Journal of Hydraulic Research

ISSN: 0022-1686 (Print) 1814-2079 (Online) Journal homepage: http://www.tandfonline.com/loi/tjhr20

Exact solutions for normal depth problem

Rajesh Srivastava

To cite this article: Rajesh Srivastava (2006) Exact solutions for normal depth problem, Journal
of Hydraulic Research, 44:3, 427-428, DOI: 10.1080/00221686.2006.9521694

To link to this article: http://dx.doi.org/10.1080/00221686.2006.9521694

Published online: 26 Apr 2010.

Submit your article to this journal

Article views: 72

View related articles

Citing articles: 6 View citing articles

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=tjhr20

Download by: [Universite Laval] Date: 06 May 2016, At: 22:45

 Journal of Hydraulic Research Vol. 44, No. 3 (2006), pp. 427–428
© 2006 International Association of Hydraulic Engineering and Research
Discussion
Exact solutions for normal depth problem
P.K. SWAMEE and P.N. RATHIE, Journal of Hydraulic Research, Volume 42, 2004, Issue 5, pp. 541–547
Discusser:
RAJESH SRIVASTAVA, Associate Professor, Department of Civil Engineering, Indian Institute of Technology Kanpur,
Kanpur 208016, India. Tel.: +91-512-2597755; fax: 2597395; e-mail: rajeshs@iitk.ac.in
The authors have proposed infinite series expressions which help
in evaluating the normal depths in various open channel sections
for a given discharge. The discusser would like to mention the
following points regarding the paper:
Downloaded by [Universite Laval] at 22:45 06 May 2016
1. For any infinite series, the convergence and the number of
terms to be used for a reasonable accuracy must be men-
tioned. Similarly, when there are two different expressions
given for a section (e.g. Eqs (28) and (32) for rectangular
section), it would be prudent to suggest a range of applicabil-
ity. In absence of this information, the authors’ hope that the
equations are useful to water resources engineers may not be
fulfilled.
2. The discusser performed a detailed error analysis using all the
terms listed in Eqs (28) and (32). Figure 1 shows the relative
error in βn over a wide range of values of Nb . The small error
of the proposed infinite series is clearly seen and it is also
obvious that Eq. (28) should be used for Nb < 1 and (32) for
Nb > 1. It may also be noticed that Eqs (28) and (32) result
in larger error near Nb = 1, with the error being as high as
0.6%. It was found that six to seven terms of the series are
required to keep the maximum error <1%. This behaviour is
similar to the one observed (Srivastava, in press) for a closely
related problem (Swamee and Rathie, 2004).
3. A more practical approximation could be obtained by trun-
cating the series (28) and (32) after a finite number of terms
and obtaining the coefficients which provide minimum error
(Srivastava, in press). On the other hand, if one decides to use
an infinite series, a computationally simpler algorithm could
be obtained based on a convergent iterative solution of Eq.
(25) as follows:
y0 = 0
3/5 3/5
y1 = Nb (1 + 2y0 )2/5 = Nb
3/5 3/5
3/5 2/5
y2 = Nb (1 + 2y1 )2/5 = Nb 1 + 2Nb
3/5
y3 = Nb (1 + 2y2 )2/5
3/5  3/5 3/5 2/5
= Nb 1 + 2Nb (1 + 2Nb )2/5
and so on.
427
4. However, from an engineer’s point of view, it would be more
useful to have a simple and reasonably accurate expression
rather than one which is more complicated, and less accurate
for some cases. Therefore, an attempt was made by the dis-
cusser to arrive at an expression based on truncation of the
iterative algorithm described above. The following approxi-
mation was found to have a maximum error of about 0.06%
in estimation of βn :

0.3929 2/5
3/5
βn = N b 1 + 2.404Nb0.6321 1 + 2.030Nb0.9363
(74)
Figure 1 shows the performance of Eq. (74). Clearly, Eqs (28)
and (32) are more accurate over a large range but Eq. (74)
is simpler to use, less computationally intensive, reasonably
accurate, and applicable over the entire range of Nb values.
Therefore, the discusser feels that it should be preferred over
Eqs (28) and (32) for the computation of normal depth in
rectangular channels. Similar methodology using convergent
iterative schemes might be attempted for other sections to
obtain a finite series approximation, and is likely to produce
usable results, especially if split-range approximations are
allowed.
5. There appears to be typographical errors in Eqs (53)–(57)
and (63)–(66) and the terms of the form cos−1 (1 − x) should
probably be cos−1 (1 − 2x). Also, for circular channels, the
multiplicity of normal depth for the same discharge for ηn >
0.82 results in two normal depths for ND >0.312. The dis-
cusser presumes that Eq. (71) will provide the smaller normal
depth in such cases, but would like the authors to comment
on it.
6. The last comment is more a matter of personal prefer-
ence. Above Eq. (73) the authors use n = 0.015 s/m1/3 .
The discusser feels that this gives the impression that n
would be different in FPS units (i.e. n = 0.015 ×
0.30481/3 s/ft1/3 = 0.010 s/ft1/3 ). It might be better to assume
that n is dimensionless and the coefficient in the Manning’s
equation (1 for SI units or 1.486 for FPS units) has the
necessary dimensions (m1/3 /s or ft1/3 /s).
 428 Swamee and Rathie

10 2. Within the above stated specified applicability limits of

Eqs (28) and (32) the corresponding series are convergent.
1
Howerver, the convergence is weak near Nb = 0.966998.
Percent Error in βn

0.1 3. The objective of the paper is to obtain exact solutions for nor-
mal depth. The end result of numerical solution is a number,
0.01
which is devoid of physical picture. A closed form solution
0.001 gives physical picture of functional dependence of normal
depth with the independent variables.
0.0001
4. Equation (74), though yields high accuracy, however, for
0.00001 Nb → ∞ does not yield the equation βn = 22/3 Nb which
0.01 0.1 1 10 100 is the asymptotic form as obtained from Eq. (24). Thus the
Nb exponents occurring in Eq. (74) of the discusser should be
Figure 1 Percentage error in the prediction of βn for a given Nb . Solid
suitably modified.
line: Eq. (28); dotted line: Eq. (32); dashed line: Eq. (74). 5. The correct form of Eqs (58), (62), (67) and (71) has been
given in the replies of the discussion by Kouchakzadeh and
Vatankhah. Yes, equations are applicable to the lower range
of ND . The upper range is only of academic importance.
Since the upper range is very small, with some increased error
Reference
equations can be applied to upper range as well.
Downloaded by [Universite Laval] at 22:45 06 May 2016

6. Treating the roughness coefficient dimensionless is not cor-

1. Swamee P.K. and Rathie P.N., (2004). “Exact Solutions
for Sequent Depth Problem”. J. Irrig. Drain. Engng. ASCE
130(6), 520–522.
Closure to “Exact solutions for normal depth problem”
By Prabhata K. Swamee, Emeritus Fellow, Department of Civil
Engineering, Indian Institude of Technology, Roorkee, Roorkee
247 667, India
Pushpa N. Rathie, Professor, Department of Statistics, Univer-
sity of Brasilia, Brasilia 70910-900, Brazil
The authors are very much grateful to Rajesh Srivastava for his
discussion. The replies of the points raised by the discusser are
as follows.
1. It is sufficient to use six terms in Eqs (28) and (32) for obtain-
ing normal depth with reasonable accuracy. Using Braaksma
(1964) it is found that Eq. (28) converges for Nb < 0.966998
whereas Eq. (32) converges for Nb > 0.966998.
rect. Considerable literature has grown on this controversy.
By attaching dimensions to 1 in 1/n, the roughness coefficient
occurring in the Manning’s equation can be made dimension-
less. In such an eventuality 1 is changed to 1.486 for converting
the Manning’s equation from SI units to fps units. Thus, the
values n for various types of surface materials remain same
as their original values in SI units. Such a manipulation for
conversion of an equation from one system of units to another
system of units sets a dangerous precedence. Thus, by a sim-
ilar logic, area A of a square of side a given by A = 1a2 , say
in metric units, can be rewritten to accept a in meter and yield
A in ft2 by changing 1 to some other numerical value. One
may, thus, think that the side a of a square is dimensionless!
Reference
1. Braaksma, B.L.J. (1964). “Asymptotic Expansions and
Analytic Continuations for a Class of Barnes-Integrals”.
Compos. Math. 15, 239–341.