Acta Meehanica 54, 157--179 (1985)

ACTA bIECHANICA
@ by Springer.Verlag 1985

Critical Flow Condition in Open Channel Hydraulics

By

W. H. Hager, Lausanne, Switzerland

With 13 Figures

(Received July 4, 1984)

Summary

Transition between sub- and supercritical flows in open channel occurs at the so-called
critical point, for which critical flow conditions appear. This particular flow state has been
originally introduced for flows with parallel streamlines. However, as streamlines are
considerably sloped and inclined in the vicinity of the critical point, these effects have to be
accounted for. The present investigation explores in detail these effects. In particular,
present investigations include i) first order analysis expressing discharge in terms of up-
stream head and channel geometry (calibration of discharge measurement structures) and
if) determination of surface profiles for transitional flow states by accounting for the
effective pressure and velocity distributions. Analysis is based on a first order model with
restriction to typical channel bottom and sidewall geometry. Results are compared with
observations, and a fair agreement between the two is noted.

1. Introduction

Flow in open channels m a y have two distinctly different flow conditions,

which are distinguished b y the F r o u d e n u m b e r F = V2/(gh) in which V is average
cross-sectional velocity, g gravitational acceleration and h flow depth. F o r s t e a d y
flow conditions q u a n t i t y F attributes to one-dimensional flows, for which V and h
are only functions of the longitudinal coordinate x, thus F = F(x). Flows with
F < ( > ) 1 are refered to as subcritical (supercritical), and the limit between the
two, F = 1, corresponds to critical flow conditions. Flows with F < 1 m u s t be
c o m p u t e d against the flow direction, while these with F > 1 are to be c o m p u t e d
in flow direction, w h a t e v e r the Channel g e o m e t r y looks like [4]. This c o m p u t a t i o n a l
principle becomes particularly i m p o r t a n t for b a c k w a t e r profiles, and it m u s t be
a t t r i b u t e d to Chow [4] having standardized the possible surface profiles in pris-
matic straight channels. Transitions between the two respective flow conditions
158 W.H. Hager:

may either be from sub- to supercritical flow (which are known to be smooth) or
from super- to subcritical flow (which become discontinuous). The above compu-
tational principle holds only for transitions from sub- to supercritical flow states
while hydraulic jumps must be analysed b y a modified procedure [7], [8].
In open channel hydraulics critical flow state plays an outstanding role since
discharge and flow depth are related uniquely for given channel geometry. As has
been shown b y Jaeger [12] critical flow condition implies either minimum head
H = h + V~/(2g) for constant discharge Q, or discharge Q becomes a maximum for
constant head H. This particular property allows determination of:

-- discharge-head relation using critical flow condition (discharge measurement

structures), and
-- surface profiles at the up- und downstream zones of the critical point (for
which F -- 1) by initiation with critical flow condition (control point).

The present investigation considers the two of these problems b y accounting

for the effects of streamline slope and curvature (which become important in the
vicinity of the critical point). In section 2, the governing equations for transitional
flow state are presented, and proof is given for the existence of critical flow. Its
geometrical locus is one of the following possibilities [11] :

-- maximum o f the bottom geometry (weir crest, for instance) in prismatic

channels with constant discharge),
-- m i n i m u m o f cross-sectional area (contractions, for instance) in a channel with
constant bottom slope and discharge,
-- e n d of lateral inflow (side-channel spilhvays, for instance) in prismatic channels
with locally increasing discharge and constant bottom slope,
b e g i n n i n g of lateral outflow (side-weirs, for instance) in prismatic channels
with locally decreasing discharge and constant bottom slope.

Evidently, there may be combinations of the above cases, for which modi-
fications are to be considered. Note too, that critical condition is not necessary for
transitions from sub- to supercritical flow but that downstream submergence must
be relatively small in addition.
In section 3 derivatives of the surface profile are deduced in the vicinity of
the critical point. These are computed using the lowest order of approximation.
Typical channel geometries are given in section 4. Considerations are restricted
on flows with locally constant discharge, for which results become relatively
simple. Furthermore, channel bottoms will be either parabolic or straight, and
cross-sectional profiles trapezoidal. Note, however, that the present approach
allows determination of critical flow conditions for arbitrary channel geometry,
but that analysis then becomes extremely laborious.
In section 6, the generalized critical flow conditions are formulated b y account-
ing for the effects of streamline slope and curvature. These equations are applied
 Critical Flow Condition in Open Channel Hydraulics 159

in section 7 on rectangular profiles with either Venturi (local constriction) or

weir shape. Results yield expressions for discharge in terms of upstream flow
depth and channel geometry. Solutions are compared with observations in
section 8, and a fair agreement in found. In section 9 trapezoidal Venturi channels
and weirs are explored in detail. In part 10, finally, surface profile computations
at the up- and downstream zones of critical point are presented. I t will be shown
that the present approach allows for a complete description of flow in channels
with arbitrary but relatively smooth channel geometry under steady flow condi-
tions.

2. E n e r g y Principle

Flows in open channels with predominate flow direction may be analysed using
the one-dimensional mass, energy and momentum balances. When restricting
considerations on steady, inviscid flow conditions, the first two become particu-
larly simple. Let V be the average velocity and A the wetted cross-sectional
area, then
Q = VA (1)

E=Q z+h@ (2)

in which Q discharge, E energy, z the channel bottom elevation measured from an
arbitrary but fixed reference level, h pressure height and g gravitational accel-
eration.
For flows with constant discharge in time and space, and for nearly parallel
horizontal streamlines, Eq. (2) simplifies to

V2
H = z + h + 2g (3)

in which H is energy head and h corresponds to the flow depth, being identical
with pressure head. For flows with slightly inclined and curved streamlines, (3)
must be expanded to yield [9]

H = z + h + ~Q~( ]+ 2hh" -- h '~ 4- hz" -- h' z' - - z ' ~

3 !
(4)

in which primes denote ordinary differentiation with respect to the longitudinal

coordinate x. Note that Eqs. (3) and (4) become identical for parallel, straight
streamlines, k' -- h . . . . z' -- z --= 0. Note too that H according to (3) and (4) m a y
only be determined at a particular position x of the channel, but that a second
relation must prescribe longitudinal variations of H = H(x) in order to determine.
the surface profile h(x).
 160 W.H. Hager:

Let us pick one of the cross-sections at position x ---- xo, for which A : Ao.
Then
Q~
/t0 = h + - - (5)
2gAo ~

is corresponding energy head with H = zo + Ho. For fixed discharge Q, and with
Ao = Ao(h), Ho depends only on flow depth h. The extreme of this function is

dH ~ Q2 dA ~
-- ---- 1 -- = 0. (6)
dh gAo s dh

To decide whether this corresponds to a minimum or m a x i m u m in H, the second

derivative is considered

~dh gAo
-4 \ dh / gA0a dh ~ , (7)

which, for extreme value considerations, simplifies to

d2Ho 3 dAo d2Ao/dh 2

(s)
dh ~ A o dh dAo/dh

According to Appendix I, this expression remains always positive; therefore, the

above extremum corresponds always to a minimum.
I n open channel hydraulics it is customary to denote

Q2 dA
F 2 -- - - (9)
gA~ dh

in which F is the Froude number; note that index " 0 " has been dropped for con-
venience. Note also that this result holds only for flows with nearly parallel,
horizontal streamlines.
The conditions F = 1 will be refered to as critical/low condition, and Y < (:>) 1
correspond to subcritical (supercritical) flow states. Further, as has been demon-
strated b y Jaeger [12], F = 1 represents the minimum for fixed discharge Q with
variable cross-section A0 (as shown above) and for fixed energy head H0 with
variable discharge Q.

3. Extended Energy Considerations

Consider open channel flows which are accurately described b y Eq. (3) and for
which discharge Q and energy head H are constant. With
Q2
H = z + h + - - (10)
2gA ~
 Critical Flow Condition in Open Channel Hydraulics 161

t h e first following derivatives are

Q2A'
H ' =: z' + h' - - - - = O, (11)
gA~

3Q2A ' Q~A"

H " = z" + h" + - - 0, (12)
gA ~ gA s

12Q2A,3 9Q2A'A" Q~A'"

I t ' " = z ' " + Id" - - + - - = 0, (13)
gA ~ gA 4 gA s

60Q~A'a 72@A'~A" 9Q~A"2 12Q~A,A,,, Q2AIV

I/IV = zlV _1_ hlV d- @- - -~- -- : - O.
gA s gA 5 gA 4 gA 4 9.43
(14)

Condition for critical flow results from Eq. (11), implying t h a t z' = 0. Inserting
the critical condition Q~ = gAah'/A ' into the following relations yields expressions
for the derivatives in h = h(x), n a m e l y

h'A" 3h'A'
h" = z", (15)
A' A

h'A'" 9h'A" 12h'A '2

h. . . . . . + z'", (16)
A' A A2

h'A TM 12h'A'" 9h'A ''2 72h'A'A" 60h'A '3

h Iv -- - - + z Iv. (17)
A" A AA' A2 A3

N o t e t h a t these are valid only at the so-called critical point, at which Y -- 1, and
t h a t t h e y m a y subsequently be used in order to determine flow characteristics at
the transition from sub- to supercritical flow conditions.

4. C h a n n e l Geometry

Preceeding analysis m a y only be quantified b y specifying the governing

channel b o t t o m g e o m e t r y z(x) and the cross-sectional area F = F(x, h). Present
considerations will be restricted to typical cases often met in practice, including
parabolic b o t t o m g e o m e t r y
z(x) = - - x 2 / 2 R (18)

in which R is the radius of curvature at the origin x = 0, and symmetrical

trapezoidal cross-sections with
A -- bh + mh 2 (19)
 162 W.H. Hager:

in which b ~ b(x) is channel width at the bottom and m = re(x) the cotangent of
the side-angle.
From (18) one has for z . . . . z/R, z" =- - - 1 / R and zi ~ 0 with i > 2. Let
x = 0 be the critical point, then zc = 0, z c ' = 0 (from Eq. (11)), z c " = - - 1 / R
and z~ ~- 0.
The derivatives of Eq. (19) may be shown to obtain

A ' = h'(b + 2mh), (20)

A " = b"h + bh" + m " h 2 + 2mh '2 + 2 m h h " , (21)

A"" = 3b"h' + bh'" 4- 6m"hh' + 6mh'h" + 2 m h h ' " , (22)

A TM ~ 6b"h" -+ bh TM + 12m"h '2 + 12m"hh" + 6mh ''~ + 8mh'h"" + 2mh,h Iv (23)

in which all derivatives in b(x) and re(x) exceeding order two have been dropped
(see also below).

5. Flow Surface at Critical Point

Equations (15) to (17) can be further developped using the geometrical

relations deduced above. Let

mh m"h h2b ''

s=---~-, t-- b" ' u-- b ' w z"h (24)

be typical scalings, then

h '2 = u(1 + s) (1 + t) -- w(1 + s) (1 + 2s)

(25)
3 + lOs + lOs 2

hh"=--u(l+s){2+5s+t(l+4s)} +2h'2(l+2s)(2+5s+5s 2) (26)

(1 4- 8) (3 4- 108 -4- 1082)

h~h''' {6uhh"(1 + 2t) + 6sh2h"24- 12tub '2} (1 + s) a -- 60h'4(1 4- 2s) 4

h, 4h'2(1 + 8)2 (3 + 10a + 10s ~)

36h'Z(1 @ 2s) {u(1 + 2t) + 2shh"} + 9{u(1 4- t) 4- hh"(1 4- 2s) 4- 2sh'2}a

4h'2(3 + 10s + 10s 2)

72h'Z(1 + 2s)" {u(1 + t) + hh"(1 -4- 2s) + 2sh '~}

q- 4h'2(1 + 2) (3 + 10s + 10s 2) (27)

Note that all quantities relate to critical flow conditions (e.g. s in (24) is combined
b y m = me, h =- hc and b = be). Further, it is to be noted that h '~ and hh" in
Eq. (27) must be expressed b y the results obtained ih (25) and (26), respectively.
 Critical Flow Condition in Open Channel Hydraulics 163

R e c t a n g u l a r (m ~ 0) a n d triangular (b ~ 0) profiles are special eases of the

trapezoidal profile. Using Eqs. (25) to (27) results for rectangular channels are

h'2 = (u - - w ) / 3 , (2S)

hh" : - - 2 ( u -~ 2w)/9, (29)

h~h ''' 4u(3h '2 - - h h " ) - - 3(u 2 ~- h2h ''2) 24h'~hh '' - - 20h '4
h' 4h '2 , (30)

while these for t r i a n g u l a r c h a n n e l s m a y be expressed with 7 ~ m " h ~ / m as

h '~ : (7 - - 2w)/10, (31)

hh" : - - ( 7 ~- 2w)/5, (32)

h2h ''' 3(207h '2 -- 47hh" - - 5h2h ''2 ~- 60h'2hh '' - - 70h '4 - - 3v~/2)
(32)
h' 20h '2

I t is noted t h a t a r b i t r a r y derivatives in h(x) a t the critical point m i g h t be com-

puted, b u t t h a t c o m p u t a t i o n s b e c o m e the more and more laborious.

6. Extended Bernoulli Equation

Transitions f r o m sub- to supercritical flow states are continuous, while these

f r o m super- to subcritical flow states (resulting in a hydraulic jump) are dis-
c o n t i n u o u s [7]. Consider a channel with a local, continuous contraction in width.
At the u p s t r e a m zone channel width B and b o t t o m slope J,,o < Js,c are c o n s t a n t
(Js,c denotes critical b o t t o m slope for which critical flow occurs at uniform flow
conditions), while b o t t o m slope of the d o w n s t r e a m zone is such t h a t J~,~ > Js,c,
see Fig. 1. F o r c o n s t a n t channel roughness uniform flow depths h~,o and h~v,~
result f o r x --> =J=~, a n d transition from sub- to supercritical flow state occurs a t
the critical point x = 0 of the channel. F a r up- a n d downstream, streamlines are

Jso --
Fig. I. Transition from sub- to supercritical flow states by a local, continuous channel
contraction and subsequent expansion. Note that up- and downstream bottom slopes,
,J~,oand Js,~, are such that uniform flows are sub- and supereritical, respectively (dashed
lines) (see also Fig. 14 for plan view)
164 W. lt. ltager:

slightly inclined but parallel to the channel bottom. Therefore, Eq. (3) may be
applied by specifying H = H(x). At the transitional zone, however, Eq. (4) must
be accounted for since effects of streamline inclination and curvature become
significant. Minimum energy head m a y be computed by setting again

_ __ { 2hh" -- h '2
H ' = h' Q2A' 1 + -~ hz" -- h'z' -- z '2)
gA a ~ 3
(34)
A- ~ A- hz'" -- h"z' -- 2z'z" = O.

At the critical point necessary condition for critical flow is zc' = 0 from Eq. (11),
and su]]icient conditions are

2hh" -- h '2 \
H = h q- 2 - ~ 1-k 3 + hz"), (35)

@ A ' _-- (1 -~ 2hh" -- h '~ hz" hh"'A-t-1

gA3h ' 3 + 3A' ] (36}

in which z'" has been dropped (see Eq. (18)). Equation (36) relates discharge Q
and critical depth h c, while (35) links critical depth hc with critical head H e. In-
formations concerning hc and Hc in terms of discharge and cross-sectional area
m a y be computed when inserting Eqs. (25) to (27) into (35) and (36).

7. Critical Flow in Rectangular Profiles

Rectangular profiles are characterized by channel width b = b(x) and bottom

geometry z = z(x). Two particular cases are of interest i) width variation in a
straight, horizontal channel, z ~-- 0 (simplest of Venturi channels), and ii) bottom
variation z(x) according to Eq. (18) in a prismatic channel (parabolic, two-
dimensional weir).
Venturi Channel

Restricting considerations on the case in which z --= 0, (thus w = 0 in (24))

Eqs. (28) to (30) yield
~2h"H
h '2 = u/3, hh" = --2u/9, h' = --61u/36 ~ - - 5 u / 3 (37)

from which
Q2A ' ( 8__..~1--1 Q2
(38)
 Critical Flow Condition in Open Channel Hydraulics 165~

according to (36) and

Q~ ( 7u) (39).
H = h + 2g--d~ 1 - - ~

according to Eq. (35). Eliminating discharge Q then yields

H=h
{1 + ~ -1( 1 - - 7u~/{1+ 8u)} ~(5u)
~-- 1-- (4o)

in which first order expansions according to [9] have been performed. Owing:
to the significant variation of h(x) at the transitional zone, critical flow depth
h is difficult to measure. Consequently, it is more convenient to consider the new
(measurable) parameter
H~b,,
v = --; (41)
b
u and U are related by

u = --9-- 1 -~ (42)
from Eq. (40), thus
3h 1-- (43),
H= y 2431"
Discharge Q may finally be expressed with the aid of Eq. (38) to yield

Q : (2/3) 8/2 b [/gH-'~ 1 q- (44)

Evidently, U ---- 0 for parallel streamlines (lowest order of approximation), and

the resulting expression
Q0= (2/3)8f~b l / ~ (45)
gives a unique relation between discharge Q and head H at the critical cross-
section with width b. The more accurate expression (44) accounts in addition
for the shape of the contraction (U c~ B ' ) , and it is noted that Q/Qo > 1 with
b/' > 0. Consequently, discharge is underestimated when computed with Eq. (45).
Note that (44) results from a first order model, whence U must be smaller than
unity in order that quadratic terms in U are insignificant.

Circular Weir

By setting u ~ 0, Eqs. (28) to (30) yield for

~ rpt

h '2 ~ --w/3, hh" ~ --4w/9, - - 5w/9 (46),,

h'
166 W.H. Hager:

and Eqs. (35), (36) become

~=h+2g--~
(1+ 22w
27/' (47)

Q2
= (1 + w) -1, (48)
gb~h3
whence
H~-~-
3 (5w)1--~- . (49)

It is convenient to set
--W = z " H -~ H / R (50)

in analogy to U as defined in Eq. (41), thus w -~ - - 2 W / 3 and

22W~
Q = (2/3) a/2 b ~/g-~ 1 + - - ~ - ) . (51)

Since zc" ~ 0 for weir flow (W ~ 0) Eq. (51) expresses that discharge Q is always
larger than when computed by the hydrostatic theory approach (45) (in which
W = 0).
When combining the two effects, whence considering a Venturi channel with
a weir-shaped bottom profile, the results

h 2{ 2(20U~--2UW- 15W2)~
= -ff 1 + 243(2U -- 3W) J' (52)

14U 22U
Q/Qo = 1 + ~-~ + 81 (53)

may be deduced. Note that the two effects add linearly in the discharge relation
(53).

8. Experiments

Before passing over to more complex channel geometries, the preceeding

results will be compared with observations regarding i) slope of free surface
at critical point and ii) discharge relation.

Venturi Channel

Investigations concerning Venturi channels analyse usually discharge-head

relation. However, surface profiles have been observed only by a few researches,
notably by Khafagi [13], [14] and de Marchi [15] (the latter omitting indication
 Critical Flow Condition in Open Channel Hydraulics 167

30 ~
-h'
20 o 9

10 c ~/-"

U
0 T
0 1 2

Fig. 2. Comparison between Eq. (54) and Khafagi's data [13], [14] with respect to the free
surface inclination at the critical cross-section. (~1) -~c = 54.4 cm, (V) bc/bo ~ 0.5, (Ik) be~
b o -~ 0.4, ('~) bc/b o = 0.29 [13], and ( m) "short-throated", ( 9 ) "long throated" Venturi
channels [14]

of geometrical parameters). Fig. 2 compares Khafagi's data with Eq. (28) for
straight channel bottom w = 0. Inserting (42), this may equally be expressed as

h/=--y2 (54)

in which only first order terms have been retained as before. It is observed that
agreement with experiments is favorable, provided U ~< 1. For U ~ 1, however,
f
data approach an asymptotic value of the order hc, o~ = --0.4.
An evaluation of discharge coefficient q ~ Q/Qo with Q and Q0 according
to (44) and (45) fa]s when neglecting energy dissipation between critical and
upstream cross-sections (q ~ 1.0 =[= 0.05). Therefore, data has been reconsidered
using Manning-Strickler's formula

V2
J/- K2Rhd/a (55)

in which K is roughness coefficient and R~ hydraulic radius. The average friction

slope is approximately J/.~ = J/,c/2 (upstream ve]ocity Vo ~ Vc).~ thus He
H0 -- AH with A H = J/,a~lX, A x being the distance between the two sections
(K----85 mZl3/s, A x - ~ 0.5 m have been used). From Eq. (44), discharge m a y
b y expressed as
14U
q = Q/Qo = 1 -~ 243 (56)

in which U ~ Hc2/(bcrc) with rc as radius of curvature of the channel side-wall

at the criticall cross-section. I t is noted from Fig. 3 that agreement between
(56) and observations is fair for U ~ 1 (scatter of ~ 1 % ) , but that q (U>~ 1)

12 A c t a M e c h . 54/~--4
168 W. It. !Iager:

1.08
q 9

1.04

;~ It, q 9 9

U
I
0 1 2
Fig. 3. Comparison between Eq. (56) and Khafagi's data [13], [14] regarding discharge
coefficient q; notation as in Fig. 2

again becomes independent of U. The dashed curve in Fig. 3, corresponding to

14U
243
q =-- 1 - t - ~ , (57)
1+--
2

is likely to be valid for U < 4 (as is substantiated by 2 other experiments of

Khafagi not contained in Fig. 3).
It is interesting to note that Bos [3] tries to correlate q with a different param-
eter, namely H/rc, but resulting evaluations scatter widely (4-10%). Obviously,
the present investigation indicates that both (Hire) and (H/bc) are to be considered.
Nevertheless, discharge characteristics in a Venturi channel may approximately
be computed by the traditional approach (Eq. (45)), if a presicion of less than
5% is sufficient. However, precision is raised when taking into account the effects
of frictional losses between upstream (section of measurement) and critical
section, as well as discharge augmentation due to non-parallel streamlines. The
first of these effects is usually eliminated by choosing an upstream channel slope
being approximately half of the critical slope. Fig. 3 then yields resulting head-
discharge relation immediately.

Weir Flow

Among various weir shapes, standard spillways are widely employed in hy-
draulic practice. Owing to the pressure distribution on the weir surface other
weir types as circular or parabolically shaped may only be used for relatively
small dams. The present approach holds for small W = H/R; consequently
the three of weir shapes have analogous hydraulic characteristics for identical W.
From Eq. (46) one has

3
1/ R (ss)
 Critical Flow Condition in Open Channel Hydraulics 169

30 ~

20 ~ -

10 ~

W
0 =
0 1 2
Fig. 4. Free surface slope at cril~ica] point, comparison between Eq. (58) and observations
of (m) Biau [2] (parabolic weir, R-~ 0.88 m), and (4) Ali [1] (st~rtdard shaped weir,
R 1,62Hd) =

in which z c " = --1/R. Fig. 4 shows a comparison between tests on parabolic

and standard shaped weirs according to All [1], Blau [2] and Eq. (58). Note t h a t
deviations between the two are small for W ~ 0.5.
Circular shaped weirs have further been tested in a channel of width b ---- 0.30m
and length 5 m ; test equipment is extensively described in [7]. Investigations
aimed to determine flow surfaces and velocity distributions for weirs of radius
R -~ 4.44 cm, 5.70 cm and 9.02 cm. Typical surface profiles are given in Fig. 5

25

20 40
r-

~ 4 0 ~
15 30- -- ~ 3 0
"- - - 20
-- - - 2 - -

-- - - I - -

10
t
o (f/s}
O(I/s) / ~ i -
10- ~ __

R=4:i--
h (crn}
4Icm~))
4
l ,.- x(cm)
l
v
l
" - x (cm) ~ /

13o 13s 14o ,~s 13o ~o i~o I~o

Fig. 5. Surface profiles for flows over circular weirs with r~dius 1~ = 4.44 c m (left) and
R = 9.02 c m (right) (see also Appendix If)

12"
170 W . g . Hager:

for discharges upto 45 1/s. Tables 1 to 3 (see Appendix II) contain the complete
data. Typical velocity distributions at the crest-section are shown in Fig. 6.
Surface inclinations at the critical cross section are given in Fig. 7 as funct o n
of W = H/R. Note that deviations from our first order prediction (58) (solid
surve) are small for W < 0.5, but that test yield a maximum of h' for W ~ 1.5.
Evidently, the weir becomes insignificant for W ~ ~ , thus h' (W -+ ~ ) ~ 0.

(m/s)
r" i
0 1 2

0
Fig. 6. Typical velocity distributions at the circular weir crest section (R = 4.44 cm,
aQ=151/s, bQ=301/s, cQ=351/s)

.0~
20~ I'l_ 9 9
0 r t i I
0 1 2 3 4
Fig. 7. Free surface slope hc' according to Eq. (58) (solid curve) and from tests on circular
shaped weirs, ( 9 R = 4.44 era, (Q) R = 5.70 era, ( 9 R = 9,02 em

Discharge characteristics over weirs may again be expressed by

Q= q (__:)3,2 b VgH a (59)

in which q = q(W).
Fig. 8 compares the present prediction (51) (in which q = 1-~-22W/81)
with observations described above and with that given by Bos [3]. Evidently,
upstream head must have a certain value Hmi n in order that effects of surface
tension become insignificant (Hmi. ~--0.05 m, say). For smaller heads, one has
to account for the parameter a/~R 2 in which a is surface tension and ~ specific
weight (see [16]).
I t is noted that Bos' curve lies above the present tests for large W; this
must be attributed to the different definition of the head on the weir. While
 Critical Flow Condition in Open Channel Hydraulics 171

1.3 // Oi .....
/ /~ ..............d'"= ,

1,2 / .~...., ..... .,

;I../."' ,,
1.1 /I
A

1 I i I W
0 1 2 3 4
Fig. 8. Comparison between Eq. (59) (solid curve) and tests on circular shaped weirs
(Notation as in Fig. 7); also indicated is experimental curve of Bos [3] (dashed curve).
Open signs refer to flows in which surface tension effects become significant and the dotted
curve corresponds to (60)

Bos takes the upstream flow depth as head on the weir, the present investigation
accounts for the energy head. Further, Bos's curve is valid only for H/(2R) < 1.5,
while some of the present tests exceed this limit.
As is observed from Fig. 9, the present experimental investigation allowed
air to enter from the downstream zone; therefore, the minimum weir crest pressure
will be atmospheric. For standard shaped spillawys, q =- 1.28 under design head
(Cd =- 0.49 when expressing discharge by Q = Cab]/2gH~, [5]), which ,corresponds
to the limiting value in Fig. 8.
For moderate values of W, one may set

22W/81
q(W) : 1+ aw' (60)
1+--
4
which is plotted as dotted curve in Fig. 8.

Fig. 9. Typical experimental arrangement for flow over circular weir in a rectangular
channel (b = 0.3 m, Q = 201/s, R = 4.44 em)
172 W.H. Hager:

Venturi Channels with Weir Shaped Bottom

According to Eq. (53), effects of width contraction in straight channels

and weir shaped bottom in prismatic channels simply add. Therefore, it is rea-
sonable to extend these features on flows having U and W of order unity.
With Eqs. (57) and (60) discharge coefficient may be expressed as

14U/243 22W/81
q(U, W) --= 1 + ~ + 3W" (61)
1+ -- 1+--
2 4

This simple relation permits accurate determination of discharge for given

chi~nnel geometry (b, r, R) and (rueasured) head H. F o r known discharge and
channel geometry, in contrast, one will determined head H approximately with
Eq. (45) (zero order solution), compute the values of U and W, and find a first
order solution for H. This iterative process must be repeated until convergence
is sufficient.

1.1 1.25
q ex @11
/
/
9 /
+/
/

w 4 qth
1 1 /~
a) 0 0.5 b) 1 125

Fig. 10. Comparison of Eq. (61) with Blau's tests in a) Channel KI having combined bottom
and width variation; ( 9 ) b = 0,40 m, ( 9 ) b = 0.60 m, (@) b = 0.97 m; and b) in channel
PVdwithr = 0 . 8 m , R ~ 3 . 2 m , b = 0 . 6 m f o r H < 0.86m

Eq. (61) is tested in Fig. 10 with Blau's data from Karlshorst [2]. Channel
K1 consists of a combination of Venturi channel and local hump with crest
at the minimum width section (r = 1.6 m, R = 1.7 In, approx.). Heads varied
between 0.13 ~ H ~ 0.61 m, and minimum width were b = 0.40 m, b = 0.60 m
and b = 0.97 m. Therefore, Ureax --~ 0.612/(0.4. 1.6) = 0.6, for which q = 1.025
according to (61) by setting W = 0. The effect of width variation on discharge
coefficient q m a y be neglected. Consequently, data has been considered as a
function of W only by multiplying it with unity for W = 0 and by 1.025 for
W = 0.5.
Blau determines the governing head H at the critical section by assuming
an inviscid fluid. The present evaluation, shown in Fig. 10, has corrected heads
Hc b y accounting for the frictional head losses J1 (K = 75 ml/3/s, AL -~ 2 m).
A second comparison of Blau's tests with (61) is shown in Fig. 10b). Data
refers to "channel PVd", of which the critical channel width is b ~ 0.60 m,
 Critical Flow Condition in Open Channel Hydraulics 173

the bottom radius R ---=3.2 m and the radius of side-walls r = 0.80 m. qe, and
(~th refer to observed and computed q-values, and it is noted that agreement
between the two is fair for relatively small values of both, U and W, but that
effective q-values must be expected to become higher for U > 1 and W > 1
as predicted. This fact must be attributed to the neglect of quadratic terms
( U . W) in Eq. (61).

9. Critical Flow in Trapezoidal Channels

Let us pass over to channels of trapezoidal cross-section with straight bottom

profile. Derivatives of the flow profile h(x) at the critical cross-section may
be deduced from Eqs. (25) to (27) by setting w = 0 resulting in

h '~ = u(1 s) (1 + t)
3 -~ 10s -~ 10s 2' (62)

hh"=-u {2~- 1% + 40s 2 ~-(330sa~-10s

+ t(--l~_
10s2)
e + 4s -~ 20s 2 -~ 2083)}.. (63)

Inserting this into Eq. (35) yields

Q2 { u{7 ~_ 47s ~100s2 ~ 70sa ~_ t(1~_ 21s ~_ 60s2 ~ 50sa)}

H=-h~- ~ 1-- 3(3~- 10s~-lOs2) 2 "
(64)

Determination of the third order derivative is lengthy. Note, however, that

h~h'"/(h'u) may be expressed only as a function of s and t. This is evaluated
in Fig. 11 for 0 ~-- s --~ 2 and --0.6 ~ t ~ ~-0.6. I t is observed that all curves
approach an asymptotic value for large s, and that variation of the third order
derivative is significant for s --~ 0.
One of the cases to be treated simpler assumes t ---- 0, thus m = m c = constant
but b/' ~ 0. The curve plotted in Fig. 11 may be approximated by the relation

I I
h2 h"'/(h'u)

-1

S
-2
0 0.5 1 ' 1.5 2
Fig. 11. Term h2h'"/(h'u) as a function of s and t in trapezoidal, nearly horizontal channels
174 W.H. Hager:

h'----u --
5(
3 1 1 + 4.48]" (65)
Eq. (36) now reads

Q2(1+28) {1_] 8 u ( 1"68 )} -1

gb2h3(1+ s)a = ~ t 0.1 -t- s (66)

in analogy to (38) for 8 = 0. Inserting this into (64), the result becomes

1 - - ~-~
7u( 1
8)} ,+s
1.2"~- s 2(1~-28)
H
--=1+
h
1+~-
8~( 1.68)
1 o.1+8

1 -]- -------~s) 1 -- 1 (67)

2(1 + -~ 1 -~-8

( 1 + - -2) 8] 1 + 8 { / 1 - - 5u( 1 2___~/~

- s<2.
-- 2(1 + ~- 1-~ 811

Note that term in brackets is small for u < 1, so that H/h may be expressed
by the governing term
H 1+8
-- ~ 1 -}- . (68)
h 2(1 + 28)

gecalling the definition of s from Eq. (24), one deduces for

s = 4S -- 3 + ]/910
-~ 16S ~- 16S2 ~_ Y2S(1 + ]-'--~]2S/9
t (69)

in which S = mH/b. In analogy, one has for

u=--~- 1+ 1~ ] (70)
so that

-z- -~ 1 + 58/~1 -~ ~+ ~ ] l

To lowest order of approximation, streamlines are horizontal and parallel (U = 0),

whence
g(1 + 8)3 h3 3;5 b i/2~ i s < 2
(72)
S]'
 Critical Flow Condition in Open Channel Hydraulics 175

when inserting the above results into (66). To first order of approximation,
the effect of U on q =: Q/Oo obtains

14U( 4~_.~-/ (73)

q = 1 + -~-~ l 1 + 0.48I"

Note that the effect of U augments discharge, while this of s lowers it. Con-
sequently, discharge may be estimated using equation (72) provided U and S
are smaller than unity.
Finally, from Eq. (57) (which is valid for S = 0), relation (73) m a y be
expressed as

q = 1 + -----------if- 1 -7--O-AS/" (74)

1-4---
2

This has been compared with tests executed by Blau [2] for a trapezoidal channel
having bc =- 0.066 5 m, r e -- 0.3785 m, m~-0.30, and 1 . 5 < Q < 7 5 I / s , for
which heads are 0.05 < H < 0.40 m (Fig. 12). Corresponding U-values lie
between 0.1 and 6.3, and these for S between 0.2 and 1.8. The plot indicates
fair agreement for S < 1, but suggests

14U/243 1 (75)
q=l+ U 1 S
1+--
2
for S > 1.

0.9 ~

0.8 J
0 1 2

Fig. 12. Comparison of Eq. (74) (solid curve) with tests of Blau [2] (e) for trapezoidal
Venturi channel; the dashed curve (Eq. (75)) reproduces observations for S > 1

10. Longitudinal Surface Profile

Critical point is the origin of one-dimensional surface profile computations,

provided flow conditions are subcritical at its upstream and supercritical at
its downstream zone. The usual approach assumes hydrostatic pressure and
 176 W.H. Hager:

uniform velocity d'stribution, such that Eq. (5) is assumed to be valid at a

particular section x = x0. Let z = z(z) be the channel bottom profile, thus
bottom slope Js = --z'. Further, for continuous turbulent flows, energy dissi-
pation is given b y Eq. (55). The governing expression of the surface profile may
be deduced using H ' = J~ - - J/, resulting in [6]

h' - - (76)
1 - - F ~"

Evidently, this becomes singular for F = 1 provided J8 =4=J/.

Slopes of the surface profile have been derived in seeion 3, and it has been
noted that this and corresponding curvature are significant. Consequently,

J3o

80 1()0 1;~0 1~10 160 180 250 220 24o

:Fig. 13. Comparison of first order model (Eq. (4)) with observations of Khafagi [13] (O) for
one-dimensional flow profile in a Vcnturi channel (plan above). All lengths in (em)

the above traditional theory of backwater profiles holds not or only as a first
approximation in the vicinity of the ciritieaI point of the channel. Preciser results
are to be expected using Eq. (4) and H ' = ,/8 - - J / . In order to test not only
the local behaviour of the present approach, but to investigate also the global
flow characteristics, this has been applied to a Venturi channel shown in Fig. 13.
Details of the computational procedure will be explained in detail in a subsequent
investigation, here it is only intended to compare observations of Khafagi [13]
with the numerical model. I t is noted that an excellent agreement between
the present model equations and observational data is achieved. Far upstream,
a train of standing waves is noted, which are described extensively in [10]. At
the downstream zone, submergence produces a hydraulic jump. This particular
phenomenon has not been reproduced by the present approach, since significant
.additional dissipation had to be included into the model equations, see e.g. [8].
 Critical Flow Condition in Opel{ Channel Hydraulics 177

11. Conclusions

The present investigation deals with critical flow condition in open channel
flows. Using the lowest order of approximation for the head-discharge relation,
derivatives of the surface profile at the critical point of the channel are determined.
Inserting these into the first order critical conditions (35) and (36), relations
for discharge in terms of critical flow depth and corresponding energy are deduced.
The present approach includes not only informations at but also in the vicinity
of the critical point of the channel (effect of channel shape on critical flow con-
ditions). Computations refer to trapezoidal cross-section with parabolically
shaped bottom profiles. I t is possible, in principle to investigate arbitrary channel
geometry, but analysis then becomes unwieldy.
Comparison of predictions With observations on Venturi channels and cy-
lindrical weirs indicates fair agreement except for high relative streamline
curvature. Finally, the first order model equation is applied on global flow
features in a Venturi channel, and solutions fit favorably with observational
data.

Appendix I

Energy head H0 given in Eq. (5) corresponds to an extremum with respect

to flow depth provided F = 1. Ho(h) becomes a minimum provided d2Ho/dh 2 > 0
in Eq. (8), thus
3 (dAI2 d2A
-A ~ d h / -- --'7
dh > 0 (77)

in which :ndex " 0 " has been dropped.

Evidently, cross-section at a particular location x is a unique function of
flow depth h, A = A(h) with d A / d h > 0. Depending on the geometrical shape,
the second derivative is either positive, negative or vanishes. Arbitrary cross-
sections may be approximated (at least piece-wise) b y

A = ah m (78)

with a and m as coefficients independent from flow depth h. With dA/dh = amh m-1
and d2A/dh ~ = a m ( m - - 1) h m-2 Eq. (77) obtains

d2H
- - arnhm-2(2m -[- 1) > 0. (79)
dh ~

Since h ~> 0 and dA/dh > 0, one has a > 0 and m > 0; consequently, d2H/dh 2 > 0
and energy minimum always exists.
178 W. H. Hager:

~,
,~.~ ~.
~"~ ~.
~ ~~:, ~.
,..~ I|

11
II oo

I:",

r
H

o~

cr

~
 Critical Flow Condition in Open Channel Hydraulics 179

Acknowledgement

I would like to thank t o my brother, K. Hager, Civil Engr., for having developed the
numerical model used in section 10.

References

[1] Ali, K. H. M. : Flow over rounded spillways. Prec. ASCE, J. Hydraulics Division 98,
I-IY2, 365--379 (1972).
[2] Blau, E.: Die modellm~Bige Untersuchung yon Venturikan~len verschiedener GrSBe
und Form. (Ver6ffcntlichungen Forsehungsanstalt Schiffahrt, Wasser- und Grundbau,
Nr. 8.) Berlin: Akademie-Verlag 1960.
~3] Bos, M. G.: Discharge measurement structures. Laboratorium veer Hydrauliea
en Afvoerhydrologie, Nr. 4. Nederland: Wageningen 1976.
.[4] Chow, V. T. : Open channel hydraulics. Tokyo: McGraw-Hill 1959.
~[5] Dierseh, H.-J., Schirmer, A., Busch, K.-F.: Analysis of flow with initially unknown
discharge. Prec. ASCE, g. Hydraulics Division 103, H u 213--232 (1977).
[6] Hager, W. It. : Stau- und Senkungskurven im Kanalbau. Gas -- Wasser -- Abwasser 61,
157--167 (1981).
~[7] Hager, W. H. : Die Hydraulik yon Verteilkan~len. Thesis presented at the Federal
Institute of Technology, ETH, Ziirich, for the requirements of Doctor of Sciences.
Vol. 6948, Parts I & II, Zfirich, 1981.
[8] Hager, W. H., Hutter, K.: Approximate treatment of hydraulic jump with separation
zone above the flow zone. J. Hydraulic Research 21, 195--204 (1983).
[9] Hager, W. H., Hurter, K.: Approximate treatment of two-dimensional channel,
flow. Acta Mechanica 51, 31--48 (1984).
~10] Hager, W. H., Hurter, K.: On pseudo-uniform flow in open channel hydraulics.
Acta Mechanica 53, 183--200 (1984).
![11] Hager, W. H.: Abflul~erscheinungen in offenen Kan~len. Schweizer Ingenieur und
Architekt (to be published).
i[12] Jaeger, C. : Teehnische Hydraulik. Basel: Birkh~user 1949.
![13] Khafagi, A.: Der Venturikanal. Mittl. Versuchsanstalt Wasserbau, ETH Ziirich,
Vol. I. Ziirich--Leipzig: Leemann & Co. 1942.
i[14] Khafagi, A., Hammad, S. Z. : Velocity and pressure distributions in curved stream-
line flow. Water and Water Engineering 9, 106--115, 436--445 (1954).
i[15] Marchi, G., de: Dispositivi per la misura delle portata dei canali con minime perdite
di quota. L'Encrgia Eiettrica 14, 189--216 (1937).
~16] Matthew, G. D. : Discussion to: Potential flow over spillways at low heads, by Lenau,
C. W., Prec. ASCE, J. Engineering Mechanics 94, EM1, 354--359 (1968).

Dr. W. H. Hager
Chaire de Constructions Hydrauliques, CCH
Institut de Travaux Hydrauliques, I T H , E P F L
CH-1015 Lausanne
Switzerland