Ocean Modelling 38 (2011) 41–70

Contents lists available at ScienceDirect

Ocean Modelling
journal homepage: www.elsevier.com/locate/ocemod

Accurate Boussinesq oceanic modeling with a practical, ‘‘Stiffened’’ Equation of State

Alexander F. Shchepetkin ⇑, James C. McWilliams
Institute of Geophysics and Planetary Physics, University of California, 405 Hilgard Avenue, Los Angeles, CA 90095-1567, United States

a r t i c l e i n f o a b s t r a c t

Article history: The Equation of State of seawater (EOS) relates in situ density to temperature, salinity and pressure. Most
Received 9 August 2008 of the effort in the EOS-related literature is to ensure an accurate fit of density measurements under the
Received in revised form 25 December 2010 conditions of different temperature, salinity, and pressure. In situ density is not of interest by itself in oce-
Accepted 31 January 2011
anic models, but rather plays the role of an intermediate variable linking temperature and salinity fields
Available online 13 February 2011
with the pressure-gradient force in the momentum equations, as well as providing various stability func-
tions needed for parameterization of mixing processes. This shifts the role of EOS away from representa-
Keywords:
tion of in situ density toward accurate translation of temperature and salinity gradients into adiabatic
Seawater Equation of State
Seawater compressibility
derivatives of density.
Barotropic–baroclinic mode splitting In this study we propose and assess the accuracy of a simplified, computationally-efficient algorithm
Boussinesq and non-Boussinesq modeling for EOS suitable for a free-surface, Boussinesq-approximation model. This EOS is optimized to address
Numerical stability all the needs of the model: notably, computation of pressure gradient – it is compatible with the monot-
onized interpolation of density needed for the pressure gradient scheme in sigma-coordinates of Shche-
petkin and McWilliams (2003), while more accurately representing the pressure dependency of the
thermal expansion and saline contraction coefficients as well as the stability of stratification; it facilitates
mixing parameterizations for both vertical and lateral (along neutral surfaces) mixing; and it leads to a
simpler, more robust, numerically stable barotropic–baroclinic mode splitting without the need of exces-
sive temporal filtering of fast mode. In doing so we also explore the implications of EOS compressibility
for mode splitting in non-Boussinesq free-surface models with the intent to design a comparatively accu-
rate algorithm applicable there.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Role of EOS in oceanic modeling the local column thickness), both of which participate in baro-
tropic–baroclinic mode-splitting algorithm of Shchepetkin and
The Equation of State (EOS) relates the in situ density of seawa- McWilliams (2005,);
ter with its temperature (or potential temperature), salinity, and  Stability of stratification as well as external thermodynamic
pressure (H, S, P, respectively), forcing (surface buoyancy flux) needed for mixing and plane-
tary boundary layer parameterizations;
q ¼ qEOS ðH; S; PÞ: ð1:1Þ
 Slope of neutral surfaces needed for horizontal (along-isopyc-
In a Boussinesq-approximation oceanic modeling code, in situ den- nal) mixing (Griffies et al., 1998).
sity does not appear by itself (since it is replaced by a constant ref-
erence density). Instead EOS and EOS-related quantities are needed The purpose of this study is to review the present oceanic mod-
for the following computations: eling practices for using EOS in these roles, focusing on the effects
associated with seawater compressibility, and it assesses the con-
 Pressure-gradient force (PGF); sequences of the Boussinesq approximation in the context of a
 Vertically averaged density q
 and the effective dynamic density realistic EOS. The present study extends the analysis of conse-
for the barotropic mode q⁄ (vertically integrated pressure nor- quences of Boussinesq approximation from Shchepetkin and
malized by gD2/2 where g is acceleration of gravity and D is McWilliams (2008).
This paper is organized as follows: Section 2 makes a overview
of the Boussinesq approximation, analyzes the errors associated
with using realistic seawater EOS, and introduces stiffening of
⇑ Corresponding author. EOS as a method to reduce these errors. Section 3 examines the
E-mail addresses: alex@atmos.ucla.edu (A.F. Shchepetkin), jcm@atmos.ucla.edu consequences of finite compressibility of seawater for the four
(J.C. McWilliams). algorithmic roles outlined above, to establish requirements for

1463-5003/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ocemod.2011.01.010
42 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70
the most suitable functional form of EOS. Special attention is given
to barotropic–baroclinic mode-splitting since this subject is rarely
discussed in the literature. Section 4 introduces a practical form of
EOS and makes estimates of its accuracy. Section 5 explores the
consequences of eliminating the Boussinesq approximation with
the emphasis on accuracy of barotropic–baroclinic mode-splitting
algorithm in the context of finite compressibility of seawater. Sec-
tion 6 is the conclusion.
2. Boussinesq approximation and EOS stiffening
The Boussinesq approximation (e.g., Spiegel and Veronis, 1960;
Zeytounian, 2003, among others) replaces in situ density q with a
constant reference value q0 in all places where it plays the role
of a measure of inertia, i.e., everywhere except where multiplied
by the acceleration of gravity g. The velocity field becomes non-
divergent (incompressible) because the continuity equation
changes its meaning from mass to volume conservation. This re-
verses the role of EOS from computing specific volume in a
mass-conserving model to density q in a volume-conserving one.
The conservation laws (momentum, energy, tracer content, etc.)
are changed from mass-to volume-integrated; the thermodynam-
ics is reduced to Lagrangian conservation (advection and diffusion)
of H and S, while heating/cooling of fluid by adiabatic compres-
sion/expansion is neglected (except in the definition of potential
temperature itself) and mechanical energy dissipated by viscosity
is considered ‘‘lost’’ rather than converted into heat (Mihaljan,
1962). If EOS is a linear function of H and S, mass is conserved
(along with volume) as a consequence of, and to the same degree
as, the conservation of H and S; external heating/cooling produces
a decrease/increase in mass, while keeping volume constant; and
nonlinear EOS causes a Boussinesq model to conserve only volume,
but no longer mass, even in the absence of external forcing.
The Boussinesq approximation brings simplifications by elimi-
nating mass-weighting in a finite-volume code and limiting the
role of EOS to the four purposes stated in Section 1. Except in the
part of pressure-gradient term associated with perturbation of free
surface in a free-surface model, q itself can be replaced with its
perturbation, q0 = q  q0, because it appears only inside spatial
derivatives (ultimately linked to the gradients of H and S) and only
in the context of buoyancy commonly defined as gq0 /q0, i.e., nor-
malized by q0, whereas the equations can always rewritten in such
a way that q0 itself appears only in the context of this normaliza-
tion and nowhere else. The Boussinesq approximation facilitates
the barotropic–baroclinic mode-splitting in a pair of vertically-
integrated free-surface and momentum equations, effectively
uncoupling EOS from the barotropic mode. The incompressibility
assumption eliminates acoustic waves regardless of whether the
hydrostatic approximation is made.
However, physically important effects associated with seawater
– cabbeling and thermobaricity – fundamentally require pressure-
dependency in EOS and lead to the common practice of using the
full non-approximated EOS, thus retaining its full compressibility
even in an otherwise dynamically incompressible Boussinesq mod-
el. This obscures the concept of buoyancy because it can no longer
be equated to the density (or potential density) anomaly, and can
no longer be viewed as a Lagrangianly-conserved properly of the
fluid, even thought H and S are. A related, frequently used approx-
imation is the replacement of the full in situ pressure P in EOS (1.1)
with its bulk hydrostatic reference value, P ? q0gz, when com-
puting q from the model prognostic variables, H and S, hence
neglecting pressure variation due to baroclinic effects and essen-
tially decoupling EOS pressure from the dynamic.
Under most offshore oceanic conditions q varies by  ± 3% rel-
ative to its reference value; this leads to errors associated with the
Boussinesq approximation which can be subdivided into two
categories:
(i) errors relative to not using the Boussinesq approximation
including not only quantitative, but conceptual as well, i.e.,
excluded physical processes and/or missing/altered conser-
vation laws; and
(ii) conflicts and internal inconsistencies caused the use of a
realistic seawater EOS with full compressibility effects
within a Boussinesq oceanic model.
Errors of type (i) are widely discussed in the literature
(McDougall and Garrett, 1992; Dukowicz, 1997; Lu, 2001;
Greatbatch et al., 2001; Huang and Jin, 2002; McDougall et al.,
2002; Greatbatch and McDougall, 2003; Losch et al., 2004; Griffies,
2004; Young, 2010; Tailleux, 2009; Tailleux, 2010).
An example of an internal inconsistency of type (ii) can be illus-
trated by considering a barotropic compressible fluid layer whose
density is a function of P alone (i.e., because H, S are spatially uni-
form) and in hydrostatic balance,
q ¼ qEOS ðPÞ and @ z P ¼ g q: ð2:1Þ
Integration of (2.1) yields mutually consistent vertical profiles for P
and q,
 ( 0
P ¼ Pðf  zÞ; P ðf  zÞ ¼ gRðf  zÞ;
such that
q ¼ Rðf  zÞ  qEOS ðPðf  zÞÞ; Pð0Þ ¼ Pjz¼f ¼ 0;
ð2:2Þ
where P and R are universal functions of a single argument, i.e.,
their structure depends only one the properties of qEOS(P) in (2.1)
but not directly on the local dynamical conditions, such as pertur-
bation of the free-surface f. P 0 denotes derivative of P with respect
to its argument (note that the sign is correct as stated above: both p
and q increase with increase of f  z, meaning increase downward).
The condition Pð0Þ ¼ 0 is the free-surface pressure boundary condi-
tion (for simplicity the atmospheric pressure is presumed to be con-
stant and subtracted out).
A gradient of f induces a pressure gradient,
rx P ¼ P 0 ðf  zÞ  rx f; ð2:3Þ
and creates acceleration
1 P 0 ðf  zÞ
 rx P ¼   rx f  g rx f; ð2:4Þ
q Rðf  zÞ
independently of vertical coordinate z regardless of the specific
functional form qEOS(P) in (2.1) and of the magnitude of the density
variation within the column. Eq. (2.4) is derived without the use of
Boussinesq approximation. Its Boussinesq analog is
1 Rðf  zÞ
 rx P ¼ g  rx f: ð2:5Þ
q0 q0
The presence of the multiplier R=q0 , which increases with depth, is
clearly an artifact of the Boussinesq approximation. It results in an
unphysical vertical shear in the acceleration, hence a spurious
downward intensification of a geostrophically-balanced baroclinic
current generated by a rxf. Both these spurious effects are caused
by the standard algorithmic chain in a Boussinesq model,
1
H; S ! q ¼ qEOS ðH; S; P ¼ q0 gðf  zÞÞ !  rx P: ð2:6Þ
q0
To estimate the significance of this error, consider for simplicity a
linear analog of (2.1),
qEOS ðPÞ ¼ q1 þ P=c2 ; ð2:7Þ
 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70
where q1 is the density at P = 0 and c is speed of sound. Then the
counterpart of (2.2) becomes
2 q1 g 2
Pðf  zÞ ¼ q1 c2 ½egðfzÞ=c  1  q1 gðf  zÞ þ ðf  zÞ2 þ    ;
2c2
2 qg q g2
Rðf  zÞ ¼ q1 egðfzÞ=c  q1 þ 12 ðf  zÞ þ 1 4 ðf  zÞ2 þ    ;
c 2c
ð2:8Þ
hence
R q1  q0 q1 g
1þ þ  ðf  zÞ þ   
q0 q0 q0 c2
Here  = g(f  z)/c2 plays the role of a small parameter justifying the
replacement of q with q0. Assuming c  1500 m/s and a total depth
of 5500 m, we estimate   0.025, which is a typical level for Bous-
sinesq-approximation errors. This also indicates a strong domi-
nance of the leading-order linear term in Taylor-series expansions
over the remaining terms. A comparison of (2.5) with (2.4) suggests
that the best choice of Boussinesq reference density is
Rf
q0 ¼ q ¼ D1 h Rðf  zÞdz, D = h + f to have the vertical integral of
(2.5) match its non-Boussinesq counterpart (2.4). In the case of
EOS (2.7) this leads to
 
c  gD=c2
2  1 gD
q ¼ q1  e  1  q1 1 þ  2 þ  ;
gD 2 c
which makes (2.9) into ðR=q0 Þ ¼ 1 þ gðf  z  D=2Þ=c , effectively 2
minimizing its deviation from unity. However, this cannot be done
universally, since q  changes horizontally mainly due to topography,
nor does it eliminate the spurious vertical shear.
The presence of shear in (2.5) depends on whether or not the
EOS pressure is computed as in situ pressure, which takes place
in (2.2), or is approximated by its bulk hydrostatic value propor-
tional to depth, and furthermore, whether the depth is calculated
from the dynamic or rest-state free surface, Pbulk = q0g(f  z) or
q0gz,
8 8
< qEOS ¼ q1 þ q0 gðf  zÞ=c2 ; < qEOS ¼ q1  q0 gz=c2 ;
  vs:  
1 q1 gðfzÞ
:  q rx P ¼ g q þ c2 rx f; :  q1 rx P ¼ g qq1  cgf2 rx f:
0 0 0 0
½vertical shear is present ½no shear
This sensitivity was noted by Dewar et al. (1998) (their Case A), who
estimate that the magnitude of the difference in near-bottom geo-
strophic currents may reach 5 cm/s for Gulf Stream conditions,
resulting in an overall difference of about 3 out of 50 Sv for verti-
cally integrated transport. They recommend using full dynamic
in situ pressure in EOS, which requires simultaneous vertical inte-
gration to compute PEOS and q. Paradoxically, they do so using Bous-
sinesq approximation, interpreting the difference as the recovered
physical effect, rather than a spurious one to be eliminated. Com-
parison of (2.2)–(2.4) with (2.11) suggests that using the rest-state
(f = 0) bulk pressure in EOS is more suitable for a Boussinesq model,
while a non-Boussinesq model is better using in situ pressure. (One
can also verify that the opposite combination – bulk rest-state EOS
pressure in a non-Boussinesq model – yields a reversal of the spu-
rious shear: it tends to make otherwise barotropic flows decrease
with depth.) Another aspect pointed out by Dewar et al. (1998) is
that in most oceanic situations the baroclinicity naturally tends to
cancel the free-surface PGF resulting in weak flows in the abyss.
Therefore, it is advantageous to retain baroclinic effects in EOS pres-
sure rather than use PEOS = q0g(f  z), (e.g., left side of (2.11)) that
does not provide such compensation. Using q0g(f  z) is a common
practice in terrain-following-coordinate free-surface models, POM
and ROMS, motivated primarily by the convenience: these models
do not store the unperturbed geopotential coordinate in a run-time
array.
43
For the situation where the fluid is stratified (i.e., (2.1) is no
longer true) but the structure of temperature and salinity fields
are such that the resultant isosurfaces of in situ density are parallel
to the free surface, one can verify that (2.4) still holds and the
acceleration generated by a perturbation in f does not depend on
the vertical coordinate as long as the Boussinesq approximation
is not used. Once again, the Boussinesq approximation destroys
the true vertical uniformity. The baroclinic PGF – i.e., associated
with a physically correct vertical shear – appears only when den-
sity isosurfaces become non-parallel to the free surface.
ð2:9Þ
Dukowicz (2001) made a proposal to reduce Boussinesq errors
by realizing that the largest q variation is caused by the changes
in P rather than in H and S. EOS (1.1) may be rewritten as1
q ¼ rðPÞ  qEOS ðH; S; PÞ; ð2:12Þ
where the multiplier r(P) is chosen as a universal function that does
not depend on local H and S. qEOS ðH; S; PÞ is known as thermobaric
or ‘‘stiffened’’ density. It only weakly depends on P, thus has much
smaller variation than the original in situ density. Using (2.12) one
can renormalize pressure,

dP 1
¼ ; P jP¼0 ¼ 0; ð2:13Þ
dP rðPÞ
ð2:10Þ
hence the hydrostatic balance and PGF become
@P 1 1
¼ q g and  rx P ¼  rx P  : ð2:14Þ
@z q q
These retain the same functional form as in the original non-Bous-
sinesq equations. Because r(P) is a monotone function, (2.13) is
invertible, P = P(P), making it possible to rewrite EOS as
q ¼ qEOS ðH; S; PðP ÞÞ=rðPðP ÞÞ ¼ qEOS ðH; S; P Þ: ð2:15Þ
The entire set of equations is then expressed in terms of dotted vari-
ables, q and P.
A Boussinesq-like approximation is subsequently applied to the
renormalized system (2.13)–(2.15) by replacing q ! q0 wherever
the replacement q ? q0 occurs in the standard Boussinesq approx-
imation (e.g., ð1=q Þrx P  ! ð1=q0 Þrx P  ). In addition, one can
replace EOS pressure P in (2.15) with its bulk value, q0 gz. qEOS
ð2:11Þ is much less sensitive to this replacement than qEOS in (1.1),
whether the integration of EOS pressure starts from a perturbed
or unperturbed free surface (cf., (2.11)) since @ qEOS =@P jH;S 
@ qEOS =@PjH;S and P tends to be closer to a linear function of depth
than P. The multiplier r(P) does not appear in the momentum equa-
tions written in advective form, but it does appears in the non-
Boussinesq continuity equation along with q. This leads to the
two options, either.
(i) to replace r(P) ? 1 along with q ! q0 resulting in a vol-
ume-conserving system of Boussinesq-type. In this case
EOS stiffening can be viewed just as a measure to remove
internal contradiction of using a compressible EOS in a
model that already makes the incompressibility assumption;
or
 
(ii) keep r(P), while simplifying it by rðPÞ ! r PðP  ¼ q0 gzÞ ¼
rðzÞ, but still replacing q ! q0 in continuity and, in fact,
everywhere as it would be in the Boussinesq case.
The latter leads to a model intermediate in its physical approx-
imations between Boussinesq and non-Boussinesq, while main-
taining the simplicity of a Boussinesq code. One should note that
1
We have changed the original notation of Dukowicz (2001) by replacing his q⁄,
P⁄ ? q, P to avoid notational conflict with vertically averaged densities q
 and q⁄ that
appear in (3.19). For the same reason we also changed notation relatively to SM2005,
so q⁄ appearing in this article has the same meaning as q⁄ in SM2005.
44 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70
most oceanic models are discretized using a finite-volume ap-
proach with a stretched vertical grid, so they already have an array
to compute and store control-volume heights. A non-Boussinesq
model uses density weighting over control volumes, resulting in
an additional feedback loop arising from the dependency of density
on model dynamics by EOS pressure which adds extra complexity.
On the other hand, retaining r(P) approximated as r(z) only slightly
modifies the procedure of calculating control volumes, while pre-
serving the overall logical flow of a Boussinesq code. This means
that the model is still volume-conserving, but now control volumes
are rescaled to include r(z)-weighting that can be viewed just as a
geometrical property of the grid since it does not depend on model
dynamics. In effect, this option approximates in situ density as
q0  rðzÞ instead of the constant q0 in the standard Boussinesq
approximation.
To illustrate the effect of modifying the Boussinesq approxima-
tion in this way, we reconsider (2.7)–(2.11). In this case it is natural
to chose q0 ¼ q1 and r(P) = 1 + P/(q1c2). Then the stiffened EOS
(2.15) becomes q ¼ qEOS ðP  Þ ¼ q1 , which is just a constant value,
i.e., qEOS becomes infinitely ‘‘stiff’’ and therefore insensitive to
how EOS pressure is computed (the opposite of (2.11)). The renor-
malized pressure is P = gq1(f  z), resulting in ð1=q0 Þrx P ¼
gðq1 =q0 Þrx f ¼ g rx f, which coincides with the non-Boussinesq
version (2.4) and does not contain any spurious vertical shear. Note
 2  2
that PðP Þ ¼ q1 c2 ðeP =ðq1 c Þ  1Þ and rðP Þ ¼ eP =ðq1 c Þ ; this translates
2
into rðzÞ ¼ egðfzÞ=c .
The preceding example is rather trivial because H, S fields are
uniform which makes it possible to express the entire density var-
iation in terms of r(P), so the subsequent replacement q ! q0
leads to an equivalence of the modified Boussinesq and non-Bous-
sinesq models. In the general case H, S are not uniform, and the
approximation is not exact. However, (2.12)–(2.15) is still a very
effective measure to reduce Boussinesq-approximation errors (by
90%) because for realistic conditions the variations of speed of
sound c are relatively small (i.e., c = 1480  1540 m/s) and a large
portion of that variation is due to changes in pressure alone, which
can be excluded from contributing to the errors. EOS in (2.1) is
nonlinear with respect to P; however, (2.4) still holds and r(P)
can absorb the entire density variation in this case as well. The
effectiveness of using (2.12) is facilitated by the fact that variations
of H, S tend to decrease significantly with depth, where EOS non-
linearities due to P are strongest.
Finally, we should note that McDougall et al. (2002) proposed
an alternative version of Boussinesq approximation, where they
interpret Boussinesq velocity not in its original meaning, but as
re-normalized mass flux per unit area, u ~ ¼ qu=q0 . Their system
does not produce non-physical shear in geostrophically-balanced
flow in the Boussinesq case (2.6), because now it is u ~ , not u be-
comes proportional to vertical density profile, which means that
the unscaled velocity in vertically uniform, hence is correct. They
concluded that the new Boussinesq makes EOS stiffening of Duko-
q01 ,
wicz (2001) unnecessary, because it can be avoided by their ap-
q01 ; the absolute depth (z) serves the role of EOS pressure
proach. Unfortunately their system has an undesirable side effect
of altering scaling relationship between advection and Coriolis
term which makes it impossible to derive a potential vorticity
equation for a shallow layer of barotropic compressible fluid, see
Appendix A. This puts their system into disadvantage relatively
to Boussinesq equations with stiffened EOS, and negates their crit-
icism of Dukowicz (2001).
3. Pressure effects in EOS
In this section we examine the specific requirements on EOS for
the purposes listed in Section 1 to find a functional form compat-
ible with the PGF calculation in sigma-coordinates, Shchepetkin
and McWilliams (2003,); re-evaluate the accuracy of the mode-
splitting algorithm of SM2005 focusing on compressibility effects;
and inspect the EOS needs for subgrid-scale mixing
parameterizations.
3.1. Pressure-gradient force in sigma-coordinates
Following SM2003, the baroclinic PGF in r coordinates can be
expressed entirely in terms of adiabatic derivatives of in situ den-
sity: a density-Jacobian scheme is written as
J x;s ðq; zÞ ¼ a  J x;s ðH; zÞ þ b  J x;s ðS; zÞ; ð3:1Þ
which is then integrated vertically to compute PGF. J x;s denotes the
Jacobian of its two arguments,
@ q @z @ q @z
J x;s ðq; zÞ ¼    ; ð3:2Þ
@x s @s @s @x s
% 0; z!f
where s ¼ sðx; zÞ
& 1; z ! h;
@s
> 0; s is a generalized sigma-coordinate;2 and
@z
a ¼ aðH; S; PÞ ¼ @ q=@ HjS;P¼const ;
b ¼ bðH; S; PÞ ¼ @ q=@SjH;P¼const ð3:3Þ
are thermal expansion and saline contraction coefficients.3 Geopo-
tential height z in (3.1) is used as a proxy for EOS pressure – replac-
ing it with gq0z, hence neglecting the non-uniformity of density
and the contribution due to f – 0. This is a common approximation
in Boussinesq codes. Eq. (3.1) has as its only EOS requirement on the
computation of a and b. Because a and b depend on pressure (depth),
the weighting of J x;s ðH; zÞ and J x;s ðS; zÞ in the r.h.s. changes with
depth as well, so that the same pair of gradients of H and S yields
differing contributions in PGF with depth (a thermobaric effect).
Eq. (3.1) involves two separate Jacobians for H and S. Comput-
ing them separately and combining them at the last stage is unde-
sirable because applying discretized monotonicity constraints to H
and S separately does not guarantee positive stratification of inter-
polated density, even if the density values corresponding discrete
H and S are positively stratified. This can lead to large errors and
possible numerical instability if the density field is not smooth
on the grid scale. Instead, for reasons explained below, the pre-
ferred form of EOS is
X
n
 
q ¼ qð0Þ 0
qð0Þ 0 m
1 þ q1 ðH; SÞ þ m þ qm ðH; SÞ  ðzÞ ; ð3:4Þ
m¼1
ð0Þ
where q1 ¼ q1 þ q01 ðH; SÞ is density with a given potential temper-
ature and salinity at surface pressure; the compressibility coeffi-
ð0Þ ð0Þ ð0Þ ð0Þ
cients qm ¼ qm þ q0m ðH; SÞ do not depend on z; q1 ; q1 ; . . . ; qn are
ð0Þ
constant bulk values chosen in such a way that q1
ð0Þ
q1
(when it appears inside EOS it increases downward, counting from
either the rest-state free surface z = 0, or the actual free surface
z = f). Then J x;s ðq; zÞ can be computed as
2
Throughout this study we use s to denote an arbitrary, non-separable, surface-
and terrain-following coordinate z=z(x, y, s). In contrast, r is strictly reserved for the
proportional (non-stretched) sigma-coordinate, r = (z  f)/(h + f). However, in any
case the displacements of coordinate surfaces caused by free-surface movement are
proportional to the distance from the bottom, otzjs=const = otf(z + h)/(h + f).
3
In a more standard definition a and b are normalized by density, hence a = (1/
q)oq/oHjS,P=const and b = (1/q)oq/oSjH,P=const, where q is either in situ (non-Boussinesq
version), or q = q0 is Boussinesq reference density (if Boussinesq approximation is
used). Since for the purpose of the present study it is important to keep track of the
q ? q0 replacement on every occasion, we chose to exclude (1/q) from the definition
of a and b.
 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70 45

  Xn
leads to
J x;s ðq; zÞ ¼ J x;s q01 ; z þ ðzÞm  J x;s ðq0m ; zÞ; ð3:5Þ Z 0
m¼1 @P @f 0
¼ g qjz¼f þ g J x;s ðq; zÞds
@x z @x s
where (z)m are treated as coefficients not subject to differentia-
ð0Þ @f
X
n
@f @f
tion. For this reason this can also be viewed as a form of adiabatic ¼ g q1 þ qð0Þ  ðf  zÞm  ¼ gq : ð3:10Þ
ð0Þ ð0Þ ð0Þ
differentiation. The contribution of terms with q1 ; q1 ; q2 ; . . ., can- @x m¼1 m @x @x
cels out identically here; i.e., they are dynamically passive. To en- In a Boussinesq model the last term in the r.h.s. is divided by q0,
sure that the stratification corresponding to the interpolated which leads to a spurious vertical shear in a purely barotropic flow
density field stays non-negative (non-oscillatory), the algorithm of (cf., (2.5) in Section 2). From this point of view, it is desirable to ex-
SM2003 relies on a measure of the smoothness of density field ð0Þ ð0Þ
clude q1 ; q2 ; . . . from (3.4):
based on the ratio of consecutive adiabatic differences. The scheme X
n
is based on cubic polynomial fits and requires evaluation of density q ¼ q0 þ q0 where q0 ¼ q01 ðH; SÞ þ q0m ðH; SÞ  ðzÞm ; ð3:11Þ
derivatives at the same discrete locations as density itself, qi,j,k, m¼1
which in its turn needs some kind of averaging of elementary differ- after which using (z) or f  z no longer makes a noticeable differ-
ences over the two adjacent grid intervals. If the two elementary ence SM2003. EOS stiffening of Dukowicz (2001) can be imple-
differences are of the same sign, but differ by more than a factor mented in this context by choosing
of three, a simple algebraic averaging overestimates the derivative
 X
n
at the location of qi,j,k and causes spurious oscillation of the cubic ð0Þ
rðzÞ ¼ 1 þ 1=q1 ð0Þ
qm  ðzÞm ð3:12Þ
polynomial interpolant. If EOS has no dependency of pressure, a m¼1
conventional monotonization algorithm for the interpolation of
density would suffice, but for a realistic EOS the comparison of ele- and applying it to (3.4). This leads to cancellation of terms contain-
ð0Þ ð0Þ
mentary differences is meaningful only if they are defined in an adi- ing q1 ; q2 ; . . ., etc., and, after expansion in powers of (z), the
abatic sense, resultant q0 can be expressed in the same functional form as
(3.11), but with a modified set of coefficients: the q0m ðH; SÞ are
slightly reduced to account for 1/r(z) 6 1. Obviously, r(z) commutes
Dq0ðadÞ 0 0
iþ1=2;j;k ¼ q1iþ1;j;k  q1i;j;k with the Jacobian operator,
Xn  
z m
iþ1;j;k þ zi;j;k
þ q0miþ1;j;k  q0mi;j;k  : ð3:6Þ J x;s ðrðzÞ  q ; zÞ ¼ rðzÞ  J x;s ðq ; zÞ; ð3:13Þ
m¼1
2
however, the corresponding discrete relationship holds only within
The two adjacent adiabatic differences are averaged using harmonic the order of accuracy, not exactly.
mean, and (if needed) the compressible part is computed and added The above choice of r(z) is on par with Sun et al. (1999), Hallberg
separately, (2005), but differs from Dukowicz (2001), who derives it from
globally-averaged H,S-profiles from observations to make
0ðadÞ 0ðadÞ q0  rðzÞ as close as possible to in situ density: this is incompatible
@ q 1 2Dqiþ1=2;j;k  Dqi1=2;j;k with the algorithm (3.6) and (3.7) for avoiding spurious oscillations
di;j;k  ¼ 
@n i;j;k Dn Dq0ðadÞ þ Dq
0ðadÞ
in stratification, which leads to the alternative fit in Section 4.
iþ1=2;j;k i1=2;j;k
" # In summary we identify the following requirements on EOS for
  @z
0 0 accurate computation of PGF in a sigma-model: accurate depen-
þ q1i;j;k þ 2q2i;j;k  ðzi;j;k Þ þ     ; ð3:7Þ
@n i;j;k dency of a and b on H, S, and pressure (depth); non-appearance
ð0Þ ð0Þ
of in situ density and suppression of the bulk terms q1 ; q1 , . . .,
0ðadÞ 0ðadÞ
which is applicable if Dqiþ1=2;j;k and Dqi1=2;j;k have the same sign; etc. in PGF; and an EOS form that facilitates computation of adia-
otherwise the first term in the r.h.s. is replaced with zero. Because batic differences in all directions.
the harmonic averaging never exceeds twice the smaller of the
two operands, it acts as monotonization algorithm for interpolation. 3.2. Barotropic–baroclinic mode-splitting
The algorithm (3.6), (3.7) requires that q01 , q01 , q02 , . . . be available
separate from the EOS calculation. This is practical only if (3.4) is re- The barotropic–baroclinic mode-splitting algorithm of SM2005
stricted to a very few coefficients q0m . In Section 4 we will present a makes the Boussinesq approximation without analyzing its conse-
sufficiently accurate approximation to EOS with the functional form quences for mode-splitting. It assumes that the fluid is incom-
of (3.4) that contains only two coefficients, q01 ¼ q01 ðH; SÞ and q02 , pressible and that all q variations occur due to baroclinicity,
with the latter being just a constant multiplied by q01 . which is not the realistic case where most of the q change is caused
From the above one might get an impression that the passive- by P rather than (H, S) variation.4 In this section we provide a de-
ð0Þ ð0Þ ð0Þ
ness of the bulk terms q1 ; q1 ; q2 ; . . . in (3.4) is just a consequence tailed analysis of how mode splitting is affected by the compressibil-
of neglecting the contribution disturbance of f in EOS pressure. ity effects.
This is not correct. Including f in EOS pressure would cause a Bous- The purpose of mode-splitting is to take advantage of the differ-
sinesq error of the type (ii) in (2.5)–(2.11). Consider an EOS with ence of time scales between external gravity-wave propagation
the form, and other, slower processes. Its key element is expressing the ver-
tically-integrated PGF,
X
n
Z
q ¼ q1ð0Þ þ ð0Þ
qm  ðf  zÞm þ    ð3:8Þ 1 f
F ¼ r? P dz ¼ F ½r? f; f; r? qðx; zÞ; qðx; zÞ; ð3:14Þ
m¼1 q0 h

where    denotes the primed terms in (3.4) that depend on H and S. in two parts: one that can be efficiently computed from only the
The contribution of the bulk terms barotropic variables (e.g., f) and another that is independent (or

X
n
m1 @z @f 4
This situation is partially addressed by removing the dynamically passive bulk-
J x;s ðq; zÞ ¼ qð0Þ
m  m  ðf  zÞ   ; ð3:9Þ compressibility terms from EOS in SM2003, but the main motivation was to reduce
m¼1
@s @x
the PGF error, rather than to improve mode splitting.
46 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70

as weakly dependent as possible) on f. With F the vertically inte-
grated momentum equation is rewritten as
Þ þ    ¼
@ t ðDu dF þ ½F  dF  :
|{z} |fflfflfflfflfflffl{zfflfflfflfflfflffl}
@ðdF Þ=@f – 0 @=@f0
The first r.h.s. term, dF , is identified as the barotropic PGF. In split-
explicit time-stepping it is treated as ‘‘fast’’ and recomputed at each
barotropic time step. Conversely, the second is ‘‘slow’’ and accord-
ingly kept constant during the barotropic stepping and only up-
dated during the baroclinic step. F must be computed at every
baroclinic step before the barotropic stepping begins. When com-
puting F by (3.14), there is no other choice than to use the latest
available f, which at this moment is at the previous time step. Con-
sequently, the residual dependency of F  dF on f results in a r.h.s.
term of a hyperbolic nature treated by using an effectively forward-
Euler baroclinic time-stepping. This may lead to a numerical insta-
bility, imposing an additional non-physical restriction on the baro-
clinic time-step size or may require extra damping. The difficulty
comes from the fact that (3.14) is generally a nonlinear functional,
involving products of density and, in principle, dependence of q on f
through EOS compressibility. To address these issues, dF is derived
as a variational derivative of F (SM2005, also Shchepetkin and
McWilliams, 2008)
@F @F
dF ¼ dðr? fÞ þ df;
@ ðr? fÞ @f
as an approximation for barotropic PGF (Berntsen et al., 1981;
Blumberg and Mellor, 1987; Killworth et al., 1991). These are
Boussinesq-approximation models that assume the smallness of
ð3:15Þ
q(x,z)  q0, and their mode-splitting procedures rely on this small-
ness as well. Isopycnic models tend to be non-Boussinesq and use a
different mode-split, essentially linking dF to the bottom pressure
of the resting state (f = 0) (Bleck and Smith, 1990). Since the bot-
tom pressure is computed by vertical hydrostatic integration, this
translates into dF ¼ gDr? ðq  fÞ, where q is the vertically-aver-
aged density. (Note that this dF is dimensionally different from
the similar term in (3.15) because non-Boussinesq models use den-
sity-weighting to compute u  .) This split is also empirical and
approximate. Higdon and de Szoeke (1997) propose a more accu-
rate split for a non-Boussinesq isopycnic model, which replaces q 
with a different value computed from the local vertical profile of
density (somewhat similar to q⁄, but expressed in terms of specific
volume rather than density). In all cases, with or without the Bous-
sinesq approximation, the fluid was considered as incompressible
for the purpose of mode-splitting.7
The goal of replacing gDr\f with (3.18), (3.19) is to achieve
mutual consistency between the 3D and 2D parts of the model–
mainly in the ability of the split 2D part to yield the same barotrop-
ic gravity-wave phase speed ~c as the full 3D model. This consis-
tency may or may not provide an improvement in the overall
physical accuracy. The sources of mode-splitting error from using
ð3:16Þ
gDr\f as the barotropic PGF are the following:

where f and r\f are viewed as independent state variables.5 The

(i) ‘‘Horizontal baroclinicity’’: Suppose the fluid is incompress-
use of (3.16) ensures that when two different time-stepping algo-
ible and there is no vertical stratification, but the density
rithms are applied separately to the barotropic and baroclinic com-
varies horizontally, hence q ¼ q ¼q  ðxÞ. Because of the
ponents, the resultant sum of the updated barotropic PGF (i.e.,
Boussinesq approximation, F and ~c are seenp by the 3D part
ffiffiffiffiffiffiffiffiffiffiffi
computed from the new f) and the original 3D ? 2D forcing term
of the model as proportional to q  =q0 and q  =q0 , respec-
½F  dF  is sufficiently close to the vertically-integrated PGF com-
tively. The spatial variation of q  makes it impossible to
puted from the new f,
choose q0 in such a way that the use of the gDr\f-term
@F @F 0 for the barotropic mode accurately approximates F , espe-
dF 0 þ ½F  dF  ¼ F þ r? ðf0  fÞ þ ðf  fÞ cially in a very large basin, thus contributing to mode-split-
@ðr? fÞ @f
ting error using  gDr\f. The small parameter associated
¼ F 0 þ Oððf0  fÞ2 Þ; ð3:17Þ
with this type of error is ðq   q0 Þ=q0 .
0 0
where F ¼ F ½f; r? f; r? q; q and F ¼ F ½f ; r? f ; r? q; q corre- 0 (ii) Vertical stratification: ~c is smaller in a stratified compressible
spond to the old f and the updated f0 each with same density field. fluid than in a stratified incompressible fluid with the same
The Oððf0  fÞ2 Þ estimate of the mismatch between F 0 and the l.h.s. depth,
comes from the fact that (3.17) is essentially a Taylor-series expan- ~c2 ¼ gh  ^c21 ; ð3:20Þ
sion of F 0 in powers of f0  f. Assuming that q does not depend on f
and remains frozen during barotropic time-stepping,6 (3.16) yields where ^c1 is the gravity-wave phase speed of the first baro-
" ! # clinic mode (Appendices C and D). The relevant small param-
g q f2 eter for estimating this mode-splitting error is the ratio of the
dF ¼  hr? ðq fÞ þ r? þ ðq  q
 Þfr? h ; ð3:18Þ
q0 2 squares of the modal phase speeds, ~c21 =ðghÞ.
(iii) Seawater compressibility: In the case of constant bottom
8 Z
> 1 f topography ~c is reduced relative to an incompressible fluid
>
> q
 ðxÞ ¼ qðx; zÞdz
< D h with the same depth by
where Z f Z f ð3:19Þ  
>
> 1 1 gh
> q
:  ðxÞ ¼ qðx; z 0
Þdz 0
dz ~c2 ¼ gh 1   þ    ð3:21Þ
D2 =2 h z 2 c 2

are two-dimensional horizontal (with x = (x,y)) fields kept constant
during barotropic time-stepping. This implies that the splitting
algorithm relies on the independence of q  and q⁄ from f, which in
turn relies on the assumption of independence of EOS from f.
A property of (3.18) is that if density is uniform, q ¼ q
 ¼ q0 , dF
reverts back to gDr\f that is used by most of the oceanic models
(Appendices A and B for non-stratified and stratified cases).
The mechanism of this reduction is non-Boussinesq and is
associated with the fact that the volumetric divergence of
horizontal barotropic velocity causes a smaller response in
f than in an incompressible fluid. The use of (3.18), (3.19)
qualitatively captures the influence of compressibility–by

5 7
A finite-difference version of (3.14) is a function of two adjacent grid-point In isopycnic coordinates this approximation is also needed to circumvent a
values, fi,j and fi+1,j. Their average and difference are analogous to f and r\f. theoretical difficulty with the assumption that potential density is a Lagrangian
6
More precisely, the density field is ‘‘frozen in r-space’’ during barotropic time- conserved property, which implies that it can be defined globally relative to a given
stepping. This is because the entire vertical coordinate system ‘‘breathes’’ when f reference pressure. In fact, the seawater EOS does not have this property (Jackett and
moves up and down, so each vertical grid box also changes proportional to its height, McDougall, 1997), allowing only a local definition of potential density with respect to
while the grid-box averaged values of density remain unchanged. a local reference pressure.
 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70
distinguishing q⁄ and q  with q < q
 in this case–but the anal-
yses below indicate an underestimation of the effect, and
therefore a contribution to mode-splitting error with an asso-
ciated small parameter of gh/c2. Since both corrections to the
barotropic phase speed ~c, (3.20) and (3.21), are small, they
are approximately additive (Appendix C).
(iv) Sea-level contribution to EOS pressure: In principle, PGF is sen-
sitive to how EOS pressure depends on f (cf., (2.11)), hence
F , depends on f through compressibility in EOS in addition
to the explicit dependence from r\f. In the case of constant
topography the impact on barotropic phase speed ~c due to
this effect is small relative to that of (iii)–Oðgf=c2 Þ vs.
Oðgh=c2 Þ –in fact, it formally vanishes when linearizing with
respect to small f/h  1 (Appendices A and B). However,
since f used in EOS pressure is always taken from the previ-
ous baroclinic time step and is kept constant during the
barotropic time stepping, this leads to the appearance of
hyperbolic r.h.s. terms that receive effectively a forward-
in-time treatment. This potentially makes it a source of
  ð3:22Þ
mode-splitting error and even instability. It should also be
noted that the effect can be amplified in the presence of
steep topography where the last term in (3.18),
ðq  q Þfr? h, dominates over hr\(q⁄f), even though the
influence of f into q⁄ and q  via EOS partially cancels due to
subtracting one from the other.
The mode-splitting procedure of SM2005 is designed to address
items (i) and (ii) within the Boussinesq approximation, while (iii)
and (vi) are given much less consideration.
Notice that replacing gDr\f with (3.18), (3.19) motivated by
item (i) is an artifact of the Boussinesq approximation. pffiffiffiffiffiffi From (2.4)
the non-Boussinesq local barotropic phase speed gh does not de-
pend on local vertically-averaged density q  . (The inertial terms in
the non-Boussinesq momentum equation use local density rather
than q0, resulting in cancellation of density when computing accel-
eration and phase speed; Appendix B.) Paradoxically, using
gDr\f, rather than (3.18), (3.19) may look more suitable in this
case because it yields the correct ~c. Conversely, the use of (3.18),
(3.19) can be viewed as an introduction of an a priori physical dis-
tortion into the barotropic mode to match an existing distortion in
the 3D F to minimize mode-splitting error, which is the higher pri-
ority. EOS stiffening provides some relief in this conflict by merely
transforming ðq   q0 Þ=q0 ! ðq   q0 Þ=q0 , which is expected to be
somewhat smaller; however, practical experience with ROMS
SM2005 indicates that the largest deviations of local vertically-
averaged density are due to local H,S anomalies in shallow mar-
ginal seas, often semi-enclosed by topography. Under such condi-
tions the bulk compressibility effect is not very strong, so
stiffening does not make a large difference. Closely related to item
(i) is the sensitivity of the numerical stability of the code to the
choice of q0, and (3.18), (3.19) eliminates this problem.
The reduction of ~c due to stratification, item (ii), occurs in both
non-Boussinesq and Boussinesq cases, and (3.18), (3.19) provide an
accurate accounting of it (Appendix D). To estimate its importance,
note that the ratio of phase speeds also sets the mode-splitting ra-
tio (i.e., the number of barotropic time steps per baroclinic step,
typically in the range 30–70 and closer to the upper limit for an
open-ocean configuration).8 This implies an estimated erroneous
reduction of ~c by 103 of its value from (3.20). Although this might
seem to too small to be a concern, recall that the phase error due to
the slightly different ~c values from using gDr\f is accumulated
over many barotropic time steps performed at each baroclinic step,
8 ficients from EOS,
In practice the splitting ratio is adjusted by the ratio of stability limits of the time-
stepping algorithms for the 2D and 3D parts. For ROMS the splitting ratio is typically
1.4 times larger than the ratio of phase speeds.
47
which makes the numerical stability of the code sensitive to whether
(3.18), (3.19) is used or not.
The physical explanation of item (iii) is non-Boussinesq in
nature: despite the coincidence of (2.4) with the PGF due to the
gradient of f in a barotropic incompressible flow, the value of ~c is
smaller in a compressible fluid (cf., (B.6) in Appendix B), where a
convergence of horizontal velocity causes a change of both f and
density with relative proportions of 1  gh/c2 and gh/c2,
respectively.
Eq. (3.18), (3.19) are derived using the Boussinesq approxima-
tion, but nevertheless they also predict a smaller phase speed than
with gDr\f. When the density increases with depth, q < q  , so
the PGF caused by a given r\f is slightly smaller than in the case
of vertically-uniform density with the same water-column mass
(n.b., this implies that the choice of q0 ¼ q
 ). Substitution of (2.8)
into (3.19) yields
 
c2 gD=c2 1 gD
q ¼ q1  ðe  1Þ  q1 1 þ  2 þ   
gD 2 c

c2 c2 gD=c2 1 gD
q ¼ q1  ðe  1Þ  1  q1 1 þ  2 þ    :
gD gD 3 c
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffi
With a flat bottom this implies ~c ¼ gh  q =q0 instead of just gh.
While this qualitatively captures the effect of reducing ~c relative to
that of an incompressible fluid, even if we make the best possible
choice for reference density (q0 ¼ q  ), the correction is underesti-
mated by a factor of 1.5 relative to its true, non-Boussinesq magni-
tude in (B.9) in Appendix B. This discrepancy is not surprising since
the non-Boussinesq explanation for the reduced phase speed is dif-
ferent than in (3.18).
Eq. (3.21) provides an estimate for the
pffiffiffiffiffi ffi barotropic speed reduc-
tion due to effect (iii) of 0.6% of gh assuming h = 5500m and
c = 1500 m/s. This is one-and-half orders of magnitude larger than
the effect of baroclinicity (ii), and therefore it creates a stronger
requirement for replacing gr\f with (3.18), (3.19) even thought
correcting for item (iii) was not the initial motivation. EOS stiffen-
ing dramatically reduces the difference between q  and q⁄. This
eliminates the non-physical mechanism of reduced barotropic
phase speed due to compressibility via (3.22). It also brings the
mode-splitting procedure into the realm of assumptions for its ori-
ginal derivation.
The effect of (iv) is expected to be much smaller than (iii); how-
ever, Robertson et al. (2001) observed a numerical instability in
POM while simulating tides under near-resonant conditions. The
source of the instability was traced back to the influence of f on
EOS pressure through the use of old-time-step f due to the specifics
of the mode-splitting algorithm. Although they did not classify it as a
mode-splitting instability, the mechanism of its appearance is
essentially the same as in (iv). They proposed to eliminate the insta-
bility by dropping pressure effects in EOS altogether: they split EOS
additively as q = qN(H,S) + qp(P) (starting with their Eq. (16)), and
showed the dynamic passiveness of qp, which they subsequently ex-
cluded. This leads to complete loss of the thermobaric effect, and
that may be not acceptable in many large-scale applications. EOS
stiffening of Dukowicz (2001) helps to reduce (but not to eliminate
entirely) the splitting error while keeping thermobaricity intact.
3.3. Vertical mixing parameterization
The vertical mixing parameterizations are based on the stratifi-
cation stability, expressed as the Brundt–Väisälä Frequency
N(x,z,t), and the translation of surface heat and freshwater fluxes
into a buoyancy flux. Both steps use the a(H,S,P) and b(H,S,P) coef-
N2 ¼ g½a  @ z H  b  @Sz : ð3:23Þ
48 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70

Thus, the requirement for EOS is to accurately reproduce a and b, updated version of which is available today as Locarnini et al.
but now with a shifted emphasis on accurate computation of their (2006), Antonov et al. (2006)]. Although this approach yields the
ratio, a/b. The primary sensitivity here comes from detecting when minimum possible range of variation for q, and thus has the max-
stratification and/or buoyancy forcing change sign, because this cor- imum effect in reducing the Boussinesq–approximation errors, it
responds to a drastic transition from one stable to convective mix- interferes with the ability to monitor stratification smoothness
ing regime. The Boussinesq—non-Boussinesq and EOS stiffening by (3.6), (3.7) because the spatial differences of q constructed this
issues are not a concern here because both a and b are normalized way unavoidably contain a contribution associated with the se-
by the same density–whether in situ or replaced by q0 or q0 , lected mean density profile as long as the direction of differencing
is not strictly horizontal (unlike the horizontal differencing in z-
1 @ q 1 @ q 1 @ q 1 @ q
a¼ !  ; b¼ ! ; ð3:24Þ coordinate models, for which the method of Dukowicz (2001)
q @ H S;P q0 @ H S;P q @S H;P q0 @S H;P was originally intended). As a result, a change in sign for the differ-
hence their ratio is not affected. Because negative stratification is ences no longer corresponds to the change from positive to nega-
intolerable in a hydrostatic model and parameterized vertical mixing tive physical stratification. This also alters the ratios of
is the only mechanism that prevents it (by entrainment of water from consecutive differences in (3.7) ultimately negating its monoton-
adjacent positively stratified layers above or below), it is desirable ization properties. Therefore we alternatively choose constant ref-
that the procedure for detecting negative stratification in the mixing erence values, Href and Sref, that are representative of the abyssal
scheme be consistent with the monotonized adiabatic differencing in part of the ocean, and then construct a multiplier rð^zÞ as
the PGF algorithm (3.6), (3.7). In particular, the stratification may 1
rð^zÞ ¼ ; where K ref ð^zÞ ¼ K 00 þ K ref ref
0 þ K1 ^ z þ K ref
2 ^ z2 ;
change abruptly near the interior edge of the top and bottom bound- 1  0:1^z=K ref ð^zÞ
ary layers. Accurate determination of the layer thickness requires
ð4:2Þ
density interpolation where it is often non-smooth on the vertical
ref ref
grid-scale; this leads to similar requirements on EOS as for PGF. with K ref
0¼ K 0 ðH ; S Þ, ref
K ref
¼ K 1 ðH ; S Þ, and1 ¼ K 2 ðH ; Sref Þ.
ref
K ref
2
ref

In the results presented here we choose H = 3.50C and ref

3.4. Neutral surfaces S = 34.50/00, which imply K ref

0 ¼ 2924:921, K ref
1 ¼ 0:34846939,
ref 5 ref ref
K 2 ¼ 0:145612  10 , and q1(H ,S ) = 1027.43879. The stiff-
The definition of potential density depends a reference pres- ened EOS (2.15) with this Boussinesq-approximation rð^zÞ becomes
sure, and for a realistic EOS it cannot be defined globally in the 1  0:1^z=K ref ð^zÞ
sense of a scalar function of local H,S,P with spatial derivatives q0 ðH; S; ^zÞ ¼ ½q0 þ q01 ðH; SÞ   q0 : ð4:3Þ
1  0:1^z=KðH; S; ^zÞ
equivalent to adiabatic derivatives of in situ density (Jackett and
McDougall, 1995). Instead, one should either construct a neutral q0 is the perturbation of q relative to a constant reference value q0 ,
density (Jackett and McDougall, 1997), which is inherently non- which is naturally chosen as q0 ¼ q1 ðHref ; Sref Þ. After cancellation of
local, or, alternatively, one could use a local procedure relying on the large terms, the stiffened EOS is rewritten as
a  , and b-weighting of temperature and salinity gradients (see q0 þ q01 ðH; SÞ
Eqs. (31), (32) in Section 5 in Griffies et al. (1998)). EOS in the form
q0 ðH; S; ^zÞ ¼ q01 ðH; SÞ þ 0:1^z 
K 00 þ K ref ref
0 þ K1 ^ z þ K ref
2 ^ z2
(3.4) and the associated adiabatic differencing (3.6) allow direct    
K ref  K 0 ðH; SÞ þ K ref
1  K 1 ðH; SÞ  ^ z þ K ref z2
2  K 2 ðH; SÞ  ^
computation of the local slopes of neutral surfaces. 
0

K 00 þ K 0 ðH; SÞ þ ðK 1 ðH; SÞ  0:1Þ^z þ K 2 ðH; SÞ^z2

^0 ðH; S; ^zÞ  ^z;
 q1 ðH; SÞ þ q
4. A practical ‘‘stiffened’’ EOS
ð4:4Þ
The seawater EOS of Jackett and McDougall (1995) has the form, so far without any approximation. This is already close to the de-
q1 ðH; SÞ sired functional form (3.4); however q ^0 ðH; S; ^zÞ still explicitly de-
qðH; S; ^zÞ ¼ ; pends on ^z, although weakly as we show below. Eq. (4.4) implies
1  0:1  ^z=½K 00 þ K 0 ðH; SÞ þ K 1 ðH; SÞ  ^z þ K 2 ðH; SÞ  ^z2 
that q0 ðHref ; Sref ; ^zÞ  0 and q
^0 ðHref ; Sref ; ^zÞ  0 for all ^z. This ensures
ð4:1Þ
that the range of variation for q0 is expected to be small in general,
where H is potential temperature; S salinity; q1(H,S) is the density and even diminish with depth since variations of H and S also
at a surface pressure of 1atm, fit by a 15-term polynomial contain- diminish.
ing various products of powers of Hn, n = 1, . . . , 5, and Sm, m = 1, 3/ A Taylor series expansion of (4.4) in powers of ^z yields
2, 2 (not in all permutations). The divisor K = K00 + K0(H,S) +   
q0 ðH; S; ^zÞ ¼ q01 ðH; SÞ þ q01 ðH; SÞ  ^z; ð4:5Þ
has K00 = const, and the remaining three coefficients K0, K1 K2 de-
pend on H and S with similar polynomial fits with powers of where
H, . . . , H4, and S, S3/2 that add up to a 26-term polynomial for K.  
q01 ðH; SÞ ¼ 0:1  q0 þ q01 ðH; SÞ
The UNESCO EOS has the same functional form, but it is expressed
in terms of in situ temperature; hence q1 is identically the same, but K ref
0  K 0 ðH; SÞ
 : ð4:6Þ
the coefficients for K0,K1,K2 are different. In addition, EOS pressure in ðK 00 þ K 0 ðH; SÞÞ  ðK 00 þ K ref
0 Þ
(4.1) is remapped into depth ^z.9 The complexity of the functional
form of UNESCO EOS motivates oceanic modelers to derive simplified This has exactly the functional form of (3.4). It is an approximation
versions for more efficient calculation (Mellor, 1991; Wright, 1997; to (4.4) where all K1 and K2 terms (14 out of 26 coefficients of K) are
Brydon et al., 1999; McDougall et al., 2003; Jackett et al., 2006). discarded. In practice we use a slightly modified form,
Dukowicz (2001) chooses the multiplier r(P) to fit an in situ den- q0 ðH; S; ^zÞ ¼ q01 ðH; SÞ þ q01 ðH; SÞ  ^zð1  c^zÞ; ð4:7Þ
sity profile computed from the globally horizontally averaged tem-
perature and salinity profiles based on Levitus WOA94 data [an
9
Throughout this section we adopt the convention that ^z increases downward, the
same way as hydrostatic pressure. We use notation ^z to distinguish it from z
(increasing upward) as it is used elsewhere in this article.
with c = 1.72  105 m1 for the Href and Sref values listed above.
This captures some of the nonlinear dependency of the thermobaric
effect without significant increase in complexity, and it also reduces
errors relative to (4.5). Consistently with (4.7), the algorithm for
computing elementary adiabatic differences (3.6) becomes
 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70 49

  ^z
iþ1;j;k þ ^ zi;j;k Isolines of q0 from (4.7) are plotted in the left column of Fig. 1 for
Dq0ðadÞ
iþ12;j;k
¼ q01iþ1;j;k  q01i;j;k þ q01iþ1;j;k  q01i;j;k

2  five different depths: 0, 200, 1000, 2500, and 5000 m. The shading
^ziþ1;j;k þ ^zi;j;k in Figs. 1,2 indicates the global distribution of hydrographic data
 1c : ð4:8Þ
2 values (Levitus WOA data, Antonov et al., 2006; Locarnini et al.,
2006). At each depth the data samples are combined into

Fig. 1. ‘‘Stiffened’’ density perturbation q0 defined by (4.7) (left); the corresponding thermal expansion, a (middle), and saline contraction, b (right), coefficients as functions
of potential temperature H and salinity S at five different depths ^z ¼ 0; 200; 1000; 2500 and 5000 m. The units are kg/m3, (kg/m3)/°C, and (kg/m3)/(‰) respectively (note that
here a = @ q/@ HjS,P=const and b = +@ q/@ SjH,P=const without the usual division by q). Shading indicates the distribution of observed (H, S) values (see text). (The overall format is
similar to Fig. 19 in SM2003.)
50 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70

Fig. 2. Errors in a (left), b (middle), and their ratio, a/b (right) due to the approximate EOS (4.7) compared to the complete EOS (4.4) at four different depths:
^z ¼ 200; 1000; 2500, and 5000 m. The shading is the same as in Fig. 1, and the dashed box surrounds the domain of interest. (The format is similar to Fig. 20 in SM2003.)

DH  DS = 0.250C  0.250/00 bins to compute the probability distri-
bution. This is shown by the three hues of shading: the lightest indi-
cates that this pair of H,S values is observed at least once in the
dataset, the medium shading contains 99% of the observed data at
that depth, and the darkest contains 90%. There is a significant
reduction of variation of the observed H,S with depth as manifested
by the shrinking of the shaded area, which was utilized by McDou-
gall et al. (2003) and SM2003 to simplify treatment compressibility
effects in EOS (note that only a very little seasonal variation of H, S
remains at the depths of ^z ¼ 200 m, which is illustrated by a rather
narrow strip shaded area, with the lightest shade becomes indistin-
guishable). Furthermore, as highlighted by McDougall and Jackett
(2007), there is a strong tendency for the probability distribution
to align itself with the isolines density at each depth – which they
call ‘‘thinness in S  H  P space’’.
Isolines of q0 in the (H, S)-plane tend to rotate counterclock-
wise with increasing depth (thermobaric effect) and exhibit a gen-
tle curvature (cabelling). However, as assured by the construction
of EOS, the q0 does not have tendency to increase with depth.
The density range for observed (H,S) is much reduced; e.g., at
^z ¼ 5000 m, H  3.50, S = 34.50/00, and q0 stays close to zero, in
contrast with qin situ  1050 kg/m3 (see Fig. 19, left column, in
SM2003; note that field plotted there is qin situ  1000).
The coefficients a and b are plotted in the center and right col-
umns of Fig. 1. a shows a significant increase with depth, while b
decreases slightly. One notices some differences in values and in
patterns compared to Fig. 19 in SM2003, especially for b. For
example, at ^z ¼ 5000 m, H  3.50 C and S = 350/00 b = 0.760 in
that Fig. 19, while here b = 0.745 in Fig. 1. The difference of 2%
is explained by rð^zÞ, which is responsible for the density differ-
ence between qin situ  1050 vs. q0  1027:4 here, so that
b½Fig: 19=qinsitu  b½Fig:1=q0 . Similar adjustment occurs for a,
and, as the result, adiabatic differences of stiffened density normal-
ized by q0 (using Boussinesq-like rules) are now close to that
computed from a full EOS (4.1), but normalized by qin situ (non-
Boussinesq rules).
As shown in Section 2, the most important requirement for EOS
is to accurately reproduce a and b as functions of depth within the
subspace of naturally occurring H and S values. Furthermore,
because hydrostatic models are especially sensitive to the occur-
rence of negative stratification, the accuracy of representation of
the ratio of a/b is even more valuable than the accuracy of a and
 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70
b taken separately (e.g., this occurs in southern part of Atlantic
Ocean characterized by the fresh, but cold near-surface layer on
top of warmer, but more saline deeper water, resulting in a ther-
modynamically delicate situation with very week overall stratifica-
tion, and the possibility of cabbeling and thermobaric instabilities,
Paluszkiewicz et al., 1994; McPhee, 2000). This is also true for
modeling deep currents in the presence of H and S anomalies,
which tends to cancel effects of each other, e.g., Mediterranean out-
flow. Therefore, we evaluate the differences between a, b, and a/b
computed using (4.6), (4.7) against its prototype (4.4), which is the
unapproximated stiffened EOS (Fig. 2). On each panel we indicate
the domain of interest with a dashed-line box surrounding ob-
served (H,S) values. The error in computing a within the domain
of interest can be estimated as ± 0.0001 at ^z ¼ 1000 m; ± 0.0005
at ^z ¼ 2500 m; and ± 0.002 at ^z ¼ 5000 m. For b similar errors
are ± 0.0004 at ^z ¼ 1000 m; ±0.001 at ^z ¼ 2500 m; and ±0.005 at
^z ¼ 5000 m. These error estimates are conservative because the ob-
served (H,S) areas (shaded) occupy only a fraction of the domains
of interest we have defined. Normalized by the relevant values of a
and b, the relative errors for both quantities stay within ±0.5%. This
is substantially more accurate than in Fig. 20 of SM2003, and what
is essentially new here is that zero-error line is always placed close
to the middle of the shaded area. The computation of a/b a better
alignment of the zero-error line with the shaded area. This is a
somewhat unexpected benefit of EOS stiffening: the multipliers
rð^zÞ essentially plays the role of a preconditioner to EOS of Jackett
and McDougall (1995) by removing most of the nonlinear depen-
dency from depth (pressure). The remaining q can be more accu-
rately fit by the simple function in (4.7). (Note that the for small
depth, ^z ¼ 200m, the error pattern of a/b-ratio has a saddle point
in the vicinity of observed data, which makes a second-order
smallness of error as H, S depart from it.)
5. Does Boussinesq approximation still offer any simplification?
During the review process we were challenged why a Bous-
sinesq code should still be used, given the recent theoretical enthu-
siasm and advocacy for the non-Boussinesq alternative (McDougall
et al., 2002; Greatbatch and McDougall, 2003; Marshall et al., 2004;
Griffies, 2004). While we acknowledge that some aspects of oceanic
modeling become more natural in the non-Boussinesq framework,
notably the response to thermodynamic forcing (Mellor and Ezer,
1995; Huang and Jin, 2002), the value of Boussinesq approximation
is due to accuracy and dynamical parsimony in retaining the impor-
tant behaviors in oceanic circulation while excluding extraneous ef-
fects (e.g., acoustic waves) that can be computationally difficult
(Spiegel and Veronis, 1960; Mihaljan, 1962; Zeytounian, 2003).
The non-Boussinesq generalization does not bring in any important
additional circulation behaviors (Greatbatch, 1994; Dukowicz,
1997) and a posteriori comparison between Boussinesq and non-
Boussinesq results has shown that other aspects of the modeling
codes (specifics of initialization, version of EOS, etc.) produce com-
parable or even greater sensitivities (Losch et al., 2004).
If one seeks a non-hydrostatic generalization of an oceanic sim-
ulation code for smaller, faster dynamical phenomena, the Bous-
sinesq approximation offers a major simplification by excluding
dynamic pressure from EOS, which allows a decomposition of
pressure into hydrostatic free-surface, hydrostatic baroclinic, and
non-hydrostatic components that then can be treated in quasi-
independent ways (e.g., Kanarska et al., 2007). For hydrostatic
codes, the simplifications due to Boussinesq approximation are
somewhat less pronounced (besides the logistical efficiency of
not having to compute q-weighted control volumes) which leads
to the conclusion that Boussinesq approximation is not useful at
all, because one can solve an non-approximated system instead
(e.g., de Szoeke and Samelson, 2002; Marshall et al., 2004). Still,
51
there are fewer feedback loops and interdependencies among the
the model equations in the Boussinesq case, and accordingly, fewer
decisions about numerical splittings are to be made. In this and
previous studies we have described generally successful treatment
of computationally delicate couplings between mode-splitting,
time-stepping, and realistic EOS algorithms in a Boussinesq code
with accurate, numerically stable and efficient mode-splitting pro-
cedure without the necessity of excessive time filtering, with tracer
conservation and constancy preservation and with acceptable
EOS-related errors in the PGF. In this section we examine how
these coupling issues are manifested in a non-Boussinesq algo-
rithm like the ones presently used (e.g., in MOM4; Griffies, 2004)
and suggest some directions for further non-Boussinesq model
improvements.
Realistic oceanic simulation brings two particular issues that
are outside the standard consideration of the Boussinesq approxi-
mation: the free surface f and the necessity to reintroduce
compressibility and pressure into EOS in order to obtain a correct
description of the thermobaric effect and other EOS-related nonlin-
earities. These issues raise subtle questions, most importantly
whether to use the full dynamic pressure P in EOS or approximate
it with, e.g., the reference value P0 = q0gz; the classical Boussinesq
does not offer any guideline. Dewar et al. (1998) found that there is
a quantitatively significant sensitivity to this choice and advocated
the use of full dynamic pressure in EOS computed self-consistently
between q and P related by hydrostatic balance. Dukowicz (2001)
pointed out the error discovered by Dewar et al. (1998) is self-
canceling in a non-Boussinesq model, because the pressure gradi-
ent and density are essentially multiplied by a common multiplier,
and the modification proposed by Dewar et al. (1998) actually
brings error into the Boussinesq code they used10 [see also Section 2
following (2.11)]. Dukowicz (2001) proposed to correct the PGF error
while keeping both the Boussinesq approximation and the the use of
reference pressure in EOS by stiffening the latter.11 However, his ap-
proach received limited support. At first, because of the alternative
proposal to re-interpret Boussinesq velocity as normalized mass flux
per unit area qu/q0 as described by McDougall et al. (2002).12 They
advocate full P in EOS and demonstrated that no EOS stiffening is re-
quired to obtain correct geostrophically-balanced flow. A side effect
of their approach is that it disturbs the relationship between Coriolis
and advection terms, which interferes with the derivation potential
vorticity conservation even in the simplest case of barotropic, but
slightly compressible fluid with free surface (Appendix A). Secondly,
by noting that in a hydrostatic model it is actually not very difficult
to compute density via EOS using in situ pressure self-consistent
with the density being computed,13 therefore negating one of the
starting arguments of Dukowicz (2001). The third reason for little
interest in Dukowicz (2001) comes from the realization that a
free-surface, hydrostatic oceanic code can be relatively easy
10
Although Dewar et al. (1998) did not explicitly identify in their formulas that they
used a Boussinesq code, this choice is evident from their Footnote 2, which discusses
non-existence of sound waves and the absence of a time derivative in the continuity
equation, as well as from the general context of the MOM2 code available at that time,
which they used.
11
Note that while Dukowicz (2001) suggests using both reference pressure
P0 = q0g z in EOS and stiffening of EOS at the same time, either measure taken
alone is sufficient to solve the Case A dilemma of Dewar et al. (1998) in a Boussinesq
model, cf., (2.11). However both measures are needed in the more general baroclinic
case.
12
It is interesting to note that the idea of increasing/decreasing velocities depending
on local density to eliminate approximation errors can be traced back to Oberbeck
(1888, Sec. II, paragraph (5)) who was interested in explaining steady motions in the
atmosphere.
13
This requires knowledge of the state of free-surface (or bottom pressure in a non-
Boussinesq code) which becomes available only after the advance of barotropic mode
is complete. This obstacle is typically addressed by one-time-step lag of the pressure
field in EOS, e.g., Griffies and Adcroft (2008, see Sec. 11.4 there).
52 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70
converted from Boussinesq to non-Boussinesq by replacing the dis-
cretized vertical volume factor q0Dz to qDz (Greatbatch et al.,
2001), or by noting that non-Boussinesq equations can be trans-
formed into pressure coordinates where they resemble Boussinesq
written in z-coordinates (de Szoeke and Samelson, 2002; Huang and
Jin, 2002; Losch et al., 2004; Marshall et al., 2004). Non-Boussinesq
model requires the use of full dynamic P in the fully compressible,
realistic EOS: otherwise it would reintroduce the error pointed out
by Dewar et al. (1998). This is natural and requires no extra effort in
pressure coordinates (in fact, it is impossible to separate ‘‘coordinate’’
pressure, which is also dynamic pressure from EOS pressure), but a
special effort should be taken if using approach of Greatbatch et al.
(2001). The use of dynamic P does not reintroduce acoustic waves
because they are excluded by the hydrostatic approximation alone.
What remains somewhat unsettled in the literature is that the
choice of using reference q0gz vs. full pressure P in EOS is not
arbitrary, but should be tied to whether the model is Boussinesq
or not. Thus, following the advice of Dewar et al. (1998), Losch
et al. (2004, Sec. 2a. Initialization) advocate the use of P in EOS
for both Boussinesq and non-Boussinesq versions of MITgcm they
have compared and criticize Huang and Jin (2002) for doing it
otherwise (P for non-Boussinesq and z for Boussinesq) which is
actually a better choice. Similarly, P-dependent EOS appears in
Boussinesq equations of Marshall et al. (2004), even thought it
clearly breaks the completeness of non-Boussinesq-P—Bous-
sinesq-z-coordinate isomorphism, and changing to z-dependent
EOS repairs this without causing any downstream contradiction
(i.e., EOS depends on coordinate in both cases). The same applies
to de Szoeke and Samelson (2002) who in their remark in Section 3,
p. 2197, left column, indicated Dewar et al. (1998) as the reason for
this choice. Furthermore, the classical Boussinesq (with pressure-
independent EOS) equations have proper potential-to-kinetic en-
ergy conversion resulting in total energy conservation. This prop-
erty is lost if a nonlinear pressure-dependent EOS is introduced
into Boussinesq (or anelastic) equations, cf., Ingersoll (2005). Re-
cently Young (2010) showed that to maintain the energetic consis-
tency in the case of non-linear EOS the dynamic part of pressure in
EOS must be excluded, providing yet another argument for using
 q0gz in EOS in Boussinesq models.
Once this correspondence is respected, the only surviving argu-
ments for EOS stiffening in Boussinesq model are:
 conceptually it brings the model closer to the original Bous-
sinesq physical framework where the only density variations
which matter are the ones which contribute to buoyancy varia-
tions (Oberbeck, 1879, 1888; Boussinesq, 1903, see p. VII, pp.
154–161, and pp. 172–176 there; Mihaljan, 1962). Bulk com-
pressibility does not translate into buoyancy and therefore must
be excluded. Thermobaric EOS pressure dependency does,
hence is to be kept. EOS pressure is decoupled from the
dynamic, and becomes basically a function of location (in the
classical Boussinesq EOS pressure is constant or not needed at
all). While these guidelines help to simplify theoretical inter-
pretations, they actually have very little practical consequences
in the hydrostatic models, and mostly related to the treatment
of free surface. This is not surprising, given that in a Boussinesq
model the absolute density (as opposite to density perturba-
tion) appears only in the context of free-surface part of pressure
gradient term. Our analysis reveals that while bulk compress-
ibility is mainly dynamically passive (Sections 3.1, 3.3, and
3.4), its re-introduction into a Boussinesq code brings changes
which cannot be interpreted as correct physical effects (e.g.,
proper decrease of barotropic wave phase speed if compressibil-
ity is taken into account; also the appearance artificial multipli-
ers q/q0), are dependent on subtle code decisions (e.g.,
qEOS(H, S, z) vs. qEOS(H, S, f  z)) and, not surprisingly again,
are associated with the barotropic mode (Section 3.2). In a non-
hydrostatic code these guidelines bring a substantial simplifica-
tion by allowing decoupling of the non-hydrostatic pressure
from buoyant, and treating the former one as Lagrange multi-
plier to enforce nondivergence;
 as pointed out by Dukowicz (2001), because of nonlinearity of
EOS pressure-dependency (only bulk r(P) nonlinearity is of con-
cern for this purpose), remapping P = P(P) via (2.15) results in a
more accurate referencing of bulk pressure in comparison with
just approximating it with q0gz. In practice this translates into
re-tuning of pressure-related coefficients in z-dependent EOS if
they are originally tuned for using in situ pressure, but no extra
computational cost;
 if, due to specific needs of a particular ocean model, there is a
preferred functional form of EOS different from the available
standard EOS. Separating the bulk pressure effect from EOS
‘‘preconditions’’ it to the extent that it can be fitted more easily
or more accurately with the desired form. Similarly, after iden-
tifying that bulk compressibility is dynamically passive, exclud-
ing it before computing pressure-gradient terms is preferable
over relying on numerical cancellation within the PGF scheme
in a sigma-model; and
 all existing mode-splitting algorithms in split-explicit models
can be subdivided into two categories:
(i) a priori ‘‘physicist’s’’ splitting by extracting a shallow-water
like ‘‘fast’’ term (for Boussinesq models this is typically,
gDr\f or  (gD/q0)r\(qsurff), where qsurf is surface
density at surface; and for non-Boussinesq ðg=q0 Þpb r? p0b
where pb and p0b are bottom pressure and its perturbation rel-
atively to static reference) from the vertically-integrated PGF,
while the remainder is treated as slow-time ‘‘forcing’’ (e.g.,
Berntsen et al., 1981; Blumberg and Mellor, 1987; Bleck and
Smith, 1990; Killworth et al., 1991, see also Sec. 7.7 in Griffies,
2009, for an overview). In this approach no effort is made to
take into account the influence of stratification within the
terms recomputed during fast-time stepping14; and
(ii) ‘‘mathematical’’ splittings motivated by the operator-
splitting theory, where the fast-time pressure-gradient
terms are designed to capture as close as possible (subject
to practicality of implementation and cost) the time tenden-
cies of the non-split 3D system. In practice this translates in
introduction of a special density-weighting into ‘‘fast’’ terms.
The residual ‘‘forcing’’ terms have very minor dependency on
the state of free surface field. Higdon and de Szoeke (1997),
Hallberg (1997), and SM2005 belong to this category.
Both techniques rely on essentially the same small parameters,
notably separation of time scales between the baroclinic and baro-
tropic processes which is ultimately linked to the smallness of den-
sity perturbations, however the second captures the leading-order
corrections due to non-uniform density and treats them as ‘‘fast’’,
while the first one keeps them within ‘‘forcing’’. Naturally, algo-
rithms from the second category produce more accurate splitting
(second-vs. first-order errors with respect to the associated small
parameters) and require less fast-time filtering for the numerical
stability. However, they are also tend to be more code-specific,
and, more importantly, none of the available to date is designed
to be compatible with fully-compressible EOS.15 For Boussinesq
14
This influence was noticed and suspected to be the cause of numerical instability
by Killworth et al. (1991, see Eq. (31) and the discussion in Section 3 c. Coupling
between modes there).
15
Here we emphasize that although Higdon and de Szoeke (1997) is formally non-
Boussinesq, one of their starting assumptions is that barotropic changes in free
surface (or bottom pressure) cause proportional changes in thicknesses of each layer.
This assumption breaks down if the fluid is compressible.
 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70 53

model like ROMS it is acceptable, if EOS is stiffened. Conversely, re- leaves the slow-time continuity Eq. (5.1) no other role than com-
introduction of bulk compressibility depletes its order of accuracy puting vertical velocity wk+1/2, which in this context has the mean-
from the second to the first, thus negating the advantage in splitting ing of finite-volume fluxes across the interfaces between
accuracy relatively to the first group. vertically-adjacent grid boxes moving with f. The compatibility
For non-Boussinesq models with fully compressible EOS a split- conditions (5.4) ensure that the bottom and top kinematic bound-
ting algorithm comparable in accuracy with that of Higdon and de ary conditions for vertical velocity computed are satisfied. while
Szoeke (1997) or SM2005 is yet to be designed: using dynamic P is there is some freedom in choosing how to compute the flux-vari-
EOS also means that f influences EOS via P, which means that ables with time index n + 1/2 (n.b., multiple variants of ROMS code
dependency must be somehow reflected within the ‘‘fast’’ terms. exist), two important properties of (5.1)–(5.5) must always be re-
Parameter estimate in Section 3.2 suggests that the influence of spected: formally substituting qk  1 into (5.5) turns it into (5.1),
compressibility on barotropic phase speed is at least as large as and vertical summation of (5.1) results in (5.3). These guarantee
the influence of stratification, thus neglecting the former would that the tracer Eq. (5.5) have simultaneous conservation and con-
negate any accuracy advantage of Higdon and de Szoeke (1997) stancy-preservation properties. Eq. (5.1)–(5.5) are solved explicitly
or SM2005 in comparison with simple splitting, if fully compress- and sequentially (non-iteratively) as described in SM2005: com-
ible EOS is used. pute r.h.s. terms for the 3D momentum equation; compute forcing
(coupling) terms for the barotropic mode; update barotropic vari-
5.1. Comparison of discrete time stepping in Boussinesq and non- ables and their fast-time-averaged values hf; U; Vi and hhU; Vii; up-
Boussinesq models date 3D momentum equations and enforce the compatibility
conditions (5.4); advance tracer concentrations with (5.5); then
The ROMS time-stepping algorithm guarantees that the follow- roll the procedure over to the next time step. Notice that EOS does
ing semi-discretized equations hold exactly: not participate anywhere in (5.1)–(5.5) it is needed only at the
stage of computing the PGF for the momentum equation. We refer
 nþ1=2 this type of algorithm as a logically-Boussinesq, hydrostatic, free-
Dznþ1
k ¼ Dznk  Dt  r? ðDzk uk Þ þ wkþ1=2  wk1=2 ð5:1Þ
surface code.
   
Dznþ1 ¼ Dzk Dzk ; hfinþ1 ¼ Dzk 1 þ hfinþ1 =h
ð0Þ ð0Þ
ð5:2Þ Next, consider the non-Boussinesq version of (5.1)–(5.5):
k
h
hfi nþ1 n
¼ hfi  Dt  r? hhUii nþ1=2
ð5:3Þ
qnþ1 nþ1
k Dz k ¼ qnk Dznk  Dt  r? ðqk Dzk uk Þ þ qkþ1=2 wkþ1=2
inþ1=2
X
N
nþ1=2 nþ1=2 qk1=2 wk1=2 ð5:6Þ
Dzk ¼ h þ f and hhUii  hhU; Vii
k¼1  
( )nþ1=2 Dzk ¼ Dzk Dzð0Þ
k ;f ðunchangedÞ ð5:7Þ
XN X
N
¼ Dzk uk ; Dz k v k ð5:4Þ  
k¼1 k¼1
q nþ1 h þ fnþ1 ¼ q n ðh þ fn Þ  Dt  r? hhq Uiinþ1=2 ð5:8Þ
h
Dznþ1 nþ1
¼ Dznk qnk  Dt  r? ðqk Dzk uk Þ þ qkþ1=2 wkþ1=2 where
k qk

inþ1=2 1 X
N X
N

qk1=2 wk1=2 ; ð5:5Þ q ¼  Dzk qk and Dz k ¼ h þ f ð5:9Þ

h þ f k¼1 k¼1
( )nþ1=2
XN X
N
where for symbolic simplicity we omit horizontal indices, tracer dif-  Uiinþ1=2  hhq
hhq  Viinþ1=2 ¼
 U; q qk Dzk uk ; qk Dzk v k ð5:10Þ
fusion terms, and heat and fresh-water fluxes at the ocean surface. k¼1 k¼1
Dt is the ‘‘slow’’-time step (for the 3D ‘‘baroclinic’’ mode); time indi-

ces n and n + 1 correspond to slow-time as well; r\ is the two-

dimensional, horizontal-coordinate divergence operator; Dzk are
qnþ1 nþ1 nþ1
k Dzk qk ¼ qnk Dznk qnk  Dt  r? ðqk  qk Dzk uk Þ
the vertical grid-box heights which depend on f and are therefore nþ1=2
time-dependent and also depend on the horizontal coordinates þqkþ1=2  qkþ1=2 wkþ1=2  qk1=2  qk1=2 wk1=2 ð5:11Þ
through f and topography; and q 2 {H, S, . . .} is material concentra-
tion (tracer). Eq. (5.2) states that Dzk are computed by perturbing
X
N
1
Dzkð0Þ which correspond to f = 0. ROMS uses the specific way of per- qk ¼ qEOS ðHk ; Sk ; Pk Þ where Pk ¼ g qk0 Dzk0 þ g qk Dzk :
turbing them given by the rightmost part of (5.2), but in principle it 0 2
k ¼kþ1
may be by any other choice that satisfies the left condition (5.4)
ð5:12Þ
[e.g., the rescaled-height version of the MITgcm (Adcroft and Cam-
pin, 2004); an implementation of the coastal coordinate of Stacey All vertical integrations are now done with density weighting. Nev-
et al. (1995); and models that allow only the uppermost grid-box ertheless, similarly to (5.1)–(5.5), substitution of qk  1 into (5.11)
to change with f while keeping all others fixed, as in MOM/POP turns it into (5.6), and the non-Boussinesq free-surface Eq. (5.8) is
(Griffies, 2004) and TRIM (Casulli and Cheng, 1992; Rueda et al., derived by vertical summation of (5.1). Because of this, the system
2007)]. conserves mass and the integral content and constancy for each tra-
In ROMS the barotropic mode is integrated forward explicitly in cer. In comparison with its Boussinesq counterpart, the new system
‘‘fast’’-time using much smaller time steps than the rest of the sys- exhibits a more sophisticated coupling among the equations. The
tem. A special fast-time-averaging procedure is employed to en- major difficulty is associated with the free-surface Eq. (5.8): one
sure that the averaged hfi and hhUii are exactly matched to satisfy needs to know q  nþ1 in advance in order to solve it. But knowing
the slow-time free-surface Eq. (5.3). This can be, in principle, q nþ1 , hence qnþ1
k , also requires Hnþ1
k and Snþ1
k because qnþ1
k is related
replaced with an implicit free-surface algorithm using special to them through EOS (5.12). Ultimately this precludes a sequential
precautions (Campin et al., 2004) to ensure that (5.3) is still re- algorithm comparable to that for (5.1)–(5.5).
spected. In either case, once fn+1 is known, all grid-box heights There are two known approaches to address this dilemma. The
Dznþ1
k within each vertical column (k = 1, . . . , N) are slaved to it. This first one is by Greatbatch et al. (2001) who proposed to break the
54 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70

cyclic dependency by replacing qnþ1

k and qnk with their forward- where pb is identified as ‘‘bottom pressure’’–the total weight of
extrapolated values, water column. Vertical velocity x ~ kþ1=2 computed from (5.16) has
the meaning of mass flux per-unit-area through, generally speaking,
nþ1
ðeÞn moving pressure surface. The kinematic boundary conditions at the
qðeÞ
k ¼ 2qnk  qn1
k and; similarly qk ¼ 2qn1
k  qn2
k ; ð5:13Þ
bottom and free surface (x ~ ¼ 0 at both) are respected automatically
because vertical summation of (5.16) yields (5.18). All other vari-
everywhere throughout (5.6)–(5.12) where they appear in the con-
 , qk = {Hk, Sk, . . .}, remain the same in both systems. Nei-
ables, uk, u
trol volumes. Subsequently this approach was adopted in MOM4
ther density qk, nor grid-box heights D zk appear anywhere in
(Griffies, 2004). A similar extrapolation was applied to convert a
(5.16)–(5.21), while substitution of q = const into (5.21) converts it
sigma-coordinate model (Mellor and Ezer, 1995), with credit given
into (5.16) which means that simultaneous constancy and conser-
to then unpublished work by Greatbatch and Lu). Despite the fact
vation of tracers (now in mass-weighted sense) can be achieved
that qnk is available before the update of (5.6)–(5.12) begins, it can-
independently from EOS as long as the compatibility conditions
not be used in Dznk qnk , because doing so implies that during the next
ðeÞnþ1 (5.18)–(5.20) are respected. Another similarity between the non-
time step qk will be replaced with qnþ1 k when it becomes avail-
ðeÞnþ1 Boussinesq pressure-coordinate system and its Boussinesq counter-
able;because qk –qnþ1
k
, this results in loss of mass conservation.
part is that EOS is involved only in computation of horizontal
With this precaution taken, however, the discrete mass and tracer
pressure-gradient terms (also stratification and isopycnic slopes
conservation properties exist in form of
  for the purpose of subgrid-scale mixing parameterization), but not
X
DAi;j Dzni;j;k 2qn1 n2 for density weighting of grid-box heights like in (5.6)–(5.12). How-
i;j;k  qi;j;k ¼ const
i;j;k ever, the role of EOS is reversed: now it is needed to compute grid-
X   box heights, geopotentials Uk+1/2, and, ultimately, free surface,
DAi;j Dzni;j;k qni;j;k 2qn1 n2
i;j;k  qi;j;k ¼ const; ð5:14Þ
i;j;k
Dpk
Dzk ¼  aEOS ðHk ; Sk ; P k Þ;
where DAi;j is the horizontal area of grid element i,j. This is not g
as accurate as mass, heat, and salt content conservation X k
UNþ1=2
Ukþ1=2 ¼ g Dzk0  gh; and f ¼ : ð5:23Þ
expressed in terms of simultaneous variables (i.e., computed as g
P 0
k ¼1
DAi;j Dzni;j;k qni;j;k qni;j;k with qni;j;k and qni;j;k ¼ ðH; SÞni;j;k related by EOS).
However (5.14) still guarantees that the associated errors are The pressure increments Dpk and EOS pressures Pk above are known
bounded with no tendency of accumulation with time. Additionally, as long as pb is known,
extrapolation with some weights being negative, as in (5.13), is
undesirable if monotonicity properties are required (e.g., near sharp Dpk ¼ pb  Dzkð0Þ =h where Dzð0Þ ð0Þ
k ¼ Dz ðx; yÞ
fronts or extreme salinity anomalies associated with river out- ð0Þ ð0Þ
¼ zkþ1=2 ðx; yÞ  zk1=2 ðx; yÞ; ð5:24Þ
flows); and, as we will show in the next section, extrapolation
(5.13) is a source of mode-splitting error and may cause numerical
after which
instability.
The second, and actually more promising approach to address X
N
P kþ1=2 þ Pk1=2
the solvability of (5.6)–(5.12) in the case of a hydrostatic model Pk ¼ and Pk1=2 ¼ Dpk0 : ð5:25Þ
is to convert them into pressure coordinates, (de Szoeke and 2
k0 ¼k
Samelson, 2002; Marshall et al., 2004). The principal idea is to note
ð0Þ ð0Þ ð0Þ
that qk Dzk can be interpreted as vertical pressure increment over Above zkþ1=2 ¼ zkþ1=2 ðx; yÞ, k = 0, 1, . . . , N, such that z1=2 ¼ h and
ð0Þ
grid box Dzk, hence zNþ1=2 ¼ 0 is an a priori defined generalized terrain-following coor-
dinate mapping function (cf., Eq. (1.8) from SM2005. Note that in
qnk Dznk ¼ ð1=gÞDpnk and qnþ1 nþ1
k Dzk ¼ ð1=gÞDpnþ1
k ; ð5:15Þ the case of uniform density q = q0 hence pb = q0g(h + f), the general-
after which (5.6)–(5.11) can be rewritten as ized pressure-sigma coordinate reverts back to the original z-sigma,
  Eqs. (1.9)–(1.10) from SM2005).
Dpnþ1
k ¼ Dpnk  Dt  r? ðDpk uk Þ þ x ~ k1=2 nþ1=2
~ kþ1=2  x ð5:16Þ In summary, the principal distinction between the Boussinesq
  and non-Boussinesq cases is that the Boussinesq free-surface Eq.
Dpk ¼ Dpk Dpð0Þ
k ; pb ð5:17Þ (5.3) entirely belongs to the barotropic mode (hence does not re-
quire any mode-splitting decision about which terms are fast and
pnþ1  iinþ1=2
¼ pnb  Dt  r? hhpb u ð5:18Þ
b which are slow), while all EOS-related computations occur entirely
where within the slow mode. This is no longer the case in (5.8) as it ap-
pears in the approach of Greatbatch et al. (2001). In the pres-
X
N
sure-coordinate aproach of de Szoeke and Samelson (2002) the
pb ¼ Dpk ð5:19Þ
equation for ‘‘bottom pressure’’ pb, Eq. (5.18) is entirely ‘‘fast’’,
k¼1
( )nþ1=2 however, despite the terminology, the vertically-integrated pres-
XN X
N
sure gradient term cannot be computed from pb alone, but requires
 iinþ1=2  hhpb u
hhpb u  ; pb v iinþ1=2 ¼ Dpk uk ; Dpk v k ð5:20Þ
k¼1 k¼1
the knowledge of geopotential heights and free-surface, which in
h inþ1=2
their turn require the use of EOS (5.23). Because EOS is pressure
Dpnþ1
k qk
nþ1
¼ Dpnk qnk  Dt  r? ðqk Dpk uk Þ þ qkþ1=2 x
~ kþ1=2  qk1=2 x
~ k1=2
dependent, and pressure-coordinate levels are scaled by pb (which
ð5:21Þ brings ‘‘fast’’ dependency in EOS), the barotropic system is closed
which are isomorphic to their Boussinesq counterparts (5.1)–(5.5) via EOS, which means that in order to develop an efficient and
subject to the correspondence between the variables, accurate solver mode-splitting unavoidably involves splitting of
EOS as well. This is inherently related to the incompleteness of iso-
Dpk $ Dzk morphism associated with free surface: the complete similarity
pb $ D ¼ h þ f ð5:22Þ between z-coordinate Boussinesq and pressure-coordinate non-
Boussinesq system would be only if both are rigid-lid (de Szoeke
x
~ kþ1=2 $ wkþ1=2
and Samelson, 2002), see their Section 4), which, in fact, decouples
 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70 55

barotropic (i.e., rigid lid) pressure from EOS. In the remainder of The solution procedure for (5.28), (5.29) uses values of f at the
Section 5 we re-examine known mode-splitting techniques in or- previous time steps (up to n) to compute the r.h.s. term. (In an oce-
der to explore the implication for splitting non-Boussinesq ocean anic code this happens through EOS.) Using fn and u  n at time t = tn
models with fully-compressible EOS, while avoiding introduction as the initial conditions, advance them using fast-time stepping to
of additional splitting errors. t = tn + Dt. The final values are accepted as fn+1 and u  nþ1 , and the
r.h.s. is recomputed in proceeding to the next baroclinic step.
5.2. Stability analysis of mode splitting algorithm in Greatbatch et al. We now demonstrate that this procedure is computationally
(2001) unstable unless some temporal filtering is employed. Without loss
of generality consider a one-dimensional analog of (5.28), (5.29),
Using the identity, which makes it possible to obtain the solution as a sum of left-
and right-traveling Riemann invariants:
q nþ1 þ q n sffiffiffi ! sffiffiffi !
q nþ1 ðh þ fnþ1 Þ  q n ðh þ fn Þ  ðfnþ1  fn Þ pffiffiffiffiffiffi @ h pffiffiffiffiffiffi @ h
2 ! @ t R gh  @ x R ¼

f  gh 
u f 
u
fnþ1 þ fn @t g @x g
þ ðq nþ1 n
q Þ hþ ;
2 1 gh 2fn  3fn1 þ fn2
¼  2  : ð5:30Þ
2 c Dt
one can introduce fast-and-slow splitting into (5.8) as follows (cf.,
Eq. (12.56) in Griffies, 200416): This is just a linear combination of one-dimensional versions
of (5.28), (5.29). Rþ and R can be solved for separately
 pffiffiffiffiffiffi
qðeÞ nþ1 þ qðeÞ n @f qðeÞ nþ1  qðeÞ n as Rþ ðt n þ Dt; xÞ ¼ Rþ tn ; x  Dt gh þ Q þ ðt n þ Dt; xÞ and R ðt n þ
 þ  ðh þ fÞ
2 @t Dt  pffiffiffiffiffiffi
! Dt; xÞ ¼ R tn ; x þ Dt gh þ Q  ðt n þ Dt; xÞ, where Q±(t,x) are re-
qðeÞ nþ1 þ qðeÞ n 
þ r?  ðh þ fÞu ¼ 0; ð5:26Þ sponses to the r.h.s. forcing in (5.30) with Q±(tn,x) = 0. Consider a
2  n  
f ^
Fourier component,  ¼ ðkÞn f eikx where the step multiplier
where qðeÞ is vertically averaged density computed from (5.13). For u ^
u
the purpose of the subsequent analysis, we discretize the slow- k is not yet known. Ideally, without the errors introduced but
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
time, but keep fast-time continuous (i.e., assume that barotropic numerical discretization, its values should be k ¼ e icDt 1e
gh=c 2
,
time steps are sufficiently small that the associated discretization hence jkj = 1, which corresponds to the finite-time-step phase incre-
errors are negligible in slow-time). Suppose density is computed ments for non-damped left-and right-traveling waves (cf., (B.8) in
using the non-approximated EOS with full P. In this case q depends Appendix B). The phase speed is slightly smaller than that for
on f through hydrostatic pressure. This dependency is taken into ac- incompressible fluid. The actual solution of (5.30) is
count within the slow-mode when computing terms containing qðeÞ , sffiffiffi !
but these terms are kept constant during the barotropic time-  pffiffiffiffi

R ðt; xÞ ¼ k n ^f hu ^ eik xðttn Þ gh
stepping. g
For further simplicity of analysis, we now restrict ourselves to pffiffiffiffi
the situation where density depends on pressure alone (Appendix 1 gh ^ n n1 n2
 ikx eikðttn Þ gh  1
 f 2k  3k þk e  pffiffiffiffiffiffi ;
B). In this case EOS is q = q1 + P/c2, where q1 = const is density at 2 c2 ikðt  tn Þ gh
the surface atmospheric pressure, and c is speed of sound. This
2
translates into the vertical profile q ¼ q1 egðfzÞ=c and which after reaching t = tn+1 = tn + Dt becomes
sffiffiffi !
2
q ¼ q1 ½egðhþfÞ=c  1=½gðh þ fÞ=c2  h

R ðt þ Dt; xÞ ¼ k n ^f ^ eikx eia
u
g
 q1 ½1 þ gðh þ fÞ=ð2c2 Þ þ   : ð5:27Þ
1 ^ n  eia  1
Substituting this into (5.26), assuming a flat bottom at  f 2k  3kn1 þ kn2 eikx  :
2 ia
z = h = const, linearizing about the rest state, and using the extrap-
olation rule (5.13), we obtain
nþ1 n
@f 1 gh fðeÞ  fðeÞ
¼ 
þ hr? u 
@t 2 c2 Dt
1 gh 2fn  3fn1 þ fn2
¼  2  : ð5:28Þ
2 c Dt
This and the barotropic momentum equation (Appendix B),
 þ g r? f ¼ 0;
@t u ð5:29Þ

comprise the barotropic mode. Unlike the more usual situation with
splitting in a stratified Boussinesq model, now it is the free-surface
Eq. (5.28) that receives r.h.s. the forcing term from the slow mode,
while there is no slow r.h.s. in the momentum equation. Density q is
still allowed to change due to compressibility, and (5.29) does not
imply smallness of density variation in vertical direction. Fig. 3. Absolute value of characteristic roots of (5.32) corresponding to the physical
modes as a function of a for  = 0.025, corresponding to h = 5500 m and
c = 1500 ms1. Because the mode-splitting ratio Dt/Dtfast is expected to be
16
The original code of Greatbatch et al. (2001) uses leap-frog stepping; later a more significantly larger than unity, a is allowed to exceed p, hence entering the range,
advanced, staggered time placement of tracers and momenta was introduced in where barotropic motions are aliased, if sampled with the baroclinic time steps,
MOM4. In this approach stepping of tracers and free surface is from n to (n + 1) th tn, tn+1 , . . .. Only two periods are shown; the continuation in a is nearly periodic with
baroclinic step. damping toward jkj = 1; jkj > 1 means numerical instability.
56 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70

pffiffiffiffiffiffi
The quantities a ¼ kDt gh and  = gh/c2 haveq the
ffiffi same meaning as dependency in EOS into the fast mode. Without any approximation
in Appendix B. Since R ðt þ Dt; xÞ ¼ knþ1 ^f hgu
^ eikx , EOS can be rewrittem as
sffiffiffi ! sffiffiffi !  
 eia  1
q ¼ r  q0 þ q0 ; ð5:33Þ
h ^f hu 1 
k ^f ^ ¼
u ^ eia  ^f 2  3k1 þ k2  :
g g 2 ia where r = r(f  z) chosen to absorb most of the EOS compressibility,
but is also simple to compute; q0 ¼ const is chosen in such that
ð5:31Þ
q0  q0 , and q0 is only weakly dependent on f. With (5.33) only
This pair of equations is homogeneous for f^f; u^ g, which means that the second parenthetical factor need be extrapolated in time. Then
it admits non-trivial solutions only for the specific values of k deter- Z f Z 0
@ @
mined by the characteristic equation, q dz ¼ q  ðh þ fÞdr
@t h @t 1
Z 0

1 sin a @f   @r
k2  2k cos a þ 1 þ ð2  3k1 þ k2 Þ ¼ 0: ð5:32Þ ¼ q0 þ q0 r þ ðh þ fÞ dr
2 a @t 1 @f
Z 0
±ia
The limit of  ? 0 recovers k = e , and one can also verify that @ q0
þ r ðh þ fÞdr: ð5:34Þ
k  e±ia(1/4) for a,  1, which matches (B.9) in Appendix B. For 1 @t
large a but still small , the splitting procedure distorts the ampli- Choosing r ¼ egðfzÞ=c
2
(which in r-coordinate becomes
tude, resulting in a possible numerical instability, jkj > 1 (Fig. 3). 2
r ¼ ergðhþfÞ=c ¼ er , where c is a representative constant value
This mechanism is generic to mode-splitting instability, and it is for speed of sound) converts the above into
associated with phase-delay of the slow-time r.h.s. terms—com-
Z 0  Z 0
puted only once at the beginning of the fast-time-stepping and kept @f @ q0
 q0 ð1 þ Þ þ q0 ð1  2rÞdr þ ðh þ fÞ  ð1  rÞdr
constant thereafter—relative to the more rapidly changing phase of @t 1 1 @t
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
fast-time terms. Because a has the meaning of phase increment for \fast" \mostly slow"
the barotropic wave during one baroclinic step Dt, a may exceed ± p ð5:35Þ
by several multiples, depending on the mode-splitting ratio. The
r
instability occurs even if the fast-and slow-time algorithms are both where we have expanded r = e  1  r +    for   1.
stable if taken separately, and even if there is numerical dissipation Even in the simplest case of barotropic compressible fluid,
within the slow-mode algorithm. q = q1 + P/c2 (hence q0 ¼ q1 and q0  0) the expression inside
The remedy for instability is a more accurate splitting; i.e., elim- [. . .] in the ‘‘fast’’ term is not equivalent to vertically averaged den-
inate or reduce the -terms, or to introduce artificial damping in sity q  –as follows from (5.27) q  ¼ q1 ð1 þ =2Þ in this case instead of
the fast-time-stepping, essentially preventing it from entering into q1(1 + ). This non-equivalence leads to the replacement of (5.28)
the aliasing range jaj > p while still maintaining jkj near unity. In with
the MOM4 code this is achieved by uniform averaging of both f ð1 þ Þ@ t f þ ð1 þ =2Þhr? u
 ¼ 0; ð5:36Þ
and u over a 2Dt interval, or 1Dt in the case of staggered time-step-
ping (Sec. 12.5 of Griffies, 2004). This introduces damping compa- which in combination with (5.29) yields the correct barotropic
rable to that of a backward-Euler implicit step and reduces the phase speed in agreement with (B.9). Unlike (5.28) the above does
temporal accuracy to first-order. A gentler fast-time-averaging not have any r.h.s. at all, hence produces no splitting error.
should take into account this instability mechanism, which is com- In the general case of q0 – 0, splitting (5.35) requires forward
parable in its strength with that associated with the splitting of extrapolation of q0 as in (5.13) with subsequent computation of
PGF terms (items (iii) and (iv) in Section 3.2). Furthermore, because the two-dimensional fields,
of its specific form of fast-time averaging, MOM4 does not have ex- Z 0 Z 0
act discrete consistency between the averaged free surface and q0 ¼ q0 dr and qc0 ¼ 2 q0 r dr; ð5:37Þ
1 1
averaged volume or mass fluxes (for the Boussinesq or non-Bous-
sinesq variants) in the slow-time step. Thus, the content conserva- with their discretized formulas,17
tion or constancy preservation properties for tracer advection X
N
1X N
cannot be maintained simultaneously, and one must chose be- D¼ Dz k ; q0 ¼ Dzk q0k ;
tween them. The selected choice is constancy, implemented by k¼1
D k¼1
( ! )
an auxiliary slow-time step for f in Eq. (12.95) of Griffies (2004)). 2 X
N
Dz k XN

This f is then used to compute new-time-step control-volumes qc0 ¼ Dz k þ Dz0k q0k : ð5:38Þ
D2 k¼1
2 0
k ¼kþ1
for the tracer update but thereafter discarded. The loss of content
conservation comes from the fact that the f updated this way is With (5.35), the non-Boussinesq nonlinear free-surface equation is
not equivalent to that obtained from the barotropic mode, result-
 h þ f @f

ing in a difference in control volumes. (Another source of non-con- q0 þ q0 þ q0 þ qc0 g 
servation is the Robert-Asselin filter needed in the case of leap-frog c2 @t

   h þ f 
time-stepping.) The choice of vertical grid that allows a change of
þ r? q0 þ q0 þ q0 þ q c0 g ðh þ fÞ 
u
only the top-most grid-box height with f only (while all others are 2c2
fixed) results in a slight redistribution of discrete values of density " #
@q 0 @qc hþf
0
within the vertical column in the case of purely barotropic mo- ¼ ðh þ fÞ þ g ; ð5:39Þ
tions. The existing non-Boussinesq splitting algorithms neglect this @t @t 2c2
effect, which leads to some non-conservation of mass.

17
5.3. A method for stabilizing algorithm of Greatbatch et al. (2001) by These sums do not rely on a particular structure of the vertical coordinate and are
self-normalizing in the sense that substituting q0 c0
k  1 yields q ¼ q ¼ 1. Even
0
splitting EOS
though Dzk depends on f, the resulting q0 and qc0 are only weakly dependent through
EOS pressure, but the f-dependency through Dzk cancels out (cf., Eq. (3.23) in
Dukowicz’ idea factoring of EOS provides a guideline to improve SM2005). This makes it possible to use the most recently available Dzk to compute q0
stability and splitting accuracy in (5.26) by moving most of f and q c0 .
 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70
where we back-substituted  = g(h + f)/c2 to expose all f dependen-
cies. The most accurate (but also the most complicated) time-split-
ting procedure precomputes the r.h.s. time derivatives of q0 and q c0
  
using slow-time variables, and keeps them constant during baro-
tropic time-stepping, while all f-terms, including the ones in
g(h + f)/c2 in both l.h.s. and r.h.s., are allowed to change at every
barotropic step. The splitting error of this algorithm can be esti-
mated as Oð2 Þ and Oð  Dðc2 Þ=c2 Þ. It does not rely on the smallness
of q0 =q0 . This algorithm replaces (5.14) with
X   sinesq models because it appears as a single prognostic variable un-
DAi;j Dzni;j;k r ni;j;k q0 þ 2q0n1 0n2
i;j;k  qi;j;k ¼ const
i;j;k
X   ð5:40Þ
DAi;j Dzni;j;k qni;j;k rni;j;k q0 þ 2q0n1 0n2
i;j;k  qi;j;k ¼ const:
i;j;k
These are closer to the simultaneous field-conservation laws since
r ni;j;k is computed using the same f as used to compute Dzni;j;k and
all f-terms in (5.39) evolve in fast-time with r evolving synchro-
nously with f without any splitting.
A simplified, but slightly less accurate procedure for (5.39) can
be devised by precomputing q c0 gðh þ fÞ=c2 -terms at the beginning
of barotropic stepping using fn and combining them with q0 þ q0 .
This admits additional Oð  f=h  q0 =q0 Þ ¼ Oðgf=c2  q0 =q0 Þ-errors
(definitely negligible in any practical case). Finally, precomputing
q0 gðh þ fÞ=c2 using fn brings Oð  f=hÞ ¼ Oðgf=c2 Þ errors. This is
only marginally more computationally demanding than the non-
linear analog of (5.28). Even in this case the splitting is more accu-
rate than (5.28), which admits OðÞ error. However, these
simplifications cause loss of synchronization between rni;j;k and
Dzni;j;k , resulting in some compromising of the conservation proper-
ties (5.40).
Splitting of PGF terms in the non-Boussinesq case follows Sec-
tion 3.2 using EOS (5.33). The multiplier r can be moved outside
the pressure gradient operator (2.14) and furthermore can be sep-
arated from the influence of f via (3.10). This means that q⁄ and q 
are now expressed entirely in terms of q0 and q0 , so the difference
from the Boussinesq model is in the appearance of r-terms as a
multiplier in the PGF when the momentum equations are rewritten
in conservation form.
5.4. An overview of mode-splitting algorithms in non-Boussinesq
models
Regardless of the type of vertical coordinate, vertical integration
of non-Boussinesq momentum and mass-conservation equation
yields,
" ! # SM2005 (cf., Eq. (3.15) there) addressed the latter issue by can-
@ q D2
ðqDuÞ þ g r?
   qDr? h ¼ advection; Coriolis;

@t 2 (5.41) as
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}
¼F
dissipation; forcing terms;
@ t ðq
 DÞ þ r? ðq  Þ ¼ fresh-water flux;
 Du
R
1 f
Rf Rf
where D ¼ h þ f; q  ¼ D h q dz; q ¼ 22 h z0 q dz0 dz; u
 ¼ q1D h qudz
D
are the total depth of water-column, vertically-averaged density,
vertically-averaged ‘‘dynamic’’ density [both are the same as in
(3.19)], and vertically-averaged density-weighted velocity, respec-
tively. The derivation of pressure-gradient term in this form can
be found in Section 3.1 of SM2005, up to Eq. (3.15), with the only
difference is that q
 ! q0 in terms containing time derivatives and
in the definition of u if Boussinesq approximation is applied. Anal-
ogous derivations, but expressed in terms Montgomery potential
can be found in Higdon and de Szoeke (1997, see their Section 2.1,
note the appearance of triangular sum in the second unnumbered
57
equation following (2.11) there), and also in Hallberg (1997). The
above can be rewritten in terms of bottom pressure, pb  g q D,
@
 Þ þ g r?   b
q p ðh þ fÞ
ðpb u  pb r? h ¼   
@t q 2
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} ð5:42Þ
¼F
Þ ¼   
@ t pb þ r? ðpb u
where the new prognostic variable pb is more natural for non-Bous-
der time derivative in the second equation, however, the presence
of f in PGF terms separately from q or pb cannot be eliminated com-
pletely. It should be noted that despite the common meaning F ’s
have different dimensions in (3.14), (5.41) and (5.42) – relatively
to (5.41), F is divided by q0 in (3.14) and is multiplied by an extra
g in (5.42). Nevertheless, for notational simplicity we retain the
same symbol F for all versions, and, consistently with SM2005
and with (3.15) in Section 3.2, we choose the sign of F as it appears
in the r.h.s. of barotropic momentum equations, hence negative sign
in (5.41) and (5.42).
Mode-splitting requires decomposition of vertically integrated
PGF into
F ¼ g½. . . ¼ \fast" þ \slow" ð5:43Þ
where ‘‘fast’’ term can be efficiently computed from either pb, q D, or
f field – whichever plays role of prognostic variable in the barotrop-
ic continuity equation, while ‘‘slow’’ must be made as weakly
dependent as possible from this variable. The inherent dilemma in
(5.41) and (5.42) is that while q
 D or pb/g may be considered as a sin-
gle variable, in the general case of compressible fluid changes in pb
translate in changes in both free surface f (hence changing D) and q 
(via dependency from pressure through EOS). In pressure-based
coordinated, knowing pb means that the distribution of pressure
throughout the water column is known, so one can use EOS to com-
pute specific volumes at each grid box, after which grid-box heights
become known as well. Vertical summation of heights yields D, and
therefore free surface f = D  h. However, this would work only if
there is no mode splitting, because it involves using EOS to compute
a 3D field, and EOS depends on pb, which changes in fast time.
Therefore a computationally efficient time-split model should em-
ploy some sort of approximation to allow computation of f via pb
in fast time solely from two-dimensional barotropic fields. The
second aspect is that in the presence of stratification and non-uni-
form topography F consists parts which tend to cancel each other,
so any splitting which disregards this balance is doomed to be
inaccurate.
celing the large terms by hand, first by rewriting the upper Eq.

D
@ t ðq  Þ þ gD q r? f þ r? q þ ðq  q
 Du  Þr? h ¼    ð5:44Þ
2
ð5:41Þ and, subsequently, substituting D = h + f into gD{. . .} term and sep-
arating terms which depend on f from those which do not. The
Rf
resultant f-dependent part is given by (3.18) [In SM2005 this is
F ¼ F ð0Þ þ F 0 decomposition via Eqs. (3.31)–(3.32) there.] In the
Boussinesq case this is all what is needed because f is also prognos-
tic variable of the barotropic continuity equation because of the
q ! q0 replacement in terms with time derivatives, hence
@ t ðq
 DÞ ! @ t ðq0 DÞ ¼ q0 @ t f. In doing so it is also assumed that both
q and q⁄ are independent of f, which in its turn imposes two restric-
tions: (i) either exclusion of f from EOS pressure, or incompressibil-
ity of EOS; and (ii) specific functional dependency of how vertical
coordinate system is adjusted by changes in free-surface elevation
– proportional stretching.
58 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70

The same two issues must be dealt with in the non-Boussinesq it vertically. Since there is no use for q
 and q⁄ within the fast term,
case using bottom pressure instead of free surface. A ‘‘physicist’s’’ there is no need to compute them as well, however we explicitly in-
split extracts the g pbr\f term18 from the PGF term in (5.42), clude all the terms in r.h.s. of (5.49) to facilitate
pthe
ffiffiffiffiffiffi error analysis.

   Neglecting all terms except ‘‘fast’’ yields gh for the phase
q p ðh þ fÞ
F ¼ g r?   b  pb r? h speed of external waves. Unlike in the Boussinesq case, it does
q 2
not depend on the choice of q0 and it should not in non-Boussinesq

  
q p ðh þ fÞ models. Although this is mainly correct, three effects are not cap-
¼ gpb r? f  g r? 1  b
q 2 tured: the slowdown of barotropic speed due to bulk compressibil-
1 ity (Appendix B), stratification (Appendices C and D), and
 g ½pb r? ðh þ fÞ  ðh þ fÞr? pb : ð5:45Þ topographic coupling between free-surface and stratification if
2
Then, utilizing the fact density variations are small relative to its r\h – 0 (cf., the second line in Eq. (3.32) in SM2005). As all three
mean value, pb is dominated by bulk part which does not change effects are present in the 3D mode, and are also depend on the fast
in time, so it is convenient to split pb into its reference value and variables, the differences contribute to mode-splitting error. The
anomaly, which in the simplest case19 are ‘‘mixed’’ term is baroclinic because it vanishes if density is uniform
due to ððq =q  Þ  1Þ ! 0, but it also contains p0b which evolves in
q ¼ q0 þ q0 q0  q0 ;
ð5:46Þ fast time. This term is mainly responsible for baroclinic slowdown
pb ¼ q0 gh þ pb p0b  q0 gh
0
(recall that q < q  for positively stratified fluids) and for topo-
after which (5.45) can be rewritten as graphic coupling, if h – const. It should also be noted that in the
"  !# case of compressible fluid the ratio q =q  also depends on f and
q q0 gh2 p0b þ q0 gf p0b q0 gfÞ p0b due to pressure dependency in EOS, so the ‘‘slow’’ term also con-
F ¼ gpb r? f  g r? 1  þ hþ
q 2 2 2 tains some traces of p0b -dependency. The accuracy of this split relies
1    on the following smallnesses:
 g p0b  q0 gf r? h  hr? ðp0b  q0 gfÞ þ p0b r? f  fr? p0b
2 |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} q q
 q0  r? h  1;
 gh
¼0; if f¼p0b =ðq0 gÞ  1  1; 1 ð5:50Þ
q q  c2
ð5:47Þ
and mode-splitting error has terms proportional of each of the
so far without any approximation. This still needs a relationship to
three. One of the main difficulties specific to the non-Boussinesq
translate pb into f within the barotropic mode (hence using baro-
case is the definition of the reference state for pb: while (5.46) by
tropic variables alone) in order to close the system. In the simplest
itself is merely a variable change and does not constitute any
case it is approximated as
approximation, (5.46) in combination with (5.48) is a rather strong
f ¼ p0b =ðq0 gÞ; ð5:48Þ assumption: free-surface elevation f computed by 3D mode via EOS
which turns (5.42) into does not necessarily agree with (5.48),

  
@   p0 q  
 Þ þ g q0 gh þ p0b r? b ¼  g r?
ðpb u  1  p0b h pb q p0 p0b
@t qg q f¼Dh¼ h¼h 01 þ b vs: f ¼ : ð5:51Þ
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl0ffl} |fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} q g q q g q0 g
\fast" \mixed" |fflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflffl}
"  # bias
q q0 gh2
 g r? 1 þ In the case of flat bottom the bias mostly differentiates out from the
q 2
|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl} ‘‘fast’’ term g pbr\f leaving only a minor error of Oððq   q0 Þr? fÞ,
  
baroclinic;\slow" and mostly cancels out from all other terms in (5.47). However, in
@ t p0b þ r? q0 gh þ p0b u
 ¼ ; the case of non-flat topography the bias does affects the split (hence
ð5:49Þ the second parameter in (5.50)) and is not easy to remove by includ-
ing the ‘‘mostly slow’’ into the terms recomputed during fast-time
where dots (. . .) in the first equation indicate other slow terms, such
stepping: if computed as a part of 3D algorithm, it uses the correct
as advection, Coriolis, forcing, viscous dissipation terms, as well as
values of free surface, but once treated as fast, there is no other
minor 3D correction terms in pressure-gradient associated with
choice than to use an approximate f form (5.48). For the same rea-
the fact that (5.48) is only approximately match the actual free sur-
son it is not easy to modify this algorithm to account for the effect of
face as it is computed by the 3D mode using the actual EOS. Typi-
slower barotropic phase speed due to the influence of stratification
cally only the ‘‘fast’’ term is recomputed at every barotropic time
within the ‘‘fast’’ term.20 The other aspect specific to non-Boussinesq
step, while all other terms are computed by the 3D mode and play
case is that for a fully compressible EOS it is no longer possible to
the role of 3D ? 2D forcing. The particular form of ‘‘fast’’ term as in
interpreted the ratio of q =q as purely due to stratification: it is af-
(5.49) appears in Griffies and Adcroft (2008, see Eqs. (182)-(183));
fected by bulk compressibility as well. In the next section we will de-
also Griffies (2009, see Section 7.7.4, Eqs. (7.137)–(7.139) there). A
sign a more accurate split which avoids the use of a statically defined
similar approach of considering perturbation p0b relative to a stati-
reference state altogether.
cally-defined rest-state reference field, with subsequent approxi-
mate closure to compute geopotential was taken by Marshall
et al. (2004, see Eq. (27) and the two paragraphs after it, Eqs. 5.5. Mode splitting using incremental variables: The Boussinesq case
(40)–(42), and, finally, Ul = bsrs at the end of Section 3a) for an
atmospheric model. The actual codes do not compute ‘‘slow’’ and Because the theoretical rationale for splitting the total fluid mo-
‘‘mixed’’ terms literally as in (5.49), but instead compute the full tion into barotropic and baroclinic modes mainly comes from the
baroclinic 3D PGF term using an appropriate scheme and integrate orthogonality of wave motions in the case of linearized equations
over flat bottom, and, accordingly, virtually all theoretical studies
18
about numerical stability of splitting algorithms are done in the
This particular form is motivated by the PGF term of a barotropic layer of
compressible fluid, see Appendix B, Eq. (B.2), note the mutual placement of pb and f.
19 20
A more elaborate decomposition involves using reference profile for density This aspect is somewhat parallel to the initialization issues for pressure-
(specific volume) instead of constant value can be found in Appendix B of de Szoeke coordinate models, Griffies (2009, see Sections 7.2, 7.3 there, especially the difficulties
and Samelson (2002). with the sigma-pressure model, Section 7.3.3.7.).
 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70
flat-bottom framework, the justification of splitting in the presence
of bottom topography is more obscure: while the separation of
time scales remains well-defined, the orthogonality principle is
no longer valid. Nevertheless, the splitting must be dealt with in
the general case of non-flat bottom despite the fact that a complete
analysis of numerical stability is no longer possible. SM2005 (and
even more so in Section 3.2 Shchepetkin and McWilliams (2008))
advance the following principle: the modification of vertically-
integrated pressure gradient term by advancing the barotropic
mode from 3D time step n to n + 1 should resemble as close as pos-
sible the difference between the original and the new vertically-
integrated pressure gradient as it would be computed by the 3D
mode using the new free-surface field, but the same baroclinic
density distribution. The intuitive rationale for this principle is
the presumption that if the barotropic mode drives the barotropic
fields toward a new equilibrium when advancing from n to n + 1,
that equilibrium should be disturbed as little as possible during
the next time step. The resultant form of the barotropic pressure
gradient term coincides with that of Higdon and de Szoeke
(1997) in the case of flat bottom (subject to non-Boussinesq to
Boussinesq translation), but it also leads to specific form of topog-
raphy-and-free-surface terms, and practical experience of gaining/
loosing stability of the model when switching between treating/
not-treating these as ‘‘fast’’ terms.
This leads to an equivalent, but alternative to SM2005 interpre-
tation of the splitting procedure if we rewrite (5.41) in terms of
incremental variables – basically changes caused by the barotropic
mode starting from the initial state corresponding to 3D time step
n. Thus we introduce df = f  hfin, and, consequently, substitute
f ! f þ df  hfin þ df hence D ! D þ df  h þ hfin þ df
into (5.41) while algebraically separating all terms containing df.
The Boussinesq version of it becomes
@
ðq ðD þ dfÞu
 Þ  F þ g ðD þ dfÞr? ðq dfÞ þ g q dfr? f
@t 0
df2
g r? q þ g ðq  q
 Þdfr? h ¼   
2
@
Þ ¼   
df þ r? ððD þ dfÞu
@t
where now only df and u  are changing during the barotropic time
stepping (e.g., f is kept constant in fast-time in gq⁄df r\f – the last
term in the first line of the first equation).
The solution procedure is therefore as follows: before starting
the barotropic mode the full vertically-integrated pressure-gradi-
h  2 i
ent term F ¼ g r? q2D  q  Dr? h is precomputed via the actual
pressure gradient scheme of 3D mode using the state of free-surface
at time step n and the predicted or extrapolated density field at
n + 1/2. Note, this is not equivalent to computing q  and q⁄ first
and then computing F as suggested by this formula using an ad
hoc spatial discretization – when considering continuous equations
these two forms are equivalent to each other, but discrete forms
they are not. In a sigma-model not exercising proper care here
would result in an unacceptable pressure gradient error passed into
the barotropic mode. Instead, the precomputed q  and q⁄ are to be
used only in terms containing df. Once all preparations are
complete, the barotropic variables are advanced and fast-time
averaged,
   
0 1 0 1
nþ1 hfinþ1 ¼ hfin þ hdfinþ1
  hdfi
B C
df ¼ 0 B C
! @ hðD þ dfÞu  inþ1 A ! B Dnþ1 ¼ h þ hfinþ1
  in
u ¼ hu @ A
inþ1
 iinþ1=2
hhðD þ dfÞu  inþ1 ¼ hðDþdfÞunþ1
hu Dþhdfi
59
where, obviously, the initial state for df = 0, and the right-most
operations symbolize setting new state of the ‘‘slow’’ variables,
which are also used as the initial state for the barotropic mode dur-
ing the next 3D time step.
In comparison with (3.18) the incremental form (5.53) com-
pletely ‘‘hides’’ the baroclinic-topographic hydrostatic balance into
the 3D term. Assuming spatially uniform free surface, hence setting
f = const – 0 in (3.18) turns it into gf½ðh þ f=2Þr? q þ
ðq  q Þr? h. It must vanish in the case of horizontally uniform
(flat) stratification. However its vanishing requires cancellation of
the two terms inside square braces [] because in the presence of
topography flat stratification results in non-uniform q  and q⁄.
The incremental form (5.53) ensures this cancellation for any
f = const (not necessarily f = 0) as long as the vertically integrated
pressure gradient term computed by 3D mode vanishes: recall that
3D computing of pressure gradient term is inherently more accu-
rate that via q  and q⁄. It should also be noted that smallness of
df/D is more justifiable than smallness of f/D (or, similarly f/h) be-
cause in deep areas where phase barotropic speed is large both are
small, but comparable, since f may change rapidly per time step, if
barotropic is running using time step close to the maximum al-
lowed by stability. On the other hand, in shallow areas where
free-surface changes may no longer be very small in comparison
with depth, the barotropic motions may resolved in 3D time step
resulting is smooth changes, hence df is expected to be smaller
than f. For this reason, if linearization is required, e.g., in the case
of implicit treatment of free surface using a third-party elliptic sol-
ver, the use of df in (5.53) provides a more natural approach. It
should be noted that the most commonly used approach retains
only the gDr\(q⁄df) term, and usually with q⁄ replaced with q0.
ð5:52Þ
Eq. (5.53) suggests two more terms, and notably the very last of
the first equation couples stratification with topography and free
surface, but still the whole approach yields essentially the same
5-diagonal elliptic equation for df.
5.6. Mode splitting using incremental variables: non-Boussinesq case
with fully compressible EOS
Another consequence of incremental form (5.53) is that it can
be generalized to the case of non-Boussinesq model with com-
ð5:53Þ pressible EOS. Similarly to (5.52) we introduce incremental
variables
q D ! q D þ dm;
f ! f þ df hence D ! D þ df
ð5:55Þ
q ! q þ dq
q ! q þ dq
after which the barotropic continuity equation becomes
@
dm þ r? ððq Þ ¼   
 D þ dmÞu ð5:56Þ
@t
which identifies dm as the natural fast-time prognostic variable.
The incremental form of the non-Boussinesq version of (5.41)
can be derived by noting that
!  
q D2 q q D  D
r? q
 Dr? h  r?  q
 Dr? h
2 q 2
q q ðq
 D þ dmÞ  ðD þ dfÞ
! r? þd    ðq
 D þ dmÞr? h;
q q 2
C ð5:57Þ
where, we must provide some the means to compute responses df
ð5:54Þ and dðq =q
 Þ to changing dm. Both relationships involve EOS,
60 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70

   
however they are needed during fast-time stepping, so they must q r q0 þ q0 1 q0  q0
¼    1   1 ; ð5:66Þ
be expressed somehow in terms of 2D fields only. q r q0 þ q 0 6 q0
The principal assumption that at the ‘‘slow’’ time step n the
which leads to an estimate of how ðq =q
 Þ responds to perturbation
free-surface field f and the three-dimensional density are in hydro-
in free surface df, and dm,
static equilibrium with each other as governed by fully compress-  
ible EOS, e.g., given a distribution of masses throughout each q r q q 1 1 df
¼  ! 1  
vertical column, hydrostatic pressure is computed by integration q r q q 6 6 D
from surface downward, after which specific volume and height  
q 1 1 dm
of each grid box is computed via EOS, and vertical summation of   1    ; ð5:67Þ
q 6 6 qD
grid box heights results in f. Once the barotropic mode departs
from the state corresponding to time step n, the change in dm re- where we have neglected quadratically small terms Oð2 Þ and
0 0
sults in change of both free surface df and vertically averaged den- Oð  q qq Þ.
0
sity dq in such a way that the hydrostatic equilibrium is Substitution of (5.58) and (5.67) into r.h.s. of (5.57) turns it into
maintained. This means that an increase in dm results in propor- 2    3
    ðq
 D þ dmÞ D þ 1  12  dm
tional increases in both df and dq, q 1 1 dm q
r? 4  1      5
  q 6 6 qD 2
1 dm 1  12  dm 1 dm
dq
¼  ; df ¼   1    ; ð5:58Þ
2 D 1 þ 12   dm
q D
q 2 q  ðq
 D þ dmÞr? h; ð5:68Þ

where the 12   dm which, after some algebraic transformation yields

q D term in denominator quadratically small, and is
   
merely to keep @ q 1
ððq
 D þ dmÞu  Þ  F þ gDr?   dm
q D þ dm ¼ ðq þ dq ÞðD þ dfÞ ð5:59Þ @t q 2
     
q 1 q 1
as an exact identity. Above   1 is effective vertically integrated þ g    dmr? f þ g   1   dmr? h
q 2 q 2
compressibility parameter computed via EOS. In the simplest case
   
(see Appendix B)  = gD/c2, where c is speed of sound, resulting in q 5 dm2 1 dm3
þ g r?     2 ¼ 
an estimate   0.025 for deep ocean. In a more general case we as- q 6 q 6 q D
sume EOS in form of @
dm þ r? ððq  D þ dmÞuÞ ¼    ð5:69Þ
  @t
q ¼ rðPÞ  q0 þ q0 ðH; S; PÞ ; ð5:60Þ
where we have kept terms with all powers of dm, but have dropped
where q0 ¼ const, q0  q0 , and ð@ q0 =@PÞjH;S¼const  dr=dP, so most all terms with power of  higher than the first, and terms which in-
Rf
of pressure effect is absorbed into r(P). Since q  ¼ D1 h q dz ¼ volve quadratically-small product of  by ðq =q  1Þ.
1
Rf     1
Rf 1
Rf
D h
r q0 þ q dz ¼ q0  D h r dz þ D h r q dz it is natural to
0 0
The entire splitting algorithm is therefore formulated as fol-
express lows: bottom pressure pb is considered to be a prognostostic vari-
Z
  1 f able, and it evolves in slow time by adding increments g  dm,
q ¼ r  q ¼ r q0 þ q0 where r ¼ r dz and which are computed (fast-time averaged) by the barotropic mode.
D h
Z f Before starting fast-time stepping of barotropic mode from n to
1
q0 ¼  r q0 dz: ð5:61Þ n + 1 all necessary terms notably the state of free-surface f and full
rD h
vertically integrated PGF terms are precomputed using full 3D
Change in the total depth of water-column causes proportional algorithms with compressible EOS. Also computed at this stage
stretching of q profile (hence q is not affected by the change), and kept constant thereafter until the next baroclinic time step
while in contrast, r(P) profile moves up-and-down with free surface are the stiffened density ratios q =q needed by (5.69). After these
without any stretching at all (since it is function of pressure alone) preparations, the barotropic mode is advanced toward n + 1,
dm ¼ rbott qbott df  r bott q0 df; ð5:62Þ 0 1
0 1
  hdminþ1 nþ1 n nþ1
dm ¼ 0 B C hpb i ¼ hpb i þ ghdmi
where rbott and qbott are bottom values of r and q. On the other ! @ hðq  inþ1
 D þ dmÞu A!@  inþ1
A;
 in
 ¼ hu
u hu inþ1 ¼ hðq DþdmÞ u
hand, nþ1=2 hpb inþ1 =g
hhðqD þ dmÞuii
 
   
1 1 ð5:70Þ
dm ¼ 1   r q  df  1   r q0  df ð5:63Þ
2 2
where the first column symbolizes setting of initial conditions, the
which leads to an estimate second in time stepping of barotropic mode via (5.69) performing
 ¼ 2ð1  r=rbott Þ: ð5:64Þ fast-time averaging on the way, and the third column is translation
back to slow-time variables.21
In the simplest case of vertically uniform speed of sound r is a linear
function of pressure, r ¼ 1 þ =2, cf., (3.22). A more general (5.64) 21
To prevent the accumulation of roundoff errors it is useful to extract a constant-
accounts for the nonuniformity of c due to its pressure dependency, in-time bulk part from pb,
leaving aside only the temperature effect in the upper ocean. Z f¼0
ð0Þ ð0Þ ð0Þ
^b ;
p b ¼ pb þ p where pb ¼ pb ðx; yÞ ¼ g q0 r dz;
The ratio ðq =q
 Þ is affected by both compressibility and barocli- z¼h
nicity, so its behavior is controlled by on EOS. Similarly to (5.61), and use p ^b as the prognostic variable in instead of pb, meaning that the increments
Z Z hdmin+1 are added to p ^b in (5.70), not to pb. Then, whenever needed, hpbin+1 is com-
  2 f f
puted afterward by adding back the bulk part, hpb inþ1 ¼ pb þ hp
ð0Þ
^b inþ1 , which can be
q ¼ r q ¼ r q0 þ q0 ; where r ¼ r dz0 dz and
D2 h z used everywhere in the code except time differencing and time integration: these
Z f Z f must still use p ^b . It can be easily estimated that for typical oceanic conditions pð0Þ
2 ^b , while the contrast between pð0Þ
b
q0 ¼ r q0 dz0 dz; ð5:65Þ is 3  4 orders of magnitude larger than p b and
r  D2 h z increments ghdmi is even greater – so the roundoff errors cannot be ignored, even
with double precision. However, we emphasize that p ^b is merely for bookkeeping,
where r⁄ can be estimated as r⁄ = 1 + /3 in the case of uniform c, cf., and from the mathematical point of view, the splitting algorithm does not rely on
(3.22). Combining the above, the smallness of p ^b relatively to pð0Þ
b
.
 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70 61

 
The overall structure of (5.69) is similar to its Boussinesq coun- gh D c2
terpart (5.53) with an obvious one-to-one correspondence be-
¼ -0:025;  0:04;
c2 c2
tween all the terms, with the exception of dm2- and dm3-terms in f q0
r.h.s. of the first Eq. (5.69). As expected, (5.69) reverts back to shal-
low-water equations if ðq =q Þ ! 1 and  ? 0. What is essentially
new is that in the general case of q =q –1 and  – 0 linearization
of (5.69) around the rest state, f ¼ u  ¼ 0, captures the correct
phase speed for the barotropic waves, where both effects – slow-
down due to stratification (cf., (3.20), also Appendices C and D)
and seawater compressibility (cf., (3.21), also Eq. (B.9) from Appen-
dix B.1) – are now accurately represented within the fast mode
alone, and the phase speed does not depend on the choice of refer-
ence constants such as q0 – and it should not in non-Boussinesq
models. The derivation procedure for (5.69) also illustrates that
the contributions due to the two physical effects – baroclinicity
and compressibility – cannot be combined into a single effective
field: ðq =q Þ and  appear in different combinations in r.h.s. of
(5.69), so the baroclinic ratio ðq =q Þ must be computed from the
‘‘stiffened’’ density field separately from bulk compressibility effect
accounted by . This confirms the assumption made in Section 3.2
that the slowdown of barotropic wave speed due to vertical strat-
ification (3.20) can be correctly estimated by the barotropic mode
in a Boussinesq model as long as q  and q⁄ in (3.19) are ‘‘stiffened’’.
Conversely, using non-stiffened version of them does not yield the
correct estimate of influence of compressibility (3.21), not the
combination of both effects, resulting in overall incorrect barotrop-
ic phase speed.
The case with   0 while keeping q =q – 1 corresponds to
‘‘stiffened’’ non-Boussinesq model. In this case the modification
of pressure gradient terms relatively to uniform-density shallow-
water equations is controlled entirely by the ratio q =q . This is
comparable to splitting algorithm of Higdon and de Szoeke
(1997), Hallberg (1997), which is also non-Boussinesq, however
assuming that all variations of density are solely due to stratifica-
tion. In terms of complexity and computational cost (while main-
taining the same level of mode-splitting errors) this is very
similar to the Boussinesq version (5.53): dm and df can be easily re-
lated in fast time as dm ¼ q  df, where q does not change in fast
time. This, in principle, makes it possible to express everything di-
rectly in the original rather than incremental variables.
Clearly, most of the complexity of (5.69) relative to its Bous-
sinesq counterpart is associated with the bulk compressibility -
terms: because of EOS pressure dependency and, correspondingly,
the feedback of the state of free-surface f into density distribution,
f can be accurately computed from bottom pressure pb only by
using full 3D algorithm. Since f is needed to compute barotropic-
pressure gradient term, a simplified algorithm is needed to express
it using 2D variables only during the fast-time stepping. This
necessitates approximations which may admit splitting errors.
Using incremental variables offers relief: now f is computed using
full 3D algorithm, but only once per baroclinic time step (at the
beginning of fast-time stepping), while it is increment df which
is computed using a simplified 2D algorithm [essentially via
(5.58) leading to (5.63) and ultimately to (5.69)]. This avoids
mode-splitting errors, but brings extra cost associated with the
additional terms in (5.69) recomputed at each fast-time step and,
an extra 2D field –  – to be computed once and stored. It is essen-
tial that EOS compressibility can be split into the bulk and the
much smaller residual parts – a property of seawater EOS pointed
out by Dukowicz (2001).
5.7. Summary for Boussinesq and non-Boussinesq mode splitting
The preceding analysis shows that the existing non-Boussinesq
models essentially rely on the same assumptions as the Boussinesq
models, with the relevant smallness parameters estimated as
 2  104    0:2;  0:005    0:01; ð5:71Þ
h q0
where the extreme parameter values typically do not occur simul-
taneously which limits the algorithmic error estimates that contain
their products. We have assumed a range of values of c = 1480–
1540 since most of this variation is purely due to pressure effect,
and therefore can be ‘‘absorbed’’ into r if a nonlinear function of
pressure/depth is used. The assumed range of depths varies from
mid-ocean to the inner continental shelf. The estimate for q0 =q0
(cf., Fig. 1) is significantly smaller than typical values of q0 /q0 due
to stiffening.
The mode-splitting error of the Boussinesq ROMS is estimated
as Oððq0 =q0 Þ2 Þ with all other parameters (5.71) being irrelevant
(see Section 3.2). The error increases to the first-order Oðq0 =q0 Þ
if q
 and q⁄ in (3.18), (3.19) are replaced with q0 ; and an additional
Oðgh=c2 Þ error is introduced if a non-stiffened, realistic EOS is used
in an ad hoc way.
Non-Boussinesq models must use full dynamic pressure in EOS
which implies that density (specific volume) within each vertical
grid box (or layer in layered model) depends on the state of free
surface. At this time we are not aware of any reference where
one takes into account the finite compressibility of seawater into
the context of mode-splitting – the density field is usually com-
puted using the sate of free surface at the latest available baroclinic
step and kept constant during barotropic time stepping from that
baroclinic step to the next. This causes mode-splitting errors and,
in principle, opens a possibility for the numerical instability either
via a mechanism similar to one described in Higdon and Bennett
(1996), or in our Section 5.2. With some effort, splitting in a non-
Boussinesq model can also be made predominantly second-order
accurate in sense that the error estimate depends on pair-wise
products of parameters (5.71). This can be done in both implemen-
tations of non-Boussinesq model – the density extrapolation meth-
od of Greatbatch et al. (2001) (our Section 5.2, following Eq. (5.33))
and if using pressure-based coordinates (Section 5.6). The splitting
error is essentially avoided by including the f ? EOS pressure feed-
back into the fast-time stepping. This is facilitated by factoring EOS
in a manner used by Dukowicz (2001), even though it was not orig-
inally intended for non-Boussinesq modeling. A complete account-
ing of all f dependencies within the barotropic mode results in a
substantial increase in complexity, most likely beyond the point
of diminishing return, especially if keeping in mind that a model
with stiffened EOS already captures all important physical effects
(thermobaricity, and also the steric effect – the principal non-Bous-
sinesq effect), so re-introduction of bulk compressibility brings
only minor quantitative changes.
Existing oceanic codes like MOM4p1 use simpler mode-splitting
procedures resulting in stronger reliance on the smallness of
parameters (5.71), e.g., smallness of f relative to total depth and
even to the uppermost grid-box if there is strong stratification
within the upper portion of the domain. The later is due to vertical
redistribution of density by barotropic motions in the case of ver-
tically fixed grid. This can be traced back to the long-standing vi-
sion of free-surface pressure as ‘‘pressure on the rigid lid’’, where
the Poisson equation was considered primarily as a more efficient
replacement for the original stream-function method, and subse-
quently, the split-explicit free-surface was motivated more by
the ease of implementation in the era of parallel computing (or
lack of efficient parallel elliptic solvers at that time) rather than
by the interest to the physical phenomena associated with the free
surface itself. This view is inherited from the historical de-empha-
sis of temporal accuracy for barotropic motions (first-order) and
the practice of rather heavy-handed temporal filtering of them.
62 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70
In this case conversion to a non-Boussinesq model with new
sources of mode-splitting error does not usually lead to an instabil-
ity because sufficient numerical damping is already present. Since
fine-resolution regional modeling gives more interest in barotropic
processes (e.g., tides and topographic amplification of tidally-in-
duced motions), the accuracy can be restored by using a more
time-scale selective fast-time averaging filter, but this also brings
sensitivity to splitting errors. One should note that legacy MOM
code does not formally guarantee discrete finite-volume conserva-
tion property for tracers, even in their simpler Boussinesq versions
(Griffies et al., 2001; Griffies, 2004), and therefore have somewhat
less to lose when converting to non-Boussinesq in a simple way
(Greatbatch et al., 2001). This situation has been improved with
conversion to pressure-based coordinates (with relative ease for
z-coordinate model, somewhat more difficult for sigma), but this
also commits the model to use hydrostatic approximation and
there is no obvious way to overcome this limitation. When consid-
ering barotropic mode splitting, the vertical coordinate ‘‘integrates
out’’ exposing essentially the original dilemma – density (specific
volume) depends on the state of free surface via EOS pressure –
which must be dealt with during fast-time stepping or accept addi-
tional mode-splitting errors. For hydrostatic models it is still sim-
pler to make a Boussinesq code more self-consistent in its
discretized properties than a non-Boussinesq one. Conversely, a
more general, nonhydrostatic, non-Boussinesq code for flows with
low Mach number (e.g., Gatti-Bono and Colella, 2006) would be
more computationally expensive, hence less competitive for realis-
tic, large-scale oceanic simulations.
Another class of hydrostatic oceanic models uses an isopycnic
vertical coordinate. As in the case of pressure-based coordinates,
these models also avoid the density extrapolation (5.13) by the
specific design of their coordinate due to Lagrangian or predomi-
nantly Lagrangian movement in the vertical direction. The use of
the Montgomery potential in the horizontal pressure-gradient
force makes it natural to have a non-Boussinesq formulation. Orig-
inally these models were derived using the assumption of incom-
pressible EOS and Lagrangianly conserved (potential) density
(Bleck and Boudra, 1981; Bleck and Smith, 1990). When the ther-
mobaric effect was included later by Sun et al. (1999), it introduced
EOS stiffening for the first time. This also leads to the necessity of
using ‘‘thermobaric references’’ to avoid pressure gradient errors of
an essentially sigma-type (despite using isopycnic coordinates (!)),
resulting in a mechanism for numerical instability and a delicate
treatment (Hallberg, 2005). [This approach (along with this partic-
ular reason for stiffening) may become obsolete after a recent
alternative proposed by Adcroft et al. (2008).] In addition to the
above, the mode-slitting algorithm of Higdon and de Szoeke
(1997), Hallberg (1997), still used today, was derived assuming
incompressible EOS. For these reasons, and because of their prac-
tices these models can be classified as ‘‘stiffened’’ non-Boussinesq
models.
6. Summary
The desire for a physically correct representation of EOS effects
(e.g., thermobaricity) in oceanic modeling requires the use of the
nonlinear, realistic seawater EOS. It is often implanted in an
‘‘add-on’’ fashion into a Boussinesq-approximation code. This
brings a set of internal inconsistencies with spurious effects (e.g.,
vertical dependency of acceleration created by a purely barotropic
pressure gradient) and interference with barotropic–baroclinic
mode-splitting, including potential numerical instability. This pa-
per shows in Sections 2–4 how a combined approach of Dukowicz
(2001) and SM2003 leads to an accurate Boussinesq model with a
computationally-practical form of stiffened EOS in the specific con-
text of r-coordinate modeling. In essence, it comes from the real-
ization that the bulk of the seawater compressibility effect is
dynamically passive and can be separated in the EOS from the
dynamically relevant parts. Then the constant Boussinesq refer-
ence density q0 is much closer to the varying q = q(H, S, P) com-
pared to simply replacing q ? q0 in the standard Boussinesq
approximation. This significantly reduces the Boussinesq errors
while retaining all the necessary physical effects outlined in Sec-
tion 3. The output of the EOS routine is not the in situ density or
its anomaly, but q01 ¼ q01 ðH; SÞ and q01 ¼ q01 ðH; SÞ. The former one
essentially retains its original meaning derived from Jackett and
McDougall (1995)), while the latter is now defined by (4.6). The
functional form of EOS is modified to include the cz2 term in
(4.7), which causes a corresponding change in the algorithm for
adiabatic differencing (4.8) and a minor revision of the pressure-
gradient algorithm.
The limitations of the Boussinesq approximation are widely dis-
cussed in the literature, and sometimes its complete abandonment
is advocated. In Section 5 we show that doing so leads to a more
complicated code: the mode-splitting procedure interferes with
the density computation via EOS since EOS pressure depends on
the state of free-surface field, and consequently EOS can be no
longer considered as belonging entirely to the slow mode, resulting
in a more complicated splitting algorithm to avoid additional split-
ting errors and/or extra temporal filtering for the barotropic mode
to control numerical instability. This computational aspect of EOS
compressibility is usually overlooked in theoretical studies.
Remarkably, though originally intended for a Boussinesq code,
the factoring of EOS as q ¼ rðPÞ  qEOS ðH; S; PÞ by Dukowicz (2001)
provides a useful framework for designing accurate mode-splitting
in a hydrostatic, non-Boussinesq, free-surface model that retains
all the compressibility effects in EOS. Removing the bulk compress-
ibility from EOS, but keeping thermobaric part of it (i.e., stiffening)
eliminates the need for splitting EOS pressure into fast and slow
components, thus simplifying the algorithm, however departure
from the fully realistic EOS also means that the model can no long-
er be considered as a fully non-Boussinesq. Notably, among the
existing non-Boussinesq models, modern isopycnic-coordinate
models always use stiffened EOS (Sun et al., 1999; Hallberg,
2005). This allows retention of the principal non-Boussinesq ef-
fects, notably steric sea-level changes.
Post facto we note a hierarchy of four approximations, all of
which are in practical use in hydrostatic oceanic modeling today.
They differ only by the treatment of compressibility effects in
EOS: full non-Boussinesq ? non-Boussinesq with stiffened
EOS ? Boussinesq with stiffened EOS ? standard Boussinesq (q0
in combination with full EOS). We recommend that the last of
these be discontinued; i.e., if the Boussinesq approximation is cho-
sen, it should be applied uniformly to the entire code, including
EOS. This means stiffening.
Acknowledgments
This research is supported by the Office of Naval Research
through Grants N00014-05-10293 and N00014-08-10597.
Appendix A. Alternative variants of Boussinesq approximation
and potential vorticity (PV) equation in a shallow barotropic
layer of compressible fluid
McDougall et al. (2002,) proposed an alternative version of
Boussinesq approximation – their Eqs (27)–(29) – which have ex-
actly the same form as the original Boussinesq equations, with the
~ is not the usual velocity, but is the aver-
exception that velocity u
aged normalized mass flux per unit area,
 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70 63

~ ¼ quq =q0 ¼ qu=q0 ;

u where uq ¼ qu=q; ðA:1Þ pb @ t A þ pb u  r? A  A@ t pb  Au  r? pb ¼ 0 ðA:7Þ
see their Eq. (23), where overbar means Reynolds averaging and, for which ultimately leads to PV equation
simplicity, we have changed their notation u~ !u ~ . Their motivation
comes from subgrid-scale parameterization. They noticed that f þ r?  u
ð@ t þ u  r? Þ ¼ 0: ðA:8Þ
defining mean velocity as density-weighted average uq instead of pb
a more conventional Eulerian average u, leads to a more physically
So the principle dynamics of this compressible, but barotropic fluid
interpretable turbulent correlation terms. Following these
is (i) vertical uniformity of u (as a consequence, vertical uniformity
guidelines they derived a set of averaged non-Boussinesq equations
of relative vorticity), and (ii) Lagrangian conservation of barotropic
– their Eqs. (24)–(26) – to which they applied Boussinesq approxi-
PV as stated above. The above derivation can be performed for uni-
mation to derive their alternative Boussinesq. MGL2002 argue, and
form (on an f-plane) or non-uniform Coriolis parameter, and in hor-
provide some support for their claim, that their alternative Bous-
izontal orthogonal curvilinear coordinates as well. The fact that
sinesq equations are more accurate than the standard – notably,
fluid is compressible adds almost nothing – one can easily repeat
the stationary parts of continuity and tracer equations are equiva-
the above with constant-density layer (in this case it does not mat-
lent to their non-Boussinesq counterparts, the geostrophic-balance
ter whether it is Boussinesq or non-Boussinesq) the only difference
is their momentum equations is equivalent to non-Boussinesq (see
is replacement pb ? gq0(h + f).
their Section 5). They also use this claim to dismiss the EOS
In the case of conventional Boussinesq approximation with
stiffening approach of Dukowicz (2001) as unnecessary, because
pressure-dependent EOS the pressure-gradient term in the first
their alternative Boussinesq equations already eliminate the non-
Eq. (A.2) becomes (1/q0)r\p = gq/q0r\f, the gq = P0 (f  z),
physical shear in a geostrophically-balanced flow without any need
which is no longer vertically uniform. This precludes the transition
for adjustment in EOS.
from (A.2) to (A.3) and the subsequent derivation of PV equation
Unfortunately the alternative Boussinesq equations of MGL2002
similar to above. If one considers only the geostrophic balance in
also alter the relationship between the advection and Coriolis terms
Boussinesq version of (A.2), the resultant velocity becomes propor-
making it impossible to derive barotropic PV equation, which is
tional to (q/q0)-profile as well, which is an artifact of Boussinesq
possible to derive from non-Boussinesq or stiffened Boussinesq
approximation.
sets.
Excluding dynamic pressure from EOS equation of (A.2) by
Consider a layer of non-stratified, but slightly compressible
replacing it with q ¼ q0 þ q0EOS ðq0 gzÞ eliminates the vertical
fluid, hydrostatic, non-Boussinesq,
dependency,
@ t u þ u  r? u þ w@ z u þ f k  u ¼ ð1=qÞr? p; Z f
1 q0 þ q0 jz¼f g
@ t q þ r?  ðquÞ þ @ z ðqwÞ ¼ 0;  r? p ¼ g r? f  r? q0EOS ðq0 gzÞ dz0
ðA:2Þ q0 q0 q0 z |fflfflfflfflfflfflfflfflfflfflfflffl
ffl{zfflfflfflfflfflfflfflfflfflfflfflfflffl}
@ z p ¼ g q subject to pjz¼f ¼ 0; ¼0
! !
q ¼ qEOS ðpÞ EOS @ q0 gf2
 g 1  g EOS  f r? f  g r? f 2 ; ðA:9Þ
@p z¼f 2c
along with proper kinematic boundary conditions at free surface
z = f and bottom z = h, where h = h(x, y) is bottom topography. while introducing an insignificant non-physical term due to the
Above u, w = (u, v, w) is 3D velocity vector, r\ is horizontal (two- artificial assumption that q = q0 at z = 0 instead of free surface
dimensional) gradient or divergence operator. As discussed in Sec- z = f (above c is speed of sound, so the associated smallness param-
tion 2 from (2.1) to (2.4), under these circumstances EOS and hydro-
static equations can be solved resulting in self-consistent profiles
for density and pressure, q ¼ ð1=gÞP 0 ðf  zÞ and p ¼ Pðf  zÞ (where
P 0 means derivative of P with respect to its argument f  z), and, as
follows from (2.4), the acceleration due to pressure gradient is sim-
ply gr\f and is independent of z regardless of the functional form
of P ¼ Pðf  zÞ. This means that(A.2) admits solutions where hori-
zontal velocities are independent of z,
eter gf/c2 2  106 assuming c = 1500 m/s and f = 1 m). With this
choice made it is possible to derive an analog of (A.8) while using
Boussinesq approximation. Note that the extra term in (A.9) does
not preclude elimination of pressure gradient term when taking
curl of (A.3).
EOS stiffening by Dukowicz (2001) (in this case trivially revert-
ing EOS in (A.2) to constant density, q = q0, making it insensitive to
the choice of which pressure is used in EOS), also eliminates the
vertical dependency in (1/q0)r\p, allowing derivation of baro-
@ t u þ u  r? u þ f k  u ¼ g r? f; tropic PV Eq. (A.8).
ðA:3Þ
@ t pb þ r?  ðpb uÞ ¼ 0; If the alternative Boussinesq system is used – the hydrostatic
Rf version of Eqs. (27)–(29) from MGL2002 in combination with
where pb ¼ g h q dz ¼ Pðf þ hÞ is identified as bottom pressure. It EOS using in situ pressure, q = qEOS(p) as in (A.2) – then it is not
is assumed to be invertible, h þ f ¼ ½P1 ðpb Þ. The pressure gradient the ordinary velocity, but the density-scaled velocity u ~ becomes
term r\f can be eliminated by taking horizontal curl of u-equation proportional to (q/q0), which means that unscaled velocity is ver-
(A.3), tically uniform as it should. However, despite the correct geo-
@ t ½r?  u þ u  r? ½r?  u þ f r?  u þ ½r?  ur?  u ¼ 0; strophic balance, the modified Boussinesq set of MGL2002 also
does not produce a counterpart of (A.8). Inspection of the deriva-
ðA:4Þ
tion above shows that there must be proper scaling relationships
or among all three terms: Coriolis, pressure-gradient, and advection.
In the original Boussinesq the scaling between Coriolis and advec-
@ t A þ u  r? A þ Ar?  u ¼ 0; ðA:5Þ
tion is correct, but pressure gradient receives non-physical verti-
where A is absolute vorticity, A ¼ f þ r?  u. On the other hand,
pb-equation (A.3) can be dressed up as
@ t pb þ u  r? pb þ pb r?  u ¼ 0:
Multiplying the A-equation by pb, the latest pb-equation by A, and
subtracting them to cancel the r\  u-terms,
cally-dependent multiplier. In the modified case, rescaling of
velocity by q/q0 corrects the relationship between Coriolis and
pressure gradient, but at the expense of sacrificing the relationship
ðA:6Þ
between Coriolis and advection – it is ‘‘unscaled’’, rather than den-
sity-weighted (hence vertically-dependent) vorticity should be
combined with f in order to form absolute vorticity, and then PV.
64 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70

Conversely, the alternative non-Boussinesq set–Eqs. (24)–(26) which admits wave solutions with phase speed
from MGL2002–contains the reverse density multiplier q0/q in sffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffi
the nonlinear terms, which makes it possible to derive the PV q pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi % gh; gh=c2  1
~c ¼ gh ¼ c  1  egh=c2 ðB:8Þ
equation above merely because the alternative set can be trans- qb & c; gh=c2
1:
formed back into the original non-Boussinesq written in terms of
the non-scaled u. (Since ghq ¼ P b can be identified as bottom pressure, this result
In contrast, stiffening of EOS by Dukowicz (2001) repairs the coincides with Eq. (29) in Dukowicz (2006) for a non-stratified case,
relationship between pressure gradient and Coriolis terms without i.e., by there setting N2 = 0.) For typical oceanic conditions of
disturbing the already correct mutual relationship between Corio- h = 5500 m, speed of sound c = 1500 m/s, and acceleration of gravity
lis and advection. g = 9.81 m/s2, we estimate  = gh/c2 = 0.025  1. This means that
the inclusion of the compressibility effect leads to a slightly smaller
Appendix B. Surface gravity waves in a compressible barotropic ( 0.6%) phase speed than that of a layer of the same thickness filled
shallow-water layer by an incompressible fluid,
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffi 1  e pffiffiffiffiffiffi 

B.1. Non-Boussinesq case ~c ¼ gh  ¼ gh 1  þ    ;   1: ðB:9Þ
 4
Eq. (2.1)–(2.4) from Section 2 show that the horizontal acceler- The physical explanation for the reduced phase speed comes from
ation generated by a free-surface gradient does not depend on the the fact change in q  D caused by divergence of vertically-integrated
vertical coordinate when vertical variation of density is due only to fluxes results in a lesser, by a factor of  1  /2, change in f than in
compressibility and there is hydrostatic balance, the case of incompressible fluid, while the remaining /2 fraction of
@ t u þ    ¼ g r? f: ðB:1Þ q D translates into change in density q , which does not contribute to
acceleration due to pressure gradient in (B.7).
This makes it possible to consider motions where u is vertically uni- Eq. (B.8) indicates that the gravity-wave phase speed cannot be
form. In this case (B.1) integrates vertically in a trivial manner, and larger than the speed of sound. The opposite asymptotic limit, gh/
together with vertically integrated mass-conservation equation it c2
1, has meaning as well, ~c ! c. This corresponds to a lowest-
yields, vertical-mode wave of compression in an isotropic atmosphere,
@ t ðq  Þ þ    ¼ g q
 Du  Dr? f which does not have a well-defined free surface; rather its density
@ t ðq
 DÞ þ r?  ðq 
 DuÞ ¼ 0; ðB:2Þ decreases with height, so that the e-folding scale c2/g plays the role
of its effective thickness. The hydrostatic approximation alone ex-
where cludes acoustic waves without making an assumption of incom-
Z f Z pressibility or small density variation; in fact, acoustic and
1 1 f
¼
u q u dz ¼ u dz ¼ u; gravity waves become indistinguishable in the hydrostatic case.
q D h D h
Z f
1
q ¼ q dz; and D ¼ h þ f: ðB:3Þ B.2. Boussinesq case with q = qEOS(P)
D h
Note that the difference between its density-weighted and volume As follows from (2.5), the Boussinesq counterpart of (B.1) with
averaged u disappears because of vertical uniformity of u. Solution pressure-dependent EOS q = q1 + P/c2 is
of (B.2) implies that the prognostic variable q D (to be interpreted 2
here as a whole symbol having the meaning of total weight of water @ t u þ    ¼ ðg=q0 Þ  q1 egðfzÞ=c  r? f; ðB:10Þ
column) needs to be translated into f in order to close the system. where u can no longer be vertically uniform. However it is still pos-
This is done through EOS. For simplicity we assume that EOS has sible to derive an analog of (B.7) by substituting u ¼ u~  egðfzÞ=c2 ,
the form of (2.7), viz., q = q1 + P/c2 with q1 and c constant. Com- hence
bined with hydrostatic balance, this leads to the vertical profile
2   2
q = q1 exp{g(f  z)/c2}, which results in egðfzÞ=c  @ t u ~  g=c2  egðfzÞ=c  @ t f þ   
~ þu
2 2 2 2
egðfþhÞ=c  1 egD=c  1 egD=c  1 ¼ ðg=q0 Þ  q1 egðfzÞ=c  r? f; ðB:11Þ
q ¼ q1  ¼ q1  ; hence q
 D ¼ q1 :
gðf þ hÞ=c2 gD=c2 g=c2
or
ðB:4Þ  
@t u ~ g=c2 @ t f þ    ¼ g ðq1 =q0 Þ  r? f;
~ þu ðB:12Þ
This depends on f through D.
Now we consider small motions in a compressible fluid layer ~ independent of z as long as the non-
which admits solutions with u
over a flat bottom. Substituting expression for q D from (B.4) into linear terms (denoted as dots) are vanishingly small. Equation for
the second equation (B.2) with h = const yields free surface becomes
2
Z !
2 egD=c  1 f
egD=c  1
2
q1 egD=c  f@ t f þ u  r? fg þ q1  ¼ 0;
r?  u ðB:5Þ @ t f ¼ r?  ~e
u gðfz0 Þ=c2 0 ~
dz ¼ r?  u ðB:13Þ
g=c2 g=c2
h
or
  r? f þ ðq
@t f þ u  =qb ÞDr?  u
 ¼ 0; ðB:6Þ Linearization of (B.12) and (B.13) assuming h = const, yields
2 ~ ¼ gðq1 =q0 Þ  r? f and @ t f þ h½ðe  1Þ=r?  u
@t u ~; ðB:14Þ
where we identify qb ¼ q1 egD=c as the value of the density near the
bottom. To derive (B.6) we have used the EOS for the purpose of 2
where, once again,  = gh/c . The resultant phase speed of barotropic
specific volume, i.e., to express f from mass content q
 D. This situa- waves is
tion is common for non-Boussinesq models, and it reverses the role qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
of EOS in a Boussinesq-approximation model. Linearization of (B.1) ~c ¼ ghðq1 =q0 Þ  ðe  1Þ= ðB:15Þ
and (B.6) yields the system,
where it is worth
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi to note, cf., (B.4), that q1 ðe  1Þ= ¼ q
 , hence
 ¼ g r? f and @ t f ¼ ðq
@t u  =qb Þhr?  u
; ðB:7Þ ~c ¼ gh  q  =q0 . The choice of Boussinesq reference density q0 ¼ q 
 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70 65

pffiffiffiffiffiffi
makes ~c ¼ gh, however it should be noted that in the case of non- ^ 1 ¼ ik  g^f; ix  ^f þ ix  g
 ix  u ^ ¼ ik  H1 u
^1 ;
uniform topography this choice cannot be made universal because  
q q
^ 2 ¼ ik  g  1  ^f  ik  g 1  1  g
q depends on h. The effect of slowdown of barotropic waves by  ix  u ^ ; ix  g ^2 :
^ ¼ ik  H2 u
q2 q2
compressibility is lost.
ðC:2Þ
B.3. Boussinesq with reference pressure in EOS ^ 1 and u
The right pair of equations can be used to exclude u ^ 2 from
the left pair:
Everything is the same as in the previous case, except that EOS
pressure is computed assuming f  0, and is not allowed to change ð~c2  gH1 Þ^f ¼ ~c2 g
^;
 
with f. Then q = q1 exp{  gz/c2}, so the counterpart of (B.1) q  q q ðC:3Þ
~c2  gH2 2 1 ^
g ¼ gH2 1 ^f;
becomes q2 q2
Z f
1 0 2 where we have introduced ~c ¼ x=k which is identified as phase
@t u þ    ¼  r? g q1 egz =c dz0
q0 z speed for the vertical modes, external or internal.22 This has non-
Z   trivial solutions if
g g f
2
q r? f 
0
¼ r? q1 egz =c dz0 ðB:16Þ  
q0 q0 z |fflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflffl} q  q1 q
z¼f
ð~c2  gH1 Þ  ~c2  gH2 2 ¼ ~c2 gH2 1 : ðC:4Þ
¼0 q2 q2
or @ t u þ    ¼  gqq1 e r? f   gqq01 r? f which permits solutions
gf=c2
It can be rewritten as
0
with vertically uniform u, and we note that gf/c2 5  106
assuming c = 1500 m/s. The analog of the linearized system (B.7) ~c4  ~c2 ~c20 þ ~c20 ^c21 ¼ 0; ðC:5Þ
becomes where
@ t u ¼ g ðq1 =q0 Þ  r? f and @ t f ¼ hr?  u: ðB:17Þ H1 H2 q2  q1
~c20 ¼ g ðH1 þ H2 Þ and ^c21 ¼ g  : ðC:6Þ
which yields phase speed H1 þ H 2 q2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Eq. (C.5) has two solutions,
~c ¼ gh  q1 =q0 : ðB:18Þ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
~c20 ~c40 ~c2 ~c2 qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðe ~c2 ¼  ~c20 ^c21 ¼ 0 0 1  4^c21 =~c20
For the choice of q0 ¼ q ffiffiffiffiffiffi   q1 ð1 þ =2Þ, where  = gh/
 ¼ q1
p1Þ=

2 4 2 2
c2  1, the above turns into ~c  gh  ð1  =4Þ, which is similar to
~c20 ~c20
(B.9). This result is merely a coincidence: the Boussinesq system   ^c1 ; as long as ^c21  ~c20 ;
2
ðC:7Þ
(B.17) does not provide a proper physical mechanism for phase 2 2
speed reduction due to compressibility effects (note that in compar- where ~c2þ ¼ ~c20  ^c21 is associated with the barotropic mode, and
ison with (B.7) the density multiplier appears in front of pressure ~c2 ¼ ^c21 with the baroclinic. This indicates that the presence of strat-
gradient term, rather than in free-surface equation). This result also ification reduces the square of phase speed of the barotropic mode
suggests that using the near-surface value of density to compute relative to that of a constant-density layer of the same thickness,
pressure gradient term in barotropic mode is an optimal choice H1 + H2, by the square of the baroclinic phase speed. This conclusion
for a simple mode-splitting algorithm (simple means that it does is true even beyond the assumption of a small density difference,
not take into account the influence of stratification onto barotropic q2  q1  q2: (C.5) has the property that the sum of its two roots
phase speed) in a Boussinesq model with reference EOS pressure cf., is always ~c20 .
Griffies (2004, see Sections 12.3–12.4 there), Griffies (2009, see Sec- Structure of the barotropic mode: ^f and g ^ are moving in
tion 7.7.3, Eq. (7.129)). proportion,

g^ gH1 H2  HH11þH
H2
2
 q2qq1
Appendix C. Surface gravity waves in a stratified two-layer fluid ¼1 ¼ 2
q2 q1
^f c2 H1 þ H2  HH11þHH2
2
 q2
"  #
C.1. Incompressible case H2 2
H1 q1
 1 1  ; ðC:8Þ
H 1 þ H2 ðH1 þ H2 Þ2 q2
The simplest system with both barotropic and baroclinic modes
is a two-layer model. It is also the simplest way to illustrate the which is slightly smaller than the fraction of the distance from the
influence of stratification on the phase speed of surface gravity bottom to the interface between the two layers relative to the total
waves. For small oscillations around the resting state, the two- depth. Correspondingly,
layer model has a set linear equations:  
^2
u ^c2 H1 q
¼1 1 1 1 1 ; ðC:9Þ
@ t u1 ¼ g  rx f; @ t ðf  gÞ þ H1  rx u1 ¼ 0; ^1
u gH2 H1 þ H 2 q2
 
q1 q indicating that the bottom layer moves slower than the top.
@ t u2 ¼ g  rx f  g 1  1  rx g; @ t g þ H2  rx u2 ¼ 0;
q2 q2 Direct splitting: Eqs. (C.8) and (C.9) indicate that neither the vol-
ðC:1Þ ume-nor density-weighted vertically averaged velocity,

¼
H1 u1 þ H2 u2
¼
q1 H1 u1 þ q2 H2 u2
where uk, qk, and Hk are velocities, densities, and unperturbed thick- u or u ; ðC:10Þ
H1 þ H2 q1 H1 þ q2 H2
nesses of the top (k = 1) and the bottom (k = 2) layers; f is free-
surface elevation; g is interface elevation (both are relative to their
unperturbed states, z = 0 and z = H1). This system assumes hydro-
22
static balance for an incompressible fluid, but it does not use the Because later in this appendix we will consider the case of compressible fluid
with a finite speed of sound, to avoid potential conflict in notation, we denote the
Boussinesq approximation. phase speed of gravity waves as ~c or ^c, using ~c for both modes, if they appear together,
^ 1 ; ^f; u
Fourier analysis: Let ðu1 ; f; u2 ; gÞ ¼ ðu ^ Þ  ekxixt , then (C.1)
^2 ; g or for barotropic, if separately, while ^c is used for baroclinic only if it appears
becomes separately. Plain c (without any symbol on top) is reserved for speed of sound.
66 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70

8
can be identified as a depth-independent barotropic flow. However, < q egðfzÞ=c21 ;
1
the proportionality property of perturbations, g^ vs. ^f and u
^ 2 vs. u
^1 , q¼ 
: q þ P =c2 egðgH1 zÞ=c22 ;
provides a path for splitting (C.1) into independent barotropic and 2 1 2
baroclinic subsystems. To do this multiply the upper-left Eq. (C.1) 8 h 2
i ðC:19Þ
>
< q1 c21 egðfzÞ=c1  1 top;
by H1 and the lower-left by bH2 and combine the layers; similarly,
multiply the lower-right by b and combine it with the upper-
P¼   h i
>
: P1 þ P1 þ q c2 egðgH1 zÞ=c22  1 bottom;
right: 2 2

   2
q q  q1 where P1 ¼ q1 c21 ½egðfgþH1 Þ=c1  1 is pressure at the interface be-
@ t ðH1 u1 þ bH2 u2 Þ ¼ rx g H1 þ bH2 1 f þ bgH2 2 g ;
q2 q1 tween the layers, z = H1 + g. Eq. (C.19) yields continuity of pres-
@ t ½f þ ðb  1Þg ¼ rx ðH1 u1 þ bH2 u2 Þ; sure P across the interface. Similar to (2.4), the acceleration due to
PGF (1/q)rxP is independent of the vertical coordinate within
ðC:11Þ each layer and equal to
where b is yet to be determined. Note that u1 and u2 already appear 1
in identical combinations under @ t and rx in the first and the sec-  rx P ¼ g rx f and
q
ond equations, and the goal is to achieve the same for f and g. This 2
leads to the proportionality condition, 1 q1 g ðrx f  rhx gÞegðfgþH1 Þ=c1 i
 rx P ¼     g rx g ðC:20Þ
  q 2
q2 þ q1 c21 =c22 egðfgþH1 Þ=c1  1
g H1 þ bH2 qq1 bgH2 q2qq1
2 1
¼ ; ðC:12Þ
1 b1 for the top and bottom layer, respectively. This leads to the momen-
tum equations,
and the  selection
 of b as a root of the quadratic equation
b2  qq2 1  HH12 b  qq2  HH12 ¼ 0 or, @ t u1 ¼ g rx f;
1
8
1
9
   
" 2  #1=2 = q q gH q c2
1 q2 < H1 H1 4H1 q1 @ t u2 ¼ g rx f  g 1  1  1  21  1  1 12
q2 q2 c1 q2 c2
 ðrx g  rx fÞ;
b¼  1 1þ   1 : ðC:13Þ
2 q1 : H2 H2 H2 q2 ;
ðC:21Þ

Assuming that 1  q1/q2  1, this leads to where we have expanded (C.20) in a Taylor series for gH1 =c21  1
  and retained only the leading-order terms. The equations are linear-
H2 q2
bþ ¼ 1 þ 1 and ized assuming f, g  H1, H2 and neglecting the advection terms. As
H1 þ H2 q1

  expected, the second Eq. (C.21) reverts to (C.1) if c1 ? 1, while the
H1 H2 q2 ratio c1/c2 remains finite. Furthermore, the additional terms due to
b ¼   1 1 ; ðC:14Þ
H2 H 1 þ H 2 q1 compressibility vanish if gH1 =c21 remains finite, but q1c1 = q2c2, indi-
which corresponds to the barotropic and baroclinic modes respec- cating their baroclinic and thermobaric nature.
tively. Once b is known, (C.11) can be turned into two independent The top-layer thickness equation is derived following the path
systems (choosing either + or  sign index below), of transition from (B.2) to (B.6) in Appendix B. This leads to
 
8 1 gH1
@ t ðf  gÞ ¼ H1 1  rx u1 ; ðC:22Þ
@ t U ¼ ~c2 rx f ; < f ¼ f þ ðb  1Þg;
> 2
2 c1
where U ¼ H1 u1 þ b H2 u2 ðq1 =q2 Þ;
@ t f ¼ g rx U ; >
: ~2 which is linearized, and only leading-order corrections due to finite
c ¼ g½H1 þ b H2 ðq1 =q2 Þ:
compressibility are retained. The bottom-layer thickness equation
ðC:15Þ stems from mass conservation,
It can be shown that the phase speeds ~c2 for each mode are the @ t ðq
 2 ðH2 þ gÞÞ þ rx ðq
 2 ðH2 þ gÞu2 Þ ¼ 0; ðC:23Þ
same as (C.7) for b = b±, respectively. One can also verify that (C.8)
or
and (C.9) can be obtained from the conditions,
ð@ t þ u2  rx Þðq
 2 ðH2 þ gÞÞ þ q
 2 ðH2 þ gÞrx u2 ¼ 0; ðC:24Þ
f þ ðb  1Þg ¼ 0 and H1 u1 þ b  H2 u2 ðq1 =q2 Þ ¼ 0; ðC:16Þ
where, as follows from (C.19), the vertically averaged density with-
that have the meaning of non-excitation of baroclinic degrees of
in the bottom layer is
freedom by purely barotropic motions. Since
  2
1 < bþ < q2 =q1 ; ðC:17Þ P1 egðH2 þgÞ=c2  1
q 2 ¼ q2 þ : ðC:25Þ
c22 g=c22
the weighting of u1 and u2 for the barotropic mode is within the
bounds of set by (C.10). It leans toward q2/q1 when the top layer Since the interface pressure P1 depends on the thickness of the top
is shallow relative to the bottom layer, H1  H2. layer, q
 2 depends on it as well. This leads to the participation of f in
the bottom thickness equation. Substitution of the expression for P1
C.2. Compressible two-layer non-Boussinesq case just after (C.19) into (C.25), with subsequent substitution into (C.24)
leads to
Consider a two-layer fluid with densities in the upper and lower  
q1 gH2 1 gH2
layers dependent on pressure, @t g   @ ðf  gÞ þ H2 1  rx u2 ¼ 0; ðC:26Þ
q2 c22 t 2 c22
q ¼ q1  P=c21 and q ¼ q2  P=c22 ; ðC:18Þ
where we retain only leading-order compressibility corrections. The
with q1 – q2 due to stratification and speeds of sound c1 – c2 to above reverts to the corresponding equation in (C.1) in the incom-
simulate the dependencies on temperature, salinity, and pressure. pressible limit, gH2/c2 ? 0.
Using the hydrostatic balance the above leads to self-consistent Eqs. (C.21), (C.22), and (C.26) comprise a closed system that de-
profiles of density and pressure in the top, H1 + g < z < f, and bot- scribes small oscillations in a compressible two-layer fluid. Its Fou-
tom, H1  H2 < z < H1 + g, layers: rier analysis similar to the transition from (C.1) to (C.2) yields
 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70 67

^ 1 ¼ ik  g^f;
 ix  u ix  ^f þ ix  g^ ¼ ik  H01 u
^1 ; So the counterparts of (C.12) and the left side (C.14) become
 
 ix  u 0 ^ 0
^ 2 ¼ ikg ð1  d Þ  f  ikgd  g
^; ðC:27Þ q2 H 1 H1
b2   b ¼ 0; bþ
^  ix  ^f ¼ ik  H0 u
 ixð1  Þ  g ^2 ; q1 H 2 H2
2  
q2 H 2 q2
¼1þ 1 : ðC:33Þ
where we define q1 H1 þ q2 H2 q1
     
q q gH q c2 1 gH1 This indicates, once again, that the velocity weighting parameter b+
d0 ¼ 1  1  1  21 1  1 12 ; H01 ¼ H1 1  ; corresponding to the barotropic mode is neither 1 (volume weight-
q2 q2 c1 q2 c 2 2 c21
  ing) nor q2/q1 (density weighting).
q1 gH2 0 1 gH2
 ¼  2 ; H 2 ¼ H2 1  ;
q2 c2 2 c22
Appendix D. Surface gravity waves in a stratified shallow-water
ðC:28Þ
layer
and expect that d0 ,  are sufficiently small and H01 ; H02 deviate only
Consider a Boussinesq fluid layer, where, for simplicity, we re-
slightly from H1, H2. This leads to the characteristic equation,
  strict our attention to a two-dimensional case in an x–z plane with-
~c4  g H02 þ H01 ð1  Þ ~c2 þ g 2 d0 H01 H02 ¼ 0; ðC:29Þ out Coriolis force:
where ~c ¼ x=k. This essentially repeats (C.5) and (C.6), but with a @ t u þ    ¼ ð1=q0 Þ@ x P;
modified set of coefficients. The resulting phase speeds for the baro- @ t q þ u@ x q þ w@ z q ¼ 0;
tropic and first baroclinic modes are ðD:1Þ
@ z P ¼ g q;
  gd0 H01 H02 @ x u þ @ z w ¼ 0:
~c2þ ¼ g H02 þ H01 ð1  Þ  ^c2
H02 þ H01 ð1  Þ
We rewrite this with a r-coordinate that follows f:
gd0 H01 H02
¼ 0 ; ðC:30Þ z  f % 0; z ! f;
H2 þ H01 ð1  Þ r¼
h þ f & 1; z ! h;
where, similar to (C.7), we keep only leading-order corrections aris- % f; r ! 0;
ing from stratification and here compressibility. conversely; z ¼ f þ rðh þ fÞ ðD:2Þ
When gH1 =c21 ! 0 and gH2 =c22 ! 0, both phase speeds revert to
& h; r ! 1:
their incompressible limits in (C.7). On the other hand, for a com- We assume bottom topography is flat, h = const, so that the only dis-
pressible fluid without stratification, q1 = q2, and the speed of tinction between @ x—z and @ x—r is due to f – 0. Thus,
sound is the same in both layers (c1 = c2); hence, the barotropic
@P @P @z @P @P @f @P
phase speed from c+ from (C.30) becomes ¼   ¼  ð1 þ rÞ  : ðD:3Þ

 @x z @x r @x r @z @x r @x @z
1 gðH1 þ H2 Þ Rf R0
~c2þ ¼ gðH1 þ H2 Þ 1   þ    ; ðC:31Þ Since P ¼ g q dz ¼ gðh þ fÞ q dr0 , we can further transform this
2 c2 z r
into
where the dots denote higher-order terms with respect to  Z 0 Z 0
g(H1 + H2)/c2  1. This coincides with (B.9) where h = H1 + H2. @P @f @ q
¼ g qð1 þ rÞ þ q dr0  þ gðh þ fÞ 0
dr : ðD:4Þ
In the general case with both stratification and compressibility @x z r @x r @x r
effects, as follows from (C.30), the stratification still causes the The density equation in r-coordinates is
reduction of phase speed of the barotropic mode relative to the
non-stratified case in the same manner as in (C.7), i.e., the square @ t jr q þ u  @ x jr q þ ðW=ðh þ fÞÞ  @ r q ¼ 0; ðD:5Þ
of barotropic phase speed is reduced by the square of the first baro- where @ t—rq is the time tendency relative to the moving r-surface.
clinic mode speed. In addition, both are reduced by the compress- The vertical velocity in r-coordinates W is related to its Eulerian
ibility (finite speeds of sound), due to the replacement of layer counterpart as W ¼ w  ð1 þ rÞ  ½@f @f
þ u  @x , which automatically
@t
thicknesses H1, H2 with their primed counterparts H01 ; H02 . The baro- satisfies the no-flux boundary condition Wjr¼0 ¼ 0 at the free sur-
clinic mode speed is also affected by the second term in d0 , which face as a consequence of the Eulerian kinematic surface boundary
involves the ratio of q1 c21 =q2 c22 . In principle, this admits the exis- condition, w = @ tf + u@ xf. The non-divergent continuity equation is
tence of internal waves even if q1 = q2, but the speed of sound in
the upper layer is greater than in the lower to yield a positive @ t f þ @ x jr ððh þ fÞuÞ þ @ r W ¼ 0; ðD:6Þ
d0 > 0. For typical oceanographic conditions we expect the ratio of which leads to equations for f,
phase speeds for the first baroclinic and the barotropic modes to Z 0
be within the range of 20–100, which leads to an estimate of  Þ ¼ 0;
@ t f þ @ x ððh þ fÞu ¼
where u u dr; ðD:7Þ
^c2 =~c2þ 103 . This causes a substantially smaller reduction of the 1
barotropic wave speed than the influence of bulk compressibility
and for W
(B.9) and makes stratification less of a concern as the source of
mode-splitting error.  ÞÞ þ @ r W ¼ 0:
@ x jr ððh þ fÞðu  u ðD:8Þ

C.3. Boussinesq case The first equation in (D.1) with its PGF expressed by (D.4), along
with (D.7), and (D.8), comprise a closed system which describes
The Boussinesq version of (C.1) is evolution of a layer of stratified fluid with a free surface. Since we
are interested in small oscillations around the mean resting state,
@ t u1 ¼ g ðq1 =q0 Þrx f; @ t ðf  gÞ þ H1  rx u1 ¼ 0; f  u  W  0 and q = q(r), and we linearize the system using
q  q1 the variables,
@ t u2 ¼ g ðq1 =q0 Þrx f  g 2 rx g; @ t g þ H2  rx u2 ¼ 0:
q0
ðC:32Þ f ¼ f0 ;  þ u0 ;
u¼u W ¼ W0; q ¼ qðrÞ þ q0 ; ðD:9Þ
68 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70

where all except q(r) are considered as small perturbations. Note (D.18) is that for q(r) = const, the only possible non-trivial solution
that splitting of horizontal velocity into u  and u0 , such that is ^f – 0, as allowed by x2 = k2gh  q/q0, and W c  0. This corre-
R0 0
1
u d r  0, is a matter of convenience, rather than an assump- sponds to familiar barotropic gravity waves. It also makes q0 = q
tion that the second (primed) term is smaller than the first. The sys- the natural choice for the Boussinesq pffiffiffiffiffiffi reference density, resulting
tem becomes in a correct phase speed x=k ¼ gh. Conversely, assuming ^f  0
 Z 0 but dq(r)/dr < 0 results in a Sturm–Liouville problem for W c alone,
@ g @f0
 þ u0 Þ ¼   ð1 þ rÞqðrÞ þ
ðu qðr0 Þdr0  which leads to a set of baroclinic modes.
@t q0 r @x In the general case a dilemma is that the expression inside
Z 0 0
gh @q {.[. . .]} in l.h.s. of (D.18) depends on r, while ^f does not. To resolve
 dr0 ; ðD:10Þ
q @x
0 r r this, as in the case of (D.15), vertical integration of (D.18) yields
@f0 
@u @ q0 W 0 dqðrÞ  Z Z
þh ¼ 0; þ  ¼ 0; x2 q ^ k gh 0 0 dqðr0 Þ
ik   gh f¼  c
W dr0 dr; ðD:20Þ
@t @x @t h dr 2 q0 x q0 dr
k 1 r
@u0 1 @W 0
þ  ¼ 0; ðD:11Þ
@x r h @ r where p the expression
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi inside {. . .} no longer depends on s. Setting
x=k ¼ gh  q =q0 allows for a non-trivial ^f, provided that W c ¼ 0,
 , ob-
where the first equation can be split into separate ones for u however, substitution of this x/k back into (D.18) results there in
tained by vertical integration of (D.10), non-vanishing expression {.[. . .]} and ultimately leads to a nontriv-
Z  Z 0 ial profile of W c¼W c ðrÞ, which is proportional, and in fact, is driven

@u g @f0 0
¼   ð1 þ rÞqðrÞ þ qðr0 Þdr0 dr by ^f.
@t q0 @x 1 r
Z Z This dilemma has two consequences. First, it suggests an itera-
gh 0 0
@ q0 0 tive method for finding x/k for the barotropic mode: use

q0 dr dr; ðD:12Þ
1 r @x r x2 = k2  gh  q⁄/q0 as the initial guess and substitute it into
and for u0 , the residual of (D.10) after subtraction of (D.12). With gh dqðrÞ c k gh
x @2 W dqðrÞ
ik  r  ^f ¼   þ  c
W ; ðD:21Þ
Z 0 q0 dr k @ r2 x q0 dr
0 0
PðrÞ ¼ qðr Þdr ; ðD:13Þ
r which is the vertical derivative of (D.18). Then the l.h.s. of (D.21)
can be treated as known. This leads to a problem for W c alone. This
integration of the first term inside the left integral in the r.h.s. of
(D.12) yields problem is well posed, because this value of x/k is far away from
Z Z   the eigenvalues of the operator in the r.h.s. of (D.21) (i.e., its eigen-
0 0
dP values are associated with internal modes and have much lower fre-
ð1 þ rÞqðrÞdr ¼ ð1 þ rÞ  dr
1 1 dr quencies). The solution results in a profile, W c¼W c ðrÞ, proportional
Z 0 ^
to f. Substitute it into the r.h.s. of (D.20) and recompute x/k, after
¼  ð1 þ rÞPðrÞj01 þ PðrÞdr: ðD:14Þ which the iterative cycle is repeated. Convergence is guaranteed be-
|fflfflfflfflfflfflfflfflfflfflfflffl{zfflfflfflfflfflfflfflfflfflfflfflffl} 1
¼0 cause (D.21) can be rewritten as

Then the entire (D.12) an be rewritten as c

@2 W
ix2 s  ^f ¼ c;
þ 2 W ðD:22Þ
Z Z @ r2

@u g q @f0 gh 0 0
@ q0
¼   0
dr dr; ðD:15Þ 2 2 2 dqðrÞ
@t q0 @x q0 1 r @x r where 2 ¼ xk 2  h N2 ¼ xk 2  gh  ð q1  dr Þ is of the same order as the
0
square of the ratio of the first-baroclinic and barotropic mode phase
where we have identified 2
speeds, hence   1. The scaling of W c relative to ^f is estimated as
Z 0 Z 0 c ix2^f, and the scaling of the mismatch between this l.h.s. and
W
q ¼ 2 qðr0 Þdr0 dr; ðD:16Þ r.h.s. of (D.20) is estimated as
1 r
 2
which coincides with (3.19) from Section 3.2. x q x
ik 2  gh  vs: ix   4 ; ðD:23Þ
Let k q0 k
 
 ; u0 ; W 0 ; q0 Þ ¼ ^f; u
^; u c; q
^  eikxixt ; which implies that the corrections are indeed sufficiently small.
ðf0 ; u ^0 ; W ðD:17Þ
The second consequence is that, even though the system is lin-
c; q
^; W
where the last three, u ^ are functions of r. Substitution into ear, it did not split into two independent subsystems: barotropic
(D.10) leads to and baroclinic. This does not mean that the it is not splittable in

Z  principle, but rather that the actual structure of the barotropic
0
x2 gh ^f mode cannot be represented by a vertically uniform velocity: it
ik   ð1 þ rÞqðrÞ þ qðs0 Þdr0
k
2 q0 r has a non-constant profile that depends on stratification q(r). Sim-
Z c slaved to ^f indicates than
ilarly, the presence of a non-trivial W
c k gh
x @W 0 dqðr0 Þ
¼    c
W dr0 ; ðD:18Þ vertical density profile is not ‘‘frozen’’ in r-coordinates in purely
k @ r x q0 r dr
barotropic motions, but it changes following f. This indicates that
c using a Fou-
where we have eliminated all variables except ^f and W the ideal stretching of the r-coordinate by f should not just be pro-
rier-transform of (D.11), viz., portional to the distance from the bottom,

x ^f 1 Wc dqðrÞ 1 @Wc z ¼ zð0Þ þ f  ðzð0Þ þ hÞ=h; ðD:24Þ

^ ¼
u  ; q^ 0 ¼   ; ^0 ¼ 
and u  : ðD:19Þ
k h ix h dr ikh @ r but should depend on the density profile as well to cancel W asso-
Eq. (D.18) comprises a Sturm–Liouville problem with unknowns ciated with barotropic motions, and thus cancel changes of density.
x (playing the role of an eigenvalue), ^f (an arbitrary non-trivial This determines the level of error in the mode-splitting algorithm of
Fourier amplitude), and W c¼W c ðrÞ (a vertical profile multiplied SM2005 because the mismatch in (D.20) indicates that the phase
by an arbitrary non-zero constant). The profile is subject to a pair speed of the barotropic mode is different between thep2D and 3Dffi
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
cj
of boundary conditions, W c parts of the code. The 2D part sees it as just x=k ¼ gh  q =q0 ,
r¼0 ¼ 0 and W jr¼1 ¼ 0. A property of
 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70
and the 3D part sees it as the complete solution of (D.20) and (D.21).
This is a mode-splitting error, because there are contributions in the
vertically-integrated r.h.s. of the 3D momentum equations that de-
pend on f and have a hyperbolic character, yet are frozen during the
barotropic time-stepping, hence are computed using effectively a
forward-in-time Euler algorithm.
References
Adcroft, A., Campin, J.-M., 2004. Rescaled height coordinates for accurate
representation of free-surface flows in ocean circulation models. Ocean
Modeling 7, 269–284. doi:10.1016/j.ocemod.2003.09.003.
Adcroft, A., Hallberg, R., Harrison, M., 2008. A finite volume discretization of the
pressure gradient force using analytic integration. Ocean Modeling 22, 106–
113. doi:10.1016/j.ocemod.2008.02.001.
Antonov, J.I., Locarnini, R.A., Boyer, T.P., Mishonov, A.V., Garcia, H.E. 2006. World
Ocean Atlas 2005. Salinity. Levitus, S. (Ed.), vol. 2. NOAA Atlas NESDIS 62, US
Government Printing Office, Washington, DC, 182 pp.
Berntsen, H., Kowalik, Z., Sælid, S., Sørli, K., 1981. Efficient numerical simulation of
ocean hydrodynamics by splitting procedure. Modeling, Identification and
Control 2, 181–199.
Bleck, R., Boudra, D.B., 1981. Initial testing of a numerical ocean circulation model
using a hybrid (quasi-isopycnic) vertical coordinate. Journal of Physical
Oceanography 11, 755–770.
Bleck, R., Smith, L.T., 1990. A wind-driven isopycnic coordinate model of the north
and equatorial Atlantic Ocean: 1. Model development and supporting
experiments. Journal of Geophysical Research 95C, 3273–3285.
Blumberg, A.F., Mellor, G.L., 1987. A description of a three-dimensional coastal
ocean circulation model. In: Heaps, N.S. (Ed.), Three-Dimensional Coastal Ocean
Models. Pub. AGU, Washington, DC, pp. 1–16.
Boussinesq, J., 1903. Théorie analytique de la chaleur. Tome II. In: Analytic theory of
heat. Vol. 2, Gauthier-Villars, Paris, 625pp. <http://gallica.bnf.fr/ark:/12148/
bpt6k61635r>.
Brydon, D., Sun, S., Bleck, R., 1999. A new approximation of the equation of state for
seawater, suitable for numerical ocean models. Journal of Geophysical Research
104, 1537–1540.
Campin, J.-M., Adcroft, A., Hill, C., Marshall, J., 2004. Conservation of properties in a
free-surface model. Ocean Modeling 6, 221–244. doi:10.1016/S1463-
5003(03)00009-X.
Casulli, V., Cheng, R.T., 1992. Semi-implicit finite-difference methods for 3-
dimensional shallow-water flow. International Journal for Numerical Methods
in Fluids 15, 629–648.
de Szoeke, R.A., Samelson, R.M., 2002. The duality between the Boussinesq and non-
Boussinesq hydrostatic equations of motion. Journal of Physical Oceanography
32, 2194–2203. doi:10.1175/1520-0485(2002)032<2194:TDBTBA>2.0.CO;2.
Dewar, W.K., Hsueh, Y., McDougall, T.J., Yuan, D.I., 1998. Calculation of pressure in
ocean simulations. Journal of Physical Oceanography 28, 577–588.
Dukowicz, J.K., 1997. Steric sea level in the Los Alamos POP code – Non-Boussinesq
effects. In: Lin, C., Laprise, R., Ritchie, H. (Eds.), Numerical Methods in
Atmospheric and Oceanic Modeling: The Andre Robert Memorial Volume.
NRC Research Press, pp. 533–546.
Dukowicz, J.K., 2001. Reduction of density and pressure gradient errors in ocean
simulations. Journal of Physical Oceanography 31, 1915–1921.
Dukowicz, J.K., 2006. Structure of the barotropic mode in layered ocean models.
Ocean Modeling 11, 49–68. doi:10.1016/j.ocemod.2004.11.005.
Gatti-Bono, C., Colella, P., 2006. An anelastic allspeed projection method for
gravitationally stratified flows. Journal of Computational Physics 216, 589–615.
doi:10.1016/j.jcp.2005.12.017.
Greatbatch, R.J., 1994. A note on the representation of steric sea level in models that
conserve volume rather than mass. Journal of Geophysical Research 99 (12),
76712,77.
Greatbatch, R.J., McDougall, T.J., 2003. The non-Boussinesq temporal residual mean.
Journal of Physical Oceanography 33, 1231–1239.
Greatbatch, R.J., Lu, Y., Cai, Y., 2001. Relaxing the Boussinesq approximation in
ocean circulation models. Journal of Atmospheric and Oceanic Technology 18,
1911–1923. doi:10.1175/1520-0426(2001)018<1911:RTBAIO>2.0.CO;2.
Griffies, S.M., 2004. Fundamentals of Ocean Climate Models, vol. 496. Princeton
University Press. 496 pp. ISBN-13#09780691118925.
Griffies, S.M., 2009. Elements of MOM4p1. GFDL Ocean Group Technical Report No.
6. NOAA/GFDL, April 21, 2009, 444 pp.
Griffies, S.M., Adcroft, A.J., 2008. Formulating the equations of ocean models. In:
Hecht, M.W., Hasumi, H. (Eds.), Ocean Modeling in an Eddying Regime,
Geophysical Monograph 177. American Geophysical Union, Washington, DC,
pp. 281–318.
Griffies, S.M., Gnanadesikan, A., Pacanowski, R.C., Larichev, V.D., Dukowicz, J.K.,
Smith, R.D., 1998. Isoneutral diffusion in a z-coordinate ocean model. Journal of
Physical Oceanography 28, 805–830.
Griffies, S.M., Pacanowski, R.C., Schmidt, M., Balaji, V., 2001. Tracer conservation
with an explicit free surface method for z-coordinate ocean models. Monthly
Weather Review 5, 1081–1098.
Hallberg, 1997. Stable split time stepping schemes for large-scale ocean modeling.
Journal of Computational Physics 135, 54–65.
69
Hallberg, 2005. A thermobaric instability of Lagrangian vertical coordinate ocean
models. Ocean Modeling 8, 279–300. doi:10.1016/j.ocemod.2004.01.001.
Higdon, R.L., Bennett, A.F., 1996. Stability analysis of operator splitting for large-
scale ocean modelling. Journal of Computational Physics 123, 311–329.
Higdon, R.L., de Szoeke, R.A., 1997. Barotropic–baroclinic time splitting for ocean
circullation modeling. Journal of Computational Physics 135, 31–53.
Huang, R.X., Jin, H.Z., 2002. Sea surface elevation and bottom pressure anomalies
due to thermohaline forcing. Part I: Isolated perturbations. Journal of Physical
Oceanography 32, 2131–2150.
Ingersoll, A.P., 2005. Boussinesq and anelastic approximations revisited: Potential
energy release during thermobaric instability. Journal of Physical Oceanography
35, 1359–1369. doi:10.1175/JPO2756.1.
Jackett, D.R., McDougall, T.J., 1995. Minimal adjustment of hydrostatic profiles to
achieve static stability. Journal of Atmospheric and Oceanic Technology 12,
381–389.
Jackett, D.R., McDougall, T.J., 1997. A neutral density variable for the world’s oceans.
Journal of Physical Oceanography 27, 237–263.
Jackett, D.R., McDougall, T.J., Feistel, R., Wright, D.G., Griffies, S.M., 2006. Algorithms
for density, potential temperature, conservative temperature, and the freezing
temperature of seawater. Journal of Atmospheric and Oceanic Technology 23,
1709–1728.
Kanarska, Y., Shchepetkin, A.F., McWilliams, J.C., 2007. Algorithm for non-
hydrostatic dynamics in ROMS. Ocean Modeling 18, 143–174. doi:10.1016/
j.ocemod.2007.04.001.
Killworth, P.D., Stainforth, D., Webb, D.J., Paterson, S.M., 1991. The development of a
free-surface Bryan–Cox–Semtner ocean model. Journal of Physical
Oceanography 21, 1333–1348.
Locarnini, R.A., Mishonov, A.V., Antonov, J.I., Boyer, T.P., Garcia, H.E., 2006. World
Ocean Atlas 2005. Temperature. Levitus, S. (Ed.), vol. 1, NOAA Atlas NESDIS 61,
US Government Printing Office, Washington, DC, 182 pp.
Losch, M., Adcroft, A., Campin, J.-M., 2004. How sensitive are coarse general
circulation models to fundamental approximations in the equations of motion?
Journal of Physical Oceanography 34, 306–319. doi:10.1175/1520-
0485(2004)034<0306:HSACGC>2.0.CO;2.
Lu, Y., 2001. Including non-Boussinesq effect in Boussinesq ocean circulation
models. Journal of Physical Oceanography 31, 1616–1622.
Marshall, J., Adcroft, A., Campin, J.-M., Hill, C., White, A., 2004. Atmosphere-ocean
modeling exploiting fluid isomorphisms. Monthly Weather Review 132, 2882–
2894. doi:10.1175/MWR2835.1.
McDougall, T.J., Garrett, C.J.R., 1992. Scalar conservation equations in a turbulent
ocean. Deep-Sea Research 39, 1953–1966.
McDougall, T.J., Jackett, D.R., 2007. The thinness of the ocean in S  H  p space and
the implications for mean diapycnal advection. Journal of Physical
Oceanography 37, 1714–1732. doi:10.1175/JPO3114.1.
McDougall, T.J., Greatbatch, R.J., Lu, Y., 2002. On conservation equations in
oceanography: How accurate are Boussinesq ocean models? Journal of
Physical Oceanography 32, 1574–1584. doi:10.1175/1520-
0485(2002)032<1574:OCEIOH>2.0.CO;2.
McDougall, T.J., Jackett, D.J., Wright, D.G., Feistel, R., 2003. Accurate and
computationally efficient algorithms for potential temperature and density
and of seawater. Journal of Atmospheric and Oceanic Technology 20, 730–
741.
McPhee, M.G., 2000. Marginal thermobaric stability in the ice-covered upper ocean
over Maud Rise. Journal of Physical Oceanography 30, 2710–2722. doi:10.1175/
1520-0485(2000)030<2710:MTSITI>2.0.CO;2.
Mellor, G.L., 1991. An equation of state for numerical modeling of oceans and
estuaries. Journal of Atmospheric and Oceanic Technology 1991, 609–611.
Mellor, G.L., Ezer, 1995. Sea level variations induced by heating and cooling: An
evaluation of the Boussinesq approximation in ocean models. Journal of
Geophysical Research 100, 20557–20565.
Mihaljan, J.M., 1962. A Rigorous exposition of the Boussinesq approximations
applicable to a thin layer of fluid. Astrophysics Journal 136, 1126–1133.
doi:10.1086/147463.
Oberbeck, A., 1879. Ueber die Wärmeleitung der Fls̈sigkeiten bei Berc̈ksichtigung
der Strm̈ungen infolge von Temperaturdifferenzen. Annalen der Physik 243,
271–292. doi:10.1002/andp.18792430606.
Oberbeck, A., 1888. On the phenomena of motion in the atmosphere (first comm).
In: The Mechanics of the Earth’s Atmosphere, Translated by Cleveland Abbe.
publ. Smithsonian Inst., Washington, 1891. pp. 176–187. <http://
books.google.com/books?id=kGBDAAAAIAAJ>.
Paluszkiewicz, T., Garwood, R.W., Denbo, D.W., 1994. Deep convective plumes in the
ocean. Oceanography 7, 3744.
Robertson, R., Padman, L., Levine, M.D., 2001. A correction to the baroclinic pressure
gradient term in the Princeton ocean model. Journal of Atmospheric and
Oceanic Technology 18, 1068–1075.
Rueda, F.J., Sanmiguel-Rojas, E., Hodges, B.R., 2007. Baroclinic stability for a family
of two-level, semi-implicit numerical methods for the 3D shallow water
equations. International Journal of Numerical Methods in Fluids 54, 237–268.
doi:10.1002/fld.1391.
Shchepetkin, A.F., McWilliams, J.C., 2003. A method for computing horizontal
pressure-gradient force in an oceanic model with a non-aligned vertical
coordinate. Journal of Geophysical Research 108, 3090–3124. doi:10.1029/
2001JC001047.
Shchepetkin, A.F., McWilliams, J.C., 2005. The regional ocean modeling system: a
split-explicit, free-surface, topography-following-coordinate oceanic model.
Ocean Modeling 9, 347–404. doi:10.1016/j.ocemod.2004.08.002.
70 A.F. Shchepetkin, J.C. McWilliams / Ocean Modelling 38 (2011) 41–70
Shchepetkin, A.F., McWilliams, J.C., 2008. Computational kernel algorithms for fine-
scale, multi-process, long-term oceanic simulations. In: Ciarlet, P.G. (Ed.),
Temam, R., Tribbia, J., (Guest Ed.), Handbook of Numerical Analysis.
Computational Methods for the Ocean and the Atmosphere. vol. XIV, Elsevier
Science, pp. 119–182. doi:10.1016/S1570-8659(08)01202-0.
Spiegel, E.A., Veronis, G., 1960. On the Boussinesq approximation for a compressible
fluid. Astrophysics Journal 131, 442–447.
Stacey, M.W., Pond, S., Nowak, Z.P., 1995. A numerical model of the circulation in
knight inlet, British Columbia, Canada. Journal of Physical Oceanography 25,
1037–1062.
Sun, S., Bleck, R., Rooth, C., Dukowicz, J., Chassignet, E., Killworth, P., 1999. Inclusion
of thermobaricity in isopycnis-coordinate ocean models. Journal of Physical
Oceanography 29, 2719–2729. doi:10.1175/1520-0485(1999)029<2719:
IOTIIC>2.0.CO;2.
Tailleux, R., 2009. On the energetics of stratified turbulent mixing, irreversible
thermodynamics, Boussinesq models and the ocean heat engine controversy.
Journal of Fluid Mechanics 638, 339–382. doi:10.1017/S002211200999111X.
Tailleux, R., 2010. Identifying and quantifying nonconservative energy production/
destruction terms in hydrostatic Boussinesq primitive equation models. Ocean
Modeling 34, 125–136. doi:10.1016/j.ocemod.2010.05.003.
Wright, D.G., 1997. An equation of state for use in ocean models: Eckart’s formula
revisited. Journal of Physical Oceanography 14, 735–741.
Young, W.R., 2010. Dynamic enthalpy, conservative temperature, and the seawater
Boussinesq approximation. Journal of Physical Oceanography 40, 394–400.
doi:10.1175/2009JPO4294.1.
Zeytounian, R.Kh., 2003. Joseph Boussinesq and his approximation: a contemporal
view. Comptes Rendus Mecanique 331, 575–586. doi:10.1016/S1631-
0721(03)00120-7.