Benchmarking Physics-Informed Time-Series Models for Operational Global Station Weather Forecasting

View PDF HTML (experimental)

Abstract:The development of Time-Series Forecasting (TSF) models is often constrained by the lack of comprehensive datasets, especially in Global Station Weather Forecasting (GSWF), where existing datasets are small, temporally short, and spatially sparse. To address this, we introduce WEATHER-5K, a large-scale observational weather dataset that better reflects real-world conditions, supporting improved model training and evaluation. While recent TSF methods perform well on benchmarks, they lag behind operational Numerical Weather Prediction systems in capturing complex weather dynamics and extreme events. We propose PhysicsFormer, a physics-informed forecasting model combining a dynamic core with a Transformer residual to predict future weather states. Physical consistency is enforced via pressure-wind alignment and energy-aware smoothness losses, ensuring plausible dynamics while capturing complex temporal patterns. We benchmark PhysicsFormer and other TSF models against operational systems across several weather variables, extreme event prediction, and model complexity, providing a comprehensive assessment of the gap between academic TSF models and operational forecasting. The dataset and benchmark implementation are available at: this https URL.

Comments: 34 pages, 20 figures

Subjects:

Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV); Atmospheric and Oceanic Physics (physics.ao-ph); Machine Learning (stat.ML)

Cite as: arXiv:2406.14399 [cs.LG]

(or arXiv:2406.14399v3 [cs.LG] for this version)

https://doi.org/10.48550/arXiv.2406.14399

arXiv-issued DOI via DataCite

Submission history

From: Tao Han [view email] [v1] Thu, 20 Jun 2024 15:18:52 UTC (19,019 KB) [v2] Fri, 11 Oct 2024 18:34:16 UTC (11,314 KB) [v3] Tue, 31 Mar 2026 14:54:47 UTC (21,201 KB)