Advancing seasonal prediction of tropical cyclone activity with a hybrid AI-physics climate model

Machine learning (ML) models are successful with weather forecasting and have shown progress in climate simulations, yet leveraging them for useful climate predictions needs exploration. Here we show this feasibility using neural general circulation model (NeuralGCM), a hybrid ML-physics atmospheric...

Full description

Saved in:
Bibliographic Details
Main Authors: Gan Zhang, Megha Rao, Janni Yuval, Ming Zhao
Format: Article
Language:English
Published: IOP Publishing 2025-01-01
Series:Environmental Research Letters
Subjects:
Online Access:https://doi.org/10.1088/1748-9326/adf864
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849390315012620288
author Gan Zhang
Megha Rao
Janni Yuval
Ming Zhao
author_facet Gan Zhang
Megha Rao
Janni Yuval
Ming Zhao
author_sort Gan Zhang
collection DOAJ
description Machine learning (ML) models are successful with weather forecasting and have shown progress in climate simulations, yet leveraging them for useful climate predictions needs exploration. Here we show this feasibility using neural general circulation model (NeuralGCM), a hybrid ML-physics atmospheric model developed by Google, for seasonal predictions of large-scale atmospheric variability and Northern Hemisphere tropical cyclone (TC) activity. Inspired by physical model studies, we simplify boundary conditions, assuming sea surface temperature and sea ice follow their climatological cycle but persist anomalies present at the initialization time. With such forcings, NeuralGCM can generate 100 simulation days in ∼8 min with a single graphics processing unit while simulating realistic atmospheric circulation and TC climatology patterns. This configuration yields useful seasonal predictions (July–November) for the tropical atmosphere and various TC activity metrics. Notably, the predicted and observed TC frequency in the North Atlantic and East Pacific basins are significantly correlated during 1990–2023 ( r = ∼0.7), suggesting prediction skill comparable to existing physical GCMs. Despite challenges associated with model resolution and simplified boundary forcings, the model-predicted interannual variations demonstrate significant correlations with the observed sub-basin TC tracks ( p < 0.1) and basin-wide accumulated cyclone energy (ACE) ( p < 0.01) of the North Atlantic and North Pacific basins. These findings highlight the promise of leveraging ML models with physical insights to model TC risks and deliver seamless weather-climate predictions.
format Article
id doaj-art-26a3297fc474419f93278ce4b0b0f5a6
institution Kabale University
issn 1748-9326
language English
publishDate 2025-01-01
publisher IOP Publishing
record_format Article
series Environmental Research Letters
spelling doaj-art-26a3297fc474419f93278ce4b0b0f5a62025-08-20T03:41:43ZengIOP PublishingEnvironmental Research Letters1748-93262025-01-0120909403110.1088/1748-9326/adf864Advancing seasonal prediction of tropical cyclone activity with a hybrid AI-physics climate modelGan Zhang0https://orcid.org/0000-0002-7323-3409Megha Rao1Janni Yuval2Ming Zhao3Department of Climate, Meteorology, and Atmospheric Sciences, University of Illinois at Urbana-Champaign , Urbana, IL 61801, United States of AmericaDepartment of Climate, Meteorology, and Atmospheric Sciences, University of Illinois at Urbana-Champaign , Urbana, IL 61801, United States of AmericaGoogle Research , Mountain View, CA 94043, United States of AmericaGeophysical Fluid Dynamics Laboratory, National Oceanic and Atmospheric Administration , Princeton, NJ 08540, United States of AmericaMachine learning (ML) models are successful with weather forecasting and have shown progress in climate simulations, yet leveraging them for useful climate predictions needs exploration. Here we show this feasibility using neural general circulation model (NeuralGCM), a hybrid ML-physics atmospheric model developed by Google, for seasonal predictions of large-scale atmospheric variability and Northern Hemisphere tropical cyclone (TC) activity. Inspired by physical model studies, we simplify boundary conditions, assuming sea surface temperature and sea ice follow their climatological cycle but persist anomalies present at the initialization time. With such forcings, NeuralGCM can generate 100 simulation days in ∼8 min with a single graphics processing unit while simulating realistic atmospheric circulation and TC climatology patterns. This configuration yields useful seasonal predictions (July–November) for the tropical atmosphere and various TC activity metrics. Notably, the predicted and observed TC frequency in the North Atlantic and East Pacific basins are significantly correlated during 1990–2023 ( r = ∼0.7), suggesting prediction skill comparable to existing physical GCMs. Despite challenges associated with model resolution and simplified boundary forcings, the model-predicted interannual variations demonstrate significant correlations with the observed sub-basin TC tracks ( p < 0.1) and basin-wide accumulated cyclone energy (ACE) ( p < 0.01) of the North Atlantic and North Pacific basins. These findings highlight the promise of leveraging ML models with physical insights to model TC risks and deliver seamless weather-climate predictions.https://doi.org/10.1088/1748-9326/adf864climate predictionmachine learningtropical cycloneclimate modelmodel evaluation
spellingShingle Gan Zhang
Megha Rao
Janni Yuval
Ming Zhao
Advancing seasonal prediction of tropical cyclone activity with a hybrid AI-physics climate model
Environmental Research Letters
climate prediction
machine learning
tropical cyclone
climate model
model evaluation
title Advancing seasonal prediction of tropical cyclone activity with a hybrid AI-physics climate model
title_full Advancing seasonal prediction of tropical cyclone activity with a hybrid AI-physics climate model
title_fullStr Advancing seasonal prediction of tropical cyclone activity with a hybrid AI-physics climate model
title_full_unstemmed Advancing seasonal prediction of tropical cyclone activity with a hybrid AI-physics climate model
title_short Advancing seasonal prediction of tropical cyclone activity with a hybrid AI-physics climate model
title_sort advancing seasonal prediction of tropical cyclone activity with a hybrid ai physics climate model
topic climate prediction
machine learning
tropical cyclone
climate model
model evaluation
url https://doi.org/10.1088/1748-9326/adf864
work_keys_str_mv AT ganzhang advancingseasonalpredictionoftropicalcycloneactivitywithahybridaiphysicsclimatemodel
AT megharao advancingseasonalpredictionoftropicalcycloneactivitywithahybridaiphysicsclimatemodel
AT janniyuval advancingseasonalpredictionoftropicalcycloneactivitywithahybridaiphysicsclimatemodel
AT mingzhao advancingseasonalpredictionoftropicalcycloneactivitywithahybridaiphysicsclimatemodel