GEOPHYSICA
POLONICA 1987
Vot. XXXV, no. 1
@
APPLICABILITY OF DIFFUSION ANALOGY IN FLOOD ROUTING
James C. L DOOGE
Civil Engineering Department, University College Dublin EarIsfort Terrace, Dublin 2, Ireland
y
Jaroslaw J. NAPIÓRKOWSKI
Institute of Geophysics, Polish Academy of Sciences 00973 Warsaw, P. a. BoK 155, Pasteura 3
Abstract
A number of forms of the diffusion analogy approximation to the linearised St. Venant equation for Hood routing in channels are compared. It is suggested that the diffusion analogy based on the kinema.tic wave approximation of certain terms be used. The upper limits for the dimensionless wave number are given for certain prescribed levels of error in the diffusion analogy predictions for the phase velocity and for the attenuation per unit length. 
1. INTRODUCTION
1
.
, by a diffusion analagy method
The degree of approximatian involv~d in replacing the camplete S1. Venant equations
ar by a kinematic wave method bas been discussed by a : ~ber of authors. Same of these authars have approached the subject through harmonie ana.ysis. This caD be dane either 'by means af a frequency analysis"of the linearised S1.Veflaut equations with harmonie boundary canditions (Grijsen and Vreugdenhil, 1976; Vreugdenhil, 1972, 1977) or by wave number analysis of the linearised equations (Menendez and Norscini, 1982; Ponce and Simon s, 1977; Pance et al., 1978). The present fiole commen~s brieflyon thiee points:ll) the existence of mate than one form of the djffusion analogy; (2) the selection af that form of diffusion analagy which approximates most closely to the complete St. Venant equatians; (3) the possibility af formulating a simple criterion for the lange of applicability of ibis model af the diffusian analogy. The discussion will, for the sake of brevity, cancentrate on the special case of a wide rectangular channel with Chezy friction and on wave number analysis.
5 AGP 1/87
I
66
J. C. I. DOOGE, J. J. NAPIORKOWSKI
[2)
I
2. COMPLETE LINEAR EQUATIONS
r l
The linearised St. Venant equations for onedimensional unsteady Bow in a broad rectangu1ar channel with Chezy friction may be written as (Deymie, 1935; Ponce and Simons, 1977):
oy oy
+vo+yo
ot
ox
=0, ox
ov
(t)
oy+ Vo~+~
oX g ox
g at
ov=So ~
(Yo Vo),
2V
(2)
" .
where y is the perturbation indepth flam the reference depth Yo, v is the perturbation i~ velocity flam the reference velocity vo, g is the acceleration due to gravity, So is the bottom slope ofthe channel, t is the elapsed time, and x is the distance along the diannel. By e1iminating v(x, t) flam equations (1) and (2) we caD obtain a single second ordei",!t'
equations in leX, t) given by: 02y 2
(gyovo)
02y oy 2gSo oy 2vo =3gSo+' OX2 ox at ox Vo at
(3)
1 ~
The aboveequation caD be expressed convenient1y in dimensionless form (Woolhiser and Liggett," 1967) byusing the bottom slope of the channel (So) and the depth and velocity (Yo, Do)of the steady uniform reference condition about which perturbations ale taken. Thus we caD write , y (4) y =" Yo x x,'=, Yo So "(S)
, . .
I
t'=
t
Yo SoDo
(6)0;;+,
and transform equation (3) to 2 o'2y' 2 02y' ,2 02y' al' al' O:S(tF o) uX '2 Fo,; at O.SJi o at ,z = l.S~uX +,;;, ~ ox ut where Fo=vo/Jgyo is the Froude number for the reference Bow condition.
3. VARIETY OF DIFFUSION ANALOGY MODELS
,J: '",
(7)
Inthe case of maDYriver channe1s, the second and third terms in equatión (2) ale or an order of magnitude smaller than the other three termsIn the equation (Cunge et al., 1980; Hender.son, 1966; Kuchment, 1972). If these twa smaller terms ale neglected so that
.,.
r

r
(3]
DIFFUSION
ANALOGY IN FLOOD ROUn:NG
67
equation (2) becomes OJ =SO
ax
(~2~ )
Yo Vo
(8)
and v(x, t) is naw eliminated between equation (1) and equation (8) we obtain,
+1.5vo'=at ax 2So ax2
ay
ay
VoYo a2y
(9)
instead of equation (3). The approximation represented by equation (9)is of the same form as the convectivediffusionequation ay ay +c=Dat ax a2y ax2 (10)
,
t and consequent1y known solutions of the latter equation caD be applied to flood routing~. (Hayami, 1951; Schonfeld, 1948). For this approximation based on neglecting terms in .(requation (2) (Ponce and Simons,. 1977; Ponce et al., 1978) the parameters of the convectivediffusion equation are related to the parameters of the channel and the reference flow conditions by c= 1.5vo (11)
which is the kinematic wiwe speed and
D = Vo,Yo 2So
(12)
which is equivalent hydraulic diffusivity of the channel. If, alternatively, we examine the dimensionless form of the equation, a somewhat different appn;>ximation is suggested. As, the Froude number approaches zero, then the \ second and third term in equation (7) obviously tend to zero and the same must be true or the corresponding terms in equation (3). If these terms are neglected then we algo have a convectivediffusion equation. In this.case the convective parameter cis again given by y equation (11) but the equivalent channel diffusivity is
~
,.
~
'
.D=(lF~)
VoYo
(13)
2So! ,
which will differ from that given by the fiISt approximation particularly for higher Froude
numbers.
"
Vet another form of diffusion analogy caD be derived ifinstead ofneglecting smallterms entire1y, they are represented by one of other of the surviving terms on the basi s of the kinematic wave approximation (Dooge and Harley, 1967). For the kinemati~ wave approximation (Lighthill and Whitham, 1955) we caD write the solution for the perturbation in depth as: , ~=f(x1.5vo Yo t). (14)
.
\

..
68
J. C. l. DOOGE, J. J. NAPIÓRKOWSKI
[4]
Insertion of this solution in equation (1) reveals that to satisfy continuity we JllUst write v =0.5f(xl.5vo Vo . t) (15)
where the same function caD be used since there is by definition no phase shift for the kinematic wavesolution (Menendez and Norscifii, 1982). ThisJower order solution caD be used to approx.imate either the terms neglected in equation (2) or the terms neglected in equation (3), Ifwe wish to approximate the se~ond term in equation (2) on the basis of the kinematic appro,omation then we write
.
Vo av v~  =f
,
g ax
2g
(x1.5vot)=0.5
.
v~ ay
(16)
gY ax
~
(17)
~imilar1y we write'the third term as 1 av  =
g
1.5f
at
v~ , (x1.5vot)= 2g'
0.75
v~ ay .
r"
gyoax
lnserting these twa approximations juto equation (2) we obtain
" . If v (x, t) is eliminated
"'l'4 (l_F~ )ay=80 (
ax
between equations ay +1.5v ay
2V
,
_(18)
Yo vo. )
(1) and (18) we get
at
'
ax
=
( 4)
'
1 F~ VOYO a2y
2So ax2
(19)
which lead to a value of
voYo D= I F~ 4 280
( )
2
(20)

for the equivalentchanne1diffusivity. If alternativelywe wish to make the appr~ximation in equation (3) then we write the} ~ secondterm as 2vo=3vo ~at
and the, third term as
'ay
2 .
f( x1.5vot)=3voz ~
2 a2y
"
2ay
2
(21)
~
~
...
 y=2.25vo at
a2y
2 "
f
(xl.5vot)=2,25vO"2' ax
(22)
Substitution of these approximations in equation (3) algo gives the convective diffusion equation in the form of equation (19),
v
r
[5], DIFFUSION ANALOGY IN FLOOD ROUTING
69
l
4. CHOICE OF FORM OF DIFFUSION ANALOGY
The' three forms of the diffusion analogy discussed in the last section a11agree in predicting the convective parameter as
, e=1.5vo
(11)
which is the kinematic wave velocity. Accordingly a11of lbem will predict {he fust moment or lag of the linear channel response (LCR) as x
U'l(LCR) = 1.5vo
(23)
l
which is identical to the value for the complete linearised equation (Dooge and Harley ,.. 1967). They differ however in their prediction ofthe equivalent channel aiffusivity,D excep . for the limiting case of F=O. Accordingly the question arises ofwhether aDYone particula form ofdiffusion analogy approximation caD be shown to be 'preferabIe to the others. .l. On general grounds one could expect. that the model s based on the approximations of terms through the kinematic wave approxim~tion would be preferable to those based on complete neglect of these terms. These general considerations are reinforced by comparing SOfie properties of equation (19) with other knowa resuIts in open channel hydraulics. Firstly, the model represented by equation (19) unlike the other two modeIs, indicates that diffusivity will become negative (Le. disturbances will amplify) for Fo> 2. Secondly, the second moment about the centre of the area of the soIution of 'the complete St. Venan't equation represented by equation (3) for a delta function input is given by Dooge and Harley (1967):
X 2 2 Yo u2=(10.25Fo)3 Sox ( 1.5vo )
2
(24)
and for the diffusion analogy represented by equation (10) is given by ~
U2
= 2D ~
. ex
(e )2.
(25)
Using the value of c erom equation (11) which is common to a11forms of the diffusion ana
~.
logy and equating these two values/or the second moment we get'
D = (1 0.25F~) VoYo 2So
.
(26)
'"
10
which is the:value already obtained by using the kinematic wave solution to approximate terms in equations (2) or (3). tt is suggested that aDYdiscussion of the range of applicability of the diffusion analogy should be confined to this form ofthe analogy which reproduces exactly the fiest and second moments of the compIete linear solution. In the final section of this note, the wave analysi~ method used by Ponce and Simon s (1977) will be applied to the question ofthe rangI\:of applicability of this particular model.
70
J. C. I. DOOGE, J. J. NAPIÓRKOWSKl
[6]
5. WAVE NUMBER ANAL YSIS OF DIFFUSION ANALOGY
The solution for a harmonie perturbation in depth of either the complete equation or or the diffusion analogy approximation caD be sought in the form
Y= Ymexp[i(uxpt)],
solution caD be written,in terms of a dimensionless wave num ber
U=,U SO
I
(27)
whereu is the real wavenumber and p is the complex propagation factor. Alternatively ibis
Yo
(28)
and of a dimensionless propagation factor P' given by
p'=so that the solution caD be written as
Yo
Sovo
P
(29)~
.....
..
.t
:,
Y= Ymexp [i(UIX' P't')].

(30)
Either the subs~itution of equation (27) juto equation (3) and the use of equations (28) and (29) or the substitution of equation (30) juto equation (7) gives F~(p')2 2(u'F~i)P'  [u'(1F~) +3u'i] =0. (31)
Equation (31) corresponds to the characteristic equation derived by Ponce and Sim on s (1977) but in present paper is derived without using the assumption of a zero phase shift. between depth and velocity which js only true for the limiting case ot u' = O(the kinematic wave approximation). \
:
The dimensionlessphase velocityfor the completeequation is givenby
I
c'c = =CC Vo
P~ u''
(32)
..r
The logarithmic decrement (the attenuation over a single wave length) is given by
. c5'=21t p; ,
c P~
(33)
}
where P~ and P; are the real and imaginary parts of the complex propagation factor P' which is defined by equation (31). Substitution of equation (27) juto the general equation (10) for diffusionanalogy gives P=CUiDU2 which from equations (11) and (12)  or equation (13), orequation as P=1.5vo ufi (1rF~)u2, (20) (34)

caD be written
, (35)
where the value of r depends on the particular form of the diffusion analogy model.
.
(7]
DIFFUSION
ANALOGY IN FLOOD ROUTING
71
Accordingly, the dimensionless phasevelocity corresponding to equation (32) is
c~ =
1.5
(36)
and the dimensiopless logarithmic decremen.1 correspond to equation (33) is
(1'
c5~=21t(1rF~).
3
,
(37)
where r depends on the model used. The nitio ofthe attenuation over a single wave length for.a diffusion analogy model to the attenuation for the co.Ihplete solution wiII be given by
I
"
'""
eXP(~~c5;)=exp{ 21t[(1rF~)
~':~IJ}'
(38)
'f ...
Fig. 1 shows the attenuation ratio for the primary wave as a function of the dimensionless wave number (1' for the diffusion analogy model used by Ponce and Simon s (1977) for which r=O for the case of subcritical flow. Fig. 2 shows the same ratio for the diffusion
1.4
1.3
1.2
,<1: er:
o j::1.1
:z 1.0. o ><1: 0..9
) .
UJ >>
~
,0.3
<1: 0.8
~
0..7
l
i,
j
.
0..6 0..0.1
0..1
.1
e;'
10.
Fig.!.
Attenuation ratios fordiffusion analogy with r=O
analogy model proposed by Dooge and HarIey (1967) in which r=0.25. The fact that the latter model gives a closer approximation over a wideT range ofdimensionless wave numbers at aDY given Froude nwnber reinforces the arguments put forward to support ibis model in the last section. There wilI be no difference in attenuation for the limiting case of . .
72
1.4
J. C. l. DOOGE, J. J. NAPIÓRKOWSKI
[8]
1.3
1.2
o
er:
~ 1.1
2: 01.0 i= <1: :::.
I<[
~ 0.9 l0.8 0.7
0.6 0.01 01
03
j
f
r 6
10
Fig. 2. Attenuation ratios for diffusion, analogy with r=0.25 TableI Range pf applicability for diffusion analogy models Level of error (1) 1% 5% 10% Limiting value of a'
I
for' r=O (2) 0.02 0.097 0.19
I for r=0.25
l
(3)
0.31 0.51 0.64
Fo =O and the maximum differenc e in attenuation in subcritical flow will be for Fo =1. ~. Table 1 compares the range of the iwo models for different levels of prescribed errorfor Fo= 1. The nmge for the model with r=0.25 is seen to be mafiJ times that for the model '1 with r=O particularly at low lev eIs or prescribed erraT. ;l
6. RANGE OF APPLICABILITY OF DIFFUSION ANALOGY
The phase velocity for the complete equation is obtained by solving equation (31) for the complex propagation factor and applying equation (32). The resulting value of c; for the complete equations is presented graphically by Ponce and Simon s (1977). The erraT in phase velocity for the diffusion ailalogy caD bedetermined from .
,
CR=
Cd I' Cc
(39)
'[9]
DIFFBSION
ANALOGY
IN FLOOD
ROUTING
73
1.00
~ f
0.95 <J: er: 0.90 >fer: ~0.85 o u
0.80 0.1 1 (5 r Fig. 3. Error in diffusion celerity for subcritical flow
10
'l
....
'
This ratio is pl~tted in Fig. 3 as a functionof the dimensionless ~ave number. From ibis. figure it can b e seen that aU forms ofthe diffusion analogy give a good approximation of the
"
1
.
,
phase velocity for wave numbers less than the limit defined by a'<0.62.
'
.
(40)
to ensure that the errat is less than 5 %.The criteria for other levefsare given in the second column of Table 2. The above results are applicable to aU versions of the diffusion analogy discussed earlier. In the case of the attenuation the results will differ for the various models and the discussion is confined to the Dooge and fbrley (1967) model which a number of criteria
.
Table
2
Error levels in phase velocity and attenuation per unit length Level of error (1) 1% 5% 10%
Limiting value of, (f'
I
for c (2)
~ , .
~
,
"
0.27
0.62 0.89
I
I
for i5(3)
0.41 0.60 0.71.
l
indicate to be the most accurate approximation. for a singkwave length wasof the diffusion analogy attenuation to the complete equation In the lasi section the ratio approximated and plotted as a function of the dime'nsionless wave number in Fig. 2. In practice we need an estimate of the errat over a fixed length of channel rather than the errat over individual wave lengths. The logarithmic decrement for a fixed length is given by
\
,
~ u=a
"P~ X .
(41)
P~
The ratio of the twa attenuations for a fixed length is given by
/' (
.
a'
P'
,
exp[«(jd(jC)a'x']=exp { 2nu'x' [(lrF~)}
 P~J } .
(42)
74
1.4
J.
.
C. I. DOOGE, J. J. NAPIÓRK.OWSKI
[10}
13 O.f 1.2
~1.f f<I: CI::
2: 1.0 O f<I:
LU
..
~
0.9
ff<I: 0.8 0.7
~
.f :.
0.1
006 0.01
6'
10
Fig. 4. Error in diffusion analogy attenuation per unit length for subcritical flow
°The ratio is shawa in Fig. 4 for the case or x' = 1 which is usually taken as repr.esenting a relatively short channel. For multiplies of ibis length the attenuation ratio can be obtained by raising the Talio for unity length to the appropnate power. Fig. 4 suggests an appropriate criterion for 5 % erraT in a unit length as u' <0.60. (43)
The maximum dimensionless wave length for other levels oferror attenuati?n Talio per unit oJengthare sho~ in the third column of Table 2. . 'f 7. CONCLUSIONS I Three alternate forms or a convectiondiffusion equation suitable for fl.ood routing applications bas been presented. The first is ~he c1assic one, based on neglecting inertia altogether in linearised St. Venant equations, in which the hydraulic diffusivity is independent or Frol1de number. The second form is based on the partial neglect of inertia. The equivalent diffusivity differs from that given by the c1assic form particularly forhigher Froude numbers. The third form ot diffusion analogy is derived by approximating the inertia terms on the basis of the kinematic wave solution. This form reproduces exactly the nrst and second moments of the complete linear solution. This indicates, that ibis form of the diffusion analogy is the most suitable of the three alternative forms. o .' .. f ~.
t
'[II]
DIFFUSION
ANALOGY IN FLOOD ROUTING
'
75
The range of applicability of third form for flood routing applications is discussed in terms of the wave analysis method for a numbef of Froude numbers between Oand l and a number of dimensionless wave numbers «(J'y~(So) etween 0.01 and lO. b
Manuscript received 12 May 1986
REFERENCES
'Cunge J. A., Holly F. M., Verwey A., 1980, Practical Aspects of Computational River Hydraulies. Pitman, London. , ,Deymie Ph., 1935, Propagation d'une intumescence allongee;Revue Generale I'Hydraulique (France), 3, 138  142. Dooge J. C. I., Har1ey B. M., 1967,Linear routing in uniform open channels,Proc. Intern. Hydrology
..
~"'"
Symp.,
Fort
Collins,
Colorado,
September
6
'Orijsen
J. G., Vreugdenhil
Laboratory,
C. B., 1976, Numerical representation of jlood waves in rivers, Delft
'
 8,
57
 63.
Hydraulics
Publ. No. 165.
".. >Hayami Sh., 1, 1 16.
l
I
I .1
I
I I
,
1951, On the propagation of jlood waves, Kyoto Univ., Disaster Prevention Res. Inst.. \ . Henderson F. M., 1966, Open Channel F/ow, MacMillan Comp., New Yórk. !Kuchment L. S., 1972, Matematicheskoe modelirovanie rechnogo stoku (Mathematical Modelling ol River Flow), Gidrometeoizdat., Leningrad. :Lighthill M. J., Whitham G. B., 1955, On kinematic waves.] F/ood movement in lonu rivers, Proc. R. Soc., London, A229, 281  316. .Menendez A. N., Norscini R., 1982, Spectrum ofshallow water waves:an analNis, J. Hydrau!. Div., ASCE, HY1, 75  94. ,Ponce V. M.; Simops D. B., 1977, Shallow water propagation in open channeljlow, J. Hydrau!. Div., ASCE, HYI2, 1461  1476. .Ponce V. M., RuhMing Li, Simon s D. B., 1978, Applicability of kinematic and dijjusion models, J. Hydraul. Div., ASCE, HY3, 353  360. . :Schonfeld J. C., 1948,Voortplanting en Verzwakking von Hoogwatergolvenop een Rivier (Propagation . I and attenuation ofjlood wavesin a river), De Ingenieur, 60, Bouw En Waterbouwkunde, 1. Vreugdenhil C. B., 1972, Mathematical methods for jlood waves, Delft Hydraulics Laboratory, Research Rep. S89IV.
I J i
I
"",v Vreugdenhil C. B., 1977, Useof computersforhydraulic engineeringproblems, Delft Hydraulic Labo'1 ratory, Pub!. No 185. .Woolhiser D. A., Liggett J. A., 1967, Unsteady onedimensionaljlow over a piane  the rising hydro
~
graph, Water Resources Res., 3, 753  771. ZASTOSOWANIE ANALOGU DYFUZYJNEGO W MODELOWANIU TRANSFORMACJI FALI POWODZIOWEJ
Streszczenie
l
1 . h
.I
Porównan,o trzy modele dyfuzji konwekcyjnej, aproksymujace zlinearyzowane równania St. Ve. nanta.Do symulowania transformacji fali powodziowej zaleca sie 'stosowac model otrzymany przez ,oszacowanie skladników inercyjnych za pomoca fali kinematycznej. Przeanalizowano bledy tlumienia i predkosci fazowej tego modelu. Podano graniczne wartosci bezwymiarowej 'liczby falowej dla wybranych poziomów bledów. .