How accurate is Google's new AI hurricane forecast model?
TAMPA, Fla. (WFLA)—A few weeks ago, Google DeepMind and Google Labs released their new AI hurricane model to the public. The Google team claims its AI model performs better than traditional hurricane models on both track and intensity forecasts.
We will examine the accuracy claims more closely, but first, let's discuss what this AI model is.
This AI model is not a traditional physics-based model like the ECMWF (European) or the GFS (American) model.
Models like the ECMWF (European) use numerical weather prediction, solving fluid dynamics, thermodynamics, and radiation equations on a high-resolution global grid. This requires immense computational resources and supercomputer infrastructure. Just one run can take several hours for the supercomputers to finish. (It is worth mentioning that ECMWF also has an AI model)
The Google AI Model is built on a trained neural network, which mimics the human brain, that can make inferences almost instantly after training. It learns from decades of vast historical weather data – essentially doing very advanced pattern recognition, thus it outputs forecasts without solving the complex differential equations of physics. So the process takes just a minute to complete a 15-day forecast.
Just like the ECMWF (European) model, Google's AI model produces 50 ensemble members. The ensemble members are solutions that are each slightly perturbed. Think of it as a family of solutions rather than just one track and intensity.
Google claims that in tests for 2023–24 storms in the North Atlantic and East Pacific, its 5-day track forecasts were ~85 miles closer to actual tracks than ECMWF's ENS ensemble and their AI model outperformed NOAA's best intensity model – the HAFS model – on intensity forecasts, matching or exceeding high-resolution physics-based accuracy.
We took Google's AI model for a spin and looked into the results for our two most impactful storms last year – Helene and Milton – and found that on track forecasts, a few days ahead of time, the model did extremely well.
In both cases, the Google AI model was just miles away from the actual track. See the comparison below.
It should be noted that in both of these cases, the track was pretty straightforward, with many models in the ballpark of the actual track. Still, the Google AI results are very impressive.
On intensity forecasts, however, the results were not quite as spot on. They were decent in Helene, but way under-forecast the strength in Milton. You can compare the results below. To be fair, intensity forecasting is much harder than track forecasting.
To truly be able to judge the accuracy of the Google AI model – or any model – we need lots more data. So, this 2025 hurricane season, we will monitor the model to see how it performs. We will post the model and also show it on WFLA-TV when storms threaten.
Copyright 2025 Nexstar Media, Inc. All rights reserved. This material may not be published, broadcast, rewritten, or redistributed.
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Yahoo
40 minutes ago
- Yahoo
Genentech Provides Update on Astegolimab in Chronic Obstructive Pulmonary Disease
– The pivotal Phase IIb ALIENTO study met the primary endpoint of a statistically significant reduction in the annualized exacerbation rate (AER) at 52 weeks when astegolimab was given every two weeks – – The Phase III ARNASA study did not meet the primary endpoint of a statistically significant reduction in the AER at 52 weeks – – The safety profile of astegolimab was consistent with previously reported data, with no new safety signals identified – – Analysis of the ALIENTO and ARNASA data will be discussed with regulatory authorities and shared at an upcoming medical meeting – SOUTH SAN FRANCISCO, Calif., July 21, 2025--(BUSINESS WIRE)--Genentech, a member of the Roche Group (SIX: RO, ROG; OTCQX: RHHBY), announced today topline results from the pivotal Phase IIb ALIENTO (n=1,301) and the Phase III ARNASA (n=1,375) trials investigating astegolimab compared to placebo, on top of standard of care maintenance therapy in people with moderate to very severe chronic obstructive pulmonary disease (COPD). The studies included a broad population: both former and current smokers, regardless of blood eosinophil count, who have a history of frequent exacerbations. The pivotal Phase IIb ALIENTO study met its primary endpoint and showed that astegolimab reduced the annualized exacerbation rate (AER) by a statistically significant 15.4% at 52 weeks, when given every two weeks. However, the Phase III ARNASA study did not meet its primary endpoint of a statistically significant reduction in the AER, demonstrating a numerical 14.5% reduction at 52 weeks when astegolimab was given every two weeks. The results were generally consistent across secondary endpoints in both studies. The total number of exacerbations was lower than prospectively anticipated in both trials. The safety profile of astegolimab was consistent with previously reported data, with no new safety signals identified. "While COPD remains the third leading cause of death worldwide, patients and families have limited treatment options for managing this debilitating and complex disease," said Levi Garraway, M.D., Ph.D., chief medical officer and head of Global Product Development. "This was the first set of studies in an 'all-comers' COPD population, and we will discuss these data with regulatory authorities to evaluate next steps for astegolimab." Detailed results from ALIENTO and ARNASA will be shared at an upcoming medical meeting. About the ALIENTO and ARNASA studies Astegolimab is an investigational, fully human anti-ST2 monoclonal antibody designed to bind with high affinity to the ST2 receptor, thereby blocking the signaling of IL-33. The astegolimab COPD pivotal program consists of two registrational studies, the Phase IIb ALIENTO (NCT05037929) and Phase III ARNASA (NCT05595642) studies. Both ALIENTO and ARNASA are double-blinded, placebo-controlled, multi-center studies that evaluate the efficacy and safety of astegolimab administered every two or every four weeks in patients with COPD on top of standard of care maintenance therapy. Patients in the studies included former and current smokers, regardless of blood eosinophil count, who have a history of frequent exacerbations. The primary analysis is based on the initial phase of the study, which consisted of 1,301 patients for ALIENTO and 1,375 patients for ARNASA. The primary endpoint is the reduction in the annualized rate of moderate and severe COPD exacerbations (AER) over the 52-week treatment period. AER is the total number of exacerbations (a sudden worsening in airway function and respiratory symptoms) occurring over the relevant treatment period, divided by the total number of patient years. Standard of care maintenance therapy for both studies was one of the following combinations - inhaled corticosteroid (ICS) plus long-acting beta-agonist (LABA); long-acting muscarinic antagonist (LAMA) plus LABA; ICS plus LAMA plus LABA. About Genentech Founded more than 40 years ago, Genentech is a leading biotechnology company that discovers, develops, manufactures and commercializes medicines to treat patients with serious and life-threatening medical conditions. The company, a member of the Roche Group, has headquarters in South San Francisco, California. For additional information about the company, please visit View source version on Contacts Media Contact:Lindsey Mathias, (650) 467-6800 Advocacy Contact:Del Conyers, (571) 684-1616 Investor Contacts:Loren Kalm, (650) 225-3217Bruno Eschli, +41 61 68-75284 Error in retrieving data Sign in to access your portfolio Error in retrieving data Error in retrieving data Error in retrieving data Error in retrieving data
Yahoo
2 hours ago
- Yahoo
Man who died after being pulled into MRI machine was wearing 9kg weight-training chain, wife reveals
A man who died after being pulled into an MRI machine in New York was wearing a large weight-training chain around his neck, his wife has said. Keith McAllister, 61, entered a room at the Nassau Open MRI clinic while a scan of his wife's knee was under way. The machine's strong magnetic force drew him in by the 9kg metal chain around his neck, according to Nassau County Police. His wife, Adrienne Jones-McAllister, said she had called out to her husband to help her off the table. "I yelled out Keith's name, [shouting] Keith, come help me up," she said in an interview with News 12 Long Island. She said her husband entered the room wearing the chain, which he uses for weight training. "I saw the machine snatch him around and pull him into the machine," Ms Jones-McAllister said as tears streamed down her face. "He died, he lost, he went limp in my arms." Police said that the accident last Wednesday "resulted in a medical episode" and left Mr McAllister in a critical condition in hospital. Ms Jones-McAllister said her husband had suffered a series of heart attacks after he was freed from the MRI machine. He was later pronounced dead. MRI stands for Magnetic Resonance Imaging. The machines use strong magnetic fields and radio waves to create detailed images of the inside of the body. Due to the magnetic fields, "very powerful forces" are exerted on objects made of iron, some steels, and other magnetic materials, the National Institute of Biomedical Imaging and Bioengineering says. It says the forces are "strong enough to fling a wheelchair across the room". Read more from Sky News:Trump diagnosed with chronic venous insufficiencyLA nightclub crowd crash may have been intentional, police say Sky News' US partner network NBC New York reported that MRI accidents are rare but can be fatal. It is not the first time someone has been killed by an MRI machine in New York. In 2001, six-year-old Michael Colombini died at the Westchester Medical Centre when an oxygen tank flew into the chamber, drawn in by the MRI's 10-ton electromagnet.
Yahoo
2 hours ago
- Yahoo
Northwest DC flooding traps drivers, several rescued
WASHINGTON (DC News Now) — Several people were rescued in Northwest D.C. after multiple drivers were trapped in floodwater Saturday, according to the DC Fire and EMS Department. MCFRS: Multiple people rescued after being trapped in floodwaters Around 5:40 p.m., three people were rescued unharmed from a car trapped in high water near 3rd Street and Underwood Street NW. Crews also reported multiple disabled vehicles in the 7100 block of Piney Branch Rd NW, where several people were safely removed from a car that was submerged in the water. No injuries have been reported so far in the flooding. Copyright 2025 Nexstar Media, Inc. All rights reserved. This material may not be published, broadcast, rewritten, or redistributed.
