_version_ 1866916960451690496
author Dunstan, Tom
Strickson, Oliver
Bennett, Thusal
Bowyer, Jack
Burnand, Matthew
Chappell, James
Coca-Castro, Alejandro
Dale, Kirstine Ida
Daub, Eric G.
Eftekhari, Noushin
Janmaijaya, Manvendra
Lillis, Jon
Salvador-Jasin, David
Simpson, Nathan
Chan, Ryan Sze-Yin
Elmasri, Mohamad
France, Lydia Allegranza
Madge, Sam
Bokeria, Levan
Brown, Hannah
Dodds, Tom
Ellis, Anna-Louise
Llewellyn-Jones, David
McCaie, Theo
Moreton, Sophia
Potter, Tom
Robinson, James
Scaife, Adam A.
Stenson, Iain
Walters, David
Bett-Williams, Karina
van Zeeland, Louisa
Yatsyshin, Peter
Hosking, J. Scott
author_facet Dunstan, Tom
Strickson, Oliver
Bennett, Thusal
Bowyer, Jack
Burnand, Matthew
Chappell, James
Coca-Castro, Alejandro
Dale, Kirstine Ida
Daub, Eric G.
Eftekhari, Noushin
Janmaijaya, Manvendra
Lillis, Jon
Salvador-Jasin, David
Simpson, Nathan
Chan, Ryan Sze-Yin
Elmasri, Mohamad
France, Lydia Allegranza
Madge, Sam
Bokeria, Levan
Brown, Hannah
Dodds, Tom
Ellis, Anna-Louise
Llewellyn-Jones, David
McCaie, Theo
Moreton, Sophia
Potter, Tom
Robinson, James
Scaife, Adam A.
Stenson, Iain
Walters, David
Bett-Williams, Karina
van Zeeland, Louisa
Yatsyshin, Peter
Hosking, J. Scott
contents 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 \textit{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.
format Preprint
id arxiv_https___arxiv_org_abs_2509_17601
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle FastNet: Improving the physical consistency of machine-learning weather prediction models through loss function design
Dunstan, Tom
Strickson, Oliver
Bennett, Thusal
Bowyer, Jack
Burnand, Matthew
Chappell, James
Coca-Castro, Alejandro
Dale, Kirstine Ida
Daub, Eric G.
Eftekhari, Noushin
Janmaijaya, Manvendra
Lillis, Jon
Salvador-Jasin, David
Simpson, Nathan
Chan, Ryan Sze-Yin
Elmasri, Mohamad
France, Lydia Allegranza
Madge, Sam
Bokeria, Levan
Brown, Hannah
Dodds, Tom
Ellis, Anna-Louise
Llewellyn-Jones, David
McCaie, Theo
Moreton, Sophia
Potter, Tom
Robinson, James
Scaife, Adam A.
Stenson, Iain
Walters, David
Bett-Williams, Karina
van Zeeland, Louisa
Yatsyshin, Peter
Hosking, J. Scott
Atmospheric and Oceanic Physics
Machine Learning
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 \textit{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.
title FastNet: Improving the physical consistency of machine-learning weather prediction models through loss function design
topic Atmospheric and Oceanic Physics
Machine Learning
url https://arxiv.org/abs/2509.17601