🔗 Share this article

As Developing Cyclone Melissa swirled off the coast of Haiti, meteorologist Philippe Papin felt certain it would soon escalate to a monster hurricane.

Serving as primary meteorologist on duty, he forecasted that in a single day the storm would become a category 4 hurricane and start shifting in the direction of the Jamaican shoreline. Not a single expert had ever issued such a bold forecast for rapid strengthening.

However, Papin possessed a secret advantage: artificial intelligence in the guise of the tech giant’s new DeepMind hurricane model – released for the first time in June. And, as predicted, Melissa did become a storm of remarkable power that tore through Jamaica.

Growing Reliance on Artificial Intelligence Forecasting

Meteorologists are increasingly leaning hard on the AI system. During 25 October, Papin explained in his public discussion that Google’s model was a primary reason for his certainty: “Approximately 40/50 Google DeepMind ensemble members indicate Melissa becoming a Category 5 storm. Although I am unprepared to forecast that strength yet given path variability, that remains a possibility.

“There is a high probability that a period of quick strengthening will occur as the system moves slowly over very warm sea temperatures which represent the highest marine thermal energy in the whole Atlantic basin.”

Surpassing Conventional Models

Google DeepMind is the pioneer AI model focused on tropical cyclones, and now the first to outperform traditional weather forecasters at their own game. Across all tropical systems so far this year, the AI is the best – surpassing experts on path forecasts.

The hurricane ultimately struck in Jamaica at category 5 intensity, one of the strongest coastal impacts ever documented in almost 200 years of record-keeping across the region. The confident prediction probably provided residents extra time to prepare for the disaster, possibly saving lives and property.

The Way The Model Works

The AI system works by identifying trends that conventional lengthy scientific weather models may miss.

“They do it much more quickly than their physics-based cousins, and the processing requirements is less expensive and time consuming,” said Michael Lowry, a former meteorologist.

“What this hurricane season has proven in short order is that the recent artificial intelligence systems are competitive with and, in certain instances, superior than the slower physics-based forecasting tools we’ve relied upon,” he said.

Understanding Machine Learning

To be sure, Google DeepMind is an instance of machine learning – a technique that has been employed in data-heavy sciences like weather science for years – and is distinct from generative AI like ChatGPT.

AI training takes mounds of data and pulls out patterns from them in a manner that its system only requires minutes to come up with an answer, and can operate on a desktop computer – in strong contrast to the flagship models that governments have used for decades that can require many hours to run and require some of the biggest supercomputers in the world.

Expert Reactions and Future Advances

Nevertheless, the fact that Google’s model could outperform previous gold-standard legacy models so rapidly is nothing short of amazing to weather scientists who have spent their careers trying to forecast the most intense storms.

“It’s astonishing,” said James Franklin, a former expert. “The data is sufficient that it’s pretty clear this is not a case of chance.”

Franklin said that while Google DeepMind is beating all other models on predicting the future path of storms worldwide this year, similar to other systems it occasionally gets high-end intensity predictions wrong. It struggled with another storm earlier this year, as it was similarly experiencing quick strengthening to maximum intensity north of the Caribbean.

During the next break, Franklin stated he intends to talk with Google about how it can enhance the DeepMind output even more helpful for experts by providing additional internal information they can use to evaluate the reasons it is producing its conclusions.

“A key concern that troubles me is that while these forecasts seem to be really, really good, the output of the system is kind of a opaque process,” remarked Franklin.

Wider Industry Developments

There has never been a private, for-profit company that has developed a top-level weather model which allows researchers a view of its techniques – unlike nearly all other models which are provided at no cost to the general audience in their entirety by the authorities that designed and maintain them.

The company is not the only one in starting to use AI to solve challenging weather forecasting problems. The US and European governments also have their respective artificial intelligence systems in the development phase – which have demonstrated better performance over previous non-AI versions.

Future developments in artificial intelligence predictions appear to involve new firms tackling formerly difficult problems such as long-range forecasts and improved advance warnings of tornado outbreaks and sudden deluges – and they are receiving US government funding to pursue this. One company, WindBorne Systems, is even deploying its own weather balloons to address deficiencies in the US weather-observing network.