Google has presented its new machine learning (ML) approach that can rapidly, efficiently and accurately simulate Earth’s atmosphere in the latest issue of Nature.
In the article, Neural general circulation models for weather and climate, the company outlines NeuralGCM, a ML-based approach to simulating Earth’s atmosphere that combines traditional physics-based modeling with ML for improved simulation accuracy and efficiency.
Developed in partnership with the European Centre for Medium-Range Weather Forecasts (ECMWF), NeuralGCM generates 2-15 day weather forecasts that are more accurate than the current gold-standard physics-based model, and reproduces temperatures over a past 40-year period more accurately than traditional atmospheric models.
Although Google has not yet built NeuralGCM into a full climate model, it believes the solution marks a significant step toward developing more powerful and accessible climate models. “We hope it will eventually enable a more accurate and actionable understanding of how our climate is changing,” Stephan Hoyer, senior staff software engineer at Google Research said in a blog on its website.
Transforming climate modeling
Like a traditional climate model, NeuralGCM divides the Earth’s atmosphere into cubes and runs calculations on the physics of large-scale processes like air and moisture movement. But instead of depending on parameterizations formulated by scientists to simulate small-scale aspects like cloud formation, it uses a neural network to learn the physics of those events from existing weather data.
“A key innovation of NeuralGCM is that we rewrote the numerical solver for large-scale processes from scratch inJAX. This allowed us to use gradient-based optimization to tune the behavior of the coupled system ‘online’ over many time steps,” Hoyer wrote. “In contrast, prior attempts to enhance climate models with ML struggled greatly with numerical stability, because they used ‘offline’ training, which ignores critical feedback between small- and large-scale processes that accumulates over time.
“Another bonus of writing the entire model in JAX is that it runs efficiently on TPUs and GPUs, in contrast to traditional climate models that mostly run on CPUs.
“We trained a suite of NeuralGCM models using weather data from ECMWF from 1979 to 2019 at 0.7°, 1.4° and 2.8° resolution. Although our models were trained on weather forecasts, we designed NeuralGCM to be a general-purpose atmospheric model,” Hoyer added.
Expansion plans
NeuralGCM currently models just Earth’s atmosphere but Google has said it hopes to eventually include other aspects of Earth’s climate system, such as oceans and the carbon cycle, into the model.
“By doing so, we’ll allow NeuralGCM to make predictions on longer timescales, going beyond predicting weather over days and weeks to making forecasts on climate timescales,” Hoyer wrote.
