Overview

Climate downscaling estimates local weather and climate patterns from coarse global models or reanalysis, helping assess regional impacts such as extreme heat, heavy precipitation, and infrastructure risk. Approaches range from statistical correction and U-Net-style convolutional networks to continuous-field operators and probabilistic diffusion or flow-matching models, sometimes guided by terrain or physical constraints. Evaluation commonly uses ERA5, regional climate simulations, and station or satellite observations, measuring not only pointwise errors but also extremes, spatial structure, uncertainty, and consistency across variables. A central challenge is that methods that improve average accuracy or fine-scale texture may still misrepresent extremes, dynamics, or future climates, and results can vary across regions and evaluation metrics.

Latest papers

CardsList

32 papers3 papers in the last four weeks, down 57% on the four weeks before. 0.0% of all new papers.

  1. GCD: GCM-consistent Diffusion for Zero-shot Downscaling across Heterogeneous GCMs

    Jan 29, 2026Ruian Tie, Wenbo Xiong, Zhengyu Shi +4Climate DownscalingClimate Modeling

  2. Generate the Forest before the Trees - A Hierarchical Diffusion model for Climate Downscaling

    Jun 24, 2025Declan J. Curran, Sanaa Hobeichi, Hira Saleem +2Climate DownscalingClimate Modeling