Google DeepMind Introduces WeatherNext 2 AI Model For Accurate Global Weather Forecasts
In Brief
Google DeepMind has launched WeatherNext 2, delivering faster, higher-resolution, and multi-scenario global weather forecasts to support more accurate and actionable decision-making.
Google DeepMind, the AI division of Google, has introduced WeatherNext 2, its most advanced system to date for generating global weather forecasts with improved accuracy and higher resolution.
WeatherNext 2 can produce forecasts up to eight times faster, with temporal resolution as precise as one hour, thanks to a new model capable of generating hundreds of potential scenarios. This approach has been used to assist weather agencies in decision-making, including experimental cyclone predictions.
The system is now being made accessible to users, with forecast data available through Google Earth Engine and BigQuery. Additionally, an early access program has been launched on Google Cloud’s Vertex AI platform to allow custom model inference.
Integration of WeatherNext technology has already enhanced weather forecasts across Google Search, Gemini, Pixel Weather, and the Google Maps Platform Weather API, and in the coming weeks, it will also support weather information within Google Maps.
WeatherNext 2 Introduces AI-Powered Functional Generative Networks For Better Weather Forecasts
Accurate weather forecasting requires capturing the full range of possible outcomes, including extreme scenarios that are critical for planning. WeatherNext 2 is capable of generating hundreds of potential weather outcomes from a single initial condition, with each prediction taking under a minute on a single TPU—an operation that would require hours using traditional physics-based supercomputer models.
The system delivers highly skillful, high-resolution forecasts down to the hour, outperforming the previous WeatherNext model on 99.9% of variables, including temperature, wind, and humidity, across lead times of 0 to 15 days. This enables more precise and actionable predictions.
The enhanced performance is achieved through a new AI modeling approach known as a Functional Generative Network (FGN), which introduces controlled ‘noise’ directly into the model architecture, ensuring that forecasts remain physically realistic and internally consistent.
This methodology is particularly effective for predicting both ‘marginals’—individual weather elements such as temperature at a location, wind speed at a certain altitude, or humidity—and ‘joints,’ which are complex, interconnected systems that depend on the relationships between these individual elements. Although the model is trained only on marginals, it can accurately infer joints, allowing it to forecast large-scale patterns, such as regions experiencing extreme heat or the expected power output of an entire wind farm.
With WeatherNext 2, advanced research is being applied to practical, high-impact weather forecasting. Efforts continue to refine and enhance the technology while making the latest tools accessible to the global community.
Future work includes exploring additional data sources and expanding availability to reach more users. By providing robust tools and open data, the initiative aims to support scientific discovery and enable researchers, developers, and organizations worldwide to make informed decisions on complex challenges and drive innovation for the future.
Disclaimer
In line with the Trust Project guidelines, please note that the information provided on this page is not intended to be and should not be interpreted as legal, tax, investment, financial, or any other form of advice. It is important to only invest what you can afford to lose and to seek independent financial advice if you have any doubts. For further information, we suggest referring to the terms and conditions as well as the help and support pages provided by the issuer or advertiser. MetaversePost is committed to accurate, unbiased reporting, but market conditions are subject to change without notice.
About The Author
Alisa, a dedicated journalist at the MPost, specializes in cryptocurrency, zero-knowledge proofs, investments, and the expansive realm of Web3. With a keen eye for emerging trends and technologies, she delivers comprehensive coverage to inform and engage readers in the ever-evolving landscape of digital finance.
More articles
Alisa, a dedicated journalist at the MPost, specializes in cryptocurrency, zero-knowledge proofs, investments, and the expansive realm of Web3. With a keen eye for emerging trends and technologies, she delivers comprehensive coverage to inform and engage readers in the ever-evolving landscape of digital finance.
