Generative Lagrangian data assimilation for ocean dynamics under extreme sparsity

2025-07-08

AI Summary

Oceanic sub-surface and remote-region observations are extremely sparse (>99% missing), irregularly distributed, and Lagrangian in nature, posing severe challenges for accurate reconstruction of mesoscale turbulence and eddy dynamics using conventional data assimilation and deep learning methods. Method: We propose a coupled Neural Operator–Denoising Diffusion Probabilistic Model (NO-DDPM) framework: a neural operator encodes global physical constraints to guide conditional generative reconstruction by DDPM; multi-source supervision integrates synthetic drifter trajectories with real satellite remote sensing data. Results: Under 99.9% observational sparsity, NO-DDPM accurately recovers eddy structures, eddy shedding processes, and extreme wave features—outperforming state-of-the-art deep learning baselines. It demonstrates strong generalization and physical consistency on both synthetic and real-world datasets.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Systems and Infrastructure for Web, Mobile and WoT: Data management and stream processing for Web, mobile and wireless applications

Abstract

Reconstructing ocean dynamics from observational data is fundamentally limited by the sparse, irregular, and Lagrangian nature of spatial sampling, particularly in subsurface and remote regions. This sparsity poses significant challenges for forecasting key phenomena such as eddy shedding and rogue waves. Traditional data assimilation methods and deep learning models often struggle to recover mesoscale turbulence under such constraints. We leverage a deep learning framework that combines neural operators with denoising diffusion probabilistic models (DDPMs) to reconstruct high-resolution ocean states from extremely sparse Lagrangian observations. By conditioning the generative model on neural operator outputs, the framework accurately captures small-scale, high-wavenumber dynamics even at $99%$ sparsity (for synthetic data) and $99.9%$ sparsity (for real satellite observations). We validate our method on benchmark systems, synthetic float observations, and real satellite data, demonstrating robust performance under severe spatial sampling limitations as compared to other deep learning baselines.

Problem

• Reconstructing ocean dynamics from sparse Lagrangian data
• Overcoming limitations in forecasting eddy shedding and rogue waves
• Improving mesoscale turbulence recovery with deep learning

Innovation

• Combines neural operators with denoising diffusion models
• Reconstructs ocean states from extremely sparse data
• Accurately captures small-scale dynamics at high sparsity