Skip to main content
QUICK REVIEW

[论文解读] Field-Space Autoencoder for Scalable Climate Emulators

Johannes Meuer, Maximilian Witte|arXiv (Cornell University)|Jan 21, 2026
Generative Adversarial Networks and Image Synthesis被引用 0
一句话总结

介绍在 HEALPix 球面网格上运行的 Field-Space Autoencoder(FS-AE),具备 Field-Space Attention,用于高效的气候数据压缩、零-shot 超分辨率,以及在压缩空间中的扩散式生成仿真。

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.

研究动机与目标

  • 通过开发几何一致的压缩方法,解决千米尺度气候数据的存储与 I/O 挑战。
  • 直接在球面 HEALPix 网格上工作以保留物理结构和拓扑保真度。
  • 启用零-shot 超分辨率,以在不重新训练的情况下跨尺度映射。
  • 通过在压缩场中实现基于扩散的生成仿真器,连接低分辨率数据与高分辨率数据。

提出的方法

  • 使用 Field-Space Attention 与 Field-Space Compression/Decompression 在 HEALPix 球面网格上原生操作。
  • 实现多尺度残差分解以实现分辨率不变的压缩。
  • 训练基于 Transformer 的 Field-Space Autoencoder,瓶颈提供 4^N 的压缩比。
  • 将 FS-AE 与 CNN-VAE 基线在 ERA5 表面气温上按多种压缩比进行对比。
  • 在压缩空间开发 Compressed Field Diffusion,以从低分辨率集合中学习气候变率,并从高分辨率数据中学习局部物理过程。
Figure 1: Field-Space Autoencoder: overview and multi-scale processing. (a) The Encoder $E$ compresses multi-scale inputs on the HEALPix sphere at a target coarse HEALPix level. The Decoder $D$ reconstructs fields at the original input HEALPix level. We denote native inputs by $x^{(n)}$ , residuals
Figure 1: Field-Space Autoencoder: overview and multi-scale processing. (a) The Encoder $E$ compresses multi-scale inputs on the HEALPix sphere at a target coarse HEALPix level. The Decoder $D$ reconstructs fields at the original input HEALPix level. We denote native inputs by $x^{(n)}$ , residuals

实验结果

研究问题

  • RQ1 Field-Space Autoencoder 在高压缩比下能否比卷积基线更忠实地压缩球面气候场?
  • RQ2多尺度、分辨率不变的设计是否能够实现跨不同网格分辨率的零-shot 超分辨率?
  • RQ3在压缩场空间中的扩散模型是否能够合成时间上连贯、物理可行的集合,结合低分辨率信息和高分辨率信息?

主要发现

  • FS-AE 在所有测试的压缩比中均优于 CNN-VAE 基线的重建保真度。
  • 在 64× 压缩下,FS-AE 以相似或更好的 PSNR 实现比 CNN-VAE 高 4× 的压缩效率。
  • 压缩空间在 t-SNE 中呈现具有物理含义的结构,反映季节和长期变暖趋势,并将火山事件标记为离群点。
  • 该模型对多变量具有泛化能力,并通过跨变量注意力在梯度较高区域(如地形)维持保真度。
  • 零-shot 超分辨率在推理时通过屏蔽细尺度剩余量实现对输入的局部放大,无需微调。
  • Compressed Field Diffusion 产生的集合在光谱清晰度方面更接近 ERA5/MPI-ESM-HR,而非低分辨率模型,同时保持内部变率模式。
Figure 2: Reconstruction performance and compressed space visualization. (a) Dual-axis plot showing root-mean-square error (RMSE; left vertical axis, °C) and peak signal-to-noise ratio (PSNR; right vertical axis, dB) as functions of compression ratio for both the Field-Space Autoencoder (FS-AE) and
Figure 2: Reconstruction performance and compressed space visualization. (a) Dual-axis plot showing root-mean-square error (RMSE; left vertical axis, °C) and peak signal-to-noise ratio (PSNR; right vertical axis, dB) as functions of compression ratio for both the Field-Space Autoencoder (FS-AE) and

更好的研究,从现在开始

从论文设计到论文写作,大幅缩短您的研究时间。

无需绑定信用卡

本解读由 AI 生成,并经人工编辑审核。