🔗 Share this article

When Developing Cyclone Melissa was churning south of Haiti, weather expert Philippe Papin felt certain it would soon grow into a major tropical system.

As the primary meteorologist on duty, he forecasted that in a single day the storm would become a category 4 hurricane and start shifting towards the Jamaican shoreline. Not a single expert had ever issued such a bold forecast for quick intensification.

However, Papin had an ace up his sleeve: artificial intelligence in the form of the tech giant’s new DeepMind cyclone prediction system – launched for the first time in June. And, as predicted, Melissa did become a storm of astonishing strength that tore through Jamaica.

Growing Dependence on Artificial Intelligence Forecasting

Forecasters are heavily relying upon Google DeepMind. On the morning of 25 October, Papin explained in his official briefing that Google’s model was a primary reason for his certainty: “Approximately 40/50 AI ensemble members show Melissa becoming a most intense storm. While I am not ready to forecast that strength yet given track uncertainty, that remains a possibility.

“There is a high probability that a phase of quick strengthening will occur as the storm moves slowly over exceptionally hot ocean waters which is the most extreme oceanic heat content in the entire Atlantic basin.”

Outperforming Conventional Models

The AI model is the pioneer AI model focused on hurricanes, and currently the first to outperform standard weather forecasters at their own game. Through all 13 Atlantic storms this season, Google’s model is top-performing – even beating human forecasters on path forecasts.

The hurricane ultimately struck in Jamaica at maximum intensity, among the most powerful landfalls recorded in almost 200 years of record-keeping across the Atlantic basin. Papin’s bold forecast probably provided residents extra time to prepare for the disaster, potentially preserving people and assets.

How Google’s System Functions

The AI system operates through spotting patterns that traditional lengthy scientific prediction systems may overlook.

“The AI performs far faster than their physics-based cousins, and the processing requirements is more affordable and demanding,” said Michael Lowry, a former forecaster.

“This season’s events has proven in short order is that the newcomer artificial intelligence systems are on par with and, in certain instances, more accurate than the less rapid traditional forecasting tools we’ve relied upon,” Lowry added.

Understanding Machine Learning

To be sure, the system is an example of AI training – a method that has been used in research fields like weather science for years – and is distinct from generative AI like ChatGPT.

AI training processes large datasets and extracts trends from them in a such a way that its system only takes a few minutes to generate an result, and can operate on a desktop computer – in strong contrast to the flagship models that authorities have used for years that can require many hours to process and need some of the biggest high-performance systems in the world.

Expert Reactions and Future Advances

Still, the reality that the AI could outperform earlier top-tier legacy models so quickly is nothing short of amazing to weather scientists who have dedicated their lives trying to forecast the most intense weather systems.

“It’s astonishing,” commented James Franklin, a former forecaster. “The data is now large enough that it’s pretty clear this is not just beginner’s luck.”

Franklin said that although Google DeepMind is beating all other models on forecasting the trajectory of storms globally this year, similar to other systems it sometimes errs on high-end intensity forecasts wrong. It struggled with Hurricane Erin previously, as it was also undergoing quick strengthening to maximum intensity above the Caribbean.

In the coming offseason, Franklin stated he plans to talk with Google about how it can make the DeepMind output more useful for forecasters by providing additional internal information they can use to assess exactly why it is coming up with its conclusions.

“The one thing that nags at me is that although these forecasts appear really, really good, the output of the model is essentially a black box,” said Franklin.

Broader Industry Developments

There has never been a commercial entity that has produced a high-performance forecasting system which grants experts a peek into its techniques – unlike nearly all other models which are provided free to the public in their entirety by the authorities that designed and maintain them.

Google is not the only one in adopting artificial intelligence to address difficult meteorological problems. The authorities also have their own artificial intelligence systems in the development phase – which have also shown better performance over earlier non-AI versions.

The next steps in AI weather forecasts appear to involve new firms taking swings at previously tough-to-solve problems such as long-range forecasts and better advance warnings of severe weather and sudden deluges – and they are receiving US government funding to pursue this. One company, WindBorne Systems, is also launching its own weather balloons to fill the gaps in the US weather-observing network.