Google Launches WeatherNext 2, Results Faster And More Accurate Weather Forecast

JAKARTA Google, through Google DeepMind and Google Research, introduced the latest WeatherNext 2 weather forecast model.

In its official broadcast, Google stated that this AI model can produce weather forecasts eight times faster with a resolution of up to an hour. This can be done thanks to hundreds of possible weather scenarios.

From the test results, WeatherNext 2 showed a much better performance improvement. The model surpassed the previous model in 99.9 percent of weather variables, including temperature, wind and humidity. The model is able to produce more accurate predictions in a matter of hours.

WeatherNext 2 can predict hundreds of possible weather results from one starting point in less than a minute at one TPU. This process usually takes hours on supercomputers that use traditional physics-based models.

This whole performance improvement is supported by a new AI modeling called Functional Generative Network (FGN). This approach is able to predict 'joint', namely a large, complex, and interconnected weather system.

Google is committed to integrating this new technology into its various products. In the coming weeks, this technology will be used to strengthen weather information presented on Google Maps.

"By combining WeatherNext technology, we have now increased weather forecasts in Searches, Gemini, Pixel Weather, and the Google Maps Platform Weather API," Google said, quoted on Tuesday, November 18.

Google states that data access to this AI model is opened via Earth Engine and BigQuery. Google will provide research results from the development of WeatherNext 2 for a wider community.

"By providing sophisticated devices and open data, we hope to accelerate scientific discoveries and empower the global ecosystem," Google added.