Training-Free Data Assimilation with GenCast

2025/09/23 by Thomas Savary, François Rozet, Savary, Thomas +3 · 1 voice · 1 citation
Computer Science · Earth and Planetary Sciences · Physics and Astronomy · #Atmospheric and Oceanic Physics (physics.ao-ph) #Computer Graphics and Visualization Techniques #FOS: Computer and information sciences #FOS: Physical sciences #Machine Learning (cs.LG) #Meteorological Phenomena and Simulations #Model Reduction and Neural Networks #cs.LG #physics.ao-ph

paper · pdf · doi:10.48550/arxiv.2509.18811

openalex publication_date 2025/09/23 · arxiv published 2025/09/23 · arxiv updated 2025/10/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Data assimilation is widely used in many disciplines such as meteorology, oceanography, and robotics to estimate the state of a dynamical system from noisy observations. In this work, we propose a lightweight and general method to perform data assimilation using diffusion models pre-trained for emulating dynamical systems. Our method builds on particle filters, a class of data assimilation algorithms, and does not require any further training. As a guiding example throughout this work, we illustrate our methodology on GenCast, a diffusion-based model that generates global ensemble weather forecasts.

Citations

• Lost in Latent Space: An Empirical Study of Latent Diffusion Models for Physics Emulation
• Align-DA: Align Score-based Atmospheric Data Assimilation with Multiple Preferences
• Diffusion-LAM: Probabilistic Limited Area Weather Forecasting with Diffusion
• Probabilistic weather forecasting with machine learning
• On conditional diffusion models for PDE simulations
• Learning Diffusion Priors from Observations by Expectation Maximization
• Nonlinear ensemble filtering with diffusion models: Application to the surface quasi-geostrophic dynamics
• DiffDA: a Diffusion Model for Weather-scale Data Assimilation
• Learning skillful medium-range global weather forecasting
• DPM-Solver++: Fast Solver for Guided Sampling of Diffusion Probabilistic Models
• Diffusion Posterior Sampling for General Noisy Inverse Problems
• Elucidating the Design Space of Diffusion-Based Generative Models
• High-Resolution Image Synthesis with Latent Diffusion Models
• High-Resolution Image Synthesis with Latent Diffusion Models
• Score-Based Generative Modeling through Stochastic Differential Equations
• Denoising Diffusion Probabilistic Models
• The ERA5 global reanalysis
• Learning to Simulate Complex Physics with Graph Networks
• Data Assimilation in the Geosciences - An overview on methods, issues and perspectives
• Accelerating Eulerian Fluid Simulation With Convolutional Networks
• Deep Unsupervised Learning using Nonequilibrium Thermodynamics
• Tweedie’s Formula and Selection Bias
• Numerical solution of the Navier-Stokes equations

Cited by

• DAISI: Data Assimilation with Inverse Sampling using Stochastic Interpolants

Discussions

• Finally, at the #CCAI workshop, Thomas will show how, without retraining, GenCast can be embedded in a particle filter for data assimilation. That is, no initial state x0 is required anymore, observat [bsky, 4 points, 0 comments]

Related

• Using Diffusion Models to do Data Assimilation
• DiffDA: a Diffusion Model for Weather-scale Data Assimilation
• Generating observation guided ensembles for data assimilation with denoising diffusion probabilistic model
• FlowDAS: A Stochastic Interpolant-based Framework for Data Assimilation
• Flow Matching for Efficient and Scalable Data Assimilation