🔗 Share this article

When Tropical Storm Melissa was churning south of Haiti, weather expert Philippe Papin had confidence it would soon escalate to a major tropical system.

As the lead forecaster on duty, he predicted that in just 24 hours the storm would intensify into a severe hurricane and start shifting in the direction of the Jamaican shoreline. No forecaster had ever issued this confident forecast for quick intensification.

But, Papin possessed a secret advantage: AI technology in the form of the tech giant’s recently introduced DeepMind hurricane model – released for the first time in June. True to the forecast, Melissa did become a storm of remarkable power that tore through Jamaica.

Increasing Reliance on Artificial Intelligence Forecasting

Meteorologists are heavily relying upon Google DeepMind. On the morning of 25 October, Papin explained in his public discussion that Google’s model was a primary reason for his confidence: “Roughly 40/50 AI simulation runs indicate Melissa reaching a Category 5 hurricane. While I am not ready to forecast that strength yet due to track uncertainty, that remains a possibility.

“It appears likely that a phase of rapid intensification is expected as the system drifts over exceptionally hot ocean waters which is the highest marine thermal energy in the whole Atlantic basin.”

Outperforming Conventional Systems

The AI model is the pioneer AI model dedicated to tropical cyclones, and now the first to beat traditional meteorological experts at their own game. Across all tropical systems this season, the AI is the best – surpassing human forecasters on path forecasts.

Melissa eventually made landfall in Jamaica at category 5 intensity, among the most powerful landfalls recorded in nearly two centuries of record-keeping across the region. The confident prediction probably provided people in Jamaica additional preparation time to get ready for the disaster, potentially preserving people and assets.

How The Model Works

Google’s model operates through spotting patterns that conventional lengthy scientific prediction systems may miss.

“The AI performs far faster than their physics-based cousins, and the computing power is less expensive and demanding,” said Michael Lowry, a former meteorologist.

“This season’s events has demonstrated in quick time is that the recent AI weather models are competitive with and, in certain instances, superior than the less rapid traditional forecasting tools we’ve relied upon,” Lowry said.

Understanding Machine Learning

It’s important to note, Google DeepMind is an example of machine learning – a technique that has been used in data-heavy sciences like weather science for years – and is distinct from generative AI like ChatGPT.

AI training takes large datasets and extracts trends from them in a manner that its model only takes a few minutes to generate an answer, and can operate on a desktop computer – in sharp difference to the flagship models that authorities have used for decades that can require many hours to process and need the largest supercomputers in the world.

Expert Responses and Future Advances

Nevertheless, the reality that Google’s model could exceed earlier gold-standard traditional systems so rapidly is nothing short of amazing to weather scientists who have dedicated their lives trying to predict the world’s strongest weather systems.

“I’m impressed,” commented James Franklin, a former forecaster. “The sample is now large enough that it’s pretty clear this is not a case of beginner’s luck.”

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

During the next break, he said he plans to talk with the company about how it can enhance the DeepMind output even more helpful for forecasters by offering additional internal information they can utilize to assess the reasons it is coming up with its conclusions.

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

Wider Industry Developments

Historically, no a private, for-profit company that has produced a top-level forecasting system which allows researchers a peek into its techniques – unlike most other models which are provided at no cost to the general audience in their full form by the authorities that created and operate them.

Google is not the only one in adopting artificial intelligence to address difficult weather forecasting problems. The US and European governments also have their respective artificial intelligence systems in the works – which have demonstrated better performance over previous traditional systems.

Future developments in artificial intelligence predictions seem to be new firms tackling previously difficult problems such as sub-seasonal outlooks and improved advance warnings of tornado outbreaks and sudden deluges – and they have secured US government funding to do so. A particular firm, WindBorne Systems, is even deploying its proprietary weather balloons to fill the gaps in the US weather-observing network.