Google DeepMind and Google Research have introduced the world’s most modern and accurate weather model powered by artificial intelligence, ‘WeatherNext 3’. Unlike older systems, this new model has the ability to make forecasts every hour with the help of live satellite and weather station data, which is 5 times more detailed than previous models.
According to a recent independent assessment by ‘Brightband’, ‘WeatherNext’ is currently the world’s most reliable and accurate global weather model available.
Older artificial intelligence models, which generally relied on numerical weather data, were different from this new system, which learns directly from ground and satellite observations.
The new model can forecast important weather changes such as temperature and humidity at an impressive resolution of 5 kilometres. For other ground variables, a resolution of 10 kilometres has been used, while a resolution of 25 kilometres has been used to assess wind speed.
According to ‘Google’, this system is 5 times more detailed than ‘WeatherNext 2’, which provided data every 6 hours on a 25-kilometre grid.
Special Features for Rain, Snowfall and Renewable Energy
The new model has particularly improved the forecasting of rain and snowfall. This system has been trained on NASA’s ‘IMERG’ and its own radar data.
During testing, it showed an improvement of up to 60 per cent in medium-range forecasts and up to 30 per cent in initial forecasts.
Furthermore, ‘WeatherNext 3’ has also been specifically designed for renewable energy production. It can accurately measure wind speed at a height of 100 meters (which is the height of wind turbines) and cloud density and sunlight for solar energy projects.
Availability and Warning
Google has made the data from this model available to researchers, developers and businesses through ‘BigQuery’, ‘Earth Engine’ and ‘Google Cloud Storage’. From today, this system will also start working on ‘Google Search’, the ‘Gemini app’ and ‘Google Maps’, which will bring an improvement of up to 50 percent in long-term forecasts.
However, the company has warned that weather is naturally uncertain; therefore, for official warnings regarding emergencies and disasters, the public should rely only on their local meteorological department.
Due to climate change, sudden floods, storms and incidents of heavy rainfall have increased around the world. Traditional weather systems run on supercomputers, in which data processing can take up to 6 hours, making it difficult to detect suddenly changing weather in a timely manner.
Particularly in regions such as Latin America, Africa and the Asia-Pacific, where the use of expensive traditional models was limited, there was a strong need for cheaper and immediate forecasting models.