![\"\"](\"https:\/\/images.newscientist.com\/wp-content\/uploads\/2024\/07\/19145025\/SEI_213323563.jpg\")
<\/div>AI could help predict the weather more accurately<\/p>\n
Ranimiro Lotufo Neto\/Alamy<\/p>\n<\/div>\n<\/figcaption><\/figure>\n<\/p>\n
Google researchers have built an artificial intelligence that they say can forecast weather and climate patterns just as well as current physics models while also requiring less computer power.<\/p>\n
Existing forecasts are based on mathematical models run by enormously powerful supercomputers<\/a> that deterministically predict what will happen in the future. Since they were first used in the 1950s, these models have grown more and more detailed, requiring ever more computer power.<\/p>\n Several projects have aimed to replace these intense calculations with much less demanding AI, including a DeepMind tool to forecast rain locally<\/a> on short timescales. But like most AI models, these are a \u201cblack box\u201d whose inner workings are a mystery, and the inability to explain or replicate their methods is problematic. Climate scientists also point out that if the models are trained on historical data, they will struggle to predict unprecedented phenomena now occurring due to climate change.<\/p>\n Now, Dmitrii Kochkov<\/a> at Google Research in California and his colleagues have created a model called NeuralGCM that they believe strikes a balance between the two approaches.<\/p>\n <\/p>\n Typical climate models divide Earth\u2019s surface into a grid of cells up to 100 kilometres across; the limits of computing power make it impractical to simulate at higher resolutions. Phenomena like clouds, air turbulence and convection inside those cells are merely approximated by computer code that is continually tweaked to more accurately match observational data. This approach, called parameterisation, hopes to capture, at least partially, the small-scale phenomena that the wider physics model cannot.<\/p>\n NeuralGCM is trained to take over this small-scale approximation, making it less computationally intensive and more accurate. In a paper, the researchers say that the model can process 70,000 days of simulation in 24\u2009hours using a single chip called a tensor processing unit (TPU).<\/a> In comparison, a competing model called X-SHiELD<\/a> uses a supercomputer with thousands of processing units to process just 19 days of simulation.<\/p>\n The paper also claims that NeuralGCM produces forecasts with accuracy comparable to, and sometimes better than, best-in-class models. Google didn\u2019t respond to an interview request from New Scientist.<\/em><\/p>\n