Multimodal Spatiotemporal Atmospheric Data Assimilation with Latent Video Flow-matching
Dibyajyoti Chakraborty, Romit Maulik
cs.LG, math-ph, physics.ao-ph, physics.flu-dyn
Submitted: 2026-08-05
Code: https://github.com/ISCLPurdue/era5_videogen
License: http://creativecommons.org/licenses/by/4.0/
The gist: Data assimilation (DA) uses Bayesian inference to update the state of a numerical forecast model with observed data.
Terminology
Abstract
Data assimilation (DA) uses Bayesian inference to update the state of a numerical forecast model with observed data. In this study, we propose a fundamentally different, unified approach to atmospheric data assimilation. We use latent video flow-matching to sample temporally consistent trajectories from a prior trained using ERA5 reanalysis (69 variables over an 8-day window). We also use posterior sampling to assimilate real observation sources, such as those from the NOAA Integrated Global Radiosonde Archive and the Integrated Surface Database. Because the prior generates a continuous trajectory, it naturally propagates information between observed and unobserved frames. Therefore, we can perform various DA tasks, such as filtering and smoothing, simply by changing the observed frames. Moreover, we generate full-state ensemble forecasts directly from sparse observations, achieving performance competitive with state-of-the-art observation-to-forecast models.
Sources
- Score-Based Generative Modeling through Stochastic Differential Equations
- Flow Matching for Generative Modeling
- Diffusion Posterior Sampling for General Noisy Inverse Problems
- GraphDOP: Towards skilful data-driven medium-range weather forecasts learnt and initialised directly from observations
- Physics-Informed CNNs for Super-Resolution of Sparse Observations on Dynamical Systems
- Averaging Weights Leads to Wider Optima and Better Generalization
- Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow
- Zero-Shot Image Restoration Using Denoising Diffusion Null-Space Model
- Generative Diffusion-based Downscaling for Climate
- DiffDA: a Diffusion Model for Weather-scale Data Assimilation
- Generative assimilation and prediction for weather and climate
- PhyDA: Physics-Guided Diffusion Models for Data Assimilation in Atmospheric Systems
- Ensemble score filter with image inpainting for data assimilation in tracking surface quasi-geostrophic dynamics with partial observations
- STeP: A Framework for Solving Scientific Video Inverse Problems with Spatiotemporal Diffusion Priors
- Appa: Bending Weather Dynamics with Latent Diffusion Models for Global Data Assimilation
- Classifier-Free Diffusion Guidance
- Guidance with Spherical Gaussian Constraint for Conditional Diffusion
- Adam: A Method for Stochastic Optimization
- Decoupled Weight Decay Regularization
- HealDA: Highlighting the importance of initial errors in end-to-end AI weather forecasts
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks