Fluid Dynamics Research 33 (2003) 41 55

A multiple scales analysis of the undular hydraulic jump in

turbulent open channel !ow
H. Steinr&uck , W. Schneider, W. Grillhofer
Institute of Fluid Mechanics and Heat Transfer, Technical University of Vienna, Resselgasse 3,
A-1040 Vienna, Austria
Received 23 July 2002; accepted 21 January 2003
Communicated by T. Mullin

Abstract
The undular hydraulic jump in turbulent open channel !ow is considered in the double limit of very large
Reynolds numbers, i.e. Re , and Froude numbers approaching the critical value, i.e. Fr = 1 + 32  with
 0. Fully developed turbulent !ow far upstream is assumed. Owing to the large Reynolds number the
constant inclination angle of the bottom, , is small. The undular jump is associated with a distinguished limit,
which is characterized by the similarity parameter A==2 =O(1). Since a wavy solution with a slowly varying
amplitude is expected, a multiple scales expansion is performed. A new independent variable is introduced
such that the wave length becomes constant and normalized to one. The perturbation equations of the orders
; 3=2 ; 2 , and 5=2 are considered in order to obtain a complete 8rst-order solution. Analytical results for the
amplitude of the 8rst wave and the 8rst wave length are obtained. They are compared with measured data
and reasonable agreement is observed.
The analysis also shows that a previously given ordinary di:erential equation describing the shape of the
free surface is uniformly valid.
c 2003 Published by The Japan Society of Fluid Mechanics and Elsevier Science B.V. All rights reserved.

PACS: 47.27.Jv; 47.27.Pa; 44:35:+i; 47:60:+i
Keywords: Open Channel !ow; Turbulent !ow; Hydraulic Jump

1. Introduction
The undular hydraulic jump is a peculiar change of state that can be observed in steady open
channel !ows if the upstream Froude number is but slightly above the critical value 1, cf. B&oA (1927),

Corresponding author. Tel.: +43-1-58801-32231; fax: +43-1-58801-32299.

E-mail addresses: herbert.steinrueck@tuwien.ac.at (H. Steinr&uck), wilhelm.schneider@tuwien.ac.at (W. Schneider).

c 2003 Published by The Japan Society of Fluid Mechanics and Elsevier Science B.V.
0169-5983/03/$30.00 
All rights reserved.
doi:10.1016/S0169-5983(03)00041-8

42

H. Steinr*uck et al. / Fluid Dynamics Research 33 (2003) 41 55

Chow (1959), Henderson (1966), Hager and Hutter (1984), Hager (1992), Chanson (1993), Chanson
and Montes (1995), Reinauer and Hager (1995), Ohtsu et al. (2001). As far as the observations are
concerned, the undular hydraulic jump is characterized by a wavy shape of the free surface, with
wave lengths that are much larger than the depth of the liquid and amplitudes that decay but slowly
in main !ow direction.
From a theoretical point of view there is the diHculty that, though the viscosity e:ects are very
weak, an inviscid-!ow solution does not exist, cf. Benjamin and Lighthill (1954). While the analysis
is rather subtle, the non-existence of the inviscid-!ow solution can easily be understood on the basis
of the over-all conservation laws. For, as it is well-known from hydraulics, any hydraulic jump that
satis8es the conservation laws of mass and momentum is necessarily associated with mechanical
energy losses, i.e. viscous dissipation. For laminar !ow, Johnson (1972) was able to cope with these
diHculties and provide an asymptotic analysis in terms of Froude numbers near 1 and large Reynolds
numbers. His main result is a steady-state version of the Korteweg-de Vries-Burgers equation that
governs the perturbations of the surface elevation.
For turbulent !ow, attempts have been made to analyze the undular hydraulic jump on the basis of various ad hoc approximations (e.g. Andersen, 1978; Hager and Hutter, 1984; Kaufmann,
1934; Lau:er, 1935; Montes and Chanson, 1998) or by analogy to the laminar-!ow case (Johnson,
1972). More detailed comments on those previous investigations can be found in Grillhofer (2002)
and Grillhofer and Schneider (2002, 2003).
In Grillhofer (2002) and Grillhofer and Schneider (2002, 2003), an asymptotic analysis of the
undular jump in turbulent !ow was given. Plane !ow over a bottom of constant slope was considered
in the double limit of very large Reynolds numbers, i.e. Re , and Froude numbers approaching
the critical value, i.e. Fr = 1 + 32  with  0. Fully developed turbulent !ow far upstream was
assumed. Owing to the large Reynolds number the inclination angle of the bottom, , is small.
The undular jump was considered in a distinguished limit, which is characterized by the similarity
parameter A==2 =O(1). Since the wave lengths are expected to be much larger than the water depth,
a shallow water limit was considered. Without making use of any turbulence model, a non-linear
third-order ordinary di:erential equation was obtained for the shape of the free surface. Reasonable
agreement with data from measurements was obtained. However, the damping parameter = 13 A1=2 ,
which appears in the ordinary di:erential equation, is half an order of magnitude smaller than the
other terms. Though this resembles the non-existence of the inviscid-!ow solution, it might be seen
as a weakness of the analysis. Thus in the present paper a more elaborate, multiple scaling approach
is pursued in order to provide a self-consistent asymptotic analysis of the problem.
2. Governing equations and scaling
We consider a plane open channel !ow of a !uid with constant density  and constant kinematic
viscosity  over a plane bottom with constant angle of inclination , cf. Fig. 1. The coordinate
system is such that the x-axis is in the bottom plane and the y-axis is perpendicular to it. The
corresponding velocity components are u and v, respectively.
As a reference state, we choose the fully developed turbulent open channel !ow with volumetric
mean velocity uL r and height of the free surface hLr , as shown in Fig. 1. Having a shallow water
approximation in mind we choose a typical wave length l to non-dimensionalize the x-coordinate. The

H. Steinr*uck et al. / Fluid Dynamics Research 33 (2003) 41 55

43

Fig. 1. The undular hydraulic jump in turbulent open channel !ow.

velocity components are scaled accordingly. The pressure p is referred to the hydrostatic pressure
ghLr at the bottom of the channel. For fully developed !ow the wall shear stress must balance
gravity. Thus we have for the reference wall shear stress
wr = ghLr ;

(1)

where sin  has been replaced by , as  is very small. Scaling the Reynolds stresses (u
2 ), (u
v
)
and (v
2 ) with the square of the reference friction velocity, which is given by
wr

= ghLr = 2 uL2r ;
(2)
u2r =

Fr
we introduce the following dimensionless variables:
x
X= ;
l

Y=

(U
2 ) =

(u
2 )
;
u2r

y
;
hLr

uL
UL = ;
uL r

(U
V
) =

vL
VL =
;
uL r

(u
v
)
;
u2r

(V
2 ) =

hL
HL = ;
hLr
u
U =
;
u r

pL
PL = 2 Fr 2 ;
uL r

(3)

(v
2 )
;
u2r

(4)

where the dimensionless parameters

hLr
uL r
Fr =
;  = ; 
l
ghLr

(5)

are the Froude number, the shallow water parameter and the angle of inclination, respectively.
For high Reynolds numbers, the !ow 8eld can be decomposed into a defect layer and a viscous
sub-layer. Since a universal solution exists for the viscous sub-layer, cf. Kluwick (1998), Schlichting
and Gersten (2000), it is suHcient to consider the defect layer only.
Using the scaling (3) and (4), the continuity equation for the dimensionless Reynolds averaged
velocities becomes
UL X + VL Y = 0:
The momentum equations for the defect layer in dimensionless form are
 
Fr 2 (UL UL X + VL UL Y ) = PL X + ((U
2 )X + (U
V
)Y );
 
2 Fr 2 (UL VL X + VL VL Y ) = PL Y 1 ((U
V
)X (V
2 )Y ):

(6)
(7)
(8)

44

H. Steinr*uck et al. / Fluid Dynamics Research 33 (2003) 41 55

We de8ne the averaged free surface Y = HL (X ) by a particular streamline of the Reynolds averaged
velocity 8eld (u;
L v).
L This particular streamline corresponds to the averaged volume !ow rate
 H (X;Z; t)
 HL (X )
L

V :=
U (X; Y; t) dY =
UL (X; Y ) dY:
(9)
0

Using the above de8nition of the mean free surface, the common kinematic boundary condition for
the averaged velocities holds
VL (X; HL (X )) = UL (X; HL (X )) HL X (X ):

(10)

In the absence of a general theory of turbulent free surfaces, the dynamic boundary conditions are
also formulated following common approaches, cf. Rodi (1993). Thus




(U V ) HL X [PL + (U
2 )] = 0

at Y = HL :
(11)

2






L
L
[P + (V )] H X (U V ) = 0
Matching with respect to the viscous sub-layer is accomplished by requiring
VL (X; 0) = 0

(12)

and, furthermore, making use of the logarithmic friction law (cf. Schlichting and Gersten, 2000,
p. 524, Eq. (17.54)), which may be written in terms of
ur hLr
Rer =
(13)

as follows:



1

+
L
L
L
(14)
U
ln(Rer U H ) + C + C
U S (X ) =
Fr
!
with
CL =


0

HL

Fr d UL
1

!Y
U dY

dY;

where ! is von Karmans constant and C + is another empirical constant, which, as it will turn out,
does not appear in the 8nal results.
For the purpose of comparison with other work, we note that Rer is related to  and Fr via the
logarithmic friction law
Fr
1
= ln Rer + C + + CL + CLL
(15)
 !
with CLL being a further empirical constant, cf. Schlichting and Gersten (2000), p. 535, Eq. (17.59).
Our intention is to study the !ow for slightly supercritical upstream !ow conditions. Thus we
introduce the small perturbation parameter  according to the relation
Fr = 1 + 32 ;
where the coeHcient

(16)
3
2

serves for later convenience.

H. Steinr*uck et al. / Fluid Dynamics Research 33 (2003) 41 55

45

We perform an asymptotic analysis with respect to three small parameters, namely ,  and . The
relative sizes of these three parameters have to be chosen such that one can use the shallow water
approximation and take turbulence e:ects into account, but limit their magnitude in such a way that
the well-known results for fully developed !ow can be applied locally to the disturbed !ow. Thus
no turbulence model will be required to derive the results for the free surface.
These requirements are satis8ed if the shallow water parameter  and the angle of inclination ,
which is a measure of the magnitude of the e:ects of turbulence are chosen to be
 = 31=2 ;

 = A2 ;

(17)

where the coeHcient 3 is introduced for the sake of simplifying the 8nal equations, while A is a
coupling parameter of order 1.
3. Asymptotic expansions
To motivate the correct form of the asymptotic expansion we have to anticipate the expected
results. It is required that far upstream there is the fully developed slightly supercritical !ow. Downstream an undular !ow with slowly changing wave length and amplitude is expected. The latter
requirement is typical for a multiple scales analysis. Since the wave length is supposed to change,
it is useful to replace the independent variable X by a new coordinate # such that the solution has
constant wave length as a function of the new variable. The wave length 1=! and the amplitude are
assumed to depend on the slowly varying variable % as follows:
1
# = %(X); X = &(X X0 ()); d# = ! dX;
(18)
&
where & is the ratio between the fast and slow scale variables, with & = &() 0 as  0, while
X0 is a shift of the origin between the fast and slow variables with &X0 () 0 as  0.
The !ow is a:ected by turbulence in two ways. Firstly, in the fully developed !ow the velocity
deviation from the volumetric mean is known to be of the order of the friction velocity ur , (cf.
Schlichting and Gersten, 2000, p. 534, Eq. (17.51)). Secondly, the !ow is a:ected by turbulence via
the derivative of the Reynolds shear stress (u
v
)Y , which is of order = = O(3=2 ). The latter term
turns out to be necessary to yield an undular free surface. The deviation from critical !ow conditions
is also of order O(). Therefore, one expects the expansion of the solution to start with terms of
order O(). Since the shear stresses are of order O(3=2 ), a term of this order must be contained in
the expansion.
In a multiple scales expansion, the slowly changing variables are usually determined from a
solvability condition of the next-order terms. Since the next-order terms are 1=2 times smaller then
the 8rst-order terms, the stretching factor has to be of order O(1=2 ):
&() = 1=2 :

(19)

L as follows:
Thus we expand all dependent variables, e.g. the non-dimensional pressure P,
L Y; ; ) = PL 0 (Y ) + PL 1 (#; %; Y; A) + 3=2 PL 3=2 (#; %; Y; A)
P(X;
+ 2 PL 2 (#; %; Y; A) + 5=2 PL 5=2 (#; %; Y; A) + O(3 ):
The expansions of the other dependent variables are analogous.

(20)

46

H. Steinr*uck et al. / Fluid Dynamics Research 33 (2003) 41 55

3.1. The fully developed 4ow

In view of the dimensionless form of the governing equations (6)(14), we have for the fully
developed !ow (cf. Schlichting and Gersten, 2000, p. 534, Eq. (17.51)),

UL (Y ) = 1 +  A RUL (Y ); PL (Y ) = 1 Y + 2 A(V
2 );
(U
V
) (Y ) = Y 1;

HL = 1;

(21)

where RUL (Y ) = O(1) is the non-dimensional velocity defect with respect to the volumetric mean
of the velocity. Thus the integral over the velocity defect vanishes:
 1
RUL (Y ) dY = 0:
(22)
0

3.2. Terms of order O()

Using momentum equation (8) and dynamic boundary condition (11) we obtain
PL 1 (#; %; Y ) = HL 1 (#; %):

(23)

The momentum equation (7) yields

UL 1; # = PL 1; #
with the solution
UL 1 (#; %; Y ) = HL 1 (#; %) +

(24)

A RUL 1 (Y; %):

(25)

Note that the asymptotic analysis allows the velocity defect to depend on the slowly varying
variable %.
From the continuity equation (6) we get the y-component of the velocity:
VL 1 (#; %; Y ) = !HL 1; # (#; %)Y:

(26)

From the dynamic boundary condition (11) and the relation for the surface velocity (14), the
perturbation of the Reynolds shear stress is obtained:
(U
V
)1 (#; %; 1) = HL 1 (#; %);

(U
V
)1 (#; %; 0) = 2HL 1 (#; %):

(27)

The comparison of terms of order O() leaves, however, the perturbation HL 1 of the height of the
free surface undetermined. An equation for HL 1 has to be derived from the higher-order equations.
3.3. Terms of order O(3=2 )
The analysis of the O(3=2 ) terms follows the same lines as the analysis of the O() terms. Formally
in Eqs. (23)(27) the subscript 1 has to be replaced by the subscript 3=2. We note that in order
to avoid secular terms RUL 1; % has to vanish. Thus we have RUL 1 (Y; %) = RUL (Y ).
Up to now we have not used the kinematic boundary condition at all. Inspection shows that it is
already satis8ed.

H. Steinr*uck et al. / Fluid Dynamics Research 33 (2003) 41 55

47

In what follows, the discussion of the terms of order O(2 ) and O(5=2 ), respectively, will run
along the same lines, with the exception that the kinematic boundary condition will provide solvability conditions that will serve as equations to determine the perturbation quantities HL 1 and HL 3=2 ,
respectively.
3.4. Terms of order O(2 )
Using the y-momentum equation (8), the O(2 ) pressure perturbation PL 2 can be expressed in terms
of HL 2 as
PL 2 (#; %; Y ) = HL 2; % + 92 HL 1; ## (1 Y 2 )!2 :

(28)

Inserting PL 2 into the x-momentum equation (7), using the continuity equation (6) and integrating
with respect to Y gives:
VL 2 (X; %; 1) = 3HL 1; # ! HL 2; # ! 3HL 1; ### !3 HL 3=2; % !

 1
L
L
L
RU dY:
(H 1 + A RU )H 1; # ! 2H 1; # ! A
0

(29)

However, VL 2 has to satisfy the kinematic boundary condition as well, i.e.

VL 2 + VL 1; Y HL 1 = (HL 2; # + UL 1 HL 1; # + HL 3=2% )!:

(30)

Thus we obtain the solvability condition

!3 HL 1; ### + !HL 1; # (HL 1 1) = 0;

(31)

which is an equation for HL 1 as a function of the fast variable #. Constants of integration, which
may depend on the slow variable, remain undetermined. They have to be determined from avoiding
secular terms in the equation for the O(5=2 ) terms.
3.5. Terms of order O(5=2 )
Using the y-momentum equation (8) and the dynamic boundary condition (11) we obtain the
O(5=2 ) pressure perturbation PL 5=2 :


9
d!
!2 HL 3=2; ## + !2 HL 1; #% + HL 1; # !
(1 Y 2 ):
PL 5=2 = HL 5=2 +
(32)
2
d%
In the balance of the O(5=2 ) terms of the x-momentum equation the 8rst perturbation of the Reynolds
shear stress (U
V
)1 has to be taken into account, i.e.
(UL 5=2; # + UL 2; % )! + 3(UL 3=2; # + UL 1; % )! = (PL 5=2; # + PL 2; % )! 13 A(U
V
)1; Y
UL 1 UL 1; % ! UL 1 UL 3=2; # ! UL 3=2 UL 1; # ! VL 3=2 UL 1; Y VL 1 UL 3=2; Y :

(33)

48

H. Steinr*uck et al. / Fluid Dynamics Research 33 (2003) 41 55

Using the continuity equation (6), the dynamic boundary condition (11) and Eq. (27) we obtain
VL 5=2 3(HL 3=2; # + 2HL 1; % )!
d! L
= HL 5=2; # ! 3HL 3=2; ### !3 HL 2; % ! 9HL 1; ##% !3 9!2
H 1; ##
d%

3
(HL 21 )% ! + (HL 1 HL 3=2 )# ! + (HL 1; % + HL 1; # ) A RU
2
 1
A
L
L
RU dy ((U
V
)1 |Y =1 (U
V
)1 |Y =0 ):
+ 2(H 3=2; # + H 1; % )! A
3
0

(34)

Finally, with the help of the kinematic boundary condition for the O(5=2 ) terms, i.e.
VL 5=2 + VL 1; Y HL 3=2 + VL 3=2; Y HL 1
=HL 5=2; # ! + HL 2; % ! + UL 1 HL 3=2; # ! + UL 3=2 HL 1; # ! + UL 1 HL 1; % !;

(35)

we obtain the following equation for HL 3=2 :

!2 HL 3=2; ## + HL 3=2 HL 1 HL 3=2 = r

(36)

with
!r# =

A L
d!
H 1 2HL 1; ##% !3 HL 1; ## !2
R% !:
3
d%

(37)

3.6. Di6erential equations for the slowly varying quantities

Integrating (31) twice we obtain
1
2

!2 HL 21; # + 16 HL 31 12 HL 21 = RHL 1 + S;

(38)

where R(%) and S(%) are constants of integration which depend on the slowly varying variable %.
Considering R and S given, ! has to be determined such that HL 1 is periodic with period 1 as a
function of #. In the following, we derive di:erential equations for the slowly varying constants R
and S. Integrating (36) with respect to # over one period gives the following di:erential equation
for R:

A 1 L
!R% =
(39)
H 1 (#; %) d#:
3 0
Taking the derivative of (38) with respect to #, multiplying the result with HL 3=2; # and adding (36)
multiplied by HL 1; # we get
(!2 HL 3=2; ## + HL 3=2 HL 1 HL 3=2 r)HL 1; # + (!HL 1; ## + 12 HL 21 HL 1 R)HL 3=2; # = 0:
Integrating (40) with respect to # over one period, and integrating the result by parts yields
 1
r# HL 1 d# = 0:
0

(40)

(41)

H. Steinr*uck et al. / Fluid Dynamics Research 33 (2003) 41 55

49

Taking the derivative of (38) with respect to % and using (37) and (41) we obtain the following
di:erential equation for S:

A 1 L2
!S% =
(42)
H d#:
3 0 1
As a result, the slowly varying quantities can be determined as follows: Let h1 (R; S) 6 h2 (R; S)
6 h3 (R; S) be the three zeros of the polynomial
p(h; R; S) = 16 h3 12 h2 Rh S
and let
Ij (R; S) =

h3 (R;S)

h2 (R;S)

(43)

hj dh
;
(h h1 (R; S))(h h2 (R; S))(h3 (R; S) h)

Then the equations for R, S and ! are

2A
2A
R% = I1 (R; S); S% = I2 (R; S);
3
3

j = 0; 1; 2:

1
:
!=
2 3I0 (R; S)

(44)

(45)

It remains to determine appropriate initial conditions for R and S. As the origin of the X -axis we
choose the location of the 8rst crest of the free surface, i.e. the 8rst maximum of HL 1 . Since HL 1 has
to decay to zero for X , the period becomes in8nite as X . This gives the following
initial conditions for R and S:
R(0) = S(0) = 0:
of

(46)

The main result of the analysis can now be summarized as follows: Let H(X ; R; S) be the solution
H2X + 13 H3 H2 = 2RH + 2S;

HX (0; R; S) = 0;

HXX (0; R; S) 0;

(47)

where R = R(%) and S = S(%) are the solutions of (45) and (46). Then the 8rst-order perturbation
of the free surface, i.e. HL 1 , is given by


 %
%
d%
#

1=2
; R; S ; X =
; #=
(48)
HL 1 (X ) = H
!
!

0
with
1
X (#; ) = X + X0 ();



X0 () =

1=2

!( #)

d#:

(49)

The shift X0 of the fast variable is introduced such that the 8rst maximum occurs at X = 0. We
will see later that X(0) = 0 and X0 is 8nite. Thus the multiple scales approximation (48) fails for
X . However, an asymptotic analysis using the original coordinate X yields the inviscid
solution
HL 1 (X ) = H(X ; 0; 0);
(50)
which is a homo-clinic orbit in the phase plane. It is the limiting solution for the 8rst elevation and
describes the behavior for X correctly, but fails to approximate the undular behavior.
In Fig. 2, the multiples scales solution (48), evaluated for =104 ; A=3 and the homo-clinic orbit
are plotted. Moreover, the solution of the equations for the slowly varying variables represented by

50

H. Steinr*uck et al. / Fluid Dynamics Research 33 (2003) 41 55

Fig. 2. Multiple scales solution (48) for A = 3,  = 104 and the homo-clinic orbit (A = 0) according to Eq. (50). The
solution for the slowly varying variables is represented by h2 and h3 , i.e. the local minima and maxima, respectively.

Fig. 3. Phase portrait of the multiple scales solution (48) for A = 3;  = 104 and the homo-clinic orbit (A = 0) according
to Eq. (50).

the zeros h2 and h3 of the polynomial p(h; R; S) are also shown in Fig. 2. In Fig. 3, the multiple
scales solution (48) and the homo-clinic orbit (50) are shown in the phase plane.
3.7. Asymptotic behavior far downstream (%; X )
To discuss the behavior for X , it is appropriate to consider the slowly varying quantities
hm = 12 (h3 + h2 );

Rh = 12 (h3 h2 )

(51)

H. Steinr*uck et al. / Fluid Dynamics Research 33 (2003) 41 55

51

instead of R, S. Expanding the integrals Ij for Rh=hm 1 we obtain the following di:erential equations for Rh and hm :


1
Rh
A
1+
(52)
RhX A
; hm; X
4hm
3
hm
with the solution
hm



A
A
X + ln 1 + X ;
3
3

Rh = O(X3=4 ):

(53)

Thus the amplitude of the oscillation decays and the mean value over an oscillation period increases
linearly with increasing X.
3.8. Behavior near the origin (% 1)
Although the polynomial p(h), which occurs in the denominator of I1 and I2 , respectively, has a
double root at h2 = h1 = 0 for R = S = 0, these integrals exist for R = S = 0:

(54)
I1 (0; 0) = 2 3; I2 (0; 0) = 4 3:
Using
R R% (0)% = 4A%;

S S% (0)% = 8A%;

we obtain for the zeros of the polynomial p(h):

h1 4 A%; h2 4 A%; h3 3 + 83 A% + :
Expanding I0 for % 1 gives
1
1
ln
+ 2 ln 6:
!
A%
With


 %
1
1
X(%) =
d% % ln
+1+2 ;
!
A%
0
we get X(0) = 0 and obtain the shift

 1=2

1
1
1
3

1

X0 () = X
=
d# ln A  ln 2 ln 3 :
2
2
2
2

!( #)
0

(55)
(56)

(57)

(58)

(59)

4. Comparison with experiments

In the literature there is a large amount of experimental data for the undular hydraulic jump, but
to obtain a meaningful comparison with the present analysis the experiments have to satisfy the
following criteria:
(i) A fully developed !ow far upstream is required. This is only possible in an inclined channel.

52

H. Steinr*uck et al. / Fluid Dynamics Research 33 (2003) 41 55

Table 1
Experimental data, selected from Chanson (1993)
#

Exp.

V
(m2 =s)


(rad)

hLr
(m)

w
(m)

+
(m)

hLC
(m)

Fr
()

()

+as =+ex
()

5=2 =
()

1
2
3
4
5

HCUJ8e
HCUJ10c
HCUJ4b
HCUH1b
HCUJ4a

0.028
0.080
0.119
0.0416
0.120

4:10 103
3:99 103
4:99 103
3:7 103
4:33 103

0.041
0.084
0.105
0.053
0.109

0.25
0.25
0.25
0.25
0.25

0.310
0.600
0.750
0.450
0.660

0.052
0.100
0.144
0.068
0.138

1.077
1.049
1.119
1.088
1.062

0.115
0.225
0.075
0.086
0.171

0.94
1.07
0.87
0.83
1.20

0.15
0.04
0.35
0.23
0.08

(ii) The theory is based on an asymptotic expansion in terms of the small parameter . Thus we
consider only experiments with 0  0:1, i.e. 1 Fr 1:15.
(iii) The relative error of neglected terms has to be small. The smallest
terms that in!uence the
5=2
leading order result for the wave length are of order O(A )=O( ), whereas terms of order
O(3 ) are neglected. Thus the relative truncation error is of O(5=2 =), and only experiments
with 5=2 = being suHciently small are considered.
(iv) The height of the 8rst crest and the 8rst wave length, i.e. the distance between two successive
crests, must be among the measured quantities.
In Table 1 experiments selected to meet these criteria are listed.
In the analysis, the maxima of the surface elevation occur at # = 12 ; 32 ; : : : ; and so on. Thus the

8rst maximum is HL 1; max = h3 ( 2 ) = 3 + 4 A 3  , cf. (56), and for the height hC of the 8rst crest we
obtain:
hC = hLr 1 + 23 (Fr 1)HL 1; max = 1 + 2(Fr 1)(1 + 83 ):
(60)

The parameter = A =3 has been introduced in Grillhofer (2002) and Grillhofer and Schneider
(2002,2003) and is used also here for convenience.
Referring the 8rst wave length + to l = hLr = = hLr =3 , we obtain from (18)
  3=2
 
3 
1

1

X
=
d#:
(61)
-() = X
2
2

!(
#)
1=2
In the limit  0 Eq. (61) becomes

8
8 3
-() ln + 1 = ln + 1:
A 
3

(62)

In Figs. 4 and 5, a comparison of measured data with the results (60), (61) and (62), respectively,
is given. The volume !ux in the experiments is measured with an accuracy of 5% (cf. Chanson,
1993, p. 2-1). Therefore, the e:ect of a variation of the volume !ux by 5% is also shown in
the 8gures. It can be seen that the results are very sensitive with respect to uncertainties in the
volume !ux. In Fig. 5, results are givenfor both a numerical evaluation of - and the asymptotic
approximation for small values of = A =3.

H. Steinr*uck et al. / Fluid Dynamics Research 33 (2003) 41 55

53

Fig. 4. Comparison of calculated and measured amplitudes hLC = hLr of the 8rst elevation. The dotted lines indicate how the
non-dimensionalized height of the 8rst crest changes for a variation of the volume !ow rate within 5% around the value
given in Table 1.

Fig. 5. Comparison of calculated and measured distances +=l between the 8rst two maxima. The dotted lines indicate how
the non-dimensionalized 8rst wave length changes for a variation of the volume !ow rate within 5% around the value
given in Table 1.

Taking into account that the aspect ratio w= hLr , where w is the width of the channel, is larger
than 0.16 in all experiments, so that side wall e:ects may in!uence the results considerably, the
agreement between the measured data and the theoretical results for plane !ow is reasonable.

54

H. Steinr*uck et al. / Fluid Dynamics Research 33 (2003) 41 55

5. Conclusions
The multiple scaling approach as proposed in this paper permits a self-consistent asymptotic
analysis of the undular hydraulic jump in turbulent !ow. Results for the perturbation of the height
of the !uid are obtained without making use of a turbulence model. Ordinary di:erential equations
have been derived for the slowly varying quantities, i.e. the amplitude and the wave length, as well
as for the rapidly varying height of the free surface.
The price for the rigorous analysis is a limited range of parameters where the requirements for
the analysis are satis8ed. In view of the high sensitivity of the wave length and the amplitude, the
agreement with experimental data appears reasonable.
Appendix A. Uniformly valid di#erential equationrederived
For the numerical evaluation it is much easier to solve a uniformly valid di:erential equation
instead of evaluating (48). For that purpose we consider the sum of the 8rst two terms of the
expansion for HL , i.e.
HL 1+ = HL 1 + 1=2 HL 3=2 :

(63)

Using (31) and (36) and neglecting the terms of order O() we get the ordinary di:erential equation
A
;
(64)
HL 1+; XXX + HL 1+; X (HL 1+ 1) HL 1+ = 0 with =
3
which has been given previously by Grillhofer (2002) and Grillhofer and Schneider (2002). One
can easily verify that (64) is equivalent to
HL 21+; X + 13 HL 31+ HL 21+ = 2RHL 1+ + 2S;

(65)

RX = HL 1+ ;

(66)

SX = HL 21+ :

Thus an asymptotic analysis of (64) using the multiple scales method gives the leading-order result
(48), and Eq. (64) can be viewed as an uniformly valid di:erential equation. A multiple scales
analysis of Eq. (64) can be found in Steinr&uck (2002).
References
Andersen, V.M., 1978. Undular hydraulic jump. J. Hydraul. Div. 104 (HY8), 11851188.
Benjamin, T.B., Lighthill, M.J., 1954. On cnoidal waves and bores. Proc. R. Soc. A 224, 448460.
B&oA, P., 1927. Berechnung der Wasserspiegellage. Forschungsarbeiten VDI, 284.
Chanson, H., 1993. Characteristics of undular hydraulic jumps. University of Queensland, Department of Civil Engineering
Research Report Series, CE146.
Chanson, H., Montes, J.S., 1995. Characteristics of undular hydraulic jumps: experimental apparatus and !ow patterns. J.
Hydraul. Eng. 121, 129144.
Chow, V.T., 1959. Open-Channel Hydraulics. McGraw-Hill, New York.
Grillhofer, W., 2002. Der wellige Wassersprung in einer turbulenten Kanalstr&omung mit freier Ober!&ache. Dissertation,
Technische Universit&at Wien.

H. Steinr*uck et al. / Fluid Dynamics Research 33 (2003) 41 55

55

Grillhofer, W., Schneider, W., 2002. The undular hydraulic jump in turbulent open channel !ow. Paper Presented at the
GAMM Annual Meeting, Augsburg, 26 March 2002.
Grillhofer, W., Schneider, W., 2003. The undular hydraulic jump in turbulent open channel !ow at large Reynolds numbers.
Phys. Fluids 15, 730735.
Hager, W., 1992. Energy Dissipators and Hydraulic Jumps. Kluwer, Dordrecht.
Hager, W.H., Hutter, K., 1984. On pseudo-uniform !ow in open channel hydraulics. Acta Mech. 53, 183200.
Henderson, F.M., 1966. Open Channel Flow. Macmillan, New York.
Johnson, R.S., 1972. Shallow water waves on a viscous !uid the undular bore. Phys. Fluids 15, 16931699.
Kaufmann, K., 1934. Hydromechanik II. Springer, Berlin.
Kluwick, A. (Ed.), 1998. Recent Advances in Boundary Layer Theory. CISM Courses and Lectures No. 390. Springer,
Berlin.
Lau:er, H., 1935. Wassersprung bei kleinen Sprungh&ohen. Wasserwirtsch. Tech., pp. 137140.
Montes, J.S., Chanson, H., 1998. Characteristics of undular hydraulic jumps: experiments and analysis. J. Hydraul. Eng.
124, 192205.
Ohtsu, I., Yasuda, Y., Gotoh, H., Iahr, M., 2001. Hydraulic condition for undular-jump formations. J. Hydraul. Res. 39,
203209.
Reinauer, R., Hager, W., 1995. Non-breaking undular hydraulic jump. J. Hydraul. Res. 33, 683698.
Rodi, W., 1993. Turbulence Models and their Application in Hydraulics, A State-of-the-Art Review, 3rd Edition. Balkema,
Rotterdan, Netherlands.
Schlichting, H., Gersten, K., 2000. Boundary Layer Theory, 8th Edition. Springer, Berlin.
Steinr&uck, H., 2002. Multiple scales analysis of the undular hydraulic jump. Paper presented at the GAMM Annual
Meeting, Augsburg, 26 March 2002 (Also: Multiple scales analysis of an equation describing undular hydraulic jumps.
Z. Angew. Math. Mech., submitted).