![[2602.13416] High-Resolution Climate Projections Using Diffusion-Based Downscaling of a Lightweight Climate Emulator](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.13416] High-Resolution Climate Projections Using Diffusion-Based Downscaling of a Lightweight Climate Emulator
Summary
This article presents a novel approach to high-resolution climate projections using a diffusion-based downscaling framework applied to a lightweight climate emulator, LUCIE, achieving 25 km resolution from coarse outputs.
Why It Matters
Accurate climate projections are essential for understanding regional impacts of climate change. This research addresses limitations in existing models by providing a method that enhances resolution while preserving essential climatic dynamics, which can inform policy and adaptation strategies.
Key Takeaways
- LUCIE, a lightweight climate emulator, provides accurate long-term climate statistics.
- The new downscaling framework utilizes diffusion-based generative models for improved resolution.
- The model achieves a downscaling from ~300 km to ~25 km, enhancing regional climate assessments.
- Performance is validated through various metrics, ensuring reliability of the projections.
- This approach can significantly aid in climate impact studies and decision-making.
Computer Science > Machine Learning arXiv:2602.13416 (cs) [Submitted on 13 Feb 2026] Title:High-Resolution Climate Projections Using Diffusion-Based Downscaling of a Lightweight Climate Emulator Authors:Haiwen Guan, Moein Darman, Dibyajyoti Chakraborty, Troy Arcomano, Ashesh Chattopadhyay, Romit Maulik View a PDF of the paper titled High-Resolution Climate Projections Using Diffusion-Based Downscaling of a Lightweight Climate Emulator, by Haiwen Guan and 5 other authors View PDF HTML (experimental) Abstract:The proliferation of data-driven models in weather and climate sciences has marked a significant paradigm shift, with advanced models demonstrating exceptional skill in medium-range forecasting. However, these models are often limited by long-term instabilities, climatological drift, and substantial computational costs during training and inference, restricting their broader application for climate studies. Addressing these limitations, Guan et al. (2024) introduced LUCIE, a lightweight, physically consistent climate emulator utilizing a Spherical Fourier Neural Operator (SFNO) architecture. This model is able to reproduce accurate long-term statistics including climatological mean and seasonal variability. However, LUCIE's native resolution (~300 km) is inadequate for detailed regional impact assessments. To overcome this limitation, we introduce a deep learning-based downscaling framework, leveraging probabilistic diffusion-based generative models with conditional and...
