Artificial Intelligence Weather Forecasting

[GordMay]

The AI weather forecast has arrived

A revolution is rocking long-range weather forecasts.
Thought infeasible, only a few years ago, Artificial Intelligence, trained on weather data and models can now make forecasts on par with supercomputers, up to 10 days in advance, taking only minutes to run, on a robust desktop computer.

Some of the world’s biggest tech giants are jockeying to claim the most skillful model, including Google DeepMind, which describes its GraphCast model in Science last week [1], and Huawei, which published a similar model, called Pangu-Weather, in Nature earlier this year [2]. Google also has a short-term AI weather model, MetNet-3 [3], that makes rolling 24-hour predictions that are more accurate than nearly any weather agency’s.

The models have grown so accurate that the European Centre for Medium-Range Weather Forecasts [ECMWF], the world’s top forecaster, has embraced the technology, developing its own model, which it began publishing last month. [4a & b]

But these AI forecasters also struggle with some extremes, including hurricane intensity. Still, researchers now expect the cheap computing required by these models will open up the field to many new innovations going forward.

[1] “Learning skillful medium-range global weather forecasting” ~ by Remi Lam, et al
➥ https://www.science.org/doi/10.1126/...et_cid=4985621

[2] “Accurate medium-range global weather forecasting with 3D neural networks” ~ by Kaifeng Bi, et al
➥ https://www.nature.com/articles/s41586-023-06185-3

[3] “MetNet-3: A state-of-the-art neural weather model available in Google products”
➥ https://blog.research.google/2023/11...l-weather.html

[4a] Artificial Intelligence/Integrated Forecasting System (AIFS) ~ by ECMWF
➥ https://www.ecmwf.int/en/newsletter/.../aifs-launched

[4b] ECMWF Experimental Model
➥ https://charts.ecmwf.int/?facets=%7B...odels%22%5D%7D


Traditional weather models start by feeding a snapshot of current conditions, based on observations from satellites, weather stations, and buoys, into a gridlike computer model that divides the atmosphere into millions of boxes. The snapshot is run forward in time by applying the physical laws of fluid dynamics to each box—at great computational expense. The models can take several hours to run on supercomputers with 1 million processors, and weather agencies typically produce updates just four times a day.

The new AI models skip the expense of solving equations in favor of “deep learning.” They identify patterns in the way the atmosphere naturally evolves, after training on 40 years of ECMWF “reanalysis” data—a combination of observations and short-term model forecasts that represents modelers’ best and most complete picture of past weather. When fed a starting snapshot of the atmosphere based on the same combination of observations and modeling, GraphCast can outperform the ECMWF forecast out to 10 days on 90% of its verification targets, including hurricane tracks and extreme temperatures. Although it took 32 computers 4 weeks to train the AI model, the resulting algorithm is lightweight enough to work in less than 1 minute on a single desktop computer, says Rémi Lam, lead author of the GraphCast paper. “It is fast, accurate, and useful.”