Ain Shams Engineering Journal (2013) 4, 625632

Ain Shams University

Ain Shams Engineering Journal

www.elsevier.com/locate/asej
www.sciencedirect.com

CIVIL ENGINEERING

Accurate gradually varied ow model for water

surface prole in circular channels
Ali R. Vatankhah

a,*

, Said M. Easa

Department of Irrigation and Reclamation Engineering, University College of Agriculture and Natural Resources,
University of Tehran, P.O. Box 4111, Karaj 31587-77871, Iran
b
Department of Civil Engineering, Ryerson University, Toronto, Ontario, Canada M5B 2K3
Received 12 October 2012; revised 13 January 2013; accepted 24 January 2013
Available online 7 March 2013

KEYWORDS
Gradually varied ow;
Circular channels;
Analytical solution;
Dimensionless variables

Abstract The paper presents an accurate approximation of the Froude number (F) for circular
channels which is part of the gradually varied ow (GVF) equation. The proposed approximation
is developed using optimization technique to minimize the relative error between the exact and estimated values, resulting in a maximum error of 0.6% compared with 14% for the existing approximate method. The approximate F is used in the governing GVF equation to develop an exact
analytical solution of this equation using the concept of simplest partial fractions. A comparison
of the proposed and approximate solutions for backwater length shows that the error of the existing
approximate solution could reach up to 30% for large normal ow depths.
 2013 Ain Shams University. Production and hosting by Elsevier B.V.
All rights reserved.

1. Introduction
Accurate estimation of the water surface prole in gradually
varied ows is important in the planning and design of water
works. Numerous studies on gradually varied ow (GVF) have
been conducted for different types of open channels [111].
However, analytical and semi-analytical studies of GVF related to circular open channels are limited. Chow [12] developed a semi-analytical solution for circular channels using
* Corresponding author. Tel.: +98 02632221119.
E-mail addresses: arvatan@ut.ac.ir (A.R. Vatankhah),
@ryerson.ca (S.M. Easa).
Peer review under responsibility of Ain Shams University.

Production and hosting by Elsevier

seasa

constant hydraulic components. To improve Chow solution,

Zaghloul [13] derived mathematical expressions for the
hydraulic exponents as functions of the GVF depth using
numerical integration. Keifer and Chu [14] and Nalluri and
Tomlinson [15] used a method that did not include hydraulic
components. Nalluri and Tomlinson [15] modied Keifer and
Chus method using the subtended angle at the center instead
of the relative depth. Both solutions used tabulated ow functions and their interpolated values which are not vary accurate.
Zaghloul [16] developed a computer model to compute the
GVF length for circular sections. The results were presented
in a tabular form. Zaghloul and Shahin [17] also developed a
package using the Lotus 1-2-3 to calculate the geometric and
hydraulic properties of an open circular channel. The results
were expressed as dimensionless quantities and were presented
in tabular and graphical forms. Zaghloul [18], using variable
pipe roughness, developed a computer model to calculate the

2090-4479  2013 Ain Shams University. Production and hosting by Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.asej.2013.01.005

626

A.R. Vatankhah, S.M. Easa

length of the GVF prole using the step and integration

methods.
Hager [19,20] presented approximate expressions for uniform and critical ows and obtained a simple form for the
GVF differential equation that can be analytically integrated.
However, he implemented three types of approximations
related to the cross section area, the GVF equation, and the
integration procedure. This paper presents an accurate approximation of the Froude number for circular channels along with
an exact analytical solution of the water surface prole and
compares the results with Hagers approximate (semi-analytical) solution.
The basic governing equation of the steady GVF in open
channels is given by [22,23]
dy S0  Sf

dx
1  F2

F2

Q2 4
g ;
gD5

Hagers model

where Q = discharge of the channel (m3/s), n = Manning

roughness coefcient (m1/3 s), and D = diameter of the channel. From Eq. (2) the uniform ow depth, yn, can be expressed
as [21]
(
s!)1=2
2
nQ
3  9  28 1=2 8=3
yn D
7
S0 D

2. Accurate GVF approximation for circular channels

The exact formulas for T and A of a circular channel can be
obtained as
T D sinh=2

A D h  sinh=8

h 2 cos1 1  2g

where h is the central angle subtended by the end points of the

water surface. Substituting for T and A from Eqs. (6) and (7)
into Eq. (4) gives
F2e

where g = gravitational acceleration (m/s2), T = width of

water surface (m), and A = cross section area of ow (m2).

Figure 1

exact model

To aid the integration of the GVF equation, the exact formula

of Eq. (9) was approximated by
F2

Q2
ag5 bg3 cg4
gD5

10a

where a, b, and c are coefcients determined by optimization. Noting that h of Eq. (9) is a function of g, different values of g ranging
from 0.1 to 0.92 were used and the coefcients that minimized the
maximum absolute relative error between the exact and approximate Froude numbers of Eqs. (9) and (10a), (Fe  F)/Fe, were
determined. The approximate formula is given by
F2

Q2
0:64g5 0:405g3
gD5
0:834g4 ;

proposed approximate model

10b

The maximum relative error of Eq. (10b) is about 0.6% in the

range of 0.1 < y/D < 0.92 (Fig. 2).
From Eq. (10b) the critical ow depth, gc = yc/D (where
F = 1) can be determined by
1

Q2 T
gA3

Q2 sinh=2


gD5 hsin h 3
8

where the subscript n denotes the condition of the uniform/

normal state of ow. This variable is expressed in dimensionless form as gn = yn/D.
The exact Froude number is given by

where g = dimensionless ow depth (y/D). The maximum relative error, e, of Eq. (5) is about 14% in the range
0.1 < g < 0.92 (Fig. 2), where e = 100(Fe  F)/Fe. Clearly,
the approximation of Eq. (4) is inaccurate.

where y = depth of ow (m), x = distance along the channel,

measured positive in the downstream direction (m), dy/
dx = slope of the free surface at any location x, S0 = longitudinal slope of the channel bottom, Sf = friction slope, and
F = Froude number. For F < 1 the ow is subcritical, and
for F > 1 the ow is supercritical. For F = 1 the ow is critical
and dy/dx tends to innity (provided Sf S0) and thus the
water-surface prole approaches the critical-depth line vertically. When Sf = S0, the ow is uniform and dy/dx tends to
zero and hence the surface prole approaches the normaldepth line asymptotically. A schematic of the water surface
prole in a circular channel is shown in Fig. 1.
The friction slope in a circular open channel has been determined using experimental observations of Sauerbrey as [20,21]


2
n2 Q2 3 y2
7 y2
Sf 16=3
1

2
4 D2
12 D2
D

F2e

Using an approximate formula for the cross section area, Hager [20] approximated the Froude number as


Q2 
4
0:64g5c 0:405g3
c 0:834gc
5
gD

11

where the subscript c denotes the condition of the critical

state of ow. Substituting for Sf and F from Eqs. (2) and
(10b) into Eq. (1) yields

Water surface prole in a circular channel.

Accurate gradually varied ow model for water surface prole in circular channels


627

p 2
7
7 p 2
dg 1  g2  1 
dg  x1 g2  x2
12
12


p 2
7
7 p 2
dg  x3 g2  x4
dg 1  g2 1 
12
12

18
19

where

Figure 2 Existing and proposed approximations for square

Froude number.
16n2 Q2

n 2

y2

20

x2

6
7

21

x3

6
7

x4

6
7

o2

S0  9D16=3 Dy 2 1  127 D2
dy

dx 1  Q25 0:64g5 0:405g3 0:834g4

gD
Eq. (12) can be written in dimensionless form as

2
0:64g9 jg4  0:405g  0:834 1  127 g2
jdv
dg


2
dg4 1  127 g2  1

12

16n2 Q2 x
9D19=3
9S0 D16=3
d
16n2 Q2
4
j 0:64g5c 0:405g3
c 0:834gc

22
23

Substituting Eqs. (18) and (19) into Eq. (17) yields

p
0:64g9 jg4  0:405g  0:83412  7g2 2
dg
168 djdv
g2  x1 g2  x2


13

where v = dimensionless distance along the channel,

d = function of channel geometric and ow parameters,
j = function of dimensionless critical depth and g = dimensionless ow depth (g = y/D and gc = yc/D). The v,d and j
variables are given by
v

s!
7
1  p
3 d
s!
7
1  1  p
3 d
s!
7
1 1 p
3 d
s!
7
1  1 p
3 d

6
1
x1
7

14

0:64g9 jg4  0:405g  0:83412  7g2 2

dg
g2  x3 g2  x4
24

Breaking the right-hand side of Eq. (24) into simplest partial

fractions, then
0:64g9 jg4  0:405g  0:83412  7g2 2
g2  x1 g2  x2
0:64g9 jg4  0:405g  0:83412  7g2 2
g2  x3 g2  x4


jx21 0:8347x01  122
1

g2  x1
x1  x2


jx22 0:8347x02  122
1

g2  x2
x1  x2


jx23 0:8347x3  122
1

x3  x4
g2  x3


jx24 0:8347x4  122
1

x3  x4
g2  x4


0:64x41  0:4057x1  122
g

g2  x1
x1  x2


0:64x42  0:4057x2  122
g

g2  x2
x1  x2


0:64x43  0:4057x3  122
g

g2  x3
x3  x4


0:64x44  0:4057x4  122
g

g2  x4
x3  x4


15
16

Eq. (13) is a rst-order ordinary differential equation based on

Manning roughness coefcient that is generally used in
hydraulic engineering practice. According to Eq. (13), the
dimensionless water surface prole, g, depends on three independent dimensionless parameters, namely, v, gc, and d. As
noted, Eq. (13) which gathers all the variables of the gradually-varied-ow equation into four dimensionless parameters
(v, gc, d, and g) is suitable for analytical integration.
3. Exact analytical solution for water surface prole
Consider a circular open channel with a positive slope (d > 0).
Then, Eq. (13) can be analytically integrated if its right-hand
side (integrand) could be broken into simplest partial fractions.
This integrand can be broken as

2
0:64g9 jg4  0:405g  0:834 1  127 g2
p

2jdv
dg


dg2 1  127 g2  1

2
0:64g9 jg4  0:405g  0:834 1  127 g2
p 
dg 17


dg2 1  127 g2 1
Since the denominators of the integrand are polynomials of degree 4, the two denominators can be factorized, respectively, as

j
107:52
 168 p  p g5
d
d
Substituting Eq. (25) into Eq. (24) and integrating yields

25

628

A.R. Vatankhah, S.M. Easa


p
jx21 0:8347x01  122 g  x1 
I1 g 
ln 
p
p
2x1  x2 x1
g x1 

p
jx22 0:8347x02  122 g  x2 
ln 

p
p
g x2 
2x1  x2 x2

p
jx23 0:8347x3  122 g  x3 
ln 

p
p
2x3  x4 x3
g x3 


jx24 0:8347x4  122
g
1
tan

p
p
x3  x4 x4
x4

p
jx21 0:8347x01  122
168 djv 
x1  x2
Z
1

dg
g2  x1
jx22 0:8347x02  122
x1  x2
Z
1

dg
g2  x2

jx23 0:8347x3  122

x3  x4
Z
1

dg
g2  x3

jx24 0:8347x4  12
x3  x4
Z
1

dg
g2  x4


0:64x41  0:4057x1  122  2
ln g  x1 
2x1  x2


0:64x42  0:4057x2  122  2
ln g  x2 
2x1  x2


0:64x43  0:4057x3  122  2
ln g  x3 
2x3  x4

0:64x41
Z

 0:4057x1  12
x1  x2

0:64x44  0:4057x4  122

ln jg2  x4 j  168
2x3  x4
j
17:92
 p g  p g6
d
d

g
dg
g2  x1

Considering the boundary condition for the control section

where g = g0 at v = v0, the integration constant is eliminated
as follows

0:64x42  0:4057x2  122


x1  x2
Z
g

dg
g2  x2

v  v0 I1 g  I1 g0

0:64x43  0:4057x3  122

x3  x4
Z
g

dg
g2  x3

0:64x44  0:4057x4  122

x3  x4
Z
g
j
17:92

dg  168 p g  p g6
g2  x4
d
d

Note that for positive values of x


p
Z
g  x
1
1

p

p
dg

ln
g2  x
2 x g x
For negative values of x, one gets


Z
1
1
g
1
p

dg

tan
g2  x
x
x

26

27

This relationship is valid for 0 6 1  7/(3d1/2) or

5.444 6 d(gn = yn/D 6 0.926). For 5.444 6 d, x1, x2 and x3
are always positive and x4 is always negative. Thus, Eq. (26)
becomes

where I1(g) is the ow function which is given by

Using Eq. (32), the gradually varied ow proles of a circular

channel can be determined analytically for a given boundary
condition. The uniform ow is usually considered established
if the ow depth g deviates by 1% from the uniform ow depth
gn. That is, g0 = 1.01gn for backwater curves and g0 = 0.99gn
for drawdown curves. Since the nominal uniform ow condition g0 = gn occurs only as x 1, it is assumed for practical purposes that g0 = gn 0.01 [21].
For the special case of a horizontal circular open channel
with a longitudinal slope of zero (S0 = 0), Eq. (15) yields
d = 0 and Eq. (13) reduces to

2
7
jdv 0:64g9 jg4  0:405g  0:834 1  g2 dg 33
12

jv I2 g constant
28

where tan denotes inverse tangent.

To complete the integration procedure, the signs of x1, x2,
x3 and x4 should be determined. For this, the normal ow
depth of Eq. (3) is rewritten in terms of d as
s!)1=2
(
p
6
7
29
D x2
yn D
1  1  p
7
3 d

1
p  I1 g constant
168j d

32

which can be analytically integrated as follow,

1

31

30

34

where the ow function I2(g) is given by

7 14
14 12
8 10
49
1
g
g 
g 
jg9 jg7
450
225
125
1296
6


147 6 1
6811 5
189 4
g 
j
g 
g

6400
5
24000
1600
973 3
81 2 417
g
g
g

3000
400
500

I2 g 

35

The integration constant can be eliminated similarly as shown

in Eq. (32).
To illustrate the implementation of the proposed solution
of Eq. (32), the characteristics of the water surface prole were
established for different values of gn, as shown in Fig. 3. The
gure shows the prole curves for different values of gc. Dividing curves for subcritical and supercritical ow regimes (i.e.
critical ow depth curve g = gc) and for hydraulically mild
and steep slopes (critical slope curve gn = gc) are shown. The

Accurate gradually varied ow model for water surface prole in circular channels

Figure 3

629

Dimensionless water surface proles for different gn.

horizontal dashed line represents the uniform ow depth. The

water surface curves are tangents to the vertical as they intersect with the critical ow depth curve and approach the uniform ow depth asymptotically.
4. Comparison with existing approximate method
The existing approximate solution for backwater curves is given by [19,20]

Table 1 Three cases used for comparison with the existing

approximate method.
Case

1
2
3

Channel variables
gn

So

gc

0.7014
0.7500
0.8022

1.25
1.25
1.25

0.0111
0.0111
0.0111

0.96
1.05
1.1

0.0005
0.000515
0.0005

0.4189
0.4381
0.4484

630

A.R. Vatankhah, S.M. Easa

 

1
Y 1 Yr  1

X  Xr Y  Yr  1  Y4c ln
4
Y  1 Yr 1
2arctan Y  arctan Yr 


X 1 

1:1y2N 1=2 X;

yN < 0:9

36
37

where
 
Y = y/yn, Yc = yc/yn, yN = yn/D, X = Sox/yn, and
Xr ; Yr = boundary conditions which equal (0, 1.01) for backwater curves and (0, 0.99) for drawdown curves. Note that
some dimensionless variables are dened differently from
those of the proposed solution.
For comparison, the backwater length was calculated for
different values of g for three cases with gn = 0.70, 0.75, and
0.80 using the proposed accurate solution of Eq. (32) and
the approximate solution of Eq. (36). The corresponding values of the other variables are shown in Table 1. The percentage
difference in backwater lengths of the proposed and approximate solutions is shown in Fig. 4. As noted, the error of the
approximate method increases as the value of gn increases
and could reach up to about 30%, where the existing approximate solution underestimates the backwater length. The error
also increases as the control ow depth g increases. For example, for gn = 0.75 the error ranges from 17% to 22%. For this
case, for g = 0.90 for example, the lengths of the proposed
accurate approximate solutions are 2954.00 m and
2339.21 m, respectively, representing a difference of 20.8%.
The respective backwater proles are shown in Fig. 5. For verication, the exact backwater length was calculated using

Figure 6 Prole length for different number of segments of the

direct step method.

numerical integration of Eq. (13) as 2954.00 m which is identical to the value of the proposed solution. The backwater
curves for g = 0.85 are also shown in Fig. 5. As noted, the error in the estimated backwater length decreases for smaller
control ow depths.
For g = 0.90, the steps of applying the proposed accurate
solution are as follows: (a) yn = 0.9375 m from Eq. (3), thus
gn = 0.9375/1.25 = 0.7500, (b) j = 27.4541 from Eq. (16),
(c) d = 7.0013 from Eq. (15), (d) g0 = 1.01 0.7500 =
0.7575, (e) I1(0.90) = 864.12 and I1(0.7575) = 1258.75 from
Eq. (31), and (f) v = 0.0708 and v0 = 0.1031 from Eq. (30),
thus, v  v0 = 0.17395 and from Eq. (14), x = 2954.00 m.
5. Comparison with the Direct Step Method

Figure 4 Difference between backwater lengths of proposed

(exact) and existing (approximate) solutions for gn = 0.70, 0.75,
and 0.80.

To show the accuracy and efciency (substantially reduced

computational effort) of the proposed method, a comparison
between the lengths of the water surface prole computed
using the direct step method and the proposed semi-exact
method is performed as shown in the example below.
For g = 0.90, the proposed accurate solution gives
x = 2954.00 m, as described earlier. For the direct step method, the computations start with a known depth, yb, at the control structure and proceed in the upstream direction. By
considering the location at the control structure as xb = 0,
the computed values of x will be negative. Using the boundary
condition y(xb) = yb, prescribed water depth, the direct step
method will be as follows [21,22]
x2 xb

E2  Eb
S0  12 Sf2 Sfb

38

in which
Z

D
E y  F2 dy Dg 0:106667g6 0:2025g2 0:278g3 39
j
 
2
n2 Q2 3
7
40
Sf 16=3 g2 1  g2
4
12
D

Figure 5 Backwater lengths of proposed and approximate

solutions for two ow depths (g = 0.85 and 0.90) for gn = 0.75.

where E is specic energy and (Sf2 + Sfb)/2 is mean friction

slope. Note that for a correct comparison, the specic energy
and friction slope are considered according to this study. Thus,
given a specied ow depth, y2, the location of Section 2, x2,
can be determined using Eqs. (38)(40). This is the starting value for the next step and thus the total length of the water surface prole can be calculated.
Substituting Eqs. (39) and (40) into Eq. (38), yields a
dimensionless form as

Accurate gradually varied ow model for water surface prole in circular channels
(

 )
3
0:106667g62 0:2025g2
2 0:278g2
1


g2  gb j
3
 0:106667g6b 0:2025g2
b 0:278gb
n
o
v2 vb
 
2 3 2 
2
4 gb 1  127 g2b
d  329 34 g22 1  127 g22
41
In this example, equal decrements of the dimensionless ow
depth are used, where the decrement size equals the difference
between the starting and ending depths divided by the number
of segments per prole J, that is (0.9  0.7575)/J. Increasing J
will increase the accuracy of the prole calculation, but will increase the calculation time. Fig. 6 shows variation of the total
length of the backwater curve with the number of segments.
The results show that the near exact solution x = 2954.00 requires 500 segments, clearly representing a huge computational
effort.
6. Conclusions
A very accurate approximation of the Froude number of circular channels having an error in the order of less than 1%
is presented in this paper. The Froude number was used in
the governing GFV equation to develop an exact analytical
solution of the water surface prole. The proposed solution
has substantially improved the accuracy of estimating the
water surface prole compared with the existing approximate method which underestimates the backwater length.
The characteristics of the water surface prole based on
the proposed solution have been illustrated using sample
analysis graphs that may be useful for practical purposes.
The optimization concept underlying the proposed approximation of the Froude number may be used for other types
of open channels.

References
[1] Gunder DF. Prole curves for open-channel ow. Trans Am Soc
Civ Eng 1943;108:4818.
[2] Gill MA. Exact solution of gradually varied ow. J Hydraul Div
ASCE 1976;102(HY9):135364.
[3] Kumar A. Integral solutions of the gradually varied equation for
rectangular and triangular channels. Proc Inst Civ Eng ASCE
1978;65(2):50915.
[4] Dubin JR. On gradually varied ow proles in rectangular open
channels. J Hydraul Res 1999;37(I):99105.
[5] Ramamurthy AS, Saghravani SF, Balachandar R. A direct
integration method for computation of gradually varied ow
proles. Can J Civ Eng 2000;27(6):13005.
[6] Venutelli M. Direct integration of the equation of gradually varied
ow. J Irrigat Drain Eng ASCE 2004;130(1):8891.
[7] Vatankhah AR. Exact sensitivity equation for one-dimensional
steady-state shallow water ow (application to model calibration).
J Hydrol Eng 2010;15(11):93945.
[8] Vatankhah AR. Analytical integration of the equation of gradually varied ow for triangular channels. Flow Meas Instrum
2010;21(4):5469.
[9] Vatankhah AR. Direct integration of gradually varied ow
equation in parabolic channels. Flow Meas Instrum 2011;22(3):
23541.

631

[10] Vatankhah AR, Easa SM. Direct integration of Manning based

gradually varied ow equation. Proc ICE J Water Manage 2011;
164(5):25764.
[11] Vatankhah AR. Direct integration of manning-based GVF
equation in trapezoidal channels. J Hydrol Eng 2012;17(3):
45562.
[12] Chow VT. Integrating the equation of gradually varied ow. In:
Proceedings Paper No. 838, vol. 81. ASCE; 1955; p. 132.
[13] Zaghloul NA. Hydraulic exponents M and N for gravity ow
pipes. Adv Water Res 1998;21(3):18591.
[14] Keifer CJ, Chu HH. Backwater functions by numerical integration. Trans ASCE 1955;120:42942.
[15] Nalluri C, Tomlinson JH. Varied ow functions for circular
channels. J Hydr Div ASCE 1978;104(7):9831000.
[16] Zaghloul NA. A computer model to calculate varied ow
functions for circular channels. Adv Eng Soft 1990;12(3).
[17] Zaghloul NA, Shanin MM. Computer simulation of gradually
varied ow proles in circular sections. Adv Eng Software 1993;
16(1):3746.
[18] Zaghloul NA. Gradually varied ow in circular channels with
variable roughness. Adv Eng Software 1992;15(1):3342.
[19] Hager WH. Backwater curves in circular channels. J Irri Drain
Eng 1991;117(2):17383.
[20] Hager WH. Wastewater hydraulics: theory and practice. New
York, NY: Springer; 2010.
[21] Chow VT. Open-channel hydraulics. New York, NY: McGrawHill; 1959.
[22] Subramanya K. Flow in open channels. 3rd ed. Singapore:
McGraw-Hill; 2009.
[23] Sauerbrey M. Abuss in Entwasserungsleitungen. Wasser und
Abwasser in Forschung und Praxis, 1, Inst. Wasserversorgung,
Abwasserbeseitigung und Stadtbauwesen, TH Darmstadt, E.
Schmidt Verlag; 1969 [in German].
Ali R. Vatankhah earned his M.Eng. (1999)
and Ph.D. (2008) from University of Tehran.
He is currently an assistant professor in the
Department of Irrigation and Reclamation
Engineering, University of Tehran, Tehran,
Iran. He has authored more than 60 international refereed journal articles. He received
the best-discussion award for Discussion of
Applying Particle Swarm Optimization to
Parameter Estimation of the Nonlinear
Muskingum Model, Journal of Hydrologic
Engineering (ASCE), 15(11), 949952, 2010. His research interests
include investigation of hydraulic behavior of irrigation structures and
their optimum design, and mathematical modeling in open channel
hydraulics.

Said M. Easa earned his B.Sc. (1972) from

Cairo University, M.Eng. (1976) from
McMaster University, and Ph.D. (1982) from
University of California at Berkeley. He is
currently professor of civil engineering,
Ryerson University, Toronto, Canada. He is
Associate Editor of the Canadian Journal of
Civil Engineering. He has published more
than 210 refereed journal articles. His work
received national and international lifetime
and best paper awards, including 2001 Frank
M. Masters Transportation Engineering Award and 2005 Arthur M.

632
Wellington Prize from American Society of Civil Engineers (ASCE),
2010 James A. Vance Award and 2003 Sandford Fleming Award from
Canadian Society for Civil Engineering (CSCE), and 2010 Award of
Academic Merit from Transportation Association of Canada. He is

A.R. Vatankhah, S.M. Easa

Fellow of Canadian Academy of Engineering, Engineering Institute of
Canada, and CSCE. His research interests include highway drainage
and transportation management.