Physics-Constrained Neural Networks for Improved Short-Term Weather Forecasting: A Case Study over the South Pacific

Egor Bugaev, Fedor Buzaev, Dmitry Efremenko, Denis Derkach, Fedor Ratnikov

Keywords: Weather Forecast, Neural Networks, Physics, NWP, physics-informed neural networks, WeatherBench, WeatherGFT

TL;DR: Two novel hybrid architectures for middle-range Weather Forecasting

Abstract: This study introduces enhancements to physics-constrained neural networks (PCNNs) that improve the accuracy and stability of hybrid short-term weather forecasting models. Building on the WeatherGFT architecture, three innovations are proposed. First, an upgraded numerical solver—combining a fifth-order weighted essentially non-oscillatory scheme (WENO-5), a $\beta$-plane approximation, and subgrid-scale viscosity—permits a fourfold increase in the integration time step (to 1200 s) while reducing the daily mean squared error by up to 26%. Second, a unified autoregressive hybrid block replaces the original chain of 24 specialised modules, eliminating overfitting to specific lead times; and third, the physical core is integrated with two state-of-the-art neural backbones, resulting in PI-PredFormer and PI-IAM4VP. Evaluation on the WeatherBench South Pacific subset (2000–2004) shows that these hybrids reduce root mean squared error at 1–12 h lead times by 8–22% compared to purely neural counterparts, while better preserving physical consistency. These results demonstrate that incremental refinement of hybrid components offers a practical route toward more accurate and efficient short-range weather forecasting.

Journal Opt In: No, I do not wish to participate

Submission Number: 116