Dynamical Potential Energy: A New Approach to

FEBRUARY 2013
457
ROQUET
Dynamical Potential Energy: A New Approach to Ocean Energetics
FABIEN ROQUET
Meteorological Institute of the Stockholm University, Stockholm, Sweden
(Manuscript received 6 June 2012, in final form 15 October 2012)
ABSTRACT
The concept of available potential energy is supposed to indicate which part of the potential energy is
available to transform into kinetic energy. Yet it is impossible to obtain a unique definition of available
potential energy for the real ocean because of nonlinearities of the equation of state, rendering its usefulness
largely hypothetical. In this paper, the conservation of energy is first reformulated in terms of horizontal
anomalies of density and pressure for a simplified ocean model using the Boussinesq and hydrostatic approximations. This framework introduces the concept of ‘‘dynamical potential energy,’’ defined as the horizontal anomaly of potential energy, to replace available potential energy. Modified conservation equations
are derived that make it much simpler to identify oceanic power input by buoyancy and mechanical forces.
Closed budgets of energy are presented for idealized circulations obtained with a general circulation model,
comparing spatial patterns of power inputs generated by wind and thermal forcings. Finally, a generalization of the framework to compressible fluids is presented, opening the way to applications in atmosphere
energetics.
1. Introduction
The concept of energy has historically played a major
role in the development of science, as it is one of the
most fundamental, general, and significant principles
of physical theory (Callen 1985). Applied to the study of
physical oceanography, the two fundamental laws of
thermodynamics (respectively, conservation of energy
and maximization of entropy under specific conditions)
have the potential to provide a deep understanding of
what shapes the observed large-scale circulation for a
given distribution of mechanical and thermohaline forcings. In particular, energy considerations should help to
determine if ocean circulation is primarily forced thermally (i.e., by imposed density contrasts) or mechanically
(from external sources of momentum).
In practice, the relative contribution of wind, buoyancy fluxes, and internal mixing in setting the general
circulation remains difficult to assess objectively. The
wind forcing is an essential driver of the circulation
through Ekman pumping, while buoyancy forcing is a
condition sine qua non for the existence of an interior
stratification (e.g., Gnanadesikan 1999). However,
Corresponding author address: Fabien Roquet, Meteorological
Institute of the Stockholm University, Stockholm, Sweden.
E-mail: [email protected]
DOI: 10.1175/JPO-D-12-098.1
Ó 2013 American Meteorological Society
buoyancy forcing (i.e., heat and freshwater fluxes)
alone is not efficient at driving an interior circulation
(Sandström 1908), thus many authors have argued that
a mechanical source of energy was needed to sustain the
observed deep circulation, most likely related to wind
and tidal mixing (Munk and Wunsch 1998; Huang 1999;
Wunsch and Ferrari 2004). Ekman pumping is not very
efficient either as follows: of roughly 60 TW of wind
power input, only about 1 TW powers the interior circulation by Ekman pumping, with the rest of it going
into waves and turbulence (e.g., Wang and Huang 2004).
To make further progress, a refined understanding of
ocean energetics is needed, preferentially at the local
scale.
A major difficulty in ocean energetics lies in the definition of the gravitational potential energy (PE),
EP [ rgz,
(1)
where r is the in situ density, g is the local gravity, and z
is the height relative to an arbitrary reference level. This
definition fails to take into consideration that (i) the
relative variations of density are very small in the ocean
and (ii) mass conservation imposes global constraints on
rates of change of PE: any ascending mass flux is always
largely compensated by a descending mass flux elsewhere (when using the Boussinesq approximation, this
458
JOURNAL OF PHYSICAL OCEANOGRAPHY
latter constraint becomes one about volume conservation and vertical velocities as will be seen later). The
concept of available potential energy (APE) was introduced precisely so as to overcome these limitations
in the PE definition.
APE is defined as the difference between the total
potential energy and the minimum total potential energy, which could result from any adiabatic redistribution
of mass (Lorenz 1955), corresponding to the fraction of
PE that would be released in the ocean circulation while
reaching adiabatically a resting hydrostatic state. APE
is nowadays a widely used concept in the study of ocean
energetics (e.g., Oort et al. 1989; Winters et al. 1995;
Toggweiler and Samuels 1998; Huang 1998; Tailleux
2009; Hughes et al. 2009). However, APE is not uniquely
defined for a realistic equation of state because of the
nonlinear dependence of seawater density on both temperature and salinity (see Tailleux 2012, for a review).
This limitation is a serious one as nonlinear effects of
the equation of state are important in the ocean (e.g.,
Gnanadesikan et al. 2005; Iudicone et al. 2008; Klocker
and McDougall 2010; Hieronymus and Nycander 2013).
Algorithms have been proposed to diagnose APE in
numerical simulations with a realistic equation of state
(Huang 2005; Ilicak et al. 2012); however, it is unclear
whether the APE reference state used in these methods
is actually a state of minimum PE.
Here a modified framework is proposed for the discussion of ocean energetics, based on what will be called
dynamical potential energy (DPE). For a model using
Boussinesq and hydrostatic assumptions with a linear
equation of state, DPE is simply defined as the horizontal anomaly of PE, that is, the difference between PE
and its horizontal average vertical profile. While the
justification of DPE relies heavily on the applicability of
the hydrostatic approximation, Boussinesq approximations and linearity of the equation of state can be relaxed
provided that definitions of PE and DPE are suitably
modified.
After introducing the concept of DPE in section 2,
a quick comparison with APE will be proposed in section 3. The modified framework will be derived in section 4. To demonstrate the utility of this new energy
framework, local balances of kinetic energy (KE) and
DPE will be presented in section 5, comparing the energetics of three simulated circulations differing only in
their surface forcings—wind forcing alone, thermal
forcing alone, and wind and thermal forcings together.
A discussion of the origin and nature of the power
converted from PE to KE will be proposed in section 6.
Finally, methods to generalize the DPE framework to
cases with nonlinear equation of state and compressible
fluids are described in section 7.
VOLUME 43
2. Introduction of dynamical potential energy
Roquet et al. (2011), discussed the wind-provided input of KE to the large-scale circulation using Ekman
theory. Two different representations were used, mapping the following.
(i) The direct rate of wind work on the geostrophic
circulation, obtained from scalar product of the
wind stress and surface geostrophic velocity (Stern
1975), t o ug,o, suitably time and space averaged.
(ii) The input of pressure work into the ocean interior,
as the product of (minus) the surface pressure
elevation times the vertical velocity induced by
Ekman pumping, 2po*wEk , again suitably averaged.
The surface pressure elevation is proportional to the
sea surface height defined relative to the global
mean sea level, po* 5 ro gh, and the Ekman pumping
velocity, wEk, is related to the curl of the wind stress.
Although the two representations have markedly
different spatial patterns, their integrals over the global
ocean are strictly equal for a steady-state ocean—provided
that the horizontal Ekman transport vanishes across
coastlines. The direct input of wind work to the geostrophic circulation (first view) is redistributed laterally
into the Ekman layer before entering the geostrophic
interior as an input of pressure work (second view).
Because this input of pressure work is proportional
to the horizontal anomalies of sea surface height (i.e.,
sea surface height relative to global mean sea level), it
can be reinterpreted as the power locally generated by
Ekman pumping to maintain the local anomaly of sea
surface height relative to its equilibrium—which is global
mean sea level.
The flux of pressure work is in fact simply expressed as
a flux of pressure, highlighting the role of pressure in
transporting energy in the ocean interior. Pressure has
the same dimension as an energy per unit volume:
N m22 [ (N m) m23 5 J m23, and a flux of pressure
indeed represents a power input per unit area. Pressure,
and more specifically its horizontal anomalies, appear
as central to understanding oceanic energetics.
The framework developed in this paper is in part
a generalization of Roquet et al. (2011), exploring the
role of horizontal anomalies of pressure and density.
For the sake of clarity, we will now focus on a simplified
ocean model using the Boussinesq and hydrostatic approximations (e.g., Vallis 2006), with a linear equation
of state. Introduce the notation
T(x, y, z, t) 5 T(z, t) 1 T*(x, y, z, t)
(2)
for any tracer T. The horizontal mean T is a function of
depth and time
FEBRUARY 2013
T(z, t) 5
1
A(z)
ðð
T(x, y, z, t) dx dy ,
(3)
ÐÐ
where A(z) 5
dx dy is the ocean area at depth z.
The hydrostatic balance can be rewritten by separating horizontal mean and horizontal anomalies of both
pressure p and density r
›p*
›p
5 2r*g 2
1 rg .
›z
›z
(4)
The two last terms are bracketed, as they nearly cancel
(exact cancellation with vertical sidewalls) and are both
constant in the horizontal. We will see later how a new
pressure variable can be defined, the so-called dynamical pressure, to account for these two horizontal mean
contributions. For now, note only that a reduced horizontal hydrostatic balance can be obtained, where horizontal anomalies of the gravity term nearly balance the
vertical gradient of pressure anomalies.
Define the DPE as the horizontal anomaly of PE
EP* [ r*gz .
(5)
Note that this definition applies only for an idealized
fluid assuming Boussinesq and hydrostatic approximations and a linear equation of state (or at least an equation
of state independent of depth), yet an equivalent definition can be obtained for non-Boussinesq fluids and nonlinear equations of state (see section 7).
The rate of conversion of DPE to KE is 2r*gw.
Compared to the rate of conversion from PE to KE
(2rgw), this expression proves advantageous as follows.
(i) For a resting fluid with flat isopycnals, 2rgw increases greatly with depth as it is proportional to
the full hydrostatic pressure gradient, while 2r*gw
simply vanishes everywhere.
(ii) From volume conservation (using Boussinesq assumption), the horizontal mean vertical velocity is
always exactly zero, w 5 0. This is why global integrals of both forms of conversion rates are always
strictly equal,
ððð
459
ROQUET
2r*gw dV 5
ððð
2rgw dV .
(6)
In other words, the rate of conversion of DPE to KE is
more readily associated with the observed stratification,
while representing the same global amount of conversion as does the more conventional definition. This
property constitutes the main justification for focusing
on DPE to study ocean energetics.
3. Comparison with available potential energy
The difference in definition between APE and DPE
can be best viewed when comparing their global integrals over a closed ocean basin. The global reservoir of
APE is always strictly positive for a stable stratification,
except when isopycnals are flat at every depth. In contrast, for DPE, the global reservoir is always exactly
zero,
ððð
(7)
EP* dV 5 0.
A simple example of DPE and its contrast with APE
is shown in Fig. 1a, for a density distribution resembling the idealized pycnocline model proposed by
Gnanadesikan (1999). A thermocline separates a shallow, light lens of water from a deeper and denser water
mass. The densest water mass is able to reach the surface
in the right corner of the domain, creating a horizontal
gradient of density where the thermocline is vertical.
Reference profiles of density used to calculate both DPE
and APE are shown in Fig. 1b, with the resulting density
anomalies used to calculate DPE and APE in Figs. 1c and
1d, respectively.
Because DPE always integrates to zero globally, a
more useful measure of the magnitude of regional anomalies is the integral over the regions of positive DPE,
DPE1 [
ððð
E*.0
P
EP* dV 5
ððð
jEP*j
dV .
2
(8)
The global size of the APE reservoir is
ðð
APE 5 (1 2 b)L
5 r0 g
ð0
abH
r0 gz dz 2 bL
ð abH
aH
r0 gz dz (9)
b(1 2 b)
L(aH)2 ,
2
(10)
where a is the ratio of light water-layer depth to fulldepth H, b is the ratio of light-layer length to length of
the basin L, and r0 is the difference of density between
the two layers.
In this example, it is found that the global reservoir of
positive DPE values is strictly identical despite qualitative differences in the distribution of density anomalies,
DPE1 5 bL
ð0
aH
2(1 2 b)r0 gz dz 5
ðð
APE.
Note that equality between global sizes of DPE1 and
APE reservoirs will not hold in the general case as, for
example, introducing a thermocline tilt destroys it. Yet,
the example presented here indicates that they should
460
JOURNAL OF PHYSICAL OCEANOGRAPHY
VOLUME 43
FIG. 1. (a) Comparison of DPE and APE for a simple density distribution is shown. (b) Reference vertical profiles
used to compute DPE and APE are shown. Density anomalies obtained as the difference between the density
distribution and the reference profile are presented for (c) DPE and (d) APE. The global integral of DPE is exactly
zero, while it is strictly positive for APE. However, the global integral of positive values of DPE (DPE1) is strictly
equal to the global integral of APE (not true in general).
be comparable for realistic distributions of mass in the
ocean (in the model study cases presented later, we
found differences of order 20%, see Table 1).
Here, it must be emphasized that APE and DPE are
very different in nature. APE is fundamentally a thermodynamic concept, making use of an adiabatic rearrangement of fluid parcels to define a state of reference.
It relies on the weakness of interior mixing to provide
a valuable estimate of the fraction of PE available to
the general circulation. In contrast, DPE is a purely
dynamical concept finding its justification in the predominance of the hydrostatic equilibrium in the vertical
momentum balance, allowing removal of the large static
contribution to the hydrostatic pressure in the dynamical equations. Contrary to the situation with APE, no
consideration of water mass mixing or adiabatic fluid
motion is involved in the definition of the reference state
of DPE.
APE works well for adiabatic processes, such as wave
propagation, baroclinic instability, or the effect of wind
pumping on the stratification, since the reference state is
then constant through time. However, when mixing is
considered, the usefulness of APE becomes questionable as its reference state is constantly changing with
time and rarely corresponds to the resting state that
would eventually reach the system by switching off all
forcings. With a nonlinear equation of state, the APE
reference state is not even well defined. Like in the case
TABLE 1. Size of energy reservoirs and rates of work integrated
globally in the WIND, RELAX, and MIXED experiments. Abbreviations for reservoirs are as in the core text. For the total rates
of work, WIND is the surface input of frictional power, CONV is
the total conversion rate, FRICTION is the total rate of frictional
work minus the WIND input, and CORIOLIS is for the Coriolis
term, which is slightly working in the model. The total KE trend
(sum of the four working rates) is negligible, as circulations are in
equilibrium.
WIND
KE
PE
DPE1
APE
WIND
CONV
FRICTION
CORIOLIS
RELAX
Reservoir size (PJ)
0.77
3.79
27.6 3 108
27.6 3 108
0
8.8 3 103
0
6.3 3 103
Total rate of work (GW)
1.73
0
0
17.99
21.71
220.03
20.02
2.04
MIXED
7.87
27.6 3 108
10.3 3 103
8.0 3 103
25.89
10.42
239.13
2.82
FEBRUARY 2013
461
ROQUET
of APE, the DPE reference state is varying in time in the
general case of an unbalanced circulation, yet in practice, DPE is much easier to diagnose and well defined
for any equation of state (although its definition has to
be slightly changed for a nonlinear equation of state,
see section 7). The concept of DPE will now be used to
derive a modified framework for the study of energy
conservation.
DEP
5 rgw 1 WD ,
Dt
(17)
Consider now the conservation of energy for a simplified ocean model using the Boussinesq and hydrostatic
approximations (e.g., Vallis 2006), assuming that the
equations of state is linear.
where EP 5 rgz is the potential energy and WD 5 Dgz
represents the rate of production of PE, that is, sources
and sinks of PE related to diabatic processes. The reference level z 5 0 is arbitrary—taken at the surface in
the following.
Equations (16) and (17) define the energetics for an
hydrostatic Boussinesq ocean model. Note that the
global constraint of volume conservation has not been
used to derive these conventional equations. The continuity equation was only used locally to transform the
advection operator into a flux divergence. We turn now
to formulating similar equations using DPE.
a. Standard formulation of energetics
b. Modified KE balance
Equations of motion are standard, for momentum
conservation, the hydrostatic assumption, continuity, and
density conservation
To replace the PE to KE conversion term by a DPE to
KE conversion term, expressed as 2r*gw in the equations of energy conservation, the ‘‘dynamical pressure’’
variable is defined as follows:
4. Derivation of the framework
ro
Duh
1 ro f k 3 uh 5 2$h p 1 F ,
Dt
(11)
(12)
$ u 5 0, and
(13)
(14)
with
D/Dt [ ›/›t 1 u $.
(15)
Here, u 5 (uh, w) is the velocity vector, ro is a constant
reference density, f is the Coriolis frequency, and F is
the frictional force. Here, D is the local rate of production of density, representing sources and sinks of
density due to diabatic processes, such as diffusion, convection, and surface density fluxes.
The standard equation for KE is obtained as the scalar
product of the momentum equations with the velocity
vector
DEK
5 2$ ( pu) 2 rgw 1 WF ,
Dt
ð0
z
›p
5 2rg,
›z
Dr
5 D,
Dt
pdyn 5 p* 2
(16)
where EK [ ro juh j2 /2 is the hydrostatic kinetic energy.
The term WF 5 uh F represents the rate of work of
frictional forces. Note that the pressure divergence term
includes both horizontal and vertical pressure fluxes.
The equation for PE is obtained from the multiplication of the density Eq. (14) by gz,
›p
1 rg dz,
›z
(18)
where z 5 0 corresponds to the surface level, using
a rigid lid assumption for simplicity. The vertical axis is
directed upward, so z # 0 in the interior. A method to
account for a linear free surface is presented in appendix
B. It suffices to note here that it does not fundamentally
modify the results presented below. The energetics of
a fully nonlinear free surface is non trivial, so it is left
aside here.
For a basin with vertical sidewalls, dynamical pressure
is simply the horizontal anomaly of pressure. In the
more general case of sloping bottom topography, the
two expressions differ by a function of depth only (which
is typically small compared to the magnitude of pressure
anomalies). From Eq. (4), dynamical pressure has the
same horizontal gradient as the full pressure, and satisfies a reduced hydrostatic balance
$h pdyn 5 $h p,
›pdyn
5 2r*g
›z
(19)
producing the modified KE equation
DEK
5 2$ (pdyn u) 2 r*gw 1 WF ,
Dt
(20)
which is very similar to the conventional form in Eq.
(16), except that it employs only anomalies about the
horizontal average, with the horizontal mean contribution to the hydrostatic gradient suppressed.
462
JOURNAL OF PHYSICAL OCEANOGRAPHY
c. DPE balance
In the PE Eq. (17), a near cancellation is found between horizontal mean contributions of conversion and
the PE advection terms
›r
rgw 2 $ (EP u) 5 2gz w
5S.
›z
(21)
The residual S is named the ‘‘stratification term,’’ as
it is related to the horizontal mean stratification. The
stratification term has neither the form of a conversion
nor a flux divergence, and looks like a source or sink of
PE, although physically, there is no corresponding local
source or sink. Note that S is proportional to w at each
depth level, so its horizontal mean always vanishes. This
term represents a transfer of power between different
areas on the same depth level, redistributing horizontally sources and sinks of PE.
Define the rate of production of DPE as the sum of the
PE production rate and the stratification term
›r
0
.
5 WD 1 S 5 gz D 2 w
WD
›z
(22)
The rate of production of DPE depends on a competition between local rates of production of density by
mixing and diffusion processes, and the vertical advection of the horizontal mean density.
PE balance becomes
›EP
0
1 $ (EP*u) 5 r*gw 1 WD
.
›t
(23)
Note that the time derivative term involves the full PE,
but the flux divergence term operates on DPE only, as
advection of horizontal mean PE is embedded in the
stratification term.
The DPE framework is made of Eqs. (20) and (23).
The global integral of this system of equations is identical to the global integral of the standard Eqs. (16) and
(17). More interestingly, their respective horizontal
means are also identical. Differences appear only when
comparing their local expressions, with advantage in the
modified formulation for mapping and interpreting the
energy contributions. The model study in the next section will further illustrate this important point.
The case of a hydrostatic Boussinesq ocean model
using a nonlinear equation of state will be presented in
section 7. The major difference with the linear case is in
the definition of PE (Young 2010; Nycander 2011). Otherwise, equations for energy conservation are strictly
identical. The case of a compressible hydrostatic model
using a nonlinear equation of state will also be treated,
VOLUME 43
demonstrating that only the hydrostatic assumption is
required to derive this framework.
5. Energy diagnostic of MITgcm simulations
To illustrate the use of the DPE framework, a lowresolution idealized configuration of the Massachusetts
Institute of Technology general circulation model
(MITgcm) (Marshall et al. 1997) is used, consisting of
a Northern Hemisphere basin spanning 458 latitude and
608 longitude over 3 km in depth (40 3 48 3 15 grid
points). Sidewalls are vertical from the surface to 1200 m,
then have a 2% slope until they reach a flat bottom at
3000-m depth—see Hughes and DeCuevas (2001), to see
why the slope matters to obtain a realistic quasi-inviscid
circulation. Boundary conditions are no-slip on the
sidewalls and bottom, with a linear free surface; no
mixed-layer scheme is used; simple friction and diffusion
schemes (no eddy parameterization mixing scheme) are
employed.
A linear equation of state is used, that depends only
on temperature. Because density and temperature variations are proportional, the numerical model satisfies
equations presented in section 4, with
r 5 ro (1 2 aT) ,
(24)
F 5 Az ›2z u 1 Ah Dh u, and
(25)
D 5 2ro a(kz ›2z T 1 kh Dh T) ,
(26)
with ro 5 103 kg m23, a 5 2 3 1024 8C21, Ah 5 5 3
103 m2 s21, Az 5 10 m2 s21, kh 5 400 m2 s21, and kz 5
1024 m2 s21, increased to 1 m2 s21 in case of static
instability.
Three experiments were performed, where the model
was run from a resting homogeneous state to near equilibrium (2000 years) using different configurations of
forcing fields (see Fig. 2a). In the WIND experiment, a
sinusoidal zonal wind stress field was applied at the surface, with a 0.1 N m22 maximum value at 358N. In the
RELAX experiment, the surface temperature was relaxed toward a meridional profile of reference ranging
from 58C in the north to 208C in the south, with a restoring time scale of 30 days. Wind stress and temperature relaxation were applied together in the MIXED
experiment, mimicking North Atlantic conditions.
a. Overview of simulated circulations
The balanced state of the WIND experiment consists
in a classic double-gyre circulation with a western intensification of currents forming a northeast jet between
the two gyres (Fig. 3a). In this experiment, the most
FEBRUARY 2013
463
ROQUET
FIG. 2. (a) Meridional distribution of wind stress (gray curve) and
temperature relaxation profile (black curve) used in the MITgcm
simulations. (b) Global-mean stratification of the balanced state
for the three experiments.
notable feature is that the ocean is completely unstratified,
permitting simplifications in the energy equations. In
particular, r* 5 0 everywhere. Consequently, the dynamical pressure is constant with depth, as expected for a
barotropic circulation, and the rate of DPE to KE conversion, as well as all terms in the DPE equation, are
identically null.
The circulation simulated in the RELAX experiment
corresponds to what Stern (1975) called horizontal
convection (see review of Hughes and Griffiths 2008),
where warm and cold sources are applied on the surface
geopotential. A pattern of density anomalies is generated at the surface, requiring the existence of a largescale circulation to establish a stationary state. Although
the RELAX circulation features a double-gyre structure
like in the WIND experiment (cf. Figs. 3a and 3c), numerous fundamental differences exist between the two
circulations. In particular, surface heat fluxes induce a
subsurface thermocline separating the surface-intensified
gyres from a slower deep reaching flow (see Fig. 2b
and 3d).
A strong asymmetry now occurs between the northern
and southern gyres. The northern gyre presents a zonally
elongated shape that follows closely the northern boundary of the domain. By contrast, the southern gyre is less
intense, shallower but more wide-spread, with its western boundary current touching the western wall. Overall, the circulation induced in the RELAX experiment is
far more intense than in the WIND experiment: the total
reservoir size of KE is 3.8 PJ, 5 times larger than the
0.8 PJ reservoir in the WIND case (see Table 1). This occurs because KE is quadratic in the velocity, so even with
a similar depth-integrated transport, a surface intensified
circulation is more energetic than a barotropic one.
Overall, the MIXED experiment strongly resembles
the RELAX experiment, with a similar circulation
pattern consisting in two asymmetric gyres. However,
MIXED circulation is surface-intensified because of wind
forcing, resulting in a KE reservoir with a doubled size
compared to RELAX (7.9 PJ). The thermocline is slightly
deeper than in the RELAX case (see Fig. 2b and 3f), seen
as a 20% increase of the DPE1 reservoir size. This
deepening is consistent with wind stress suction in the
north, and pumping in the south, thus accentuating the
magnitude of horizontal anomalies of density.
The size of the APE reservoir can be easily estimated
for the three experiments because the equation of state
is fully linear. The APE density reference profile is obtained by sorting water parcels by density. In section 3, it
was argued that APE and DPE1 reservoirs should
generally be of comparable size. Indeed, typical differences are of 20% in the simulations, with a similar increase
between the RELAX and the MIXED experiments using
either of the definitions.
b. Energetics of the WIND experiment
The steady-state balance of KE in the WIND simulation consists of a simple balance between the divergence
of dynamical pressure fluxes and the work of frictional
forces
$ ( pdyn u) 5 WF ,
(27)
where advection of KE is neglected, as it is typically one
order of magnitude smaller than other terms.
Globally, 1.7 GW of wind stress power input enters at
the surface, which is then removed in the interior
through friction (Table 1). A small contribution comes
from the Coriolis term in the model, as its discretized
form is not perfectly orthogonal to the velocity vector.
The horizontal mean budget is shown in Fig. 4, including
the vertical profile of dynamical pressure flux (Fig. 4a),
464
JOURNAL OF PHYSICAL OCEANOGRAPHY
VOLUME 43
FIG. 3. Distribution of (left) dynamical pressure and (right) horizontal density anomalies for the three experiments.
Positive (negative) anomalies are given in red (blue). For each panel, the value of isosurfaces is given in the subtitle
(same value used for the surface contour interval). For example, (a) the 60.01 dbar isosurfaces of dynamical pressure
for the WIND experiment are shown, and surface contours are superimposed with a 0.01 dbar contour interval.
whose vertical gradient balances the frictional work
(Fig. 4b). This budget is exactly the same when considering the standard form of energy equations or the
modified one, as pdyn w 5 pw and r*gw 5 rgw (50 in
this case).
The value of the modified framework appears when
considering the local budget. Because the DPE to KE
conversion term is zero everywhere, the local description of energetics is greatly simplified. Frictional forces
work in two main areas (Fig. 5a). Sources of KE are
found in the first vertical level of the model, and sinks
are found at every depth below, mainly along the
southern and western boundaries where the currents are
strongest.
Power is transported between wind-driven sources
and dissipative sinks as fluxes of dynamical pressure,
moving downward while rotating around the two gyres.
Although not shown here, the precise region of dissipation lies close to regions of frictional work as typical
distances over which viscous stress fluxes are transported
FEBRUARY 2013
ROQUET
465
FIG. 4. Horizontally integrated budget of energy ÐÐ
for (a),(b) WIND, (c),(d)
RELAX, and (d),(e) MIXED experÐÐ
Horizontally
iments. (left) Vertical fluxes of dynamical pressure ( pdyn w)ÐÐand DPE ( Ep* w) are presented. (right)
ÐÐ
integrated rates of work for (friction)
the
frictional
forces
(
W
),
(conv)
the
conversion
term
2r*gw,
and (PE
F
ÐÐ
prod)
ÐÐ the PE production term ( WD ) are shown. Note that the stratification term has no contribution on the horizontal
( S 5 0). Advection of KE is not shown, because it is a small contribution that can safely be neglected here.
before being dissipated are small compared to ocean
basin scales, along both vertical and horizontal directions
(see appendix A).
Figure 6 compares the theoretical distribution of inputs of pressure work induced by Ekman pumping as
defined by Roquet et al. (2011) with vertical fluxes of
dynamical pressure (i.e., horizontal anomalies of pressure
in this case) taking place below the first vertical level of
the model, that is at a 50-m depth. As expected, the two
distributions compare favorably, although numerical noise
at the grid scale is apparent in the model. Note that the
comparison is successful because dynamical pressure is
considered instead of full hydrostatic pressure. Otherwise, the large hydrostatic contribution to the pressure
466
JOURNAL OF PHYSICAL OCEANOGRAPHY
VOLUME 43
FIG. 5. Distribution of (left) rates of frictional work and (right) rates of DPE into KE conversion for (a),(b) WIND,
(c),(d) RELAX, and (e),(f) MIXED experiments. Regions of power sources are connected to regions of power sinks
by large-scale fluxes of dynamical pressure (see Fig. 3), thus following geostrophic contours on the horizontal.
field would indeed hide correlations existing between
sea surface height and subsurface isobars.
c. Energetics of the RELAX experiment
1) KE BALANCE
In the RELAX experiment, power comes necessarily
from the conversion of DPE into KE, as there is no
surface wind stress
$ ( pdyn u) 5 2r*gw 1 WF ,
(28)
where advection of KE is again neglected. The resulting
circulation is more vigorous than in WIND, both in
terms of KE reservoir size and rates of work. The total
conversion rate (18.0 GW, see Table 1) is 10 times larger
than the wind-power input in the WIND circulation.
The horizontally integrated budget shows a similar
balance to the WIND case, except that the source of KE
is now provided through the DPE to KE conversion rate
(see Fig. 4). Horizontal mean conversion is positive at all
depths (Fig. 4d), while the frictional power is negative.
The vertical flux of dynamical pressure, which transports
FEBRUARY 2013
ROQUET
467
FIG. 6. Comparison of spatial patterns of (left) upward Ekman-induced inputs of pressure work as defined by
Roquet et al. (2011) (po*wEk , see section 2), and (right) dynamical pressure flux below the first vertical level for the
three model experiments ([pdynw]50m). The two patterns compare favorably for (top) the WIND case. (middle) In the
RELAX case, the buoyancy-driven flux of dynamical pressure is mainly localized along coastal boundaries. (bottom)
The MIXED case pattern is the superposition of an Ekman-induced input with a buoyancy-driven pattern as in the
RELAX case. Bathymetric contours are superimposed (doted lines).
power between sources and sinks, reverses direction at
about 500-m depth, from upward above to downward
below (Fig. 4c).
Spatial distributions of frictional work and DPE to
KE conversion rates are shown in Figs. 5c and 5d, respectively. Areas of large DPE to KE conversions are
confined to thin regions along western, northern, and
eastern boundaries, extending deep in the water column,
up to 1000 m in the weakly stratified northern gyre (see
Fig. 5d). DPE to KE conversion rates are locally of
both signs, as the vertical velocity shows. Two surfaceintensified regions of dissipation are found, the first south
468
JOURNAL OF PHYSICAL OCEANOGRAPHY
of the western boundary and the second north of the
eastern boundary. Another less active region of dissipation is found at depth along the northern flank of
the basin, associated with the deep circulation of the
northern gyre.
2) DPE BALANCE
The DPE to KE conversion rate results from a balance involving the divergence of DPE fluxes and the
DPE production rate
0
.
$ (EP*u) 5 r*gw 1 WD
(29)
Taking its global integral, it appears that the total
conversion rate is powered by the total DPE production
rate, equal to the total PE production rate (the global
integral of the stratification term is always zero). The
horizontally integrated budget indicates that DPE is
produced in the interior (see PE production in Fig. 4d),
and is then advected upward before being converted
(see Fig. 4c). As for KE, the horizontally integrated
DPE balance is strictly identical to the standard PE one
(EP w 5 EP*w and WD 5 WD0 ).
Consider now the local expression of DPE conservation. The spatial distribution of zonally integrated contributions to the DPE balance are presented in Fig. 7
(left panels). The PE production term is positive wherever diffusion acted to increase the local temperature,
mainly within the southern thermocline. In contrast,
it is negative below the convectively mixed layer in
the northern part of the basin. The stratification term
is locally of the same order of magnitude as the PE
production term, nearly balancing the positive production rates of PE in the thermocline. Recall that the
horizontal mean of the stratification term is always
zero, thus in effect, this term induces a large horizontal
redistribution of the PE produced by diffusion and
mixing, concentrating most of it along the northern
boundary.
Distributions of conversion and DPE production
terms are strikingly similar, especially when compared
to the distribution of PE production areas. A large part
of the conversion occurs in the northernmost sector of
the basin. Positive conversion rates are also found along
the western boundary (see Fig. 5d), powering the weak
southern gyre. This somewhat counterintuitive result
illustrates the importance of the stratification term in
redistributing horizontally PE production before its actual conversion. Although PE is largely produced in
stratified areas, where diffusion and mixing erode efficiently local density gradients, most of the conversion
occurs in weakly stratified areas mainly because the
largest vertical velocities are observed there.
VOLUME 43
d. Energetics of the MIXED experiment
One striking feature of the MIXED experiment is
that, although it is forced with the same wind stress
distribution as the WIND experiment, it has an input of
wind power 15 times larger (see Table 1). Added to the
power input from DPE to KE conversion, the MIXED
circulation receives twice as much power input as does
the RELAX experiment. This illustrates the complexity of energy diagnostics, where wind and buoyancy forcings strongly interact to generate the observed
circulation.
To first-order, the heat forcing seems to control the
large-scale distribution of density, as the MIXED stratification is very similar to the RELAX one. The wind
forcing controls the intensity of the surface circulation.
Because of the stratification, the wind forcing acts on
a shallower layer than in the baroclinic case, thus able
to produce faster currents, hence the larger wind-power
inputs and increased size of the KE reservoir.
The DPE to KE conversion rate is modulated by the
presence of wind (cf. Figs. 5d and 5f) as Ekman pumping
modifies the depth of isopycnals. Bands of negative conversion rates are seen at the surface in both gyres as
pumping is directed upward in the northern gyre, and
downward in the southern one. As a result, the horizontalmean conversion rate is negative near the surface (Fig.
4f), and the total conversion rate is almost halved compared to the RELAX case (see also Fig. 7, right panels).
Using a depth-density streamfunction to analyze the
global overturning circulation, Nycander et al. (2007)
found that the subsurface circulation was on average
thermally indirect, that is, powered by wind, and the
deeper circulation was thermally indirect (i.e., driven by
heat conduction and small-scale mixing) in a similar
simulation. This is consistent with our findings, showing
how Ekman pumping has two separate contributions on
the KE balance: a direct frictional power input, and an
indirect effect on conversion rates through the vertical
displacement of isopycnals. In the MIXED experiment,
these two effects are acting against each other, yet this is
not necessary the case.
This simulation illustrates why the common practice
of describing separately a thermohaline circulation and
a wind-driven one in the ocean has some justification,
although this separation is not exact. Wind and heat
forcings act differently on the ocean, producing distinct
patterns of circulation. This is clearly seen in the distribution of vertical fluxes of dynamical pressure (see Fig.
6). The MIXED pattern is made of the superposition
of a wind-driven Ekman pattern and a buoyancy-driven
pattern similar to the one in the RELAX case. Again,
the wind-driven pattern of pressure flux compares well
FEBRUARY 2013
ROQUET
469
FIG. 7. Zonally integrated distributions of the different contributions to the DPE balance diagnosed for (left) the
RELAX and (right) MIXED experiments: (top to bottom) the PE production rate WD , the stratification term S, the
DPE production rate WD0 5 WD 1 S, and the conversion rate 2r*gw.
with the theoretical expression discussed in Roquet et al.
(2011). However, this pattern is significantly different
from the one in the WIND case, although the wind forcing
is the same. This is another illustration of the leading role
of stratification on the circulation.
6. Conversion and mixing energy supply
The total conversion rate is the only contribution to the
PE balance that does not depend on the reference level
used to define PE, but it is not directly related to the
distribution of mixing and surface buoyancy fluxes. On
the other hand, the rate of production of PE has a
straightforward relation to mixing and diffusion, but its
expression does depend on the reference level. However, there is a strong relationship between the conversion rates and PE production rates that will be
discussed here.
For a balanced state, the total conversion and total PE
production are strictly equal
470
JOURNAL OF PHYSICAL OCEANOGRAPHY
VOLUME 43
FIG. 8. Vertical distribution of sources and sinks of density for (a) RELAX and (b) MIXED experiments. In the
Bousinesq approximation, density sources/sinks are included to represent contraction/expansion of water parcels. As
the circulation is balanced, the total amount of production and destruction of density are matching in absolute value.
The difference in their respective mean depths is the source of conversion power, consistent with the classical
Sandström (1908) argument.
ððð
2rgw dV 5
ððð
WD dV .
(30)
The total PE production is then independent of the
reference level.
The total production of density is zero in steady
state. Divide the ocean interior into two volumes, one
with sources of density (i.e., regions where diffusion
and mixing tends to increase the local density), and
another one with sinks of density. The two volumes contain the same amount of density variation, with opposite
signs
D[
ððð
jDj
dV 5
2
ððð
ððð
D dV 5 2
D.0
D dV . (31)
D,0
Now consider the mean depth of each
1
z [
D
1
ððð
Dz dV,
D.0
1
z [2
D
2
ððð
Dz dV . (32)
D,0
Finally, the total conversion rate, which is equal to the
total PE production rate, is proportional to the total
amount of density sources and the mean depth difference between density source and sink volumes,
ððð
(33)
2rgw dV 5 Dg(z1 2 z2 ) .
Hence, the total conversion rate available to power the
circulation depends on both the intensity of diffusion
and its vertical distribution.
To have a positive total conversion, the center of gravity
of density sources (cooling) must be above the center of
gravity of density sinks (heating) (Sandström 1908). This
result can be reinterpreted in terms of expansion and contraction areas, as the Boussinesq approximation essentially
consists in expressing volume expansion as a density sink
and contraction as a density source. Although heat flux is
applied at ocean surface, regions of contraction and expansion are found at depth, and a buoyancy driven circulation can be generated (Jeffreys 1926).
In the RELAX experiment, expansion occurs at a
mean depth of 217 m compared to 334 m for contraction,
thus Dz 5 117 m (Fig. 8). In comparison, Dz is only 46 m
in the MIXED experiment, as Ekman pumping induces
a sink of density at the surface, which explains the reduction of total conversion rate. In both cases, the height
difference is small compared to the basin depth, consistent with the Sandström (1908) argument that horizontal
convection is inefficient. Yet, this small difference in
depth is sufficient to produce a total conversion rate
comparable (though smaller) to the wind power input
in the MIXED experiment, and far bigger than the
power input in the wind only case. In this sense, Ekman
pumping is not very efficient either. Note also that the
decrease in MIXED total conversion as compared to
FEBRUARY 2013
471
ROQUET
RELAX is mainly due to shoaling of the mean expansion depth, due to large wind-related sinks of density at
the surface. This wind-induced effect on conversion rates
is thought to be large in the real ocean, due to the open
zonal channel in the Southern Ocean, resulting in a negative total conversion rate (Toggweiler and Samuels 1998).
The origin of power, which is converted from PE to KE,
can now be better understood, as it comes directly from
diffusion and mixing. The type of energy supply depends
on the nature of the mixing involved. In the real ocean,
mixing of density can result of a large variety of processes,
including molecular diffusion, eddy stirring, wave–current
interaction, double-diffusion or convective adjustments
(Wunsch and Ferrari 2004). Each process has its own
power supply. For example, convection depends on static
stability properties of the water-column and is powered by
PE conversion. Internal mixing can be greatly enhanced
by turbulence and wave activity, such as internal tides,
which are mechanically powered. Nonlinearities of the
equation of state also provide large sources and sinks of
density associated with the isoneutral mixing.
In the model, constant vertical and horizontal diffusivity coefficients were specified. These parameters are
unrealistically large compared to molecular diffusivities
to ascertain numerical stability and to account for subgridscale turbulence and stirring. The origin of the mixing
power source remains unspecified in our model, as the
processes involved in this diffusion are not resolved.
More sophisticated models generally use subgrid scale
parameterization trying to mimic the diffusion induced
by these unresolved scales, often using energetic arguments to be as realistic as possible. That subgrid scale
mixing is parameterized means that sources and sinks
of PE are locally applied directly in the interior of the
simulated ocean.
Note that knowledge of mixing energy supplies and
mixing efficiency is not sufficient to determine the total
conversion. One must know the spatial distribution of
density sources and sinks, which depend also on geometrical constraints and buoyancy forcings. In the limit
case of no buoyancy forcing, the ocean eventually becomes homogeneous, that is, spatial anomalies of density
(hence DPE) vanish. No conversion can then happen,
regardless of the intensity of mixing.
7. Generalization to seawater Boussinesq and
compressible models
In earlier sections, a DPE framework has been derived using the simple Boussinesq hydrostatic model,
using an assumption of linear equation of state. This assumption will now be relaxed using the so-called seawater
Boussinesq approximation (Vallis 2006) where the equation
of state is any given function of conservative temperature (McDougall 2003), salinity, and depth. Young
(2010) and Nycander (2011) showed how this model is
strictly equivalent to the compressible primitive equations in the limit of small relative density perturbations.
The DPE framework is in fact not limited to models
using the Boussinesq approximation. An equivalence
can be obtained for compressible hydrostatic models,
noting that this latter model is isomorphic to the Boussinesq hydrostatic model when formulated in pressure
coordinates (e.g., de Szoeke and Samelson 2002; Marshall
et al. 2004; Vallis 2006). Derivations given in this section
are here to demonstrate that the concept of DPE is potentially very general, requiring only the hydrostatic assumption to be valid.
a. Seawater Boussinesq model
The conservation of energy is here stated using the
seawater Boussinesq approximation (Vallis 2006; Young
2010; Nycander 2011). Contrary to the simple Boussinesq
approximation, the equation of state can be nonlinear. In
the seawater Boussinesq model, the density conservation
Eq. (14) is replaced by an equation of state, and two
equations for temperature and salinity conservation
r 5 r~(u, S, z) ,
Du
5 Du ,
Dt
(34)
and
(35)
DS
5 DS ,
Dt
(36)
where diffusion, mixing and surface fluxes are included
in the local terms Du and DS.
The procedure to obtain the equation of PE is described in Young (2010) and Nycander (2011). Define
the Boussinesq potential energy (BPE) as the vertical
integral of the equation of state times g
fb (u, S, z) [ 2
EbP 5 E
P
ð0
r~g dz.
(37)
z
This definition corresponds to a variant of the
Boussinesq dynamic enthalpy of Young (2010) and
the effective potential energy of Nycander (2011).
Note that for an equation of state independent of z,
the BPE definition reduces to the standard PE definition
EP 5 rgz.
Differentiating BPE definition gives
fb DEbP ›E
P
5
Dt
›z u,S
fb Dz ›E
P
1
Dt
›u S,z
fb DS
Du ›E
P
1
.
Dt
›S Dt
u,z
(38)
472
JOURNAL OF PHYSICAL OCEANOGRAPHY
VOLUME 43
›f
5 2a,
›p
fb /›z) 5 rg [see Eq. (37)], and
By definition, (›E
u,S
P
Dz/Dt 5 w, so their product gives (minus) the rate of
conversion. Using also Eqs. (35) and (36), the equation
for BPE is found to be identical to the PE Eq. (17)
(43)
=p up 5 0,
(44)
a5a
~ (u, S, p) ,
DEbP
Dt
5 rgw 1 WD ,
(39)
Du
5 Du ,
Dt
where
fb ›E
P
WD [
›u S,z
fb ›E
u
P
D 1
DS
›S DS
5 DS ,
Dt
(40)
u,z
is the new expression for the rate of PE production by
diabatic processes. Hence, using BPE instead of the
standard definition of PE, the conservation of energy
takes the same form as in the case of a linear equation of
state. If the standard PE definition was used instead of
BPE, the equation for internal energy should be explicitly considered, adding a lot of complexity in the set
of energy equations.
The DPE-based framework is then obtained using
horizontal anomalies of BPE as the definition for DPE.
Note that the stratification term
›Eb
S 5 w rg 2 P
›z
!
(41)
cannot be directly related to ›r/›z anymore because of
nonlinearities of the equation of state (and more specifically, because the equation of state depends on z).
Using the isomorphism between hydrostatic Boussinesq
equations of motion and primitive equations in pressure
coordinates (e.g., de Szoeke and Samelson 2002; Marshall
et al. 2004; Vallis 2006), it is straightforward to generalize the framework to the study of compressible fluids,
paving the way for an application of this framework to
the study of atmospheric energetics.
Let us express primitive equations of motion in the
pressure coordinate system, where pressure is used as
the vertical coordinate. We define the velocity vector
up 5 (uph , wp ), and the gradient operator =p 5 (=ph , ›/›p),
in this coordinate system. Here, wp 5 Dp/Dt is the
cross-pressure pseudovertical velocity (often referred
as omega), which has dimension of a power per unit
volume, J s21 m23.
Equations of motion are
(42)
(46)
(47)
with
D/Dt [ ›/›t 1 u $ 5 ›/›tjp 1 up =p .
(48)
Here, f 5 gz is the geopotential, and a 5 1/r is the
specific volume. Note that the momentum equations are
given for a unit mass parcel of fluid, instead of a unit
volume parcel in the Boussinesq formulation. This is a
more natural form for compressible fluids, as the conserved quantity is mass, not volume. In the following, the
subscript m is used for a quantity given in unit mass, as
for the frictional force noted F m. The equation of state is
a function of conservative temperature, salinity (or humidity for the atmosphere) and pressure, and gives the
specific volume instead of density.
For the compressible hydrostatic model, KE per unit
mass is given as EK,m [ juph j2 /2, and PE per unit mass is
equal to the geopotential value, EP,m 5 f. By analogy
with the seawater Boussinesq hydrostatic model, we
introduce an effective PE variable, tentatively called
compressible potential energy (CPE),
b. Compressible model
Duph
p
p
1 f k 3 uh 5 2=h f 1 F m ,
Dt
and
(45)
c (u, S, p) [
EcP,m 5 Eg
P,m
ðp
a
~ dp
(49)
po
where Po is an arbitrary chosen pressure of reference.
Note that CPE corresponds to the dynamic enthalpy
defined by Young (2010).
Take the scalar product of momentum equations with
the velocity vector for the KE equation, and the total
differential of CPE for the PE equation
DEK,m
Dt
5 2=p (fup ) 2 awp 1 WF ,m , and (50)
DEcP,m
5 awp 1 WD,m .
Dt
(51)
The rate of conversion of PE to KE is given by (minus)
the product of the specific volume by the cross-pressure
velocity 2awp, WF ,m is the rate of frictional work, and
FEBRUARY 2013
ROQUET
WD ,m is the rate of CPE production by diabatic
processes
c ›Eg
P,m WD,m [
›u c ›Eg
P,m D 1
›S u
S,p
DS .
(52)
u,p
To retrieve analogous modified equations of energy
conservation, we modify the definition of the horizontal
mean, taking it along isobars instead of along geopotentials. Horizontal (isobaric) means are here functions of pressure and time, using the notation
^ t) 1 T 0 (x, y, p, t) for a tracer T. We
T(x, y, p, t) 5 T(p,
0
define DPE as the isobaric anomaly of CPE, EcP,m .
A dynamical geopotential variable is also defined, in
analogy with the dynamical pressure
1
0
^
›
f
@ 1a
^ A dp .
fdyn 5 f0 1
po ›p
ðp
(53)
Finally, the modified set of energy equations is obtained
DEK,m
Dt
5 2=p (fdyn up ) 2 a0 wp 1 WF ,m ,
and
›EcP,m
0
1 =p (EcP,m up ) 5 a0 wp 1 WD,m 1 S m ,
›t
(54)
(55)
where the stratification term per unit mass is
Sm 5 w
p
a
^2
›EPc^,m
!
›p
.
(56)
Note that if the equation of state is independent of p,
the stratification term simply reduces to
S m 5 2(p 2 po )wp
›^
a
.
›p
(57)
The DPE to KE conversion rate is given as (minus)
the spatial correlation between the isobaric anomaly of
specific volume a0 and the cross-isobar vertical velocity
wp. In analogy with the Boussinesq case, the global
^ p 5 0 derived from the continuity equation
constraint w
is implicitly used here to remove the large static contribution 2^
awp from the conversion term.
8. Discussion and conclusions
In this paper, a new framework is proposed to study
ocean energetics, based on the concept of dynamical
potential energy. DPE is defined as the horizontal anomaly
473
of PE (where definitions of horizontal anomaly and PE
depends on the considered model).
For an incompressible fluid, the standard form of PE
to KE conversion rate is replaced by a more dynamically
relevant expression, the DPE to KE conversion rate,
which is proportional to the spatial correlation between
vertical velocities and horizontal anomalies of density.
This definition is motivated by the fact that the horizontal mean vertical velocity is then zero, and therefore
horizontal means of both expressions for conversion
rates are exactly equal. The DPE framework provides
modified expressions for KE and PE balance, which are
easier to map and interpret, as most of the hydrostatic
contributions have been removed. At the same time, this
framework simply reduces to standard energy conservation when integrated globally or horizontally. So standard interpretations of ocean energetics are conserved,
while gaining a physical understanding of energy balances
directly at the local scale.
A new pressure variable, tentatively called the dynamical pressure, is introduced so as to satisfy a reduced
hydrostatic balance function of the horizontal anomaly
of density. Dynamical pressure, which is essentially equal
to the horizontal anomaly of pressure, accounts for all
dynamical effects embedded in the pressure variable in
such a way that the static component of the hydrostatic
pressure field is optimally removed. The dynamical pressure proves to be very practical to visualize the large-scale
circulation. In a barotropic case, dynamical pressure is
by definition vertically constant, thus simply showing
the spatial pattern of sea surface height extending from
surface to bottom.
Fluxes of dynamical pressure represent the main way
to transport power on long distances, linking KE sources
and sinks. There are mainly two types of sources and
sinks of KE, related to the following.
(i) The rate of work of frictional forces. This power
source locally accelerates the flow when it is positive
or acts as a brake (internal and bottom friction,
eventually dissipated). The wind-power input is a
mechanical source of power, with an overall accelerating effect, while bottom and internal friction are
most generally sinks of power, transforming KE
into Joule heating.
(ii) The rate of conversion of DPE. The flow is locally
accelerated when relatively light waters are upwelled or when relatively heavy waters are downwelled, and decelerated otherwise. This power source
comes from the PE generated by diffusion and
mixing, destabilizing the large-scale distribution of
mass, which is eventually converted after being spatially redistributed through PE advection.
474
JOURNAL OF PHYSICAL OCEANOGRAPHY
To illustrate the use of the DPE framework, and to
demonstrate the technical feasibility of its numerical
diagnostics, several simulations using a state-of-the-art
ocean general circulation model were analyzed. Using
a simple squared basin configuration, the energetics of
circulations induced by wind and/or buoyancy forcings
was characterized. The wind forcing alone generated a
circulation through a net input of power at the surface,
transported in the interior as a flux of dynamical pressure where it was eventually dissipated. In contrast, the
buoyancy forcing induced a large-scale circulation from
global positive conversion rates. When wind and buoyancy forcings were combined, both patterns of energetics were superimposed, with the surface circulation
being primarily controlled by the wind, and the deeper
circulation being more influenced by the buoyancy
fluxes and conversion rates. However, response to both
wind and buoyancy forcings was not a simple addition
of responses to each forcing taken separately (see also
Saenz et al. 2012). The wind power input was greatly
enhanced by the presence of a subsurface thermocline,
while the total conversion rate was modulated by an
indirect effect of wind on isopycnal depths.
To better understand how buoyancy forcings power
a circulation, a simple relation has been derived to link
internal mixing with conversion rates [see Eq. (33)]. It
shows that the total conversion rate available to power
the circulation not only depends on the intensity of
mixing acting to erode density gradients, but also on
its vertical distribution. Hence, the presence of strong
mixing coefficients is not a sufficient condition for large
amounts of conversion. The mean depths where expansion and contraction occur must also be sufficiently
different. A negative total conversion rate is even possible, if the mean depth of density sources is pushed
below the density sinks depth, for instance due to the
wind. The DPE framework should allow a better understanding of what sets the distribution and magnitude
of conversion rates, as it provides an objective way to
analyze each term of the energy budget directly at the
local scale.
A straightforward method to generalize this framework to a compressible fluid in hydrostatic balance has
been presented, using the well-known isomorphism between hydrostatic Boussinesq equations of motion and
primitive equations in pressure coordinates (e.g., de Szoeke
and Samelson 2002; Marshall et al. 2004; Vallis 2006).
This opens the way for potential applications of the DPE
framework to the study of atmosphere energetics.
The question of how DPE and APE frameworks compare has been briefly mentioned. APE and DPE definitions are strictly different except for the trivial case
of a resting ocean with flat isopycnals. The general
VOLUME 43
philosophy leading to their definition is also dramatically different, as APE derives from purely thermodynamic considerations, while DPE comes from dynamical
constraints imposed by the hydrostatic balance. Yet, as
seen in section 3 and in the model study, the amount of
positive DPE values compares well with the total APE
reservoir, indicating that both definitions are somehow
connected.
DPE presents several advantages over APE, namely,
DPE has a simpler and more practical definition, and it
is perfectly well defined for any equation of state. In
comparison, the definition of APE rapidly becomes
overly complicated as the determination of the minimum PE state of reference is a difficult and computationally expensive task. Moreover, APE is not uniquely
defined for a realistic nonlinear equation of state (i.e.,
a nonlinear function of both temperature and salinity),
making its use for the real ocean circulation highly hypothetical. Nonetheless, APE is very popular as it has
provided a nice framework to explain fundamental
processes such as baroclinic instability or the effect of
wind on the stratification. The question of whether DPE
can fully replace APE to understand ocean (and atmosphere) energetics is now open and will require further
investigations.
Acknowledgments. Many thanks go to Carl Wunsch,
Jonas Nycander, and Gurvan Madec for extremely
helpful discussions and comments. Jean Michel Campin
was also of precious help, in particular to carry out model
experiments. This work started at MIT (Cambridge,
Massachusetts) during a post-doctoral period supported
by the National Ocean Partnership Program (NOPP)
through funding of NASA to the ECCO-GODAE
Consortium.
APPENDIX A
Work of Viscous Forces
Frictional forces can be written using the viscous
stress tensor, t 5 [t ij], where t ij is ith component of the
viscous stress along the jth direction
F i 5 2 å ›j t ij .
(A1)
j
Parameterizing the stress tensor using a Fickian diffusion of momentum, t ij 5 2ronj›jui, the work of frictional forces can be expressed as
WF 5 ro å ui ›j (nj ›j ui )
i,j
(A2)
FEBRUARY 2013
475
ROQUET
The sink of kinetic energy owing to viscous dissipation
is transformed into internal energy by Joule effect (increasing locally the temperature). However its effect is
supposedly so small in the ocean that the Joule effect is
most often neglected in ocean general circulation models
(see e.g., Griffies 2004).
APPENDIX B
Energetics of a Linear Free Surface
FIG. A1. Frictional work and dissipation for a one-dimensional
normal jet. The velocity profile follows a standard Gaussian distribution, with a standard deviation of 1. Frictional work rate,
dissipation rate and meridional flux of KE are normalized, using
ronx 5 1.
5
å ›j (nj ›j EK ) 2 ro å nj (›j ui )2 ,
j
(A3)
i,j
with uh 5 (u1, u2). The first right-hand side term is given
as the divergence of a viscous flux, function of the KE
gradient, and the second term is a negative-definite
quantity corresponding to the viscous dissipation, function of squared velocity shears.
The wind stress acts as surface boundary condition for
the stress tensor, thus providing energy as a vertical
viscous flux. This power input is then transported in the
Ekman boundary layer where it is dissipated while
generating the velocity shear within Ekman spiral.
To better understand the relation between frictional
work and viscous dissipation, these quantities are diagnosed for a Gaussian jet (see Fig. A1). Most of the
frictional work enters at the core of the jet, where velocity
and curvature are both largest. In contrast, dissipation is
concentrated on the flank of the jet, where the shear is
strongest. Power is transported from the center to the
flanks of the jet as a viscous stress flux proportional to the
gradient of KE.
The typical distance between areas of frictional work,
and dissipative areas, is given by the width of the jet. In
the real ocean, jet width is typically of the order of
100 km horizontally, and 50 m vertically. These spatial
scales are thus negligible compared to the size of ocean
basins.
Let us consider a linear free surface, modeled as
a layer enclosed between the surface geopotential (z 5
0) and the sea surface height (z 5 h). Variations of
density within this layer are neglected, r 5 ro, as well as
horizontal velocities, uh 5 0.
A first consequence is that the vertical velocity is
constant vertically within this layer to satisfy continuity,
and equal to the rate of change of sea surface height
wh 5
›h
.
›t
(B1)
Neglecting horizontal velocities within the layer means
that the hydrostatic KE is always equal to zero. Divergence of the pressure flux term can be integrated vertically, yielding
ðh
$ ( pu) dz 5
0
ðh
0
›( pw)
dz 5 2po wh ,
›z
(B2)
where po 5 rogh is the pressure along the surface
geopotential.
A simple manipulation shows that
2po wh 5
ðh
0
2ro gwh dz 5
›
›t
ðh
0
EP dz .
(B3)
In other words, the pressure divergence term is always
exactly compensated by the amount of conversion taking place within the surface layer, which is itself powered
by changes in free surface PE content.
Finally, in this simple yet physically relevant model,
the budget of energy within the free surface is trivial. To
treat free surface within the more general problem of
full ocean energetics, it suffices to note that the source
of the power entering the ocean interior as a flux of
pressure along the surface geopotential is generated by
a change in PE converted within the free surface.
More realistic models of free surface will require
greater care in the expression of their energetics, however the linear free surface already gives an excellent
first-order picture of the importance of free surface
476
JOURNAL OF PHYSICAL OCEANOGRAPHY
energetics in the context of large-scale oceanography,
still widely used in many ocean general circulation
models.
REFERENCES
Callen, H. B., 1985: Thermodynamics and an Introduction to
Thermostatics. John Wiley and Sons, 493 pp.
de Szoeke, R. A., and R. M. Samelson, 2002: The duality between
the Boussinesq and non-Boussinesq hydrostatic equations of
motion. J. Phys. Oceanogr., 32, 2194–2203.
Gnanadesikan, A., 1999: A simple predictive model for the structure of the oceanic pycnocline. Science, 283, 2077–2079.
——, R. D. Slater, P. S. Swathi, and G. K. Vallis, 2005: The energetics of ocean heat transport. J. Climate, 18, 2604–2616.
Griffies, S., 2004: Fundamentals of Ocean Climate Models. Princeton
University Press, 518 pp.
Hieronymus, M., and J. Nycander, 2013: The buoyancy budget with
a nonlinear equation of state. J. Phys. Oceanogr., 43, 176–186.
Huang, R. X., 1998: Mixing and available potential energy in
a Boussinesq ocean. J. Phys. Oceanogr., 28, 669–678.
——, 1999: Mixing and energetics of the oceanic thermohaline
circulation. J. Phys. Oceanogr., 29, 727–746.
——, 2005: Available potential energy in the world’s oceans.
J. Mar. Res., 63, 141–158.
Hughes, C. W., and B. A. DeCuevas, 2001: Why western boundary
currents in realistic oceans are inviscid: A link between form
stress and bottom pressure torques. J. Phys. Oceanogr., 31,
2871–2885.
Hughes, G. O., and R. W. Griffiths, 2008: Horizontal convection.
Annu. Rev. Fluid Mech., 40, 185–208, doi:10.1146/annurev.
fluid.40.111406.102148.
——, A. M. C. Hogg, and R. W. Griffiths, 2009: Available potential
energy and irreversible mixing in the meridional overturning
circulation. J. Phys. Oceanogr., 39, 3130–3146.
Ilicak, M., A. J. Adcroft, S. M. Griffies, and R. W. Hallberg, 2012:
Spurious dianeutral mixing and the role of momentum closure. Ocean Modell., 45–46, 37–58.
Iudicone, D., G. Madec, B. Blanke, and S. Speich, 2008: The role of
Southern Ocean surface forcings and mixing in the global
conveyor. J. Phys. Oceanogr., 38, 1377–1400.
Jeffreys, H., 1926: On fluid motions produced by differences of temperature and humidity. Quart. J. Roy. Meteor. Soc., 51, 347–356.
Klocker, A., and T. J. McDougall, 2010: Influence of the nonlinear
equation of state on global estimates of dianeutral advection
and diffusion. J. Phys. Oceanogr., 40, 1690–1709.
Lorenz, E. N., 1955: Available potential energy and the maintenance of the general circulation. Tellus, 7, 157–167.
Marshall, J., A. Adcroft, C. Hill, L. Perelman, and C. Heisey, 1997:
A finite-volume, incompressible Navier–Stokes model for
VOLUME 43
studies of the ocean on parallel computers. J. Geophys. Res.,
102 (C3), 5753–5766.
——, ——, J.-M. Campin, C. Hill, and A. White, 2004: Atmosphere–
ocean modeling exploiting fluid isomorphisms. Mon. Wea. Rev.,
132, 2882–2894.
McDougall, T. J., 2003: Potential enthalpy: A conservative oceanic
variable for evaluating heat content and heat fluxes. J. Phys.
Oceanogr., 33, 945–963.
Munk, W. H., and C. Wunsch, 1998: The moon and mixing: Abyssal
recipes II. Deep-Sea Res. I, 45A, 1977–2010.
Nycander, J., 2011: Energy conversion, mixing energy and neutral
surfaces with a nonlinear equation of state. J. Phys. Oceanogr.,
41, 28–41.
——, J. Nilsson, K. Döös, and G. Broström, 2007: Thermodynamic
analysis of ocean circulation. J. Phys. Oceanogr., 37, 2038–
2052.
Oort, A. H., S. C. Ascher, S. Levitus, and J. P. Peixoto, 1989: New
estimates of the available potential energy in the world ocean.
J. Geophys. Res., 94, 3187–3200.
Roquet, F., C. Wunsch, and G. Madec, 2011: On the patterns of
wind-power input to the ocean circulation. J. Phys. Oceanogr.,
41, 2328–2342.
Saenz, J. A., A. M. Hogg, G. O. Hughes, and R. W. Griffiths, 2012:
Mechanical power input from buoyancy and wind to the circulation in an ocean model. Geophys. Res. Lett., 39, L13605,
doi:10.1029/2012GL052035.
Sandström, J. W., 1908: Dynamicsche versuche mit meerwasser.
Ann. Hydrogr. Martimen Meteor., 36, 6–23.
Stern, M. E., 1975: Ocean Circulation Physics. Academic Press,
246 pp.
Tailleux, R., 2009: On the energetics of stratified turbulent mixing,
irreversible thermodynamics, Boussinesq models and the
ocean heat engine controversy. J. Fluid Mech., 638, 339–382.
——, 2012: Available potential energy and exergy in stratified
fluids. Annu. Rev. Fluid Mech., 45, 35–58.
Toggweiler, J. R., and B. Samuels, 1998: On the ocean’s large-scale
circulation near the limit of no vertical mixing. J. Phys. Oceanogr., 28, 1832–1852.
Vallis, G. K., 2006: Atmospheric and Oceanic Fluid Dynamics.
Cambridge University Press, 745 pp.
Wang, W., and R. X. Huang, 2004: Wind energy input to the surface
waves. J. Phys. Oceanogr., 34, 1276–1280.
Winters, K. B., P. N. Lombard, J. J. Riley, and E. A. D’Asaro, 1995:
Available potential energy and mixing in density-stratified
fluids. J. Fluid Mech., 289, 115–128.
Wunsch, C., and R. Ferrari, 2004: Vertical mixing energy and the
general circulation of the oceans. Annu. Rev. Fluid Mech., 36,
281–314.
Young, W. R., 2010: Dynamic enthalpy, conservative temperature,
and the seawater Boussinesq approximation. J. Phys. Oceanogr., 40, 394–400.