(THU

UlE tISENllOWtn l.lllllAir

Dm

3 115

02875 9748

35'

THE JOHNS HOPKINS UNIVERSITY Department of Mechanics Baltimore, Maryland

AugustJS^SlUTlinT, ITMB

> ) 'O

A SIMILARITY SOLUTION FOR THE GULF STREAM

By

Onrald S. Janowitz

Technical Report No. 30 (WB-E55A Series)

Sponsored by THE U. S. DEPARTMENT OF COMMERCE ENVIRONMENTAL SCIENCE SERVICES ADMINISTRATION

The research reported in this document was supported by the U. S. Department of Commerce, Environmental Science Services Administration under Grant No. E22-50-68(G).

THE JOHNS HOPKINS UNIVERSITY Department of Mechanics Baltimore, Maryland

August, 1968 I !

A SIMILARITY SOLUTION FOR THE GULF STREAM

By Gerald S. Janowitz

I i
i
'i

Technical Report No. 3 0 (WB-E$SA Series)

Sponsored by THE U. S. DEPARTMENT OF COMMERCE ENVIRONMENTAL SCIENCE SERVICES ADMINISTRATION

The research reported in this document was supported by the U. S. Department of Commerce, Environmental Science Services Administration under Grant No. E22-50-68(G).

A SIMILARITY SOLUTION FOR THE GULF STREAM

Abstract A simple model of the Gulf Stream is considered which allows us to study in detail the inertial, viscous, and rotational effects in a narrow two-dimensional current, flowing northwards along the western edge of a rectangular ocean basin, under the influence of a uniform inflow from the interior of the basin. It is found that in the region of maximum A method for

northwards velocity all of the forces are significant. estimating the eddy viscosity is also discussed.

A SIMILARITY SOLUTION FOR THE GULF STREAM

1.

Introduction Previous detailed studies of the Gulf Stream have either considered

linear frictional models, e.g., Munk (1950), or non-linear inviscid models, e.g., Charney (1955) and Morgan (1956). Moore (1963) constructed

a model of the general circulation in an ocean basin which includes viscous effects and inertial effects through an Oseen linearization. His

y f uV^^
paper demonstrates the significance of the Reynolds number -^ j-j-i where 1 is a velocity typical of the overall circulation. / does not attempt to consider details of the flow. This paper considers a two-dimensional model of the Gulf Stream with constant density and eddy viscosity, in which the effect of the interior is to supply a uniform inflow to the current. We shall be able to study in detail the relative significance of the various forces which are acting and, thus, hope to provide information and direction for more realistic, i.e., stratified, models. The ocean basin is Idealized as rectangular and of uniform depth, H, the fluid is of uniform density and uniform eddy viscosity, and the flow is parallel to the surface. The origin of a rectangular cartesian His theory ,

coordinate system is placed in the southwestern corner of the basin with the y-axis running northwards along a rigid wall and the x-axis eastwards, A uniform flow of speed U east. approaches the wall from the

The northwards velocity along the x-axis is assumed to vanish

(Fig, 1). Making the beta-plane approximation, the governing equations

i
Figure 1,

Geometry of the problem

become

'^ Ul

(la)

^ir

(Ib)

(Ic)

The boundary conditions become

a(oyCj) ^ i^r<^/yl=- c)
(Id)

and as

X / L

" ^ ^

where L is the width of the stream. ^ : ^

We now assume that within the stream

We define a atreamfunction by

a -- - S^^ y

V - = ^K-

(2)

Differentiating (lb) and (la) with respect to x and y respectively and subtracting the results leads to the following equation for the streamfunction

or

We seek the most general similarity solution of this equation which is of the form

s^= 5r^W (\)

where

(5a)

'}f=(KK
(5) into (4) leads to

l^C'j)

(5b) Substituting Eqs,

and < < is a constant to be chosen for convenience. ?'

where

13

? = oC f^Sj^

(6b)

and to the following equation for TCI^) (7) where C/ , Cp and C o as well as (x^ are arbitrary constants. To simplify Eq, (7) we choose

^ = c,
Applying conditions (Id), we find

if

0^-0,

(8a)

^'(ao)=0,

(c)

z/3

We choose

CCi Cl)
so

-^Jy^T/^
(8d)

Equation (6b) simplifies to

?-f4)
We make the following definitions:

1/2 (9)

^2

(;L>C

U"'"~

,3/2.

(10)

Equations (6a), (2), and (7) become

(11)

and

This governing equation for T(\) and we obtain

may be integrated from zero to

Letting

Tf yCo , and applying the boundary conditions, we find

Finally the governing equation for ji'h

) bee icomes

Equation (12) has been discussed by Moore (1963), under a different set of boundary conditions, where the flow near a latitude of vanishing wind stress curl was under consideration. He discussed the asymptotic

solution of this equation and did not attempt to integrate it. The pressure field can be computed from Eqs, (la) and (lb) and we find

pGc.^')-F^foc,^) _ ^i^^_^^yi! ] _ ^,.^) ii^.^ _

and

(13)

where 2.

PX^CG^M)

^^ * h pressure at the outer edge of the current, ^e

Solutions for large "}] We now wish to study the asymptotic nature of the solution. We

define '?C']j) by

^(\)-=-\
Assuming that V f ^ ] /

+ ^CX)

(15)

is a solution to Eq. (12) under conditions (8),

for sufficiently large y , ^CTI) ^ ^ i ^^s derivatives must become small. '' For these values of. yi y 'l^^j must satisfy

Seeking solutions of

JCTI)

which remain small and are of the form ~^

^,

requires that o ^ have a negative real part and satisfy <

We may readily show the following. (a) If f^ <r (^ -^ ,2.. 6 0 , the decaying root is complex, and

we may say that a countercurrent exists. (b) If i \ > H Z i , the decaying root is real and no counter-

current exists. Hence, it may be possible that the real Gulf Stream exhibits a countercurrent at some times and none at others, depending on the value of r\ , 3. A Linear Solution In this Section we rederive Munk's solution to obtain numerical values to compare with the results of the integration of (12), We neglect inertial terms on the left-hand side of (la) and (lb) and make the boundary-layer approximation. The governing equation becomes

We seek a solution of the form

which satisfies the conditions

^fo,<^) = ^^ ro,^)-o,
The solution i s

%fco^^)^o^ ^j, (oo,^= a

~KJL
"y"
r

1/-5

This is Munk's (1950) solution.

This solution cannot directly be O j

thought of as the low Reynolds number solution of (12), since as R-^ r (?[)-^> O (12) to ^ . However, if we change the independent variable of

, where

?M-ft'"^ = r-fy''x.
)te e i and denote ^ ^
by ( ), Eq. (12) becomes (19) Then clearly

is the solution of (19) as

txi ~^

O.

Equation (18) provides numerical information which we can use as a comparison with the solutions of (12). First, we define the strength of the countercurrents, ^ , as the va lue of 'J-('^^J at the first zero of J f'h). For Munk's solution

The maximum value of ^ mum northwards velocity is

(i / , which is proportional to the maxi"

'/3
(21)

This maximum of '

Q ) occurs at

10

-1/3
The relative vorticity at the wall is proportional to T (O). We find

The information contained in Eqs. (20)-(23) will be compared to the numerical results. 4. The Numerical Integration Equation (12) and the conditions

^ f o ) = f V o l = O^
form a two-point boundary value problem.

ffoo)=

l-O

We chose to guess at

f'Vo )

and to integrate (12) outwards to a large value of /^ f ^ S- 0J, The solution was assumed to be obtained when "j ( 5"-0/'^ /, C and f c > /y * -^ ^^f ^. O^ <^~^ 'Ty I . For small R , Eq, (19) was integrated out to

' \ , 5". 0. Actually, it was found that if the integration of (12)

were continued^for sufficiently large ^ , the solution blew up. However, numerical integrations of (12) corresponding to bracketing values of T Co) blew up much more rapidly and in different manners (Fig. 2),

and it was not difficult to find the solution. 5The Values of R chosen for the Integration. We chose to integrate (12) for (^ 0 . \^ I . O j S~. 0 j I 0. Q I ,

This range enabled us to consider cases with countercurrents ( Q-0' \.0j as well as more realistic low eddy viscosity cases. We now

il

f"{o) =2.350 7.0 f"(o)-2.297 6.0 5.0 4.0 3.0 2.0 1.0 f"(o)-2.250

-1.0 -2.0 -3.0 -4.0 -5.0 -6.0

Figure 2.

Plot of

-f(li^

vs. 'h

for R = 5,0 and three values of 7

fo)

12

cons ider the ranges of H

and V

that this range of R .

encompasses.

We first estimate t/ in terms of n Gulf Stream has been flowing for about 20

At 35 N. latitude/the of latitude ['^3.0x-/0 c^m.) Thus,

LL 12 3 is of depth rr, and carries approximately 74 x 10 cm of fluid.

or

H
if rr is expressed in kilometers, we find Taking /(^ ^=^

_^^
/ ^ )C/^0 /Lytn.S^C

7
and

^ 4 ' /

i^
and

Taking two reasonable values for H , Fig, 3 gives the range of V L ) for the range of K ( considered.

H (Kn..)
o, s~o

V C^yr>^/s^c.)
R-o,f 9.-10.0 4 . Gy^lO

i-x
CZ.Z

CK^.)

4A^lo iX^lS

/.

00

I.^^IO^

4^.j
V. and H

Figure 3,

The range of // and

Thus, the range of /? considered allows a wide variation in V to be considered.

13

6.

The Results After a number of iterations, solutions to Eq, (12) under condi

tions (8) were obtained. countercurrents exist.

As expected, for f^-

0. I

and ^

hD as

The strengths of these countercurrents, -S ,

defined in Section 3 are given in Fig. 4.

THeo/ey

K=o.i O-lll

R^ l>o

S^

0.1C3

Figure 4.

The strengths of the countercurrents. -> T ( ^ 7 ~ / ^ 7 Qfy O. monotonically, and no counter-

For higher K. ^ "riTj^) approaches I'V currents exist.

The velocity profiles are plotted in Figs. 6-9.

In Fig, 5 we show the maximum northwards velocity and the result according to the linear theory (Eq, (21)), We see that the linear theory overestimates this velocity.

/?=
e ?CACT

0. / 0.2.'5)S> 2^4

/.O (9.4^^3 O-^^l

5". o O.Go^ O.'^ZQ

10,0 0,G1^ I.17S~

TH-eojey 0,

Figure 5.

The maximum value of

" ff) f

(.

^^00,^) ) ' ^rooj^)

14

II

II

o II
peJ

U O

C ^

Hc
CO

M-l O

4J O
(X.

(U

1-1

15

3lr)
I

1/

X5

II
^

II Pi

Lf^

u o
M-i

T3

d
CO

M-^
M-l O

O i-l PM

M 00
1-1

16

:^D

X -J

o
in II

a! u o

Pc
TO

r*:^

4J

00
OJ

M
bO i-i

17

^ti

(^ ^

o q

^r^ o
X
-J

to

fvj

It
^-h
o o
II PS

. o

II

I-l

o
^

o
C O

o
ro
O
CJ

r^

^
y-i

o
4J
1-1

P M

O
as

(U
I-l

60
r4

PM

18

One measure of the width of the current is given by the value of 71 where 1/ is a maximum. This is given in Fig. 10 along with the result The linear result again overestimates

from the linear theory (Eq, (22)). the value obtained numerically.

R=

0. 1 2.-5"^
2,^/

/. 0

5~'C 0.62. (D.~7l

( D^O

/. /3
1.2.1

O^H
<5.5~6

Figure 10, The value of TJ where 1/ is a maximum. Another result of interest is 7 (o) since this is proportional

to the relative vorticity at the wall as well as the viscous force exerted by the fluid on the wall. begin the numerical integration. It is also the value which we use to The actual value, as well as the linear

result (Eq. 23), is shown in Fig. 11.

0. (

1.0

^.0

(0.0 3.39 7 4.(^4^

l?BSULT
L ikJEn^

0. 2Losr5" 0. S-^b' 7

2. zey
^,^20

<5.2./5X

/ . Oo O

Figure 11. The values

O f'M^lr^-aV/^a). f

19

The linear result overestimates the actual value but gives a reasonable starting point for the integration. We now define the width of the current, ^^^ by requiring

: f f y = 0.9,
The numerical integration gives the following (Fig. 12), It is

R.=

0. /

S~". 0

/O^O

1 = 4-. 3C)
change in "^

Z.'S^Z-

2.43

12,47

Figure 12, The width of the stream, '^^, . ^("Vj)-

^' ^^

Interesting to note that increasing R from 5 to 10 makes less than a 1% . A physical measurement of the width of the stream, 2.50 gives a value for ^-X ,

combined with the assumption that ^ j^ ^ hence for U

, A physical measurement of the point of maximum north/-x gives the value

wards velocity together with the deduced value of of R for the flow and, hence, an estimate of V ,

We now note that Eq, (12) is identical with the momentum equation in the y-direction, which we can readily see if we substitute (11) into (lb). In Eq, (12) we can identify /"f with the sum of the Coriolis " YT with the inertial effect, and -^ and

and pressure forces, " ?

with the viscous force. We designate these three effects as/O/ _L V respectively. Equation (12) can be written syabolically

To determine the relative importance of these three effects, we plot the numerical results for these terms versus *[ in Fig. 13, for the range

20

I
> *^
H

^
r-l

a o
CO

cr
(U 60 H M 60

<U X!
P! 1-1

tf?
4-1

O H

CD

Od

>
0) X!

4-1

14-1

\
\

(0 (U

1
4-1 r4

c
60

0) M

a
60 H

21

of R considered.

Near the wall the balance is essentially between /3

and

V . For R = 1.0, 5.0, and 10.0, all three effects are significant in the range of y across the region of maximum northwards velocity. Thus,

the hope of matching a linear viscous boundary layer near the wall to an inviscid inertial layer at the outer edge of the stream does not seem to be a reasonable method of studying the details of the motion. The Oseen approximation has been used by Moore to consider the overall nature of the flow in an ocean basin. To determine how close to the W-l/x-tlrVM

wall this approximation remains valid, in Fig, 14 we plot

versus 7f . We see that for R = 5,0 and 10,0, the Oseen approximation, as measured by this ratio, is valid for y 1,0,

To summarize, the similarity solution indicates that any realistic model of the Gulf Stream should contain both inertial and viscous effects. The relative insensitivity of the width of the stream to the value of R, at least if R > 5,0, may lead to estimates of the eddy viscosity based and the point of maximum

on measurements of tha width ^^..,*^ki^,^p;^m velocity.

22

0/N

R = I0.0
VMAX
1

/ ^

10 .

2.0

10 .
R=5.0 0/N
'MAX

10 .

2.0

10 .
R = I.O 0/N
'MAX

()
1

< 1.0

, 2.0

1.0

0/N

R=O.I
VMAX
1. ,,

10 .

2.0

VM

Figure 14, The validity of the Oseen approximation as measured by the ratio O/f^,

0/U -

- \} Vx /(ouVx 4 -Vl/tj ^

23

References

Charney, J. G., 1955 The Gulf Stream as an inertial boundary layer. Proc. U. S. Nat. Acad, Sci.. Washington. 41, 731-740, Moore, D, W., 1963 Rossby waves in ocean circulation. Deep-Sea Res,, 10, 735-747. Morgan, G. W,, 1956 On the wind-driven ocean circulation, Mech. , 12^, 49-80. Munk, W,, 1950 79-93. On the wind-driven ocean circulation, J. Fluid

J. Met,. 2