Last updated on August 28th, 2025 at 06:11 am
Someone I know quite well worked for a weather forecasting AI startup, and it did not have a tremendous amount of success. It’s hard to crunch numbers on the ephemeral. But big storms are different. They are visible, influences such temperature, pressure and surrounding weather systems are more clear, and with seven to ten day windows and very specific timelines, we should be able to do more; if not forecasting, then management and recovery.
When Typhoon Kajiki made landfall in Vietnam in August 2025, it wasn’t the sheer size of the storm that caught people off guard, but the speed of its intensification. Within hours, what began as a tropical depression had strengthened into a destructive system that battered the north-central coast.
- 3 lives lost
- ~600,000 residents evacuated
- 7,000 homes destroyed
- 28,800 hectares of rice flooded
Hanoi’s streets turned into rivers, airports shut down, schools closed, and more than 16,500 soldiers plus 107,000 paramilitary personnel were mobilized in the emergency response. Kajiki’s devastation wasn’t unprecedented—but it was a wake-up call. Old forecasting methods are no longer enough.
A New Storm Reality
Vietnam faces 10–12 typhoons each year, but recent storms have been harder to predict. Kajiki intensified near land and shifted quickly, leaving less warning time than ever before.
Traditional models—while advanced—sometimes fail to capture micro-climate triggers like subtle changes in sea surface temperature, which can spark sudden surges in storm strength.
Why AI Could Make a Difference
AI can complement and accelerate weather science.
- Rapid intensification alerts → process terabytes of satellite and ocean data in near real-time.
- Hyper-local flood forecasts → simulate impact down to the neighborhood level.
- Evacuation optimization → predict which areas are most at risk, reducing unnecessary disruption.
- Agricultural resilience → forecast crop losses and guide insurance or subsidies.
AI-powered forecasting is about giving meteorologists sharper tools—not replacing them.
Global Lessons Vietnam Can Learn
- United States → AI models now outperform traditional ones in short-term hurricane predictions.
- India → Google’s AI flood alerts reach millions, saving lives along the Ganges and Brahmaputra.
- Japan → AI simulations drive more effective municipal drills.
Vietnam—with its mobile-first culture and growing digital infrastructure—is well-positioned to adopt similar systems. Imagine Zalo push alerts or AI-driven loudspeaker warnings giving families extra hours to prepare.
Vietnam’s Opportunity
Kajiki showed that Vietnam is capable of moving hundreds of thousands of people quickly. But to stay ahead of storms, predictive intelligence must pair with that reactive strength. With machine-learning hurricane models, flood-predictive dashboards, and early-warning chatbots communicating with the public, high-impact forecasting is changing disaster management. How could these be solidified?
- University partnerships to refine local AI storm models.
- Public-private collaborations with telecoms for mobile alerts.
- Agricultural pilots to forecast storm-related crop losses.
- International climate funds to finance AI-driven resilience projects.
Kajiki is also a warning shot. With climate change, storms will form faster, hit harder, and flood deeper. The choice is simple: continue to react to every storm, or build smarter AI-enhanced systems that protect lives and livelihoods before disaster strikes.Possible directions for AI use include:
Advanced Analysis via AI data modelling: (e.g. real-time satellite feeds, sensor networks, flood modeling) These could help authorities pre-stock evacuation zones, adjust reservoir releases, and deploy personnel more efficiently—saving precious time and lives. Modelling rainfall and city layouts based on relationship to sea level should make it possible for cities to do more forecasting around which areas are most likely to be flooded, both to alert urban populations and to ensure resources are there to lessen the duration of the impact.
What systems are likely to fail at which points? Where are power, internet and water systems’ greatest vulnerabilities and how hard are they likely to be hit? In areas where storms are frequent, the impact on populations and commerce could potentially be mitigated and preparation focused on long term consequence management at different levels of rainfall and wind.
Communications with the Most Up to Date Information Across the Country:
- Using storms like Kajiki and previous typhoons and their data can offer tremendous advance understanding of how a storm will function, and communicate it. AI-integrated early-warning systems make proactive evacuation models possible, along with predictive infrastructure and comms planning, to answer questions people are asking most frequently. with the most up to date information. How long should people expect to wait out a storm? Will their area be prone to flooding? Will power, water and hydro be affected? What should they stock? When can they leave their homes?
- A well informed population is a better prepared population, and a less impacted one.
