How AI has made hurricane forecasting easier

Hurricane forecasting is witnessing a major transformation with the increasing use of artificial intelligence (AI) technology. The change was evident when Hurricane Isaias hit Florida in early October. Google DeepMind's AI model accurately predicted the storm's path, outperforming traditional forecasting methods that had initially projected a different course toward southeastern Louisiana.

Michael Lowry, a veteran meteorologist and hurricane expert, praised Google DeepMind's model as the gold standard.

He said it has outperformed all other computer models in predicting hurricanes.

The National Hurricane Center (NHC) started using the Google DeepMind model in its official forecasts last year.

By the end of the 2025 Atlantic hurricane season, it had become one of the best at predicting a storm's path and strength.

Unlike traditional physics-based hurricane models that use real-time atmospheric data to predict a storm's path and strength, AI models rely heavily on historical data.

They recognize patterns by analyzing thousands of past storms along with atmospheric data.

This approach is different from conventional methods as it doesn't depend on the governing equations of the atmosphere or physics to make predictions.

Most developers of these AI models aren't meteorologists but software engineers who create neural networks.

A neural network is a set of algorithms designed to identify patterns in large datasets.

However, this can make the models appear like a black box, as we don't know why they might get a forecast right or wrong.

This uncertainty can be unsettling for meteorologists who rely on these systems during storms when millions are depending on their forecasts.

While some AI models have proven to be accurate, many unproven weather forecasting AI models still exist.

The most accurate ones so far are from Google's DeepMind and the UK's European Centre for Medium-Range Weather Forecasts (ECMWF).

The NHC uses a blend of several proven models, both AI and traditional, for its official forecasts.

However, human forecasters still interpret these model results before deciding what information to relay to the public.

The Google DeepMind model still needs some improvements, especially in determining a storm's intensity.

NHC Director Michael Brennan said his team regularly provides feedback to Google DeepMind for this purpose.

Despite its shortcomings, the speed at which AI is evolving is remarkable.

Unlike traditional models that are updated annually at best, AI updates can take just months or even weeks.