Velocity Transition and Flow

Geometry in Rectangular Expanding

Channels: Incompressible Steady
State 1-D Navier-Stokes Approach
Sushil Devkota

Department of Civil Engineering, Nepal College of Information

Technology, Pokhara University, Nepal

April 2025

Abstract
This research presents an analytical study of velocity transition in expand-
ing rectangular open channels under steady, incompressible, viscous flow.
By simplifying the one-dimensional Navier–Stokes equation with a constant
depth gradient assumption, a velocity distribution equation is derived in-
corporating viscous diffusion. Using Airy functions and their asymptotic
properties, the solution simplifies for large transition parameters, ultimately
resembling inviscid flow behavior. Moreover, transition bed width geometry
is also derived, clarifying continuous expansion from narrow starting section
to expanded ending section under the control of velocity variations. A head
loss estimation method based on the hydraulic slope at the midpoint section
is also proposed, effectively handling varying bed widths and depths. The
study confirms that in expanding, mild-slope channels, the depth gradient
exceeds the bed slope, validating the constant gradient assumption. The de-
rived relationships provide a structured framework for optimizing transition
designs, enhancing both theoretical understanding and practical hydraulic
efficiency.

1
1 Introduction
The transition of water flow from narrow to wider channels presents a fun-
damental problem in fluid mechanics, with implications across hydraulic en-
gineering, sediment management, and environmental hydrology. Such tran-
sitions immanently involve spatial variations in both velocity and depth,
governed by the principles of mass continuity and momentum conserva-
tion [7, 8, 23]. These noticable alternations can be particularly observed in
systems such as natural rivers, culverts and irrigation canals where channel
geometry is non-uniform [26].

Although Navier’s pioneering derivation of the Navier–Stokes equation, based

on Laplace’s molecular forces and extended through elastic theory and La-
grange’s method, initially went unrecognized, it remarkably provided the
first predictive model for viscous fluid motion, despite its early limitations to
capillary flows [6].The dynamics of flows are scrupulously described by the
Navier-Stokes equations, which encapsulate the momentum conservation for
Newtonian fluids, accounting for pressure gradients, gravitational forces, and
viscous effects [4, 5, 31].The total energy head—comprising elevation, pres-
sure, and velocity components—is redistributed due to expansion-induced
deceleration and boundary shear, contributing to head losses and flow stabil-
ity [28]. It is generally known that the Gradually Varied Flow (GVF) theory
becomes primary when there is a purpose of modeling transitions in an open
channel flow. It predicts the spatial change of flow depth and velocity due
to geometric variations of width and bed slope changes [14, 28]. While GVF
particularly implies a non-uniform flow regime, simplifying assumptions are
sometimes applicable to simplify analytical tractability in expanding chan-
nels [25, 27]. Such approximations, though idealized, provide useful insight
into the underlying physics and can approximate actual behavior in mildly
varied transitions.

The different methods of solution for 1-D and 3-D Navier-Stokes equations
and GVF equation with certain conditions have been developed since early
days. Solutions by using numerical methods have been becoming less complex
yet rather than analytical solutions. Many researchers have used methods
like finite difference method [18], three level finite element method [13], finite
volume method [9] and others. However, the primary focus is still in perfect
analytical solution, different conditional solutions have been developed since

2
early period. Direct transform method with perturbation method reduces the
equation and gives appropriate analytical solutions with boundary conditions
applied at infinity [24]. Even though the complexity of 3-D Navier-Stokes
problem cannot be neglected, perturbation technique can be applied to 2-
D Navier-Stokes problem considering its equivalence as 3-D Navier-Stokes
problem [16]. Although the unsteady solutions of incompressible 3-D Navier-
Stokes equation have less possibility physically, they are suitable for testing
and analytical observation of flow [11].

Not only in 3-D but also in 1-D, the solution of corresponding equation be-
comes quite complex due to its non-linear behavior. There exists a solution of
Navier-Stokes equation for compressible viscous fluid in one dimension, which
is strong globally, with the help of large initial data [15]. The reduction of
1-D Navier-Stokes equation to viscous Burgers equation the effect of pressure
gradients and Reynolds numbers on fluid flow can be analyzed using Orlowski
and Sobczyk transformation [3]. Global smooth solutions for 1-D barotropic
Navier-Stokes equations with degenerate viscosity coefficients are also pos-
sible in existence, encapsulating the uniqueness [19].GVF equation has also
been analyzed for the sake of solution of Navier-Stokes equation. Direct inte-
gration on GVF equation is applicable on wide rectangular channel utilizing
Manning’s formula so that simplification on computation of flow profiles can
be achieved resembling averaged parametric values [29, 30].Newton-Raphson
method is vital for GVF computations emphasizing the averaged hydraulic
parameters [17].

From a theoretical perspective, analytical solution to the 1-D Navier-Stokes

equation is intricate due to the non-linearity of the system. The absence
of exact non-stationary solutions for the non-linear Navier–Stokes equations
has inevitably led to the advancement of statistical and numerical methods,
which rely on an infinite open-loop framework to analyze the behavior of
their solutions.( [12, 20]. Nonetheless, specific cases such as compressible
isothermal flow with density-dependent viscosity admit exact formulations
and notable solutions [10]. These solutions serve as benchmarks for numer-
ical solvers and offer valuable analytical perspectives.The findings from the
solution of Navier-Stokes contribute to improve the hydraulic design, reduce
energy loss, and more effective water resource management [26].

The present research builds on these foundations to explore velocity tran-

3
sitions and channel geometry in gradually expanding open-channel flows.
While the depth gradient in GVF is not strictly constant, approximation
can be done for the reduction of complexities. By assuming a nearly con-
stant depth gradient—a simplification that aligns with mild expansion and
milder slope cases—the derivation of analytical expressions for velocity us-
ing the 1-D Navier-Stokes framework. Finally, approximation under certain
considerations leads to the simplified expressions of velocity and bed width
geometry. This approach not only bridges the gap between theoretical and
applied hydraulics but also enhances an ability to predict flow behavior in
engineered and natural systems.

2 Governing Equations and Assumptions

Navier-Stokes equation for incompressible unsteady state 3-D flow [4, 5, 8,
31]is:  
∂u ∇p
+ (u · ∇)u = − + ν∇2 u + g (1)
∂t ρ
Where:

• u is the velocity vector in 3-D.

• ρ is the density of the fluid.

• p is the pressure.

• ν is the kinematic viscosity.

• g is the gravitational acceleration vector (assuming that gravity acts in

a vertically downward direction).

• ∇2 u is the Laplacian of the velocity vector.

But for incompressible steady state 1-D flow, Navier stokes equation [4, 5, 8]
is:
du 1 dp d2 u
u =− + ν 2 + gx (2)
dx ρ dx dx
Where x is measured along the direction of flow and gx is the gravitational
acceleration in the direction of x. In terms of depth gradient and slope of

4
bed,Eq. 2 can be written as:

du dh d2 u
u = −g + ν 2 + gS0 (3)
dx dx dx
Where h and S0 are depth and bed slope of channel respectively. The bed
slope (S0 ) is considered milder for the appropriate assumption of depth vari-
ation. The assumption of a constant depth gradient is taken. So,
dh
=k (4)
dx
For expanding channel having mild slope, k > S0 is taken under assumption
which is consistent with gradually varied subcritical flow. The value of the
depth gradient constant k can be calculated by using initial narrow section
depth h1 , final wider section depth h2 and length of transition L as,
h1 − h2
k= (5)
L

3 Analytical Solution for Velocity Transition

The equation Eq. 3 can be written as:

du d2 u
u = ν 2 + g(S0 − k) (6)
dx dx
The solution of Eq. 6 can be proceeded using particular method. In this
2
governing equation, the viscous diffusion term ν ddxu2 plays a crucial role in
balancing both the non-linear convective term u du
dx
and constant term g(S0 −
2
k). Based on this observation, the equation is written in the form where ddxu2
is expressed in terms of u du
dx
combined with a certain constant, leading to a
solvable formulation. Now, the viscous diffusion term can be expressed as,

d2 u du
2
= c1 u + c2 (7)
dx dx
Here, c1 and c2 are constants. Substituting Eq. 7 in Eq. 6 ,
 
du du
u = ν c1 u + c2 + g(S0 − k) (8)
dx dx

5
 du
(1 − νc1 )u = g(k − S0 ) − νc2 (9)
dx
It is clear that u du
dx
is variable term. So, comparing the respective coefficients
from Eq. 9,
1 g(k − S0 )
c1 = and c2 = (10)
ν ν
Substituting Eq. 10 in Eq. 7,
d2 u u du g(k − S0 )
2
= + (11)
dx ν dx ν
 
du u g(k − S0 )
d = du + dx (12)
dx ν ν
Integrating Eq. 12,
du u2 g(k − S0 )x
= + + c3 (13)
dx 2ν ν
Where c3 is an integration constant. Since this study considers an expand-
ing channel where k > S0 , it is expected that velocity decreases along the
2
flow direction ( du
dx
< 0). However, both u2ν and g(k−S ν
0 )x
are positive terms
contributing to an increase in velocity. To satisfy the expected behavior of
decreasing velocity, c3 must be negative and sufficiently large in magnitude
to counterbalance these positive terms ensuring dudx
remains negative.

So, c3 = −c4 , where c4 is positive. Now, Eq. 13 can be written as,

1 du u2 g(k − S0 )x c4
= 2+ − (14)
2ν dx 4ν 2ν 2 2ν
Introducing a new variable v and a constant a such that,
u g(k − S0 )
v= and a3 = (15)
2ν 2ν 2
Substituting Eq. 15 in Eq. 14,
dv c4
− v2 + − a3 x = 0 (16)
dx 2ν
Eq. 16 can be written as,
1 dv v 2 c4
2
− 2 + 2 − ax = 0 (17)
a dx a 2a ν
6
Again, introducing other variables t and z such that,
v c4
t= and z = − ax (18)
a 2a2 ν
Substituting Eq. 18 in Eq. 17,
dt
+ t2 − z = 0 (19)
dz
Assuming a variable w such that,
1 dw
t= (20)
w dz
Substituting Eq. 20 in Eq. 19, finally it becomes,
d2 w
− wz = 0 (21)
dz 2
This Eq. 21 is the Airy differential equation [1,2,21,22]. The general solution
of Airy differential equation is expressed in the form of Airy functions of first
kind and second kind respectively.
So, the solution of Eq. 21 [1, 2, 22] can be expressed as,

w = a1 Ai(z) + a2 Bi(z) (22)

Where Ai(x) and Bi(x) are the Airy functions of first kind and second kind
respectively. a1 and a2 are the constants. Using Eq. 19 the solution given by
Eq. 22 can be expressed as,
a1 Ai′ (z) + a2 Bi′ (z)
t= (23)
a1 Ai(z) + a2 Bi(z)

Here Ai′ (z) and Bi′ (z) are the first derivatives of Airy functions of first kind
and second kind respectively. The solution given by Eq. 23 can be expressed
using Eq. 18 as,
a3 Ai′ (z) + Bi′ (z)
 
v=a (24)
a3 Ai(z) + Bi(z)
Where a3 = aa21 . Similarly, using Eq. 15, the solution from Eq. 24 can be
reduced to:
a3 Ai′ (z) + Bi′ (z)
 
u = 2νa (25)
a3 Ai(z) + Bi(z)

7
This is the velocity distribution along the length of channel or flow.
For large values of z, the Airy functions exhibit distinct asymptotic behav-
ior. The function Ai(z) decays exponentially, whereas Bi(x) grows rapidly.
Specially for sufficiently large z, Ai(z) approaches zero and its contribution
to the velocity distribution becomes negligible. Conversely, Bi(z) remains
dominant in this regime, leading to the velocity distribution being primarily
governed by terms involving Bi(z) and Bi′ (z).
So, according to this approximation, Eq. 25 becomes,

2νaBi′ (z)
u= (26)
Bi(z)

For large values of z, the ratio of Bi′ (z) and Bi(z) follows the asymptotic
behavior(From Eq. A6),
Bi′ (z) 1
≈ z2 (27)
Bi(z)
Using this behavior from Eq. 27 and Eq. 18, Eq. 26 reduces to:
 c  12
4
u = 2νa − ax (28)
2a2 ν
Again, substituting value of a defined by Eq. 15, Eq. 28 becomes,
1
u = [2νc4 − 2g(k − S0 )x] 2 (29)

The two constants, c4 and k are to be determined. To determine c4 , de-

termination of k is necessary. As k depends upon h2 referring to Eq. 5, it is
mandatory to find value of h2 .

Energy equation which accounts flow conditions and head loss in transition
is used between two rectangular sections(narrow and expanded) to find h2 in
an approximated way. Using energy equation [7, 8, 23, 28, 31] between those
two sections,
Q2 Q2
h1 + + LS 0 = h2 + + hL (30)
2gB12 h21 2gB22 h22
Where,

• h1 and h2 are the flow depths at narrow and expanded sections respec-
tively as stated earlier in the assumption section.

8
 • Q is the flow discharge.

• B1 and B2 are the bed widths at narrow and expanded sections respec-
tively.

• hL is the head loss between the transition, formulated approximately

as(Eq. B5) ,
4
16
2 2 [B1 + B2 + 2(h1 + h2 )] 3
hL = 2 n Q L
3
10 10 (31)
(B1 + B2 ) 3 (h1 + h2 ) 3
Substituting Eq. 31 in Eq. 30,
4
Q2 Q2 16 [B1 + B2 + 2(h1 + h2 )] 3
h1 + 2 2
+LS 0 = h2 + 2 2
+2 3 n2 Q2 L 10 10 (32)
2gB1 h1 2gB2 h2 (B1 + B2 ) 3 (h1 + h2 ) 3

Solving Eq. 32, h2 can be determined knowing all other parameters of the
equation. Once h2 is determined, depth gradient(k) can be estimated ac-
cording to Eq. 5.

For the determination of the constant c4 , both boundary conditions can be

applied.
At x = 0(starting narrow section,B1 , h1 ) using Eq. 29:

Q 1
u1 = = (2νc4 ) 2 (33)
B1 h1
At x = L(ending expanded section,B2 , h2 ):

Q 1
u2 = = (2νc4 − 2g(k − S0 )L) 2 (34)
B2 h2
Dividing Eq. 33 by Eq. 34,c4 can be determined as,

g(k − S0 )L(B2 h2 )2
c4 = (35)
ν[(B2 h2 )2 − (B1 h1 )2 ]

The depth of flow at distance x from starting narrow section can be written
as function of x expressed in linear form because depth gradient is assumed
constant.So,
h(x) = h1 − kx (36)

9
Using Eq. 35, Eq. 34 finally can be written in terms of function of x as,
 21
2g(k − S0 )L(B2 h2 )2

u(x) = − 2g(k − S0 )x (37)
(B2 h2 )2 − (B1 h1 )2
This is the final velocity transition between narrow and expanded section.
Knowing all other parameters of the measure, the velocity can be determined
at every distance from the starting section.
For the transition bed geometry, Eq. 37 can be written in terms of functions
of bed width and flow depth as,
 21
2g(k − S0 )L(B2 h2 )2

Q
u(x) = = − 2g(k − S0 )x (38)
B(x)h(x) (B2 h2 )2 − (B1 h1 )2
Using Eq. 36 and Eq. 38, bed width geometry can be expressed as,
Q
B(x) = h i 12 (39)
2g(k−S0 )L(B2 h2 )2
(h1 − kx) (B2 h2 )2 −(B1 h1 )2
− 2g(k − S0 )x

This gives the bed width of channel at distance x from starting narrow sec-
tion.

4 Discussion and Interpretation of Results

The derived velocity transition equation(25) provides critical insights in the
flow behavior in expanding channel assuming a constant depth gradient. For
the removal of complexity from the GVF equation, dh dx
is assumed constant so
that the energy line effect also merges on it. The velocity expression, initially
formulated with viscous diffusion effects, ultimately reduces to a form inde-
pendent of viscosity after simplification of Airy functions in an approximate
way, which can be observed from equation(38). This suggests that, under
steady conditions, the primary governing factor in velocity distribution is
geometric transition rather than viscous effects, aligning with the behavior
observed in inviscid flow models.

The derived formulations(equation 38 and equation 39) consider a nearly

constant depth gradient, which remains reasonable approximation in milder
slopes where gradually varied flow conditions occurs. The assumption of

10
constant depth gradient greater than bed slope( dh dx
> S0 ) aligns with ex-
pected behavior of GVF equation in milder slope channels. This theoretical
approach of the solutions provides valuable insights into the velocity tran-
sition under steady flow, its straight application on channels having steeper
slopes may be limited due to the increasing significance of rapidly varied flow
phenomena and variation of depth, where assumptions of constant depth gra-
dient and milder slope may no longer hold.
The value of c4 from equation(35) undeniably manifests that the assertion of
c3 being largely negative is true. This can be insisted because the viscosity
of most flowing fluids is generally small in value, which is expressively in-
cluded in the denominator with the presence of a larger value than it in the
numerator, as can be seen from equation(35).

Navier-Stokes equation inherently accounts for viscous diffusion, eliminat-

ing the need for an explicit energy line slope term, unlike the inviscid ap-
proach. However, after applying large value approximation of Airy function,
the velocity expression resembles the inviscid case(equation 38 without the
energy line slope. This occurs because for a large value of z(which depends
on the flow and transition parameter), the dominant term in the solution be-
haves similarly to the inviscid formulation. Essentially, viscosity smoothens
the transition, but its influence becomes embedded in the derived velocity
profile, making it appear as if the energy slope term was not explicitly needed.

The transition bed width function B(x) ensures a smooth expansion from
B1 to B2 , maintaining flow continuity while controlling the velocity varia-
tion. The derived relationship between velocity, depth and bed width con-
firms that the expansion induced deceleration follows a predictable pattern,
dictated by the imposed depth gradient k. This strengthens the validity of a
nearly constant dh
dx
, which, where an approximation remains justifiable given
the gradual nature of transition flow.

In estimating the head loss between two sections in this gradually varied
flow, an approximate method was employed, evaluating the hydraulic slope
at a hypothetical midpoint section derived from the average of bed widths
and flow depths. While this specific midpoint approximation technique is
not explicitly cited in standard literature, it follows the conceptual basis of
interpolating hydraulic parameters in the standard step method, as described
by [8,23]. This approach provides a practical, logically consistent estimate of

11
head loss in the absence of discrete measurement points within the transition
reach. The formulation details of this calculation are presented in . Given the
smooth, gradual nature of the expansion and the assumed constancy of the
depth gradient, this simplification remains reasonable and aligns with typical
practices used in engineering approximations of gradually varied flows.

Comparing results with traditional empirical methods, the derived equations,

offer a more structured approach capturing the interplay between velocity,
depth variation and transition geometry. The formulation provides a the-
oretical foundation for optimizing transition design ensuring minimal flow
disturbances while maintaining hydraulic efficiency.

5 Conclusion
This research presents a velocity transition model and transition bed geom-
etry for expanding rectangular open channel flow, derived from 1-D steady
incompressible Navier-Stokes equation with gradual variation in depth. The
governing equation, which includes the effects of viscous diffusion and grav-
itational forces, is transformed and solved using Airy functions. Through
asymptotic approximation for large values of the argument, the velocity sim-
plifies to a form similar to that of inviscid flow, despite accounting for viscous
diffusion. The study also establishes method to determine the transition bed
width variation using derived velocity distribution. The result expression
for velocity transition and geometry provides an analytical framework for
understanding flow expansion in engineering applications such as aqueduct
transitions, diffusers and gradually expanding channels.

Appendix A: Airy Functions and Their Asymp-

totic Properties
For large positive z, the asymptotic expansions of the Airy functions are
given by:

12
   
1 2 3/2 5 385 85085
Ai(z) ∼ √ 1/4 exp − z 1− + − + ···
2 πz 3 48z 3/2 4608z 3 2211840z 9/2
(A1)

z 1/4
  
′ 2 3/2 7 455 145495
Ai (z) ∼ − √ exp − z 1− + − + ···
2 π 3 48z 3/2 4608z 3 2211840z 9/2
(A2)

  
1 2 3/2 5 385 85085
Bi(z) ∼ √ 1/4 exp + z 1+ + + + ···
πz 3 48z 3/2 4608z 3 2211840z 9/2
(A3)

z 1/4
  
′ 2 3/2 7 455 145495
Bi (z) ∼ √ exp + z 1+ + + + ···
π 3 48z 3/2 4608z 3 2211840z 9/2
(A4)

The ratio of Bi′ (z) to Bi(z) is then:

Bi′ (z)
 
1/2 1 1 139
=z 1+ + + + ··· (A5)
Bi(z) 12z 3/2 288z 3 103680z 9/2

To leading order, for very large z:

Bi′ (z)
≈ z 1/2 (A6)
Bi(z)

Appendix B: Approximate Head Loss Calcu-

lation Within the Transition
In this study, the head loss between two sections of a gradually varied flow
(GVF) was approximated by evaluating the hydraulic slope at the midpoint

13
between the two sections. The hydraulic parameters — such as flow area,
wetted perimeter, and hydraulic radius — were determined at this midpoint
based on the average of bed widths and flow depths. This representative
friction slope was then used in the head loss equation to estimate the energy
loss between the two points. For a gradually varied flow (GVF) between two
sections with varying depth and width, the approximate head loss hL can
be computed by considering the energy slope at the midpoint section. The
expression for head loss from [7, 8, 23] is given by:

hL ≈ L · Sf,mid (B1)
where:

• Sf,mid is the energy slope at the midpoint between narrow starting

section(1) and expanded end section(2).

• L is the length of the transition between the two sections.

The midpoint slope from [7, 23, 28, 31] is approximated as:

Q2 4
Sf,mid = 10 n2 Pmid
3
(B2)
g Amid
3

where:

• Q is the discharge.

• g is the gravitational acceleration.

• Amid is the cross-sectional area at the midpoint, given by:

Amid = Bmid · hmid (B3)

• Pmid is the wetted perimeter at the midpoint, approximated as:

Pmid = (Bmid + 2hmid ) (B4)

Where:

• Bmid = B1 +B2
2

• hmid = h1 +h2
2

14
 • n is Manning’s roughness coefficient.
Using Eq. B4, Eq. B3 and Eq. B2, head loss(hL ) can be expressed according
to Eq. B1 as:
4
16
2 2 [B1 + B2 + 2(h1 + h2 )] 3
hL = 2 3 n Q L 10 10 (B5)
(B1 + B2 ) 3 (h1 + h2 ) 3

References
[1] Milton Abramowitz and Irene A Stegun. Handbook of mathematical
functions with formulas, graphs, and mathematical tables, volume 55.
US Government printing office, 1948.
[2] George B Arfken, Hans J Weber, and Frank E Harris. Mathematical
methods for physicists: a comprehensive guide. Academic press, 2011.
[3] MAK Azad and LS Andallah. Explicit exponential finite difference
scheme for 1d navier-stokes equation with time dependent pressure gra-
dient. GANIT: Journal of Bangladesh Mathematical Society, 36:79–90,
2016.
[4] George Keith Batchelor. An introduction to fluid dynamics. Cambridge
university press, 2000.
[5] Peter S Bernard. Fluid dynamics. Cambridge University Press, 2015.
[6] Sylvio R Bistafa. On the development of the navier-stokes equation by
navier. Revista Brasileira de Ensino de Fı́sica, 40(2):e2603, 2018.
[7] Hubert Chanson. Hydraulics of open channel flow. Elsevier, 2004.
[8] M Hanif Chaudhry et al. Open-channel flow, volume 523. Springer,
2008.
[9] Amaresh Dalal, V Eswaran, and G Biswas. A finite-volume method
for navier-stokes equations on unstructured meshes. Numerical heat
transfer, part B: fundamentals, 54(3):238–259, 2008.
[10] Jianwei Dong and Litao Zhang. Analytical solutions to the 1d compress-
ible isothermal navier–stokes equations with density-dependent viscos-
ity. Journal of Mathematical Physics, 62(12), 2021.

15
[11] C Ross Ethier and DA Steinman. Exact fully 3d navier–stokes solu-
tions for benchmarking. International Journal for Numerical Methods
in Fluids, 19(5):369–375, 1994.

[12] Vitaly L Ginzburg. Ld landau. Ferroelectrics, 267(1):69–75, 2002.

[13] Volker Gravemeier, Wolfgang A Wall, and Ekkehard Ramm. A three-

level finite element method for the instationary incompressible navier–
stokes equations. Computer Methods in Applied Mechanics and Engi-
neering, 193(15-16):1323–1366, 2004.

[14] Sanaz Hadian, Hossein Afzalimehr, and Sajjad Ahmad. Effects of chan-
nel width variations on turbulent flow structures in the presence of three-
dimensional pool-riffle. Sustainability, 15(10):7829, 2023.

[15] Boris Haspot. Existence of global strong solution for the compress-
ible navier–stokes equations with degenerate viscosity coefficients in 1d.
Mathematische Nachrichten, 291(14-15):2188–2203, 2018.

[16] Dragoş Iftimie. The 3d navier-stokes equations seen as a perturbation

of the 2d navier-stokes equations. Bulletin de la Société Mathématique
de France, 127(4):473–517, 1999.

[17] Farhoud Kalateh, Mohammad Jahandar Malekabadi, and Yousef Has-

sanzadeh. A novel method for numerical simulation of gradually varied
flow profile based on adaptive newton raphson method. Irrigation Sci-
ences and Engineering, 40(1):167–176, 2017.

[18] DC Lo, Krishnan Murugesan, and DL Young. Numerical solution of

three-dimensional velocity–vorticity navier–stokes equations by finite
difference method. International journal for numerical methods in fluids,
47(12):1469–1487, 2005.

[19] Antoine Mellet and Alexis Vasseur. On the barotropic compressible

navier–stokes equations. Communications in Partial Differential Equa-
tions, 32(3):431–452, 2007.

[20] Andrei S Monin, Akiva M Yaglom, and TS Lundgren. Statistical fluid

mechanics, vol. ii. 1976.

16
 [21] Charles N Moore. Gn watson, a treatise on the theory of bessel functions.
1945.

[22] FWJ Olver. Unsolved problems in the asymptotic estimation of special

functions. Theory and application of special functions, pages 99–142,
1975.

[23] Byrne Perry. Open-channel hydraulics. ven te chow. mcgraw-hill,

new york, 1959. xviii+ 680 pp. illus. 17. Science, 131(3408) : 1215 −
−1215, 1960.

[24] Mohammad Mehdi Rashidi and Ganji Domairry. New analytical solution of
the three-dimensional navier–stokes equations. Modern Physics Letters B,
23(26):3147–3155, 2009.

[25] H Prashanth Reddy, M Hanif Chaudhry, and Jasim Imran. Computation of

gradually varied flow in compound open channel networks. Sadhana, 39:1523–
1545, 2014.

[26] Hunter Rouse. Open channel flow. by fm henderson. macmillan, 1966. 522
pp. 14.95. Journalof F luidM echanics, 29(2) : 414 − −415, 1967.

[27] Hunter Rouse et al. Engineering hydraulics: Proceedings of the fourth hy-
draulics conference iowa institute of hydraulic research june 12-15, 1949.
1950.

[28] Victor L Streeter, E Benjamin Wylie, and Keith W Bedford. Fluid mechanics.
Mc Graw Hill Education, 2014.

[29] Ali R Vatankhah and Said M Easa. Direct integration of manning-based

gradually varied flow equation. In Proceedings of the Institution of Civil
Engineers-Water Management, volume 164, pages 257–264. Thomas Telford
Ltd, 2011.

[30] Maurizio Venutelli. Direct integration of the equation of gradually varied

flow. Journal of irrigation and drainage engineering, 130(1):88–91, 2004.

[31] Frank M White. Fluid mechanics, mcgraw-hill. New York, 1994.

17