Intl. J. River Basin Management Vol. 6, No. 1 (2008), pp.

1321
2008 IAHR, INBO & IAHS
Discharge estimation for equatorial natural rivers with overbank ow
LAI SAI HIN, Lecturer, River Engineering and Urban Drainage Research Centre (REDAC), Universiti Sains Malaysia,
Engineering Campus, 14300 Nibong Tebal, Penang, Malaysia. E-mail: redaclai@eng.usm.my
NABIL BESSAIH, Department of Civil Engineering, Faculty of Engineering, Universiti Malaysia Sarawak (UNIMAS),
94300 Kotta Samarahan, Sarawak, Malaysia
LAW PUONG LING, Department of Civil Engineering, Faculty of Engineering, UNIMAS,
94300 Kota Samarahan, Sarawak, Malaysia
AMINUDDINAB. GHANI, Professor and Deputy Director, REDAC, Universiti Sains Malaysia,
Engineering Campus, 14300 Nibong Tebal, Penang, Malaysia. E-mail: redac02@eng.usm.my
NORAZAZI ZAKARIA, Professor and Director, REDAC, Universiti Sains Malaysia,
Engineering Campus, 14300 Nibong Tebal, Penang, Malaysia. E-mail: redac01@eng.usm.my
MAHYAU SENG, Department of Civil Engineering, Faculty of Engineering, UNIMAS,
94300 Kota Samarahan, Sarawak, Malaysia
ABSTRACT
The estimation of discharge capacity in river channels is complicated by variations in geometry and boundary roughness. Estimating ood ows
is particularly difcult because of compound cross-sectional geometries and because of the difculties of ow gauging. Results are presented of a
eld study including the stage-discharge relationships and surface roughness in term of the Darcy-Weisbach friction factor, f
a
for several frequently
ooded equatorial natural rivers. Equations are presented giving the apparent shear force acting on the vertical interface between the main channel
and oodplain. The resulted apparent friction factor, f
a
is shown to increase rapidly for low relative depth. A method for predicting the discharge
of overbank ow of natural rivers is then presented, by means of a composite friction, f
c
, which represents the actual resistance to ow due to the
averaged boundary shear force and the apparent shear force. Equations are also presented giving the composite friction factor from easily calculated
parameters for overbank owof natural rivers. The results obtained using the methods proposed showthat a signicant improvement has been achieved
compare to the discharge obtained using traditional methods, with an averaged error of 2.7%.
Keywords: Discharge estimation; ow resistance; friction factor; momentum transfer; natural river; overbank ow.
1 Introduction
Whether by nature or design, many rivers exhibit a compound or
two-stage geometry, consisting of a deep central main channel
anked by one or two oodplains. For this type of channel geom-
etry, previous research has shown that the use of conventional
methods for overbank ow hydraulic calculation are incorrect,
as they fail to allow for the loss in conveyance arising from inter-
action between main channel and oodplains (Zheleznyakov,
1965; Knight and Demetriou 1983, Wormleaton et al. 1982,
Prinos and Townsend 1984, and Myers 1987, 1991, Acker,
1992, Lambert and Myers 1998, Myers et al. 2001). This has
become the subject of considerable research in the past 30 years,
covering various aspect of compound channel e.g. ow distribu-
tion, stage discharge relationship, surface roughness, apparent
shear, and discharge estimation. Various methods as well as
Received on February 17, 2007. Accepted on May 25, 2007.
13
empirical formulas have been proposed for discharge calcula-
tion of overbank ow, however none yet commands wide spread
acceptance, due to most of the research carried out for overbank
ow have been laboratory based with idealized conditions.
In natural rivers, the oodplains are often heavily wooded
or vegetated with very complex congurations, the shapes are
irregular, and the ow is usually turbulent with a considerable
mixing. These leading to a completely different roughness con-
ditions than that modeled in laboratory umes. Thus, despite
the extensive researches carried out, full-scale eld experiments
would be the best way to further understanding of overbank ow
in compound river channels, as well as evolving accurate meth-
ods of overbank discharge prediction. At the present stage, eld
study is rare due to overbank ow conditions occur typically
under ood conditions when acquisition of data is difcult and
sometimes dangerous.
14 Lai Sai Hin et al.
As many ood improvement schemes consist of a main chan-
nel with associated oodplains or berms. An overestimate of
discharge capacity at the design stage may lead to ooding
at more frequent intervals than contemplated in these cases.
Moreover, in ood routing through complex river systems, it
is essential that the interaction between the channel and ood-
plain sections be properly modeled. Therefore, the need for
accurate and preferably simple methods of discharge calculation
in compound sections is thus very important. Therefore, a eld
study on discharge estimation of overbank owwas carried out to
derive a method for accurate discharge predictions during ood
events and for a reliable stage-discharge relation for ood control
measures and management schemes.
2 Review
In analyzing the ow through open channels of regular sectional
shape and hydraulic roughness, it is sufcient, in general, to
use the overall hydraulic radius as the parameter, which charac-
terizes the properties of the cross section. It is then possible to
calculate the discharge through the channel fromone of a range of
well-known uniform ow formulas such as the Chezy, Manning,
and Darcy-Weisbach equations, in termof the channel roughness,
slope and hydraulic radius.
However, if the cross-sectional shape is irregular, this can
lead to considerable errors. One particularly important exam-
ple of this occurs when we encounter a compound cross-section
consisting of a deep main channel with associated shallow ood-
plains or berms. In this case, there may be a sudden change of
depth at the transition between the main channel and the ood-
plain. Moreover, the hydraulic roughness of the oodplain is
often greater than that of the main channel. The combined effects
of the greater depth of ow and smaller hydraulic roughness
of the main channel can lead to signicantly higher velocity
than those occurring on the oodplain. This velocity gradient
inevitably results in a lateral mass and momentumtransfer mech-
anismas shown in Figure 1, which can greatly reduce the channel
discharge capacity.
Sellin (1964), was the rst to investigate the momentum
transfer mechanism, which was manifested by a series of vor-
tices having vertical axes, form along each channel-oodplain
interface (Figure 2). Zheleznyakov (1965), (1971), conrmed the
presence of the momentum transfer mechanism, which he called
the kinematic effect, and he demonstrated, under laboratory
conditions, the effects of the mechanismin decreasing the overall
rate of discharge for oodplain depths just above bank-full. As the
oodplain depth increased, the importance of the phenomenon
diminished. Barishnikov and Ivanov (1971) reached similar con-
clusions and found a reduction in the section discharge capacity
of up to 16%, due to momentum transfer effect.
Townsend (1968) undertook similar research in a channel with
a single oodplain and studied the longitudinal and transverse
characteristics of the ow. His results highlighted the tendency
of the turbulence at the interface to disperse laterally across the
oodplain. This illustrated the ability of the series of vortices
Figure 1 Mechanisms of overbank owin a straight compound channel
(after Shiono and Knight, 1991).
Figure 2 Vortices form at the interface of main channel and oodplain.
present in the mixing zone to transport the nal sediment fractions
from the fast owing deep channels of rivers on to the ood-
plains in time of ooding. Based on the results, he concluded
that for small oodplain depths, both the stream-wise and cross-
stream turbulence intensities at the channel-oodplain interface
were signicantly higher under interacting (compound channel)
ow conditions than the corresponding values obtained for the
equivalent isolated (separate channels) conditions.
Myers & Brennan (1990) found that the mechanism retards
channel velocity and discharge, while increasing the correspond-
ing parameters on the oodplain. The most notable feature of
Discharge estimation for equatorial natural rivers with overbank ow 15
these relationships is the discontinuity at bankfull depth, with a
reduction in discharge as depth rises just above the bankfull value.
If ow depth continues to rise, the oodplain discharge and ow
velocity will increase rapidly, to a point where main channel and
oodplain are roughly equal in carrying capacity. This equaliza-
tion of discharge and velocity results in a consequent decrease
in momentum transfer from main channel to oodplain and may
lead to a reversal in the direction of momentum transfer at larger
depths.
Prinos and Townsend (1985) presented that when the ow
in a river channel rises above the bankfull stage and inundates
the adjacent oodplain areas, momentum is transferred across
the junction regions separating the deep and shallowzones. If the
velocity gradient across the junction region is large, this trans-
fer mechanism will inuence both velocity and boundary shear
force distributions and also the turbulence characteristics of the
junction regions. Under such conditions, the accuracy of most
conventional discharge estimation methods is reduced. This is
largely because conventional discharge estimation methods do
not account for the appreciable apparent shear force occurring at
the main channel-oodplain interfaces.
The studies of Wormleaton et al. (1982), Myers (1987) and
Myers et al. (2001) illustrated the error encountered by vari-
ous traditional calculation methods for discharge estimation in
compound channels due to momentum transfer. The single chan-
nel method (SCM) is found seriously in error when the ow
is just overbank. The divided channel method with horizontal
(HDCM) and vertical (VDCM) division lines are either under or
over-estimated the ow in compound channels, with errors of up
to 45% in the rough oodplain case. Smooth main channel and
oodplain are more accurately modeled by this method with 10%
error. While the river discharge errors peak at almost 30%. It is
clear that more reliable and preferably simple method of analysis
is needed for discharge estimation in such rivers.
3 Field study and data collection
The study was carried out in three natural rivers namely River
Senggai, River Senggi (B) and River Batu located in Kuching, the
capital cityof Sarawakstate, Malaysia. These rivers were selected
due to serious oods occurrence during Monsoon season in the
past few years. Extensive ood data from River Main (2002) in
North Ireland has also been obtained for comparison.
The selected rivers are shown in Figures 36. It presents that
the rivers are almost straight and uniform in cross section, free
from backwater and tidal effect. Table 1 shows the geometrical
properties and surface conditions of the rivers at the gauging
stations for comparison.
Flow gauging of the rivers was carried out from an adjustable
bridge built across the rivers, using the velocity-area method, in
which an electromagnetic ow meter was used to measure point
velocity at 20%, 40%, 60% and 80% of ow depth at up to 20
verticals across the sections. The ow depths and point veloc-
ities were measured to an accuracy of 0.0005 m (0.5 mm) and
0.0001 m/s respectively. For each measuring point, 36 reading
Figure 3 Morphological cross-section of River Senggai.
Figure 4 Morphological cross-section of River Senggi (B).
Figure 5 Morphological cross-section of River Batu.
were taken and averaged to give a mean point velocity to reduce
the error due to variation in water ow. Some 20 discharges were
recorded for each river, covering a wide range of inbank and
overbank ows.
16 Lai Sai Hin et al.
Figure 6 Morphological cross-section of River Main.
4 Theory considerations
The Darcy-Weisbach equation is commonly used to express ow
in open channels, following the recommendation of theAmerican
Society of Civil Engineers (ASCE) in 1963. Therefore it has been
adopted in this study, in which
V =

2gDS
0
f

1/2
(1)
where V is the average cross-sectional velocity, D (=
P
B
R)
is the hydraulic diameter, S
o
is the bed slope, g is grav-
itational acceleration, and f is the Darcy-Weisbach friction
factor.
In applying the Darcy-Weisbach equation above, the great-
est uncertainty lies in the determination of the Darcy-Weisbach
friction factor, f; for there is no exact method of selecting the
f value, especially under overbank ow conditions of natural
rivers, as limited published data is available and heterogeneous
oodplains may be difcult to characterize. The picture is fur-
ther complicated by the existent of an apparent shear due to
the momentum transfer at the interface region as mentioned
above.
In order to account for the increase of resistance due to
momentum transfer in discharge calculation of overbank ow,
Table 1 Geometrical properties and surface conditions.
Geometrical properties River Senggai River Senggi (B) River Batu River Main
Bankfull depth, h(m) 1.060 1.306 1.544 0.900
Top width, B (m) 5.285 5.500 5.150 13.700
Aspect ratio, B/h 4.986 4.211 3.335 15.222
Bed slope main channel , S
0
0.0010 0.0010 0.0016 0.0030
Bed slope left oodplain, S
L
0.0010 0.00085 0.0013 0.0030
Bed slope right oodplain, S
R
0.0010 0.00085 0.0013 0.0030
Surface condition main channel Erodible soil Erodible soil large boulder coarse gravel
Surface condition side bank Erodible soil long vegetation Erodible soil large boulder
Surface condition oodplain long vegetation long vegetation long vegetation short vegetation
the invisible apparent shear can be quantied as shown below:
Assuming uniform ow in a simple channel section with
trapezoidal shape,
Driving force = Resisting force
W sin =
0
Px
but
0
= C
f

V
2
2
, C
f
=
f
4
, W = gA, and sin = S
0
. For a unit
distance,
fPV
2
= 8gAS
0
In the case of overbank ow with vertical interface, an addi-
tional resisting force due to apparent shear must be added in terms
of an apparent friction factor, f
a
, velocity gradient at the interface
region of main channel and ood plain, V
2
and the interface
perimeter, y (Christodolou and Myers, 1999). Then, considering
the balance of forces along the owdirection in the main channel
leads to:
f
am
=
8gA
m
S
0
f
mp
PV
2
m
2y

V
2
(2)
Similarly, for the oodplain region,
f
af
=
f
fp
PV
2
f
8gA
f
S
0
y

V
2
(3)
For Equations 2 and 3 above, the owvelocities V
m
and V
f
are
frommeasured values, the channel slopes S
0
, the interface wetted
perimeter y, the sub-sectional area Aand the sub-sectional wetted
perimeter P can be obtained easily from geometrical measure-
ment. The boundary friction factors, f
mp
and f
fp
can be obtained
by extrapolation using inbank data, assuming that no interac-
tion existed between the main channel and oodplain. Since
the measured velocities are strongly inuenced by the ood-
plain and main channel interaction, the velocity differences (V)
between subsections were obtained, not from the measured val-
ues, but fromestimated velocities for each sub-sections using the
Darcy-Weisbach equation.
The values of f
am
and f
af
calculated using Equations 7 and
8 above should in principle have the same values (= f
a
) but
acting in opposite directions, and depend on the geometrical
cross section (B, b), owdepth (H, Hh), geometrical relation-
ships (M
f
, M
m
, R
f
, R
m
), and boundary roughness (f
fp
, f
mp
)of
Discharge estimation for equatorial natural rivers with overbank ow 17
the main channel and oodplain. In dimensionless form:
f
a
,

B
b
,
H h
H
,
M
f
M
m
,
R
f
R
m
,
f
fp
f
mp

(4)
This relationship can be sorted using a multiple non-linear regres-
sion analysis approach to give the relationships on which f
a
, are
depends. This approach is preferred as it allows the evaluation of
f
a
in a predictive sense, i.e. without the need for measuring the
velocities of the subsections.
5 Velocity distribution
The lateral distributions of averaged depth velocity at the gauging
site of the rivers are shown in Figures 79. These gures clearly
show that the maximum ow velocity occurs in the central of
main channel region, which decreases towards the side banks
direction, and increases with the increase of owdepth. Whereas,
the ow velocity on the oodplains is found near to zero in all
cases even at high overbank ow due to the retention effects of
the oodplain vegetations. As a result, a large velocity gradient
0
0.2
0.4
0.6
0.8
1
1.2
59 64 69 74
Lateral distance (m)
A
v
e
r
a
g
e
d

d
e
p
t
h

v
e
l
o
c
i
t
y
,
V

(
m
/
s
)
H=1.048
H=1.068
H=1.174
H=1.128
H=1.155
H=1.228
H=1.288
H=1.55
H=1.587
H=1.643
main channel floodplain floodplain
Figure 7 Averaged depth velocity for overbank ow of River Senggai.
0
0.2
0.4
0.6
0.8
1
1.2
1.4
54 56 58 60 62 64 66 68 70
Lateral distance (m)
A
v
e
r
a
g
e
d

d
e
p
t
h
v
e
l
o
c
i
t
y
,
V

(
m
/
s
)
H=1.543
H=1.695
H=1.92
H=2.02
H=2.243
H=2.42
main channel floodplain floodplain
Figure 8 Averaged depth velocity for overbank ow of River Batu.
0
0.1
0.2
0.3
0.4
0.5
0.6
2 4 6 8 10 12 14 16 18 20
Lateral Distance, (m)
A
v
e
r
a
g
e
d

d
e
p
t
h

v
e
l
o
c
i
t
y
,
v

(
m
/
s
)
h=0.698
H=1.088
H=1.35
H=1.475
H=2.13
H=2.278
main channel floodplain floodplain
Figure 9 Averaged depth velocity for overbank ow of River
Senggai (B).
is found between the main channel and oodplain, due to the
different in depth and surface roughness.
At the interface region between the main channel and ood-
plain, the velocity is found to decrease rapidly, i.e. from very
high main channel velocity to near or sometimes smaller than
the oodplain velocity. This is due to the signicant momentum
transfer and apparent shear existed between the two zones. These
interactions tend to retard the ow at the interface region of main
channel, and reduced the cross sectional discharge capacity.
6 Flow resistance results
By substituting D = (P/B)R to Equation 1, the resistance to ow
for the main channel region of the investigated rivers has been
calculated in term of the Darcy-Weisbach friction factor, f
m
as
shown in Figure 10. For inbank ow, i.e. (H h)/H < 0, the
f value for the selected equatorial rivers was found to be in the
range of 0.2 at low ow, and it decreases linearly with ow
depths towards the bankfull level, due to the decrease in relative
roughness in the main channel region. An exception to this is
in River Senggi (B), which experienced a slight increment of
f
m
value due to the vegetation at the side banks. The respective
values for River Main were found comparatively small with f
m
values equal to 0.04.
The overbank ow is characterized by an increased of rough-
ness value. As the surface properties in the main channels
remained the same, such an increment can be considered due
to the apparent shear mentioned earlier, which slows down the
ow in main channel. For River Senggai and River Batu with
obvious roughness differences between the main channel and
oodplain, the increase in roughness starts when the ow is just
overbank. For example, the f values for River Senggai increased
from 0.157 at the bankfull level to 0.208 at (H h)/H = 0.082,
before they continue to reduce at higher depths. For River Senggi
(B) and River Main, the increase only starts after a certain stage
of overbank ow, i.e. (H h)/H = 0.166 for River Senggi (B)
and (H h)/H = 0.302 for River Main due to the bank vegeta-
tion and cross-sectional geometry which prevent the interaction
occurring before the effective bankfull level.
For the oodplain regions, the velocities collected from eld
measurements during overbank ow are always close to zero,
except under very high overbank ow depth. As a result, the
f
f
values obtained are very high. Such values are known to be
0
0.2
0.4
0.6
-0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Depth Ratio, (H-h)/H
D
a
r
c
y
-
W
e
i
s
b
a
c
h
,

f
Senggai
Batu
Senggi (B)
Main
Figure 10 Variation of Darcy-Weisbach friction factor, f with depth of
ow.
18 Lai Sai Hin et al.
seriously affected by the ponding effects of the oodplain veg-
etation, and they are not suitable for use in representing the actual
oodplain roughness.
7 Apparent friction factor, f
a
The apparent friction factor, f
a
calculated using Equation 2 is
shown in Figure 11. These results show that a large apparent
shear exists at the interface region especially when the ow is
just overbank. In an extreme case, the value of apparent friction
factor at bankfull level for River Senggai is found equal to 151.11.
For other rivers such as River Senggi (B), River Batu and River
Main the maximum apparent friction factors found are 41.37,
15.34 and 9.08 respectively.
When the ow continues to rise, the value of f
a
is found to
decrease with depth, while the velocity difference is increased
with depth in all cases as shown in Table 2. This supports the
nding that the apparent friction factor is inversely proportionate
to V
2
as reported by Christodolou and Myers 1999.
When the apparent friction factor, f
a
is compared with the
averaged boundary friction factor, f
m
as shown in Table 2, the
maximumf
a
/f
m
ratios obtained for River Senggai, River Senggi
(B), River Batu, and River Main are, 962.5, 177.58, 56.16, and
444.95, respectively. This implies that the apparent friction factor
at the interface region is many times greater than the averaged
boundary friction factor.
0
40
80
120
160
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Depth Ratio, (H-h)/H
A
p
p
a
r
e
n
t

F
r
i
c
t
i
o
n

F
a
c
t
o
r
,

f
a
River Senggai
River Senggi (B)
River Batu
River Main
Figure 11 Variation of apparent shear with depth for overbank ow of
natural rivers.
Table 2 Comparison of averaged boundary friction factor (f
m
) and apparent friction factor (f
a
) for overbank ow of natural rivers.
River Senggai River Batu River Senggi (B) River Main
(H h)/H f
a
f
a
/f
m
V (H h)/H f
a
f
a
/f
m
V (H h)/H f
a
f
a
/f
m
V (H h)/H f
a
f
a
/f
m
V
0.008 151.1 962.5 0.277 0.089 9.076 56.2 0.476 0.033 41.370 177.6 0.269 0.053 15.341 444.9 1.167
0.060 18.509 140.8 0.297 0.119 6.530 41.1 0.489 0.0777 17.697 75.6 0.276 0.062 12.021 371.4 1.176
0.082 12.366 102.2 0.319 0.196 3.596 23.8 0.527 0.1153 11.966 50.9 0.279 0.109 6.631 207.0 1.262
0.097 9.585 84.1 0.334 0.214 3.200 21.4 0.537 0.1894 7.051 29.7 0.288 0.163 4.450 140.8 1.277
0.113 7.682 72.1 0.350 0.236 2.817 19.1 0.547 0.2701 5.391 22.5 0.300 0.240 2.955 95.7 1.306
0.137 5.549 57.9 0.376 0.266 2.368 16.3 0.565 0.3284 4.326 18.0 0.310 0.302 2.051 67.9 1.323
0.177 3.310 42.1 0.429 0.312 1.881 13.3 0.592 0.3795 3.648 15.1 0.321 0.375 1.730 59.2 1.342
0.223 1.861 30.5 0.510 0.338 1.445 10.3 0.608 0.4027 3.413 14.1 0.326 0.471 1.294 47.1 1.368
0.284 1.007 23.2 0.643 0.362 1.262 9.1 0.626 0.4271 2.993 12.3 0.332 0.581 0.869 35.4 1.434
0.316 0.758 20.2 0.715
The results obtained also shows that the ow in compound
natural rivers is different from the ow in laboratory compound
channel, in which for compound channels, most of the results
reported in previous studies as described in Section 2 show that
the velocity on oodplain increases rapidly with depth of ow,
as a result, the velocity difference and apparent friction factor
between the main channel and oodplain becomes smaller and
approaching zero at depth (H h)/H 0.5. However, in this
study, it has beenshownthat the velocitydifference for compound
natural rivers is increasing with depth, due to the very rough sur-
face conditions and ponding effects on the ood plain regions
as mentioned above. The values of f
a
is also found to remain high
even at higher degree of overbank ow, e.g. for a relative depth
of (H h)/H = 0.3, the apparent friction factor for River Sen-
ggai, River Senggi (B), River Batu, and River Main is about 20,
20, 14, 68 time lager than the averaged boundary friction factor,
probably due to the slowdown effects from the slow-moving
ood plain ow.
A theoretical analysis of the complicated turbulent ow pat-
terns giving rise to the apparent shear force at the interface would
be intractable unless certain rather sweeping assumptions were
made. As an alternative, a statistical approach at least might be
able to indicate the important factor determining the apparent
shear force, and at best provide a simple means of determin-
ing it from easily calculated parameters of the geometrical and
hydraulic characteristics of the channel. Therefore the values of
f
a
calculated have been tested against several independent vari-
ables such as B/b, (H h)/H, M
f
/M
m
, R
f
/R
m
, and f
fp
/f
mp
.
The relationship sought in the analysis is of the form
f
a
= a

H h
H

b1

M
f
M
m

b2

B
b

b3

f
f
f
m

b4

R
f
R
m

b5
(5)
in which a, and b1 b5 are constants.
From the analysis, the correlations between f
a
and each inde-
pendent variable: (H h)/H, R
f
/R
m
, B/b, f
fp
/f
mp
, M
f
/M
m
are 0.952, 0.748, 0.270, 0.122, and 0.071, respectively. This
indicates the value of f
a
is strongly depends on the depth ratio,
and moderately strong depends on the hydraulic radius ratio
between ood plain and main channel. Whereas the rather weaker
Discharge estimation for equatorial natural rivers with overbank ow 19
R
2
= 0.9909
0
10
20
30
40
50
0 10 20 30 40 50
Observed f
a
E
s
t
i
m
a
t
e
d

f
a
Line of perfect agreement
Figure 12 Apparent friction factor, f
a
observed and predicted using
Equation 6.
correlations between f
a
and B/b, f
fp
/f
mp
, M
f
/M
m
would indi-
cate that these variables will have lesser degree of inuent to the
values of f
a
.
However, when the correlation is carried out with a combina-
tion of 2, 3, 4 and 5 variables, better results are obtained. This
shows that all the variables are signicant in the determination of
an accurate value of f
a
. Therefore, all the independent variables
are retained, and the nal regression equation obtained is:
f
a
= 0.82

H h
H

2

M
f
M
m

11.5

B
b

0.55

f
fp
f
mp

0.85

R
f
R
m

0.31
+0.1255 (6)
The coefcient of multiple correlations is found to be 0.995. The
determination coefcient of 0.991 indicates that the preceding
equation would explain 99.1% of the total deviation in f
a
.
The values of f
a
estimated by equation 6 are compared with
the observed values in Figure 12. The close agreement of the
data, fromooded natural rivers, and over a range of geometrical
conditions, is encouraging.
8 Composite friction factor
Based on the estimated boundary shear stress, f
mp
, and the appar-
ent friction factor, f
a
,the weighted ratio on how the composite
(actual) friction factor, in the main channel, f
c
depends on f
mp
andf
a
has beenfoundinterms of the associatedwettedperimeters
P
m
and y, as:
f
c
= 1.03

0.117

P
m
P
t

7.06
f
mp
+0.507

2y
P
t

1.08
f
a

0.008

(7)
in which P
t
is the total wetted perimeter
The results obtained using Equation 7 are compared with the
observed main channel friction factor in Figure 13. It can be seen
that that most of the main channel friction factors can be cal-
culated accurately, with a coefcient of determination of 0.994.
However, as the data are based on only four natural rivers, only
limited reliance can be placed on this. Additional data is needed
y = x + 3E-06
R
2
= 0.9938
0
0.1
0.2
0.3
0.4
0.5
0.6
0 0.1 0.2 0.3 0.4 0.5 0.6
Observed fc
E
s
t
i
m
a
t
e
d

f
c
Figure 13 Observed and estimated composite friction factor using
Eq. 7.
before more reliable relationships for calculating the apparent
friction factor and composite friction factor can be found. Nev-
ertheless, the writers consider that this analysis has helped to
pinpoint the important causative factors in determining the fric-
tion factors, and at best provide a simple means of determining
it from easily calculated parameters.
9 Discharge estimation
As mentioned above, the velocities on oodplain are close to zero
in most cases, and it has little contribution to the overall discharge
capacity. Therefore, the discharge estimation carried out was
focused on the main channel region, to avoid the overestimation
at the oodplain region, and to give a simple but reasonable esti-
mate of discharge for such river channels. The results of obtained
using the composite friction factor estimated, f
c
were plotted in
Figs. 1417. Also plotted are the observed data and the discharge
estimated using the conventional Manning equation with bank-
full roughness for comparison. From the results, it can be seen
that the discharges are signicantlyover-or-under estimatedusing
the conventional method with an averaged error of 12.5%, 19.2%,
16.3%, 44.9% for River Senggai, River Senggi (B), River Batu,
and River Main, respectively. In extreme case, the discharge esti-
mated for River Main (i.e. H h/H = 0.58) is almost twice of
the observed discharge with a maximum error of 91.2%. These
highlighted the danger inherent in the conventional practices of
extrapolating inbank data for the analysis of overbank ows.
0
0.5
1
1.5
2
2.5
3
3.5
4
-0.2 -0.1 0 0.1 0.2 0.3 0.4
Depth Ratio, (H-h)/H
D
i
s
c
h
a
r
g
e
,

Q
(
m
3
/
s
)
Conventional method
Observed
Composite friction factor
Figure 14 Comparison of observed and predicted main channel dis-
charge for River Senggai.
20 Lai Sai Hin et al.
0
1
2
3
4
-0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5
Depth ratio, (H-h)/H
D
i
s
c
h
a
r
g
e
,

Q

(
m
3
/
s
)
Conventional method
Observed
Composite friction factor
Figure 15 Comparison of observed and predicted main channel dis-
charge for River Senggi (B).
0
1
2
3
4
5
6
-0.2 -0.1 0 0.1 0.2 0.3 0.4
Depth ratio, (H-h)/H
D
i
s
c
h
a
r
g
e
,

Q

(
m
3
/
s
)
Traditional method
Observed
Composite frictionfactor
Figure 16 Comparison of observed and predicted main channel dis-
charge for River Batu.
0
20
40
60
80
100
-0.2 0 0.2 0.4 0.6
Depth Ratio,(H-h)/H
D
i
s
c
h
a
r
g
e
,

Q

(
m
3
/
s
)
Traditional method
Observed
Composite frictionfactor
Figure 17 Comparison of observed and predicted main channel dis-
charge for River Main.
From the graphs plotted, it can also be seen that the results
are signicantly improved using the estimated composite friction
factor, compared to those estimated using traditional methods, in
which all the discharge estimated using the composite friction is
seen to match closely to the observed data. Table 3 further shows
the signicant improvement achieved using the proposed method
in which the averaged error for River Senggai, River Senggi (B),
River Batu and River Main have been reduced tremendously to
Table 3 Comparison between discharges estimated using traditional
method, and composite friction factor, f
c
.
River Senggai River Senggi (B) River Batu River Main
Stat. Calc. Trad. f
c
Trad. f
c
Trad. f
c
Trad. f
c
Max. Error (%) 28.44 8.82 56.16 12.55 24.88 5.93 91.21 8.83
Ave. Error (%) 12.52 2.37 19.20 3.60 14.15 1.38 44.91 3.38
RMSE (%) 15.49 3.54 26.00 4.77 16.95 2.12 49.43 4.38
2.37, 3.60, 1.38 and 3.38%only. Other statistical calculation such
as the maximum error, and root mean square error (RMSE) car-
ried out also supported positively the consistency and accuracy of
the proposed method in discharge estimation of overbank ows.
10 Conclusions
Based on extensive data collected from four frequently ooded
natural rivers and results obtained, it can be concluded that:
1. A very strong apparent shear is found at the interface region
of main channel and oodplain. The apparent shear has been
quantied in the form of apparent friction factor f
a
, and it is
found that the apparent shear is maximized when the ow is
just overbank but reduced at higher depths.
2. The use of traditional method for overbank ow estimation
was found to be signicantly in error, with an averaged over-
or-under estimation of 22.7%, due to the interaction between
the main channel and oodplain ows.
3. Astatistical method is able to provide a simple means of deter-
mining the apparent friction factor and composite friction
factor from easily calculated parameters for overbank ow
of natural rivers. The results obtained using the methods pro-
posed show that a signicant improvement has been achieved
compare to those obtained using traditional methods, with
an averaged error of 2.7%. However, more data is needed to
further generalize and verify the equations.
Notation
= Fluid density

0
= Mean shear stress on the bottom
x = Flowing distance in streamwise direction
V =Velocity gradient at the interface region of main
channel and ood plain
A =Wetted area
A
m
, A
f
=Wetted areas for the main channel and ood plain
b = Bankfull width
B =Top width
C
f
= Resistance coefcient
D = Hydraulic diameter
f = Darcy-Weisbach friction factor
f
a
=Apparent friction factor
f
c
= Composite friction factor
Discharge estimation for equatorial natural rivers with overbank ow 21
f
am
, f
af
=Apparent friction factors for the main channel and
ood plain
f
f
, f
m
= Observed Darcy-Weisbach friction factors for the
main channel and ood plain
f
fp
, f
mp
= Estimated Darcy-Weisbach friction factors for the
main channel and ood plain
h = Bankfull depth
H = Depth of ow
HDCM = Horizontal Divided Channel Method
M =A geometrical parameter (= 2P/B)
M
f
, M
m
= Geometrical parameters, for ood plain and
main channel
P = Cross sectional wetted perimeter
P
m
= Main channel wetted perimeter
R = Hydraulic radius
R
f
, R
m
= Hydraulic radius for ood plain and main
channel
S
0
= Longitudinal bed slopes
SCM = Single Channel Method
V = Mean velocity
V
m
, V
f
= Mean velocities for the main channel and ood
plain
VDCM =Vertical Divided Channel Method
W = Gravitational weight of uid
y = Interface wetted perimeter
Acknowledgement
The authors would like to express their gratitude to Dr. W. R. C.
Myers, University of Ulster, U.K., for providing data from River
Main in Northern Ireland.
References
1. Zheleznyakov, G.V. (1965). Relative decit of mean
velocity of instable river ow; kinematic effect in river beds
with oodplains, Proceedings of the Eleventh International
Conference of the International Association for Hydraulic
Research, Leningard, U.S.S.R.
2. Knight, D.W. and Demetriou, J.D. (1983). Floodplain
and Main Channel Flow Interaction, Journal of Hydraulic
Engineering, ASCE, 109(8), 10731092.
3. Wormleaton, P.R., Allen, J. and Hadjipanos, P. (1982).
Discharge Assessment in Compound Channel Flow, Jour-
nal of Hydraulic Division, ASCE, 108(9), 975994.
4. Prinos, P. and Townsend, R. (1984). Comparison of
Methods for Predicting Discharge in Compound Open
Channels, Advances in Water Resources, 7(Dec), 180187.
5. Myers, W.R.C. (1987). Velocity and Discharge in Com-
pound Channels, Journal of Hydraulic Engineering, ASCE,
113, 753766.
6. Myers, W.R.C. (1991). Inuence of Geometry on Dis-
charge Capacity of Open Channels, Journal of Hydraulic
Engineering, ASCE, 117, 676680.
7. Ackers, P. (1992). Hydraulic design of two-stage chan-
nels, Proceedings of the Institution of Civil Engineering,
Water, Maritime and Energy, 96, 247257.
8. Lambert, M.F. and Myers, W.R. (1998). Estimating
the Discharge Capacity in Straight Compound Channels,
Proceedings of the Institution of Civil Engineering, Water,
Maritime and Energy, 130, 8494.
9. Myers, W.R., Lyness, J.F. and Cassells, J. (2001). Inu-
ence of Boundary on Velocity and Discharge in Compound
Channels, Journal of Hydraulic Research, 39(3), 311319.
10. Shiono, K. and Knight, D.W. (1991). Turbulent Open
Channel Flows with Variable Depth Across the Channel,
Journal of Fluid Mechanics, 222, 617646.
11. Sellin, R.H.J. (1964). A laboratory investigation in the
interaction between ow in the channel of a river and that
of its oodplain, La Houille Blanche, No. 7, 793801.
12. Zheleznyakov, G.V. (1971). Interaction of channel and
oodplain streams, proceedings, Fourteenth International
Conference of the International Association for Hydraulic
Research, Paris, France.
13. Barishnikov, N.B. and Ivanov, G.V. (1971). Role of
oodplain in ood discharge of a river channel, Proceed-
ings, 14th International Conference of the International
Association of Hydraulic Research, Paris, France.
14. Townsend, D.R. (1968). An investigation of turbulence
characteristics in a river model of complex cross-section,
Proceedings of the Institute of Civil Engineers, Vol. 40,
155175.
15. Myers, W.R.C. and Brennan, E.K. (1990). Flow
Resistance in Compound Channel, Journal of Hydraulic
Research, 28(2), 141155.
16. Prinos, P. and Townsend, R. (1985). Friction factor rela-
tionships for owin compound channels. Advances in water
resources, University of Birmingham(Tebbutt T.H.Y. (ed.)).
Elsevier, London, 332339.
17. Myers, W.R.C. (2002). River Main Data courtesy of Myers
WRC, Department of Civil Engineering and Transport,
University of Ulster at Jordanstown.
18. Task Force. (1963). Friction Factor in Open Channels,
Report of the ASCE Task Force, Journal of Hydraulic
Engineering, ASCE, 89(2), 97143.
19. Christodolou, G.C. and Myers, W.R.C. (1999).
Apparent friction factor on the oodplain-main chan-
nel interface of compound channel sections, Proceed-
ings of 28th Congress of the International Association
for Hydraulic Research, Craz, Austria, SE3, A379,
http://www.iahr.org/membersonly/grazproceedings99/doz/
000/000/367.htm.