Finetuning AI Foundation Models to Develop Subgrid-Scale Parameterizations: A Case Study on Atmospheric Gravity Waves
🤖 AI Summary
Parameterizations of unresolved processes—such as gravity waves—in climate models introduce significant uncertainty. Method: This study pioneers the adaptation of a large-scale pretrained AI foundation model (Prithvi WxC) to subgrid-scale parameterization, fine-tuning its encoder-decoder architecture and integrating high-resolution reanalysis data to develop a physically consistent gravity wave flux parameterization. Prediction distribution similarity is quantified via the Hellinger distance. Contribution/Results: The proposed method substantially outperforms an Attention U-Net baseline in both monthly-mean and instantaneous field predictions, reducing the Hellinger distance from 0.11 to 0.06. This work establishes a novel paradigm for leveraging pretrained foundation models in Earth system subgrid modeling, uniquely combining cross-regional generalizability with deep integration of data-driven learning and physical constraints.
Application Category
📝 Abstract
Global climate models parameterize a range of atmospheric-oceanic processes like gravity waves, clouds, moist convection, and turbulence that cannot be sufficiently resolved. These subgrid-scale closures for unresolved processes are a leading source of model uncertainty. Here, we present a new approach to developing machine learning parameterizations of small-scale climate processes by fine-tuning a pre-trained AI foundation model (FM). FMs are largely unexplored in climate research. A pre-trained encoder-decoder from a 2.3 billion parameter FM (NASA and IBM Research's Prithvi WxC) -- which contains a latent probabilistic representation of atmospheric evolution -- is fine-tuned (or reused) to create a deep learning parameterization for atmospheric gravity waves (GWs). The parameterization captures GW effects for a coarse-resolution climate model by learning the fluxes from an atmospheric reanalysis with 10 times finer resolution. A comparison of monthly averages and instantaneous evolution with a machine learning model baseline (an Attention U-Net) reveals superior predictive performance of the FM parameterization throughout the atmosphere, even in regions excluded from pre-training. This performance boost is quantified using the Hellinger distance, which is 0.11 for the baseline and 0.06 for the fine-tuned model. Our findings emphasize the versatility and reusability of FMs, which could be used to accomplish a range of atmosphere- and climate-related applications, leading the way for the creation of observations-driven and physically accurate parameterizations for more earth-system processes.
Problem
Research questions and friction points this paper is trying to address.
Fine-tuning AI foundation models for subgrid-scale climate parameterizations
Developing machine learning parameterizations for atmospheric gravity waves
Reducing model uncertainty in coarse-resolution climate simulations
Innovation
Methods, ideas, or system contributions that make the work stand out.
Fine-tuning pre-trained foundation model for parameterizations
Using encoder-decoder from Prithvi WxC for gravity waves
Learning subgrid fluxes from high-resolution reanalysis data
🔎 Similar Papers
No similar papers found.
