Incomplete Observations Boost Evolutionary Performance in Ocean Modeling

2026-07-21

AI Summary

This work addresses the challenge that existing data-driven ocean modeling approaches rely heavily on complete reanalysis datasets and struggle to learn effectively from sparse, noisy observations. To overcome this limitation, the authors propose a generative state-space model in which oceanic physical fields are treated as latent states. Neural networks are employed to model both the state transition dynamics and the initial state distribution, while a masked Gaussian emission distribution is introduced to accommodate incomplete observations. Leveraging the assumption of stationary ergodic Markov processes, they devise an efficient expectation-maximization (EM) algorithm that requires only state sequences of length two, substantially enhancing scalability without compromising theoretical consistency. Experiments on CMIP6 simulations and FY-3D satellite data demonstrate that the method achieves high-fidelity reconstruction and accurate prediction of ocean fields, marking the first successful end-to-end training of an ocean dynamics model directly from sparse observational data.

Abstract

Data-driven methods have revolutionized ocean modeling, yet current approaches rely heavily on complete reanalysis datasets, imposing computational constraints and limiting model performance to that of the training data. Here, we present a generative state-space model and an optimization framework that enable learning directly from sparse and noisy observations. The model is essentially a hidden Markov model with a continuous state space, where oceanic physical quantities are treated as hidden states and measurements as observations, enabling a unified representation of ocean fields and observational data. Both the initial-state and state-transition modules are implemented as neural networks to capture the complexity and temporal evolution of ocean states, while the emission module is formulated as a masked Gaussian distribution. To train the model from sparse observations, we derive an optimization framework based on the expectation-maximization (EM) algorithm. The framework alternately reconstructs high-fidelity ocean fields via Langevin dynamics and optimizes deep neural networks to capture temporal evolution. Theoretical analysis shows that the framework maximizes the likelihood of observations under the generative model. For efficiency, we assume that ocean-state evolution follows a stationary, ergodic, and Markovian stochastic process and adopt only length-two state sequences during optimization. Experiments on CMIP6 simulation data and FY-3D satellite data demonstrate high-fidelity reconstruction and accurate prediction, showing that sparse observations can directly improve the model's representation of ocean-state dynamics. This work offers a scalable pathway for next-generation Earth system models to learn directly from sparse, incomplete real-world observations.

Problem

• incomplete observations
• ocean modeling
• sparse data
• data-driven methods
• state-space model

Innovation

• generative state-space model
• sparse observations
• expectation-maximization
• neural dynamics
• ocean modeling