Ensembles of Neural Surrogates for Parametric Sensitivity in Ocean Modeling

2025-08-22

AI Summary

Low-resolution ocean simulations rely on parameterizations to represent subgrid-scale processes, yet the sensitivity of these parameterizations remains difficult to quantify—hindering model calibration and observational constraint. To address this, we propose an ensemble-learning-based neural surrogate modeling framework that integrates deep neural networks, automatic differentiation, and large-scale hyperparameter optimization. Crucially, it jointly estimates both function outputs and their gradients—even without ground-truth gradient labels—and, for the first time, quantifies epistemic uncertainty in both simultaneously. The method substantially improves accuracy, robustness, and reliability in forward prediction, long-term autoregressive simulation, and adjoint-based sensitivity estimation. It thus provides an interpretable, computationally efficient, and uncertainty-aware analytical tool for optimizing ocean model parameterizations.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Search and Optimization: Learning to Search
• Natural Language Processing: Learning & Optimization for NLP

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Accurate simulations of the oceans are crucial in understanding the Earth system. Despite their efficiency, simulations at lower resolutions must rely on various uncertain parameterizations to account for unresolved processes. However, model sensitivity to parameterizations is difficult to quantify, making it challenging to tune these parameterizations to reproduce observations. Deep learning surrogates have shown promise for efficient computation of the parametric sensitivities in the form of partial derivatives, but their reliability is difficult to evaluate without ground truth derivatives. In this work, we leverage large-scale hyperparameter search and ensemble learning to improve both forward predictions, autoregressive rollout, and backward adjoint sensitivity estimation. Particularly, the ensemble method provides epistemic uncertainty of function value predictions and their derivatives, providing improved reliability of the neural surrogates in decision making.

Problem

• Quantifying model sensitivity to uncertain parameterizations in ocean modeling
• Improving reliability of neural surrogates for parametric sensitivity estimation
• Providing epistemic uncertainty for function predictions and derivatives

Innovation

• Ensemble learning for neural surrogate reliability
• Hyperparameter search enhances forward and adjoint predictions
• Epistemic uncertainty quantification for derivatives and values