Machine Learning for Submesoscale Characterization, Ocean Prediction, and Exploration (ML-SCOPE)
Abstract
Our ocean dynamics emphasis is the exploration, characterization, and prediction of submesoscales, i.e. the scales intermediate between local turbulence and open-ocean mesoscale. We will also extract information about diverse ocean features and dynamics such as complex instability mechanisms, nonlinear interactions, coherent structures, mixing, and background turbulence. Eulerian and Lagrangian modeling methods will be investigated as they represent different modalities for ML. Our long-term goal is to obtain machine-intelligent modeling systems that seamlessly integrate stochastic ocean dynamical models and their multi-fidelity representations with Bayesian and generative learning from data-model misfits, to construct improved ocean models with more accurate parameterizations and discover invariances or differential equations, over a range of spatial and temporal scales. We will learn from heterogeneous measured data sets, multi-resolution simulated fields from three ocean modeling systems, and data-assimilative simulations in several ocean regions and basins, and the global ocean. We will use and vastly extend hierarchical differentialequation-based Bayesian learning, stochastic dynamic reduced-order methods, data-driven closure models, Bayesian Gaussian Processes, adaptive DL schemes, and generative models and adversarial networks. We will obtain submesoscale ML super-parameterizations. We will develop ML for data assimilation and ML-based adaptive sampling to identify the most informative data for model learning. We will refine and incubate methods using idealized and semi-realistic test cases. We will quantify and optimize their robustness, and verify ML results using novel metrics of success. The ML schemes will extract causality, invariances, dynamical system manifolds, and coherent structures. Ultimately, our symbolic interpretation of ML models into emergent dynamical and constitutive relations would further compress knowledge relative to deep networks, thus extending well outside the range of the training data.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Nov 26, 2019
- Source ID
- N000142012023
Entities
People
- Pierre Felix Lermusiaux
Organizations
- Massachusetts Institute of Technology
- Office of Naval Research
- United States Navy