Data-driven, deep-learning modeling frameworks have been recently developed for forecasting time series data. Such machine learning models may be useful in multiple domains including the atmospheric and oceanic ones, and in general, the larger fluids community. The present work investigates the possible effectiveness of such deep neural operator models for reproducing and predicting classic fluid flows and simulations of realistic ocean dynamics. We first briefly evaluate the capabilities of such deep neural operator models when trained on a simulated two-dimensional fluid flow past a cylinder. We then investigate their application to forecasting ocean surface circulation in the Middle Atlantic Bight and Massachusetts Bay, learning from high-resolution data-assimilative simulations employed for real sea experiments. We confirm that trained deep neural operator models are capable of predicting idealized periodic eddy shedding. For realistic ocean surface flows and our preliminary study, they can predict several of the features and show some skill, providing potential for future research and applications.

Better quantitative understanding and accurate data-assimilative predictions of the three-dimensional and time-dependent ocean acidification (OA) processes in coastal regions is urgently needed for the protection and sustainable utilization of ocean resources. In this paper, we extend and showcase the use of our MIT-MSEAS systems for high-resolution coupled physical-biogeochemical-acidification simulations and Bayesian learning of OA models in Massachusetts Bay, starting with simple empirical and equilibrium OA models. Simulations are shown to have reasonable skill when compared to available in situ and remote data. The impacts of wind forcing, internal tides, and solitary waves on water transports and mixing, and OA fields, are explored. Strong wind events are shown to reset circulations and the OA state in the Bay. Internal tides increase vertical mixing of waters in the shallow regions. Solitary waves propagating off Stellwagen Bank coupled with lateral turbulent mixing provide a pathway for exchange of surface and deep waters. Both of these effects are shown to impact biological activity and OA. A mechanism for the creation of multiple subsurface chlorophyll maxima is presented, involving wind-induced upwelling, internal tides, and advection of near surface fields. Due to the measurement sparsity and limited understanding of complex OA processes, the state variables and parameterizations of OA models are very uncertain. We thus present a proof-of-concept study to simultaneously learn and estimate the OA state variables and model parameterizations from sparse observations using our novel dynamics-based Bayesian learning framework for high-dimensional and multi-disciplinary estimation. Results are found to be encouraging for more realistic OA model learning.

We study the performance of sparse regression methods and propose new techniques to distill the governing equations of nonlinear dynamical systems from data. We start from the recently proposed generic methodology of learning interpretable equation forms from data, followed by performance of least absolute shrinkage and selection operator (LASSO) for this purpose. We first develop an algorithm that uses the dual of LASSO optimization for higher accuracy and stability. We then derive a second algorithm that learns the candidate function library in a dynamic data driven applications systems (DDDAS) manner to distill the governing equations of the dynamical system. This is achieved via sequentially thresholded ridge regression (STRidge) over a orthogonal polynomial space. The performance of the methods is illustrated using the Lorenz 63 system and a marine ecosystem model.

In fish modeling, a significant amount of uncertainty exists in the parameter values, parameterizations, functional form of model equations, and even the state variables themselves. This is due to the complexity and lack of understanding of the processes involved, as well as the measurement sparsity. These challenges motivate the present proof-of-concept study to simultaneously learn and estimate the state variables, parameters, and model equations from sparse observations. We employ a novel dynamics-based Bayesian learning framework for high-dimensional, coupled fish-biogeochemical-physical partial-differential equations (PDEs) models, allowing the simultaneous inference of the augmented state variables and parameters. After reviewing the status of ecosystem modeling in the coastal oceans, we first complete a series of PDE-based learning experiments that showcase capabilities for fish-biogeochemical-physical model equations and parameters, using nonhydrostatic Boussinesq flows past a seamount. We then showcase realistic ocean primitive-equation simulations and analyses, using fish catch data  for the Lakshadweep islands in India. These modeling and learning efforts could improve fisheries management from a standpoint of sustainability and efficiency.

Autonomous underwater vehicles (AUVs) are used to execute an increasingly challenging set of missions in commercial, environmental and defense industries. The resources available to the AUV in service of these missions are typically a limited power supply and onboard sensing of its local environment. Optimal path planning is needed to maximize the chances that these AUVs will successfully complete long endurance missions within their power budget. A time-optimal path planner has been recently developed to minimize AUV mission time required to traverse a dynamic ocean environment at a specified speed through the water. For many missions, time minimization is appropriate because the AUVs operate at a fixed propeller speed. However, the ultimate limiting constraint on AUV operations is often the onboard power supply, rather than mission time. While an empirical or theoretical relationship between mission time and power could be applied to estimate power usage in the path planner, the real power usage and availability on an AUV varies mission-to-mission, as a result of multiple factors, including vehicle buoyancy, battery charge cycle, fin configuration, and water type or quality. In this work, we use data collected from two mid-size AUVs operating in various conditions to learn the mission-to-mission variability in the power budget so that it could be incorporated into the mission planner.