Mesoscale eddy transports profoundly modulate the circulations and tracer budgets in the global ocean. In predictive ocean models, these transports are typically parameterized via the prescriptions of an eddy buoyancy diffusivity and an isopycnal eddy diffusivity. However, relatively little is known about the magnitude/structure of these eddy diffusivities over continental slopes, which hinders the understanding and prediction of shelf-open ocean exchanges. This thesis serves to fill such theoretical and numerical gaps by documenting our studies of parameterizing mesoscale eddy transports over continental slopes.

We focus first on parameterizing the eddy buoyancy diffusivity in “prograde” currents, a flow regime commonly found along continental margins under downwelling-favorable winds...[ Read more ]

Mesoscale eddy transports profoundly modulate the circulations and tracer budgets in the global ocean. In predictive ocean models, these transports are typically parameterized via the prescriptions of an eddy buoyancy diffusivity and an isopycnal eddy diffusivity. However, relatively little is known about the magnitude/structure of these eddy diffusivities over continental slopes, which hinders the understanding and prediction of shelf-open ocean exchanges. This thesis serves to fill such theoretical and numerical gaps by documenting our studies of parameterizing mesoscale eddy transports over continental slopes.

We focus first on parameterizing the eddy buoyancy diffusivity in “prograde” currents, a flow regime commonly found along continental margins under downwelling-favorable winds or occupied by buoyant boundary currents. The choice of prograde currents complements the study of Wang and Stewart [269], who specifically focused on parameterizing eddy buoyancy diffusivities across “retrograde” currents driven by upwelling-favorable winds. By conducting eddy-resolving simulations of prograde currents, we find that the diagnosed cross-slope eddy buoyancy diffusivity decays from Ο (10⁴ m²/s) in the relatively flat-bottomed region to Ο (10 m²/s) over the steep continental slope. To quantitatively capture this shelf-to-open-ocean transition of eddy buoyancy diffusivity, a couple of physicsbased, “slope-aware” scalings are developed. These scalings are adapted from existing theories for eddy buoyancy transports in the open ocean but incorporate the topographic suppression effects on eddy fluxes in prograde currents, quantified via analytical functions of the slope Burger number.

We then conducted non-eddying simulations of both prograde and retrograde current systems to prognostically evaluate the predictive skills of slope-aware scalings documented herein and in Wang and Stewart [269], respectively. Our modeling results show that the slope-aware scalings of eddy diffusivities more accurately reconstruct the mean flow state over continental slopes than parameterization schemes widely implemented in today’s ocean general circulation models.

Next, we diagnose the isopycnal eddy diffusivity in a suite of eddy-resolving simulations of retrograde currents. The diagnosed cross-slope isopycnal eddy diffusivity is suppressed in the upper open ocean occupied by strong alongshore flows, but enhanced at depths where alongshore flows are weakened. Over continental slopes, isopycnal eddy diffusivity also strengthens at mid-depths, but almost vanishes approaching the sloping seafloor. These observations motivate us to propose a full-depth slope-aware parameterization for the isopycnal eddy diffusivity in retrograde currents. Apart from incorporating the mean-flow suppression effect, this parameterization accounts for the eddy anisotropy effect induced by steep topography, which shapes both the cross-slope and vertical structures of cross-shore isopycnal eddy diffusivity.

Lastly, we employ a purely data-driven approach, which involves the construction of a fully-connected artificial neural network (ANN), to reproducing the diagnosed isopycnal eddy diffusivity in retrograde currents. Offline-mode parameterized tracer simulations are conducted to evaluate the performance of both physics-based and data-driven parameterizations of isopycnal eddy diffusivity. These parameterized simulations show that our proposed slope-aware scaling and the ANN-learnt diffusivity outperform other parameterization schemes in reproducing the tracer solutions of the eddy-resolving model.

This thesis serves as a key step toward parameterizing mesoscale eddy transport across continental slopes in predictive ocean climate models.