Fixing the Double Penalty in Data-Driven Weather Forecasting Through a Modified Spherical Harmonic Loss Function

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Subich, Christopher, Husain, Syed Zahid, Separovic, Leo, Yang, Jing
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912403074056192
author Subich, Christopher
Husain, Syed Zahid
Separovic, Leo
Yang, Jing
author_facet Subich, Christopher
Husain, Syed Zahid
Separovic, Leo
Yang, Jing
contents Recent advancements in data-driven weather forecasting models have delivered deterministic models that outperform the leading operational forecast systems based on traditional, physics-based models. However, these data-driven models are typically trained with a mean squared error loss function, which causes smoothing of fine scales through a "double penalty" effect. We develop a simple, parameter-free modification to this loss function that avoids this problem by separating the loss attributable to decorrelation from the loss attributable to spectral amplitude errors. Fine-tuning the GraphCast model with this new loss function results in sharp deterministic weather forecasts, an increase of the model's effective resolution from 1,250km to 160km, improvements to ensemble spread, and improvements to predictions of tropical cyclone strength and surface wind extremes.
format Preprint
id arxiv_https___arxiv_org_abs_2501_19374
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Fixing the Double Penalty in Data-Driven Weather Forecasting Through a Modified Spherical Harmonic Loss Function
Subich, Christopher
Husain, Syed Zahid
Separovic, Leo
Yang, Jing
Machine Learning
Atmospheric and Oceanic Physics
I.2.6; I.2.1; J.2
Recent advancements in data-driven weather forecasting models have delivered deterministic models that outperform the leading operational forecast systems based on traditional, physics-based models. However, these data-driven models are typically trained with a mean squared error loss function, which causes smoothing of fine scales through a "double penalty" effect. We develop a simple, parameter-free modification to this loss function that avoids this problem by separating the loss attributable to decorrelation from the loss attributable to spectral amplitude errors. Fine-tuning the GraphCast model with this new loss function results in sharp deterministic weather forecasts, an increase of the model's effective resolution from 1,250km to 160km, improvements to ensemble spread, and improvements to predictions of tropical cyclone strength and surface wind extremes.
title Fixing the Double Penalty in Data-Driven Weather Forecasting Through a Modified Spherical Harmonic Loss Function
topic Machine Learning
Atmospheric and Oceanic Physics
I.2.6; I.2.1; J.2
url https://arxiv.org/abs/2501.19374