betaFeedback

WorksPhysics and Astronomy

Tidally dominated flows past a three-dimensional topography: Wake vortices, turbulence, and mixing

Abstract

Oceanic turbulence influences the transport and mixing of freshwater, heat, nutrients, and other biogeochemical tracers. It also has broader implications for oceanic and atmospheric circulations. Tides contribute substantially to the mechanically driven turbulent ocean mixing through the internal waves resulting from tide-topography interactions. Tidal currents also drive turbulent wakes and shear layers when the topography is 3D. The hypothesis that seamounts are the ``stirring rods'' of the ocean has motivated considerable recent interest in turbulent flow features near 3D topography. It also motivates the present LES of tidally dominated flows (tidal oscillations superposed on a weaker mean) past an idealized steep seamount. Complex interactions occur between the topography, the near wake, and previously shed vortices, especially during the tidal phases when the flow direction is reversed. The topographic wake is shown to be a hotspot for mixing, featuring large dissipation rates in the attached shear layers, hydraulic jet, recirculation region in the near wake, and peripheries of shed vortices. The majority of the observed dissipation is due to the vertical shear. Over a tidal cycle, the volume-integrated local dissipation within the wake is at least four times greater than the internal wave flux that may be dissipated elsewhere. Furthermore, normalized dissipation rates are maximized for the purely tidal setting. Within the tidal cycle, bulk mixing efficiency (ηη) varies substantially and is maximized at η≈0.25η\approx 0.25 around flow reversals.

Generated using the official AMS LATEX template v6.1 two-column layout. This work has been submitted for publication. Copyright in this work may be transferred without further notice, and this version may no longer be accessible.

SIGNIFICANCE STATEMENT: Tidal forcing, primarily due to the gravitational attraction from the moon and the sun, constitutes a major source of mechanical energy input into the world ocean. Delineating the energy pathways that ultimately lead to dissipation is crucial to understanding the ocean circulation and mixing, which in turn have broader implications. A significant portion of energy dissipation is attributed to breaking of internal waves generated due to flow-topography interactions. We show that steep seamounts, hypothesized to be stirring rods of the ocean, not only generate internal waves, but also act as local hotspots of energy dissipation. The local dissipation in tidally dominated flows is at least four times greater than the energy transferred to the internal waves, to potentially be dissipated elsewhere.

observations reveal that strong turbulent mixing is concentrated in localized hotspots ([45]), with typical background values in the deep ocean closer to ϵK∼10−10\epsilon_K \sim10^{-10} W kg−1^{-1} ([23]).

Turbulent dissipation in the ocean interior is thought to be largely driven by the breaking of internal waves, a portion of which originates from the interaction of tidal and subinertial flows with bottom topography (e.g., [26, 25, 19]). Internal-wave-related dissipation rates of up to ϵK∼10−7\epsilon_K \sim10^{-7} W kg−1^{-1} have been observed across the global ocean (see [45], for a compilation of observations from moorings, floats, and ADCPs). The mechanisms for the cascade of wave energy leading up to turbulence have been extensively investigated ([23, 37]) and parameterized ([24]).

In addition to wave generation, flow past isolated topographic features can result in localized dissipation near the topography. For example, flow separation at three-dimensional topography can generate attached shear layers and hydraulic jets, which are sites of enhanced turbulence ([35]). The laboratory experiments of Hunt and Snyder [15] and Castro et al. [5] revealed that wake eddies are shed in the lee of dynamically tall obstacles when the stratification does not have sufficient kinetic energy to overcome the potential energy barrier imposed by the stratification. Vertically coupled wake vortices in the lee of three-dimensional topography exposed to a steady current were extensively investigated through numerical simulations, using hydrostatic models ((Dong et al.

Introduction

Turbulence plays a key role in the transport of freshwater, pollutants, nutrients, and other biogeochemical tracers in the ocean ([43]). Strong velocity gradients in turbulent motions result in enhanced dissipation rates, which is crucial to the ocean’s energy budget. Specifically, the turbulent dissipation of available potential energy results in irreversible mixing, which plays a central role in maintaining the ocean’s stratification ([50, 51]). Munk and Wunsch [25] estimated that a total of 2.1 TW of dissipation is required to maintain the global abyssal stratification. While this corresponds to a global average dissipation rate of ϵK∼10−9\epsilon_K \sim10^{-9} W kg−1^{-1}, 2007; Perfect et al. [30, 31]; Srinivasan et al. [39]; Jagannathan et al. [16]) and non-hydrostatic large eddy simulations (Puthan et al. [34]; Liu et al. [21]), and were shown to be regions of enhanced dissipation.

Wake eddies in the lee of islands and submarine ridges were revealed by a number of recent observational studies, and dissipation rates as high as ϵK∼10−5 W kg−1\epsilon_K \sim10^{-5}\ \mathrm{W\ kg^{-1}} were observed in the wake (Chang et al. [8, 7]; Zeiden et al. [54]; MacKinnon et al. 2019; Johnston et al. [17]; St. Laurent et al. [40]; Voet et al. [44]; Wynne-Cattanach et al. [52]). The wake dynamics in these oceanographic settings depend strongly on the flow impinging on the topography. Specifically, tidal oscillations can result in significant changes to the wake behavior, compared to a steady inflow. For example, the large eddy simulations (LESs) of Puthan et al. [36] revealed tidal synchronization of the wake eddies. Puthan et al. [35] observed dissipation rates of up to ϵK∼10−5 W kg−1\epsilon_K \sim10^{-5}\ \mathrm{W\ kg^{-1}} in the turbulent near wake, and noted that synchronization in tidally modulated flows occurred only if the amplitude of the tidal current is at least 20% of the mean flow.

The tidal component of the flow interacting with topography can be comparable to or even larger in velocity than the mean current. Such cases of flow-topography interaction occur in the deep ocean and have been observationally studied (Dale and Inall [9]; Voet et al. [44]) from the perspective of nonlinear internal waves and turbulence on the slopes. The maximum flow velocity was not large (0.7 m s−1\mathrm{m\ s^{-1}} in Dale and Inall [9] and 0.2 m s−1\mathrm{m\ s^{-1}} in Voet et al. [44]), but the topographies had steep slopes and modest widths ∼1 km\sim1\ \mathrm{km}, so that the tidal excursion number was ∼1\sim1. In contrast, coastal regions subject to strong barotropic tides can exhibit peak flows exceeding 1 m s−1\mathrm{m\ s^{-1}} (Lu et al. 2000). Motivated by the flow regime of these examples, we perform an LES study of tidally dominated flows where the relative magnitude of the mean is systematically varied from 0 (pure tide) to 1 and the tidal excursion number exceeds unity. We will show that the turbulent wake dynamics in this regime of tidal domination are substantially different and constitute a different regime from previous LES work with tides. Although the focus is on wake eddies and turbulence, the internal wave flux (a surrogate for remote dissipation) will be compared to the turbulent dissipation in the wake.

The computational model used for the LES is discussed in § 2. The key non-dimensional parameters and other quantities of interest are discussed in § 2. The values of the physical parameters, and corresponding non-dimensional groups are presented in § 2. The vortex dynamics in the wake within a single tidal cycle is discussed using three representative cases of tidally dominated flows in § 3, wherein the salient features in the streamwise velocity and the vertical vorticity are highlighted. The link between observed flow features and the dissipation rate of kinetic energy is then discussed in § 4, and the role of vertical shear in the system is delineated. Tidal-phase-averaged dissipation rate is then investigated in § 5, to obtain statistics of the volume-integrated dissipation rate in the wake (“wake dissipation”), which is compared against the area-integrated internal wave flux (“wave dissipation”). The dissipation of available potential energy is also investigated, and the mixing efficiency within the wake is quantified as a function of the tidal phase. Finally, § 6 presents a discussion of the key findings and their implications.

Theory and methods

Model Setup

The computational model (see Figure 1 for a schematic) simulates the interaction of an unsteady barotropic current with an idealized three-dimensional (3D) topography, in a linearly stratified setting with background rotation. The unsteady free-stream U∞(t)=Uc+Utsin⁡(2πftt)U_\infty(t) = U_c + U_t \sin(2\pi f_t t) is modeled as a superposition of a mean current UcU_c and an oscillating tidal component Utsin⁡(2πftt)U_t \sin(2\pi f_t t), and the conical seamount has height hh and base diameter DD. The buoyancy frequency NN is a constant, and the Coriolis parameter is 2πfΩ2\pi f_\Omega.

Schematic of the problem setup

Figure 1. A schematic of the problem setup. A description of the various symbols is in § 2. The gray-shaded regions denote the three sponge layers near the western (x/D≈−16x/D \approx-16), eastern (x/D≈16x/D \approx16), and top (z/h≈10z/h \approx10) boundaries. The light gray lines correspond to cell edges in the computational grid — a line is plotted every 128, 128, and 20 edges in the xx-, yy-, and zz-directions, respectively.

The numerical model solves the three-dimensional Navier-Stokes equations filtered for large eddy simulations, under the Boussinesq approximation in a rotating framework, expressed using the index notation as:

∂ui∂xi=0,(1)\frac{\partial u_i}{\partial x_i}=0, \tag*{(1)}
∂ui∂t+∂(uiuj)∂xj−ϵij3(2πfΩ)(uj−U∞δj1)=−1ρ0∂p∗∂xi−gρ∗ρ0δi3+F∞(t)δi1+∂τij∂xj,(2)\frac{\partial u_i}{\partial t}+\frac{\partial(u_i u_j)}{\partial x_j}-\epsilon_{ij3}(2\pi f_\Omega)(u_j-U_\infty\delta_{j1}) =-\frac{1}{\rho_0}\frac{\partial p^*}{\partial x_i}-\frac{g\rho^*}{\rho_0}\delta_{i3}+F_\infty(t)\delta_{i1}+\frac{\partial\tau_{ij}}{\partial x_j}, \tag*{(2)}
∂ρ∂t+∂(ujρ)∂xj=∂Λj∂xj.(3)\frac{\partial\rho}{\partial t}+\frac{\partial(u_j\rho)}{\partial x_j}=\frac{\partial\Lambda_j}{\partial x_j}. \tag*{(3)}

Here, uiu_i represents the filtered velocity component in the ithi^{\mathrm{th}} direction, with i=1,2,3i=1,2,3 corresponding to the streamwise/zonal (xx), spanwise/meridional (yy), and vertical (zz) directions, respectively. The forcing term

F∞(t)=dU∞dt=2πftUtcos⁡(2πftt)(4)F_\infty(t)=\frac{\mathrm{d}U_\infty}{\mathrm{d}t}=2\pi f_t U_t\cos(2\pi f_t t) \tag*{(4)}

sustains the barotropic tidal oscillations in the system. The total density ρ\rho is modeled as a sum of a reference density ρ0\rho_0, a vertically varying ρr\rho_r, and a deviation ρ∗\rho^*:

ρ=ρ0+ρr(z)+ρ∗(x,y,z,t)=ρb(z)+ρ∗(x,y,z,t).(5)\rho=\rho_0+\rho_r(z)+\rho^*(x,y,z,t)=\rho_b(z)+\rho^*(x,y,z,t). \tag*{(5)}

The total pressure pp is a sum of the ambient pressure pap_a, a hydrostatic component php_h that balances the contribution from ρb\rho_b, a meridionally varying pgp_g that is in geostrophic balance with the barotropic U∞(t)U_{\infty}(t), and a deviation p∗p^{*}:

p=pa+pb(z)+pg(y,t)+p∗(x,y,z,t).(6)p = p_a + p_b(z) + p_g(y,t) + p^{*}(x,y,z,t). \tag*{(6)}

The stress tensor τij=(ν+νSGS)(∂ui/∂xj)\tau_{ij} = (\nu+\nu_{\mathrm{SGS}})(\partial u_i/\partial x_j) incorporates both the molecular viscous stress and the subgrid-scale (SGS) stress, calculated using the Wall-Adapting Local Eddy-viscosity (WALE) model [27]:

νSGS=(CwΔ)2(SijdSijd)3/2(SijSij)5/2+(SijdSijd)5/4,(7)\nu_{\mathrm{SGS}} = (C_{\mathrm{w}}\Delta)^2 \frac{(\mathcal{S}^{d}_{ij}\mathcal{S}^{d}_{ij})^{3/2}}{(S_{ij}S_{ij})^{5/2}+(\mathcal{S}^{d}_{ij}\mathcal{S}^{d}_{ij})^{5/4}}, \tag*{(7)}

where

Sijd=SikSkj+ΩikΩkj−13δij(SmnSmn−ΩmnΩmn),(8)\mathcal{S}^{d}_{ij} = S_{ik}S_{kj}+\Omega_{ik}\Omega_{kj}-\frac{1}{3}\delta_{ij}(S_{mn}S_{mn}-\Omega_{mn}\Omega_{mn}), \tag*{(8)}
Sij=12(∂ui∂xj+∂uj∂xi),Ωij=12(∂ui∂xj−∂uj∂xi),(9)S_{ij} = \frac{1}{2}\left(\frac{\partial u_i}{\partial x_j}+\frac{\partial u_j}{\partial x_i}\right), \qquad\Omega_{ij} = \frac{1}{2}\left(\frac{\partial u_i}{\partial x_j}-\frac{\partial u_j}{\partial x_i}\right), \tag*{(9)}

Cw=0.5C_{\mathrm{w}}=0.5 is a constant, and Δ=(ΔxΔyΔz)1/3\Delta=(\Delta x\Delta y\Delta z)^{1/3}, where Δx\Delta x, Δy\Delta y, Δz\Delta z are local grid sizes in the xx-, yy- and zz- directions, respectively. The model accounts for the role of small-scale eddies in turbulent dissipation and ensures νSGS\nu_{\mathrm{SGS}} asymptotes to zero at a wall, by construction. Similarly, the scalar flux term Λj\Lambda_j is given by Λj=(κ+κSGS)(∂ρ/∂xj)\Lambda_j=(\kappa+\kappa_{\mathrm{SGS}})(\partial\rho/\partial x_j), with κSGS=νSGS\kappa_{\mathrm{SGS}}=\nu_{\mathrm{SGS}}.

The unsteady, three-dimensional governing equations are solved numerically on a staggered grid that spans x/D∈[−16,16]x/D\in[-16,16] in the streamwise direction, y/D∈[−6,6]y/D\in[-6,6] in the spanwise direction, and z/h∈[0,10.22]z/h\in[0,10.22] in the vertical.

The domain is discretized on a 2562×2562×4022562\times2562\times402 grid with a total of 2.7 billion points. A uniform grid is used near the topography, within x/D∈[−3,3]x/D\in[-3,3], y/D∈[−3,3]y/D\in[-3,3] and z/h∈[0,1.32]z/h\in[0,1.32], and a mild grid stretching (by factors of 1.0038 and 1.023, in the xx- and zz-directions, respectively) away from the topography. The topography is modeled using the immersed boundary method of Yang and Balaras [53], with no-slip and adiabatic boundary conditions on its surface. Sponge layers [41] that damp the flow to the free-stream conditions are placed at the top and the streamwise boundaries, and the domain is periodic in the spanwise direction. Spatial derivatives are discretized using a second-order central difference, and the time integration is performed using a semi-implicit, low-storage Runge-Kutta (RK-3)/Crank-Nicholson scheme [11, 3].

Key parameters and definitions

Important physical parameters that govern the dynamics of the selected problem include topographic geometry, height hh and diameter DD of the conical seamount; the current, its mean velocity UcU_c, tidal velocity amplitude UtU_t and the tidal frequency ftf_t; the background, the Coriolis frequency fΩf_{\Omega} and the buoyancy frequency NN. Associated non-dimensional parameters and their physical significance are outlined below. The aspect ratio of the conical seamount, a=h/Da=h/D, determines the slope angle β=tan⁡−1(h/(D/2))\beta=\tan^{-1}(h/(D/2)), which plays a key role in the internal wave dynamics in the system. For example, the reflection of incident internal waves propagating at an angle α\alpha depends on the slope criticality γ=tan⁡(β)/tan⁡(α)\gamma= \tan(\beta)/\tan(\alpha). The value of γ\gamma also determines turbulence during internal tide generation, e.g., resonant boundary layer response at near-critical slope with γ≈1\gamma\approx1 ([13]), breaking lee waves at supercritical (γ>1\gamma> 1) ridges ([20, 1]), and wave-beam boundary layers ([12]) at γ≥1\gamma\ge1 slopes.

The velocity ratio R∗=Uc/UtR^* = U_c/U_t, a measure of tidal strength relative to the mean, is the key non-dimensional parameter of interest in this study. To understand tidally dominated flow, we investigate the regime R∗≤1R^* \le1. Puthan et al. [35] has already considered the other regime (R=1/R∗<1R = 1/R^* < 1) of tidally modulated flow.

The tidal excursion number Ext=Ut/πftDEx_t = U_t/\pi f_t D, which is a non-dimensional distance traveled by a fluid particle due to tidal flow over half the tidal cycle (tidal excursion length), is also of interest. Winters and Armi [48], for example, discusses how a modified tidal excursion number is a measure of the relative importance of inertial and time-dependent effects near the topography. An excursion number can also be defined based on the mean velocity of a tidal current, as Exc=Uc/2ftDEx_c = U_c/2f_tD.

A useful parameter for the organization of wake vortices when Ut≳UcU_t \gtrsim U_c is the excursion number for the second half of the tidal cycle (π<ϕ<2π\pi< \phi< 2\pi),

Ex−(R∗)=Exc−Ext=Ut2ftD(R∗−2π).(10)Ex^{-}(R^*) = Ex_c - Ex_t = \frac{U_t}{2f_tD}\left(R^* - \frac{2}{\pi}\right). \tag*{(10)}

Eq. (10) implies that the net particle advection during the tidal half-cycle when the tidal current opposes the mean is negative (Ex−<0Ex^- < 0) when R∗<2/πR^* < 2/\pi. The magnitude of Ex−Ex^- quantifies relative displacements by the mean and tidal component during π<ϕ<2π\pi< \phi< 2\pi, and influences the vortices in the vicinity of the topography, as will be shown later.

Another parameter of interest is the frequency ratio f∗=fvs,c/ftf^* = f_{vs,c}/f_t (Puthan et al. [36]), where fvs,c=StcUc/Df_{vs,c} = St_cU_c/D is an estimate of the natural vortex shedding frequency in the absence of tidal oscillations. The Strouhal number Stc=0.273−1.11Rec−0.5+0.482Rec−1St_c = 0.273 - 1.11Re_c^{-0.5} + 0.482Re_c^{-1} is calculated with the UcU_c-based Reynolds number, using the Stc−RecSt_c - Re_c relationship for a cylinder wake proposed by Williamson and Brown [47].

In stratified flows with a velocity scale UU, the fluid particles have sufficient kinetic energy to overcome vertical displacements of up to U/NU/N. The topographic Froude number is a ratio of this vertical length scale to the height of the topography, and indicates the depth below which the upstream flow is blocked (Baines [2]; Winters and Armi [49]). Here, we define the Froude number as Frh=UM/NhFr_h = U_M/Nh, where UM=Uc+UtU_M = U_c + U_t is the free-stream velocity at peak tidal flow. The topography is dynamically tall if Frh<1Fr_h < 1, and nonlinear effects can occur.

The Rossby number Ro=UM/2πfΩDRo = U_M/2\pi f_\Omega D quantifies the importance of the background rotation in the system, where 2πfΩ2\pi f_\Omega is the Coriolis parameter. The earth’s rotation plays a critical role in the mesoscale regime, when Ro≲100Ro \lesssim10^0. For tidally dominated flows, even if Ro≳100Ro \gtrsim10^0, background rotation can play a significant role during parts of the tidal cycle when the tidal flow is opposite to the mean current.

The Reynolds number Re=UMD/νRe = U_M D/\nu is also defined based on the velocity at peak tidal flow. To understand the dissipation and mixing in the wake of the three-dimensional topography, we investigate Re∼104Re \sim10^4 wherein a turbulent near wake is observed downstream of the seamount (Puthan et al. [35]). Additionally, the non-dimensional numbers presented in Table 1 are to be associated with the bulk dynamics over an entire tidal cycle. The instantaneous, local magnitudes of these parameters can span a wide range of values.

R∗=Uc/UtR^* = U_c/U_t1.000.750.500.250.00
Ro=UM/2πfΩDRo = U_M/2\pi f_\Omega D10.609.277.956.625.30
Frh=UM/NhFr_h = U_M/Nh0.300.260.230.190.15
Exc=Uc/2ftDEx_c = U_c/2f_tD4.493.372.241.120.00
Ex−=Exc−ExtEx^- = Ex_c - Ex_t1.630.51-0.61-1.74-2.86
f∗=fvs,c/ftf^* = f_{vs,c}/f_t2.351.751.160.560.00
Re=UMD/νRe = U_MD/\nu2000017500150001250010000

Table 1. Important non-dimensional groups. Note that the aspect ratio of the seamount a=h/D=0.3a = h/D = 0.3, tidal excursion number Ext=Ut/πftD=2.86Ex_t = U_t/\pi f_tD = 2.86, and the Burger number Bu=(Ro/Frh)2=(Nh/2πfΩD)2=1253.5Bu = (Ro/Fr_h)^2 = (Nh/2\pi f_\Omega D)^2 = 1253.5, for all five cases considered.

The dissipation rates of kinetic energy per unit mass ϵK\epsilon_K and available potential energy per unit mass ϵP\epsilon_P are calculated as

ϵK=2(ν+νSGS)SijSij,(11)\epsilon_K = 2(\nu+ \nu_{\mathrm{SGS}})S_{ij}S_{ij}, \tag*{(11)}
ϵP=2(κ+κSGS)(g22ρ02N2)∂ρ∗∂xj∂ρ∗∂xj,(12)\epsilon_P = 2(\kappa+ \kappa_{\mathrm{SGS}})\left(\frac{g^2}{2\rho_0^2N^2}\right)\frac{\partial\rho^*}{\partial x_j}\frac{\partial\rho^*}{\partial x_j}, \tag*{(12)}

where ρ∗\rho^* is the deviation from the linear background density profile (Kang and Fringer [18]). Note, that the definition in Eq. (12) is valid for the dissipation rate of available potential energy only for a linear background stratification, which is the case in this study. To obtain statistically robust measures of the “wake dissipation,” volume integrals of the tidal-phase-averaged dissipation rates are computed. The ‘tidal phase average’ (or ‘phase average’) ξˉ(ϕ)\bar{\xi}(\phi) of a quantity ξ(t)\xi(t) at a tidal phase 0≤ϕ<2π0 \le\phi< 2\pi is defined as

ξˉ(ϕ)=∑n=n1n2ξ(t=(n−1)+(ϕ/2π)ft)/(n2−n1+1).(13)\bar{\xi}(\phi) = \sum_{n=n_1}^{n_2} \xi\left(t = \frac{(n-1)+(\phi/2\pi)}{f_t}\right)/(n_2-n_1+1). \tag*{(13)}

For our analysis, we allowed the simulations to spin up over nine tidal cycles, and used the flow fields from the next ten cycles for phase averaging, i.e., n1=10n_1 = 10 and n2=19n_2 = 19.

The volume integrals used to obtain the wake dissipation rates ϵK\epsilon_K and ϵP\epsilon_P are computed using a trapezoidal rule on the data from four horizontal transects at z/h∈{0.10,0.25,0.5,0.75}z/h \in\{0.10, 0.25, 0.5, 0.75\} and a midpoint rule on the three vertical transects at y/D∈{−0.375,0,0.375}y/D \in\{-0.375, 0, 0.375\}. The integration in the vertical direction is performed from z/h=0+=0.025z/h = 0^+ = 0.025, thus excluding the thin bottom viscous layer, up to z/h=1.2z/h = 1.2. Horizontally, the volume integration spans the entire spanwise extent of the domain (12D12D) and ±10D\pm10D in the streamwise direction. This “wake region,” within which the wake dissipation rates are calculated, encloses a volume of approximately 10.6 km310.6\ \mathrm{km}^3.

The wake dissipation is then compared against the “wave dissipation,” i.e., kinetic energy transported away from the seamount by the area-integrated internal wave flux F\mathcal{F} [32, 35], to potentially be dissipated elsewhere. The quantity F\mathcal{F}, defined as

F=∫−6D6D∫−10D10Dρ∗w∗∣z=1.2h dxdy,(14)\mathcal{F} = \int_{-6D}^{6D} \int_{-10D}^{10D} \left.\rho^* w^*\right|_{z=1.2h}\,\mathrm{d}x\mathrm{d}y, \tag*{(14)}

is computed as an integral over the 60 km2^2 area of the top of the subdomain within which the wake dissipation is calculated. Finally, the bulk mixing efficiency η=EP/(EP+EK)\eta= E_P/(E_P + E_K) [38, 33] is computed to quantify the extent of mixing occurring in the system, as a function of the tidal phase. The mixing efficiency is related to the mixing coefficient Γ≃EP/EK=η/(1−η)\Gamma\simeq E_P/E_K = \eta/(1-\eta) [6], which is usually used to estimate diffusivities from available dissipation measurements. Although Osborn [28] postulated Γ≤0.2\Gamma\le0.2 in a statistically steady state in the ocean, the inequality is often neglected in practice.

Selection of simulations

Table 1 presents the values of the different non-dimensional groups discussed above, in § 2.c. The Froude number and the aspect ratio of the topography that interacts with the simulated barotropic flow in a linear stratification are representative of abyssal seamounts that act as ‘stirring rods’ in the ocean. Since our study is motivated by the variability in boundary currents experienced by bottom topography, we investigate the effect of varying the velocity ratio R∗R^* by fixing UtU_t to a constant value and varying the magnitude of UcU_c. To investigate tidally dominated flows, we consider five values of the velocity ratio, R∗∈{1.00,0.75,0.50,0.25,0.00}R^* \in\{1.00, 0.75, 0.50, 0.25, 0.00\}, listed in decreasing order of the magnitude of UcU_c.

In dimensional terms, five values of UcU_c are considered between 0.1 and 0 m s−1^{-1}. The tidal flow has an amplitude Ut=0.1U_t = 0.1 m s−1^{-1} and a frequency ft=2.23×10−5f_t = 2.23 \times10^{-5} s−1^{-1}, which corresponds to the M2M_2 tidal component. The idealized seamount in the simulations has a height of h=150h = 150 m and a base diameter of D=500D = 500 m. The buoyancy frequency of N=4.44×10−3N = 4.44 \times10^{-3} s−1^{-1} associated with the linear stratification results in Frh≤0.3Fr_h \le0.3, and 2πfΩ=3.77×10−52\pi f_\Omega= 3.77 \times10^{-5} rad s−1^{-1} is the Coriolis parameter for a 15∘15^\circ N latitude. The Reynolds number is sufficiently large to establish a turbulent wake.

Temporal dynamics: turbulent wake, vertically coupled vortices, and lee waves

We begin with a discussion of the tidal phasing of the flow, wherein the streamwise velocity (uu) on a vertical plane at y/D=0y/D = 0 and the vertical vorticity (ωz\omega_z) on a horizontal plane at z/h=0.25z/h = 0.25 are investigated at four different phases from a representative tidal cycle (the 11th one is chosen). There is cycle-to-cycle variability in the vortex patterns owing to both wake turbulence and a low-frequency modulation by the Coriolis force. Nevertheless, there are overall trends in the vortex dynamics. These trends will be compared among the cases with R∗=Uc/Ut=1R^* = U_c/U_t = 1, 0.5, and 0 (pure tide). The distribution of wake vortices is a function of the tidal phase. To facilitate the discussion, individual vortices in figures are labeled for each case: AiR∗A_i^{R^*} and CiR∗C_i^{R^*} correspond to an anticyclonic and a cyclonic vortex, respectively, of index ii, from case R∗R^*. For example, A20.5A_2^{0.5} is an anticyclonic vortex for case R∗=0.5R^* = 0.5 and index 2, and tracked as the phase progresses.

Strength of the mean current equals the tidal amplitude (R∗=1R^*=1)

The discussion of temporal dynamics commences with the R∗=1R^*=1 case, where the free-stream current is always eastward with U∞≥0U_\infty\ge0. Insofar as wake eddies, although the value of f∗=2.35f^*=2.35, there is sufficient time for a Strouhal pair to be shed from the topography between ϕ=π/2\phi=\pi/2 to π\pi. This pair of cyclonic and anticyclonic vortices is able to escape the near wake during most tidal cycles and propagate downstream as coherent wake eddies. However, the vortices formed during the latter half of the cycle, when the tidal component is westward (negative), interact with the near wake and often are not able to propagate out of the near wake as coherent wake eddies.

  1. The tidal phasing is elaborated starting with ϕ=π/2\phi=\pi/2 when the free-stream velocity (U∞U_\infty) attains its maximum value of UM=Uc+UtU_M=U_c+U_t. A near wake with a large velocity defect (u<U∞u<U_\infty) is seen immediately downstream of the seamount (0<x/D≤10<x/D\le1) in Figure 2(a.Figure 1, below the lee waves emanating from the top of the seamount. A thin hydraulic jet with a thickness ∼UM/N\sim U_M/N is present at the interface between the lee waves and the velocity-defect region. The horizontal slice of the vertical vorticity (ωz\omega_z) in Figure 2(a.Figure 2 reveals near-wake turbulence (multiscale vorticity). Vortices shed earlier are also seen farther east, e.g., the anticyclonic vortex A−11A^1_{-1}, whose signature is also present in the velocity defect as a zone of heightened negative u−U∞u-U_\infty in Figure 2(a.Figure 1. In fact, the cyclonic and anticyclonic vortices C−11C^1_{-1} and A−11A^1_{-1}, respectively, form a Strouhal pair shed during the first half of the previous (10th) tidal cycle.

Evolution of the $R^*=1$ flow at four tidal phases

Figure 2. Evolution of the R∗=1R^*=1 flow is illustrated by four different tidal phases ϕ\phi in the 11th cycle: (a) π/2\pi/2, (b) π\pi, (c) 3π/23\pi/2 and (d) 2π2\pi. At each tidal phase, the top and bottom sub-panels present: (..1) the streamwise velocity deviation on a vertical transect at y/D=0y/D=0, and (..2) the vertical vorticity on a horizontal transect at z/h=0.25z/h=0.25. The sinusoid plotted as insets in (..1) represents the variation of the free-stream velocity U∞U_\infty. The red dot indicates the tidal phase and the dot-dashed line represents U∞=0U_\infty=0. The dashed lines indicate z/h=0.25z/h=0.25 in (..1), and y/D=0y/D=0 in (..2).

  1. Despite f∗=2.35f^*=2.35 suggesting a shedding period of approximately half the tidal cycle, the relatively large Ut=UcU_t=U_c results in the shedding of the near-wake vortices C11C^1_1 and A11A^1_1 of Figure 2(a.Figure 2 as a well-formed Strouhal pair within a quarter tidal cycle, as seen at ϕ=π\phi=\pi in Figure 2(b.Figure 2. The cyclonic vortex C21C^1_2 is better

developed, and a new anticyclonic vortex A11A_1^1 rolls up in the recirculation region. Thin, inclined layers of enhanced velocity defect are more visible in Figure 2(b.Figure 1 compared to Figure 2(a.Figure 1, and a comparison with the vorticity field in Figure 2(b.Figure 2 reveals their link to the peripheries of the vertically coupled vortices (see, for example, near x/D=1x/D = 1). Signatures of vertical overturns are seen in these thin layers, and are associated with the vertical shear at the peripheries of the coherent vortices. The velocity of the hydraulic jet above the recirculation zone is halved in Figure 2(b.Figure 1 since the U∞/NU_\infty/N scale is halved from ϕ=π/2\phi= \pi/2 to ϕ=π\phi= \pi. The wavelength of newly generated lee waves, although not very visible, is also shortened.

  1. The current decelerates to U∞=0U_\infty= 0 at ϕ=3π/2\phi= 3\pi/2. The hydraulic jet is absent and the lee waves are weakened in Figure 2(c.Figure 1. Velocity defect, albeit weaker, is still seen in the vertical transect because of the vortices that were shed earlier. The anticyclonic vortex A11A_1^1 in Figure 2(b.Figure 2 is formed in the recirculation region during the deceleration of U∞U_\infty, but it is not shed downstream and remains near the topography. From ϕ=π\phi= \pi to 3π/23\pi/2, it moves southward and interacts with the adjacent cyclonic vortex C21C_2^1 to form a vortex pair. The formation of the vortex pair before A11A_1^1 fully escapes the near wake in turn results in the upstream (westward) advection of the turbulent near wake, despite U∞U_\infty remaining non-negative throughout the tidal cycle. Note that a second cyclonic vortex C31C_3^1, carried over from the southern half of the recirculation region in Figure 2(b.Figure 2, is also advected upstream (westward) of the seamount. At phase 3π/23\pi/2, although U∞=0U_\infty= 0, strong turbulence with fine-scale variability is seen in the near wake owing to the interactions between the vortical structures generated earlier and the topography. As a result of these interactions, turbulence is also seen upstream of the seamount in Figure 2(c.Figure 2.

  1. As the current accelerates back to U∞=UcU_\infty= U_c at phase ϕ=2π\phi= 2\pi, the vortex pairs C11C_1^1-A11A_1^1 and C21C_2^1-A21A_2^1 and the turbulent near wake seen at ϕ=3π/2\phi= 3\pi/2 are advected downstream (Figure 2(d.Figure 2), along with previously shed vortices farther east of the seamount. Note that the C21C_2^1-A21A_2^1 vortex pair has moved away from y/D=0y/D = 0 because of a self-induced southward velocity. The cyclonic vorticity associated with C31C_3^1 is not shed as a coherent vortex as it interacts with A21A_2^1 and breaks down into turbulence. The streamwise velocity field in Figure 2(d.Figure 1 also reflects the downstream advection of previously existing features, except immediately downstream of the seamount. Remnants of upstream velocity defect seen in Figure 2(c.Figure 1, albeit weak, are present at the end of the tidal cycle,

Stark differences exist between Figure 2(b) and (d) although both phases have U∞=UcU_\infty= U_c. The excursion number for the latter half of the tidal cycle is Ex−(R∗=1)=1.63∼100Ex^{-}(R^* = 1) = 1.63 \sim10^0. The ∼1\sim1 value of Ex−Ex^{-} for the last half of the tidal cycle has little bearing on the flow field between ϕ=π\phi= \pi and ϕ=2π\phi= 2\pi in Figure 2, since the C21C_2^1-A21A_2^1 vortex pair, for example, has not been advected far away from the topography by ϕ=2π\phi= 2\pi. Finally, a comparison between Figure 2(d) and (a) allows us to complete the tidal cycle. In fact, the nonlinear evolution of the C21C_2^1-A21A_2^1 vortex pair, together with the turbulent breakdown of C31C_3^1, gives rise to vortical structures in cycle 12 that are analogous to the C01C_0^1-A01A_0^1 pair of cycle 11 in Figure 2(a).

In summary, while there is some cycle-to-cycle variability, as revealed by Supplemental Fig. S1, overall trends can be inferred. During the first half of the tidal cycle for the R∗=1R^* = 1 case with f∗=2.35f^* = 2.35, when the tidal component is in the direction of the mean current, a Strouhal pair is shed. Within the Strouhal pair, the second vortex tends to be larger and stronger (also see Supplemental Fig. S1) because the generation of the first is hindered by the near wake dynamics of the previous cycle. During the latter half of the tidal cycle, at least two new vortices are generated, but they are not shed as coherent vortices owing to low U∞U_\infty. Instead, they form a vortex pair close to the seamount and influence the dynamics of the near wake, even when U∞=0U_\infty= 0 at peak westward tidal flow. The vortex pair is laterally displaced from the centerline (y=0y = 0) and then advected away from the seamount as U∞U_\infty grows stronger eastward.

Strength of the mean current equals half the tidal amplitude (R∗=0.5R^* = 0.5), and no mean current/pure tide (R∗=0R^* = 0)

Moving to cases with R∗=0.5R^* = 0.5 and 0, qualitative differences in the wake eddies are observed. Unlike R∗=1R^* = 1, both cases experience a westward current (U∞<0U_\infty< 0) during parts of the tidal cycle, i.e., when the sinusoids plotted in the insets of Figure 3(a.Figure 1 and (b.1) lie below the respective dot-dashed lines. Regardless, at ϕ=π/2\phi= \pi/2, eastward U∞U_\infty peaks at UM=Uc+UTU_M = U_c + U_T for all R∗R^*.

Eight-panel comparison of flow fields for $R^{\ast}=0.50$ and $R^{\ast}=0.00$

Figure 3. Same as Figure 2, but for R∗=0.50R^{\ast}=0.50 (left column) and R∗=0.00R^{\ast}=0.00 (right column). Note the vertical shift in the U∞=0U_{\infty}=0 indicator (dot-dashed line) with respect to the sinusoid in the insets, as R∗R^{\ast} is varied.

In the R∗=0.5R^* = 0.5 case, the UcU_c-based vortex shedding frequency is half that at R∗=1R^* = 1 — see Table 1: f∗=1.16f^* = 1.16 and 2.35, for R∗=0.5R^* = 0.5 and 1, respectively. Because of the slower vortex shedding, unlike R∗=1R^* = 1, two fully formed vortex pairs are unable to escape the near wake as coherent wake eddies during each tidal cycle. Instead, at R∗=0.5R^* = 0.5, a single vortex of alternating polarity and a Strouhal pair is shed eastward every tidal cycle. Other vortices resulting from flow separation are observed, but do not remain coherent as they are advected away from the topography. For R∗=0R^* = 0, on the other hand, Uc=0U_c = 0, and hence the mean-current-based vortex shedding is irrelevant. Instead, phase eddies are formed during tidal acceleration and shed during deceleration — see observations of [4],

Pawlak et al. [29] and Wynne-Cattanach et al. [52], for example — instead of Strouhal pairs.

  1. Similar to R∗=1R^{\ast}=1, recirculation regions with large velocity defects are present for both R∗=0.5R^{\ast}=0.5 and 00, in Figure 3(a.Figure 1 and (b.1), respectively. Hydraulic jets with U∞/NU_{\infty}/N length scale and lee waves are also seen

above the recirculation regions. The near-wake ωz\omega_z fields in Figure 3(a.Figure 2 and (b.2) correspond to different stages of vortex shedding. Farther east, Figure 3(a.Figure 2 reveals a previously shed vortex A−20.5A_{-2}^{0.5} and a C−10.5C_{-1}^{0.5}-A−10.5A_{-1}^{0.5} vortex pair (y/D≈−2y/D \approx-2) that were advected eastward away from the seamount. A weaker cyclonic vortex C00.5C_0^{0.5} is also seen, but it does not remain coherent until the end of the tidal cycle.

For R∗=Uc/Ut=0R^* = U_c/U_t = 0, the absence of mean advection implies that previously shed vortices remain in the vicinity of the seamount and potentially interact with the near wake. This is evidenced in Figure 3(b.Figure 2, where a few vortices (C00,A−10,C−10C_0^0,A_{-1}^0,C_{-1}^0) that were previously shed westward, and subsequently advected back towards the seamount during 0<ϕ≤π/20 < \phi\leq\pi/2, are seen. As will be seen, the northern vortex C00C_0^0 has a strong influence on the subsequent evolution of the wake vortices.

  1. At the later phase ϕ=π\phi= \pi, Figure 3(d.Figure 2 reveals that C00C_0^0 pairs with a newly formed (during 0<ϕ<π/20 < \phi< \pi/2) A10A_1^0, resulting in the shedding of a vortex pair C00C_0^0-A10A_1^0 to the north-east of the seamount. A vortex pair A10.5A_1^{0.5}-C20.5C_2^{0.5} is also seen in Figure 3(c.Figure 2 for R∗=0.5R^* = 0.5, but the mechanism here is similar to R∗=1R^* = 1, wherein the cyclonic vorticity in the vortex pair originates from the southern half of the recirculation region, as opposed to a previously shed vortex for R∗=0R^* = 0. Inclined layers of enhanced velocity defect are more clearly seen during the deceleration stage for both R∗R^*, in Figure 3(c.Figure 1 and (d.1), as was the case for R∗=1R^* = 1.

  1. Later, at tidal phase ϕ=3π/2\phi= 3\pi/2, the current with U∞<0U_\infty< 0 attains a westward maximum for all R∗<1R^* < 1, and a near wake region with a ‘velocity defect’ is seen to the west of the seamount in Figure 3(e.Figure 1 and (f.1). Lee wave signatures, along with the hydraulic jet, with length scales ∼∣U∞∣/N\sim|U_\infty|/N, are also seen west of the seamount, above the recirculation region. As seen in Figure 3(e.Figure 2 and (f.2), the vortex pairs shed earlier to the northeast are advected westward by the negative U∞U_\infty and northward by their self-induced velocities. New recirculation regions are formed to the west of the seamount since U∞U_\infty is negative, and the near wake is better developed for R∗=0R^* = 0 because of a larger ∣U∞∣|U_\infty|, and a longer exposure to U∞<0U_\infty< 0.

The polarity of the northern and southern vortices within the recirculation region is reversed for a westward wake compared to an eastward wake — for example, the northern vortex C00C_0^0 in the recirculation region is cyclonic in Figure 3(f.Figure 2, but A10A_1^0 is anticyclonic in Figure 3(b.Figure 2. An implication of the polarity reversal, as the current gradually changes its direction, is the shedding of residual vorticity from the recirculation region on the eastern side as small vortices and/or vortex pairs, as seen immediately towards the west of the seamount in Figure 3(e.Figure 2 and (f.2). An example is the small vortex pair near (x/D,y/D=2,0.5x/D,y/D = 2,0.5) in Figure 3(b.Figure 2.

  1. At the end of the tidal cycle, the flow features seen in Figure 3(g.Figure 2 are reminiscent of Figure 3(e.Figure 2, albeit the turbulent breakdown of the northern A10.5A_1^{0.5}-C20.5C_2^{0.5} vortex pair. This is reasonable for R∗=0.5R^* = 0.5, given its low Ex−=−0.61Ex^- = -0.61, i.e., small westward advection by the current. In contrast, R∗=0R^* = 0 has an order of magnitude larger Ex−=−2.86Ex^- = -2.86. Therefore, Figure 3(h) is almost a mirror image of Figure 3(d), with a vortex pair A10A_1^0-C10C_1^0 shed to the northwest.

The vertical sections of defect velocity have distinctive characteristics. For case R∗=0.5R^* = 0.5, Figure 3(g.Figure 1 shows the persistence of large velocity defects west of the seamount at the end of the tidal cycle, despite R∗=0.5R^* = 0.5 experiencing westward free-stream velocities for only a fraction of the tidal cycle. For case R∗=0R^* = 0, there are well-formed wakes at ϕ=π/2\phi= \pi/2 (Figure 3b.Figure 1) and ϕ=3π/2\phi= 3\pi/2 (Figure 3f.Figure 1) on the eastward and westward sides, respectively, with similar geometry but of opposite sign. Although the free-stream current is zero at ϕ=π\phi= \pi (Figure 3d.Figure 1) and 2π2\pi (Figure 3h.Figure 1), there is significant local velocity near the topography.

In summary, over the course of the 11th tidal cycle, a cyclonic vortex and a vortex pair were shed downstream, to the south and north of the y/D=0y/D = 0 line, respectively, for the R∗=0.5R^* = 0.5 case (f∗=1.16f^* = 1.16). Similar to R∗=1R^* = 1, the coherent vortex was shed during the eastward tide, and the vortex pair was generated when the tide flowed westward. For R∗=0.5R^* = 0.5, the subsequent (12th) tidal cycle (figures not shown) shows mirrored (around the y=0y = 0 line) vortex shedding: an anticyclonic vortex shed to the north of y/D=0y/D = 0 and a vortex pair to the south.

For the pure tide (R∗=0R^* = 0), however, a potentially bistable state was seen. Anticyclonic and cyclonic vortices were shed to the north of y/D=0y/D = 0, during the first (eastward tide) and second (westward tide) halves of the tidal cycle, respectively. The newly shed vortices combined with previously shed vortices to form vortex pairs that advected away in the yy direction due to self-induced velocities. During either half of the cycle, the presence of previously shed vortices north of the seamount influenced the recirculation region, locking the vortex shedding to the north of y/D=0y/D = 0 across multiple tidal cycles. The sustained northward vortex pair shedding is likely one of two possible scenarios — we believe the system is equally likely to lock onto a vortex-pair shedding towards the south of y/D=0y/D = 0.

Temporal spectra for R∗∈{1.00,0.75,0.50,0.25,0.00}R^* \in\{1.00,0.75,0.50,0.25,0.00\}

The tidal phasing of the wake vortex dynamics demonstrates that wake dynamics vary considerably with R∗R^*, in tidally dominated flows past 3D topography. However, comprehensively mapping these complex wake dynamics in the ocean can be challenging. Localized velocity measurements are obtained from moorings and ADCPs during field experiments, and a spectral analysis can reveal links to the dominant forcing in the system (i.e., the tide). Hence, we present the temporal spectra of uu and vv measured at a point inside the near wake for all R∗R^* in Figure 4(a) and (b), respectively. To link these velocity spectra to the vortex dynamics in observed in the simulations, temporal spectra of area-averaged vertical vorticity are presented in Figure 4(c). Since dissipation/mixing is a primary focus of this work, the power spectral density of the area-integrated ϵK\epsilon_K is presented in Figure 4(d).

Temporal spectra of streamwise velocity, spanwise velocity, squared vertical vorticity, and dissipation rate

Figure 4. Temporal spectra of (a) the streamwise velocity uu, (b) the spanwise velocity vv, (c) squared vertical vorticity ωz2\omega_z^2, and (d) dissipation rate of kinetic energy per unit mass ϵK\epsilon_K, for the five R∗R^* considered in this study. In (a) & (b), the spectra are calculated from measurements at a point (x/D,y/D,z/h)=(1.5,0,0.25)(x/D,y/D,z/h)=(1.5,0,0.25). The spectra in (c) & (d) are for the horizontal-area integrals of the respective quantities, at z/h=0.25z/h=0.25. The vertical lines indicate the background rotation fΩf_{\Omega} (dotted), tidal frequency & its harmonics nftnf_t (black solid), and half the tidal frequency & its odd multiples (2n+1)ft/2(2n+1)f_t/2 (gray dashed), with n=0,1,2,3,4n=0,1,2,3,4. Red circles indicate the most dominant peak in each spectrum.

Each temporal spectrum in Figure 4 was calculated from a data set of around 2700 samples collected over 18 tidal cycles, after subtracting the temporal mean. The sampling frequency is almost 5 minutes so that the highest frequency resolved in the post-processing is 75 ftf_t (Nyquist criterion), where ft=2.237×10−5 s−1f_t=2.237\times10^{-5}\ \mathrm{s}^{-1} is the M2M_2 tidal frequency. Welch’s method was used to obtain spectra, and each data set was divided into nine segments with an overlap of 10%, with a Hanning window to reduce edge effects. Recall, that the Coriolis parameter (2πfΩ)(2\pi f_{\Omega}) at 15° N considered in the simulations corresponds to fΩ≈ft/4f_{\Omega}\approx f_t/4 (dotted line in Figure 4).

The temporal spectra of the streamwise (uu) and spanwise (vv) velocity are measured at the point (x/D,y/D,z/h)=(1.5,0,0.25)(x/D,y/D,z/h)=(1.5,0,0.25) in the near wake to compare with the previous work on tidally modulated (R∗>1R^*>1) flow (Putnam et al. 2022). The uu-spectra in Figure 4(a), for all five R∗R^*, present a dominant peak at ftf_t, the streamwise forcing frequency in the system. This is in agreement with the observations for tidally modulated flows past a seamount (Putnam et al. 2021). Spectral peaks are also seen at integer multiples of ftf_t.

The vv-spectra at the near-wake measurement location do not all present a dominant peak at ftf_t in Figure 4(b), as a result of tidal synchronization of vortex shedding (Putnam et al. 2021). This means that the intrinsic vortex shedding frequency fvs,cf_{\mathrm{vs,c}} of the mean current matters to the ν\nu-spectra. The peak for R∗=1R^{*}=1 is at the subharmonic ft/2f_t/2 and its integer multiples of ft/2f_t/2, similar to the R∗=1R^{*}=1 case of Puthan et al. [35] that overlaps with the present strong-tide regime. For the next smaller R∗R^{*} of 0.75 and 0.5, the peaks are ftf_t and 3ft/23f_t/2, respectively. The R∗=0.5R^{*}=0.5 case exhibits an additional local peak at ft/2f_t/2. For the two smallest R∗=0.25R^{*}=0.25 and R∗=0R^{*}=0 with the weakest mean current (Uc<0.5UtU_c<0.5U_t), the dominant peak is at ftf_t, similar to the uu-spectra. In all cases, spectral peaks are seen at integer multiples of the dominant frequency.

The coherent vortex dynamics near the seamount is reflected in the power spectra of the area-integrated ωz\omega_z presented in Figure 4(c). The area integral is performed on the horizontal transects at z/h=0.25z/h=0.25, presented earlier in Figure 2 and Figure 3. A dominant peak is seen at the forcing frequency ftf_t for the three cases with strong mean currents R∗=1R^{*}=1, 0.75, and 0.5. Although the two cases with the weakest mean currents, R∗=0.25R^{*}=0.25 and 0, also show peaks at ftf_t, the frequency 2ft2f_t is more dominant. This is because for both R∗R^{*}, vortices (and/or vortex pairs) are shed during the latter half of the tidal cycle as well, as suggested by the relatively large negative magnitudes of excursion number Ex−=−1.74Ex^{-}=-1.74 and −2.86-2.86, for R∗=0.25R^{*}=0.25 and 0, respectively.

The spectra for R∗=1R^{*}=1, 0.75 and 0.5 are qualitatively similar to corresponding ωz\omega_z spectra in Figure 4(c). This, however, does not mean that the horizontal shear is the primary source of dissipation. As will be demonstrated later, a significant portion of the wake dissipation occurs by vertical shear at the peripheries of the vertically coupled coherent horizontal vortices, and this link to ωz\omega_z is the reason for the qualitative similarity between the ϵK\epsilon_K and ωz\omega_z spectra and their peaks at ftf_t. The dominant peak of the ϵK\epsilon_K-spectrum for R∗=0.25R^{*}=0.25 is at ftf_t with another slightly smaller peak at 2ft2f_t (the peak in its ωz\omega_z spectrum). The dominant peak of the ϵK\epsilon_K-spectrum for R∗=0R^{*}=0 is clearly at 2ft2f_t, matching its dominant peak in the ωz\omega_z spectrum.

Instantaneous dissipation rates of kinetic energy

As a first step towards understanding mixing in such tidally dominated topographic flows, the tidal phasing of the dissipation rate of kinetic energy per unit mass (ϵK\epsilon_K) is investigated. It will be shown that there is substantial phase-dependent dissipation (ϵK∼10−5 W kg−1\epsilon_K\sim10^{-5}\ \mathrm{W\,kg}^{-1}) in the near wake (x/D≤2.5x/D\le2.5). Patches of dissipation ϵK∼10−7 W kg−1\epsilon_K\sim10^{-7}\ \mathrm{W\,kg}^{-1}, associated with wake eddy instabilities, are seen further downstream up to x/D≈6x/D\approx6 and laterally up to y/D≈±2y/D\approx\pm2.

  1. At ϕ=π/2\phi=\pi/2 (Figure 5a), the near wake displays large dissipation rates with ϵK∼10−5 W kg−1\epsilon_K\sim10^{-5}\ \mathrm{W\,kg}^{-1} as the strongest values. The vertical transect in Figure 5(a.Figure 1 reveals that the hydraulic jet, which breaks down into turbulence approximately one wavelength downstream of the seamount, is a site of intense dissipation.

Dissipation rates $\epsilon_K$ at four different tidal phases

Figure 5. Dissipation rates ϵK\epsilon_K (in W kg−1^{-1}) for R∗=1R^*=1, at four different tidal phases: (a) π/2\pi/2, (b) π\pi, (c) 3π/23\pi/2 and (d) 2π2\pi. At each tidal phase, the top and bottom sub-panels correspond to a vertical transect at y/D=0y/D=0 and a horizontal transect at z/h=0.25z/h=0.25, respectively. The dashed lines in the vertical and horizontal transects indicate z/h=0.25z/h=0.25 and y/D=0y/D=0, respectively. The ϵK\epsilon_K fields are to be compared with the velocity and vorticity fields of Figure 2. The blue contour in panel (a) corresponds to a dissipation rate ϵK=10−8\epsilon_K=10^{-8} W kg−1^{-1}.

The recirculation region below the jet displays significant ϵK\epsilon_K as well, albeit an order of magnitude smaller. Non-trivial dissipation is also seen near the site of lee wave generation, but with much smaller values, ϵK∼10−8 W kg−1\epsilon_K\sim10^{-8}\ \mathrm{W\,kg}^{-1}. The vortices shed earlier and advected eastward to x/D≈6x/D\approx6 also display ϵK∼10−7 W kg−1\epsilon_K\sim10^{-7}\ \mathrm{W\,kg}^{-1} along their peripheries in Figure 5(a.Figure 2. Thus, instability of wake eddies results in significant dissipation at a large downstream distance of x/D≈6x/D\approx6. The strongest dissipation rates (ϵK∼10−5 W kg−1\epsilon_K\sim10^{-5}\ \mathrm{W\,kg}^{-1}) in the horizontal transect are associated with the attached shear layers which form when, under strong stratification, the flow goes around the 3D topography and separates laterally.

  1. The ‘elongated’ near wake seen in Figure 5(b.Figure 1 at ϕ=π\phi=\pi, owing to the wake-vortex dynamics presented in Figure 2(b), exhibits large ϵK∼10−6 W kg−1\epsilon_K\sim10^{-6}\ \mathrm{W\,kg}^{-1}. A comparison with Figure 2(b.Figure 1 reveals that the inclined layers of enhanced velocity defect that show vertical overturns are sites of strong dissipation. The link between the inclined layers to vortex peripheries is further supported by correspondence between the ϵK∼10−7 W kg−1\epsilon_K\sim10^{-7}\ \mathrm{W\,kg}^{-1} signatures at x/D=3.5x/D=3.5 in Figure 5(b.Figure 1 and Figure 5(b.Figure 2. In general, the near-wake dissipation regions are more spread out at ϕ=π\phi=\pi (U∞=UcU_{\infty}=U_c) compared to ϕ=π/2\phi=\pi/2 (U∞=Uc+UtU_{\infty}=U_c+U_t). However, the strongest dissipative features, i.e., the hydraulic jet and the attached shear layers, do not contribute as much to wake dissipation at ϕ=π\phi=\pi as they do at ϕ=π/2\phi=\pi/2. Hence, as will be shown, the volume-integrated value of ϵK\epsilon_K at ϕ=π\phi=\pi is lower than at ϕ=π/2\phi=\pi/2.

  1. The inclined layers of ϵK∼10−7 W kg−1\epsilon_K\sim10^{-7}\ \mathrm{W\,kg}^{-1} persist when U∞=0U_{\infty}=0 at tidal phase ϕ=3π/2\phi=3\pi/2 in Figure 5(c.Figure 1. They are associated with the peripheries of previously shed vortices seen in Figure 5(c.Figure 2. The turbulent recirculation region in Figure 5(c.Figure 2, which almost surrounds the topography, shows ϵK∼10−7 W kg−1\epsilon_K\sim10^{-7}\ \mathrm{W\,kg}^{-1}. An interesting finding, as a result, is the presence of ‘rising dissipation signatures’ upstream of the seamount, despite U∞U_{\infty} remaining non-negative throughout the tidal cycle.

  1. Ultimately, at the end of the tidal cycle in Figure 5(d), the dissipative structures seen earlier are advected downstream, as the current accelerates back up to U∞=UcU_{\infty}=U_c. The attached shear layers at the boundary of the small, newly-formed recirculation region present enhanced ϵK∼10−6 W kg−1\epsilon_K\sim10^{-6}\ \mathrm{W\,kg}^{-1} (cf. Figure 2(d.Figure 2 and Figure 5(d.Figure 2). The vertical transect in Figure 5(d.Figure 1 shows a dissipative hydraulic jet, and lee waves with non-trivial ϵK∼10−8 W kg−1\epsilon_K\sim10^{-8}\ \mathrm{W\,kg}^{-1}. Signatures of nascent vertical overturns are also seen, at (x/D,z/h=1,1)(x/D,z/h=1,1), for example, with ϵK∼10−7 W kg−1\epsilon_K\sim10^{-7}\ \mathrm{W\,kg}^{-1}. The dissipative

structures seen upstream of the seamount in Figure 5(c) persist, albeit with weaker ϵK\epsilon_K magnitudes.

To understand the role played by the subgrid-scale model in the dissipation rates observed in our LES, the ratio of subgrid-scale viscosity and molecular viscosity (νSGS/ν\nu_{\mathrm{SGS}}/\nu) for R∗=1R^*=1 is presented in Supplemental Fig. S2. As indicated by the probability density functions of νSGS/ν\nu_{\mathrm{SGS}}/\nu, the median value of the ratio is vanishingly small (∼10−14\sim10^{-14}), and non-trivial νSGS\nu_{\mathrm{SGS}} is localized to the near wake. Thus, by design, the resolution of the turbulence by the chosen grid is very good.

To gain some insight into the underlying mechanisms for the dissipation seen in the system, contributions from each of the six independent components of the instantaneous strain rate tensor to the dissipation rate ϵK\epsilon_K presented in Figure 5(a) are presented in Figure 6. The blue contours in Figure 5(a), corresponding to a dissipation rate of ϵK=10−8\epsilon_K=10^{-8} W kg−1^{-1}, are also plotted in red in all 6 panels of Figure 6 to delineate regions of strong dissipation.

Six panels showing contributions from the strain-rate tensor components

Figure 6. Contribution from the six unique components of the strain-rate tensor (a) S11S_{11}, (b) S33S_{33}, (c) S22S_{22}, (d) S12S_{12}, (e) S13S_{13}, and (f) S23S_{23}, to the dissipation rate ϵK\epsilon_K at tidal phase ϕ=π/2\phi= \pi/2 for R∗=1R^* = 1, which was presented in Figure 5(a). The red contour in each panel corresponds to ϵK=10−8 W kg−1\epsilon_K = 10^{-8}\ \mathrm{W\,kg}^{-1}, and matches the blue contour in Figure 5(a).

Outside the turbulent near wake, the normal strain rate S11S_{11} in Figure 6(a) contributes primarily to the dissipation associated with the lee waves above the seamount. Contributions to the wake waves, e.g. [14], and region of ‘upstream-blocking’ [49] are also seen, but these regions correspond to low ϵK\epsilon_K in Figure 5(a). The vertical strain-rate S33S_{33} in Figure 6(b) dominates the lee wave dissipation rates, and also contributes to the wake waves that are eastward and above the seamount wake.

The horizontal shear of horizontal velocities S12S_{12} does not play a role downstream of the near wake in the vertical plane, in Figure 6(d.Figure 1. Its contribution is maximized at the centers of shed vortices (c.f. Figure 6(d.Figure 2 and Figure 2(a.Figure 2), but these are not regions of strong dissipation (c.f. Figure 6(d.Figure 2 and Figure 5(a.Figure 2). As a result, even though vertically coupled horizontally eddying motions dominate the wake, the seen dissipation rates ϵK\epsilon_K are not because of the horizontal strain S12S_{12}. Instead, the two vertical shear components S13S_{13} and S23S_{23} dominate, especially in the inclined layers of strong dissipation at the peripheries of the coherent vortices. Note, that S23S_{23} contributes to the dissipation associated with the wake waves, but not the lee waves.

Within the immediate near wake, the turbulent flow is closer to three-dimensional, and the contributions from the different strain-rate components are of the same order of magnitude. The dissipation in the attached shear layers, before they transition to turbulence, is primarily due to the horizontal shear S12S_{12}, in addition to S13S_{13}. Similarly, the laminar dissipation in the hydraulic jet, prior to transition, is due to S11S_{11}, S33S_{33}, and S13S_{13}.

For completeness, Figure 7 presents ϵK\epsilon_K for the cases with R∗=0.5R^* = 0.5 and 0 (pure tide) whose velocity/vorticity fields were presented in Figure 3. For brevity, we refrain from an elaborate description. Similar to R∗=1R^* = 1, strongest

Dissipation rate fields for $R^{*}=0.50$ and $R^{*}=0.00$

Figure 7. Same as Figure 5, but for R∗=0.50R^{*}=0.50 (left column) and R∗=0.00R^{*}=0.00 (right column). The dissipation rate fields are to be compared with the velocity and vorticity fields of Figure 3.

ϵK∼10−5 W kg−1\epsilon_{K} \sim10^{-5}\ \mathrm{W}\,\mathrm{kg}^{-1} is associated with hydraulic jets and attached shear layers with additional contributions from inclined velocity defect layers in the near wake and peripheries of wake eddies. The phase/spatial distribution of ϵK\epsilon_{K} corresponds to that of the velocity and vorticity fields shown in Figure 3. Both R∗=0.5R^{*}=0.5 and 00 show strong dissipation signatures west of the seamount by the end of the tidal cycle. These ‘upstream’ signatures are stronger for smaller

R∗R^{*} because the more dominant the tide, the stronger (and longer) the westward U∞U_{\infty} experienced by the topography.

Phase-averaged dissipation and mixing

This section is devoted to the dependence of dissipation and mixing statistics on the relative strength of the tidal component, R∗R^{*}. A tidal phase average is employed to obtain a robust statistical measure. The volume integrated dissipation — its phase dependence, and cycle average — varies systematically with R∗R^{*}. These wake dissipation values owing to the ‘stirring rod’ turbulence of 3D topographies are compared with the internal wave flux (wave dissipation) and found to be substantially larger. The mixing of the buoyancy field by turbulence is also quantified.

The wake structures and corresponding dissipation signatures discussed in § 3 and § 4 constitute one realization of an ensemble of plausible scenarios arising from vortex-vortex and vortex-topography interactions subject to turbulence. In addition, there is low-frequency modulation due to background rotation. Supplemental fig:S1 presents ten realizations of the ϕ=π/2\phi= \pi/2 tidal phase for R∗=1R^{*} = 1, taken across ten consecutive tidal periods. The ensemble exhibits fluctuations around a well-defined state: two developed wake eddies in 2<x/D<62 < x/D < 6 preceded by eddy formation at the topography, 0<x<D<20 < x < D < 2. Phase-averaged statistics, obtained here by utilizing data from ten consecutive cycles from the tenth cycle (9≤tft<199 \le t f_t < 19), will be used to characterize the flow statistics.

Figure 8 presents the phase-averaged dissipation, denoted by ϵ‾K\overline{\epsilon}_{K}, for R∗=1R^{*} = 1 at four different tidal phases, with Figure 5 showing one of the ten realizations used for the phase averaging. The ϵ‾K\overline{\epsilon}_{K} fields at all four ϕ\phi are relatively smooth compared to the instantaneous ϵK\epsilon_{K} fields, and approximately symmetric about the y/D=0y/D = 0 line (except at ϕ=2π\phi= 2\pi, to be discussed shortly), indicating that the 10-cycle phase average is adequate. The north-south symmetry of ϵ‾K\overline{\epsilon}_{K} is also in agreement with the cyclonic-anticyclonic symmetry in a vertical shear dominated regime, as discussed by Srinivasan et al. [39]. Importantly, as summarized below, the large ϵK\epsilon_{K} values encountered during the 11th cycle are representative of the phase average. The hot spots of dissipation, although somewhat spatially smeared, are robust to phase averaging. For example, at ϕ=π/2\phi= \pi/2, the ϵ‾K\overline{\epsilon}_{K} fields present the strongest dissipation rates ϵ‾K∼10−5\overline{\epsilon}_{K} \sim10^{-5} W kg−1^{-1} at the hydraulic jet and the attached shear layers in Figure 8(a.Figure 1 and (a.2), respectively. Signatures of dissipation along the peripheries of inclined vortical structures are visible (4≤x/D<84 \le x/D < 8) in both vertical and horizontal slices. Non-trivial ϵK\epsilon_K is also seen in the lee wave region above the seamount and the near wake. When U∞U_{\infty} reaches zero at phase ϕ=3π/2\phi= 3\pi/2, Figure 8(c.Figure 1 reveals clear signatures of the inclined layers corresponding to ϵK∼10−7 W kg−1\epsilon_K \sim10^{-7}\ \mathrm{W\,kg}^{-1} or greater. Robust signatures of enhanced upstream dissipation are also seen in Figure 8(c.Figure 2, owing to eddy-induced westward flow.

Phase-averaged dissipation rate fields

Figure 8. Same as Figure 5, but the quantity presented is the phase-averaged dissipation rate ϵ‾K\overline{\epsilon}_{K}, in W kg−1^{-1}.

The ϵK\epsilon_K field at ϕ=2π\phi= 2\pi in Figure 8(d.Figure 2 is somewhat asymmetric, favoring larger values south of the topography, compared to the three other ϕ\phi. This is likely related to the evolution of the vortex pairs that are mostly shed towards the south of the seamount, as the flow accelerates back up from U∞=0U_{\infty} = 0. A well-formed near wake is absent during this interval and the effect of background rotation on the vortex pair generated at ϕ=3π/2\phi= 3\pi/2 is different depending on whether it is shed towards the north or south of the seamount. It is possible a vortex pair is more likely to be shed south of the seamount due to the action of the Coriolis force, but this remains to be explored. Nevertheless, the hydraulic jet and attached shear layers are seen again in Figure 8(d), with ϵK∼10−7 W kg−1\epsilon_K \sim10^{-7}\ \mathrm{W\,kg}^{-1}, and signatures of nascent vertical overturnings are visible along inclined layers.

To delineate the importance of the dissipation in the wake, we present the volume-integral of the kinetic energy dissipated in the wake of the seamount in Figure 9. A comparison is also made with the wave dissipation, i.e., kinetic energy transported away from the seamount by the internal wave flux, to potentially be dissipated elsewhere. To understand how the wake dissipation from a bulk integral may be represented in field experiments, an area integral over a vertical transect as seen from a ship is also presented.

Four-panel plots of tidal phase dependence for wake and wave dissipation

Figure 9. Top row shows the tidal phase dependence of: (a) the wake dissipation EK\mathcal{E}_K, and (b) the wave dissipation (i.e., the internal wave flux) F\mathcal{F}. Bottom row: (c) F/EK\mathcal{F}/\mathcal{E}_K, the ratio of wave to wake dissipation, and (d) EKy0\mathcal{E}_K^{y0}, an approximation of EK\mathcal{E}_K obtained using an area integral over a vertical transect of measurements at y/D=0y/D=0, as a function of ϕ\phi. The quantities in (a)–(d) are computed from phase-averaged fields, and the standard deviations across tidal cycles are represented using error bars at each ϕ\phi. The black dashed vertical lines correspond to ϕ=π/4\phi=\pi/4 and 3π/43\pi/4, the tidal phases when U∞U_\infty attains its maximum and minimum values.

Figure 9(a) presents the variation of the volume-integrated dissipation rate EKE_K (defined in the figure legend), as a function of ϕ\phi. For R∗=1R^{*} = 1, EKE_K increases monotonically with ϕ\phi during 0≤ϕ<π/20 \le\phi< \pi/2, since newer and stronger dissipative structures are generated in the near wake as U∞U_{\infty} continues to accelerate. EKE_K continues to increase as the U∞U_{\infty} starts to decelerate until EKE_K attains a peak shortly (≈37\approx37 minutes for the M2M_2 tide in this problem) after ϕ=π/2\phi= \pi/2. The offset is because of the persistence of strong dissipative features for a short while after their generation, while new dissipative features are generated. For the remainder of the deceleration stage, from π/2<ϕ≤π\pi/2 < \phi\le\pi, ϵK\epsilon_K decreases monotonically with ϕ\phi. Similar trends are seen for all five R∗R^{*} for the first half of the tidal cycle, 0≤ϕ<π0 \le\phi< \pi.

As the tidal component starts flowing westward for ϕ>π\phi> \pi, ϵK\epsilon_K continues to decrease for R∗=0.75R^{*} = 0.75 and 11, before increasing again towards the end of the cycle. However, ϵK\epsilon_K for the three remaining R∗=0R^{*} = 0, 0.250.25, and 0.50.5, which correspond to Ex−(R∗)<0\mathrm{Ex}^{-}(R^{*}) < 0 (the net current becomes

westward), present a second peak in the latter half of the tidal cycle. For all three cases, the second peak occurs near ϕ=3π/2\phi= 3\pi/2, at maximum westward current. Interestingly, the ϵK\epsilon_K corresponding to the second maxima of R∗=0R^{*} = 0 and 0.250.25, which also revealed strong peaks at 2ft2f_t in the temporal spectral in Figure 4(c) and (d), exceeds dissipation rates seen at the same phase for the other, larger values of R∗R^{*}. The standard deviation in the volume integrals about ϵK\epsilon_K (plotted as error bars), which accounts for variability across tidal cycles, is relatively small — the maximum value is 6 W m−36\ \mathrm{W\,m}^{-3}, for (R∗,ϕ)=(1,π/4)(R^{*},\phi) = (1,\pi/4), which is only 6.67%6.67\% of the corresponding value of ϵK\epsilon_K.

Cycle-averaged ϵK\epsilon_K, denoted as ⟨ϵK⟩\langle\epsilon_K\rangle, provides an overall space-time metric and is tabulated in the first row of Table 2. These values decrease monotonically as R∗R^{*} decreases from 11 to 00. The maximum current speed, UM=Uc+UtU_M = U_c + U_t, decreases systematically with decreasing R∗R^{*} as well, and this contributor to enhanced dissipation needs to be separated out for disentangling the role of the tide from the mean current. An appropriate normalization of cycle-averaged ⟨ϵK⟩\langle\epsilon_K\rangle is the inviscid scale (U3/L)(U^3/L) of dissipation rate multiplied by a scale for the wake volume (hD2)(hD^2) and the reference density (ρ0)(\rho_0). Such normalization using U=UMU = U_M and L=DL = D leads to the fourth row of Table 2. The non-dimensionalized ⟨ϵK⟩\langle\epsilon_K\rangle increases as the tide becomes more dominant, which is the opposite of the trend in dimensional ϵK\epsilon_K. The increase is three times for the pure tide with respect to R∗=1R^{*} = 1. It is a remarkable consequence of the complex eddy-eddy and eddy-wake-topography interactions (described in previous sections) that come into play when the current reverses direction, a facet of tidally dominated flows at 3D topography.

QuantitiesR∗R^{*}1.000.750.500.250.00
⟨EK⟩\langle\mathcal{E}_{K}\rangle (kW)45.2432.0022.8318.5017.46
⟨F⟩\langle\mathcal{F}\rangle (kW)11.545.893.262.152.38
⟨F⟩/⟨EK⟩\langle\mathcal{F}\rangle/\langle\mathcal{E}_{K}\rangle (%)25.5018.4214.3011.6413.61
⟨EK⟩/ρ0UM3hD\langle\mathcal{E}_{K}\rangle/\rho_{0}U_{M}^{3}hD0.0750.0800.0900.1260.232
⟨EP⟩/(⟨EP⟩+⟨EK⟩)\langle\mathcal{E}_{P}\rangle/(\langle\mathcal{E}_{P}\rangle+\langle\mathcal{E}_{K}\rangle)0.1540.1600.1630.1650.181
⟨EP⟩/⟨EK⟩\langle\mathcal{E}_{P}\rangle/\langle\mathcal{E}_{K}\rangle0.1820.1910.1940.1970.221

Table 2. Tidal-cycle averages of various quantities, defined as ⟨ξ⟩=∫02πξ dϕ/2π\langle\xi\rangle=\int_{0}^{2\pi}\xi\,\mathrm{d}\phi/2\pi, for the five values of R∗R^{*} considered.

The internal wave flux F\mathcal{F} for R∗=0.75R^{*} = 0.75 and 11 in Figure 9(b) presents a peak shortly (≈75\approx75 minutes for an M2M_2 tide) after ϕ=π/2\phi= \pi/2 due to the persistence of the large amplitude lee waves generated when U∞=Uc+UtU_{\infty} = U_c + U_t. As was seen for ϵK\epsilon_K, the R∗≤0.5R^{*} \le0.5 cases with Ex−(R∗)<0\mathrm{Ex}^{-}(R^{*}) < 0 present more than one local maxima in F\mathcal{F}. From Table 2, the average of F\mathcal{F} over the tidal cycle also decreases from R∗=1R^{*} = 1 to 0.250.25, but increases slightly when R∗R^{*} becomes 00. As a result, the ratio of cycle-averaged F\mathcal{F} and ϵK\epsilon_K also decreases from 25.5%25.5\% at R∗=1R^{*} = 1 to 11.64%11.64\% at R∗=0.25R^{*} = 0.25, before increasing to 13.61%13.61\% at R∗=0R^{*} = 0 (Table 2). This is somewhat evident from Figure 9(c), wherein the F/ϵK\mathcal{F}/\epsilon_K values for R∗=0R^{*} = 0 are close to or greater than those for R∗=0.25R^{*} = 0.25. Additionally, F/ϵK<1\mathcal{F}/\epsilon_K < 1 throughout the tidal cycle (0≤ϕ<2π0 \le\phi< 2\pi), for all five values of R∗R^{*}.

Finally, to emulate observational measurements from a boat during field experiments, Figure 9(d) presents ϵKy0\epsilon_K^{y0}, obtained from dissipation rate measurements on a vertical transect passing through y/D=0y/D = 0. The integral over the transect area is multiplied by DD (assumed lateral scale for the wake) to enable direct comparison of ϵKy0\epsilon_K^{y0} and ϵK\epsilon_K. The ϵKy0\epsilon_K^{y0} curves are qualitatively similar to the ϵK\epsilon_K curves in Figure 9(a), but with the peaks offset further to the right. of ϕ=π/2\phi=\pi/2. This is because the vertical transect fails to sample the strong dissipation rates associated with the attached lateral shear layers, which are strongest when the current U∞U_\infty peaks. Additionally, the magnitude of EKy0\mathcal{E}_K^{y0} is weaker than EK\mathcal{E}_K at all tidal phases, for all values of R∗R^*, and on average, the ratio EK/EKy0≈3\mathcal{E}_K/\mathcal{E}_K^{y0}\approx3. The error bars are

also smaller, but the maximum value of 0.016 Wm2^2 kg−1^{-1} is 18% of the corresponding EKy0\mathcal{E}_K^{y0} value.

In the ocean, horizontal transects can also be sampled using autonomous gliders. The vertical variation of dissipation rates in the wake is nevertheless of interest, given the complex vortex dynamics observed in the wake (see Figure 2 and Figure 3, for example). The quantity EKz∗\mathcal{E}_K^{z^*}, which is the horizontal-area integral of E‾K\overline{\mathcal{E}}_{K} multiplied by hh and ρ0\rho_{0}, computed at three different z∗=z/h=0.25z^{*}=z/h=0.25, 0.5 and 0.75 are presented in Figure 10(a), (b) and (c), respectively.

Three plots of phase dependence of dissipation at different heights

Figure 10. Phase dependence of EKz∗\mathcal{E}_K^{z^*}, obtained using a horizontal-area integral, at three different heights z∗=z/hz^*=z/h: (a) 0.25, (b) 0.50, and (c) 0.75.

The EKz∗\mathcal{E}_{K}^{z^{*}} curves for all five R∗R^{*} are qualitatively similar to each other (and the bulk EK\mathcal{E}_{K} in Figure 9a), despite the vertical variability in wake-vortex dynamics. The magnitudes of EKz∗\mathcal{E}_{K}^{z^{*}} decrease monotonically with increasing z∗z^{*}, likely linked to the decrease in the horizontal size of the vortical structures with z∗z^{*}, but a clear scaling with the local diameter dz∗=(1−z∗)Dd_{z}^{*} = (1-z^{*})D is not found.

When investigating ocean mixing, the quantity of direct interest is the dissipation rate of the available potential energy per unit mass ϵP\epsilon_{P}. The spatial distribution of ϵP\epsilon_{P} is related to the distribution of ϵK\epsilon_{K} (c.f. Supplemental Fig. S3 and Figure 5). The ϵP\epsilon_{P} is maximized in regions where nearby fluid is ‘entrained’ and mixed, which usually lies close to the regions of strong shear, which in turn are regions of enhanced ϵK\epsilon_{K}. The volume integral of ϵP\epsilon_{P} in the wake region, EP\mathcal{E}_{P}, is presented as a function ϕ\phi in Figure 11(a). The EP\mathcal{E}_{P} curves are qualitatively similar to the EK\mathcal{E}_{K} in Figure 9(a), albeit with lower magnitudes and flatter peaks offset to the right of ϕ=π/4\phi= \pi/4. The error bars in EP\mathcal{E}_{P}, as a fraction of the corresponding EP\mathcal{E}_{P}, have larger magnitudes in Figure 11(a) compared to Figure 9(a).

Phase dependence of wake dissipation and mixing efficiency

Figure 11. Top row shows phase dependence of: (a) the wake dissipation of available potential energy per unit mass EP\mathcal{E}_{P}, and (b) the bulk mixing efficiency η\eta. Bottom row: phase dependence of the approximations based on a streamwise vertical transect (c) EPy0\mathcal{E}_{P}^{y0}, and (d) ηy0\eta^{y0}, of the quantities on the top. Standard deviations across tidal cycles are represented using error bars at each tidal phase. ϕ=π/4\phi=\pi/4 (maximum U∞U_{\infty}) and 3π/43\pi/4 (minimum U∞U_{\infty}) are shown using black dashed vertical lines and the zero-U∞U_{\infty} phase by lines in color (depends on R∗R^{*}).

The variation of the bulk mixing efficiency η\eta for the wake region, as a function of ϕ\phi, is presented in Figure 11(b). As the tidal component accelerates from zero, η\eta first decreases to its minimum, before increasing with ϕ\phi. This behavior is likely because EK\mathcal{E}_{K} initially increases faster than EP\mathcal{E}_{P} as the shear increases, before the shear instabilities grow to finite amplitude and start mixing the fluid, when EP\mathcal{E}_{P} takes over. The minimum in η\eta occurs earlier and becomes progressively smaller as R∗R^{*} is increased — the greater the Uc=UtR∗U_{c} = U_{t}R^{*}, the stronger the shear, and earlier the shear instabilities grow to finite amplitude. As the flow evolves and vortical structures become better developed, the rate of entrainment and mixing (EP\mathcal{E}_{P}) increases faster than EK\mathcal{E}_{K} for all R∗R^{*} until η\eta attains a local maximum. Interestingly, the maximum η\eta is attained near U∞=0U_{\infty} = 0 for all R∗R^{*} — this is because the flow features that result in the strongest ϵK\epsilon_{K}, i.e., the hydraulic jet and attached shear layers, are absent when U∞=0U_{\infty} = 0, while the existing turbulence continues to mix the buoyancy gradients created during the preceding energetic phase of the current. The η\eta curve decreases with ϕ\phi after the local maximum, before increasing again to a second maximum for R∗<1R^{*} < 1 at a phase ϕ>3π/2\phi> 3\pi/2, when U∞U_{\infty} again becomes zero as the current reverts to an eastward flow. Interestingly, a larger R∗R^{*} corresponds to a shorter time interval between the zero-crossings of U∞U_{\infty}, a greater second minimum of mixing efficiency η\eta, and a lesser first minimum of η\eta.

Within a tidal cycle, the value of η\eta varies between 0.1 and a maximum of approximately 0.25 for all five R∗R^{*}, which corresponds to standard deviations of 20–30% about the mean value ⟨η⟩\langle\eta\rangle for different R∗R^{*}. An average bulk mixing efficiency for the tidal cycle, defined as ⟨EP⟩/(⟨EP⟩+⟨EK⟩)\langle\mathcal{E}_{P}\rangle/ (\langle\mathcal{E}_{P}\rangle+ \langle\mathcal{E}_{K}\rangle), is presented in the fifth row of Table 2. This average η\eta follows the trend of the non-dimensionalized ⟨EK⟩\langle\mathcal{E}_{K}\rangle — it increases as the tide becomes more dominant, with a minimum of 0.154 for R∗=1R^{*} = 1, and a maximum of 0.181 for R∗=0R^{*} = 0. The corresponding values of an average mixing coefficient is presented in the last row, and the values all lie close to the widely used value of 0.2.

If only a vertical transect of measurements is available, the approximation EPy0\mathcal{E}_{P}^{y0} in Figure 11(c) is qualitatively similar to EP\mathcal{E}_{P}. For the available potential energy, the ratio EP/EPy0≈2\mathcal{E}_{P}/\mathcal{E}_{P}^{y0} \approx2 (averaged across R∗R^{*} and all ϕ\phi) is smaller than the ratio for the kinetic energy. The mixing efficiency calculated from the vertical transect ηy0\eta^{y0} in Figure 11(d), however, is markedly different from the bulk mixing efficiency η\eta. The ηy0\eta^{y0} magnitudes for all R∗R^{*} remain close to 0.26 for the majority of the tidal cycle, and the standard deviations for the five R∗R^{*} are only 5–17% about ⟨ηy0⟩\langle\eta^{y0}\rangle. The low values of η\eta seen during the eastward (same direction as mean) tide, especially as the flow accelerates, are not seen in the centerline-transect value ηy0\eta^{y0}. The error bars for ηy0\eta^{y0} at each tidal phase in Figure 11(d), however, are larger, indicating its values likely fall in the range 0.2≤ηy0≤0.30.2 \le\eta^{y0} \le0.3 for the majority of the tidal cycle, π/10≤ϕ≤3π/2\pi/10 \le\phi\le3\pi/2.

Discussion and Conclusions

Large eddy simulations (LESs) were performed to investigate the vortex dynamics, turbulence, dissipation, and mixing in the wake of tidally dominated currents past three-dimensional (3D) topography in a linearly stratified environment with weak background rotation. The barotropic flow is modeled as U∞(t)=Uc+Utsin⁡(2πftt)U_{\infty}(t) = U_{c} + U_{t}\sin(2\pi f_{t}t), where UcU_{c} is the mean current and UtU_{t} is the amplitude of the tidal oscillation, and the relative strength of the tidal flow is quantified using the ratio R∗=Uc/UtR^{*} = U_{c}/U_{t}. To investigate tidally dominated flows, R∗R^{*} was varied from 0 to 1 by varying UcU_{c} from 0 to UtU_{t}. The idealized 3D topography, a conical seamount with height hh and base diameter DD, is supercritical, and a hydraulic response was observed near the apex. The seamount is also dynamically tall, with the topographic Froude number Frh≤0.3Fr_{h} \le0.3 across R∗R^{*}, and lateral flow separation resulted in the generation of wake eddies.

Investigation of the flow evolution within a tidal cycle revealed lee waves emanating from the top of the seamount, a wake region below, and a thin hydraulic jet near the seamount at the interface between the two. The wake region consists of a turbulent recirculation region close to the seamount (‘near wake’) enclosed laterally within attached shear layers, and vertically coupled coherent vortices shed and advected away from the near wake (‘wake eddies/vortices’). In the R∗=1R^{*} = 1 simulation, a Strouhal pair of coherent vortices was shed during the first half of the tidal cycle, and a vortex pair was shed during the second. For a weaker mean current (R∗=0.5R^{*} = 0.5), only one coherent vortex was shed in the first half of the cycle before a vortex pair was generated during the latter half. The wake vortex dynamics in the absence of a mean current (R∗=0R^{*} = 0) was significantly different due to the purely oscillatory U∞U_{\infty}. In the LES, a coherent vortex is shed to the north of the topography during the first half of the tidal cycle. During reverse flow in the next half, the vortex is advected back towards the seamount, and its interaction with the recirculation region results in a vortex of the opposite polarity shed to the north on the other side of the seamount. Since the simulations are performed on an ff-plane, the consistent northward shedding is likely one state of a bistable system.

For reference, simulations performed for R∗=0R^* = 0 with U∞(t)=Utsin⁡(2πftt+π)U_\infty(t) = U_t \sin(2\pi f_t t + \pi), i.e., a phase difference of π\pi for the tidal oscillation, revealed consistent southward shedding (not shown). Systems are often bistable on timescales that are orders of magnitude longer than any convective timescale ([46]). Hence, the system presented here could potentially switch to southward shedding after a sufficiently long time interval, or due to the action of a sufficiently large-amplitude perturbation.

Spectral analysis of the streamwise velocity measured at a point within the near wake revealed a dominant peak at the tidal frequency ftf_t for all five R∗R^*. The dominant frequencies in the spanwise velocity spectra, however, were not all at ftf_t, suggesting tidal synchronization of the near-wake dynamics to frequencies other than ftf_t, as observed for the tidally modulated cases (R∗>1R^* > 1) in Puthan et al. [35]. The horizontal-area-integrated vertical vorticity spectra reflected the tidal synchronization of wake vortices: R∗=1R^* = 1, 0.75, and 0.5 had dominant peaks at ftf_t, whereas the two tidally most dominant cases, R∗=0R^* = 0 and 0.25, had them at 2ft2f_t. The spectra of horizontal-area-integrated dissipation rates of kinetic energy per unit mass ϵK\epsilon_K were qualitatively similar to the vorticity spectra.

The wake dissipation exhibited large values (ϵK∼10−6.5 W kg−1\epsilon_K \sim10^{-6.5}\ \mathrm{W\ kg^{-1}}) at the coherent vortex peripheries across R∗R^*. The dissipation rate field is spread out and patches with ϵK∼10−6.5 W kg−1\epsilon_K \sim10^{-6.5}\ \mathrm{W\ kg^{-1}} are encountered laterally up to 2.5D2.5D from the topography owing to vortex-vortex interactions. The lateral spread is larger for a purely tidal flow. Owing to the mean advection, patches with ϵK∼10−6.5 W kg−1\epsilon_K \sim10^{-6.5}\ \mathrm{W\ kg^{-1}} are found far downstream, e.g., 6D6D downstream of the seamount. Stronger dissipation (approximately an order of magnitude greater) was observed in the near wake with x/D<3x/D < 3 in all cases, and the largest values (ϵK∼10−5 W kg−1\epsilon_K \sim10^{-5}\ \mathrm{W\ kg^{-1}}) were consistently observed in the hydraulic jet and attached shear layers. The largest ϵK\epsilon_K regions, being highly localized, are difficult to sample in observational transects and present a challenge when obtaining field estimates of wake dissipation.

The volume-integrated wake dissipation EK\mathcal{E}_K peaked shortly after maximum U∞U_\infty (tidal phase π/2\pi/2) for all five R∗R^*, and was consistently greater than the internal wave flux — by a factor of about 4 at R∗=1R^* = 1 to about 7 at R∗=0R^* = 0. Thus, the local dissipation in this 3D topographic wake was substantially greater than the energy that is transported away to potentially be dissipated elsewhere. The magnitudes of the horizontal-area integrals of dissipation rates within the wake decreased with increasing height above the bottom, but their tidal phasing remained qualitatively similar throughout the tidal cycle. The tidal-cycle-averaged dissipation rate magnitudes decreased with R∗R^* in the simulations, but their normalized values were the greatest for tidally the most dominant cases. This is an important overall consequence of the complex eddy-eddy-wake-topography interactions that occur as the flow reverses direction.

The tidal phasing of the dissipation rate of available potential energy in the wake EP\mathcal{E}_P shared some similarities with EK\mathcal{E}_K. The magnitudes of EP\mathcal{E}_P attained a maximum after peak U∞U_\infty, but the peaks were offset further from phase π/2\pi/2 compared to EK\mathcal{E}_K. The bulk mixing efficiency η\eta varied between 0.1 and 0.25 within a tidal cycle. Interestingly, η\eta was maximized near U∞=0U_\infty= 0, for all R∗R^*. This is because the hydraulic jet and attached shear layers, which contribute significantly to ϵK\epsilon_K, are absent when U∞=0U_\infty= 0, while the existing turbulence continues to mix the buoyancy gradients created during the preceding energetic phase of the current.

Observationally, it is often easier to obtain vertical transects of measurements from a boat, compared to full three-dimensional fields. Analysis of the simulation data revealed that bulk mixing efficiencies obtained using measurements from a vertical transect alone differ from the bulk values. However, the approximations of volume-integrated dissipation rates of kinetic energy and available potential energy, obtained from a vertical transect passing through y/D=0y/D = 0, were qualitatively similar to the bulk values. Quantitatively, these approximations differed by a factor of 2–3 on average throughout the tidal cycle, for all R∗R^*. A closer investigation revealed that the vertical shear components S13S_{13} and S23S_{23} are the major contributors to ϵK\epsilon_K. In vertical transects, S13S_{13} and S23S_{23} resulted in 89% or more of the phase-averaged dissipation rates of kinetic energy at all tidal phases, across the five R∗R^* considered (not shown).

When drawing parallels between observations and the results presented here, it is important to remember that the non-dimensional groups ReRe, RoRo, and FrhFr_h presented are all defined using the velocity at peak tidal flow, UMU_M. These metrics should be associated with the bulk dynamics over an entire tidal cycle, and not the local dynamics at an instant. The latter depends on the instantaneous U∞U_\infty, which can lie in the range 0≤∣U∞∣≤UM0 \leq|U_\infty| \leq U_M for the tidally dominated cases investigated here. Other non-dimensional groups like the gradient Richardson number, vorticity Rossby number, and the angle ϕRiB\phi_{RiB} related to the potential vorticity introduced by Thomas et al. [42], for example, are pertinent to local investigations of fluid instabilities. A closer investigation of the instabilities that lead to turbulence in the near wake, coherent eddies, hydraulic jet, and attached shear layers is of future interest.

Additionally, drawing inferences about the bulk, based on the instantaneous values of non-dimensional parameters alone, is inappropriate. For example, (R∗,ϕ)=(0,π)(R^*,\phi) = (0,\pi) and (R∗,ϕ)=(1,3π/2)(R^*,\phi) = (1,3\pi/2) have the same values for ReRe, RoRo, FrhFr_h defined using instantaneous U∞U_\infty, but correspond to very different wakes, and hence cannot be compared. The wakes are qualitatively different for the same values of instantaneous non-dimensional groups for the same R∗R^{*} — (R∗,ϕR^{*},\phi) = (1, π\pi) and (R∗,ϕR^{*},\phi) = (1, 2π\pi), for example — and hence, one needs to account for the tidal phasing of the flow.

Finally, the idealized large eddy simulations presented here considered a barotropic inflow in a linearly stratified environment. The barotropic tide is modeled as a sinusoidal wave oscillating at a characteristic tidal frequency ftf_t, with the ratio R∗R^{*} determining whether the flow changes direction within a tidal cycle. While this is a reasonable approximation for diurnal and semi-diurnal tides, it is not suitable for mixed tides. The spring-neap modulation of tidal flows can also have significant effects on the observed wake dynamics and mixing. A more realistic baroclinic current in a non-linear stratification can result in modified wake dynamics, depending on the pycnocline thickness and its relative position vertically with respect to the topography, for example, and requires further investigation.

Acknowledgments. This work was supported by the ONR grant N00014-23-1-2170, and high-performance computer time and resources from the DoD High Performance Computing Modernization Program. The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability statement. The datasets generated and/or presented can be accessed at the following DOI: 10.5281/zenodo.15391881.

References

  1. [1]Alford, M. H., J. M. Klymak, and G. S. Carter, 2014: Breaking internal lee waves at kaena ridge, hawaii. Geophysical Research Letters, 41 (3), 906–912.DOI
  2. [2]Baines, P. G., 2022: Topographic effects in stratified flows. Cambridge university press.
  3. [3]Bewley, T. R., 2007: Numerical Renaissance: simulation, optimization and control. Renaissance Press.
  4. [4]Black, K., and S. Gay, 1987: Eddy formation in unsteady flows. Journal of Geophysical Research: Oceans, 92 (C9), 9514–9522.DOI
  5. [5]Castro, I., W. Snyder, and G. Marsh, 1983: Stratified flow over three-dimensional ridges. Journal of Fluid Mechanics, 135, 261–282.DOI
  6. [6]Caulfield, C., 2021: Layering, instabilities, and mixing in turbulent stratified flows. Annual Review of Fluid Mechanics, 53 (1), 113–145.
  7. [7]Chang, M.-H., S. Jan, C.-L. Liu, Y.-H. Cheng, and V. Mensah, 2019: Observations of island wakes at high rossby numbers: Evolution of submesoscale vortices and free shear layers. Journal of Physical Oceanography, 49 (11), 2997–3016.
  8. [8]Chang, M.-H., T. Y. Tang, C.-R. Ho, and S.-Y. Chao, 2013: Kuroshio-induced wake in the lee of green island off taiwan. Journal of Geophysical Research: Oceans, 118 (3), 1508–1519.
  9. [9]Dale, A. C., and M. E. Inall, 2015: Tidal mixing processes amid small-scale, deep-ocean topography. Geophysical Research Letters, 42 (2), 484–491.
  10. [11]Fletcher, C. A. J., 1991: Computational Techniques for Fluid Dynamics. 2nd ed., Springer.
  11. [12]Gayen, B., and S. Sarkar, 2011a: Boundary mixing by density overturns in an internal tidal beam. Geophysical research letters, 38 (14).
  12. [13]Gayen, B., and S. Sarkar, 2011b: Direct and large-eddy simulations of internal tide generation at a near-critical slope. Journal of Fluid Mechanics, 681, 48–79.
  13. [14]Gola, D., S. Nidhan, and S. Sarkar, 2024: Internal gravity waves in flow past a bluff body under different levels of stratification. Journal of Fluid Mechanics, 999, A23.
  14. [15]Hunt, J., and W. Snyder, 1980: Experiments on stably and neutrally stratified flow over a model three-dimensional hill. Journal of Fluid Mechanics, 96 (4), 671–704.
  15. [16]Jagannathan, A., K. Srinivasan, J. C. McWilliams, M. J. Molemaker, and A. L. Stewart, 2021: Boundary-layer-mediated vorticity generation in currents over sloping bathymetry. Journal of Physical Oceanography, 51 (6), 1757–1778.
  16. [17]Johnston, T. S., and Coauthors, 2019: Energy and momentum lost to wake eddies and lee waves generated by the north equatorial current and tidal flows at peleliu, palau. Oceanography, 32 (4), 110–125.
  17. [18]Kang, D., and O. Fringer, 2010: On the calculation of available potential energy in internal wave fields. Journal of Physical Oceanography, 40 (11), 2539–2545.
  18. [19]Kunze, E., 2017: Internal-wave-driven mixing: Global geography and budgets. Journal of Physical Oceanography, 47 (6), 1325–1345.DOI
  19. [20]Legg, S., and J. Klymak, 2008: Internal hydraulic jumps and overturning generated by tidal flow over a tall steep ridge. Journal of Physical Oceanography, 38 (9), 1949–1964.
  20. [21]Liu, J., P. Puthan, and S. Sarkar, 2024: Effect of rotation on wake vortices in stratified flow. Journal of Fluid Mechanics, 999, A44.
  21. [23]MacKinnon, J., L. St Laurent, and A. C. N. Garabato, 2013: Diapycnal mixing processes in the ocean interior. International Geophysics, Vol. 103, Elsevier, 159–183.
  22. [24]MacKinnon, J. A., M. H. Alford, G. Voet, K. L. Zeiden, T. Shaun Johnston, M. Siegelman, S. Merrifield, and M. Merrifield, 2017: Eddy wake generation from broadband currents near palau. Journal of Geophysical Research: Oceans, 124 (7), 4891–4903.
  23. [25]Munk, W., and C. Wunsch, 1998: Abyssal recipes ii: Energetics of tidal and wind mixing. Deep Sea Research Part I: Oceanographic Research Papers, 45 (12), 1977–2010.
  24. [26]Munk, W. H., 1966: Abyssal recipes. Deep Sea research and oceanographic abstracts, Elsevier, Vol. 13, 707–730.DOI
  25. [27]Nicoud, F., and F. Ducros, 1999: Subgrid-scale stress modelling based on the square of the velocity gradient tensor. Flow, turbulence and Combustion, 62 (3), 183–200.
  26. [28]Osborn, T. R., 1980: Estimates of the local rate of vertical diffusion from dissipation measurements. Journal of physical oceanography, 10 (1), 83–89.
  27. [29]Pawlak, G., P. MacCready, and R. McCabe, 2003: Evolution of vortical flow structure in an ocean-boundary process. Near-Boundary Processes and their Parameterization. Proc. Aha Huliko ‘a Hawaiian Winter Workshop, Citeseer, 131–142.
  28. [30]Perfect, B., N. Kumar, and J. Riley, 2018: Vortex structures in the wake of an idealized seamount in rotating, stratified flow. Geophysical Research Letters, 45 (17), 9098–9105.
  29. [31]Perfect, B., N. Kumar, and J. Riley, 2020a: Energetics of seamount wakes. part i: Energy exchange. Journal of Physical Oceanography, 50 (5), 1365–1382.
  30. [32]Perfect, B., N. Kumar, and J. Riley, 2020b: Energetics of seamount wakes. part ii: Wave fluxes. Journal of Physical Oceanography, 50 (5), 1383–1398.
  31. [33]Puthan, P., M. Jalali, V. K. Chalamalla, and S. Sarkar, 2019: Energetics and mixing in buoyancy-driven near-bottom stratified flow. Journal of Fluid Mechanics, 869, 214–237.
  32. [34]Puthan, P., M. Jalali, J. L. Ortiz-Tarin, K. Chongsiripinyo, G. Pawlak, and S. Sarkar, 2020: The wake of a three-dimensional underwater obstacle: Effect of bottom boundary conditions. Ocean Modelling, 149, 101 611.
  33. [35]Puthan, P., G. Pawlak, and S. Sarkar, 2022: Wake vortices and dissipation in a tidally modulated flow past a three-dimensional topography. Journal of Geophysical Research: Oceans, 127 (8), e2022JC018 470.
  34. [36]Puthan, P., S. Sarkar, and G. Pawlak, 2021: Tidal synchronization of lee vortices in geophysical wakes. Geophysical Research Letters, 48 (4), e2020GL090 905.
  35. [37]Sarkar, S., and A. Scotti, 2017: From topographic internal gravity waves to turbulence. Annual Review of Fluid Mechanics, 49 (1), 195–220.
  36. [38]Scotti, A., and B. White, 2014: Diagnosing mixing in stratified turbulent flows with a locally defined available potential energy. Journal of fluid mechanics, 740, 114–135.
  37. [39]Srinivasan, K., J. C. McWilliams, and A. Jagannathan, 2021: High vertical shear and dissipation in equatorial topographic wakes. Journal of Physical Oceanography, 51 (6), 1985–2001.
  38. [40]St. Laurent, L., T. Ijichi, S. T. Merrifield, J. Shapiro, and H. L. Simmons, 2019: Turbulence and vorticity in the wake of palau. Oceanography, 32 (4), 102–109.
  39. [41]Taylor, J. R., and S. Sarkar, 2007: Internal gravity waves generated by a turbulent bottom Ekman layer. J. Fluid Mech., 590 (1), 331–354.
  40. [42]Thomas, L. N., J. R. Taylor, R. Ferrari, and T. M. Joyce, 2013: Symmetric instability in the gulf stream. Deep Sea Research Part II: Topical Studies in Oceanography, 91, 96–110.
  41. [43]Thorpe, S., 2004: Recent developments in the study of ocean turbulence. Annu. Rev. Earth Planet. Sci., 32 (1), 91–109.
  42. [44]Voet, G., M. H. Alford, J. A. MacKinnon, and J. D. Nash, 2020: Topographic form drag on tides and low-frequency flow: Observations of nonlinear lee waves over a tall submarine ridge near palau. Journal of Physical Oceanography, 50 (5), 1489–1507.
  43. [45]Waterhouse, A. F., and Coauthors, 2014: Global patterns of diapycnal mixing from measurements of the turbulent dissipation rate. Journal of Physical Oceanography, 44 (7), 1854–1872.
  44. [46]Williams, O. J., and A. J. Smits, 2025: Asymmetries in nominally symmetric flows. Annual Review of Fluid Mechanics, 57.DOI
  45. [47]Williamson, C. H., and G. Brown, 1998: A series in 1/√Re to represent the strouhal–reynolds number relationship of the cylinder wake. Journal of Fluids and Structures, 12 (8), 1073–1085.
  46. [48]Winters, K. B., and L. Armi, 2013: The response of a continuously stratified fluid to an oscillating flow past an obstacle. Journal of Fluid Mechanics, 727, 83–118.
  47. [49]Winters, K. B., and L. Armi, 2014: Topographic control of stratified flows: Upstream jets, blocking and isolating layers. Journal of fluid mechanics, 753, 80–103.
  48. [50]Winters, K. B., P. N. Lombard, J. J. Riley, and E. A. D’Asaro, 1995: Available potential energy and mixing in density-stratified fluids. Journal of Fluid Mechanics, 289, 115–128.
  49. [51]Wunsch, C., and R. Ferrari, 2004: Vertical mixing, energy, and the general circulation of the oceans. Annu. Rev. Fluid Mech., 36 (1), 281–314.
  50. [52]Wynne-Cattanach, B. L., M. H. Alford, J. A. MacKinnon, and G. Voet, 2022: Measurements of turbulence generated by wake eddies near a steep headland. Journal of Geophysical Research: Oceans, 127 (8), e2022JC018 674.
  51. [53]Yang, J., and E. Balaras, 2006: An embedded-boundary formulation for large-eddy simulation of turbulent flows interacting with moving boundaries. J. Comput. Phys., 215, 12–40.
  52. [54]Zeiden, K. L., D. L. Rudnick, and J. A. MacKinnon, 2019: Glider observations of a mesoscale oceanic island wake. Journal of Physical Oceanography, 49 (9), 2217–2235.

Paper details

vlnc:W_r4anxmzf
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents