FourCastNet 3: A geometric approach to probabilistic machine-learning weather forecasting at scale

2025-07-16

AI Summary

Addressing challenges in spherical geometric modeling and multiscale spatial correlation for global probabilistic ensemble weather forecasting, this paper proposes a convolutional neural network architecture specifically designed for spherical geometry, along with a model-data hybrid parallel training paradigm inspired by numerical model domain decomposition. The method rigorously respects spherical topology, ensuring spectral stability and dynamical consistency. It enables efficient training on thousand-GPU clusters and achieves single-GPU inference in just 20 seconds for 90-day global forecasts at 0.25° resolution and 6-hour temporal intervals. Compared to conventional ensemble systems, the method significantly improves forecast accuracy; it matches state-of-the-art diffusion models in skill while accelerating inference by 8–60×. The approach further guarantees physical interpretability, probabilistic calibration, and computational scalability.

Technology Category

• Machine Learning: Ensemble Methods
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Computer Vision: Diffusion Models for Vision

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

FourCastNet 3 advances global weather modeling by implementing a scalable, geometric machine learning (ML) approach to probabilistic ensemble forecasting. The approach is designed to respect spherical geometry and to accurately model the spatially correlated probabilistic nature of the problem, resulting in stable spectra and realistic dynamics across multiple scales. FourCastNet 3 delivers forecasting accuracy that surpasses leading conventional ensemble models and rivals the best diffusion-based methods, while producing forecasts 8 to 60 times faster than these approaches. In contrast to other ML approaches, FourCastNet 3 demonstrates excellent probabilistic calibration and retains realistic spectra, even at extended lead times of up to 60 days. All of these advances are realized using a purely convolutional neural network architecture tailored for spherical geometry. Scalable and efficient large-scale training on 1024 GPUs and more is enabled by a novel training paradigm for combined model- and data-parallelism, inspired by domain decomposition methods in classical numerical models. Additionally, FourCastNet 3 enables rapid inference on a single GPU, producing a 90-day global forecast at 0.25°, 6-hourly resolution in under 20 seconds. Its computational efficiency, medium-range probabilistic skill, spectral fidelity, and rollout stability at subseasonal timescales make it a strong candidate for improving meteorological forecasting and early warning systems through large ensemble predictions.

Problem

• Develops scalable geometric ML for probabilistic weather forecasting
• Ensures accurate modeling of spherical geometry and spatial correlations
• Achieves fast, high-resolution forecasts with superior computational efficiency

Innovation

• Geometric ML approach for probabilistic forecasting
• Scalable convolutional network for spherical geometry
• Efficient training on 1024 GPUs via domain decomposition