![[2602.11807] PuYun-LDM: A Latent Diffusion Model for High-Resolution Ensemble Weather Forecasts](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.11807] PuYun-LDM: A Latent Diffusion Model for High-Resolution Ensemble Weather Forecasts
Summary
The paper presents PuYun-LDM, a novel latent diffusion model designed to enhance high-resolution ensemble weather forecasts, addressing challenges in modeling meteorological data.
Why It Matters
Accurate weather forecasting is crucial for various sectors, including agriculture, disaster management, and urban planning. This research introduces advanced modeling techniques that could significantly improve forecast reliability and efficiency, thereby impacting decision-making processes globally.
Key Takeaways
- PuYun-LDM improves latent diffusability for high-resolution weather forecasts.
- The model utilizes a 3D Masked AutoEncoder to enhance weather-state feature encoding.
- Variable-Aware Masked Frequency Modeling adapts to spectral energy distributions.
- Achieves superior performance at short lead times compared to existing methods.
- Generates 15-day forecasts efficiently on a single NVIDIA H200 GPU.
Computer Science > Artificial Intelligence arXiv:2602.11807 (cs) [Submitted on 12 Feb 2026 (v1), last revised 13 Feb 2026 (this version, v2)] Title:PuYun-LDM: A Latent Diffusion Model for High-Resolution Ensemble Weather Forecasts Authors:Lianjun Wu, Shengchen Zhu, Yuxuan Liu, Liuyu Kai, Xiaoduan Feng, Duomin Wang, Wenshuo Liu, Jingxuan Zhang, Kelvin Li, Bin Wang View a PDF of the paper titled PuYun-LDM: A Latent Diffusion Model for High-Resolution Ensemble Weather Forecasts, by Lianjun Wu and 9 other authors View PDF HTML (experimental) Abstract:Latent diffusion models (LDMs) suffer from limited diffusability in high-resolution (<=0.25°) ensemble weather forecasting, where diffusability characterizes how easily a latent data distribution can be modeled by a diffusion process. Unlike natural image fields, meteorological fields lack task-agnostic foundation models and explicit semantic structures, making VFM-based regularization inapplicable. Moreover, existing frequency-based approaches impose identical spectral regularization across channels under a homogeneity assumption, which leads to uneven regularization strength under the inter-variable spectral heterogeneity in multivariate meteorological data. To address these challenges, we propose a 3D Masked AutoEncoder (3D-MAE) that encodes weather-state evolution features as an additional conditioning for the diffusion model, together with a Variable-Aware Masked Frequency Modeling (VA-MFM) strategy that adaptively selects th...
