Research Trends 2026-10-09

The National Hurricane Center Leans on Google DeepMind for Hurricane Isaias -- Forecasting a Category 2 Before the Storm Had a Name

From its first advisory on October 6, the US National Hurricane Center's official discussions cite Google DeepMind's model (GDMI) as a guide for Isaias's track and intensity; by October 8 forecasters called it and a corrected consensus 'our best performing track guidance.' CNN reports the model is WeatherNext 3.

The forecast discussions the National Hurricane Center publishes with each advisory show an AI model becoming everyday forecasting guidance. Isaias began as Tropical Depression Nine on Tuesday, October 6, 2026, and by October 8 was a hurricane in the Gulf of Mexico.

What the NHC wrote

  • Discussion 1 (October 6): the track forecast was "in best agreement with the GDMI and ECAI aids" -- GDMI being the Google DeepMind model.
  • Discussion 2: the intensity forecast was "generally following the more aggressive Google DeepMind and HCCA aids."
  • Discussion 6 (October 7): the forecast "continues to lean more heavily on the better-performing corrected consensus (HCCA) and Google DeepMind (GDMI) aids."
  • Discussion 7 (October 8): GDMI and HCCA "have been our best performing track guidance."

Why forecasters trust it

CNN reports that relying on DeepMind "meant NHC forecast Isaias would become a Category 2 from the first advisory on Tuesday, when it didn't even have a name yet," while many physics-based models were "generally forecasting a weaker storm." It names the model's newest iteration as WeatherNext 3, and says forecasters' confidence "stems largely from experience": last season the center anticipated Hurricane Melissa's rapid intensification almost three days ahead "based largely on the projections from DeepMind." NHC director Michael Brennan told CNN, "The AI models have shown more consistency" from one model run to the next. He added, "We've definitely been making more aggressive intensity forecasts overall for the last several years," pointing to support from physics-based models and forecasters' own reading of a storm's environment as well. Isaias also set a record as the Atlantic's latest first hurricane of a season.

How the model is used in practice. The workflow is not one where the AI makes the forecast. Human forecasters compare several aids -- AI models, physics models and statistical consensus -- and lean on whichever has been verifying best, saying so in public text. The AI model earns weight one storm at a time.

What remains uncertain. When we checked, Isaias had not made landfall, so whether the early aggressive intensity call verifies is still open. Discussion 7 itself flags a hard final stretch: before landfall on the north Gulf coast, "vertical wind shear ramps up to 40-50 kt, and this will likely impart some weakening," while the wind field may grow broader and more asymmetric. The NHC text names "Google DeepMind" and "GDMI," not a model version; the WeatherNext 3 label comes from CNN.

Analysis: this is a good model for adopting AI in any high-stakes forecasting job -- run it beside existing methods, track its error openly, and let trust follow its record rather than its launch claims.

NHC discussions from Isaias's first advisory cite Google DeepMind's model as track and intensity guidance and by October 8 call it among 'our best performing track guidance'; human forecasters still weigh several aids, and whether the early Category 2 call verifies depends on the storm's landfall.