FastNet: Improving the physical consistency of machine-learning weather prediction models through loss function design

2025-09-22

AI Summary

MLWP models often lack physical consistency in deterministic forecasting, particularly in preserving small-scale structures and accurately predicting extreme events. To address this, we propose FastNet—a knowledge-integrated graph neural network for global weather forecasting. Our approach introduces three key innovations: (1) a corrected spherical harmonic loss and a horizontal gradient loss to enhance spectral fidelity and enforce dynamical constraints; (2) decoupled representation of wind speed and wind direction to mitigate directional bias; and (3) a multi-task loss function that jointly optimizes physical consistency and forecast accuracy. Experiments demonstrate that FastNet maintains competitive mean squared error while significantly suppressing unphysical artifacts, improving extreme wind-speed prediction accuracy and better preserving small-scale meteorological features. FastNet thus establishes a new paradigm for interpretable, high-fidelity machine learning–based weather modeling.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Computer Vision: Large Vision Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Machine learning weather prediction (MLWP) models have demonstrated remarkable potential in delivering accurate forecasts at significantly reduced computational cost compared to traditional numerical weather prediction (NWP) systems. However, challenges remain in ensuring the physical consistency of MLWP outputs, particularly in deterministic settings. This study presents FastNet, a graph neural network (GNN)-based global prediction model, and investigates the impact of alternative loss function designs on improving the physical realism of its forecasts. We explore three key modifications to the standard mean squared error (MSE) loss: (1) a modified spherical harmonic (MSH) loss that penalises spectral amplitude errors to reduce blurring and enhance small-scale structure retention; (2) inclusion of horizontal gradient terms in the loss to suppress non-physical artefacts; and (3) an alternative wind representation that decouples speed and direction to better capture extreme wind events. Results show that while the MSH and gradient-based losses extit{alone} may slightly degrade RMSE scores, when trained in combination the model exhibits very similar MSE performance to an MSE-trained model while at the same time significantly improving spectral fidelity and physical consistency. The alternative wind representation further improves wind speed accuracy and reduces directional bias. Collectively, these findings highlight the importance of loss function design as a mechanism for embedding domain knowledge into MLWP models and advancing their operational readiness.

Problem

• Improving physical consistency of machine learning weather prediction models
• Addressing physical realism issues in deterministic ML weather forecasts
• Enhancing spectral fidelity and reducing non-physical artifacts in forecasts

Innovation

• Modified spherical harmonic loss reduces blurring
• Gradient-based loss suppresses non-physical artefacts
• Decoupled wind representation improves extreme events