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

2026-06-16

AI Summary

This study addresses the limitations of purely data-driven nowcasting models in stability and generalization by proposing a hybrid forecasting framework that integrates physical constraints with deep neural networks. The approach features an enhanced numerical solver supporting larger time steps, a unified autoregressive hybrid module to mitigate temporal overfitting, and state-of-the-art neural backbones, yielding two architectures: PI-PredFormer and PI-IAM4VP. By incorporating key physical mechanisms—such as the WENO-5 scheme, beta-plane approximation, and subgrid-scale viscosity—the method achieves 8–22% lower root mean square error for 1–12 hour forecasts and up to 26% reduction in daily mean squared error on a South Pacific dataset, while significantly improving the physical consistency of predictions.

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. 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 from 2000 to 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.

Problem

• short-term weather forecasting
• physics-constrained neural networks
• hybrid modeling
• physical consistency
• forecast accuracy

Innovation

• physics-constrained neural networks
• WENO-5
• autoregressive hybrid block
• numerical solver
• short-term weather forecasting