AI is helping to create a breakthrough in weather and climate forecasting

Stay informed with free updates

Simply log in to Climate change myFT Digest – delivered straight to your inbox.

Artificial intelligence has helped lead to a breakthrough in accurate long-term weather and climate forecasting, according to research that promises advances in forecasting and the wider use of machine learning.

Using a hybrid of machine learning and existing forecasting tools, a Google-led model called NeuralGCM successfully applied AI to conventional atmospheric physics models to track decades-long climate trends and extreme weather events like cyclones, a team of scientists found.

This combination of machine learning with established techniques could provide a template for improving the use of artificial intelligence in other areas from materials discovery to engineering design, the researchers suggest. NeuralGCM was much faster than traditional weather and climate forecasts and better than AI-only models at longer-term forecasts, they said.

“NeuralGCM shows that when we combine artificial intelligence with physics-based models, we can dramatically improve the accuracy and speed of atmospheric climate simulations,” said Stephan Hoyer, principal engineer at Google Research and co-author of the paper on the published work. in nature.

The paper reported that NeuralGCM proved to be faster, more accurate and consumed less computing power in tests against a current atmospheric physics instrument-based forecasting model called X-SHiELD, which is being developed by a branch of the US National Oceanic and Atmospheric Administration.

In one study, NeuralGCM identified nearly the same number of tropical cyclones as conventional extreme weather trackers and twice as many as X-SHiELD. In another test based on temperature and humidity levels during 2020, the error rate was 15 to 50 percent lower.

NeuralGCM calculations were able to generate 70,000 simulation days in 24 hours using one of Google’s customized AI tensor processing units, the paper reported. In contrast, for comparable calculations, X-SHiELD generated only 19 simulation days and required 13,824 computing units.

Google collaborated with the intergovernmental European Center for Medium-Range Weather Forecasts (ECMWF) to develop NeuralGCM.

The European group published their model in June, and Google has made the code for NeuralGCM open access. It uses 80 years of ECMWF observational and reanalysis data for machine learning.

Last year, Google’s DeepMind unit introduced an AI-only weather forecasting model called GraphCast that outperformed conventional methods for up to 10 days ahead.

Established forecasting agencies such as the UK Met Office also have projects to integrate machine learning into their work.

Peter Dueben, head of ECMWF Earth System Modeling and co-author of the latest paper, said AI-only models were “often viewed with skepticism” by experts because they were not based on mathematical equations derived from physics.

Combining a physics-based model with a deep learning model “seems to bring the best of both worlds,” he said, adding that the approach is “a big step toward machine learning climate modeling.”

More “work” remained to be done, such as enabling NeuralGCM to estimate the impact of rising CO₂ on global surface temperatures, Dueben said. Other areas where the model needed further improvement included its ability to simulate an unprecedented climate, the paper said.

An expert not involved in the work, Cédric M. John, head of data science for environment and sustainability at Queen Mary University of London, said there was “compelling evidence” that NeuralGCM is more accurate than machine learning alone and faster than “full physical model. While there is still “room for improvement,” the possibility of error should be measurable and upgrades should be possible, he suggested.

“Importantly, this hybrid model captures the set of forecasts well, and the practical implication of this is that an estimate of the uncertainty of the forecast can be derived,” said John.

Google is involved in a growing number of environmental stewardship initiatives. It provides technology support for a satellite mission to monitor planetary warming methane emissions and partners with NASA, the US space agency, to help local governments monitor air quality.