2025-08-25 ワシントン大学(UW)
NASA Earth Observing System/Interdisciplinary Science (IDS) program under the Earth Science Enterprise (ESE)
＜関連情報＞
- https://www.washington.edu/news/2025/08/25/ai-simulates-1000-years-of-climate/
- https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2025AV001706
観測された気候の効率的なシミュレーションのためのディープラーニング地球システムモデル A Deep Learning Earth System Model for Efficient Simulation of the Observed Climate
Nathaniel Cresswell-Clay, Bowen Liu, Dale R. Durran, Zihui Liu, Zachary I. Espinosa, Raul A. Moreno, Matthias Karlbauer
AGU Advances published: 25 August 2025
DOI:https://doi.org/10.1029/2025AV001706
Abstract
A key challenge for computationally intensive state-of-the-art Earth System models is to distinguish global warming signals from interannual variability. Here we introduce Deep Learning Earth System Model (DLESyM), a parsimonious deep learning model that accurately simulates the Earth’s current climate over 1000-year periods with minimal smoothing and no drift. DLESyM simulations equal or exceed key metrics of seasonal and interannual variability—such as tropical cyclogenesis over the range of observed intensities, the cycle of the Indian Summer monsoon, and the climatology of mid-latitude blocking events—when compared to historical simulations from four leading models from the sixth Climate Model Intercomparison Project. DLESyM, trained on both historical reanalysis data and satellite observations, is an accurate, highly efficient model of the coupled Earth system, empowering long-range sub-seasonal and seasonal forecasts while using a fraction of the energy and computational time required by traditional models.
Key Points
- The coupled atmosphere-ocean Deep Learning Earth System Model (DLESyM) simulates 1,000 years of equilibrium climate in less than 12 hr
- While only trained to optimize MSE over 24 hr in the atmosphere, DLESyM produces realistic climatology and interannual variability
- The equilibrium climate simulated by DLESyM has high fidelity to observed climate and is competitive with CMIP6 models
Plain Language Summary
Machine learning (ML) based atmosphere models have recently demonstrated the ability to accurately predict weather over 10 days, and have proven superior to operational weather prediction systems based on conventional numerical methods. Yet many of these ML weather models fail to recreate our atmosphere over longer simulations during which predicted states become overly smooth, physically unrealistic, or entirely unstable. In this work we build on canonical understanding of Earth System variability by coupling an ML weather model to an ML model for sea-surface temperature. The resulting Deep Learning Earth SYstem Model, DLESyM, is both parsimonious in its design and capable of producing thousands of years of realistic atmosphere and ocean states including seasonal cycles, spontaneous tropical cyclones, strong winter storms, and even interannual variability. Our results are compared to those of conventional numerical Earth System models from CMIP6 demonstrating DLESyM’s competitive fidelity to the internal variability of the Earth System. Crucially the computational cost of running DLESyM is significantly less compared to traditional models, making it a powerful yet affordable tool for the study of weather and climate variability.
