Downscaling climate projections to 1 km with single-image super resolution

2025/09/24 by Košťál, Petr, Kordík, Pavel, Podsztavek, Ondřej
#Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Machine Learning (cs.LG)

paper · doi:10.48550/arxiv.2509.21399

Abstract

High-resolution climate projections are essential for local decision-making. However, available climate projections have low spatial resolution (e.g. 12.5 km), which limits their usability. We address this limitation by leveraging single-image super-resolution models to statistically downscale climate projections to 1-km resolution. Since high-resolution climate projections are unavailable, we train models on a high-resolution observational gridded data set and apply them to low-resolution climate projections. We cannot evaluate downscaled climate projections with common metrics (e.g. pixel-wise root-mean-square error) because we lack ground-truth high-resolution climate projections. Therefore, we evaluate climate indicators computed at weather station locations. Experiments on daily mean temperature demonstrate that single-image super-resolution models can downscale climate projections without increasing the error of climate indicators compared to low-resolution climate projections.

Citations

• Transformer based super-resolution downscaling for regional reanalysis: Full domain vs tiling approaches
• Evaluating the transferability potential of deep learning models for climate downscaling
• Kaushikreddym/CasCorrDiff: CasCorrDiff v1.0.0
• WeatherBench 2: A benchmark for the next generation of data-driven global weather models
• Multi-variable Hard Physical Constraints for Climate Model Downscaling
• SuperBench: A Super-Resolution Benchmark Dataset for Scientific Machine Learning
• Fourier Neural Operators for Arbitrary Resolution Climate Data Downscaling
• Deep generative model super-resolves spatially correlated multiregional climate data
• Increasing the accuracy and resolution of precipitation forecasts using deep generative models
• Fourier Neural Operator for Parametric Partial Differential Equations
• Enhanced Deep Residual Networks for Single Image Super-Resolution
• DeepSD: Generating High Resolution Climate Change Projections through Single Image Super-Resolution
• Image Super-Resolution Using Deep Convolutional Networks
• Image Super-Resolution Using Deep Convolutional Networks
• Image quality assessment: from error visibility to structural similarity
• Swin Transformer: Hierarchical Vision Transformer using Shifted Windows
• Swin Transformer: Hierarchical Vision Transformer using Shifted Windows

Related

• DeepSD: Generating High Resolution Climate Change Projections through Single Image Super-Resolution
• Resolution-Agnostic Transformer-based Climate Downscaling
• Hard-Constrained Deep Learning for Climate Downscaling
• Emerging AI-based weather prediction models as downscaling tools
• Identifying high resolution benchmark data needs and Novel data-driven methodologies for Climate Downscaling