

Vietnam's national forecasting centre rolled out AI-assisted typhoon and flood forecasting for the 2025 storm season, combining machine-learning models on radar and satellite data with Google's cyclone and flood models under a July 2025 MOU with Japan's Weathernews.

90-110 km
24-hour cyclone-centre position error (2025 season)
~30 %
Estimated economic-loss reduction from 24h warnings

Details

Maturity
Established
Promoter
National Centre for Hydro-Meteorological Forecasting, Vietnam Meteorological and Hydrological Administration (Ministry of Agriculture and Environment)
Period
2025-2025
Keywords
meteorology, disaster forecasting, machine learning, public safety, flood and typhoon warning

Context

After Typhoon Yagi's destructive 2024 landfall exposed gaps in track-prediction accuracy, Vietnam's National Centre for Hydro-Meteorological Forecasting (NCHMF) began integrating AI into operational forecasting for the 2025 typhoon and flood season.

Activities

Machine-learning models trained on radar, satellite imagery and automated observation stations generate short-term rainfall nowcasts and identify cyclone centres. A July 2025 MOU with Japan's Weathernews supplies AI-based typhoon-track predictions, alongside adopted Google flood/cyclone models.

Results

24-hour cyclone-centre position error now runs 90-110 km, in line with regional norms; storm and lightning alerts are issued 30 minutes to 3 hours ahead of events, and 24-hour advance warnings are associated with roughly 30% lower economic losses.

Conclusions

The centre acknowledges it still lags Japan, China and South Korea technologically, citing sparse observation networks, budget constraints and a shortage of AI specialists as ongoing limits.

Implementation
Indicative cost
Medium (€50k–€500k) — Not disclosed; largely software/partnership-based rather than new hardware.
Time to results
Medium (1–3 years) — Integrated ahead of the 2025 typhoon and flood season; MOU signed 31 July 2025.
Staffing & skills
NCHMF forecasters, Hanoi University of Science and Technology AI institute (integration support)

Conditions for success

  • International partnerships (Weathernews MOU, Google flood/cyclone models) supplementing domestic capacity

Common failure modes

  • Sparse observation networks, budget constraints and a shortage of AI specialists limit further gains
  • Still lags regional peers (Japan, China, South Korea) technologically

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Similar practices you may find useful