Unpaired Downscaling of Fluid Flows with Diffusion Bridges

We present a method to downscale idealized geophysical fluid simulations using generative models based on diffusion maps. By analyzing the Fourier spectra of fields drawn from different data distributions, we show how a diffusion bridge can be used as a transformation between a low resolution and a high resolution dataset, allowing for new sample generation of high-resolution fields given specific low resolution features. The ability to generate new samples allows for the computation of any statistic of interest, without any additional calibration or training. Our unsupervised setup is also designed to downscale fields without access to paired training data; this flexibility allows for the combination of multiple source and target domains without additional training. We demonstrate that the method enhances resolution and corrects context-dependent biases in geophysical fluid simulations, including in extreme events. We anticipate that the same method can be used to downscale the output of climate simulations, including temperature and precipitation fields, without needing to train a new model for each application and providing a significant computational cost savings.

• Publication year

  2023

• Venue

  Artificial Intelligence for the Earth Systems

• Publication date

  2023-05-02

• Fields of study

  Physics, Computer Science, Engineering

• Identifiers
  DOI 10.48550/arXiv.2305.01822arXiv 2305.01822
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Debias Coarsely, Sample Conditionally: Statistical Downscaling through Optimal Transport and Probabilistic Diffusion Models

  2023cited by this paper
• Verifying the Union of Manifolds Hypothesis for Image Data

  2022cited by this paper
• Imagen Video: High Definition Video Generation with Diffusion Models

  2022cited by this paper
• A Generative Deep Learning Approach to Stochastic Downscaling of Precipitation Forecasts

  2022cited by this paper
• Perception Prioritized Training of Diffusion Models

  2022cited by this paper
• Increasing the accuracy and resolution of precipitation forecasts using deep generative models

  2022cited by this paper
• Dual Diffusion Implicit Bridges for Image-to-Image Translation

  2022influential reference
• Understanding Diffusion Models: A Unified Perspective

  2022cited by this paper
• Contrastive Learning for Climate Model Bias Correction and Super-Resolution

  2022cited by this paper
• Generative Modelling With Inverse Heat Dissipation

  2022cited by this paper
• Solving Inverse Problems in Medical Imaging with Score-Based Generative Models

  2021cited by this paper
• Image Super-Resolution via Iterative Refinement

  2021cited by this paper
• Generative Adversarial Networks

  2021cited by this paper
• Diffusion Models Beat GANs on Image Synthesis

  2021influential reference
• GeophysicalFlows.jl: Solvers for geophysical fluid dynamics problems in periodic domains on CPUs & GPUs

  2021cited by this paper
• E. Schrödinger’s 1931 paper “On the Reversal of the Laws of Nature” [“Über die Umkehrung der Naturgesetze”, Sitzungsberichte der preussischen Akademie der Wissenschaften, physikalisch-mathematische Klasse, 8 N9 144–153]

  2021cited by this paper
• Maximum Likelihood Training of Score-Based Diffusion Models

  2021cited by this paper
• CSDI: Conditional Score-based Diffusion Models for Probabilistic Time Series Imputation

  2021cited by this paper
• Conditional Image Generation with Score-Based Diffusion Models

  2021cited by this paper
• Learning to Correct Climate Projection Biases

  2021cited by this paper
• ClimAlign: Unsupervised statistical downscaling of climate variables via normalizing flows

  2020cited by this paper
• Improved Techniques for Training Score-Based Generative Models

  2020cited by this paper
• Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains

  2020cited by this paper
• Denoising Diffusion Probabilistic Models

  2020cited by this paper
• Contrastive Learning for Unpaired Image-to-Image Translation

  2020cited by this paper
• DiffWave: A Versatile Diffusion Model for Audio Synthesis

  2020cited by this paper
• Normalizing Flows for Probabilistic Modeling and Inference

  2019cited by this paper
• Deep Learning for Image Super-Resolution: A Survey

  2019cited by this paper
• Generative Modeling by Estimating Gradients of the Data Distribution

  2019cited by this paper
• An Introduction to Variational Autoencoders

  2019cited by this paper
• Group Normalization

  2018cited by this paper
• Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks

  2018influential reference
• accepted for publication

  2018cited by this paper
• Searching for Activation Functions

  2018cited by this paper
• StarGAN: Unified Generative Adversarial Networks for Multi-domain Image-to-Image Translation

  2017cited by this paper
• Non-local Neural Networks

  2017cited by this paper
• Deep Residual Learning for Image Recognition

  2015cited by this paper
• U-Net: Convolutional Networks for Biomedical Image Segmentation

  2015influential reference
• Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015

  2015influential reference
• Deep Unsupervised Learning using Nonequilibrium Thermodynamics

  2015cited by this paper
• An intercomparison of statistical downscaling methods used for water resource assessments in the United States

  2014cited by this paper
• Statistical Downscaling Using Localized Constructed Analogs (LOCA)

  2014cited by this paper
• Assessment of Alternative Methods for Statistically Downscaling Daily GCM Precipitation Outputs to Simulate Regional Streamflow

  2014cited by this paper
• Probabilistic estimates of future changes in California temperature and precipitation using statistical and dynamical downscaling

  2013cited by this paper
• Bias correction can modify climate model simulated precipitation changes without adverse effect on the ensemble mean

  2013cited by this paper
• Technical Note: Bias correcting climate model simulated daily temperature extremes with quantile mapping

  2012cited by this paper
• A comparison of statistical downscaling methods suited for wildfire applications

  2012cited by this paper
• Impact of a Statistical Bias Correction on the Projected Hydrological Changes Obtained from Three GCMs and Two Hydrology Models

  2011cited by this paper
• A Connection Between Score Matching and Denoising Autoencoders

  2011cited by this paper
• Linking climate change modelling to impacts studies: recent advances in downscaling techniques for hydrological modelling

  2007cited by this paper
• Utility of daily vs. monthly large-scale climate data: an intercomparison of two statistical downscaling methods

  2007cited by this paper
• Review of scenario selection and downscaling methods for the assessment of climate change impacts on hydrology in the United States pacific northwest

  2007cited by this paper
• Stochastic Models for the Kinematics of Moisture Transport and Condensation in Homogeneous Turbulent Flows

  2006influential reference
• An overview of power spectral density (PSD) calculations

  2005cited by this paper
• Estimation of Non-Normalized Statistical Models by Score Matching

  2005cited by this paper
• Hydrologic Implications of Dynamical and Statistical Approaches to Downscaling Climate Model Outputs

  2004cited by this paper
• Long‐range experimental hydrologic forecasting for the eastern United States

  2002cited by this paper
• Reverse-time diffusion equation models

  1982influential reference

• SerpentFlow: Generative Unpaired Domain Alignment via Shared-Structure Decomposition

  2026influential citation
• MESMER-RCM: a probabilistic climate emulator for regional warming projections

  2026cites this paper
• Universal Diffusion-Based Probabilistic Downscaling

  2026cites this paper
• Satellite Observations Guided Diffusion Model for Accurate Meteorological States at Arbitrary Resolution

  2025cites this paper
• Predicting Forced Responses of Probability Distributions via the Fluctuation-Dissipation Theorem and Generative Modeling

  2025cites this paper
• Surface to Seafloor: A Generative AI Framework for Decoding the Ocean Interior State

  2025cites this paper
• Physics-constrained generative machine learning-based high-resolution downscaling of Greenland's surface mass balance and surface temperature

  2025cites this paper
• GEN2: A Generative Prediction-Correction Framework for Long-time Emulations of Spatially-Resolved Climate Extremes

  2025influential citation
• Statistical Parameter Calibration with the Generalized Fluctuation Dissipation Theorem and Generative Modeling

  2025cites this paper
• Reframing Generative Models for Physical Systems using Stochastic Interpolants

  2025cites this paper
• AODDiff: Probabilistic Reconstruction of Aerosol Optical Depth via Diffusion-based Bayesian Inference

  2025cites this paper
• DiffDisReg: Mitigating Forgetting in Diffusion Models with Discriminative Regularization

  2025cites this paper
• A Generative Framework for Probabilistic, Spatiotemporally Coherent Downscaling of Climate Simulation

  2024cites this paper
• Fast, Scale-Adaptive, and Uncertainty-Aware Downscaling of Earth System Model Fields with Generative Foundation Models

  2024influential citation
• Generative Diffusion-based Downscaling for Climate

  2024cites this paper
• A Generative Machine Learning Approach for Improving Precipitation from Earth System Models

  2024cites this paper
• A Probabilistic Framework for Learning Non‐Intrusive Corrections to Long‐Time Climate Simulations From Short‐Time Training Data

  2024cites this paper
• Machine learning for the physics of climate

  2024cites this paper
• Fast, scale-adaptive and uncertainty-aware downscaling of Earth system model fields with generative machine learning

  2024influential citation
• Response Theory via Generative Score Modeling

  2024cites this paper
• Diffusion Schrödinger Bridge Matching

  2023influential citation
• Easing Color Shifts in Score-Based Diffusion Models

  2023cites this paper
• When Geoscience Meets Foundation Models: Toward a general geoscience artificial intelligence system

  2023cites this paper
• Debias Coarsely, Sample Conditionally: Statistical Downscaling through Optimal Transport and Probabilistic Diffusion Models

  2023influential citation
• Residual corrective diffusion modeling for km-scale atmospheric downscaling

  2023cites this paper
• Surface to Seafloor: A Generative Machine Learning Framework for Decoding the Ocean Interior State

  year unknowncites this paper