momentum balance and eliassen–palm flux on moist
TRANSCRIPT
Momentum Balance and Eliassen–Palm Flux on Moist Isentropic Surfaces
RAY YAMADA
Courant Institute of Mathematical Sciences, New York University, New York, New York
OLIVIER PAULUIS
Courant Institute of Mathematical Sciences, New York University, New York, New York, and Center for
Prototype Climate Modeling, New York University Abu Dhabi, Abu Dhabi, United Arab Emirates
(Manuscript received 4 August 2015, in final form 29 October 2015)
ABSTRACT
Previous formulations for the zonally averagedmomentumbudget andEliassen–Palm (EP) flux diagnostics
do not adequately account for moist dynamics, since air parcels are not differentiated by their moisture
content when averages are taken. The difficulty in formulating the momentum budget in moist coordinates
lies in the fact that they are generally not invertible with height. Here, a conditional-averaging approach is
used to derive a weak formulation of the momentum budget and EP flux in terms of a general vertical
coordinate that is not assumed to be invertible. The generalized equation reduces to the typical mass-
weighted zonal-mean momentum equation for invertible vertical coordinates.
The weak formulation is applied here to study the momentum budget on moist isentropes. Recent studies
have shown that the meridional mass transport in the midlatitudes is twice as strong onmoist isentropes as on
dry isentropes. It is shown here that this implies a similar increase in the EP flux between the dry and moist
frameworks. Physically, the increase in momentum exchange is tied to an enhancement of the form drag
associated with the horizontal structure of midlatitude eddies, where the poleward flow of moist air is located
in regions of strong eastward pressure gradient.
1. Introduction
High-energy air parcels rise in the tropics and flow
poleward along the upper branch of the meridional
overturning circulation. From the conservation of axial
angular momentum, these air parcels would eventually
attain unrealistically high zonal velocities in the mid-
latitudes if it were not for the downward momentum
transport induced by the form drag associated with
baroclinic eddies. The transformed Eulerian-mean
(TEM) equations, derived by Andrews and McIntyre
(1976), successfully capture these interactions between
eddies and the mean flow. In this framework, the mean
meridional circulation is represented by the residual
circulation, which consists of a single thermally direct
cell in each hemisphere. It approximates the Lagrangian-
mean circulation and is predominantly eddy driven in the
midlatitudes. The Coriolis torque on the residual circu-
lation acts to accelerate the zonal wind in the upper
troposphere. In steady state, it is balanced by the
Eliassen–Palm (EP) flux divergence, which represents
the form drag due to the eddies (e.g., Andrews et al.
1987; Vallis 2006).
The EP flux has become a standard diagnostic tool for
studying wave–mean flow interactions (Edmon et al.
1980). Andrews and McIntyre (1976) generalize the
original EP relation (Eliassen and Palm 1961), showing
that the EP flux divergence vanishes when the waves are
assumed to be steady, conservative, and of small am-
plitude. Under such conditions, it follows from the TEM
equations that the waves cannot induce changes in the
mean flow—a statement of the nonacceleration theorem
(Charney and Drazin 1961)—and thus a divergent EP
flux provides a measure of how wave transience and
dissipation affect the zonal-mean circulation. The non-
acceleration result can be arrived at more readily if
quasigeostrophic scaling is imposed (Edmon et al. 1980).
In this case, the total eddy forcing of the mean state is
accounted for by the EP flux divergence, which acts as a
Corresponding author address: Ray Yamada, Courant Institute
of Mathematical Sciences, New York University, 251 Mercer St.,
New York, NY 10012.
E-mail: [email protected]
MARCH 2016 YAMADA AND PAULU I S 1293
DOI: 10.1175/JAS-D-15-0229.1
� 2016 American Meteorological Society
zonal forcing. The generalized Lagrangian-mean for-
mulation in Andrews and McIntyre (1978a,b) general-
izes the nonacceleration theorem to finite-amplitude
waves, but is difficult to use in practice as it involves
particle displacements. More recently, Nakamura and
Zhu (2010) extended the nonacceleration result to finite
amplitude in pressure coordinates by using a hybrid
Eulerian–Lagrangian approach.
Following the works of Andrews andMcIntyre (1976)
and Edmon et al. (1980), subsequent studies utilized
isentropic averaging (i.e., averaging on surfaces of con-
stant potential temperature u) to make further advances
to the TEMdescription of the large-scale circulation. On
the one hand, isentropic averaging was pursued to gen-
eralize the EP flux diagnostic to finite-amplitude waves
and nongeostrophic flows in an Eulerian framework.
Andrews (1983) derived a finite-amplitude nondiver-
gent EP flux from the isentropic primitive equations,
assuming nonacceleration conditions a priori. Tung
(1986) relaxed the assumption of nonacceleration con-
ditions and used mass-weighted averages to derive an
isentropic EP flux that includes diabatic effects. Iwasaki
(1989) formulated a similar finite-amplitude nongeo-
strophic EP flux using isentropic zonal-mean pressure as
the vertical coordinate instead.
Meanwhile, isentropic averaging was applied to study
the mean meridional circulation (e.g., Dutton 1976;
Townsend and Johnson 1985; Johnson 1989). In isen-
tropic coordinates, poleward fluxes of warm air are
separated from equatorward fluxes of cool air, and so a
single-celled circulation emerges naturally when the
mass flux is averaged on isentropes. The isentropic cir-
culation is qualitatively similar to the TEM residual
circulation, and the two can be shown to be equivalent to
leading order in wave amplitude (McIntosh and
McDougall 1996; Pauluis et al. 2011). Moreover, the is-
entropic circulation offers several advantages over the
TEM circulation: (i) no small wave amplitude or qua-
sigeostrophic scaling assumptions are made; (ii) the
streamlines of isentropic circulation close at the lower
boundary, while those of the TEM circulation do not
(Held and Schneider 1999; Tanaka et al. 2004; Lalibertéet al. 2013); and (iii) the vertical motion in the isentropic
streamfunction directly reflects diabatic heating.
The isentropic EP flux and circulation are dynamically
tied through zonal momentum balance, as is the case
with the TEM EP flux and residual circulation. The is-
entropic EP flux divergence is related to the eddy flux of
Ertel potential vorticity (PV; Tung 1986). Since eddies
tend to mix PV downgradient (equatorward) along is-
entropic layers that do not intersect the surface (e.g.,
Yang et al. 1990), it can be shown (Held and Schneider
1999) that eddies drive a thermally direct isentropic
circulation. In the steady-state momentum equation, the
EP flux divergence balances the Coriolis force on the
isentropic circulation. The isentropic formalism gives
the EP flux a clear physical relationship to form drag,
which is expressed in terms of correlations between is-
entropic pressure anomalies and the isentropic slope
(Andrews 1983; Juckes et al. 1994).
The aforementioned studies outline some of the sig-
nificant theoretical advances in our understanding of the
momentum cycle, wave–mean flow interactions, and the
general circulation. These momentum budget analyses
have been limited, however, to dry dynamics with a
single exception (Stone and Salustri 1984). Air parcels
with the same value of potential temperature but dif-
fering moisture contents are not distinguished when
averages are taken. Recently, Pauluis et al. (2008,
hereafter PCK08) and Pauluis et al. (2010, hereafter
PCK10) compared the dry and moist isentropic
streamfunctions, where the latter is computed by aver-
aging the meridional mass transport on surfaces of
constant equivalent potential temperature ue rather than
potential temperature.1 Like the dry isentropic circula-
tion, the moist isentropic streamfunction consists of a
single direct cell in each hemisphere, but in the mid-
latitudes, the total mass transport of the moist isentropic
circulation is about twice as large as that of the dry is-
entropic circulation. The additional mass flux on moist
isentropes arises from a low-level flow of warm,moist air
that is advected poleward by bothmidlatitude baroclinic
eddies and subtropical stationary eddies (Pauluis et al.
2011; Shaw and Pauluis 2012). Air parcels within this
moist branch of the circulation rise within the storm
tracks, precipitate, and enter the upper branch of the
circulation, where they cool radiatively and eventually
subside into the low-level return flow of the circulation.
The dry isentropic diagnostic fails to clearly differentiate
the low-level poleward moist branch from the equator-
ward return flow, since the two branches consist of air
masses with similar values of potential temperature. This
leads to partial cancellation in the mass transport when
averaged on potential temperature surfaces. In contrast,
the moist branch has a similar value of equivalent po-
tential temperature as the upper branch of the circulation,
which allows for its clear separation from the low-level
return flow when using the moist isentropic diagnostic.
The additional eddy mass flux associated with the
moist branch has significant implications for the zonal
momentum cycle. The enhanced mass flux on moist
1 Czaja and Marshall (2006) had computed a similar version of
the dry and moist isentropic streamfunction, in which they used
surfaces of constant dry and moist static energy instead.
1294 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 73
isentropes implies a stronger Coriolis torque on the
zonal flow than that previously indicated by dry diag-
nostics. For momentum balance, the form drag and EP
flux on moist isentropes must increase as well. However,
there are currently no available formulations of the
momentum budget on moist isentropes. A major chal-
lenge with moist isentropes is that, unlike potential tem-
perature, equivalent potential temperature is not invertible
with height, especially at low latitudes (Xu and Emanuel
1989). This prevents the primitive equations from being
transformed onto moist isentropes via a simple change of
variables, and so zonal averages at constant equivalent
potential temperature cannot be taken in the usual way.
Stone and Salustri (1984) modify the quasigeostrophic
TEM residual circulation and EP flux to account for the
eddy transport of both sensible and latent heat, but they
assume amoist static stability that is vertically stratified and
latitudinally independent. Their formulation is also con-
strained by the same assumptions that go into the TEM
equations, and so the accuracy of their method near the
surface is not reliable. Their results do, however, show that
when large-scale condensation is accounted for, there is a
dramatic increase in the EP flux and residual circulation in
the subtropics and midlatitudes, as compared to the origi-
nal TEM description.
The goal of this work is twofold. First, we generalize
the isentropic EP flux diagnostics (Andrews 1983; Tung
1986; Iwasaki 1989) to allow for a general vertical co-
ordinate h that may be noninvertible. The generalized
framework is discussed in section 2 and involves condi-
tionally averaging the primitive equations based on the
value of h in the zonal–height plane. It is consistent with
the conventional mass-weighted zonal-mean equations
that are obtained when the vertical coordinate is in-
vertible. Second, we apply this new method in section 3
to evaluate the momentum budget on dry and moist
isentropes using the Modern-Era Retrospective Analy-
sis for Research and Applications (MERRA) reanalysis
from 1979 to 2012. The moist isentropic EP flux is shown
to have amuch largermagnitude and greater subtropical
extent than the dry isentropic EP flux. These differences
are explained by analyzing the zonal structure of synoptic
eddies in section 4. In section 5, we discuss the decompo-
sition of the EP flux into orographic and atmospheric form
drags, which leads to a more familiar expression for the EP
flux. Conclusions are given in section 6.
2. Weak formulation of the zonally averagedprimitive equations in generalized verticalcoordinates
In this section, we formulate the zonally averaged
mass-weighted primitive equations on constant-h surfaces,
where h is a general vertical coordinate that is not as-
sumed to be invertible with height. We derive a gener-
alized EP flux and EP theorem and give a clear physical
interpretation of the EP flux and its relationship to form
drag. These results are analogous to those derived in
Andrews (1983) and Tung (1986) in u coordinates but
have been generalized to hold on arbitrary h surfaces.
Deriving these results requires the use of distributions,
since traditional zonal averaging cannot be applied
along noninvertible h surfaces. For clarity, we begin
with the primitive equations in Cartesian coordinates (x,
y, z), but one could also derive the results from
spherical–pressure coordinates (appendix B):
›r
›t1= � (rv)5 0, (1a)
›u
›t1 v � =u52
1
r
›p
›x1 f y1D , (1b)
›p
›z52rg , (1c)
›h
›t1 v � =h5 _h, and (1d)
›C
›t1 v � =C5 _C . (1e)
The above equations are the continuity equation, zonal
momentum equation, hydrostatic relationship, and
general evolution equations for h and a tracer C, re-
spectively. The meridional momentum and temperature
equations are not explicitly written out, but they can be
considered as special cases of the tracer equation. The
velocity vector is denoted by v5 (u, y,w) andD denotes
frictional drag. The notation is otherwise standard.
The domain is assumed to be a periodic, zonal channel
[0,L]3 [y1, y2]3 [zsfc(x, y), ‘) that is bounded between
two fixed latitudes y1 and y2. The surface height is de-
noted by zsfc. The velocity normal to the surface and
along the latitudinal boundaries is taken to be zero. We
will use the symbol D to refer to the domain [0, L] 3[zsfc, ‘) within the x–z plane at fixed latitude. If X is a
subset of the domain, then ›(X) will denote the
boundary of the set X.
a. Weak formulation of the mass-weighted zonalaverage
When h is invertible with height, the primitive equa-
tions can be transformed from (x, y, z) to (x, y, h) co-
ordinates. Zonal averages (denoted by a bar) can be
taken in the usual way at constant value of h:
f ( y,h, t)51
L
ðL0
f (x, y,h, t) dx . (2)
MARCH 2016 YAMADA AND PAULU I S 1295
When h is not invertible with height, this expression for
the zonal mean breaks down since f becomes multi-
valued at a given value of h. To work around this, we
use a conditional averaging approach similar to that
used in Pauluis and Mrowiec (2013) for studying the
upward mass transport in moist convection and also to
that in PCK10 for computing the meridional mass
transport on dry and moist isentropes.2 Here, we con-
sider the distribution of f over h at each latitude y and
time t, where the distribution is computed, in principle,
by integrating f in the x–z plane over those regions in
which h takes on a designated value h0:
hf i(y,h0, t)
51
L
ðL0
ð‘zsfc(x,y)
f (x, y, z, t)d[h02h(x, y, z, t)] dz dx .
(3)
The Dirac delta function d(x) is given by the derivative
of the Heaviside functionH(x) so (3) can be rewritten as
h f i5 limDh/ 0
1
2LDh
ððjh2h0j,Dh
f (x, y, z, t) dz dx . (4)
Hence, h f i involves integrating f over the x–z plane in
the regions in which h is within Dh of h0 in the limit as
Dh goes to zero. In practice, a small but finite Dh is used
to approximate h f i. We will refer to h f i as the delta
distribution of f on h, or simply the delta distribution of
f, when h is understood from context. The units of h f iare the units of f multiplied by meters per units of h.
We also define the following Heaviside distribu-
tion h�iH:
hf iH(y,h
0, t)
51
L
ðL0
ð‘zsfc(x,y)
f (x, y, z, t)H[h02h(x, y, z, t)] dz dx .
(5)
The delta distribution is given by the derivative of the
Heaviside distribution with respect to h0: hf i5 ›h0hf iH .
Although the delta distribution (3) defines a gener-
alized average on noninvertible h surfaces, hfi is not
generally equal to f even for invertible h. Instead, it can
be shown that for invertible h,
hf (x, y, z, t)i5 f (x, y,h, t)
����›z›h���� . (6)
The derivation is given in appendixA. The factor j›z/›hjis the Jacobian of transformation that relates the volume
element in (x, y, z) space to that in (x, y, h) space (i.e.,
dxdydz 5 j›z/›hjdxdydh). Hence, when f is chosen to
be a quantity per unit volume, then the delta distribution
of f is equal to the zonalmean of f on constant h surfaces.
In particular, mass-weighted zonal averages of f are
preserved by the delta distribution:
hrf (x, y, z, t)i5 rf (x, y,h, t)
����›z›h����5 r
hf (x, y,h, t) , (7)
where rh 5 rj›z/›hj is the density in (x, y, h) coordinates.
For (6) to hold on h surfaces that intersect the ground, we
must assume that f 5 0 below ground. This falls out natu-
rally for mass-weighted averages (7) if we adopt the com-
mon convention of setting the density to zero below ground
(Lorenz 1955). With this convention, mass-weighted aver-
ages contain no contribution from the portions along the
isentropes that are below ground.While such underground
conventions are necessarywhen taking zonalmeans in (x, y,
h) coordinates, the formulation for the delta distribution
(3) only involves integrating above the surface so that un-
derground conventions are not needed.
For invertible h, the conventional mass-weighted
zonal mean of f will be denoted by
~f (y,h, t)5rhf
rh
. (8)
For general, possibly noninvertible, h we define a weak
formulation of the mass-weighted zonal average of f on
h surfaces as follows:
f (y,h, t)5hrf ihri . (9)
Let f*(x, y, z, t)5 f (x, y, z, t)2 f [y, h(x, y, z, t), t] de-
note the departure of f from f . From (7), it follows im-
mediately that for invertible h the two definitions (8)
and (9) are consistent; that is, f 5 ~f . The weak form
of the mass-weighted average allows for the typical
manipulations of averages (e.g., bf*5 0,bf 5 f , andbfg5 f g1 df*g*), while the delta distribution h�i by itself
does not.
b. Weak form of the zonally averaged primitiveequations
The weak form of the zonally averaged primitive
equations can be derived by first rewriting (1) in flux
2 The meridional mass transport on a moist isentrope can be
computed as Lhryidue, which is a special case of the delta distri-
bution (3) with f 5 ry and h 5 ue. The moist isentropic stream-
function is given byC(y, ue0, t)5LhryiH, which is equivalent to theexpression given in PCK08 and PCK10 before time averaging.
1296 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 73
form (so that mass-weighted averages can be taken) and
then taking the delta distribution (3) of the equations.
The weak forms of the zonally averaged continuity,
zonal momentum, hydrostatic balance, and tracer
equation are given as follows:
›
›thri1 ›
›y(hriy)1 ›
›h0
(hrib_h)5 0, (10a)
›
›t(hriu)1 ›
›y(hricuy)1 ›
›h0
(hricu _h)52
�›p
›x
�1 f hriy1 hriD , (10b)�
›p
›z
�52hrig, and (10c)
›
›t(hriC)1 ›
›y(hricCy)1 ›
›h0
(hricC _h)5 hrib_C . (10d)
In deriving these equations, we only need to consider
the manipulation of the tracer equation, since the
other equations are all just special cases of it [e.g.,
C5 1 with _C5 0 for the continuity equation and C5 u
with _C being the rhs of (1b) for the zonal momentum
equation]. The derivation of the tracer equation is
given in appendix A and is similar to the derivation for
the moist isentropic continuity equation given in
Pauluis and Mrowiec (2013). The equations formu-
lated from spherical–pressure coordinates are given in
appendix B.
The weak formulation (10) provides the advantage of
formulating the zonally averaged momentum and tracer
budgets for a noninvertible choice of vertical co-
ordinate. When h is invertible, the weak formulation
reduces to the same set of mass-weighted zonally aver-
aged equations that would result when the vertical co-
ordinate is first transformed to h before averaging. For
example, when h is chosen to be u, (10) reduces to the
mass-weighted zonally averaged isentropic primitive
equations (e.g., Gallimore and Johnson 1981). In this
case, mass-weighted averages can be rewritten in terms
of the usual mass-weighted average (8); for example,
hriy5 ru~y. The pressure gradient can be written in terms
of the gradient of the Montgomery streamfunction
h›p/›xi5 rug›c/›x, where c 5 cpT 1 gz.
c. Generalized EP flux and form drag decomposition
The zonal momentum (10b) can be rewritten in terms
of a generalized Eliassen–Palm flux vector F, as follows:
›
›t(hriu)1 ›
›y(hriuy)1 ›
›h0
(hriub_h)5= � F1 f hriy1 hriD , (11)
where
F5 (Fy,F
h)5
�2hridu*y*,2hri du* _h*2�›p
›x
�H
�(12)
and
= � F5›
›yFy1
›
›h0
Fh0.
The momentum flux terms in (10b) were separated into
their mean and eddy components. The EP flux vector
contains the eddy momentum fluxes and a pressure gra-
dient term, where we have used the fact that h›p/›xi5›h0
h›p/›xiH .The EP flux divergence physically describes the eddy
momentum transfer into an infinitesimal h layer lying
in the x–z plane. This layer describes the region of fluid
lying within Dh of some fixed value, h0, and has width
Dy in the latitudinal direction. Andrews (1983) and
Tung (1986) provide a similar physical interpretation of
= � F for u layers. But unlike u, h is not assumed to be
invertible, and so an h0 contour could consist of disjoint
pieces with vertical folds and loops. Figure 1 shows
three general cases: (i) loop, (ii) zonal ring, and
(iii) surface-intersecting contour, where the h0 contour
is drawn as a thick solid line and the infinitesimal h layer
is denoted by the surrounding thin dashed lines. More
general setups can be broken up into combinations of
cases (i)–(iii).
The first term in the horizontal and vertical compo-
nents of = � F are the along-h and across-h components
of the eddy momentum flux convergence. Below, we
show that the pressure gradient term in the vertical
component of = � F describes the form drag (i.e., the
average pressure exerted in the zonal direction on an
infinitesimal h layer per unit h) that arises from both
topography and the surrounding atmosphere. When the
topography is flat (i.e., no zonal asymmetries in the
surface height), the only contribution to form drag
comes from atmospheric eddies.
A general form of theEP theorem follows from (11): if
the atmosphere is frictionless and adiabatic in the sense
that _h vanishes, then a necessary condition for steady
flow is that the = � F must vanish. This can be shown as
follows. Under the assumptions stated above, the con-
tinuity (10a) implies that ›y(hriy)5 0. Furthermore,
since y vanishes on the meridional boundaries and hri isnonzero this implies that y5 0; that is, there is no mean
circulation on h surfaces. Then from (11),= � Fmust also
vanish. The EP theorem stated above holds for finite-
amplitude waves and nongeostrophic flows. Its state-
ment is analogous to the versions given in Andrews
MARCH 2016 YAMADA AND PAULU I S 1297
(1983) and Tung (1986), but is now generalized for
arbitrary h coordinates.
We now give a physical interpretation of the pressure
gradient terms, 2h›p/›xi and 2h›p/›xiH, as form drags.
Consider the finite atmospheric layer in the x–z plane that is
bounded by the h0 contour and indicated by the stippled
region in Fig. 1 (i.e., this layer describes the subset
fh, h0g). The formdrag described by the term2h›p/›xiHis the average pressure exerted in the zonal direction on the
finite layer fh , h0g:
2
�›p
›x
�H
521
L
ððh,h0
›p
›xdz dx52
1
L
þ›(h,h0)
p dz , (13)
where Green’s theorem was used in the last step. The
increment dz along the boundary contour describes the
effective surface area (per unit y) on which the pressure
acts zonally. The boundary of the finite layer is illus-
trated in Fig. 1 by thick lines with arrows indicating the
direction of integration. The pressure along the bound-
ary of the layer can be separated into its surface (along
gsfc) and atmospheric (along gatm) components, as in-
dicated in Fig. 1:
21
L
þ›(h,h0)
p dz521
L
ðgsfc
p dz1
ðgatm
pdz
!.
We then separate the atmospheric component into its
mean and eddy parts, so that
2
�›p
›x
�H
521
L
ðgsfc
p dz1 p
ðh5h0
dz1
ðh5h0
p*dz
!,
(14)
where p(y, h0, t) was pulled out of the integral because
›(h , h0) 5 fh 5 h0g on gatm. The mean pressure p
describes the average pressure along the h0 contour and
p* is the deviation from thatmean. The formdrag exerted
on a finite layer therefore consists of the zonal pressure
exerted from the surface (i.e., an orographic form drag)
and from the surrounding atmospheric layers.
The pressure gradient term,2h›p/›xi52›h0h›p/›xiH ,
then describes the form drag acting on the infinitesimal
layer of air lying between h0 2 Dh and h0 1 Dh. This isillustrated in Fig. 1 in which 2h›p/›xi involves the re-
gion (between the thin dashed lines) given by the dif-
ference between two finite layers, fh , h0 2 Dhg and
fh, h01Dhg, in the limit as Dh goes to zero (4). When
the h0 contour does not intersect the surface, as in cases
(i) and (ii), or when there is flat topography (such that dz
along gsfc is zero) the orographic form drag vanishes.
From (14), the mean component of the atmospheric
form drag is just 2L21›h0(pDz), where Dz is the height
displacement between the endpoints of the h0 contour.
It also vanishes when the layer does not intersect the
surface or when there is flat topography, since Dz is zeroin these cases. Thus, without topography, the only con-
tribution to form drag comes from atmospheric eddies.
The form drag description of the pressure gradient
given here generalizes the discussion in Andrews (1983)
and Tung (1986) for isentropic layers to noninvertible
h layers.Additionally, our treatment of surface-intersecting
contours is handled differently. Tung (1986) does not
consider this case, while in Andrews (1983), surface-
intersecting isentropes are extended ‘‘just under the
surface’’ by a convention developed in Lorenz (1955), in
which the pressure along underground portions of
isentropes is taken to be the pressure at the surface. This
FIG. 1. A schematic showing three general types of constant-h contours (fh 5 h0g) that can occur: (left) case (i) loop, (center) case
(ii) zonal ring, and (right) case (iii) surface-intersecting contour. The gray-shaded region near the surface represents topography. The
finite layer fh, h0g is indicated by the stippled region. The boundary of the finite layer is outlined by the thick lines with arrows drawn
such that the curve is positively oriented. Thick solid lines indicate the atmospheric component of the boundary on which h5 h0, denoted
gatm, while thick dashed lines indicate the remainder of the boundary, including the surface component gsfc. In case (ii), integration over
the sidewalls of the boundary cancels from periodicity. The thin, dashed lines surrounding gatm indicate the infinitesimal layer fh0 2Dh , h , h0 1 Dhg.
1298 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 73
convention is mathematically convenient but has no
physical basis, and the explanation for the form drag
contribution from subterranean isentropes is unclear.
Here, underground conventions are avoided altogether
and the pressure gradient is only integrated at or above
the surface. As a result, the decomposition (14) provides
a clear and physically meaningful separation of the form
drag into its orographic and atmospheric components. In
section 5, we will discuss the form drag terms in (14) in
more detail and derive alternative expressions for them,
which will give the vertical EP flux its familiar form.
In computing the EP flux (12) in practice, we use
p0 5 p2 p (i.e., deviation from the zonally averaged
mean state) instead of p. This does not affect the pres-
sure gradient since ›p/›x 5 ›p0/›x; however, the use of
p0 becomes important in the decomposition (14) for re-
moving large cancellations between the orographic and
mean atmospheric form drags that arise from the hy-
drostatic pressure.
3. Zonal momentum budget on dry and moistisentropes
In this section, we compute the terms in the zonal
momentum budget (11) in both dry and moist isentropic
coordinates (i.e., choosing h 5 u and h 5 ue, re-
spectively) using MERRA data from 1979 to 2012
(Rienecker et al. 2011). The MERRA data are output
eight times daily on a 1.258 3 1.258 latitude–longitudegrid at 42 pressure levels. Only the lowest 31 pressure
levels, which lie between 1000 and 10hPa, are used,
since we only consider isentropes which lie between 240
and 360K. The tendency terms, _u and _ue, are computed
diagnostically from the data using finite differences in
space and time. Delta distributions (4) are computed
using a finite layer width of Dh 5 1.3K.
We first consider the EP flux divergence and the
Coriolis term. In the TEM and dry isentropic frame-
works, these terms are well known (e.g., Vallis 2006,
section 12.4) to make up the dominant balance in the
midlatitude momentum budget (11):
f hriy’2= � F . (15)
The EP flux divergence is largest in its vertical compo-
nent, which is mostly due to the downward transfer of
momentum by form drag. This implies from (12) that, at
least in the dry diagnostics = � F ’ 2h›p/›xi, and so the
leading-order relation (15) is essentially geostrophic
balance in the isentropic zonal mean. In contrast, the
zonally averaged circulation on pressure surfaces does
not capture the leading-order geostrophic flow in the
midlatitudes because the zonal mean of the meridional
geostrophic velocity is zero. Consequently, the isobarically
averagedmomentumbudget only reflects the higher-order
balance, which relates the Coriolis force by the ageo-
strophic flow, the eddy momentum fluxes, and surface
friction (e.g., Peixoto and Oort 1992, section 14.5).
The isentropic-mean circulation y is important in this
discussion and can be visualized by its streamfunction.
The dry (moist) isentropic streamfunction represents
the meridional mass flux across a given latitude for air
parcels with a value of u (ue) less than u0 (ue0). It is
computed as
C(f,h0)5
2pa cosf
t
ðt0
hrpyi
Hdt ,
where h denotes u (ue) in the dry (moist) case, t is the
time period over which the circulation is averaged, rp 5g21 is the density in pressure coordinates, and the
spherical–pressure coordinate form of h�iH is given in
appendix B. Figure 2 shows the 1979–2012 annual mean
of the dry and moist isentropic streamfunctions (Figs. 2a
and 2b, respectively). The dashed (solid) contours rep-
resent clockwise (counterclockwise) rotation. In Figs. 2a,b,
the contours are drawn at every 2.5 3 1010 kg s21,
omitting the zero contour. The black lines denote the
median value of zonal-mean surface u and ue, respec-
tively. PCK10 give a detailed analysis of the dry and moist
isentropic streamfunctions. Here we list a few important
features:
(i) In both the dry and the moist case, the stream-
function consists of a single, direct cell in each
hemisphere.
(ii) The cells tilt downward with latitude because air
parcels cool radiatively as they travel poleward and
gain heat from surface fluxes as they return from
higher latitudes. This tilt is steeper in the moist
case, since ue includes surface evaporation as well
as sensible heat fluxes. Additionally, the radiative
cooling of u is partially compensated by latent heat
release.
(iii) The dry circulation has two distinct cores: a tropical
Hadley circulation and an eddy-induced midlati-
tude circulation. Cool, air parcels subside in sub-
tropical latitudes near the edge of the Hadley cell,
while in themidlatitudes, air parcels rise in the storm
track from the latent heat released in precipitation.
In the moist circulation the cooling occurs at nearly
the same rate in the subtropics and midlatitudes.
(iv) The streamlines of the return flow tend to lie below
themedian surface value of u and ue, indicating that
the equatorward branch of the circulation consists
largely of cold-air outbreaks (Held and Schneider
1999; Laliberté et al. 2013).
MARCH 2016 YAMADA AND PAULU I S 1299
(v) In the midlatitudes, the total mass transport is
around twice as large when averaged on moist,
instead of dry, isentropes (PCK08; PCK10). Pauluis
et al. show that this difference can be attributed to a
low-level flow of moist air that is advected pole-
ward from the subtropics by midlatitude eddies.
Figure 3 shows the EP flux vectors (arrows) and di-
vergence (shading) computed on dry (Fig. 3, left) and
moist (Fig. 3, right) isentropes for the 1979–2012 annual
(Figs. 3a,b), December–February (DJF) (Figs. 3b,d),
and June–August (JJA) (Figs. 3e,f) means. The color
scale for all the panels is the same. The streamfunctions
are drawn to show which isentropes correspond to the
poleward and equatorward branches of the circulation
at a given latitude.
The dry isentropic EP flux qualitatively resembles its
quasigeostrophic TEM analog. In the time mean, the
midlatitude EP flux has a large vertical component, a
characteristic feature of growing baroclinic waves (Edmon
et al. 1980), and exhibits a primarily equatorward tilt in the
upper troposphere. TheEP flux divergence (convergence)
is strongest near the lower boundary (upper troposphere)
and indicates the downward transfer of momentum
through form drag, which is necessary to maintain the
surface westerlies (e.g., Vallis 2006, chapter 12). These
features are present in both hemispheres and for all
seasons, but the EP flux and circulation are much more
intense in the winter hemisphere than in the summer
hemisphere. The main difference between the dry is-
entropic and TEM formulations of the EP flux lies near
the surface, where the TEM version has a stream-
function that does not close and unrealistically large
values of the EP flux (Held and Schneider 1999; Tanaka
et al. 2004).
At high-ue (poleward branch) levels, the moist EP
flux convergence extends well into the subtropics, to
around 208 in both hemispheres and for all seasons. In
comparison, the dry EP flux convergence becomes
much weaker equatorward of around 408. The moist EP
flux also has a significantly stronger convergence maxi-
mum at high-ue levels. For the annual mean, the con-
vergence maximum is 1.5 (2.2) times stronger in the
Northern (Southern) Hemisphere as compared to the
dry case. In the winter, the convergence maximum is
around twice as large with a factor of 2.0 (2.2) for the
NH DJF (SH JJA). The factors are even greater in the
summer, 2.4 (NH JJA) and 2.6 (SH DJF), because
the moist EP flux and circulation still have a strong
signal in the summer, whereas the dry EP flux and
circulation do not.
At low-ue (equatorward branch) levels, the differ-
ences between the dry and moist EP flux divergence are
similar to those in the EP flux convergence at high-uelevels but are not as distinct. With the exception of the
summertime (NH JJA and SH DJF), the moist EP flux
divergence extends only slightly farther into the sub-
tropics. Its maximum is greater than the dry case by a
FIG. 2. The 1979–2012 annual-mean climatology of the (a) dry and (b) moist isentropic streamfunctions. Dashed
(solid) contours represent clockwise(counterclockwise) rotation. The contours are drawn at every 2.53 1010 kg s21,
omitting the zero contour. The black lines denote the median value of zonal-mean surface u and ue.
1300 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 73
factor of 1.1 (1.1) for the Northern (Southern) Hemi-
sphere in the annual mean and by a factor of 1.1 (1.1) in
the winter and 1.3 (1.1) in the summer.
Stone and Salustri (1984) found qualitatively similar
results (i.e., an enhancement and subtropical extension
of the EP flux) in the Northern Hemisphere. Their moist
EP flux is derived from the quasigeostrophic TEM for-
mulation that has been modified to account for large-
scale condensation but assumes a vertically stratified
moist static stability. They find that the inclusion of
condensation effects increases the TEM EP flux con-
vergence and divergence maxima by around 2.5 times in
the annual mean. In contrast, the maxima in our results
do not increase by such a large factor; we find only a 1.5
(1.1) times increase in the convergence (divergence)
maximum for the Northern Hemisphere annual mean.
This quantitative difference between their results and
ours may arise from the inaccuracy of the TEM formu-
lation, which overestimates the value of the EP flux and
residual circulation near the surface (Held and Schneider
1999; Tanaka et al. 2004). In their results, both the con-
vergence and divergence maxima occur near the surface
(their Fig. 6) at latitudes where the residual stream-
function does not close at the surface (their Fig. 8).
The enhancement and subtropical extension of the EP
flux reveals a greater and more extensive momentum
exchange between moist isentropes than between dry
isentropes. These differences are consistent with the
difference in the mass transport, discussed in point (v)
above. Dynamically, a larger poleward mass transport
onmoist isentropes implies a stronger Coriolis torque on
the zonal flow. The steady-state balance (15) can only
hold if there is also an increase in the eddy momentum
transfer between moist isentropes.
FIG. 3. The 1979–2012 (a),(b) annual-, (c),(d) DJF-, and (e),(f) JJA-mean climatologies of the EP flux vectors
(arrows) and divergence (shading) computed on (left) dry and (right) moist isentropes. The streamfunctions and
median surface value of zonal-mean surface u and ue are drawn as in Fig. 2, but the contours for the streamfunctions
are drawn at every 5.0 3 1010 kg s21 starting from 62.5 3 1010 kg s21. The low-level EP flux divergence is over-
saturated in the Southern Hemisphere for all seasons. The maximum values attained in (a)–(f) are [6.3, 6.9, 6.9, 7.7,
7.6, 8.5] 3 1022 PaK21, respectively.
MARCH 2016 YAMADA AND PAULU I S 1301
Figure 4 shows that this is indeed the case. Figures
4a,b show the annual mean of the Coriolis force on the
isentropic-mean circulation f hriy for the dry (Fig. 4a)
and moist (Fig. 4b) cases. Similarly, Figs. 4c,d show the
isentropic-mean pressure gradient h›p/›xi which is the
form drag up to a minus sign (section 2). The plots in
Figs. 4a,b and Figs. 4c,d are nearly identical, showing
that (i) where there is an increase in the Coriolis force
from dry to moist isentropes, there is also an increase in
the form drag and (ii) both the dry and moist isentropic-
mean flows are nearly geostrophic. Figures 4e,f show the
Coriolis force on the ageostrophic circulation, which is
given by the difference between Figs. 4a,b and Figs. 4c,d.
The Hadley, Ferrel, and polar cells can be clearly
identified. The Coriolis force acts to accelerate the zonal
flow in the upper branches of the direct cells and de-
celerate the winds in the upper branch of the indirect
Ferrel cell. The signs of the Coriolis force are opposite in
the lower branches of the cells. Notice that the color
scale is smaller by a factor of 10, and so in the extra-
tropics, the ageostrophic Coriolis term is around an or-
der Rossby number smaller. The separation between the
dominant balance and higher-order terms is greater in
the moist case than the dry case.
The other higher-order, steady-state terms in (11)
include the momentum flux convergence terms [of
which there are four different components: isentropic
mean, 2›y(hriuy); cross-isentropic mean, 2›h0(hriub_h);
FIG. 4. The 1979–2012 annual-mean climatology of the leading-order terms in the zonal momentum budget (11)
computed on (left) dry and (right) moist isentropes: (a),(b) Coriolis force on the isentropic-mean circulation f hriyand (c),(d) isentropic-mean pressure gradient h›p/›xi. (e),(f) The Coriolis force on the ageostrophic circulation,
which is given by the difference between (a),(b) and (c),(d), respectively. The color scale for (a)–(d) is 10 times that
for (e),(f). Oversaturation tends to occur near the surface and is greatest in (e),(f), with some values reaching
around 2–3 times the color scale maximum. The magenta and black lines are as in Fig. 3.
1302 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 73
isentropic eddy, 2›y(hridu*y*); and cross-isentropic
eddy, 2›h0(hri du* _h*)], the metric term from spherical
coordinates (B4b), and the surface drag, hriD. The
isentropic eddy momentum flux convergence shown
in Figs. 5a,b has a similar structure as that when
computed on isobars. Rossby waves transfer easterly
momentum away from the midlatitude jet, which
results in momentum convergence in the jet region
and divergence on the flanks of the jet (e.g., Vallis
2006, chapter 12). The wave propagation is strongest
in the upper troposphere, where there is little mois-
ture, and so the dry and moist computations are quite
similar.
The cross-isentropic eddy momentum flux conver-
gence (Figs. 5c,d) is noticeably stronger in the sub-
tropics on moist isentropes. There is a vertical dipole
structure in each hemisphere, which indicates there is
an easterly momentum transfer to higher-ue isentropes
near the surface. This low-level eddy momentum flux is
not seen in the dry case and may be related to surface
evaporation. From Dalton’s evaporation law (Peixoto
and Oort 1992, section 10.7.3), the evaporation is pro-
portional to the surface wind speed multiplied by the
saturation deficit of the air. Since the mean surface
winds are easterly at low latitudes, an easterly wind
anomaly would trigger stronger evaporation than a
westerly wind anomaly.
The residual (Figs. 5e,f) is computed by isolating
hrpiD on the right-hand side of (B4b) and then sum-
ming all the terms on the left-hand side of the equa-
tion. The residual mainly comprises the surface drag
that is needed to balance the Coriolis force on the
ageostrophic circulation (Figs. 4e,f) and is positive
(negative) in the regions of surface easterlies (west-
erlies). We do not discuss the mean momentum flux
convergence terms nor the metric term here, but their
structures follow intuitively from the structure of the
isentropic-mean winds.
FIG. 5. Eddy components of the momentum flux convergence: (a),(b) isentropic, 2›y(hridu*y*); (c),(d) cross
isentropic,2›h0(hri du* _h*), plotted for the 1979–2012 annual mean on (left) dry and (right) moist isentropes. (e),(f)
The residual, explained in the text. As in Fig. 4, oversaturation occurs near the surface in (e),(f) with some values
reaching around 2–3 times the color scale maximum.
MARCH 2016 YAMADA AND PAULU I S 1303
4. Physical description of EP flux enhancement onmoist isentropes
To physically explain why the moist EP flux is
stronger and has a greater subtropical extent than the
dry EP flux, we now turn to analyzing a case study of
synoptic eddies. Figure 6 shows a cyclone–anticyclone
pair over the South Atlantic Ocean on 1200 UTC
6 August 2000. Figure 6a shows in shading the geo-
potential height anomaly at 700 hPa. The centers lie
within a latitude band between 308 and 458S and are
centered about 158W and 108E, respectively. To the
east of the cyclone, warm moist air is being advected
poleward (the flow velocity is shown with arrows). The
boundary between the warm subtropical and cooler
extratropical air masses is roughly delineated by the
300-K u contour (black) and 310-K ue contour (ma-
genta). The warm front appears sharper when ue is used
instead of u. This difference reflects the lower-level
transport of water vapor by the eddies. Figure 6b shows
the relative humidity in which there is a channel-like
flow of water vapor, often referred to as an ‘‘atmo-
spheric river’’ (Newell et al. 1992), from the tropics to
the warm front. The 310-K ue contour closely follows
the sharp front between moist and dry air and pro-
trudes past the 300-K u contour near 108W.
FIG. 6. A case study of synoptic eddies over the South Atlantic Ocean at 1200 UTC 6 Aug 2000 to illustrate the
enhancement of momentum exchange between moist isentropes as compared to dry isentropes. (a) Geopotential
height anomaly at 700 hPa. The 300-K u (black) and 310-K ue (magenta) contours are shown. The flow velocity is
depicted by the arrows. The dotted black lines mark the latitude circles at 308 and 408S for reference. (c) Longitude–pressure profile of the geopotential height anomaly at 408S. The 305-K u (black) and ue (magenta) contours are
shown. The gray contours show the geostrophic meridional wind at every 10m s21, omitting the zero contour and
with equatorward (poleward) wind denoted by solid (dashed) contours. (e) As in (c), but at 308S and showing the
310-K u and ue contours. (b),(d),(f) As in (a),(c),(e), but for the relative humidity.
1304 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 73
Vertical cross sections of the height anomaly and rela-
tive humidity are shown in shading at 408S (Figs. 6c,d) and
308S (Figs. 6e,f). The latitude circles are marked for ref-
erence in Figs. 6a and 6b by the dotted line. The geo-
strophic velocity is shown in gray contours with poleward
(equatorward) wind denoted by dashed (solid) lines. The
305-K (310K) u and ue contours are shown in Figs. 6c,d for
408S (Figs. 6e,f for 308S). These contours roughly divide
the troposphere into upper- and lower-isentropic layers.
At 408S, the poleward advection of warm air is indicated
by the dip in both the u and ue contours on the eastern side
of the cyclone. The warm front is again sharper (i.e., the
contours dip further) when observed using ue instead of u.
This is because the poleward flow of warm air from the
tropics has a high moisture content, as can be seen by the
region of high relative humidity (Fig. 6d) that coincides
with the poleward geostrophic winds.
The difference in the zonal structure of dry and moist
isentropes associated with typical midlatitude eddies, such
as the one shown in Fig. 6, can explain the larger EP flux
observed onmoist isentropes. In Fig. 6c, there is a clear bias
for poleward geostrophic velocity in the upper-isentropic
layer and equatorward velocity in the lower layer, as is
expected for a thermally direct eddy-driven circulation. To
leading order, the EP flux divergence is proportional to the
geostrophic velocity (i.e., = � F ’ 2h›z/›xi } hygi), sincef, 0 in the Southern Hemisphere. Hence, there is EP flux
convergence and divergence in the upper and lower layers,
respectively. The convergence is stronger in the moist case,
since the moist upper layer includes the poleward flows of
both the low-levelmoist air, aswell as the upper-levelwarm
air. Physically, the poleward geostrophic wind lies in a re-
gion of strong eastward height gradient, which explains the
westward acceleration on the upper layer.
Further equatorward at 308S (Figs. 6e,f), the sharp
moisture front is still clearly defined by the 310-K uecontour. Thus, even at subtropical latitudes away from
the storm centers, midlatitude eddies can induce a large
momentum exchange between moist isentropic layers.
In contrast, there is only a weak zonal asymmetry in the
temperature structure, as indicated by the nearly hori-
zontal 310-K u contour, which shows that the eddies
have little effect on themomentum transfer between dry
isentropic layers in the subtropics.
A summary schematic is shown in Fig. 7. The hori-
zontal (Fig. 7, top) and vertical (Fig. 7, bottom) profiles
are shown for a Northern Hemisphere midlatitude eddy
system, consisting of three centers—high, low, and high.
In the top panel, a contour of constant pressure is shown
in green. The geostrophic flow is depicted by the black
lines and arrows. The advection of warm and cold air by
the cyclone create zonal perturbations in the 300-K
contours of u (orange) and ue (magenta). The ue
contour is marked by an asymmetry, in which high-ue air
penetrates farther poleward than high-u air, while the
cold protrusions of low-u and low-ue air are comparable.
This is because the poleward branch of the moist cir-
culation includes the contribution from low-level moist
air, while the return flow is fairly dry.
In the vertical profile, the 300-K isentrope divides the at-
mosphere into upper- and lower-isentropic layers. The
asymmetry in the ue contour is also observed, inwhich it dips
lower than the u contour in the regionofmoisture advection.
This gives themoist isentropic layers a larger vertical surface
area on which pressure acts in the zonal direction (13). The
westward phase lag of the isentropes relative to the pressure
anomaly creates a net westward pressure force acting on the
upper-isentropic layer and an equal and opposite force on
the lower-isentropic layer. Hence, the form drag induced by
atmospheric eddies acts todecelerate (accelerate) theupper-
isentropic (lower isentropic) layer.
5. Formal decomposition of the pressure gradientinto form drags
In section 2, we gave a physically motivated de-
composition (14) of the pressure gradient term h›p/›xiHinto the sum of the orographic form drag and the mean
and eddy components of atmospheric form drag on the
finite layer fh, h0g. We will refer to these components
of form drag [i.e., the three terms on the rhs of (14)] as
Mo, Mm, and Me, respectively, so that
2
�›p
›x
�H
5Mo1M
m1M
e. (16)
Here, we will formallymake this decomposition and find
alternative expressions for Mo, Mm, and Me to those
given in (14) that are computationally simple and have
the familiar forms of the orographic form drag and EP
flux. Additionally, we will show that the form drag in-
duced by atmospheric eddiesMe is the dominant means
by which momentum is transferred between dry and
between moist isentropic layers.
Using the product rule to incorporate the Heaviside
function inside the derivative, the pressure gradient can
be decomposed as follows:
2
�›p
›x
�H
521
L
ðL0
ð‘zsfc
›
›x[pH(h
02h)]1p
›h
›xd(h
02h)dzdx
521
L
ððD
›
›x[pH(h
02h)]dzdx2
�p›h
›x
�(17)
521
L
þ›DpH(h
02h)dz2p
�›h
›x
�2hribp*�1
r
›h
›x
�*.
(18)
MARCH 2016 YAMADA AND PAULU I S 1305
The first term in (17) was transformed into the first term
in (18) by usingGreen’s theorem, and the second term in
(17) was split into the last two terms in (18) by using the
mean-eddy decomposition:
2
�p›h
›x
�52hribp�1
r
›h
›x
�52p
�›h
›x
�2 hribp*�1
r
›h
›x
�*.
The three terms in (18) are equivalent to Mo, Mm, and Me,
respectively, and upon furthermanipulation can bewritten as
Mo52
1
L
ðL0
psfcH(h
02h
sfc)›z
sfc
›xdx , (19a)
Mm52
1
Lp�
n
i51
[zsfc(x
2i21, y)2 z
sfc(x
2i, y)], and (19b)
Me52hribp*�1
r
›h
›x
�*. (19c)
The details are given in appendix A. In (19b), the points
x2i21 (x2i) represent the left (right) endpoints of the
h0 contour(s) that intersect the surface, where there are
n disjoint contours.
The orographic form drag,Mo, describes the pressure
along the surface acting in the zonal direction on the
finite layer fh , h0g. The orographic form drag on an
infinitesimal layer, or h0 contour, is given by
›Mo
›h0
521
L
ðL0
psfcd(h
02h
sfc)›z
sfc
›xdx (20)
521
L�i
psfc(x
i, y, t)
›zsfc
›x(x
i, y)
����›hsfc
›x(x
i, y, t)
����21
,
(21)
where the xi denote the x positions of the points of in-
tersection between the surface and h0 contours in the
x–z plane. The factor j›hsfc/›xj21 arises since the delta
function is composed with h0 2 hsfc which has an x de-
pendence. The pressure on the h0 contour exerted by
topography is determined by the surface pressure at its
endpoints and depends on the geometric factors ›zsfc/›x
FIG. 7. A schematic diagram to illustrate the form drag on dry and moist isentropic layers
associated with midlatitude eddies. (top) Longitude–latitude and (bottom) longitude–height
profiles are shown for a Northern Hemisphere midlatitude eddy system, consisting of three
centers: high, low, and high. The 300-K u and ue contours are shown in orange and magenta,
respectively. In (top), the streamlines of the flow are depicted by black lines and arrows. A
representative pressure contour is shown in green. In (bottom), the pressure anomalies are
shown by green contours, with solid (dashed) lines denoting positive (negative) values.
1306 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 73
and j›hsfc/›xj21. These factors describe the effective
surface area for pressure acting in the zonal direction
and the thickness of the h0 contour, respectively. If the
topography is flat, then ›zsfc/›x is zero and there is no
orographic form drag. The expression (21) is similar to
that given in isentropic coordinates (Koh and Plumb
2004). In height coordinates (h 5 z), the geometric
factor is simply 61, since layers of constant height are
horizontal (Peixoto and Oort 1992, section 11.1).
From (19b), it is clear that the mean component of the
atmospheric form drag Mm can only be nonzero in the
presence of topography. But unlikeMo, it does not directly
involve surface pressure and should not be confused as an
orographic form drag. Rather, it arises from the fact that
surface-intersecting h0 contours do not form closed con-
tours and can therefore have a net height displacement
between their endpoints. And so, Mm describes the mean
atmospheric pressure p along the boundary of the finite
layer fh, h0g acting on the net vertical surface area of thelayer that is exposed to the atmosphere.
When topographic effects are small, the form drag
arising from atmospheric eddiesMemakes up the largest
contribution to the total form drag in (16). The vertical
component of the EP flux is approximately
Fh’2hri du* _h*2 hribp*�1
r
›h
›x
�*.
With h 5 0, Me describes the correlation between the
pressure and isentropic-slope anomalies and has a sim-
ilar form to the dry isentropic EP flux in Andrews (1983)
and Tung (1986).
Isentropic mass-weighted means and the isentropic EP
flux do not depend on whether they are computed from
height or pressure coordinates, but the individual terms
within the form drag decomposition (16) do depend on
the choice of coordinate when the isentropes intersect the
surface (appendix B). In pressure coordinates (denoted
by superscript p), the orographic form dragM(p)o involves
the surface height zsfc rather than the surface pressure,
and the mean and eddy components of atmospheric form
drag, M(p)m and M(p)
e , are formulated in terms of the is-
entropic mass-weightedmean and eddy heights, z and z*,
rather than those for pressure (B8).
Figure 8 shows the decomposition of the isentropic-
mean pressure gradient h›p/›xi into the orographic
(›h0M(p)
o ; Figs. 8a,b), mean (›h0M(p)
m ; Figs. 8c,d), and
eddy (›h0M(p)
e ; Figs. 8e,f) atmospheric form drags on dry
(Fig. 8, left) and moist (Fig. 8, right) isentropes. The
orographic and mean atmospheric form drags are shown
for their DJF mean, while the eddy atmospheric form
drag is shown for its annual mean. The eddy form drag is
nearly identical to the EP flux divergence shown in
Figs. 3a,b. It makes up the dominant contribution to the
pressure gradient term and provides the downward
transfer of momentum needed for geostrophic balance.
In the previous section, we showed how this form drag
arises from the zonally asymmetric structure of isen-
tropes induced by atmospheric eddies.
The orographic and mean form drags are of higher
order and are mostly localized near the surface. Here we
explain their observed features in the NH midlatitudes
during DJF. These features are also present, but weaker,
in the annual mean. Near the surface around 458N, in
both the dry and moist cases, the orographic and mean
form drags form a vertical dipole, centered around 280K,
with acceleration below and deceleration above. In con-
trast, the classic description of orographic form drag on
isobars involves a deceleration of the surface flow, similar
to friction (Peixoto and Oort 1992, their Fig. 11.12). This
difference can be explained by the temperature structure
associated with large-scale mountain ranges.
Figure 9 shows the 20 January 2000 daily average of the
geopotential height and potential temperature fields. The
top panel shows the horizontal distribution of the geo-
potential height at 500hPa. There are troughs to the east of
the major mountain ranges—Rockies, Alps, and Tibetan
Plateau—and ridges to theirwest. Similar zonal asymmetries
appear in the January climatology of the height field (not
shown). A simple potential vorticity conservation argument
(Holton 2004) shows that westerly wind over mountain
ranges tends to form troughs on their leeward side.
The middle and bottom panels of Fig. 9 show the ver-
tical profile of the potential temperature and geopotential
height anomaly at 458N, respectively. The three distinct
mountain ranges shaded in gray are, fromwest to east, the
Rockies, Alps, andMongolian plateau. The troughs to the
east of the mountain ranges bring about a northerly ad-
vection of cold air on the leeward side of the mountains.
This can be seen in the temperature profile by the domes
of cold air that form just to the east of the mountains. The
280-K contours are highlighted in blue. The temperature
280K lies near the center of the vertical dipoles in ›h0M(p)
o
and ›h0M(p)
m at 458N, and is where there is a positive
maximum in M(p)o and M(p)
m , as can be inferred from the
sign of the dipole. The positive signs ofM(p)o andM(p)
m can
be explained using (B8a) and (B8b). Here,M(p)o describes
the orographic form drag imparted along the surface un-
der the cold domes marked by the blue contours. Its
dominant contribution comes from the eastern slope of
the mountains, where ›psfc/›x is largest and positive and
the surface height anomaly z0sfc is also positive. From
(B8a), M(p)o must then be positive. The sign of M(p)
m is
determined by the sign of the average height anomaly
along the 280-K contours and the sign of the difference in
surface pressure between endpoints (left minus right) of
MARCH 2016 YAMADA AND PAULU I S 1307
the contours (B8b). The average height anomaly is neg-
ative (i.e., bz0 , 0), as can be inferred from the bottom
panel of Fig. 9 in which the 280-K contours extend mostly
through the troughs on the leeward side of the mountains.
The surface pressure difference is generally negative as a
result of the elevation difference between the endpoints,
and so from (B8b) M(p)m is positive. The terms M(p)
o and
M(p)m attain their maxima near 280K because above this
temperature, the isentropes begin to pass over the
mountains, while below it the isentropes emanate from
the mountains at lower elevations.
6. Conclusions
The quasigeostrophic TEM equations have become a
textbook standard (e.g., Holton 2004; Vallis 2006) for
their mathematical simplicity and clear exposition of the
effect of eddies on the large-scale circulation. The TEM
framework shows how eddies drive a direct meridional
circulation in the midlatitudes. The EP flux divergence
contains the explicit eddy forcing of the mean flow.
Physically, it describes a downward transfer of mo-
mentum through form drag and explains how eddies drive
the surface westerlies. Isentropic EP flux diagnostics
(Andrews 1983; Tung 1986; Andrews et al. 1987; Iwasaki
1989) were developed to extend the original TEM results
to finite-amplitude waves and nongeostrophic flows. De-
spite these advances in our understanding of wave–mean
flow interactions, the existing descriptions of the large-
scale circulation have largely neglected the role of moist
processes because of the difficulty in dealing with non-
invertible moist coordinates. One exception is the work
FIG. 8. The different form drags are plotted on (left) dry and (right) moist isentropes: (a),(b) orographic, ›h0M(p)
o ;
(c),(d) mean atmospheric, ›h0M(p)
m ; and (e),(f) eddy atmospheric, ›h0M(p)
e . Panels (a)–(d) show the DJF-mean is-
entropic streamfunctions, while (e),(f) show the annual mean.
1308 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 73
of Stone and Salustri (1984), who modify the quasigeo-
strophic TEM formulation to include the effects of
large-scale condensation, but require the choice of a
vertically stratified and latitude-independent moist
static stability parameter. PCK08 and PCK10 showed
that the midlatitude circulation is twice as large when
the mass flux is separated by its value of ue instead of u.
This implies a much larger Coriolis acceleration on the
residual circulation and greater momentum exchange by
midlatitude eddies than explained by dry diagnostics.
In this paper, we established the necessary framework
for studying the momentum budget on moist isentropes
and computed the diagnostic results using MERRA re-
analysis data (1979–2012). The issue of noninvertibility is
handled by formulating aweak coordinate transformation,
which involves conditionally averaging the mass-weighted
primitive equations on the value of a general vertical co-
ordinate h in the zonal–height plane. The generalized EP
flux describes the eddymomentum transfer into an h layer
from both eddymomentum fluxes and form drag. A finite-
amplitude nongeostrophic EP theorem holds in which the
EP flux divergence vanishes for steady flow in a frictionless
and adiabatic ( _h5 0) atmosphere. When the vertical co-
ordinate is invertible, the weak formulation reduces to the
usual set of mass-weighted zonal-mean equations that
could have alternatively been derived by first changing
variables and then taking the zonal mean. The formulation
does not rely on the use of underground conventions
(Lorenz 1955) and isentropes that intersect the surface give
rise to form drag terms with clear physical interpretations.
The leading-order balance in the momentum budget,
when computed on both dry and moist isentropes, is be-
tween the Coriolis force on the geostrophic flow and the
EP flux divergence. As compared to the dry EP flux, the
moist EP flux is much stronger in the midlatitudes and
extends well into the subtropics. Stone and Salustri (1984)
found a qualitatively similar increase between the dry and
moist TEMEPfluxes, but the increase they found ismuch
larger than the increase found here between the dry and
moist isentropic EP fluxes. This quantitative discrepancy
likely arises from the inaccuracy of the TEM formulation
near the surface, which overestimates the value of the
residual streamfunction and EP flux (Held and Schneider
1999; Tanaka et al. 2004).
The difference between the dry and moist EP fluxes can
be traced back to the moist branch of the circulation
(PCK08; PCK10), which consists of a low-level poleward
flow ofmoist air that rises in the storm track through latent
heat release. The moist branch is associated with an ad-
ditional momentum exchange and energy transport that is
FIG. 9. The 20 Jan 2000 daily average of (top) geopotential height at 500 hPa over the NH extratropics, (middle)
potential temperature at 458N, and (bottom) geopotential height anomaly at 458N. The latitude circle at 458N is
delineated in (top) by the dotted horizontal line. In (middle) and (bottom), the 280-K contour is highlighted in blue.
The contour intervals are 0.1 km, 5K, and 50m in (top), (middle), and (bottom), respectively.
MARCH 2016 YAMADA AND PAULU I S 1309
not captured on dry isentropes. The low-level moist air
flows poleward through regions of strong eastward pres-
sure gradients associated with midlatitude eddies. Its high
angular momentum is thereby transferred through form
drag to isentropic layers with lower values of ue but com-
parable values of u. Themoist branch is also closely tied to
the poleward latent heat transport. It can be shown that the
moist isentropic circulation accounts for the total moist
static energy transport, whereas the dry isentropic circu-
lation only reflects the total dry static energy transport
(Döös andNilsson 2011; Yamada andPauluis 2015). In the
midlatitudes, the moist static energy transport is around
double the dry static energy transport, and the difference
between the two is given by the latent heat transport
(Peixoto and Oort 1992; Trenberth and Stepaniak 2003).
Moist isentropes better partition the poleward and
equatorward flows in the midlatitudes but have the draw-
back of losing height information as a result of their non-
invertibility.Despite the increase in the vertical component
of the EP flux on moist isentropes, there is no corre-
sponding increase in the horizontal component of the EP
flux (Fig. 5), as one might expect based on baroclinic life
cycle studies (e.g., Edmon et al. 1980). Since moist isen-
tropic layers can extend from the surface to the upper
troposphere, the additionalmomentumexchange observed
betweenmoist isentropic layersmay be the result of lateral,
rather than downward, momentum exchange.
Another open question for future research is finding a
relationship between the moist EP flux presented here
and a moist PV. In the dry TEM framework, the EP flux
divergence is equal to the quasigeostrophic eddy PV flux
(e.g., Edmon et al. 1980). This characterizes the eddy
forcing of the mean flow in terms of mixing, since PV is
materially conserved and the flux is downgradient in the
time average. Tung (1986) derived a nongeostrophic
relationship between an isentropic EP flux divergence
and the eddy flux of Ertel PV. This particular relation-
ship is not formulated in terms of mass-weighted aver-
ages and, therefore, does not have a generalization
within the framework presented here. Furthermore,
formulating a moist PV brings its own challenges. A
simple moist generalization of the Ertel PV by replacing
uwith ue does not carry over the fundamental properties
from dry theory; namely, (i) this quantity is not mate-
rially conserved even for a moist adiabatic atmosphere
and (ii) it does not have an invertibility principle
(Schubert et al. 2001).
Czaja and Marshall (2006), PCK08, and PCK10 first
used moist coordinates to study the global meridional
mass and energy transports. In this study,moist isentropic
averaging was applied to the full momentum budget,
which led to the formulation of amoist EP flux. Theweak
formulation presented here is not limited to the mo-
mentum equation, moist isentropes, or to the global cir-
culation. The tracer (10d) allows for other large-scale
tracer budgets to be computed in general vertical co-
ordinates. Schneider et al. (2006) studied the hydrological
cycle by computing zonal-mean water vapor fields and
fluxes on dry isentropes. Likewise, the weak formulation
could be used to compute the moisture budget on dry or
moist isentropes directly from data output in pressure
coordinates. A similar approach to moist isentropic av-
eraging has also found useful application in moist con-
vection (Pauluis and Mrowiec 2013; Mrowiec et al. 2015)
and hurricanes (Mrowiec et al. 2016).
Acknowledgments. We thank three anonymous re-
viewers for their thoughtful comments on an earlier
version of this manuscript. This work was supported by
the National Science Foundation under Grant AGS-
0944058 and the NYUAbuDhabi Institute under Grant
G1102. The MERRA dataset was provided by NASA’s
Global Modeling and Assimilation Office.
APPENDIX A
Derivations
a. Equivalence of f and ~f when h is invertible
Here we derive (6), which was used in section 2 to
show that f and ~f are equivalent for invertibleh. Assume
h increases monotonically with height from hsfc to ‘,where hsfc(x, y, t) 5 h[x, y, zsfc(x, y), t]. Then,
hf (x, y, z, t)i(y,h0, t)5
›
›h0
(1
L
ðL0
ð‘zsfc(x,y)
f [x, y,h(x, y, z, t), t]H[h02h(x, y, z, t)] dz dx
)
5›
›h0
"1
L
ðL0
ð‘hsfc
f (x, y,h, t)H(h02h)
����›z›h����dhdx
#
51
L
ðL0
f (x, y,h0, t)
����›z›h���� dx
5 f (x, y,h0, t)j›z/›hj
h5h0.
1310 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 73
Wemust assume f5 0 below ground for the third equality
to hold [i.e., for those points at whichh0, hsfc, we have f(x,
y, h0, t)5 0, which makes no contribution to the integral].
b. Weak form of zonally averaged tracer equation
Using the continuity (1a), the tracer equation can be
written in flux form:
›
›t(rC)1= � (rCv)5 r _C . (A1)
We then take the Heaviside distribution (5) of (A1).
This amounts to multiplying (A1) by H(h0 2 h) to get
›
›t(rC)H(h
02h)1= � (rCv)H(h
02h)5 r _CH(h
02h)
(A2)
and then integrating this expression. The derivatives can
be taken outside the integral by moving the Heaviside
function inside the derivatives. This can be done by us-
ing the product rule and the following identities:
›
›t[rCH(h
02h)]5
›
›t(rC)H(h
02h)
2 rCd(h02h)( _h2 v � =h) and
= � [rCvH(h02h)]5= � (rCv)H(h
02h)
2 rCd(h02h)v � =h ,
where we have used the fact that the derivative of
the Heaviside function is the delta function and
(1e) in the first identity. And so (A2) can be written
as
›
›t[rCH(h
02h)]1= � [rCvH(h
02h)]
1›
›h0
[rC _hH(h02h)]5 r _CH(h
02h) . (A3)
Integrating (A3) over the domain D, one immediately
obtains ›thrCiH for the first term, ›h0hrpC _hiH for the
third term, and hr _CiH for the rhs. For the integral over
the divergence term,
1
L
ðL0
ð‘0
= � [rCvH(h02h)]H(z2 z
sfc) dz dx5
1
L
ðL0
ð‘0
= � [rCvH(h02h)H(z2 z
sfc)] dz dx
21
L
ðL0
ð‘0
rCH(h02h)d(z2 z
sfc)v � =(z2 z
sfc) dz dx
5›
›y
�1
L
ðL0
ð‘0
rCyH(h02h)H(z2 z
sfc) dz dx
�2
1
L
ðL0
rCH[h02h]v
sfc� n dx .
We first make the integration limits constant by bringing
H(z 2 zsfc) inside the integral. The integral is then re-
written using the product rule as the integral over a di-
vergence term minus an integral along the surface. The
integral over the x- and z-derivative terms in the divergence
are zero from their boundary conditions, and only the
y-derivative term remains. The last line is just ›yhrCyiH ,since the velocity normal to the surface, vsfc � n, is zero.So the integral of (A3) becomes
›
›thrCi
H1
›
›yhrCyi
H1
›
›h0
hrC _hiH5 hr _Ci
H, (A4)
fromwhich the tracer (10d) follows by taking the derivative
with respect toh0. Towrite the expression in terms ofmass-
weighted averages, we use the fact that hrf i5 hrif .c. Showing the equivalence of Mo, Mm, and Me in
(19) and (14)
Here we want to show that the expressions for Mo,
Mm, andMe given in (19) are equal to the three terms in
the pressure gradient decomposition (14), respectively.
We do this by showing that (i) the decomposition (18) is
equal term by term to the decomposition (14) and
(ii) the three terms in (18) are equal to the expressions
for Mo, Mm, and Me given in (19).
As forMo, it follows immediately that the first term in
(18) is equivalent to the first term in (14), since the
contour integral around ›D only consists of the contri-
bution from the bottom boundary because the left and
right boundaries cancel by periodicity and pressure de-
cays to zero at infinity. The bottom boundary can be
parameterized by x, and so (19a) follows from the first
term in (18).
As for Mm, we start by using Green’s theorem on the
second term in (18),
�›h
›x
�5
›
›h0
1
L
ððh,h0
›h
›xdz dx
!5
›
›h0
1
L
þ›(h,h0)
hdz
!.
(A5)
MARCH 2016 YAMADA AND PAULU I S 1311
To evaluate this integral, again consider the three general
cases for the shape of the h0 contour: (i) loop, (ii) zonal
ring, and (iii) surface-intersecting contour (Fig. 1). In the
case of the loop, h equals h0 along the boundary contour
›(h, h0) and so (A5) is zero since the contour is closed. In
the case of the zonal ring, the line integrals over the left-
and right-side boundaries cancel out. The top boundary
fh 5 h0g is closed by zonal periodicity and evaluates to
zero as in the case of the loop. The integral along the
surface is independent of h0, so (A5) is again zero.
The only contribution toMm then comes from surface-
intersecting contours. In this case, the left and right
endpoints of the contour, x1 and x2 respectively, depend
on h0. However, the term associated with the change in
the endpoints with respect to h0 appears for both the gatmand gsfc integrals with opposite signs and, therefore, has
no contribution to the total contour integral. The in-
tegrand along gsfc has no h0 dependence. Along gatm,
h equals h0, but since gatm does not form a closed contour
(A5) is in general nonzero. Thus, the only contribution to
the contour integral arises from the gatm segment and is
related to the change in surface height between the
endpoints of the h0 contour. In this case, (A5) becomes
L21fzsfc[x1(h0), y]2 zsfc[x2(h0), y]g, from which the
equivalence of Mm as given in (14) and (18) follows im-
mediately. In the general case, when there are multiple
h0 contours intersecting the surface,Mm is given by (19b).
Since the first two terms in the decompositions (14)
and (18) are equivalent, it follows that their third terms
must also be equal. So Me is given by (19c).
APPENDIX B
Weak Formulation of the Primitive Equations inSpherical–Pressure Coordinates
Equations (1) in spherical–pressure coordinates (e.g.,
chapter 3 of Peixoto and Oort 1992) become
1
a cosf
›u
›l1
1
a cosf
›
›f(cosfy)1
›v
›p5 0, (B1a)
du
dt5
tanf
auy2
1
a cosf
›F
›l1 f y1D , (B1b)
2r›F
›p5 r
pg5 1, (B1c)
dh
dt5 _h, and (B1d)
dC
dt5 _C , (B1e)
whereF5 gz is the geopotential; rp5 g21 is the density
in pressure coordinates; the components of velocity are
u 5 a cosfdl/dt, y 5 adf/dt, and v 5 dp/dt; and the
advective derivative is given by
d
dt5
›
›t1
u
a cosf
›
›l1
y
a
›
›f1v
›
›p.
The Heaviside and delta distribution operators are de-
fined as
hfiH(f,h
0, t)5
1
2p
ð2p0
ðpsfc(l,f,t)0
f (l,f, p, t)H(h02h) dp dl
(B2)
and
hf i(f,h0, t)5
›
›h0
hf iH, (B3)
where the surface pressure psfc depends on time. The
units of hfi are the units of f multiplied by pascals per
units of h. The mass-weighted average is defined as
f 5 hrpf i/hrpi.The weak form of the equations can be derived
analogously to the Cartesian case. The equations are
summarized as follows:
›
›thr
pi1 1
a cosf
›
›f(hr
piy cosf)1 ›
›h0
(hrpib_h)5 0,
(B4a)
›
›t(hr
piu)1 1
a cosf
›
›f(hr
piuy cosf)1 ›
›h0
(hrpiub_h)
5= � F1 hrpi tanf
acuy1 f hr
piy1 hr
piD ,
(B4b)�2r
›F
›p
�5 h1i, and (B4c)
›
›t(hr
piC)1 1
a cosf
›
›f(hr
picCy cosf)
1›
›h0
(hrpicC _h)5 hr
pib_C , (B4d)
where the generalized EP flux vector F is defined as
F5 (Ff,F
h0)5
�2hr
pidu*y*,2hr
pi du* _h*2 1
a cosf
�›z
›l
�H
�(B5)
and
= � F51
a cosf
›
›f(F
fcosf)1
›
›h0
Fh0. (B6)
1312 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 73
In the momentum equation, the geopotential gradient
term can be decomposed into orographic and mean and
eddy atmospheric form drags. The derivation is similar
to that in section 5:
21
a cosf
�›z0
›l
�H
5M(p)o 1M(p)
m 1M(p)e , (B7)
where
M(p)o 5
1
L
ðL0
z0sfcH(h02h
sfc)›p
sfc
›xdx , (B8a)
M(p)m 52bz0� 1
a cosf
›h
›l
�5 bz0 1
2pa cosf�n
i51
psfc[l
2i21(h
0),f, t]
2 psfc[l
2i(h
0),f, t], and (B8b)
M(p)e 52hr
pibz0* 1
rpa cosf
›h
›l
!*. (B8c)
The points l2i21 (l2i) represent the left and right end-
points of the h0 contour, which intersects topography.
The height deviation z0 5 z2 z is used instead of z to
remove large cancellations betweenM(p)o andM(p)
m in the
form drag decomposition, but the use of z0 does not af-fect the EP flux.
If we let h�i(z)H denote the Heaviside distribution
(B2) in spherical–height coordinates (l, f, z), then it is
easy to show that hrpf (l,f,p, t)iH 5 hrf (l,f, z, t)i(z)H .
Similarly, h›z/›liH 5 h›p/›li(z)H , and so the weak form
of the equations does not depend on whether we start
with the primitive equations in height or pressure
coordinates. However, the individual terms in the
form drag decomposition can differ depending on the
choice of coordinate. Going back to (14) and using
integration by parts, this difference can be attributed
to a boundary term:ðg
p dz5pzj10 2ðg
z dp ,
where pzj10 denotes the difference in pz between the
endpoints of g. If g is closed then this difference vanishes.
REFERENCES
Andrews, D. G., 1983: A finite-amplitude Eliassen-Palm
theorem in isentropic coordinates. J. Atmos. Sci., 40, 1877–1883,
doi:10.1175/1520-0469(1983)040,1877:AFAEPT.2.0.CO;2.
——, and M. E. McIntyre, 1976: Planetary waves in horizontal and
vertical shear: The generalized Eliassen-Palm relation and
the mean zonal acceleration. J. Atmos. Sci., 33, 2031–2048,
doi:10.1175/1520-0469(1976)033,2031:PWIHAV.2.0.CO;2.
——, and ——, 1978a: An exact theory of nonlinear waves on a
Lagrangian-mean flow. J. FluidMech., 89, 609–646, doi:10.1017/
S0022112078002773.
——, and ——, 1978b: On wave-action and its relatives. J. Fluid
Mech., 89, 647–664, doi:10.1017/S0022112078002785.——, J. R. Holton, and C. B. Leovy, 1987: Middle Atmosphere
Dynamics. Academic Press, 489 pp.
Charney, J. G., and P. G. Drazin, 1961: Propagation of planetary-
scale disturbances from the lower into the upper atmosphere.
J. Geophys. Res., 66, 83–109, doi:10.1029/JZ066i001p00083.Czaja, A., and J. Marshall, 2006: The partitioning of poleward heat
transport between the atmosphere and ocean. J. Atmos. Sci.,
63, 1498–1511, doi:10.1175/JAS3695.1.
Döös, K., and J. Nilsson, 2011: Analysis of the meridional energy
transport by atmospheric overturning circulations. J. Atmos.
Sci., 68, 1806–1820, doi:10.1175/2010JAS3493.1.
Dutton, J. A., 1976: The Ceaseless Wind. McGraw-Hill, 579 pp.
Edmon, H. J., B. J. Hoskins, and M. E. McIntyre, 1980: Eliassen-
Palm cross sections for the troposphere. J. Atmos. Sci., 37, 2600–
2616, doi:10.1175/1520-0469(1980)037,2600:EPCSFT.2.0.CO;2.
Eliassen, A., and E. Palm, 1961: On the transfer of energy in sta-
tionary mountain waves. Geofys. Publ., 22, 1–23.
Gallimore, R. G., and D. R. Johnson, 1981: The forcing of the
meridional circulation of the isentropic zonally averaged cir-
cumpolar vortex. J. Atmos. Sci., 38, 583–599, doi:10.1175/
1520-0469(1981)038,0583:TFOTMC.2.0.CO;2.
Held, I. M., and T. Schneider, 1999: The surface branch of the
zonally averaged mass transport circulation in the tropo-
sphere. J. Atmos. Sci., 56, 1688–1697, doi:10.1175/
1520-0469(1999)056,1688:TSBOTZ.2.0.CO;2.
Holton, J. R., 2004: An Introduction to Dynamic Meteorology. 4th ed.
International Geophysics Series, Vol. 88, Academic Press, 535 pp.
Iwasaki, T., 1989: A diagnostic formulation for wave-mean flow
interactions and Lagrangian-mean circulation with a hybrid
vertical coordinate of pressure and isentropes. J. Meteor. Soc.
Japan, 67, 293–312.
Johnson, D. R., 1989: The forcing and maintenance of global
monsoonal circulations: An isentropic analysis. Advances in
Geophysics, Vol. 31, Academic Press, 43–316, doi:10.1016/
S0065-2687(08)60053-9.
Juckes, M. N., I. N. James, and M. Blackburn, 1994: The influence
of Antarctica on the momentum budget of the southern extra-
tropics.Quart. J. Roy. Meteor. Soc., 120, 1017–1044, doi:10.1002/
qj.49712051811.
Koh, T.-Y., and R. A. Plumb, 2004: Isentropic zonal average for-
malism and the near-surface circulation. Quart. J. Roy. Me-
teor. Soc., 130, 1631–1653, doi:10.1256/qj.02.219.
Laliberté, F., T. Shaw, andO. Pauluis, 2013: A theory for the lower-
tropospheric structure of the moist isentropic circulation.
J. Atmos. Sci., 70, 843–854, doi:10.1175/JAS-D-12-066.1.
Lorenz, E. N., 1955: Available potential energy and the mainte-
nance of the general circulation.Tellus, 7, 157–167, doi:10.1111/
j.2153-3490.1955.tb01148.x.
McIntosh, P., and T. J. McDougall, 1996: Isopycnal averaging and
the residual mean circulation. J. Phys. Oceanogr., 26, 1655–1660,
doi:10.1175/1520-0485(1996)026,1655:IAATRM.2.0.CO;2.
Mrowiec, A. A., O. M. Pauluis, A. M. Fridlind, and A. S.
Ackerman, 2015: Properties of a mesoscale convective system
in the context of an isentropic analysis. J. Atmos. Sci., 72,
1945–1962, doi:10.1175/JAS-D-14-0139.1.
——, ——, and F. Zhang, 2016: Isentropic analysis of a simulated
hurricane. J. Atmos. Sci., doi:10.1175/JAS-D-15-0063.1, in
press.
MARCH 2016 YAMADA AND PAULU I S 1313
Nakamura, N., and D. Zhu, 2010: Finite-amplitude wave activity and
diffusive flux of potential vorticity in eddy–mean flow interaction.
J. Atmos. Sci., 67, 2701–2716, doi:10.1175/2010JAS3432.1.
Newell, R. E., N. E. Newell, Y. Zhu, and C. Scott, 1992: Tropo-
spheric rivers?—A pilot study. Geophys. Res. Lett., 19, 2401–
2404, doi:10.1029/92GL02916.
Pauluis, O. M., and A. A. Mrowiec, 2013: Isentropic analysis of
convective motions. J. Atmos. Sci., 70, 3673–3688, doi:10.1175/JAS-D-12-0205.1.
——, A. Czaja, and R. Korty, 2008: The global atmospheric circu-
lation onmoist isentropes. Science, 321, 1075–1078, doi:10.1126/
science.1159649.
——, ——, and ——, 2010: The global atmospheric circulation in
moist isentropic coordinates. J.Climate,23, 3077–3093, doi:10.1175/
2009JCLI2789.1.
——, T. Shaw, and F. Laliberté, 2011: A statistical generalization of
the transformed Eulerian-mean circulation for an arbitrary ver-
tical coordinate system. J. Atmos. Sci., 68, 1766–1783, doi:10.1175/
2011JAS3711.1.
Peixoto, J. P., and A. H. Oort, 1992: Physics of Climate.AIP Press,
520 pp.
Rienecker, M. M., and Coauthors, 2011: MERRA: NASA’s
Modern-Era Retrospective Analysis for Research and Applica-
tions. J. Climate, 24, 3624–3648, doi:10.1175/JCLI-D-11-00015.1.
Schneider, T., K. L. Smith, P. A. O’Gorman, and C. C. Walker,
2006: A climatology of tropospheric zonal-mean water vapor
fields and fluxes in isentropic coordinates. J. Climate, 19, 5918–
5933, doi:10.1175/JCLI3931.1.
Schubert,W.H., S.A.Hausman,M.Garcia, K. V.Ooyama, andH.-C.
Kuo, 2001: Potential vorticity in a moist atmosphere. J. Atmos.
Sci., 58, 3148–3157, doi:10.1175/1520-0469(2001)058,3148:
PVIAMA.2.0.CO;2.
Shaw, T. A., and O. Pauluis, 2012: Tropical and subtropical me-
ridional latent heat transports by disturbances to the zonal
mean and their role in the general circulation. J. Atmos. Sci.,
69, 1872–1889, doi:10.1175/JAS-D-11-0236.1.
Stone, P. H., and G. Salustri, 1984: Generalization of the quasi-
geostrophic Eliassen-Palm flux to include eddy forcing of con-
densation heating. J. Atmos. Sci., 41, 3527–3536, doi:10.1175/
1520-0469(1984)041,3527:GOTQGE.2.0.CO;2.
Tanaka, D., T. Iwasaki, S. Uno, M. Ujiie, and K. Miyazaki, 2004:
Eliassen–Palm flux diagnosis based on isentropic repre-
sentation. J. Atmos. Sci., 61, 2370–2383, doi:10.1175/
1520-0469(2004)061,2370:EFDBOI.2.0.CO;2.
Townsend, R. D., and D. R. Johnson, 1985: A diagnostic study of
the isentropic zonally averaged mass circulation during the
First GARPGlobal Experiment. J. Atmos. Sci., 42, 1565–1579,
doi:10.1175/1520-0469(1985)042,1565:ADSOTI.2.0.CO;2.
Trenberth, K. E., and D. P. Stepaniak, 2003: Covariability of
components of poleward atmospheric energy transports on
seasonal and interannual timescales. J. Climate, 16, 3691–3705,
doi:10.1175/1520-0442(2003)016,3691:COCOPA.2.0.CO;2.
Tung, K. K., 1986: Nongeostrophic theory of zonally averaged
circulation. Part I: Formulation. J. Atmos. Sci., 43, 2600–2618,
doi:10.1175/1520-0469(1986)043,2600:NTOZAC.2.0.CO;2.
Vallis, G. K., 2006: Atmospheric and Oceanic Fluid Dynamics.
Cambridge University Press, 745 pp.
Xu, K.-M., and K. A. Emanuel, 1989: Is the tropical atmosphere
conditionally unstable? Mon. Wea. Rev., 117, 1471–1479,
doi:10.1175/1520-0493(1989)117,1471:ITTACU.2.0.CO;2.
Yamada, R., and O. Pauluis, 2015: Annular mode variability of the
atmospheric meridional energy transport and circulation.
J. Atmos. Sci., 72, 2070–2089, doi:10.1175/JAS-D-14-0219.1.
Yang, H., K. K. Tung, and E. Olaguer, 1990: Nongeostrophic
theory of zonally averaged circulation. Part II: Eliassen-Palm
flux divergence and isentropic mixing coefficient. J. Atmos.
Sci., 47, 215–241, doi:10.1175/1520-0469(1990)047,0215:
NTOZAC.2.0.CO;2.
1314 JOURNAL OF THE ATMOSPHER IC SC IENCES VOLUME 73