On Some Limitations of Current Machine Learning Weather Prediction Models

Abstract Machine Learning (ML) is having a profound impact in the domain of Weather and Climate Prediction. A recent development in this area has been the emergence of fully data‐driven ML prediction models which routinely claim superior performance to that of traditional physics‐based models. We ex...

Full description

Saved in:
Bibliographic Details
Main Author: Massimo Bonavita
Format: Article
Language:English
Published: Wiley 2024-06-01
Series:Geophysical Research Letters
Subjects:
Online Access:https://doi.org/10.1029/2023GL107377
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850271810289926144
author Massimo Bonavita
author_facet Massimo Bonavita
author_sort Massimo Bonavita
collection DOAJ
description Abstract Machine Learning (ML) is having a profound impact in the domain of Weather and Climate Prediction. A recent development in this area has been the emergence of fully data‐driven ML prediction models which routinely claim superior performance to that of traditional physics‐based models. We examine some aspects of the forecasts produced by three of the leading current ML models, Pangu‐Weather, FourCastNet and GraphCast, with a focus on their fidelity and physical consistency. The main conclusion is that these ML models are not able to properly reproduce sub‐synoptic and mesoscale weather phenomena and lack the fidelity and physical consistency of physics‐based models and this has impacts on the interpretation of their forecasts and their perceived skill. Balancing forecast skill and physical realism will be an important consideration for future ML models.
format Article
id doaj-art-6a3bd5d21d574d65a0bc0c3d9d10999d
institution OA Journals
issn 0094-8276
1944-8007
language English
publishDate 2024-06-01
publisher Wiley
record_format Article
series Geophysical Research Letters
spelling doaj-art-6a3bd5d21d574d65a0bc0c3d9d10999d2025-08-20T01:52:06ZengWileyGeophysical Research Letters0094-82761944-80072024-06-015112n/an/a10.1029/2023GL107377On Some Limitations of Current Machine Learning Weather Prediction ModelsMassimo Bonavita0ECMWF Reading UKAbstract Machine Learning (ML) is having a profound impact in the domain of Weather and Climate Prediction. A recent development in this area has been the emergence of fully data‐driven ML prediction models which routinely claim superior performance to that of traditional physics‐based models. We examine some aspects of the forecasts produced by three of the leading current ML models, Pangu‐Weather, FourCastNet and GraphCast, with a focus on their fidelity and physical consistency. The main conclusion is that these ML models are not able to properly reproduce sub‐synoptic and mesoscale weather phenomena and lack the fidelity and physical consistency of physics‐based models and this has impacts on the interpretation of their forecasts and their perceived skill. Balancing forecast skill and physical realism will be an important consideration for future ML models.https://doi.org/10.1029/2023GL107377machine learningnumerical weather predictiondata‐driven forecast models
spellingShingle Massimo Bonavita
On Some Limitations of Current Machine Learning Weather Prediction Models
Geophysical Research Letters
machine learning
numerical weather prediction
data‐driven forecast models
title On Some Limitations of Current Machine Learning Weather Prediction Models
title_full On Some Limitations of Current Machine Learning Weather Prediction Models
title_fullStr On Some Limitations of Current Machine Learning Weather Prediction Models
title_full_unstemmed On Some Limitations of Current Machine Learning Weather Prediction Models
title_short On Some Limitations of Current Machine Learning Weather Prediction Models
title_sort on some limitations of current machine learning weather prediction models
topic machine learning
numerical weather prediction
data‐driven forecast models
url https://doi.org/10.1029/2023GL107377
work_keys_str_mv AT massimobonavita onsomelimitationsofcurrentmachinelearningweatherpredictionmodels