February 1966 Taroh Matsuno 25

Quasi-Geostrophic Motions in the Equatorial Area*

By Taroh Matsuno

Geophysical Institute, Tokyo University, Tokyo

(Manuscript received 15 November 1965, in revised form 11 January 1966)

Abstract

Quasi-horizontal wave motions in the equatorial area are discussed. A single layer of homo-
geneous incompressible fluid with free surface is treated. The Coriolis parameter is assumed
to be proportional to the latitude. In general, waves of two different types are obtained as
solutions, one being the inertio-gravity wave and the other Rossby wave. They are distin-
guished from each other by the difference of frequencies and by the relationships between
pressure and velocity fields.
For the solutions of the lowest mode (waves confined near the equator), however, the
distinction between the Rossby and the inertio-gravity waves is not clear. The wave moves
westward and the frequency of this wave is compared to that of the gravity wave, if wave
length is large. With the increase of the wave number the frequency decreases and ap-
proaches to that of the Rossby type wave. The pressure and wind fields of this wave show
somewhat mixed character of the two types, and change continuously with the wave number.
In this connection it seems impossible to "filter out" gravity waves from large scale motions.
Another interesting feature of the equatorial disturbances is that the low frequency
waves are trapped near the equator. It is shown that the both waves of inertio-gravity
type and of the Rossby type have appreciable amplitude only near the equator. The char-
acteristic north-south extent of the waves is (c/ ) 1/2, where c is the velocity of long gravity
waves and is the Rossby parameter. This expression is identical with that derived by
Bretherton (1964) for inertio-gravity oscillations in a meridional plane.
In the later half, "forced stationary motion" in the equatorial region is treated. Based
on the same model, mass sources and sinks are introduced periodically in the east-west
direction. Then the motions and surface topography caused by them are calculated.
As expected, high and low pressures appear where mass source and sink are given
respectively. But these high and low cells are splitted into two parts separated by troughs
or ridges located along the equator. Strong east-west current was formed along the equator.
The flow directs from source to sink and it is intensified by the turning of the circular
flow in the higher latitudes.

is borne by the latter in the middle and high

1. Introduction latitudes. Therefore, in numerical studies
It is well known that quasi-horizontal on large scale motions by use of the primitive
motions of the atmosphere or of the ocean hydrodynamical equations, it is an important
have two different types, the one being problem to get a suitable pair of wind and
inertio-gravity wave and the other the so- pressure fields which does not include inertio-
called quasi-geostrophic wave, and that gravity oscillations of unrealistically large
these two waves behave in quite different amplitude. Many methods have been devised
manner. It is an established fact that the for that purpose. (Charney, 1955; Phillips,
most part of the energy of large scale motions 1960; Hinkelmann, 1959).
The present author proposed a scheme of
* Division of Meteorology , Contribution No. 143. finite difference of time integration, which
This work was financially supported by the
may filter out gravity oscillations in the
Japan Society for the Promotion of Science as
part of the Japan-U. S. Cooperative Science process of integration. (Matsuno, 1966). All
Program. these methods are based on the fact that
26 Journal of the Meteorological Society of Japan Vol. 44, No. 1

there are two distinct modes of motions. applying the so-called beta-plane approxim-
Moreover in the case of mathematical filter- ations we can get various characteristics of
ing procedure, it was assumed that the fre- large scale motions in the middle and high
quency of the inertio-gravity oscillations are latitudes mentioned ealier.
much higher than that of quasi-geostrophic On a local Cartesian coordinate system
waves of the same horizontal scale. These (Fig. 1), the equations of motion and of the
conditions are actually satisfied in the middle mass conservation are written as ;
and high latitudes and these methods were
used successfully.
Here arises a problem how we must modify
above arguments when we treat the motions
in the lower latitudes especially those in the (1)
equatorial area. Can we get two waves of
different types in the equatorial area? Is there
quasi-geostrophic motion even at the equator?
It is possible to eliminate the gravity oscil-
lations by use of the filtering procedures where u, v, are the velocities in the x and y
mentioned previously? directions respectively and h is the small
Concerning the wave motions in the equa- deviation of the elevation of the top surface,
torial area several works have been made in the mean value of which is denoted by H.
relation to long-term variations of sea level f is the Coriolis parameter and g the ac-
in the equatorial ocean. Yoshida (1959) celeration of gravity. As shown in Fig. 1
pointed out from theoretical considerations the x-axis is taken so as to coincide with the
that low frequency gravity waves may be equator directing eastward, and the y-axis
trapped in the narrow belt along the equator. is taken northward. Here we shall assume
This problem was further discussed by Stern that the Coriolis parameter f is linearly pro-
(1963) and Bretherton (1964), though their portional to the latitude,
studies were confined only to the motions in
the meridional plane. These authors limitted
the discussion only to the inertio-gravity oscil-
lations or pure inertia oscillations. Ichiye
(1960) investigated the wave motions near
the equator, and got the result that in the
equatorial ocean there may exist both Rossby
and gravity waves, and that the frequency
of the former is much lower than the latter.
However, in his treatments he made many
assumptions. Some self-inconsistent result
were obtained concerning the Rossby wave.
The author's intention in this paper is to
discuss the behaviours of the Rossby and the
gravity waves in the equatorial area more
precisely and to answer some of the questions Fig. 1. Model and Coordinates.
mentioned earlier.

2. Model and basic equations f= y.

The simplest model suitable for the dis- Here /3 is the so-called Rossby parameter and
cussion of our interests is the so-called diver- we shall take it as a constant. In the process
gent barotropic model, i. e., a layer of incom- of mathematical analyses made in this study
pressible fluid of homogeneous density with no further approximation will be made. It
a free surface under hydrostatic balance. It means that Coriolis parameter is treated as
is known that treating such a model and a variable at any occasion.
February 1966 Taroh Matsuno 27

It is convenient to convert basic equations pic model, the equations (1) may be inter-
into non-dimensional form. At first we shall preted in some other ways. For instance,
rewrite (1) by using the geopotential height let us consider a two-layer model of the
instead of the geometrical height h. atmosphere on pressure-coordinates. Adopt-
ing the conventional notations the equations
of motion and the thermodynamic equation
for small perturbations are written as ;

(la)

(4)

Here c2 (=gH) is the square of velocity of

pure gravity waves. By taking the units of
time and length as following,
where subscripts 1, 2 and 3 stand for upper,
[T]= (1/c )1/2 [L]= (c/ )1/2 (2) middle and lower levels respectively, and
the equations (la) are transformed into non-
is the stability factor. Taking
dimensional form ;
the difference of the first and the second equa-
tions and making use of continuity relation,
a closed system of equations for the differ-
enced quantities are derived.
(3)

(4a)

Hereafter non-dimensional equations (3) will

be treated and all symbols stand for dimen-
Here symbols with subscript d are defined as
sionless quantities. The time and length
follows ;
scales are shown in Fig. 2 as function of c.
Though our model is the so-called barotro-

This set of equations (4a) is just equivalent

Fig, 2. Units of time (left scale) and length
(right scale) as functions
pure gravity wave.
to (la) and consequently (3), if we replace
(u, v, ) by (ud, vd, d) and c2 by ci2
(= S p2/2) where ca is the velocity of the in-
ternal gravity waves. It means that the
equations (la) are valid to internal mode of
motions, and in this case the velocity should
be taken as the wind shear and the geopoten-
tial height should be replaced by thickness or
the temperature.
It could be shown that our model are ap-
plicable not only to such simplified models
but to some particular cases of the stratified
fluid, if we formulate the problem in a suit-
able manner. It will be discussed in the
of velocity of Section 7.
28 Journal of the Meteorological Society of Japan Vol. 44, No, 1

number for some difinite meridional mode n.

3. The frequency equation Since (8) is a cubic equation to w, we have
We shall consider wave motions propagat- three roots when n and k are specified.
ing in the east-west direction. Assuming that It is expected that two of the three roots
all quantities have the factor eiwt+ikx, the correspond to the two inertio-gravity waves,
equations (3) turn to ; one of which is propagating eastward and the
other westward, and the last one to Rossby
iwu-yv+ik =0
wave. In fact, the approximate values of
three roots of w for very large k, are given
(5) as;

(10)

Here the same symbols u, v, and are used

The upper two roots are identified as the
to denote the y-dependent part of the cor-
frequencies of inertio-gravity waves and the
responding quantities in (3). Eliminating u
lower, w3 is that of the Rossby wave. This
and we get the equation to v as follows ;
point is confirmed if we express the above
relations in terms of phase velocity of waves
(6) using original dimensioned parameters ;

This equation reduces to the equation treated

by Bretherton when k*0. Since we are con-
sidering wave motions near the equator or (11)
y* 0 the boundary condition ;

v*0; when y* ±* (7)

Here the velocity of pure gravity wave is
may be adequate. In the actual atmospheric denoted by cg instead of c, in order to avoid
situations there is upper limit to y !, the confusion. It is noted that effects of the
rotation and meridional mode are condensed
position of the pole and boundary conditions
should be different ones. in a single term including the factor
However, approximations in the boundary
conditions have little effect on the solutions
of lower modes, as described later.
The equation (6) with boundary conditions For arbitrary values of k, we can get three
(7) poses an eigen-value problem, just the frequencies, by solving the cubic equation
same as Shrodinger equation for a simple
(8). They are shown in Fig. 3a in linear
harmonic oscillator. The conditions (7) are scale. Fig. 3b shows the same relations in
satisfied only when the constant logarithmic scale. For n>_1, these three
is equal to an odd integer ; frequencies are completely separated from
each other over the whole range of k. Each
of them are identified with the frequency of
(8)
the waves of the following three kinds.
1. The inertio-gravity wave which pro-
Then the solution of (6) is given as;
pagates eastward (indicated by thin solid line
v (y) = Ce-1/2y2
Hn (y) (9) in Fig. 3)
2. The inertio-gravity wave which pro-
where Hn (y) is the Hermite polynomial of pagates westward (indicated by thin dashed
the n'th order. line)
The equation (8) gives a relation between 3. The Rossby wave that propagates
the frequency and the longitudinal wave westward with slow velocity (indicated by
February 1966 Taroh Matsuno 29

* wave number k
Fig. 3b. Same as Fig. 3a but both the fre-
quency and the wave number in logari-
thmic scale.
functions of n. Namely, if we consider n as
a continuous parameter, the frequencies of
the three different wave for n=0 is obtained
Fig. 3a. Frequencies as functions of wave by ;
number. wl (k ; 0) =lim wl (k; n)
Thinsolid line: eastward propating inertio- n-*0

gravity waves.
Thin dashed line: westward propagating where the subscript l denotes the three types
inertio-gravity waves. of waves.
Thick solid line: Rossby (quasi-geostrophic)
waves.
From the above considerations the three
Thick dashed line : The Kelvine wave like roots of (12) are classified as follows,
wave.
(east, gravity)
thick solid line)
Conventionally the subscripts 1, 2 and 3
will be used to denote the three kinds in the
(west, gravity)
above order.
(for k < 1/*2)2
Special treatments for n=0
Putting n=0 in the equation (8) we can (for k>_/1 /*2 )
also get three roots for w. In this case the (Rossby) (13)
equation (8) is factorized as following; (for k<1//2)

(w-k) (w2+kw-1) =0 (12) (for k>_ll / 2 )

From (12) we have simple expressions for
the three roots of w, but they do not cor- In this case, for n=0, the frequency of the
respond one to one with the frequencies of westward propagating gravity wave is not
three different waves. The classification of separated from that of the Rossby wave, but
the roots is made from their behaviors as they coincide with each other at k=1/ *2 ,
30 Journal of the Meteorological Society of Japan Vol. 43, No. 1

w2(k=1/ 2) = w3(k=1/ 2) =k=1/ 2

Except this point w3 is always smaller in (14)

magnitude than w2.
Here we should claim that one of the three
roots of (12), w=k, cannot be adopted as an
Since both u and appear in the first and
eigenvalue of the original equations (5), from
the third equations and they are algebraic
the following reasons. In the process of deriv-
equations, the above equations have solutions
ing the equation (6) and consequently the
only when
frequency equation (8), we assumed implicitly
the following relation holds (w-k) (w+k) =0

Then the solutions are obtained as follows,

=u=Ce- y2, for w= -k

=Ce1/2y2, for w=k =-u (15)
Therefore it is demanded that the denomina-
tor (w-k) (w+k) does not vanish, unless the Clearly the lower one does not satisfy the
numerator is identically zero. The solution boundary condition (7), and it is rejected.
obtained by equating the numerator to zero It is interesting that the upper solution,
does not satisfy the boundary conditions, if w= -k, is obtained if we put n=-1 in (8).
we solve the equation to . So we shall label this solution by n= -1. In
From the above reason the one of the Fig. 3a and 3b the value of w of this mode
roots for n=0, w=k, should be rejected. Then, is drawn with a thick dashed line. The
from (13) we see that the westward pro- frequency of this wave reaches zero, when
pagating gravity wave does not exist for the wave number becomes zero, whereas the
k>1/ 2and the Rossby wave does not exist frequencies of the other eastward propagat-
for k<1/ 2. ing waves have the lower bound (approxima-
In other words, in the case of the lowest tely 2n+1) in magnitude.
mode, n=0, we have only two waves, one of At any rate, it seems very important that
which is considered as the intertio-gravity for the lowest mode solutions the frequencies
wave propagating eastward, and the other of Rossby type wave continues to that of
wave, that propagates westward, can be inertio-gravity wave and there is no gap in
identified neither with the Rossby wave nor the spectrum.
with the inertio-gravity wave. The frequency
of this wave ranges from the value which is 4. Eigenfunctions
compared to that of the inertio-gravity wave, In the last section we got a set of eigen-
to the value which is close to that of the values, wn1, of the equations (3). They are
Rossby wave. labelled by double subscripts n and l where
As observed from Fig. 3a and 3b this wave n stands for n in equation (8) or meridional
connects the two families of waves, and mode of solutions, while l distinguishes the
because of the existence of this wave, the three roots of (8) for definite value of n.
vacant space in the frequency diagram is lost. It decides the type of solutions, that is,
whether it is of inertio-gravity waves (east-
Special solution not included in the equation (8) ward and westward propagating) or of
Next we shall consider another solution Rossby wave.
which is not included in the general form The eigensolutions of u and will be
(8). The frequency equation (8) was obtain- obtained from (9) and the following relations,
ed by reducing the original simultaneous
equation (5) to the equation for v only.
There may exist a solution which has no
(16)
meridional velocity v. Putting v (x, y) -0 in
(5) we get
February 1966 Taroh Matsuno 31

Making use of the recurrence formulas for

Hermite's polynomials

In the above expression the operator is

eigenfunctions belonging to eigenvalue wn1 skew-Hermitic, i. e., Hermitian adjoint opera-
are written as ; tor of is the same as except that the
sign is reversed. It is known that all eigen-
values of such operators are pure imaginary,
and corresponding eigenfunctions are ortho-
(17) gonal to each other unless degeneracy occurs.
The completeness is proved by use of the
completeness of the Hermite functions.

5. Rossby waves and gravity waves in the

where equatorial area
n=e-1/2y2Hn (y) Some examples of eigenfunctions obtained
in the last section are shown in Fig. 4 through
We note that from the above expression that 8. Units of wind velocity and pressure
if n is an odd number then v is an odd func-
(surface elevation) are arbitrary. In all
tion and u and are even functions with figures half of the one wave length in the
respect to y, and if n in even, parity of each x-direction is shown. In Figs. 4 and 5 the
component is reversed. From this point it eastward propagating inertio-gravity wave,
seems adequate to consider that the solution the westward propagating inertio-gravity
obtained by putting v-0 (the upper one of wave, and the Rossby wave are indicated by
(15)) corresponds to n=-1. a, b and c, respectively.
Special attention must be paid to the At first we shall note the pattern for n=l
solutions for n=0. In this case expression shown in Fig. 4a-4c. We notice clear dis-
(17) is valid, but as w01 we have only two tinction between the Rossby-type wave and
roots as described earlier. Putting n=0 in the gravity waves. The former is characteriz-
(17) they are simplified as follows ; ed by the geostrophic relationship between
pressure (surface elevation) and velocity
fields, while the latters have the features of
(18) inertio-gravity waves. It is interesting that
in the Rossby wave solution, strong zonal
velocity is found along the equator, which is
In addition to the above mentioned family
expected from approximate geostrophic bal-
of solutions we have the solution labelled
ance between the pressure gradient and wind.
with n=-1. The situation is similar for n=2, as shown
in Fig. 5a-5c. Approximate geostrophic
balance holds for the Rossby type solution,
(19)
though somewhat curious wind field is
observed in the vicinity of the equator.
These eigensolutions including (19) form Namely counter-clockwise vortex which is
an orthogonal complete set. located at the equator has no counterpart in
The orthogonality is derived directly from the pressure fields.
Those peculiarities found in the Rossby type
(5). We can rewrite (5) in the following
way solution might simply be attributed to vanish-
ing of the Coriolis force. Anyhow it is note-
(20) worthy that we can get " quasi-geostrophic "
32 Journal of the Meteorological Society of Japan Vol. 44, No. 1

a)
a)- -
q)

b)
b)

c)
c7
c)
C)

unit length

Fig. 4. Pressure and velocity distributions of

eigensolutions for n=1
a : Eastward propagating inertio-gravity wave
b : Westward propagating inertio-gravity wave Fig. 5. Same as Fig. 4 but for n=2.
c : Rossby wave.
February 1966 Taroh Matsuno 33

a)
a}

Fig. 7. Rossby type wave (n=0 k=1.0).

frequency diagram. As observed from these

figures there is no marked difference between
the "gravity" wave of k=0.5 and the Rossby
wave of k=1.0. It is expected from the ex-
pression for eigensolution (18), where k and
w are the parameters which control the type
of patterns, and if the differences are little
amount the resultant patterns will show little
difference, too. In our case the frequency w
changes continuously with wave number, k,
and consequently eigenfunction changes its
b) form gradually with change of wave number.
If we examine these figures, we note that
the westward moving waves have somewhat
mixed characteristics of the Rossby and
gravity waves. The relationship between
the pressure and the velocity fields is ap-
proximately geostrophic in the higher lati-
tudes, while near the equator ageostrophic
Fig. 6. Pressure and velocity distributions of wind components predominate. The con-
eigensolutions for n=0 and k=0.5 figuration of wind and pressure fields near
a : Eastward moving inertio-gravity wave the equator resemble those of the gravity
b : Westward moving inertio-gravity wave. wave for n=2. With increase of wave
number k, the former character tends to
motions in the domain including the equator. predominate, therefore the overall feature of
The most interesting are the behaviours the wind and pressure fields becomes those
of Rossby and gravity waves of the lowest of the quasi-geostrophic wave.
modes. Fig. 6a and 6b show the pressure The situation is similar for the wave labeled
and wind patterns for n=0 and k=0.5. For with n = -1, the solution obtained by putting
this case, as mentioned earlier, only two v (x, y)* 0. Namely, it is impossible to classify
solutions, i. e. eastward (6a) and westward this wave as either of the two wave types.
(6b) moving waves exist. The westward The solution of this type behaves like as a
moving wave was taken as a gravity wave pure gravity wave in the x-direction, while
in this case from the frequency diagram in the y-direction the geostrophic relation
shown in Fig. 3. In this diagram westward holds between zonal velocity and meridional
moving wave is classified as the Rossby wave pressure gradient. This feature is observed
for k>1/ 2 and as the gravity wave for k< in Fig. 8, which shows pattern for n=-1
1/ 2 , from the reasons described previously. and k=0.5. At the both ends of the cell
The pressure and wind distributions for n=0, where the longitudinal pressure gradient is
k=1.0 are shown in Fig. 7. This solution large, the feature of pure gravity wave is
belongs to the Rossby type according to the marked, while in the middle part where
34 Journal of the Meteorological Society of Japan Vol. 44, No. 1

detail. As understood from the expression

for eigensolutions (17) the waves of both
gravity and Rossby type of lower modes are
confined in the region near the equator. From
the equation (6) we can recognize that the
solution v has an appreciable amplitude
only in the domain,

(21)
Fig. 8. Pressure and velocity distributions of
eigensolution for n=-1 and k=0.5.
This wave behaves like as the Kelvin wave.
Within this domain the solution becomes
wavy, and in the outer part of this domain
longitudinal pressure gradient is small and it approaches to 0. y* denotes the approximate
zonal velocity is large, the other aspect, the north-south extent of the wave. The equa-
geostrophic balance between the pressure and tion (6) is equivalent to a equation which
the wind fields is pronounced. It is plausible describes the wave motions in an inhomo-
that the former character predominates in
geneous medium and condition (21) determines
short waves and the latter becomes more the domain where refractive index is positive.
pronounced in long waves. Bretherton (1964) gave an explanation to the
In conclusion we might say that there is
trapping phenomenon of pure intertia oscilla-
no marked difference between the Rossby
tion from the view point of reflection of waves
and the gravity waves for the lowest modes, at the top and the bottom of the fluid.
the wave confined near the equator. Since In our case, the existence of trapped modes
we have no physical reason to distinguish of waves may be understood as a result of
"quasi -geostrophic wave" and "gravity wave"
refraction of primary waves. Though the
we cannot apply the concept of filtering to the
medium is not inhomogeneous, the variation
motions in the equatorial area. of the Coriolis effects give arise the varia-
6. Trapping of waves in the equatorial area tion of propagation velocity of inertio-gravity
waves. The phase velocity of inertio-gravity
As mentioned previously, one of the charac-
wave is given as ;
teristic phenomena concerning the equatorial
disturbances is trapping of the waves of low (22)
frequencies. This problem was first discus-
sed by Yoshida (1959), based upon the same where k is the vector wave number in
set of equations as treated in this paper. horizontal plane and the velocity of pure
But his analysis was not complete, because gravity wave is denoted by cg. The relation
he derived an equation of surface elevation (22) is valid in this discussion, because now
which was difficult to be dealt with and he we are concerning propagation of wavelets
suggested the existence of trapped waves from a wave front, so we may take the
from the asymptotic behaviors of the solution. Coriolis parameter is constant in such a small
Stern (1963) treated inertia oscillation in the range. Since (22) states that propagation
low latitudes and Bretherton (1964) discussed velocity is larger for higher latitudes, the
inertio-gravity oscillations in a two-layered wave generated near the equator will be re-
ocean. Either of them confined the problem fracted and reflected toward the equator.
to the motions in the meridional plane, i. e., In this meaning the equator plays the role
they assumed the motion is uniform in the of a duct in the propagations of inertio-
longitudinal direction. Therefore it might be gravity waves. The solutions of lower mode
said that they did not discuss about "waves ", we have obtained are the guided waves
but motions of fluid ring, so to say. through this duct.
Here we shall discuss about the trapping From the same view point we shall consider
of waves in the equatorial area in more the trapping of the Rossby waves. The pro-
February 1966 Taroh Matsuno 35

pagation velocity of the Rossby waves by

-plane approximation is written as 7. Validity of the approximations in con-
nection with the tidal theory
In this article in Sections 2 through 4 we
have developed a theory on free waves in the
equatorial area. Since we have been dealing
with the motions of fluid of constant depth
where H is the mean depth of the fluid. under hydrostatic balance, and which is
Here we note that the larger the f is the subject to the earth's rotation, the problem
smaller (in magnitude) cR becomes, i. e., the should be included in the theories on tides.
velocity of the Rossby waves is smaller in In fact, it will be demonstrated that the equa-
the higher latitudes. Then we cannot expect tion we have treated is an approximate form
reflection of the Rossby waves at higher of the Laplace's tidal equation which is valid
latitude, though in the solutions obtain in the near the equator, and that the solutions we
Section 4 we see that the Rossby waves of have obtained in Section 4 are the free wave
lower modes are confined in the equatorial solutions of the approximated tidal equation.
region, too. The simple explanation of trap- Therefore the waves which we called inertio-
ping phenomena in terms of refraction of gravity waves and the Rossby waves are just
waves is not correct for this case. equivalent to approximate forms of free
Next we shall discuss the nature of the oscillations of the first and the second kind,
wave labelled by n= -1. This wave is very respectively.
similar to the Kelvin wave. The Kelvin wave Here we shall examine to what extent our
is a wave which propagates along the coastal solutions are applicable as approximate solu-
line with the velocity of long gravity wave. tions of the rigorous tidal equation.
The deviation of the surface elevation is just In the mathematical analyses performed
in geostrophic balance with the motions as- in Section 2, we treated the equation to v,
sociated with the waves. Particle velocities the meridional component of velocity. For
are parallel to the coastal line and have no the sake of comparison with the tidal theory,
transverse component. In these points the it is convenient to deal with the equation to
two waves are the same. In the case of
Kelvin wave, however, coastal boundary is
essential, because the amplitude of the wave
increases exponentially towards the coast.
In the case of the equatorial Kelvin wave
there is no coastal boundary but owing to
changing of sign of the Coriolis parameter,
the wave can be confined only in the vicinity
of the equator.
Finally the author should stress that there (23)
is a possibility that the inertio-gravity waves
of long periods could propagate through the where w and k are the frequency and the
equatorial duct, so to say. It seems promis- wave number in the longitudinal direction,
ing to explain the long period variation of respectively. All quantities are non-dimen-
sea level observed in the equatorial ocean, sionalized by use of scaling given as (2).
in terms of these guided waves. Now according to the tidal theory (for in-
Concerning the problems, in what way an stance, see Siebert, 1961), the Laplace's tidal
incident wave is refracted and trapped, or by equation which determines the meridional
what cause the trapped modes of waves could distributions of horizontal velocity divergence
be excited, it is necessary to carry out more and pressure is written as;
detailed analyses and will be discussed else-
where.
36 Journal of the Meteorological Society of Japan Vol. 44, No. 1

which period are order of days or so. As to

the next point, if we use the same scaling
for corresponding quantities in the two equ-
(24) ations the two terms become equal. Then we
where f is the frequency of oscillations divided may conclude that the equation we have
treated becomes identical with the tidal
by twice of the earth's angular frequency 2 ,
equations if we make approximations valid
s is the wave number in the longitudinal
near the equator.
direction, a the radius of the earth, g the
accerelation of the gravity and i is cos e Next the boundary conditions should be
discussed. Since the rigorous tidal equation
where is the co-latitude. Here noted that
is formulated on the spherical cordinates the
h is the so-called equivalent depth, and in
the tidal theory it is an eigenvalue to be boundary conditions to (24) is;
determined, because f is a specified quantity =0 at =±1 (for s* 0) (26)
as one of the periods of the tidal oscillations.
In our problem, however, we are treating This condition implies that should vanish
free oscillations and therefore the procedure at the both poles which are the singular
is reversed, i, e., h is the depth of the fluid points of the coordinate system. In our
that is prescribed and the problem is to find treatment the boundary conditions imposed
out eigenvalues of f as function of s, by were
solving (24) with the suitable boundary con-
o* when ± (27)
ditions at =±1.
Since we are considering the problems of In order to compare the above two conditions
the equatorial region or =O, the approximate we shall consider the behaviors of the solu-
form to (24) is obtained by neglecting t tions in the region distant from the equator
against 1 as follows ; or , » 0. Considering the situations

(28)

the solution of (24) in the region >>0 behaves

(25) as a convex function and should grow ex-
ponentially with , unless f is an eigenvalue.
One would note that equation (25) is very Then the eigenvalue which is determined so
similar to (23), if one reminds that the fol- as to satisfy the boundary condition (26)
lowing correspondences hold ; f*w, s *k, - y. should not differ so much from the value
Discrepancies are found in the two points : which is determined to satisfy (27). Other-
The second term is multiplied by factor wise the exponentially growing component
(1-f2) in (25), and the last term in the will become remarkable with increase of .
brackets is multiplied by non-dimensional Summarizing the above discussions the
parameter 4a2 2/gh. Since f is the ratio of the solutions obtaind in this article would be
frequency of the oscillation considered to that justified, a posteriori, as approximate solutions
of the earth's rotation, we may neglect it of the tidal equation, if the conditions (28)
against unity so far as we concern with the are fulfilled.
low frequency oscillations, say oscillations of The validity of these assumptions for
February 1966 Taroh Matsuns 37

various kinds of motions are examined and surfaces, which is equivalent to the tem-
listed below. perature, and Q should be interpreted as the
heat sources and sinks.
8. Forced stationary motion Equations (29) are transformed in dimen-
So far we have discussed free waves in the sionless forms in the same manner as
equatorial region. In this section we shall adopted in Section 2.
examine what motion is caused when some Here we shall assume that all external
external forces are working, for one simple forces have sinusoidal variation in the x-
example. The problem is ; what motions and direction. Then the solutions will have the
surface elevations will be caused when the same variation, and factor eikx will be
mass sources and sinks are put alternatingly separable. Then equations (29) are reduced
along the equator. By solving this problem to ;
we will get some informations on particular
properties of atmospheric and oceanic motions
in the low latitude area. The methods of
mathematical analysis are explained below, (30)
in a generalized form.
Here we shall start again from the same
model as treated in Section 2. Now we are
considering stationary state resulted from
some external causes, the equations of motion here various quantities are nondimensionalized
and continuity are written as follows ; and common factor eikx is omitted. The same
symbol as in (29) are used, for no confusion
will occur.
Next we shall consider the boundary con-
ditions. It may be plausible to assume that
(29) external forces or inhomogeneous terms in
(30) are not zero only in the finite distance
from y=0. Then the solution should have
no-zero value in the finite domain, i, e., the
boundary conditions to solve (30) are ;
where u, v, h, f, g and H are the same as
u, v, *0 when y*±
in Section 2. In this case in right hand side
of the equations two sets of terms are added. The free wave solutions obtained in Section
Fx, Fy, and Q are the x and y components 4 satisfy the same boundary conditions and
of forces and mass source (or sink) respec- it is proved that the set of all eigenfunctions
tively. They are to be given as external form a complete set, that is, any arbitrary
forces and tend to cause motions and undula- set of three functions (which satisfies the
tions of the surface elevations. If such ex- condition that integration of square of abso-
ternal causes are given and the fluid is lute value over the whole domain remains
set in motion, there will appear some re- finite) can be expressed by the linear com-
stitutive effects, say frictional forces, dif- bination of eigenfunctions. Therefore it is
fusions. They will counteract the motion possible to express both solution and forcing
and at last the balance will be reached. In functions of (30) by serieses of free wave
order to simulate this situation, we add the solutions. We shall write the equations (30)
terms (- u, - v, - h) in the right side of symbolically in the following way;
(29) as the simplest form that express these
effects. ( + I) X= (31)
The equations (30) may describe the internal where X and are solution and the inhomo-
mode of the atmospheric motions, as mention- geneous terms of (30), respectively. ( + l)
ed in the last part of Section 2. In this case, is the operater which corresponds to the left
h expresses the thickness between two isobaric hand side of (30). I is the unit operator and
38 Journal of the Meteorological Society of Japan Vol. 44, No. 1

f expresses the first terms in each equation Namely we shall seek for the solution of
of (30) and expresses the rest terms. We (30) which is caused by the inhomogeneous
shall denote free wave solutions obtained in terms as following ;
the previous sections by m, of which fre-
quency is wm. Then the following relation
holds

m= -iwm m (32)
Here it is noted that 's are labelled by In this case the series of solution in terms of
one subscript m, because it is possible to re- terminates up to the forth term.
arrange the set of eigensolutions obtained in In Fig. 9 the distributions of the surface
Section 4 in a single array. elevation and the circulation are shown,
together with the distribution of mass source
Expanding x and in terms of ,
and sink given as forcing term. Numerical
x= am m = bm m values adopted are k=0.5 and =0.2.
An outstanding feature of the circulation
and inserting this expression into (31), we
pattern is the strong zonal flow confined in
have the vicinity of the equator. Associated with
this flow ridge and trough are located along
( + I) am m= bm m
the equator which divide the pressure cells
Making use of (32) the above equation turns into two petals.
to Mathematically speaking, we can explain
this result by the fact that Rossby wave solu-
am (-iwm+ ) m= bm m (33)
tion predominates while gravity waves are
Since e's are orthogonal with each other we suppressed. It is quite natural because the
have following relation between the two frequency of Rossby wave is much smaller
expansion coefficients ; than that of the gravity wave and consequent-
ly more resonant to the excitation.
In the physical point of view, the situation
(34)
may be explained in the following way. For
the sake of understanding, we shall imagine
This relation represents the "response" of
the series of events which would take place
the system (30), i, e., it shows that when bm
if the mass sources and sinks are given at a
or external causes are given, how this model
certain moment.
reacts to them. It is understood from (34)
The surface elevation will be subject to the
that the low frequency modes will have large distribution of mass sources and sinks, i. e.,
amplitudes while the high frequency modes where the mass is added the surface tends
will be suppressed if the input is of the same to be raised and where the mass is extracted
amount. It is simply due to the fact that we the surface is depressed. The fluid motions
are treating stationary motions or the input
induced by the surface inclination will be
of zero frequency. Therefore the lower the fre- deflected by the Coriolis force, in higher lati-
quency of a free wave is, the more resonant tudes. Then flow field is settled as to be
it is to the excitation. Since bm is got by
geostrophic flow corresponding to the pres-
the following relation, sure field. In this way anticyclonic or cyclonic
flow fields are established where high or low
pressure cells are located. In the vicinity of
the equator, we see that a strong zonal flow
we can get {am} and consequently the solu- exists. This flow is caused by the imp-
tion of (30) for any arbitrarily given external pression of the mass sources and sinks, i, e.,
forces, (Fr, Fy, Q). the flow is directing from the mass source to
Here we shall treat one particular solution the mass sink. It is interesting that this
of (30). equatorial zonal flow is intensified by the
February 1966 Taroh Matsuno 39

Fig. 9. Stationary circulation pattern (lower) caused by the mass source and sink (upper).

turning of flows from the higher latitude. lem ", the above process may be summerized
Namely the circulations in the high latitude as follows : In the higher latitudes the flow
regions bring converging or diverging motions field is set up so as to balance geostrophical-
towards or from the equator at the end of ly with the pressure field which was gene-
each cell, because the sense of rotation is rated by mass sources and sinks. On the
opposite in each hemisphere for the same contrary, in the equatorial region, pressure
pressure pattern. If we note, for instance, distribution tends to follow wind field, ignore-
the western edge of the low pressure cell in ing the impressed mass sources. It is note-
Fig. 9 the flow is converging towards the worthy that, in this example, the surface
equator, and this flow turns to the east. elevation pattern near the equator is not
On the other hand, horizontal velocity con- the reflection of the external forces, though
vergence brings surface elevation and makes the external forces have the maximum
a ridge along the equator. In this way magnitude at the equator. It seems to be
geostrophic balance between the pressure and very important that, in the equatorial area,
the flow fields is attained in the vicinity of pressure or temperature fields could be op-
the equator, too. posite in sense to the external heatings.
In other words, when the fluid is supplied
at some place and extracted at the other place 9. Formal development of the theory for
the compensating current prefers to flow general stratified fluid
through the equator. So far our discussions were confined to the
If we speak in terms of "adjustment prob- so-called divergent barotropic model as des-
40 Journal of the Meteorological Society of Japan Vol. 44, No. 1

cribed in the second section. It is thought the differential equations with respect to the
that this model simulates the atmosphere or horizontal coordinates only. Further they
the ocean in fundamental hydrodynamic pro- are equivalent to (1) which we have treated.
perties. We cannot apply the results obtained Namely, if we put
in this study directly to the actual atmo-
spheric or oceanic phenomena. But fortunate- (41)
ly the mathematical discussions made in the
previous sections are applicable to some par- ;40) is written as
ticular cases of stratified fluids if we formulate
the problem in the following manner.
Let us consider a stably stratified fluid on (42)
the rotating earth. With some approxima-
tions conventionally adopted, we can derive Clearly this equation corresponds to the mass
the system of equations governing small conservation equation in the one-layer model
treated in the previous sections, and (35),
perturbations of the fluid motion and the
density as follows ; (36) and (42) form an equivalent set of
equations to (1), or the set of equations for
two-dimensional quasi-horizontal motions of
(35; the homogeneous fluid with (apparent) depth
H*.
(36) Here H*is to be determined in the follow-
ing way. In solving (41), the boundary con-
ditions are
(37)
w=0 at z=0

(38) at z=H

where is the surface elevation.

(39)
These conditions are transformed into,

Here (u, v, w) are the velocities in the x,

at z=0
y, z directions respectively. p the pressure,
p the density. All these symbols stand for
small perturbation quantities, while (z) is
(43)
the basic density and o is the constant mean
density, f being the Coriolis parameter and
Then the equation (41) with the above con-
g the acceleration of gravity. By eliminating
ditions poses an eigenvalue problem, and H*
p and w between (37), (38), and (39) the
equation for p is obtained as follows, will be determined as eigenvalues. The cor-
responding eigenf unctions will form a set of
orthogonal functions. If they are complete
(40) we can express any function of z in terms
of them.
where z is the abreviation of / z, the basic The mathematical discussions mentioned
stability factor. Combining this equation above are the same as usually found in the
with (35) and (36) we get three equations theory of wave motions of long wave length
for three variables u, v, and p. (Eckart, 1960).
We note that in this set of equations, dif- Summing up, the equations (35) through
ferentiation 'with respect to z appears only (39) can be solved by the method of separa-
within the brackets in (40), and if we assume tion of variables. Then equations concerning
that this term is proportional to (p/ o), the the horizontal coordinates turn to be the
system (35), (36) and (40) are reduced to same as those for homogeneous density with
February 1966 Taroh Matsuno 41

a parameter H*, apparent depth of that fluid. ridional extent of the wave is of the order
Therefore the analyses made in the previous of (c/ ) 1/2, where c is the velocity of long
sections are valid, if we note one particular gravity waves and is the Rossby para-
mode in the vertical structure. meter. For the inertio-gravity wave this
phenomenon is interpreted as the results
10. Summary and conclusions of refraction at the both sides of the
Wave motions in the equatorial area are equator. In this meaning the equator can
discussed. Based upon the so-called divergent be a wave guide for the propagation of
barotropic model, linearized equations for long period gravity waves. It seems that
small perturbations are solved. this effect may play some roles in the
The following results are obtained from the maintenance of the atmospheric or oceanic
mathematical analyses. disturbances in the equatorial areas.
(1) Even in the equatorial area there are (4) One particular example of the stationary
two types in the waves motions. The one circulation in the equatorial area is obtained.
is the inertio-gravity waves and the other Considering the same model (one layer of
is the Rossby waves. They are distinguished homogeneous fluid) we have calculated the
from each other, by the difference of their circulations and the pressure distributions
frequencies. Namely frequencies of the when mass sources and mass sinks are
inertio-gravity waves are much larger than imposed alternatingly along the equator.
that of the Rossby waves. The characteristic features of the circula-
(2) For the particular mode, however, the tions and the patterns of surface elevations
distinction between the two waves is not are :
clear. The wave of the lowest mode, which (i) In the higher latitudes the surface
is the smallest in the north-south extent tends to be raised where mass is added and
and propagates westward, has somewhat depressed where mass is extracted.
mixed characters of the Rossby wave and (ii) In the vicinity of the equator,
of the inertio-gravity wave. The frequency- however, the deviations of the surface
wave number relationship for this wave is elevation is less than that in the higher
similar to that of the Rossby wave when latitude in magnitudes. As a consequence
the longitudinal wave length is smaller high and low pressure cells are splitted into
than the meridional extent. For long wave two parts separated by the equator.
part, however, the frequency becomes very (iii) Strong zonal flow is formed along
large and approaches to that of the gravity the equator. The flow directs from the
wave of the same wave length. The re- mass source to the mass sink. This equa-
lationship between the pressure and the torial zonal flow is intensified by the turn-
wind fields also shows the mixed characters ing of the flow associated with the high
of the two types, the Rossby wave type latitude circulations.
and the inertio-gravity wave type. Either (iv) In the higher latitude region, the
of the two types becomes predominant de- velocity fields are in geostrophic balance
pending upon the wave length. In these with the pressure fields.
aspects this wave is situated in the inter- (v) If we apply the above results to the
mediate position of the two utterly different two level models of the atmosphere, we
wave regimes, and connects them con- may deduce the following things.
tinuously. If the air in the equatorial region is
It is an interesting problem that whether subject to the differential heating in the
the wave of such type exists or not in the longitudinal direction, the resultant pressure
actual atmospheric conditions, and if it and wind fields will show the following
exists what role it plays in the atmospheric features : The impressed heating and cool-
motions in the equatorial area. ing will produce the low pressure and the
(3) For lower modes the both waves of the high pressure (on the lower level), respec-
inertio-gravity type and of the Rossby type tively. The wind blows geostrophically in
are confined near the equator. The me- the high latitude region. The induced
42 Journal of the Meteorological Society of Japan Vol. 44, No. 1

vertical motions counteract to the imposed Research Institute, and to Prof. Y. Sasaki,
heat sources and sinks, and their effects University of Oklahoma, for their discussions
are stronger near the equator than in the on this work. Finally the author thanks to
higher latitudes. Consequently the warm Miss M. Onozuka and Mr. Y. Fujiki for type-
or cold air produced by the heating or cool- writing the manuscripts and drafting the
ing are splitted into the two parts, by the figures.
relatively cold or warm air belt located at
the equator. References
The present work concerns only with the Bretherton, F.P., 1964: Low frequency oscillations
mathematical analyses of the simplified trapped near the equator. Tellus, 16, 181-185.
hydrodynamical equations, and we must Charney, J.G., 1955: The use of the primitive
equations of motion in numerical prediction.
be careful in applying the results obtained
Tellus, 7, 22-26.
in this study to the actual atmospheric or Eekart, C., 1960: Hydrodynamics of Oceans and
the oceanic disturbances. Atmocpheres, Pergamon Press, New York, 1960,
It is most interesting if we find out pp. 290.
some phenomena inferred in this study, for Hinkelmann, K., 1959: Emn numerisches Experi-
instance, the trapped waves, in the actual ment mit den primitiven Gleichungen. The
atmosphere. Atmosphere and the Sea in Motion ; Scientific
Contributions to the Rossby Memorial Volume,
Acknowledgments Rockfeller Institute Press, New York, 486-500.
The author expresses his hearty thanks to Ichiye, T., On long waves in a stratified, equatorial
Prof. S. Syono, for his guidance and en- ocean caused by a travelling disturbance.
Deep Sea Res., 6, 16-37.
couragements throughout this work. This
Matsuno, T., 1966: Numerical Integrations of
work was made as a part of author's doctoral
primitive equations by use of a simulated
thesis under his guidance. The author is backward difference method. J, meteor. Soc.
deeply indebted to Prof. Y. Ogura, for his Japan, Ser. II, 44, 76-84.
many stimulative suggestions and criticism. Phillips, N.A., 1960: On the problem of initial
He also wishes to express his gratitudes to data for the primitive equations, Tellus 12,
Prof. K. Hidaka who gave him valuable com- 121-126.
ments concerning the tidal theory, and to Siebert, M., 1961: Atmospheric Tides. Advances
Prof. K. Yoshida who kindly permitted the in Geophysics Vol. 7, Academic Press, New
York, 1961, 105-187.
author to refer his unpublished manuscripts
Stern, M.E., 1963: Trapping of low frequency
and gave him many suggestions. Thanks
oscillations in an equatorial boundary layer.
are due to Dr. M. Yanai, who read the manu- Tullus, 15, 246-250.
scripts and gave the author many valuable Yoshida, K. 1959: Preprints, International Ocea-
advices. Thanks are extended to Dr. K. Ta- nographic Congress, 789-791, American Associa-
kano, the Ocean Research Institute of Tokyo tion for the Advancement of Science, Wash-
University, Dr. K. Gambo, Japan Meteorologi- ington D.C., 1959.
cal Agency, Dr. T. Murakami, Meteorological

Note added in proof : After the completion of the manuscripts, the author noticed the work by
Rosenthal (1965) *. The equatorial waves treated by him are essentially the same as the waves of n=0
in the present article, though in the Rosenthal's work uniform basic currents are considered. Rosenthal
paid attention mainly to the structures of Rossby type waves of relatively short wave length,
whereas the present author treated general properties of the waves in the equatorial region.

* S . L. Rosenthal, "Some preliminary theoretical considerations of tropospheric wave motions in

equatorial latitudes ", Mon. Wea. Rev. 93, 605-612, (1965).
February 1966 Taroh Matsuno 43

赤 道 近 く で の 準 地 衡 風 的 運 動

松 野 太 郎

東京大学理学部地球物理学教室

コ リオ リの 力 が働 か な くな る赤道 近 辺 で の大 規 模 運動 の 特性 を理 論 的 に検 討 して み た.自 由 表 面 を も った 単 層 の 流

体-い わ ゆ る発 散 順圧 モ デ ル-に つ い て 線 型 化 され た 運 動方 程 式 を扱 い 東 西 方 向 に 動 く自 由波 動 の解 を 求 め る
と,一 定 の ス ケ ー ル に対 し て3つ の解 が得 られ た.こ れ らは振 動 数 解 の形(圧 力 及 び 運 動 の 場)か ら夫 々東 向 き お
よび 西 向 きの慣 性 重 力波 お よび ロス ビー波 で あ る こ とが わ か る.但 し南 北 ス ケ 一ル最 小 の も のに 関 し ては そ の区 別 は
明 瞭 で な く一方 の型 か ら他 方 の型 に連 続 的 に かわ る.ロ ス ビー波 に相 当 す る解 は 風 と圧 力 の 関 係 が 高 緯 度 で ほ ぼ地 衡

風 的 で ある ご と,お よ び赤 道 近 くで特 異 な ふ る ま い をす る の が特 徴 で あ る.
次 に 同 じモ デ ル熱 冷源 に相 当 す る もの と して 東西 に 周期 的 なmass source,sinkを 与 え,定 常解 と求 め た.熱 源
に 相 当 す る所 は低 圧 に な る が赤 道 で分 断 され,赤 道 の ご く近 くは やや 逆 セン ス に な り,こ れ に 伴 っ て高 緯 度 と逆 向 き
の強 い 流れ が生 ず る こ とが 分 っ た.