Field-Space Autoencoder for Scalable Climate Emulators

TL;DR

This paper introduces the Field-Space Autoencoder, a scalable climate emulation framework that efficiently compresses and generates high-resolution climate data, preserving physical structures and enabling super-resolution from low- and high-resolution datasets.

Contribution

The paper presents a novel spherical compression autoencoder with Field-Space Attention, improving physical structure preservation and enabling zero-shot super-resolution in climate emulation.

Findings

Preserves physical structures better than convolutional baselines.

Enables zero-shot super-resolution from low- to high-resolution data.

Learns internal variability and fine-scale physics simultaneously.

Abstract

Kilometer-scale Earth system models are essential for capturing local climate change. However, these models are computationally expensive and produce petabyte-scale outputs, which limits their utility for applications such as probabilistic risk assessment. Here, we present the Field-Space Autoencoder, a scalable climate emulation framework based on a spherical compression model that overcomes these challenges. By utilizing Field-Space Attention, the model efficiently operates on native climate model output and therefore avoids geometric distortions caused by forcing spherical data onto Euclidean grids. This approach preserves physical structures significantly better than convolutional baselines. By producing a structured compressed field, it serves as a good baseline for downstream generative emulation. In addition, the model can perform zero-shot super-resolution that maps low-resolution large ensembles and scarce high-resolution data into a shared representation. We train a generative diffusion model on these compressed fields. The model can simultaneously learn internal variability from abundant low-resolution data and fine-scale physics from sparse high-resolution data. Our work bridges the gap between the high volume of low-resolution ensemble statistics and the scarcity of high-resolution physical detail.