Google Labs says a new artificial intelligence-based weather model can provide 15-day forecasts with unparalleled accuracy and speed, with the potential to save lives as climate change intensifies.
GenCast, invented by Google DeepMind, the London-based artificial intelligence research lab, “demonstrated better predictive ability than the current world-leading model,” the company said Wednesday.
The European Centre for Medium-Range Weather Forecasts (ECMWF) provides forecasts for 35 countries and is regarded as the global benchmark for meteorological accuracy.
But DeepMind said GenCast outperformed the center’s predictions with more than 97% accuracy in 2019 tests of 1,320 real-world scenarios.
Details of their research findings were published in the leading scientific journal Nature.
Florence Rabier, head of the European Centre for Medium-Range Weather Forecasts, told AFP the project was “a first step” toward incorporating artificial intelligence into weather forecasts. “It’s really a leap forward.”
She said at this stage it could be used to supplement their current model.
“We are making progress every year,” she added. “Against the backdrop of enormous societal pressures around climate change, any new approach that can strengthen and accelerate that progress is very welcome.”
The model, trained on four decades of temperature, wind speed and air pressure data from 1979 to 2018, can produce a 15-day forecast in just eight minutes, compared with hours currently.
“GenCast’s forecasts for both daily weather and extreme events are more accurate than top operating systems.. up to 15 days in advance,” DeepMind said in a statement.
Scientists warn that extreme weather is becoming more common and more severe due to human-caused climate change.
Last August, a series of wildfires in Hawaii killed about 100 people. Local residents criticized authorities for not receiving early warning of the fires.
This summer, a sudden heat wave in Morocco killed at least 21 people in 24 hours. In September, Hurricane Helene killed 237 people in Florida and other southeastern states.
DeepMind said GenCast “consistently outperformed” the current leading forecasting models in predicting extreme heat, extreme cold, and high wind speeds.
“More accurate predictions of extreme weather risks could help officials protect more lives, avoid damage, and save money,” DeepMind said.