MAY 2022 RAJA ET AL.

823

Near-Inertial Wave Energetics Modulated by Background Flows in

a Global Model Simulation

KESHAV J. RAJA,a MAARTEN C. BUIJSMAN,a JAY F. SHRIVER,b BRIAN K. ARBIC,c AND OLADEJI SIYANBOLAa
a
School of Ocean Science and Engineering, University of Southern Mississippi, Stennis Space Center, Mississippi
b
Ocean Dynamics and Prediction Branch, Naval Research Laboratory, Stennis Space Center, Mississippi
c
Department of Earth and Environmental Sciences, University of Michigan, Ann Arbor, Michigan

(Manuscript received 15 June 2021, in final form 23 December 2021)

ABSTRACT: We study the generation, propagation, and dissipation of wind-generated near-inertial waves (NIWs) in a
global 1/258 Hybrid Coordinate Ocean Model (HYCOM) simulation with realistic atmospheric forcing and background cir-
culation during 30 days in May–June 2019. The time-mean near-inertial wind power input and depth-integrated energy bal-
ance terms are computed for the total fields and the fields decomposed into vertical modes to differentiate between the
radiative and (locally) dissipative components of NIW energy. Only 30.3% of the near-inertial wind input projects onto the
first five modes, whereas the sum of the NIW energy in the first five modes adds up to 58% of the total NIW energy.
Almost all of the depth-integrated NIW horizontal energy flux projects on the first five modes. The global distribution of
dissipation and decay distances of NIW modes confirm that lower latitudes are a sink for NIW energy generated at higher
latitudes. The locally dissipated fraction of NIW energy qlocal is found to be uniform throughout the global ocean, with a
global mean value of 0.79. The horizontal NIW fluxes diverge from areas with cyclonic vorticity and converge in areas with
anticyclonic vorticity; that is, anticyclonic eddies are a sink for NIW energy fluxes}in particular, for higher modes. Most of
the residual energy that does not project onto modes propagates downward in anticyclonic eddies. The global near-inertial
wind power input is 0.21 TW for the 30 days, of which only 19% is transmitted below 500-m depth.
KEYWORDS: Inertia–gravity waves; Internal waves; Ocean models

1. Introduction et al. 2005; Rimac et al. 2013). Thus, the understanding of

NIW energy budget and how different near-surface processes
Diapycnal mixing in the ocean interior plays an important
interact with NIWs are essential to accurately estimate the
role in supporting the upwelling part of the meridional over-
role of NIWs in driving deep ocean mixing.
turning circulation (MOC). It is estimated that about 2 6 0.6
The wind power input to surface near-inertial motions has
TW (1 TW = 1012 W) of power is required to maintain the
been estimated by numerous studies. Early studies have all
MOC through ocean mixing (Munk and Wunsch 1998; Kunze
used the slab mixed layer model of Pollard and Millard (1970)
2017). Internal wave breaking is the dominant source of dia-
and computed the global annual-mean wind power input to
pycnal mixing in most of the ocean interior (Munk and
surface near-inertial motions ranging from 0.3 to 1.1 TW
Wunsch 1998; MacKinnon et al. 2017). Deep ocean internal
(Alford 2001; Watanabe and Hibiya 2002; Alford 2003)
waves have primarily 3 sources: internal tides generated by
depending on the spatiotemporal resolution of the wind prod-
tide–topography interaction at 1 TW of power (Egbert and
uct used (Jiang et al. 2005; Rimac et al. 2013). The slab mod-
Ray 2001; MacKinnon et al. 2017; Vic et al. 2019; de Lavergne
els, however, are prone to overestimate the wind work as they
et al. 2019; Buijsman et al. 2020), wind-generated near-inertial
do not account for the energy loss due to vertical shear insta-
motions at 0.16–0.50 TW of power (Alford 2003; Furuichi et al.
2008; Flexas et al. 2019; Alford 2020b), and topography-gen- bility, mixed layer deepening or the radiation into the ocean
erated internal lee waves at 0.06–0.51 TW of power (Scott interior (Plueddemann and Farrar 2006; Furuichi et al. 2008).
et al. 2011; Nikurashin and Ferrari 2011; Trossman et al. 2013; Alford (2020b) reviews the various definitions of near-inertial
Shakespeare 2020). The contribution from wind-generated wind work using slab models. He examines the mixed layer
near-inertial waves (NIWs) to internal wave power input has potential energy increase and compares it with the estimate
a large uncertainty. We still do not fully understand the from Argo and ERA-Interim heat flux climatology. He finds
details of NIW dynamics and how they interact with other that more than 50% of near-inertial surface wind power input
processes in the ocean. A part of the uncertainty is also associ- is lost to turbulence in the upper layers and is not available to
ated with the spatial and temporal resolution of global wind propagating NIWs.
products, in particular at high latitudes (Kunze 2017; Jiang With regard to global ocean simulations, Furuichi et al.
(2008) used a full three-dimensional primitive equation model
for a global ocean with 1/78 resolution (in longitude) forced
by 6-hourly wind stress and estimated the annual-mean
Denotes content that is immediately available upon publica- surface near-inertial wind power input to be 0.4 TW.
tion as open access.
Recently, Flexas et al. (2019) performed a high-resolution
(1/488) realistic global ocean simulation with tidal forcing to
Corresponding author: Keshav J. Raja, [email protected] estimate wind work on the oceans. They used 6-hourly winds

DOI: 10.1175/JPO-D-21-0130.1
Ó 2022 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright
Policy (www.ametsoc.org/PUBSReuseLicenses).
Unauthenticated | Downloaded 02/24/24 04:24 PM UTC
824 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 52

with 0.148 resolution and found 0.16 TW of annual mean sur- resolutions of 0.18 (resolving most of the mesoscale eddies)
face near-inertial wind power input. They concluded that and 0.48 (hardly resolving any eddies) and found enhanced
wind variability on time scales less than 6 h and spatial scales NIW kinetic energy (KE) in the interior of the ocean in the
less than 15 km is crucial to accurately represent the wind simulation with eddies. However, Rimac (2014) did not con-
power input in ocean models. sider a detailed energy analysis, nor did she correlate the
The wind-generated near-inertial motions in the surface energy of near-inertial modes with vorticity.
mixed layer have lateral variations due to vorticity gradients In this paper, we study the depth-integrated energetics of
leading to convergences and divergences that cause vertical NIWs in realistically forced global ocean simulations with the
motions at the base of the mixed layer. These vertical motions Hybrid Coordinate Ocean Model (HYCOM). We determine
generate NIWs in the ocean interior (Gill 1984; D’Asaro et al. what fraction of total near-inertial wind input, energy, and
1995). Gill (1984) developed a modal formalism to describe dissipation can be attributed to the resolved vertical modes.
these motions in the ocean interior, which take the form of We examine the global spatial patterns of the total, modal,
standing vertical oscillations, called normal modes. The hori- and residual NIW energy terms and study how these energy
zontal propagation of NIWs from their generation sites to far terms are modulated by the background vorticity. Such com-
away locations can be better understood using the method of prehensive depth-integrated NIW global energy balance has
vertical normal modes (Simmons and Alford 2012; Kelly not yet been done. Our simulations have a horizontal resolu-
2019). Observational data from the Ocean Storms Experi- tion of 1/258 and a 3-hourly wind forcing, which is an improve-
ment (D’Asaro et al. 1995) and moorings (Alford 2003) show ment over the global ocean simulations mentioned above.
that NIW energy can radiate long distances equatorward Previous model studies either have a coarser resolution or
away from its generation site as mode 1 and mode 2 waves lower-frequency wind forcing, or both. Moreover, the advan-
before dissipating. Simmons and Alford (2012) used a tage of using the isopycnal coordinate HYCOM model,
1/88 fully eddying isopycnal global ocean model to study long instead of fixed vertical coordinates, is that the effect of
range propagation of NIWs. Their model was forced by numerical diffusion is minimized (Halliwell 2004). This is ben-
6-hourly winds. They found that as much as 3%–16% of NIW eficial for studying the long-range propagation of NIWs.
energy radiates out of its generation sites and equatorward as HYCOM has been extensively validated and optimized for
low-mode internal gravity waves. Since Alford et al. (2012),
surface and internal tidal waves (Arbic et al. 2012; Ansong
there has been little effort to understand the radiative versus
et al. 2015; Arbic et al. 2018; Buijsman et al. 2016, 2020).
(locally) dissipative partitioning of NIW energy in global
However, we do not yet know how much wind energy is trans-
ocean simulations.
ferred to near-inertial gravity waves, how they propagate, and
The transfer of wind power from the ocean surface to the
where these waves dissipate in global HYCOM simulations.
interior is facilitated by several processes. The horizontal
In the next section, we describe the HYCOM model config-
scales of surface near-inertial motions can be of the order of
uration and our analysis methods. In section 3, we present the
1000 km corresponding to the wind event that generated
results of the depth-integrated near-inertial energy budget of
them. Since the vertical group velocity of near-inertial waves
the undecomposed fields and the fields decomposed into verti-
is proportional to their horizontal wavenumber, these waves
cal modes. We discuss global maps of the near-inertial wind
cannot propagate rapidly in the vertical direction unless their
input, the NIW energy, the horizontal energy fluxes, and the
horizontal scales are reduced (Gill 1984). One of the pro-
dissipation. Moreover, we analyze the influence of background
cesses that can reduce the horizontal scales of near-inertial
vorticity on the NIW energetics of the total, modal, and resid-
motions is the b-dispersion effect, whereby near-inertial
ual fields. We also examine the vertical transmission of NIW
motions are progressively dephased as a result of the plane-
tary vorticity gradient, resulting in smaller horizontal wave- energy. We discuss the implication of our results and examine
lengths and faster vertical propagation. This process has been the limitations of our methods in section 4. Our most relevant
observed by D’Asaro et al. (1995). findings are summarized in the conclusions in section 5.
Another process that has been hypothesized to shift the
local inertial frequency and modulate the propagation of 2. Methods
NIWs is the refraction of waves by the mesoscale vorticity a. HYCOM
in the ocean (Kunze 1985; Young and Jelloul 1997; Lee and
Niiler 1998; Asselin and Young 2020). An effective Coriolis We use the HYCOM (Metzger et al. 2014), which is the
parameter can be defined by feff = f 1 z/2, where f is the operational forecast model of the U.S. Navy. The model uses
Coriolis parameter and z is the relative vorticity (Mooers a hybrid vertical coordinate system, which is isopycnal in the
1975). This frequency shift can lead to the trapping of near- deep ocean and employs z coordinates to resolve the surface
inertial energy in regions of negative relative vorticity (Lee mixed layer. The model coordinates transition to terrain fol-
and Niiler 1998). Since the gradients of relative vorticity are lowing in shallow water. The simulation runs on a tripolar
much larger than the b effect, the frequency shift to feff grid at 1/258 horizontal resolution and 41 layers in the vertical
might be a stronger effect in accelerating vertical propaga- direction (approximately one-half of the levels are above
tion of near-inertial energy locally. Rimac (2014) examined 250 m, most of them 8 m apart in the open ocean). A K-profile
the influence of relative vorticity in NIW energy transmis- parameterization is used as the subgrid-scale mixing model in
sion using two global ocean simulations with horizontal the simulation. We analyze a forward (non–data assimilative)

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

MAY 2022 RAJA ET AL. 825

simulation without tidal forcing, also referred to as EXPT that project on both baroclinic and barotropic modes. How-
19.2 in HYCOM terminology. The simulation has realistic ever, we verify that the wind input to barotropic near-inertial
atmospheric forcing from the Navy Global Environmental motions is negligible relative to the wind input to baroclinic
Model (NAVGEM) (Hogan et al. 2014) with 60 atmo- near-inertial motions.
spheric levels over a height of 19 km, 0.178 horizontal reso- The depth-integrated near-inertial energy E is the sum of
lution, and 3-hourly wind forcing. The output variables are the depth-integrated kinetic energy
stored every hour. In this paper, we diagnose NIW energet-
ics from 20 May to 19 June 2019 (30 days), which coincides 0
1
with the cruises in the North Atlantic Ocean as part of the KE
r0 |u|2 dz; (2)
2 2H
Near Inertial Shear and Kinetic Energy in the North Atlantic
experiment (NISKINe). In the appendix, we also examine a and the depth-integrated available potential energy (Kang
simulation with astronomical tidal forcing, EXPT 19.0, which and Fringer 2010)
is a twin of EXPT 19.2. EXPT 19.0 is forced with the M2, S2, 0
K1, O1, and N2 tidal constituents (Buijsman et al. 2020). In the 1 b2
APE
r0 2
dz; (3)
simulation with tides, the parametric subharmonic instability 2 2H N
(PSI) of internal tides generates near-inertial motions (Haze-
winkel and Winters 2011; Ansong et al. 2018), which we do where H is the seafloor depth, b = gr/r0 is the near-inertial
not account for in our NIW energy equations. Additionally, buoyancy, and N is the buoyancy frequency.
the nonstationary diurnal internal tides cannot be fully We calculate the depth integrated NIW energy fluxes as
removed near the critical latitudes in the simulation with tides.
0
Hence, we choose to use the simulation without tides for our
FH
pu dz; (4)
analysis of NIW energetics. Both EXPT 19.0 and 19.2 are for- 2H
ward runs initiated from EXPT 21.6, a global HYCOM simu-
lation with data assimilation (DA) and tides. The tidal forcing where the near-inertial pressure perturbations p are com-
and DA were turned off on 1 April 2019 (50 days before the puted using the density anomaly r, as in Nash et al. (2005).
analysis period), providing enough time for the internal tides The depth-integrated energy flux due to self-advection of
and disturbances related to DA to dissipate. NIW energy is calculated as

b. NIW energetics 0  
1 1 b2
Fa
u r0 |u|2 1 r0 2 dz: (5)
The near-inertial fields are obtained from the model output 2H 2 2 N
after removing any remaining stationary tides using a har-
monic analysis and by applying a bandpass filter with fre- We find that global area-averaged variance of ­E=­t and
quency limits 0.8f–1/13.21 h equatorward of 568 and 0.8f–1.2f = · Fa are negligible (,1%) as compared with the variance of
poleward of 568. This filter is applied to both our simulations = · FH . Therefore, we do not present maps of ­E=­t and Fa in
with and without tides. It is designed in such a way that semi- this paper. The residual term D in Eq. (1) is dominated by vis-
diurnal internal tides that are generated poleward of 568 and cous dissipation of NIW energy and energy lost to wave drag
propagate equatorward are removed. Moreover, with this fil- (Buijsman et al. 2016) but also includes energy transfers
ter, we capture NIWs that have propagated a large distance between frequency bands through wave–wave and wave–
from their generation sites. mean flow interactions, topographic scattering, and other non-
We use the filtered near-inertial fields to calculate the time- linear processes (Polzin and Lvov 2011; Le Boyer and Alford
mean, depth-integrated near-inertial energy equation (Alford 2021). We assume that these processes lead to motions that
and Zhao 2007a; Zhai et al. 2009): dissipate locally.
The three-dimensional (3D) HYCOM fields are also
­E
 decomposed into vertical normal modes and the NIW energy
W
1 = · FH 1 Fa 1 D; (1)
­t balance for these modes is computed. We largely follow the
modal decomposition presented in Kelly (2016) and Buijsman
where the overbar indicates time-average over 30 days, W is
et al. (2020). However, in the present analysis we decompose
the near-inertial surface wind power input, E is the depth-
the baroclinic fields into vertical normal modes in the
integrated near-inertial energy, FH is the depth-integrated
horizontal NIW energy flux, Fa is the energy flux due to the HYCOM layer space itself (as opposed to Buijsman et al.
self-advection of NIW energy and D is the depth-integrated (2020), where the fields are interpolated onto an equidistant z
internal wave dissipation, considered to be a residual term. grid). The horizontal resolution of our model (1/258; 4 km
The wind power input to surface near-inertial motions is near the equator) allows the first five vertical modes to be
calculated as W = t · u(z = 0), where t is the near-inertial sur- resolved in most of the ocean (Buijsman et al. 2020).
face wind stress and u(z = 0) is the horizontal baroclinic near- The hydrostatic Sturm-Liouville equation is solved for non-
inertial velocity vector at the surface. The total wind stress is equidistant layers following Singh and Bhadauria (2009),
computed during the simulation. The near-inertial surface using 30-day mean profiles of buoyancy frequency in each
wind stress can inject power into mixed layer inertial motions horizontal grid cell. The Sturm-Liouville equation reads

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

826 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 52

­2 Wn (z) N 2 (z) We note that the group speed of a vertical mode from linear
1 2 Wn (z)
0, (6)
­z2 cn theory, cg,n = (v2 2 f2)/(knv), is difficult to compute because
it is impractical to determine the generation frequency v of
where Wn (z) is the eigenfunction of mode n, z is the vertical each remote NIW. The cg,n at a location, calculated from
coordinate, and N(z) is the buoyancy frequency. The eigen- Eq. (11) is, in a sense, an average group velocity of all the
functions obey a rigid lid boundary condition at surface wave groups (with different v) present at the location.
 (Wn
0 at z = 0, 2H). The eigenspeed
and bottom Furthermore, we estimate the time scale for the decay of a
cn
v2 2 f 2 =kn , where v is the wave frequency and k n is propagating NIW mode by Td,n = En/Dn. Hence, the decay
the horizontal wavenumber of mode n. The horizontal length scale for each near-inertial mode is
velocity eigenfunction is computed as
Ld
cg,n Td,n : (12)
­Wn (z)
U n (z)
: (7)
­z The zonally averaged Ld helps us to identify the presence
of remotely generated NIWs and quantify the distance trav-
The horizontal velocity eigenfunctions U n are projected
eled by them at each latitude.
onto the vertical profiles of the horizontal baroclinic velocities
and perturbation pressures at every time step to yield the
modal amplitudes in each horizontal grid cell; for example, 3. Results
for horizontal velocity, a. Near-inertial wind power input
0
1 The wind stress acting on the ocean surface excites motions
û n (t)
U n (z)u(z, t) dz; (8)
H 2H at the base of the mixed layer that transfer the wind power to
NIWs. The 30-day mean, near-inertial wind power input to
where û n is the modal amplitude of the nth mode along the x the total (undecomposed) fields, to modes 1–5, and to the
axis. sum of five modes are shown in Fig. 1. The wind input to the
After obtaining the modal amplitudes from the unfiltered undecomposed fields is enhanced in regions of high mesoscale
fields, we use a harmonic analysis and a bandpass filter as energy such as in the Southern Ocean and western boundary
explained earlier to remove any remaining internal tides and current regions. Regions of high mesoscale activity have high
to isolate the near-inertial field. We then compute the terms surface wind stress and extract more energy from the wind.
in the depth-integrated and time-averaged modal near-inertial The wind power input in the Arabian Sea (∼208N, 4308E)
energy equation is enhanced due to the Tropical Cyclone Vayu during 10–
17 June 2019 (Mishra et al. 2021). The wind input to the sum
­En
Wn
1 = · Fn 1 Dn : (9) of modes 1–5 in Fig. 1b is also enhanced in regions of high
­t
mesoscale activities but is smaller than the wind input to the
The modal wind power input is given by Wn = t · un(z = 0), undecomposed fields. The wind input to modes 1–5 in Figs. 1c–g
where un (z
0)
U n (z
0) · û n is the nth mode near-inertial has an almost identical distribution. Most of the wind input
velocity vector at the surface of the ocean (Furuichi et al. in these modes are in the Southern Ocean (south of 308S).
2008). The modal energy En is computed following Buijsman Mode 2 has the largest wind input in areas south of Africa
et al. (2020) as and west of South America, while modes 3–5 have wind input
throughout the Southern Ocean. The regions of the western
2
1 p̂ n boundary currents have most of the wind input going to
En
r0 H |û n |2 1 : (10)
2 r
( 0 cn )2 modes 3–5, while in the Sea of Japan, modes 1–2 dominate.
The zonal mean wind power input to modes and the unde-
The depth-integrated modal horizontal energy flux is given composed fields, as well as other energy terms, are presented
by Fn
H^u n p̂ n where ^
u n and p̂ n are the modal amplitudes of in Fig. 2a. In the following sections, we keep referencing to
the horizontal velocity vector and perturbation pressure. The this figure. The wind power input in the Southern Hemisphere
residual term Dn includes dissipation of near-inertial modal is much higher than in Northern Hemisphere for May–June
energy, and also any mode–mode or mode–mean flow energy 2019. The zonal mean wind input has distinct peaks near lati-
transfers, which we do not explicitly calculate in our analysis. tudes 508S, 388S, and 398N. These correspond to regions of
Note that the time-averaged tendency term ­En =­t is close to high mesoscale energy in the Southern Ocean and western
zero and is ignored in our analysis. In the remainder of the boundary currents.
paper, we will drop the overbars when discussing the time- The globally integrated time-mean wind power input to the
averaged terms of Eqs. (1) and (9). undecomposed fields and the modes is shown in Fig. 1h. The
The near-inertial energy propagates at the group velocity. global mean surface near-inertial wind power input in our
We can calculate the horizontal group velocity of each mode simulation for May–June 2019 is 0.21 TW. This is in agree-
following Alford and Zhao (2007b) as ment with the 0.24 TW value obtained by Simmons and
Alford (2012) for the month of May-June in their global simu-
|Fn | lation. The wind power input to the first 4 modes is almost
cg,n
: (11)
En equally distributed. The sum of wind power contributions to

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

MAY 2022 RAJA ET AL. 827

FIG. 1. Time-mean, near-inertial surface wind power input to (a) the undecomposed fields, (b) sum of modes 1–5,
and (c)–(g) modes 1–5. (h) Global integral of wind power input (numbers above the bars indicate the fraction of the
undecomposed near-inertial wind input). The 2500-m seafloor depth is plotted as a gray contour in (g), to illustrate
major oceanic ridges such as the 1: Hawaiian–Emperor seamount chain, 2: French Polynesian islands, 3: Salas y Gómez
and Nazca Ridges, 4: Walvis Ridge, and 5: Azores Plateau.

the first five modes accounts for only 30.3% (0.064 TW/0.210 activity contributes to more wind power input. The global
TW) of the undecomposed near-inertial wind power input. depth and area integrated NIW energy for the undecomposed
The remaining 69.7% (0.146 TW/0.210 TW) of near-inertial fields is 224 PJ, of which the kinetic energy makes up 85.6%
wind power input goes to motions not resolved by modes 1–5, and the available potential energy make up 14.4%.
surface near-inertial oscillations, and/or mixed layer The higher NIW energy in the North Pacific Ocean region
turbulence. that appears in the undecomposed fields and the higher
modes is most likely due to thermobaric instability (TBI;
b. NIW energy
Buijsman et al. 2020). TBI is a numerical instability that can
To better understand patterns of NIW energy and NIW occur in ocean models with isopycnal coordinates due to
propagation pathways, we present the depth-integrated, time- imperfect compensation for compressibility in the pressure
mean NIW energy E for the undecomposed fields, modes 1–5, gradient term. The pressure gradient errors can cause the
and their sum in Fig. 3. The zonally averaged NIW energy in coordinate interfaces in weakly stratified regions to migrate in
the modal and the undecomposed fields are shown in Fig. 2b. a way that amplifies the pressure gradient errors, leading to
The global distribution of NIW energy and its zonal mean an exponentially growing instability (Hallberg 2005). The dis-
shows similar patterns as the wind input in Figs. 1 and 2a. turbances due to TBI have a broadband nature and they do
Most of the undecomposed and modal NIW energy resides in not correlate well with the surface wind stress. Hence, they
the Southern Ocean with higher values in regions south of have a minimal impact on the wind power input fields in
Africa and west of South America, corresponding to a higher Fig. 1. However, TBI generates near-inertial band motions
wind input. There is enhanced NIW energy in the western that mostly project on modes 3 and higher. The regions with
boundary current regions (∼408N) where high mesoscale TBI identified by the blue polygons in Fig. 3 are excluded for

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

828 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 52

FIG. 2. Zonal averages of near-inertial (a) wind input; (b) energy density; (c) divergence of
horizontal energy flux; (d) dissipation; (e) locally dissipated fraction qlocal, remotely dissipated
fraction qremote, and total (local 1 remote) dissipated fraction qtotal; (f) modal group velocities;
and (g) modal decay length scales. In (a)–(d), the thickness of each colored layer corresponds to
the energy/power of each mode and the blue line reflects the energy/power of the undecomposed
fields.

the undecomposed fields and modes $3 in the calculation of layer deepening and dissipation, instead of NIW energy
global energy integrals and other diagnostics. The correction (Alford 2020b).
of TBI in the model is beyond the scope of this paper.
c. Horizontal propagation of NIWs
Hallberg (2005) discusses various methods by which TBI can
be avoided in ocean models. In this section, we evaluate the horizontal energy flux patterns
The modal partitioning of the global depth and area inte- for the undecomposed and modal fields. The depth-integrated,
grated NIW energy is shown in Fig. 3h. Similar to the wind time-mean NIW horizontal energy fluxes FH for the undecom-
input, mode 1 is not the most energetic near-inertial mode in posed fields are presented in Fig. 4a (vectors) along with the
the global ocean. It makes up only 10.1% of the undecom- time-mean near-inertial wind power input (in color). The
posed energy. The NIW energy in each resolved mode is mode 1 fluxes are shown in Fig. 4b. Strong near-inertial fluxes
about equal (modes 2–4 have higher energies than mode 1), are directed toward the equator from regions of high wind input
reflecting the nearly equal wind power input to these modes in both the hemispheres. In the Northern Hemisphere, both the
(Fig. 1h). The sum of NIW energy in the first five modes Kuroshio and Gulf Stream have strong wind input that cause
amounts to 58% (130 PJ/224 PJ) of the undecomposed NIW large fluxes of about 100 W m21. The NIW energy fluxes are the
energy. The remaining 42% of the undecomposed NIW strongest in southeast Pacific (∼500 W m21) and the Southern
energy goes to unresolved high wavenumber motions and Ocean (∼300 W m21) where the wind input is large.
surface near-inertial oscillations. The fraction of undecom- Although most of the mode-1 fluxes are directed toward
posed energy going to the first five modes is higher than the equator in Fig. 4b, not all low-mode fluxes reach the
the fraction of the total near-inertial wind power input equator, but rather get reflected, scattered, and/or dissi-
to the first five modes (58% and 30.3%, respectively; pated at the east–west-oriented midocean ridges such as the
cf. Figs. 3h and 1h). This suggests that a significant portion Hawaiian–Emperor seamount chain in the North Pacific,
of surface near-inertial wind power input goes to mixed Azores Plateau in the North Atlantic Ocean, the French

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

MAY 2022 RAJA ET AL. 829

FIG. 3. As in Fig. 1, but for the depth-integrated, time-mean NIW energy E. The areas enclosed by the blue polygons
are regions with TBI and are excluded from calculating global integrals.

Polynesian islands, the Salas y Gómez and Nazca Ridges in projected on the first five modes (42% of the total NIW energy) is
the South Pacific, and the Walvis Ridge off the coast of associated with nonpropagating surface near-inertial motions and
Namibia in the southeast Atlantic (Fig. 4b; the ridges are higher wavenumber waves that also propagate vertically. The
labeled in Fig. 1g). Notably, the ∼2900-km-long Salas y mode-1 flux clearly dominates among the first five modes, in con-
Gómez and Nazca Ridges that extend westward from the trast to the modal distributions of E and W. While the mode-1
west coast of South America (∼258S) block an average equa- fluxes are directed equatorward and have coherent patterns, fluxes
torward energy flux of 495 W m21. Together with the in modes 2 and higher have more chaotic flux patterns. A possible
French Polynesian islands, the Salas y Gómez and Nazca explanation for these chaotic patterns is that higher modes are
Ridges form a barrier for the equatorward propagation of NIW more strongly refracted by the background flow than lower modes
fluxes in the South Pacific Ocean between 1808 and 2908W, (Rainville and Pinkel 2006; Zaron and Egbert 2014).
blocking about 75% of the equatorward mode-1 fluxes. We compute depth-integrated horizontal flux divergences
Strong northward and diverging fluxes occur in the northern to show sources and sinks of NIW energy. Instead of show-
Pacific south of the Aleutian islands in Fig. 4a without any corre- ing noisy global maps, we discuss the zonal means of the
sponding wind power input (Fig. 1a). These fluxes are due to total and modal flux divergence shown in Fig. 2c. The zon-
TBI and mostly project on higher modes and not on mode 1 ally averaged flux divergence of modes 1–5 nearly sums up
(Fig. 4b). The broadband disturbances associated with the TBI to the undecomposed flux divergence for all latitudes. The
have superinertial frequencies that allow for poleward propagation. flux divergences of modes 1 and 2 are much higher than that
We compare the undecomposed and modal flux patterns in for modes 3–5. Low-mode NIWs propagate away from
Fig. 5 for the Gulf Stream region. The sum of the fluxes in the higher latitudes (e.g., .|408|) where flux divergence is posi-
first five modes is about 98.7% of the undecomposed flux in tive and converge in lower latitudes (e.g., ,|308|) where the
the global ocean (see Table 1 for the ratio for each mode). This divergence is negative. In the Southern Hemisphere, modes 1
means that we capture most of the horizontally propagating NIWs and 2 have high negative divergence at around 308S owing to
with the first five modes. Thus, the NIW energy that is not their reflection and eventual dissipation at midocean ridges such

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

830 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 52

FIG. 4. Depth-integrated, time-mean NIW horizontal energy flux (vectors) for the (a) undecom-
posed (total) and (b) mode-1 fields. The color shadings represent the time-mean near-inertial
surface wind power input in (a) and the magnitude of mode-1 flux in (b). The black and blue
vectors respectively represent fluxes with a magnitude less than and greater than 100 W m21 in (a)
and 50 W m21 in (b). The fluxes with magnitudes less than 10 W m21 are not shown.

as the Salas y Gómez, Nazca, and Walvis Ridges. At the onset computes global-mean qlocal values of 0.63 and 0.75 for win-
of the northern summer, the zonal-mean flux divergence in the ter and summer, respectively. We compute the locally dissi-
Northern Hemisphere is smaller than in the Southern Hemi- pated fraction in our model for the duration of 30 days and
sphere, with a minor peak close to 408N where the wind input compare with the seasonal-mean values of Alford (2020a).
has a maximum in the western boundary currents. The choice of the cutoff mode number to define the propa-
gating component of the wind input depends on the definition
d. Dissipation of NIWs
of “local.” With the horizontal resolution of ∼4 km in our sim-
The time-mean and zonally averaged NIW dissipation, that ulation, all the resolved modes propagate out of the grid cell
is, the residual calculated from the energy balance Eqs. (1) where they are generated. However, maps of qlocal at 4-km res-
and (9), is presented in Fig. 2d. Peaks in the dissipation agree olution are noisy. To reduce the grid-scale noise and to better
well with the peaks in energy and wind input in Figs. 2a and b. visualize the spatial trends, we compute qlocal in 28 3 28 bins.
Indeed, most of the wind power input is dissipated at the We can estimate the distance traveled by the modes before they
same latitude. The remainder propagates away as low-mode are dissipated by the average decay length scales Ld for modes
waves that dissipate at lower latitudes. Since the zonally aver- given by Eq. (12). The Ld for modes 4 and 5 are O(100 km) or
aged horizontal flux divergence is only around 5% of the ,28, whereas the Ld for modes 1–3 are O(1000 km), as shown
wind input in most latitudes, we may say that zonally aver- in Fig. 2g. Hence, we choose mode 3 as the cutoff to define the
aged D is almost equal to zonally averaged W. propagating component of the wind input. However, in contrast
The locally dissipated fraction of internal wave power to the computation of Alford (2020a), we assume that all the
input, qlocal = Dlocal/I (where Dlocal is the local dissipation rate wind input that does not project on modes 1–3 contributes to
and I is the internal wave power input associated with baro- local dissipation, since it is not available for propagating NIWs.
tropic to baroclinic tide conversion or wind power input) is Thus, we compute the locally dissipated fraction as
used to parameterize the internal wave driven mixing in the
3
ocean (MacKinnon et al. 2017). Although qlocal is generally
W2 Wn
evaluated for internal tides, Alford (2020a) recently discussed n
1
qlocal
: (13)
the calculation of qlocal for NIWs in the global ocean. Follo- W
wing Vic et al. (2019), Alford (2020a) assumes that NIW
modes 1–3 propagate away and modes 4–20 contribute The locally dissipated fraction qlocal is presented in Fig. 6a,
to local dissipation. Using this assumption, Alford (2020a) and its zonal average is presented in Fig. 2e (black curve).

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

MAY 2022 RAJA ET AL. 831

FIG. 5. Depth-integrated near-inertial horizontal energy fluxes FH for (a) the undecomposed fields, (b) sum of the
first five modes, and (c)–(g) modes 1–5. The magnitude of the flux is shown in color shades. Blue vectors show fluxes
with a magnitude larger than 100 W m21 in (a) and (b), 50 W m21 in (c), and 10 W m21 in (d)–(g).

The distribution of qlocal is largely uniform throughout the The global distribution of qremote is shown in Fig. 6b. The
ocean. We obtain a global mean value of qlocal = 0.79 for the value of qremote is higher at the lower latitudes (,|308|) partic-
period of May–June, which is comparable to the estimates of ularly near many oceanic ridges. The strong NIW fluxes gen-
qlocal for summer by Alford (2020a). erated at the higher latitudes of the Southern Hemisphere
The dissipation of near-inertial energy at any location in propagate equatorward and dissipate at the lower latitudes.
the ocean not only includes the dissipation of locally gener- The region west of South America has very high qremote at
ated NIWs but also the dissipation of remotely generated around 308S, owing to the reflection and dissipation of low-
NIWs that have propagated there. We can calculate a mode NIWs at the Salas y Gómez and Nazca Ridges; qremote
remotely dissipated fraction as is also higher at the French Polynesian islands in the South
 Pacific and the Walvis Ridge in the South Atlantic. The zon-
3
D2 W2 Wn ally averaged qremote in Fig. 2e (green curve) reflects the map
n
1 in Fig. 6b. The latitudes where the zonally averaged qremote is
qremote
: (14) high corresponds with the latitudes where the flux divergence
W
is negative (see Fig. 2c).
We also show the zonal average of qtotal = qremote 1 qlocal = D/W
in Fig. 2e (red curve). The zonally averaged qtotal has values less
TABLE 1. The ratio of the area-averaged flux magnitudes of than 1 at higher latitudes and increases to above 1 at low lati-
the near-inertial modes to the area-averaged flux magnitudes of
tudes. In the regions where qtotal . 1, the near-inertial dissipa-
the undecomposed fields in the global ocean.
tion exceeds the local wind input due to the dissipation of
5
n 1 2 3 4 5 remotely generated NIWs. This analysis shows that the contri-
n
1
bution of remote waves should not be ignored when developing
|Fn|/|FH| 50.5% 28.9% 11.6% 5.7% 2.0% 98.7%
parameterizations for internal wave driven mixing in the ocean.

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

832 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 52

FIG. 6. (a) The locally dissipated fraction qlocal (global mean = 0.79) and (b) the remotely dissipated
fraction qremote. The 0- and 2500-m seafloor depths are plotted as black contours.

Low-mode NIWs transport near-inertial energy from higher weaker vorticity field coincides with strong wind input due to
to lower latitudes. The speed of energy propagation is deter- local storms.
mined by the group speed. We compute the group speed cg,n We examine the distribution of the undecomposed and
for modes 1–5 using Eq. (11) and present the zonally averaged modal near-inertial wind input, NIW kinetic energy, horizon-
group velocity in Fig. 2f. Mode 1 has the highest cg,n, almost 2 tal flux divergence and dissipation as a function of binned vor-
times the mode-2 speed. In both hemispheres, the values of ticity (Fig. 8). We discuss the distribution of KE (instead of
cg,n for modes 1 and 2 increase toward the lower latitudes E) with respect to vorticity because the distribution of APE is
with a peak at about |258| before decreasing toward the equa- noisy. The energy terms are area averaged in normalized vor-
tor. The decrease toward the equator may be attributed to ticity bins with a size of Dz/f = 0.05 for the global ocean pole-
increased dissipation at midocean ridges and/or the way the ward of |78| and excluding the regions with TBI. To better
group speed is calculated, that is, it depends on waves that are isolate the vorticity effect on each mode, we also present the
generated at different latitudes with different theoretical normalized energetics in the right column of Fig. 8. The wind
group speeds. In Fig. 2g, Ld is largest for mode 1, with higher input, kinetic energy, and dissipation (Figs. 8b,d,h) are nor-
modes having gradually smaller values. The peaks of Ld corre- malized by the total area-averaged wind input, kinetic energy
spond to the zonally averaged flux divergence in Fig. 2c, with and dissipation in each mode or the undecomposed fields,
regions of negative flux divergence having high Ld. Thus, low- respectively. The flux divergence (Fig. 8f) is normalized by
mode NIWs with larger cg,n and Ld travel from higher lati- the root-mean-square flux divergence in each mode or the
tudes to lower latitudes, where they dissipate. undecomposed fields.
The modal wind input in Figs. 8a and 8b does not show any
e. Influence of background vorticity
correlation with the sign of background vorticity. The distribu-
The horizontal propagation of NIWs is not only affected by tion of wind power input as a function of normalized vorticity
the gradients in planetary vorticity but also by the gradients in is similar for all modes. The wind input in modes is higher in
relative vorticity of the background flow, which shifts the local areas with higher vorticity (positive or negative). The wind
inertial frequency as feff = f(1 1 0.5z/f) (Mooers 1975; Kunze input to the undecomposed fields (black curves in Figs. 8a,b)
1985). In this section, we examine the influence of background is slightly higher in cyclonic eddies than in anticyclonic
relative vorticity on the undecomposed and modal NIW eddies.
energy propagation and dissipation in the global ocean. The depth-integrated modal NIW KE as a function of back-
The 30-day mean surface vorticity normalized by local f is ground vorticity is shown in Fig. 8c and 8d. The KE of the
shown in Fig. 7a. The positive values correspond to cyclonic undecomposed fields (black curves in Figs. 8c,d) clearly shows
and negative values correspond to anticyclonic vorticities. preference to anticyclonic eddies. We see in Fig. 8d that for
Large mesoscale eddies occur throughout the ocean with KE, higher modes correlate more strongly with the normal-
stronger vorticities associated with major ocean currents such ized vorticity than lower modes, although all modes tend to
as the western boundary and Agulhas Currents. The regions be trapped in regions of anticyclonic vorticity.
with strong mesoscale eddies coincide with areas of enhanced Figures 8e and 8f show the divergence of NIW horizontal
near-inertial wind power input (cf. with Fig. 4). A notable energy flux as a function of normalized background vorticity.
exception is the region west of South America, where a The normalized flux divergence for all modes and for the

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

MAY 2022 RAJA ET AL. 833

FIG. 7. Time-mean (a) surface vorticity z normalized by local f, (b) residual NIW kinetic
energy DKE, and (c) NIW vertical energy flux Fz at z = 2500 m. The areas enclosed by the blue
polygons are regions with TBI and are excluded from the analyses.

undecomposed fields in Fig. 8f is positive for cyclonic vortic- surface area. Nevertheless, the trends are clearly present
ities and negative for anticyclonic vorticities. Similar to KE for |z/f| , 0.1.
(Fig. 8d), negative normalized flux divergence of the high
f. Vertical energy flux
modes correlates more strongly with anticyclonic vorticity
than the flux divergence of the lower modes. This clearly The difference between the undecomposed and modal
shows that globally, all resolved NIW modes propagate away kinetic energy, that is, the residual kinetic energy DKE, is
from cyclonic eddies and toward anticyclonic eddies. In accor- 42% of the undecomposed kinetic energy. What is the fate of
dance with the fluxes in Fig. 5, the sum of the flux divergence this energy? A global map of the DKE is shown in Fig. 7b. We
of modes 1–5 agrees well with the undecomposed flux diver- see that DKE is mostly confined to anticyclonic eddy features,
gence for all vorticity values, further confirming that the first but the largest eddies do not necessarily have the highest
five modes include most of the horizontally propagating DKE. The Southern Ocean has a high concentration of DKE,
NIWs. Last, the dissipation of the undecomposed fields and particularly to the south of Africa.
the modes in Figs. 8g and 8h reflect the same trends as for W, Since almost all of the horizontal fluxes are represented
KE, and = · FH; that is, NIW dissipation is biased toward the by the first five modes, the residual KE most likely reflects
anticyclones. Our results from a global ocean simulation are energy of high wavenumber waves that propagate vertically
in agreement with the observations of Whalen et al. (2018) downward. Hence, we calculate the vertical flux of near-
and Sanford et al. (2021), who also found that near-inertial inertial energy. The vertical energy flux is given by Fz =
energy is concentrated in anticyclones. p w, where w is the near-inertial vertical velocity calculated
The bin area as a fraction of the total surface area of offline using the continuity equation. The maximum mixed
the ocean is shown in Fig. 8b. We calculate that almost layer depths are shallower than 500 m in the global ocean
66% of ocean surface area has |z/f| , 0.025 and almost during May–June 2019. Therefore, we calculate the vertical
90% of the ocean surface area has |z/f| , 0.01. We see the energy flux at z = 2500 m to estimate NIW energy transmis-
strongest response of the NIW energy terms for the largest sion to the deep ocean. We show the time-mean vertical
vorticity values, which occur for less than 10% of the ocean flux of NIW energy at a depth of 500 m in Fig. 7c. Negative

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

834 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 52

FIG. 8. The distribution of time-mean, area-averaged near-inertial (a),(b) surface wind power input; (c),(d) kinetic
energy; (e),(f) divergence of horizontal energy flux; and (g),(h) dissipation as function of time-mean surface relative
vorticity. The variables for the undecomposed fields are shown as thick black curves, and those for the sum of
modes 1–5 are shown as thick red curves. The fractional bin area of the total ocean surface area of 2.3 3 108 km2 is
shown as a black dotted curve in (b). All values are area averaged for vorticity bins of Dz/f = 0.05. In (b), (d), (f), and
(h) the y axis variables are normalized as described in the text.

values indicate a downward energy flux. The NIW vertical TBI generates NIWs in the interior of the ocean that propa-
flux is higher in regions with strong wind input (cf. with gate toward the surface. We neglect the TBI regions in our
Fig. 4a), which are also regions of high mesoscale activity. calculations. In our simulation, the area-integrated near-
We also find high NIW transmissions in the Southern inertial power that is transmitted below z = 2500 m is 0.04
Ocean that are mostly confined to anticyclonic eddies. The TW (19% of the surface wind power input). Thus, most of
regions outlined by the blue polygons are regions with TBI, the wind power (81%) is dissipated in the top 500 m of the
where the vertical energy fluxes are also positive, that is, ocean.

FIG. 9. (a) The distribution of the time-mean depth-integrated, normalized residual kinetic energy DKE and vertical
energy flux Fz at z = 2500 m as a function of the time-mean surface relative vorticity z/f. Values are area-averaged for
vorticity bins of Dz/f = 0.05. (b) The distribution of the time-mean normalized DKE as a function of normalized Fz.
The DKE and Fz are normalized by their root-mean-square values.

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

MAY 2022 RAJA ET AL. 835

FIG. 10. Time-mean (a) surface wind input, (b) relative vorticity, (c) KE, (d) sum of KE in first five modes, and (e) DKE along a vertical
transect at 300.58W longitude. The blue contours enclose z/f , 20.1, representing anticyclonic motions. Also shown are time series of
(f) KE, (g) sum of KE in the first five modes, and (h) DKE near an anticyclone at 37.38N, 300.58W. The vertical profile for which the time
series are extracted is shown with a thin vertical black line in (b)–(e).

Both normalized DKE and Fz correlate strongly with nega- small horizontal and vertical scales. We compute a time-mean
tive relative vorticity (Fig. 9a). The area-averaged DKE and downward near-inertial flux of 0.068 W m22 at z = 2500 m at
Fz are normalized by their respective root-mean-square val- 38.58N.
ues. The residual KE that is not associated with horizontally Figures 10f–h show the time series of the undecomposed
propagating modes is trapped in anticyclonic eddies. The KE, KE in modes 1–5 and residual KE at the location of the
downward transmission is also strongest in anticyclonic anticylonic eddy (37.38N, 300.58). The undecomposed KE has
eddies. Furthermore, in Fig. 9b, we find that the normalized a maximum at z = 21000 m on day 5 and progresses down-
DKE is higher when the vertical flux of NIW energy is nega- ward as days pass. The downward propagation of KE is
tive (downward vertical flux). This clearly establishes that the marked by dashed lines in Fig. 10f. The residual KE in
vertical flux of NIW energy is associated with DKE. Fig. 10h has a vertical structure associated with large vertical
and horizontal wavenumber waves not captured by our modal
VERTICALLY PROPAGATING NIWS: A CASE STUDY analysis. From the downward propagation of KE, we compute
The vertical propagation of NIW kinetic energy is further a vertical group speed of 21.4 mm s21 (2120 m day21).
examined in an anticyclonic eddy. We take a vertical tran- The downward propagation of NIW KE is associated with
sect through an anticyclonic eddy at 37.38N, 300.58W in the short wavelength (large wavenumber) waves, as shown in
Gulf Stream region. The transect runs from 358 to 418N and Fig. 10. High wavenumber NIWs that do not project on verti-
includes a nearby region of strong wind input. Figure 10a cal modes 1–5 are modulated more by the background eddies
shows the time-mean wind input along the vertical transect. and are most likely to propagate downward as beamlike fea-
The wind input has a peak at latitudes 38.58–39.58N. The tures in anticyclonic eddies. Numerous process modeling stud-
relative vorticity in Fig. 10b shows the anticyclone near ies (Lee and Niiler 1998; Danioux et al. 2015; Asselin and
37.38N. The time-mean near-inertial KE in Fig. 10c is Young 2020) and observations (Whalen et al. 2018; Thomas
enhanced in the upper layers of the ocean in regions where the et al. 2020; Sanford et al. 2021) have established that back-
wind input is high, but is trapped at the base of the anticy- ground anticyclonic relative vorticity (negative z/f) enhance
clonic core near z = 21000 m. In contrast to the modal KE in the vertical transmission of NIW energy, and that NIWs get
Fig. 10d the residual KE in Fig. 10e shows a large variability at trapped in anticyclonic eddies, where they ultimately dissipate.

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

836 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 52

4. Discussion only 17.7% (23 PJ/130 PJ) of the sum of energy in modes 1–5.
Consequently, the energy in modes 2–5 for NIWs (107 PJ) is
The primary purpose of this paper is to study the modula-
about 1.6 times as high as for M2 internal tides (67 PJ). From
tion of NIW energetics by the background flows in a global
these results, we can conclude that while the long propagating
HYCOM simulation. For the first time in a global ocean simu-
low-mode internal wave energy is dominated by internal tides,
lation, we have examined the global spatial patterns of the
the high-mode internal wave energy that is relevant for
time-mean and depth-integrated total, modal and residual
mixing is dominated by NIWs. This difference in modal parti-
NIW energy terms and studied how these energy terms are
tioning of energy between internal tides and NIWs has also
modulated by the background vorticity. We will now discuss
been noted in observations by Alford (2010).
our results in the context of previous model studies and dis-
Although energetically weaker than internal tides, we do
cuss the merits and limitations of our simulation.
find NIWs that propagate equatorward over long distances
The analyses presented in this paper are for 30 days in
(Fig. 4b) in our simulation. The ratio of the undecomposed
May–June 2019. This period coincides with the cruises in the
fluxes that radiate equatorward of |308| to the wind input for
North Atlantic Ocean as part of Near Inertial Shear and
latitudes . |308| is 9.8% (1.4 GW/14.7 GW) in the Northern
Kinetic Energy in the North Atlantic experiment (NISKINe).
Hemisphere and 3.0% (4.6 GW/153.6 GW) in the Southern
Although 30 days of 3D data permits us to study the ener-
Hemisphere, resulting in a global value of 3.6%. The ratio we
getics of NIWs and their modulation by the background
calculate for May–June is less than the annual mean ratio of
mesoscale vorticity, we do not capture much of the seasonal-
7% calculated by Furuichi et al. (2008) and falls in the lower
ity in the NIW dynamics. May–June, at the onset of boreal
end of the annual range of 2%–16% given by Simmons and
summer, has the lowest near-inertial wind power input in
Alford (2012). The lower ratio of equatorward radiation in
the global ocean, particularly in the Northern Hemisphere.
our simulation may be due to a higher model resolution that
In the Southern Hemisphere, May–June has a wind power
facilitates more nonlinear interactions, the inclusion of wave
input that is about 83% of the peak power input (Simmons
drag in HYCOM, the presence of background flows, and a
and Alford 2012). Consequently, most of the NIW energy in
more realistic topography, which may block and reflect NIWs
our simulation is in the Southern Hemisphere. The low and
(e.g., at the Salas y Gómez and Nazca Ridges in the South
high wind speed conditions in the Northern and Southern
Pacific Ocean).
Hemispheres coincide with a significant range of stratifica-
The vertical transmission of NIW energy into the ocean
tion, mesoscale variability, and wind speeds in our simula-
interior in our simulation is in alignment with previous model
tion. Hence, we argue that our findings with respect to the
studies. The area-integrated vertical flux of near-inertial
vorticity modulation of NIW energy terms are representa-
energy at 500-m depth is 0.04 TW (19% of the surface wind
tive and not unique to our simulation period.
power input). Hence, 81% of the wind power input is dissi-
Even though the horizontal and vertical resolution of our
pated in the top 500 m of the ocean. Furuichi et al. (2008)
simulations is an improvement over past studies on global
found that about 75%–80% of surface near-inertial wind
NIW energetics, our simulations fall short in fully resolving
input is dissipated in the top 150 m using a model without
all the nonlinear interactions among waves themselves and
eddies. Rimac et al. (2016) estimated that only 10% of wind
between waves and (sub)-mesoscale flows. The resolution of
power is transmitted below the surface mixed layer in a model
our simulation (4 km in the tropics) is sufficient to resolve
simulation with mesoscale eddies. We find a greater propor-
most of the mesoscale motions, but the submesoscales
tion of NIW energy transmission to the ocean interior in
(∼1.5 km; Morvan et al. 2020), where most of the nonlinear
HYCOM, possibly due to better horizontal resolution of
interactions take place, are not resolved in our simulations.
1/258, better resolved mixed layer, and more frequent wind
Hence, we include all the energy transfers through nonlin-
forcing (3 hourly) in our model. The higher percentage of
ear interactions in the dissipation term [D in Eq. (1)],
downward flux in our study can also be due to the lower wind
assuming that all such interactions lead to motions that dis-
input in May–June. The observations by Alford et al. (2012)
sipate locally. Nevertheless, Müller et al. (2015) studied show that the downward flux of NIW energy at 800 m can
spectral energy transfers in a HYCOM simulation with simi- vary between 12% and 33% depending on the seasonal varia-
lar configuration as ours, and they have been able to repro- tion of the wind input. In a future paper, we will study the ver-
duce low vertical-mode internal wave triad interactions. In a tical propagation and dissipation of NIW energy as a function
subsequent study, Ansong et al. (2018) analyzed the global of depth in more detail.
distribution of PSI of internal tides in 4-km HYCOM simu-
lations. We too find some evidence of nonlinear interactions
5. Conclusions
in our simulation with tides, which is discussed in the
appendix. We have studied the depth-integrated energetics of NIWs
Our study shows that the modal partitioning of the depth- in realistically forced eddying global HYCOM simulations
integrated energy for NIWs is in stark contrast to the modal during 30 days in May–June 2019. We have determined what
partitioning for M2 internal tides. For M2 internal tides (see fraction of total near-inertial wind input, energy, and dissipa-
Fig. 9c in Buijsman et al. 2020), the depth-integrated energy tion can be attributed to the resolved vertical modes. We
in mode 1 is 56.2% (86 PJ/153 PJ) of the sum of energy in have examined the global spatial patterns of the total, modal,
modes 1–5, whereas for NIWs the energy in mode 1 makes up and residual NIW energy terms and studied how these energy

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

MAY 2022 RAJA ET AL. 837

terms are modulated by the background vorticity. We list the APPENDIX

main findings from our analyses below:
HYCOM Simulation with Tidal Forcing
• The wind input in the first five modes adds up to only
30.3% of the total near-inertial wind input. However, the In this section, we examine the HYCOM simulation
NIW energy in the first five modes adds up to about 58% EXPT 19.0, which is similar to EXPT 19.2 except with the
of the total NIW energy. This difference in modal projec- inclusion of tidal forcing, and we discuss the difference in
tions may imply that a significant portion of surface wind the global distribution of NIW kinetic energy with the addi-
input goes to the deepening of mixed layer and is not avail- tion of tides. Both experiments have the same wind forcing.
able for near-inertial motions as suggested by Alford We extract near-inertial fields from the experiments after
(2020b). removing stationary tides with a harmonic analysis and by
• Almost all of the depth-integrated NIW horizontal energy applying a bandpass filter as explained in section 2b. The
flux projects on the first five modes. Hence, the residual difference between the depth-integrated, time-mean, near-
energy that does not project on the first five modes is asso- inertial kinetic energies of EXPT 19.0 and EXPT
ciated with near-inertial motions that do not propagate 19.2 (KE19.0 2 KE19.2) is shown in Fig. A1a, and the zon-
horizontally. ally averaged difference of kinetic energies (excluding TBI
• The low-mode NIW fluxes diverge from higher latitudes regions) is presented in Fig. A1b. We identify several differ-
(e.g., .|408|) and converge in lower latitudes (e.g., ,|308|). ences between the experiments. The diurnal (D1) internal
The equatorward propagation of low-mode NIWs is tides are not perfectly removed in EXPT 19.0 (with tides)
impeded at many of the east–west-oriented midocean in particular in the western Pacific where D1 internal tides
ridges, where these NIWs reflect, scatter, and/or dissipate. are the strongest. In EXPT 19.0, the M2 internal tides and
• The locally dissipated fraction of NIW energy qlocal is found NIWs interact nonlinearly via PSI near the critical latitudes
to be uniform throughout the global ocean with a global (∼|308|) where the near-inertial frequency coincides with the
mean value of qlocal = 0.79. Additionally, we compute a diurnal frequency, leading to enhanced NIW energy, mostly
remotely dissipated fraction qremote. The value of qremote is in the western Pacific. The TBI is slightly worse in EXPT
higher at the lower latitudes (,|308|) than at the higher lati- 19.2 (without tides) than in EXPT 19.0. The difference map
tudes, in particular near the east–west-oriented oceanic in Fig. A1a has a noisy appearance because eddies that trap
ridges. NIW kinetic energy occur in slightly different locations in
• The energy and energy flux divergence are strongly modu- the two experiments, in particular in the Southern Ocean
lated by background vorticity. Most of the horizontal and the western boundary currents. Although EXPT 19.0
energy flux converges in anticyclones. High modes are has enhanced NIW kinetic energy in the western Pacific
more affected by the background eddies than low modes. and near the critical latitudes, globally, EXPT 19.0 has less
We observe these trends for low vorticity values NIW kinetic energy (KE19.0 = 177.5 PJ) than EXPT 19.2
(|z/f| , 0.1), which coincide with 90% of the global ocean (KE19.2 = 191.8 PJ), excluding TBI regions. The reason for
surface area. The residual kinetic energy, DKE, (42% of this difference is not directly clear, but in the simulation
total) is mostly concentrated in anticyclonic eddies. Most of with tides, internal tides, and NIWs can interact nonlinearly
this energy propagates downward, likely as waves with high and scatter energy to higher harmonics.
horizontal and vertical wavenumbers. Near the critical latitudes, the nonlinear interactions
• The global near-inertial wind power input for a 30-day between semidiurnal internal tides and NIWs result in
period in May–June 2019 is 0.21 TW. Of this, only 0.04 TW enhanced energy in the near-inertial frequency band in the
(19% of the surface wind power input) is transmitted below western Pacific to the west of Hawaii near 308N in EXPT
500-m depth. Most of the wind power input (81%) is lost in 19.0 (Figs. A1a,b). To further highlight the difference
the upper layer processes of the ocean and is not available between the no-tides and tides simulations, we show the
for deep ocean mixing. difference in velocity rotary spectra (S19.0 2 S19.2) at 308N,
1528E in the western Pacific, where we expect PSI, in Fig.
A1c. EXPT 19.0 has higher energy at the near-inertial fre-
Acknowledgments. This work was supported by the Office quency (positive values in Fig. A1c indicate higher energy
of Naval Research through NISKINe project Grants in EXPT 19.0). PSI of the semidiurnal internal tides can
N000141812801 (Buijsman and Raja), N0001420WX01883 effectively transfer energy of low-mode semidiurnal internal
(Shriver), and N000141812544 (Arbic). We thank the editor tides to high-mode near-inertial waves, which has been
and three anonymous reviewers for helpful suggestions to observed in the ocean (Van Haren 2005; Xie et al. 2011).
improve the presentation and interpretation of our results. The nonlinear interaction of M2 internal tides and NIWs
near critical latitudes can also excite waves at sum and dif-
Data availability statement. The global time-mean near- ference frequencies of f and M2 (Liang et al. 2019). In Fig.
inertial fields of modal and undecomposed wind input, A1c, EXPT 19.0 has higher energy at the f, M2, and f 1 M2
depth-integrated kinetic and available potential energies, and frequencies. This suggests the occurrence of nonlinear reso-
depth-integrated horizontal energy fluxes for EXPT 19.2 are nant interactions between NIWs, M2 internal tides, and the
published online (https://zenodo.org/record/5765439). f 1 M2 harmonic.

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

838 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 52

FIG. A1. (a) The difference of near-inertial kinetic energy between EXPT 19.0 (with tides) and EXPT 19.2 (without
tides). (b) The zonally averaged difference of energy between EXPT 19.0 and EXPT 19.2. (c) The difference in veloc-
ity rotary spectra for EXPT 19.0 (with tides) and EXPT 19.2 (without tides) at 308N, 1508E (PSI and M2 1 f interac-
tion region). The local f and M2 1 f frequencies are shown by dashed lines.

REFERENCES }}, and Coauthors, 2018: Geographical distribution of diurnal

and semidiurnal parametric subharmonic instability in a
Alford, M. H., 2001: Internal swell generation: The spatial distri- global ocean circulation model. J. Phys. Oceanogr., 48, 1409–
bution of energy flux from the wind to mixed layer near-iner- 1431, https://doi.org/10.1175/JPO-D-17-0164.1.
tial motions. J. Phys. Oceanogr., 31, 2359–2368, https://doi. Arbic, B. K., J. G. Richman, J. F. Shriver, P. G. Timko,
org/10.1175/1520-0485(2001)031,2359:ISGTSD.2.0.CO;2. E. J. Metzger, and A. J. Wallcraft, 2012: Global modeling of
}}, 2003: Improved global maps and 54-year history of wind- internal tides: Within an eddying ocean general circulation
work on ocean inertial motions. Geophys. Res. Lett., 30, 1424, model. Oceanography, 25, 20–29, https://doi.org/10.5670/
https://doi.org/10.1029/2002GL016614. oceanog.2012.38.
}}, 2010: Sustained, full-water-column observations of internal }}, and Coauthors, 2018: Primer on global internal tide and
waves and mixing near Mendocino escarpment. J. Phys. Oce- internal gravity wave continuum modeling in HYCOM and
anogr., 40, 2643–2660, https://doi.org/10.1175/2010JPO4502.1. MITgcm. New Frontiers in Operational Oceanography,
}}, 2020a: Global calculations of local and remote near-inertial- GODAE OceanView, 307–392.
wave dissipation. J. Phys. Oceanogr., 50, 3157–3164, https:// Asselin, O., and W. R. Young, 2020: Penetration of wind-
doi.org/10.1175/JPO-D-20-0106.1. generated near-inertial waves into a turbulent ocean. J.
}}, 2020b: Revisiting near-inertial wind work: Slab models, rela- Phys. Oceanogr., 50, 1699–1716, https://doi.org/10.1175/
tive stress, and mixed layer deepening. J. Phys. Oceanogr., JPO-D-15-0074.1.
50, 3141–3156, https://doi.org/10.1175/JPO-D-20-0105.1. Buijsman, M. C., and Coauthors, 2016: Impact of parameterized
}}, and Z. Zhao, 2007a: Global patterns of low-mode internal- internal wave drag on the semidiurnal energy balance in a
wave propagation. Part I: Energy and energy flux. J. Phys. global ocean circulation model. J. Phys. Oceanogr., 46, 1399–
Oceanogr., 37, 1829–1848, https://doi.org/10.1175/JPO3085.1. 1419, https://doi.org/10.1175/JPO-D-15-0074.1.
}}, and }}, 2007b: Global patterns of low-mode internal- }}, and Coauthors, 2020: On the interplay between horizontal
wave propagation. Part II: Group velocity. J. Phys. Ocean- resolution and wave drag and their effect on tidal baroclinic
ogr., 37, 1849–1858, https://doi.org/10.1175/JPO3086.1. mode waves in realistic global ocean simulations. Ocean
}}, M. F. Cronin, and J. M. Klymak, 2012: Annual cycle and Modell., 152, 101656, https://doi.org/10.1016/j.ocemod.2020.
depth penetration of wind-generated near-inertial internal 101656.
waves at Ocean Station Papa in the northeast Pacific. Danioux, E., J. Vanneste, and O. Bühler, 2015: On the concentra-
J. Phys. Oceanogr., 42, 889–909, https://doi.org/10.1175/ tion of near-inertial waves in anticyclones. J. Fluid Mech.,
JPO-D-11-092.1. 773, R2, https://doi.org/10.1017/jfm.2015.252.
Ansong, J. K., B. K. Arbic, M. C. Buijsman, J. G. Richman, D’Asaro, E. A., C. C. Eriksen, M. D. Levine, C. A. Paulson,
J. F. Shriver, and A. J. Wallcraft, 2015: Indirect evidence for P. Niiler, and P. Van Meurs, 1995: Upper-ocean inertial cur-
substantial damping of low-mode internal tides in the open rents forced by a strong storm. Part I: Data and comparisons
ocean. J. Geophys. Res. Oceans, 120, 6057–6071, https://doi. with linear theory. J. Phys. Oceanogr., 25, 2909–2936, https://
org/10.1002/2015JC010998. doi.org/10.1175/1520-0485(1995)025,2909:UOICFB.2.0.CO;2.

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

MAY 2022 RAJA ET AL. 839

de Lavergne, C., S. Falahat, G. Madec, F. Roquet, J. Nycander, Liang, C.-R., G.-Y. Chen, X.-D. Shang, X.-H. Xie, and D.-X.
and C. Vic, 2019: Toward global maps of internal tide energy Wang, 2019: Observation of enhanced nonlinear interactions
sinks. Ocean Modell., 137, 52–75, https://doi.org/10.1016/j. after severe Tropical Storm Chanchu (2004) in the western
ocemod.2019.03.010. South China Sea. J. Geophys. Res. Oceans, 124, 3837–3848,
Egbert, G. D., and R. D. Ray, 2001: Estimates of M2 tidal https://doi.org/10.1029/2018JC014839.
energy dissipation from TOPEX/Poseidon altimeter data. MacKinnon, J. A., and Coauthors, 2017: Climate process team on
J. Geophys. Res., 106, 22 475–22 502, https://doi.org/10. internal wave–driven ocean mixing. Bull. Amer. Meteor. Soc.,
1029/2000JC000699. 98, 2429–2454, https://doi.org/10.1175/BAMS-D-16-0030.1.
Flexas, M. M., A. F. Thompson, H. S. Torres, P. Klein, J. T. Farrar, Metzger, E. J., and Coauthors, 2014: US Navy operational global
H. Zhang, and D. Menemenlis, 2019: Global estimates of the ocean and Arctic ice prediction systems. Oceanography, 27,
energy transfer from the wind to the ocean, with emphasis 32–43, https://doi.org/10.5670/oceanog.2014.66.
on near-inertial oscillations. J. Geophys. Res. Oceans, 124, Mishra, K., M. Sharma, and M. Mohapatra, 2021: Performance of
5723–5746, https://doi.org/10.1029/2018JC014453. numerical weather prediction models in predicting track of
Furuichi, N., T. Hibiya, and Y. Niwa, 2008: Model-predicted distri- recurving Cyclone Vayu over Arabian Sea during June 2019.
bution of wind-induced internal wave energy in the world’s J. Earth Syst. Sci., 130, 25, https://doi.org/10.1007/s12040-020-
oceans. J. Geophys. Res., 113, C09034, https://doi.org/10.1029/ 01533-7.
2008JC004768. Mooers, C. N., 1975: Several effects of a baroclinic current on
Gill, A., 1984: On the behavior of internal waves in the wakes of the cross-stream propagation of inertial-internal waves.
storms. J. Phys. Oceanogr., 14, 1129–1151, https://doi.org/10. Geophys. Astrophys. Fluid Dyn., 6, 245–275, https://doi.
1175/1520-0485(1984)014,1129:OTBOIW.2.0.CO;2. org/10.1080/03091927509365797.
Hallberg, R., 2005: A thermobaric instability of Lagrangian verti- Morvan, M., X. Carton, S. Corréard, and R. Baraille, 2020: Sub-
cal coordinate ocean models. Ocean Modell., 8, 279–300, mesoscale dynamics in the Gulf of Aden and the Gulf of
https://doi.org/10.1016/j.ocemod.2004.01.001. Oman. Fluids, 5, 146, https://doi.org/10.3390/fluids5030146.
Halliwell, G. R., 2004: Evaluation of vertical coordinate and Müller, M., B. K. Arbic, J. G. Richman, J. F. Shriver, E. L. Kunze,
R. B. Scott, A. J. Wallcraft, and L. Zamudio, 2015: Toward
vertical mixing algorithms in the hybrid-coordinate ocean
an internal gravity wave spectrum in global ocean models.
model (HYCOM). Ocean Modell., 7, 285–322, https://doi.
Geophys. Res. Lett., 42, 3474–3481, https://doi.org/10.1002/
org/10.1016/j.ocemod.2003.10.002.
2015GL063365.
Hazewinkel, J., and K. Winters, 2011: PSI of the internal tide on a
Munk, W., and C. Wunsch, 1998: Abyssal recipes II: Energetics of
b plane: Flux divergence and near-inertial wave propagation.
tidal and wind mixing. Deep-Sea Res. I, 45, 1977–2010, https://
J. Phys. Oceanogr., 41, 1673–1682, https://doi.org/10.1175/
doi.org/10.1016/S0967-0637(98)00070-3.
2011JPO4605.1.
Nash, J. D., M. H. Alford, and E. Kunze, 2005: Estimating inter-
Hogan, T. F., and Coauthors, 2014: The Navy Global Environ-
nal wave energy fluxes in the ocean. J. Atmos. Oceanic Tech-
mental Model. Oceanography, 27, 116–125, https://doi.org/10.
nol., 22, 1551–1570, https://doi.org/10.1175/JTECH1784.1.
5670/oceanog.2014.73.
Nikurashin, M., and R. Ferrari, 2011: Global energy conversion
Jiang, J., Y. Lu, and W. Perrie, 2005: Estimating the energy flux
rate from geostrophic flows into internal lee waves in the
from the wind to ocean inertial motions: The sensitivity to
deep ocean. Geophys. Res. Lett., 38, L08610, https://doi.org/
surface wind fields. Geophys. Res. Lett., 32, L15610, https://
10.1029/2011GL046576.
doi.org/10.1029/2005GL023289. Plueddemann, A., and J. Farrar, 2006: Observations and models
Kang, D., and O. Fringer, 2010: On the calculation of available of the energy flux from the wind to mixed-layer inertial cur-
potential energy in internal wave fields. J. Phys. Oceanogr., rents. Deep-Sea Res. II, 53, 5–30, https://doi.org/10.1016/j.dsr2.
40, 2539–2545, https://doi.org/10.1175/2010JPO4497.1. 2005.10.017.
Kelly, S. M., 2016: The vertical mode decomposition of surface Pollard, R. T., and R. Millard Jr., 1970: Comparison between
and internal tides in the presence of a free surface and arbi- observed and simulated wind-generated inertial oscillations.
trary topography. J. Phys. Oceanogr., 46, 3777–3788, https:// Deep-Sea Res. Oceanogr. Abstr., 17, 813–821, https://doi.org/
doi.org/10.1175/JPO-D-16-0131.1. 10.1016/0011-7471(70)90043-4.
}}, 2019: Coastally generated near-inertial waves. J. Phys. Polzin, K. L., and Y. V. Lvov, 2011: Toward regional characteriza-
Oceanogr., 49, 2979–2995, https://doi.org/10.1175/JPO-D- tions of the oceanic internal wavefield. Rev. Geophys., 49,
18-0148.1. RG4003, https://doi.org/10.1029/2010RG000329.
Kunze, E., 1985: Near-inertial wave propagation in geostrophic Rainville, L., and R. Pinkel, 2006: Propagation of low-mode inter-
shear. J. Phys. Oceanogr., 15, 544–565, https://doi.org/10.1175/ nal waves through the ocean. J. Phys. Oceanogr., 36, 1220–
1520-0485(1985)015,0544:NIWPIG.2.0.CO;2. 1236, https://doi.org/10.1175/JPO2889.1.
}}, 2017: Internal-wave-driven mixing: Global geography and Rimac, A., 2014: The role of wind induced near-inertial waves on
budgets. J. Phys. Oceanogr., 47, 1325–1345, https://doi.org/10. the energetics of the ocean. Ph.D. thesis, Universität Ham-
1175/JPO-D-16-0141.1. burg, 117 pp., https://pure.mpg.de/rest/items/item_2075939_5/
Le Boyer, A., and M. H. Alford, 2021: Variability and sources of component/file_2075949/content.
the internal wave continuum examined from global moored }}, J.-S. von Storch, C. Eden, and H. Haak, 2013: The influence
velocity records. J. Phys. Oceanogr., 51, 2807–2823, https:// of high-resolution wind stress field on the power input to
doi.org/10.1175/JPO-D-20-0155.1. near-inertial motions in the ocean. Geophys. Res. Lett., 40,
Lee, D.-K., and P. P. Niiler, 1998: The inertial chimney: The near- 4882–4886, https://doi.org/10.1002/grl.50929.
inertial energy drainage from the ocean surface to the deep }}, }}, and }}, 2016: The total energy flux leaving the
layer. J. Geophys. Res., 103, 7579–7591, https://doi.org/10. ocean’s mixed layer. J. Phys. Oceanogr., 46, 1885–1900,
1029/97JC03200. https://doi.org/10.1175/JPO-D-15-0115.1.

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC

840 JOURNAL OF PHYSICAL OCEANOGRAPHY VOLUME 52

Sanford, T. B., B. B. Ma, and M. H. Alford, 2021: Stalling and Van Haren, H., 2005: Tidal and near-inertial peak variations
dissipation of a near-inertial wave (NIW) in an anticyclonic around the diurnal critical latitude. Geophys. Res. Lett., 32,
ocean eddy: Direct determination of group velocity and L23611, https://doi.org/10.1029/2005GL024160.
comparison with theory. J. Geophys. Res. Oceans, 126, Vic, C., and Coauthors, 2019: Deep-ocean mixing driven by small-
e2020JC016742, https://doi.org/10.1029/2020JC016742. scale internal tides. Nat. Commun., 10, 2099, https://doi.org/
Scott, R., J. Goff, A. Naveira Garabato, and A. Nurser, 2011: 10.1038/s41467-019-10149-5.
Global rate and spectral characteristics of internal gravity Watanabe, M., and T. Hibiya, 2002: Global estimates of the wind-
wave generation by geostrophic flow over topography.
induced energy flux to inertial motions in the surface mixed
J. Geophys. Res., 116, C09029, https://doi.org/10.1029/
layer. Geophys. Res. Lett., 29, 1239, https://doi.org/10.1029/
2011JC007005.
Shakespeare, C. J., 2020: Interdependence of internal tide and 2001GL014422.
lee wave generation at abyssal hills: Global calculations. Whalen, C. B., J. A. MacKinnon, and L. D. Talley, 2018: Large-
J. Phys. Oceanogr., 50, 655–677, https://doi.org/10.1175/ scale impacts of the mesoscale environment on mixing from
JPO-D-19-0179.1. wind-driven internal waves. Nat. Geosci., 11, 842–847, https://
Simmons, H. L., and M. H. Alford, 2012: Simulating the long- doi.org/10.1038/s41561-018-0213-6.
range swell of internal waves generated by ocean storms. Xie, X.-H., X.-D. Shang, H. van Haren, G.-Y. Chen, and Y.-Z.
Oceanography, 25, 30–41, https://doi.org/10.5670/oceanog. Zhang, 2011: Observations of parametric subharmonic insta-
2012.39. bility-induced near-inertial waves equatorward of the critical
Singh, A. K., and B. Bhadauria, 2009: Finite difference formulae diurnal latitude. Geophys. Res. Lett., 38, L05603, https://doi.
for unequal sub-intervals using Lagrange’s interpolation for- org/10.1029/2010GL046521.
mula. Int. J. Math. Anal., 3, 815–827. Young, W., and M. B. Jelloul, 1997: Propagation of near-inertial
Thomas, L. N., L. Rainville, O. Asselin, W. R. Young, J. Girton,
oscillations through a geostrophic flow. J. Mar. Res., 55, 735–
C. B. Whalen, L. Centurioni, and V. Hormann, 2020: Direct
766, https://doi.org/10.1357/0022240973224283.
observations of near-inertial wave z-refraction in a dipole
Zaron, E. D., and G. D. Egbert, 2014: Time-variable refraction of
vortex. Geophys. Res. Lett., 47, e2020GL090375, https://doi.
org/10.1029/2020GL090375. the internal tide at the Hawaiian ridge. J. Phys. Oceanogr.,
Trossman, D. S., B. K. Arbic, S. T. Garner, J. A. Goff, S. R. Jayne, 44, 538–557, https://doi.org/10.1175/JPO-D-12-0238.1.
E. J. Metzger, and A. J. Wallcraft, 2013: Impact of parameter- Zhai, X., R. J. Greatbatch, C. Eden, and T. Hibiya, 2009: On the
ized lee wave drag on the energy budget of an eddying global loss of wind-induced near-inertial energy to turbulent mixing
ocean model. Ocean Modell., 72, 119–142, https://doi.org/10. in the upper ocean. J. Phys. Oceanogr., 39, 3040–3045, https://
1016/j.ocemod.2013.08.006. doi.org/10.1175/2009JPO4259.1.

Unauthenticated | Downloaded 02/24/24 04:24 PM UTC