r
ijkswaterst
aar-'
~
I
~
I
,...I'
,
1
·
1
-
I
data process
i
ng
di
v
isi
on
-l
on the parabolic equation
I
method tor water wave
I
propagation
,
I
I
,--
.
1
by
a.c. radder
,
-
~
I
[uly 1978
I
.
,
1
,
I
-
~
I
'
I
I
I
:::::1 :::::1 n:::nn i;:i)1Il.. ,••• " ......u
...
".
...
.
, J:::::::: ,:!::;;::;:~;;~:;::..:::'::::!:r •••••• ." ...-•••••_1'2".' 11S••',",.•,. , '"'...
.
,,.
~
"'~\~,
,... • •••• 14',.. • •••••••• ".,,.,:a.l' ••••• , -.. ,•••• , 1, ,t.:;. . '•••• k :a... .A-:'~.' "."'.'.';;'."4" ,.':,.". ,•••"•••••" ,,-••••,to:".""., ,•••••.
.
...
,.f
. ,
'., •••,
•.
..
,..
••...
"l..
, .~"'." •••. ..
~.. " •~::::: ~:~ju
=""
~
:
:
::,
t:::::toege
pa
s
t
o
nde
r
z
o
e
k
w
a
t
ers
t
aa
t
rijkswa
t
e
r
staa
t
data
process
i
ng
d
ivi
sio
n
_-_.
- _.. -- -
-'" _-~--..:.--,-_
.L_,.~ •
I
On
the parabolle
equatiori
.
I
·
.
.
method
torwater
.
wave
'
i
·
propiigatfon
-
-
.
-
,
.
.
"
I
,
:
1
I
by
a. c. radder
I
july 1978
I
I
I
,·
I
I
I
I
), j. ,t
oegepast o
n
derzoe
k
waters
t
aa
t
I
/~_.I
I
I
I
...~;_-I
I
I
I
I
I
I
I
I
I
I
I
I Report DIV 78863 ,,!.:.":,-.~:. -~.....'..,,':._':...-r:.-On the parabolic equation method for water wave propagation.
by A.C. Radder
Ministry of Transport and Public Works, Rijkswaterstaat,
Data Processing Division, Nij_verheid$straat 1, 2288 B~ Rijswijk, The Netherlands, Tel.: (070) 906628 and 907640, telex 32683
---
--I
o.
ContentsI
I
I
I
I
I
Abstract 1 2 4 IntroductionReduction of the mxd-slope equation to the
Helmholtz-equation
Derivation of the parabolic approximation
Asymptotic analysis for the one-dimensional case
Numerical solutions for the general case Application to circular shoal
Summary and conclusions Appendix References Figures 1 - 7 5 8 13 16 19 21 23 26
I
I
I
I
I
,
I
I
Note. The present report contains the preliminary text,
submitted for publication.
I
I
I
I
I
I
I
I
1
I
I
I
I
I
I
I
-
-I
I
I
I
AbstractA.parabolic approximation to the reduced wave equation is investigated
for the propagation of periodic surface waves in shoaling water.
The approximation is derived from splitting the wave field into transmitted
and reflected components.
In the case of an area with straight and parallel bottem centourlines,
the asymptotic form of the solutien fer high frequencies is compared
with the geometrical opties appreximation.
Two numerical solution techniques are applied te the prepagatien
of an incident plane wave ever a circular shoal.
----..-_---_. ---'-" .. ---,---_. -_.-
-
-
-
-
-
-
---...-
-
-
--- .--_----
_
.
_
-
---
-
--
-
-
._------_....-2.
I
I
I
I
1. IntroductionThe propagation of periodie, small amplitude surface gravity waves
over a seabed of mild slope can he described by the solution of the
reduced wave equation
c
V'·(cc V'~) + _[ w2 ~
=
0g
c
(1)with appropriate boundary conditions. Here ~ (x,y) is the complex
two-dimensional potential function, V':: ( ~x ' ~y )
the hozä.zontia.L'gradient operator,
w
the angular frequency, and c and,I
I
I
I
Cg
are the corresponding local phase and group veloeities of thewave field. This reduced wave equation accounts for the combined
effects of refraction and diffraction, while the influences of hot tom
friction, current and wind have been neglected.
The wave equation (11 has been derived by several authors, for the
first time by Berkhoff (1972}, and by Schönfeld (1972) in a different
form.
I
I
I
Svendsen (1967).derived the equation for one horizontal dimension, as
is pointed out hy Jonsson and Brink-Kjaer (1973). Smith and Sprinks (1975)
gave a formal derivation of (11. Booij (1978] has proposed a new wave
equation, which includes the effect of a current, and which reduces to
(1} in the current-free case.
The equation (1} is essentially of elliptic type, and therefore defines
a problem which is in general properly posed only when a boundary
I
I
condition along a closed curve is given. In order to obtain a numerical
solution for short waves over a large area in the horizontal plane, a
great amount of computing time and storage is thus needed. However, in
I
I
I
I
I
I
I
I
I
t
I
I
I
I
I
I
I
I
I
propagated without appreciable reflection into a preferred direction,
and it should be natural to consider methods which make use of this
property. In the classification of Lundgren
(1976
1
,
such methods can bedistinguished as R-methods (refraction methodsL and P-methods (propagation
methodsL~ both of which represent an approximation to the mild-slope
equation
(11.
Refraction methods are based on the geometrical opties approximation,
which.fails to give a reliable solution near caustics and crossing
wave rays, where diffract~on effects become important.
Propagation methods should be able to account for such situations.
Methods of this type have been proposed by Biésel (1972), Lundgren
(lq761
and Radder(1q?7l,
but these are lacking, among other things,in the possibility of making systematic corrections which are needed
if one wants to recover the complete wave field.
In ths present work, a parabolic approximation to the reduced wave
equation (1} is derived from splitting the wave field into transmitted
and reflected components. The result is a pair of coupled equations
for the transmitted and reflected fields. By assuming that the
reflected field is negligible, a parabolic equation is obtained for the
transmitted field. Thîs procedure has been applied to opties by
Corones (1975), and to acoustics by McDaniel (19.751.The derivation
is based on the Helmholtz-equation; therefore, in § 2, a reduction of
the mild-slope equation (11 to the Helmholtz-equation is given, and
in § 3 a parabolic approximation is derived. An asymptotic form of the
solution for high frequencies is presented in § 4, in the case of an
area with straight and parallel bottom contourlines.
Finally, in § 5 and § 6 numerical solutions to the parabolie equation
are obtained in the form of two finite-difference schemes, with application
to plane wave propagation over a circular shoal with parabolic bottom
4.
I
I
I
I
I
·
I
I
I
2. Reduetion of the mild-slope eouation to the Helmholtz-equation
Although a parabolie approximation ean he direetly derived from
equation (11, it is useful, to simplify the notation and applieations,
to reduee equation (lL to the Helmholtz-equation, without 10ss of
generality.
A sealing factor is introdueed
(2)
whieh turns (11 into the Helmholtz-equation
I
I
t
I
(3)
Here the effeetive wave number kc is defined by
iJ2
lCc
k2
=
k2 _ gC ICCg
(4)
and the wave number k is the real root of the dispersion relation
I
I
I
I
I
I
I
W2=
gk tanh (kh)
(5)with h the loeal water depth and g the gravitational aeeeleration.
W dW
The phase and group veloeities are then given by C
=
k ' Cg=
dkIn shallow water, the difference k2 -k2 may become appreeiable:
C in this case one has
, C
I
ThI
~
4h
(6)
I
I
I
I
It follows, that kC may be approximated by k if
(7a)
(7b)
implying a slowly varying depth and a small bottom slope, or high
I
I
I
I
I
I
,
I
I
I
I
t
I
I
frequency wave propagation.
Unless stated otherwise, k
c
assuming that {7al and (7bl are satisfied.
will be approximated by k in this paper,
3. Derivation of the parabolic approximation
The Helmholtz-equation (3) can be written in the form
(8)
where x denotes the preferred direction of propagation, and the
sub-script c of the wavenumber k has been dropped (however, for the
derivation of the parabolic approximation the restrictions (7) are
not necessaryl.
+
The wave field ct> should be split into a transmitted field ct> and a
reflected field ct>
ct>
=
+
ct> + (9)
This can be achieved by the use of a splitting matrix T which defines
the transmitted and reflected components by
6.
,
I
I
I
I
I
I
t
I
.
,
I
The matrix T is formally arbitrary, but some genera 1 physical criteria
limit the choice of T and lead to the governing parabolic equation in
a natural way.
Firstly, equation (9) is valid for arbitrarily chosen
ep
+andep
-
onlyif T satisfies
a
+ y=
1B
+0
=
0
(11)
Using equation (81, it follows that
a<p
+ _[-
k2B
+ay
+aa
y aB
a2 ]
<p
+ +
d.X -B
ax
+ ---B ax
B-ay2
[- k2B
a2
aa
a aB
a2 ]
ep
(12a) + - - +ax
- î3
ax -
B-B
ay2
aep
[k2B
y2
+ay
y ~
+B~
]
ep
+ +
ax-
= + -B
ax
-
B ax
ay2
+[k2B
-
;:;-ay
+.."..-ay
+-a
.."..-+B-
as
a2l
eb (12b) p áXB
ex
ay2
J
.
I
Further, when k is a constant, solutions of the formI
I
I
n,+
ikx
'I'
=
e=
e-ikx
(13)should result, and the equations (12a) and (12b).should naturally
decouple in this·case. This can be achieved by choosing
+
o
+ (14)I
I
I
I
I
and th.e resulting splitting matrix is (cf. Corones (1975Il
I
I
I
while (121 reduces to <p+ +(.!_
2k (16a)I
I
I
t
I
I
Î
I
( " 1 ak i (2)_ -lk- 2k dx - 2k dy2 <p (16b)This pair of coupled equations is equ~valent to equation
(Sl.
Bye
neglecting the reflected field <p-, a parabolic equation for the
transmitted field <p+ is obtained
(17)
In a similar way, a parabolie approximation can be directly derived
~+
from equation (1), which yields ior the transmitted field ~
[ ik - -=2~k-!-e-g + 2keei
a
ay
eegay
al
4>+g .
I
I
I
Using (2t and (71, equation (17} is recovered.
I
I
I
Î
a
2By adding to the left hand sides of (14)_ me operator
S
(
)
dy2
th 1"· " d " d T ' 1 -i/A
ano er sp 1tting matrl.XlS er1ve: A
=
2 1 i/AA
=/k
2 +1,
and a closer approx~ation to equation (SLmay beay2
obtained. Unfortunately, the square root operator A makes the resulting
, where
parabolic equation practically untractable, and a satisfactory approximation
must be found for the operator A, in order to obtain numerical results
(cf. McDaniel (1975)).
In the following, the parabolic equation (17) will be considered, in
8.
I
I
I
I
a
'
t
t
I
Î
I
I
I
I
I
I
I
I
t
,
I
'
direction of the incident plane wave.
4. Asymptotic analysis for the one-dimensional case
In order to test the validi ty of the parabolie equa tion (171 as an
approximation to equation (SL solutions to both. equations will be
compared in the case of an area with straight and parallel bottom
contourlines. The problem is equivalent to pláne wave propagation in a
plane stratified medium in opties and acoustics, and the asymptotic
analysis of Seckier and Keiler (19591 will he followed here.
Dropping the
+
superscript, equation (171 can be written as2
~ on
0
k
ddX
<j> + (18)where ko denotes a constant wave number, and n
=
k/ko the indexof refraction. By introducing a new coordinate system
(p,cr)
p
=
x
cosa
+ ysina
(19)cr
=
-
x
sina
+ ycos
a
the bottom is defined through h == h (p),
n
==ne
p
);
dn
dy
=
ÀdX
dn À=
tana
(20)TI
wnere
a
is the angle of incidence, withl
al<
2 .
It will be assumed that k(P} tends to the constant value ~ (i.e. n(P) -+ 1}
I
I
I
I
*
I
t
I
ikoxNow suppose a plane wave e
is incident from x
=
_00 • The field <p can then be written in t.."1eform
<p = A (p) exp [iko (x - pfndp)] (21)
with P =
cosa
sin
2a
A satisfies
Upon inserting (21) into (18). one finds that
(22)
At P
= -
00, A ( pL is supposed to behave like·
1
I
I
I
I
I
1
I
I
I
I
I
(23)where the constant R denotes a complex reflection-coefficient.
At P =+ ·00,A(pl should satisfy aradiation condition, i.e. no
incoming wave from +00 •
The equation (221 is in general not explicitly solvabie, and the
solution must be represented by an approximation, which usually takes
on an asymptotic form for high frequencies, in the limit ko+ 00
A point at which the coefficient of A in (221 vanishes is cal led a turning point, where the character of the solution changes from
oscillatory to exponential. In the geometrical opties approximation
of tne problem, a caustic line is formed at these turning points.
If tnere is no turning point, and n + 2 À2(n
-1
»
0,the asymptotic form has an oscillatory character with R=O,
and can be found by the WKB-method (Cf. Langer (1937}). Let
-
~
A
=
In2 +2À
2n(n-
1
)
Ip (24)
10..
I
I
then ~~e WKB-approximation to ~ is given byI
I
I
I
I
I
A
ikoF
~p=
pep (26)A similar analysis for equation
(al
results in the geometrical optiesapproximation. Let (27) (28)
1
I
I
I
I
I
1
then the asymptotic form is given by
A
ikoF
~g
=
ge
g (29)In the special case
a.
=À =0,
both ~ pand ~gagree (if the sealingfactor~giS taken into account) with the classical shoaling
formula for a progressive wave
ct> :::: _1_
exp [ifx
k dx] .g
IC
Xog
(30)
~p and ~gare compared in table I, for some values of À and n.
I
I
I
I
It is assumed, that the incident wave is starting in deep water,
ko=
wJg, and a correction factor cA=
-;===n====~~_/1 + k0h(n2-1)
should be applied for the wave amplitudes, according te equation (2).
I
I
I
I
I
I
I
I
,
I
I
I
I
I
I
I
I
I
,
I
I
Table
IComparison of
w
ave amplitudes
A, w
a
v
e
num
bel's
I
V'F
I
and
wave directions
e
(
no turning point
)
.
c
A
* A
p
c:
A
*' A
g
l
V'Fp
l
I
V'F
gI
e
e
P
g0.88
0.91
2.01
2
24.
4
°
24.3°
1.01
1.07
3.03
3
,
31.5°
31.4°
0.70
0.74
2.04
2
37.4°
36.9°
0.79
0.85
3.12
3
46.8°
46.1°
-, cr. Àn
4
5°
1 2.45°
13
63.4°
2
2
-
63.4°
2
3
The
'
agreement is rather close, even for comparatively
l
arge values
of
À •Now suppose there is just one turning point at
P=
P
o ,
a point where
the coeff
i
cient of A
(
P) in(221 vanishes.
This
willoccur when n takes on the value n
p
n
=
P
1+2.À
2(31)
In
case of equation
(8},the corresponding value is given by
ng
(
3
2)
An
analysis of the turning
p
oint problem can be
f
ound
i
n
th
e artic
l
e
of Langer
(19.37
1
:
for
P
>Po
A(
p
l
takes on an exponentiall
y
decreasing
f
orm
12.
I
1
I
I
I
I
I
and for P < Po an oscillatory form
A(p)
=
2
I
Q
I
-~
cos[
f~
O
Q
dp - ~ ]where
Q
=
kopIln
2 + 2À2n(n
-1)I
Near the turninq poi.nt, the asymptotic form of the solution can be
(
33
b)
(
34
)
represented by Airy functions.
Uponinsertinq (331 into (211 one obtai.ns the asymptotic form
4>
p.
Here, only the behaviour of
4>p
at - 00 will oe qiven explicitly(35) .
I
·
1
.
I
I
I
I
I
1
I
I
.
1
·
I
I
wi.th.
I
Rpi =
1, i. e-:a fully reflected plane wave arises, with. wavenumber
kp and
wave directiona. p
qi ven byk
=
.
ko11
+ 4·/À
1+P
,
(36)
For the qeometrical opties solution, the correspondinq formulas
are qiven by
4>
(_00)=
exp[ikox]
+R
exp
[
iko(x(À
2-1) -
2
Y
À)j
(I+
À
2)]g g (37)
k
=
k
o
g 2Àtan
a.
g
= ~À1-For somevalues of À , a comparison is presented in table
Ir.
--~--- --- - - ---_-_-
---Table 11
Comparison of reflected plane waves at a turning po
i
n
t
.
n p n
g
k
P/
ko
kg/koa.
p 2 10.89
0.67
0.89
0.71
1.12.2
I
I
I
I
When À. approaches zero, the reflected waves in the solutions
<l>
p
and<P
g
deviate more and more from each other, as would be expected.
Actually, the parabolic approximation is valid provided À.2
»
1otherwise the coupling between the transmitted and reflected wave
I
I
fields in equations (161 must be taken into account, if one wants to
recover the complete wave field. For systematic corrections to the
parabolic approximation, see Corones (1975}.
I
I
I
5. Numerical solutions for the general case
The parabolic equation (171-may be solved by using finite-difference
I
,
I
techniques. In this section, two alternatives will be dealt with.
Assuming plane wave incidence
ikox
e
(38)I
then equation (181 yields for the complex potential function ~I
I
I
(39) where f=
k~n [2(n _ 1)
+ iko
d
~
n(n
)
J
dX (40)A Crank-Nicholson finite-difference equation is used for the numerical
I
I
I
I
I
solution to equation (39), cf. Richtmyer and Morton (19671:
let a rectangular grid be given with grid spacings ~x and 8y, and
be denoted by ~~ J
J
let the approximation to
~(~~x,
j~y)~,j=O,1,2, .•• The scheme I is then defined by
~~
+
1
~
+
1
+ ~ + ~~ +[-2
+
(~y)2f
~
+~] .
(~~
+ 1
+ ~~ + ~. 1 ~. 1 )+J+
1
J-
J+
J-
1
J
J
J
+4 ik
(
~
y) 2~
+~
.(~~+l-
~~
)=
0
o ~xn.
J
J
J
(41)I
I
.14. whereI
I
I
I
I
i+~n.
J i+l i=
(n. + n. )/2 ,
J J i+l.
n.
f
in(+
)/
Ax] o n. Jand with initial condition (42)
~~ =
1,
j=
0,1,2,•..
J (43)
and appropriate boundary conditions, to be specified later on.
Another solution technique, which may be preferabIe, is based
on the change of variabIe
I
I
(44)
which turns (391 into
I
I
a
2çaç.
3ç21'k
n
3ç+
f=
03y2
+3y
3y
+ 03x
It
may
be expected that the solutionç
is a less rapidly(45)
I
varying function than ~, thus providing a more accurate appraximationon the same grid.
I
I
I
However, the transformation (44)_ is singular at points where
'1'=0(branch-points, or: àmphidromic points), and a direct
application of a scheme like (41l is not possible. In order to
prevent the non-linear instabilities involved, it appears to
be useful to add to the left hand side of (451 an artificial
I
I
viscosity term of the form
I
I
I
(46)
where
S
is a dimensionless constant of the order of 1I
I
shocks, where an analogous dissipative term has been introducedI
I
I
I
to insure stabiIitYi see Richtmyer and Morton (1967), chapter 12}.
Let
(47)
then the scheme 11is defined by
I
I
I
+
R.+l [_ 2 R.+
8 i.k SË1]_2 R.+~]+
2 (R.+
rR.. )+
l;;j' gj 1 0 Sx. nj l;;j+l "'J-l (48)I
I
with initial condition
l;;~
=
0
J j
=
0,1,2, ...
(49)I
and appropriate boundary conditions.I
I
I
Both schemes being implicit, a system of simultaneous linear
equations has to be solved. For systems like (41} or (481 very
efficient methods are available.
The rate of convergence wiII be exemplified in the next section,
where numerical solutions are obtained for the case of a circular
I
I
snoal.I
I
I
I
I
I
I
I
I
..
I
I
I
I
I
I
I
I
I
I
16.
6. Application to circular shoal
As an example, the propagation of an incident plane wave will
be considered over a circular symmetrie shoal with parabolie
bottom profile. Calculations for this severe test case have been
made by Berkhoff (19761., Bettess and Zienkiewicz (19771,
Flokstra and Berkhoff (19.77)_, and Ito and Tanimoto (19721, who
additionally conducted some laboratory experiments.
The shoal is represented by the ~epth profile
for r<R ,
(50) whereh
=
ho
r
2=
(x _
X )2 + (y _ meo
=
-
(ho - h )/R
2 m andTo be definite, short wave propagation is considered, and the
assumptions (71 should apply, which amount to:
(51)
This imp lies that the curvature of the bottom is much less
than the wave number, regardless of the value of the minimum depth.
I
I
I
I
The value of the angular frequency
w
follows from the dispersionrelation (51:
w
2=
gko tanh (koho)
i denoting the correspondingwavelength by
Lo
=
2IT/ka , the problem is then defined throughthe parameters hm
IR ,
ho/R
and Lo/R.
In order to specify the boundary conditions, it is useful to
analyse the asymptotic character of the solution for large distanee x
(see appendixl. The governing equation stands for the
I
I
I
I
I
I
1
I
I
I
I
I
I
I
I
I
I
I
I
I
wave packet. The behaviour of this wave packet for large x is a
well known problem in wave mechanics: the spreading of the packet
increases linearly with the distance x, and the magnitude approaches
zero, as
11
IX .
It follows, that the required boundary conditions for schemes I
and II, in case of a shoal, can be given by the undisturbed initial
values of the solution, ~=1 and ç=O, provided these boundaries
are taken sufficiently far away from the area of interest.
In this way, the artificial reflections which may occur at the
boundaries, can be avoided.
Some calculations with the numerical schemes have been performed,
for two configurations of the shoal:
- configuration I, defined through:
hm/R = 0
.
0625
;ho/R = 0.1875
;Lo/R = 0.5
)
- configuration II, defined througn:
hm/R = 0
.
016
;ho/R = 0.116
jLo/R
=
0
.
288
The parameter eog/w2 takes on the value 0.01 for configuration I,
and the value 0.005 for configuration II, so the inequality (51]
is valid in both cases.
The constant
8
in (461 is chosen to be i, and the grid spacingshave been varied according to
~y/~ = !
J~x/Lo= 1,!,~
and1/8 .
Configuration I has been studied by Ito and Tanimoto (1972), who
use a finite difference timestep method, and by Flokstra and
Berkhoff (19.771.,who use a finite element elliptic methode Table III
demonstrates the agreement between the various methods for the
I
I
I
I
I
I
I
-
,
-
,
\I
I
! ; .---...--
-I
I
I
I
I
I
I
I
I
I
18.
rate of convergence of ~~e numerical schemes I and II.
A detailed view of the solution is given in figures 1-5, which show
a comparison of wave amplitudes between the mentioned methods,
contourlines of the amplitude and of. the phase, and energy flux
lines (waveorthogonalsl. A grid is used with 281 ~ 449 grid points;
the centre of the shoal is located at xm=33, Y =113(in grid units),
m
,
and the radius of the shoal is R=16 ~x.
-Energy flux lines are defined throuqh the energy streamfunction G:
~
====:====================
====-
-=-=-
=-
-
~=-~
=-
--=
-
-
-
_
--
-
-
_
--=----
-
=
-
_
-.
---(52)
---,.---
----
-
---
-
-
----
-
-
-
-
._---_. ---- - --- .Table 111
Comparison of maximum wave amplitudes for
a circu1ar shoa1.
~y/t.x
=
1/2
~x/La
1Scheme
I1/2
1/4-1/8
1Scheme
11
1/2
(a=
1)1/4-1/8
Ito and Tanimoto (1972)
Flokstra and Berkhoff
(1977)
Bettess and Zienkiewicz
(1977)
Configuration
IMaximum
location
Ax/La
2.17
8.5
2.08
7.0
2.05
6.8
2.05
6.6
2.10
7.0
2.03
6.5
2.04
6 •
6
.
2.05
6.6
2.1
6.3
2
.
04
6.4
Configuration
IIMaximum
location
Ax/La
2.57
9.0
3.18
7.0
3.01
6.0
2.96
5.7
2.85
6.0
2.96
5.8
2.92
5.7
2.97
5.7
3.1
5.7
2.9
5.5
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
-where amplitude A and phase F are given by ~
=
AeiF.(If the field ~ satisfies the HeLmholtz-equation (8l, it
follows that
s r .
'ilG=O,i.e. orthogonality of F and G, whichprovides another test of validity for the parabolic approximation).
Configuration 11 has been studied by Flokstra and Berkhoff (1977},
and Bettess and Zienkiewicz (1977}, using a finite element elliptic
methode In figures 6 and 7, the relative wave amplitudes on the
line -of symmetryI y=y m ' are presented.
It appears that the minimum near the rear end of the shoal cannot
be represented properly by the solution of scheme 11. This is
caused by the occurrence of branchpoints, for which
A=O.
In the vicinity of such points, the phase is a multiple valued
function, and the energy flux lines are closed. So, the application
of scheme II then results in a smoothed so~ution, which has better
convergence properties, and which is preferabie when the accuracy
requirements are not too high.
7. Summary and conclusions
For the propagation of periodic surface waves in shoaling water,
a parabolic wave equation (18) has been derived, based on the splitting
technique of Corones (1975). This method yields a pair of coupled
equations for the transmitted and reflected fields, and the
parabolic equation results from neglecting the reflected field.
In the case of an area with straight and parallel bottom
contour-lines, the asymptotic form of the solution for high frequencies
is compared with the geometrical opties approximation. There is
20:
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
In the presence of a caustic, there is a reasonable agreement
provided the angle of incidence is close enough to 90°.
Otherwise the coupling hetween the transmitted and reflected
wave fields cannot be neglected, and systematic corrections
snould be applied, if one wants to recover the complete wave field.
Finally, two numerical solution techniques are presented in the
form of finite difference schemes, each based on a different
form of the parabolie equation. As an example, wave propagation
over a circular shoal is considered, where the geometrical opties
approximation prediets a cusped caustic line. For two hottom
configurations, .the results are compared with similar calculations
in literature, showing a reasonable agreement. Which solution
technique is preferabie depends upon the required accuracy and
the available computer capacity.
The parabolic equation method may be applied to short wave
propa-gation in large coastal areas of complex bottom topography.
Acknowledgement
The author wishes to thank ir. W.F. Volker of the SWA-department,
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Appendix Motion and spreading of a wave packet
The asymptotic character of the wave field bebind a shoal
of finite extension, in water of constant depth, is governed
in tlie parabolic approximation by equation (3Ql, with n=l.
Let '1'
=
I + e: Xl=
kox
y'=
koY
i then thedis-turbance e: satisfies the Schrödinger-equation (omitting'l:
2 · de: _ 0~
-_
dX (Al)
Assuming E (and derivativesL sufficiently square integrable,
the following quantities are defined:
- the norm of the wave packet
(A2)
- the norm of the derivative
(A3)
- the mean position of the packet
<y> = -
N
If
00_00 yl E12 dy (A4)- the mean velocity d
<v>
=
dx <y> (AS)- the spreading
S
=
/«y _ <y» 2>=
l<y2> ._ <y>2 (A6)22 •.
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
and < v> are constants of the motion:
dN
=
0dx-
,
dMdx -=
0d
dx
<v> - 0 CA7)For the spreadinq S, one f~ds
(AB)
Inteqratinq
(Aal
twice, one o.bta~s the result, that for large x, S'increases linearly:s
(x »co) ::: /~ - <V>2 •x (A9)It follows then from (A21and (A9.L, that the magnj.tude
[e
I
of the wave packet approaches zeroI as1/
IX
.
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
ReferencesBerkhoff, J.C.W., 1972 Computation of combined refraction-diffraction,
Proc. 13 th Coastal Engrg. Conf., Vancouver,
ASCE 1, pp. 471-490, 1973.
Berkhoff, J.C.W., 1976 Mathematical models for simple harmonie linear
water waves, Report on mathematical investigation, Delft
Hydraulics Laboratory, Report W 154-IV •
Bettess, P. and Zienkiewicz, O.C., 1977, Diffraction and Refraction of
surface waves using Finite and Infinite Elements,
Int. J. Num. Metli.Engng. 11(8)_,1271-1290 •
Biésel, F., 1~72, Réfraction de la houle avec diffraction modérée,
Proc. 13 tfi Coastal Engrg. Conf., Vancouver, ASCE 1,
pp. 491-501, 1973.
Booij, N. 1978 A proposed equation for refraction and diffraction of
water waves under the cambined influence of current and depth,
Manuscript, Delft University of Technology •
Corones, J., 1~75, Bremmer Series that Correct Parabolie Approximations,
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
24.
Flokstra, C. and Berkhoff, J.C.W., 1977, Propagation of short waves over a circular shoal, Report W154-V (in Dutchl, Delft Hydraulics
Laboratory.
Ito, Y. and Tanimoto, K., 1~72, A method of numerical analysis of wave propagation - Application to wave diffraction and refraction,
Proc. 13 th. Coastal Engrg. Conf., Vancouver, ASCE 1, pp. 503-522, 1973
Jonsson, I.G. and Brink-Kjaer, 0., 1973, A comparison between two reduced wave equations for gradually varying depth, Inst.
Hydrodyn. and Hydraulic Engrg., Tech. Univ. Denmark, Progr. Rep. 31, pp. 13-18.
Langer, R.E., 1~37, On the Connection Formulas and the Solutions of the Wave Equation, Physical Review 51, 669-676.
Lundgren, H., 1976, Physics and Mathematics of Waves in Coastal Zones, Proc. 15 th. Coastal Engrg. Conf. Hawaii, ASCE, pp. 880-885.
McDaniel, S.T., 1975, Parabolic approximations for underwater sound propagation, J. Acoust. Soc. Am. 58, 1178-1185.
Radder, A.C., 1977, Propagation of short gravity waves in shoaling water, Refraction and diffraction, parts I and II, Rijkswaterstaat,
I
I
I
I
I
.
1
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Richtmyer, R.D. and Morton, K.W., 1967, Difference methods for
initial-value problems, Intersc. PubI., New York.
Schönfeld, J.Ch., 1972, Propagation of two-dimensional short waves, Manuscript (in Dutchl, Delft University of Technology •
Seckler, B.D., and Ke lIer, J.B., 1959., Asymptotic Theory of Diffraction
in Inhomogeneous Media,
J.
Acoust. Soc.Am~,
31, 206-216.Smith, R. and Sprinks, T., 1975, Scattering of surface waves by a
conical island,
J.
Fluid Mech. 72, 373-384.SVendsen, I.A., 1967, The wave equation for gravity waves in water of
gradually varying depth, Coastal Engrg.·Lab. and Hydraulic Lab. Tecn. Univ. Denmark, Basic Res.-Progr. Rep. 15, pp.
2-1.
I
I
I
I
I
I
I
I
I
I
I
I
Fig. 1I
I
I
I
I
26 •.Y!Lo-4
2 3 4 ~ e 7 ij 9 10 :----X/LOx/Lo;'7
:~
.
I 234 ~ 8 Ti
~y/LoX/Lo ...6
~--~~---r--~~--~--~
234 ~ ij 7 8 _Y/Lo._
-
-
-
-
--
---
-
-
--
--
_
.
_-
.._._.comparison of relative wave amplitudes for bottem
confiquration I, between results of the schemes I and II (continuous curves,
6x/Lo=1/8 )
and results of Ito and Tanimoto (19121.and of Flokstra and Berkhoff (19771I
I
I
I
I
10 2030 40 50 60 70 80 90100110120130140150160170180190200210220230240250260270280I
220 220 210 210 .'200 20!) 190 190 180 180 170' 170 ISO 160 ISO 150 .140. 140 '.ISO 130 120 120 lllt 110 100. 100 I 90 90. I i 80 80 70 70sa
se
so
:ij) 40' 40 30 30 20 ZO 10 ;0I
I
I
I
I
I
I
10 20 30 40 50 60 70 80 90100110120130140.501601701801902002102Z023024Q250260270ZS0I
I
Fig. 2Contourlines of the amplitude for confiquration
I,
I
_---_.accordinq to scheme I
-- --U
we
--/
-Lo
---=
--1/8).
-- --- ----I
I
I
I
I
·28.I
I
I
I
I
.190ISO 170 . ISO -ISO 140 130 120: 110 100 80 611 SO 40 30 -20 10 70 60 50 40 30I
I
I
I
I
I
:0
10I
I I I-
r
I I I 11''''\' 10 20 30 40 50 60 70 80 90100110120130140150160170180190200210220230240250250270250I
--_.._
-
---
---
-
---._._---- - -_...__ ._---- --- ...I
Fig. 3Contourlines of tne amplitude for confiquration I,
I
accordinq to scheme
Ir.(
~
/
Lo
=
1
/
8
).
I
I
I
I
I
I
I
I
I
-
-
-
---
-
-
-
-
-
-
-
-
--_
..---_. --_.._-_.- ._ -JQ 20 30 40'50 6~ 70 80 901001101201301401501601701801902002102%0230240250260270280 180 170 160 ISO 140 130 t20 110 100 90 80 70 ·60 50 4(] 30~t
L
.
10 ; , , ::,
_:::
t:
10.
".l"'t'iE!î'iiifi!ii4!jjr"""I''1ta
, I'
lóliiiil,ifi
'
f9
'
:?j~,
.
,.;a''"
"
iiAi
'
'''''i''',r ..
e.t:r,e.S"".-,',••4,j,f,••Jl'ii:",,,,~,,••.,,,.<.- ••~l~,,u,.ï:r.~,....r
I I l I. I -r--, IJ' I I I I I ,. t I(i I I ! I I I I I, 1 I I I 10 2030 40 60 60 70 ~O 90l001101201~0,~OlSOlS017Cleo:a020ü210(202~GZ4û2öOZ5Ct70zaO
I
220 I 210 ZOO 190 180.
170. lea: ISOua
130 IN 110 100 : se!
sa 70: i 50 50 4C 30I
I
I
I
I
I
I
-
I
, ,f f , , , , I I, f I , l , I I I , II I..I_•L•L.I. , IJ J fL,U.' I I II '_J.J. ' I I I I,' II I I, I.f~L---::
~--
--
E
--
:t ....;l
-
~-
~-1~-j
-
-,
-'t
-
-..
-
_....
-
~-
,
,-·d
-....
....; ·o.-,oE-
=s
~~--4
...~tr ~.-
""t: "1. ~-::-I
I
I
I
I
I
Fiq:- 4
Enerqy flux lines for cenfiquratien
!,accerdinq te
scheme
r
(Ax/Lo.::. 118) .29.
1.20 210 200 190I
I
I
I
I
I
I
I
I
.
1
I
I
I
-
·
·
I
~~
.r.
I
I
·
I
I
I
I
30.
---
-
-
--
---
--
-
_:__
---
--
-
--
_
.._---_
-Fig.
·
5Contour~ines of the phase for configuration I, according
toscheme
Ir
Cillc/Lo =
1/8).-I
31.I
.
.
.
.
.
10 15--
--,-
---_
.._
----
,
--
-
--
-
-25 X/L4 20 --- ---_.Fig. 6 Comparisonof re~ative waveamplitudes for configuration II,
I
I
on the.line of symmetry,between resul ts of schemeI
(continuous curves}and resul.ts of Flokstra and Berkhoff
(lS77L(circlesl
I
I
I
I
32.I
I
I
3.00t
2.60 .t
220 :~ -00o
--.---.25 --- X/LoI
I
I
I
I
I
--- ---~---_._---_. --Fig. 7 Comparison of relative wave amplitudes for configuration
::rI,on the li.neof symmetry, between resu~ts of scheme TI
(continuous curvesL and results of Flokstra and