mlsanityData to 7 Oct 2026

Controlling Ensemble Variance in Diffusion Models: An Application for Reanalyses Downscaling

arXiv 2501.14822 ↗First posted 21 Jan 2025stat.AP cs.AI cs.LGcited by 1

On the map: Diffusion models → Flow matching, time series and science

Abstract (arXiv)

In recent years, diffusion models have emerged as powerful tools for generating ensemble members in meteorology. In this work, we demonstrate how a Denoising Diffusion Implicit Model (DDIM) can effectively control ensemble variance by varying the number of diffusion steps. Introducing a theoretical framework, we relate diffusion steps to the variance expressed by the reverse diffusion process. Focusing on reanalysis downscaling, we propose an ensemble diffusion model for the full ERA5-to-CERRA domain, generating variance-calibrated ensemble members for wind speed at full spatial and temporal resolution. Our method aligns global mean variance with a reference ensemble dataset and ensures spatial variance is distributed in accordance with observed meteorological variability. Additionally, we address the lack of ensemble information in the CARRA dataset, showcasing the utility of our approach for efficient, high-resolution ensemble generation.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Placed in Flow matching, time series and science, one of 14 directions in Diffusion models. Builds on 26 of its 60 references that are in the map. First posted on arXiv 21 Jan 2025.

Builds on

26 of 60 references.

Closest ancestors · 8

Same direction, newest first

  1. 2024Dynamical-generative downscaling of climate model ensembles2 Oct
  2. 2024DiffESM: Conditional Emulation of Temperature and Precipitation in Earth System Models with 3D Diffusion Models17 Sep
  3. 2024Diffusion Model-based Probabilistic Downscaling for 180-year East Asian Climate Reconstruction2 Feb
  4. 2024Wind speed super-resolution and validation: from ERA5 to CERRA via diffusion models27 Jan
  5. 2023Residual Corrective Diffusion Modeling for Km-scale Atmospheric Downscaling24 Sep
  6. 2023Precipitation nowcasting with generative diffusion models13 Aug
  7. 2023Debias Coarsely, Sample Conditionally: Statistical Downscaling through Optimal Transport and Probabilistic Diffusion Models24 May
  8. 2023Latent diffusion models for generative precipitation nowcasting with accurate uncertainty quantification25 Apr

Other references · 18

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  3. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,604
  4. 2021Swin Transformer: Hierarchical Vision Transformer using Shifted Windowscited by 12,082
  5. 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,185
  6. 2021Improved Denoising Diffusion Probabilistic Modelscited by 3,736
  7. 2022Elucidating the Design Space of Diffusion-Based Generative Modelscited by 2,994
  8. 2023Simple diffusion: End-to-end diffusion for high resolution imagescited by 413
  9. 2023On the Importance of Noise Scheduling for Diffusion Modelscited by 193
  10. 2023GenCast: Diffusion-based ensemble forecasting for medium-range weathercited by 169
  11. 2017DeepSD: Generating High Resolution Climate Change Projections through Single Image Super-Resolutioncited by 80
  12. 2020Generative Adversarial Networks (GANs Survey): Challenges, Solutions, and Future Directionscited by 65
6 more
  1. 2022A Generative Deep Learning Approach to Stochastic Downscaling of Precipitation Forecastscited by 64
  2. 2019Stabilizing Generative Adversarial Networks: A Surveycited by 57
  3. 2023ClimateLearn: Benchmarking Machine Learning for Weather and Climate Modelingcited by 46
  4. 2022A Survey on Training Challenges in Generative Adversarial Networks for Biomedical Image Analysiscited by 24
  5. 2022Image Embedding for Denoising Generative Modelscited by 7
  6. 2024Transformer based super-resolution downscaling for regional reanalysis: Full domain vs tiling approachescited by 4

Built on by

Cited by year

Citing papers per year: 2025: 1.

1 paper.

  1. 2025Vision Transformers for Multi-Variable Climate Downscaling: Emulating Regional Climate Models with a Shared Encoder and Multi-Decoder Architecturecited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Downscaling weather forecasts from Low- to High-Resolution with Diffusion Models
  2. 2024A Generative Framework for Probabilistic, Spatiotemporally Coherent Downscaling of Climate Simulation
  3. 2024Continuous Ensemble Weather Forecasting with Diffusion models
  4. 2024TAUDiff: Highly efficient kilometer-scale downscaling using generative diffusion models
  5. 2026Universal Diffusion-Based Probabilistic Downscaling
  6. 2025Km-scale dynamical downscaling through conformalized latent diffusion models
  7. 2023Generating observation guided ensembles for data assimilation with denoising diffusion probabilistic model
  8. 2025Data-Efficient Ensemble Weather Forecasting with Diffusion Models