FourCastNet
An AI global weather forecasting model developed by NVIDIA, trained on atmospheric reanalysis datasets. It generates forecasts without solving partial differential equations, achieving inference speeds several orders of magnitude faster than traditional numerical weather prediction methods. The model is integrated into the Earth-2 platform.
Timeline (last 90 days)
Machine learning models such as GraphCast, Pangu-Weather, and FourCastNet have already demonstrated accuracy comparable to or even better than top numerical models in short- and medium-range forecasting, with inference speeds thousands of times faster than traditional models.
AI weather models such as FourCastNet, Pangu-Weather, and GraphCast learn the statistical patterns of atmospheric dynamics by training on long-term reanalysis datasets such as ERA5
In Earth-2, NVIDIA combines its in-house FourCastNet with downstream downscaling models (such as diffusion models like CorrDiff) to scale up global-scale forecasts to finer regional resolutions
Huawei's Pangu-Weather large model adopts a 3D Swin Transformer architecture, while NVIDIA's FourCastNet adopts Adaptive Fourier Neural Operators (AFNO).
All Facts (4)
Machine learning models such as GraphCast, Pangu-Weather, and FourCastNet have already demonstrated accuracy comparable to or even better than top numerical models in short- and medium-range forecasting, with inference speeds thousands of times faster than traditional models.
50%UnverifiedAI weather models such as FourCastNet, Pangu-Weather, and GraphCast learn the statistical patterns of atmospheric dynamics by training on long-term reanalysis datasets such as ERA5
50%UnverifiedIn Earth-2, NVIDIA combines its in-house FourCastNet with downstream downscaling models (such as diffusion models like CorrDiff) to scale up global-scale forecasts to finer regional resolutions
50%UnverifiedHuawei's Pangu-Weather large model adopts a 3D Swin Transformer architecture, while NVIDIA's FourCastNet adopts Adaptive Fourier Neural Operators (AFNO).
50%