Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Analytical Solutions of Nonlinear Kinematic Wave Model

Kazumasa Mizumura1

Abstract: An analytical solution of the nonlinear kinematic wave model of overland flow on a sloping plane for time-varying excess
rainfalls of sinusoidal functions is presented by the method of characteristics. To obtain the closed form of the analytical solution,
Downloaded from ascelibrary.org by University of Technology, Sydney on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

discharge per unit width in the expression of Manning’s formula is approximated by a parabolic curve. The analytical solution is also
compared with the numerical one and their agreement is found to be good. The analytical solution is applied to rainfall and runoff
processes with given hyetograph using Fourier series analysis. The computational results of the analytical solution are compared with the
observed data. It explains the observation on an impermeable slope very well. It is found that the analytical solution is suitable for
estimating design flood from excess rainfall in a simple watershed.
DOI: 10.1061/共ASCE兲1084-0699共2006兲11:6共539兲
CE Database subject headings: Overland flow; Floods; Kinematic wave theory; Analytical methods; Rainfall.

Introduction computation using the method of characteristics for kinematic


wave routing. Govindaraju et al. 共1988兲 approximately solved the
In hydrology, there are several nonlinear problems to be solved. diffusion wave model for overland flow using a weighted residual
Approaches or investigations to nonlinear problems have been method. Mizumura 共1989, 1992兲 tried to find analytical solutions
improving hydrology for a long time. One of the analyzing meth- of the kinematic wave equation for time and space varying excess
ods of nonlinearity in hydrology is to solve partial differential rainfall. The selections of the integration range for the character-
equations as the governing equation in rainfall and runoff pro- istic differential equations were not correct. Using the kinematic
cesses. Izzard 共1946兲 presented an importance of overland flow wave model, peak discharge and time to peak were predicted by
over developed surfaces such as highway pavement or airfield Cadavid et al. 共1991兲. Ponce 共1991兲 reviewed character and ap-
runway. Lighthill and Whitham 共1955兲 gave the theoretical plicability of the kinematic wave model. Tayfur and Kavvas
background of kinematic waves. Henderson and Wooding 共1964兲 共1994兲 studied overland flow in the interacting rill–interrill area
developed analytical solutions to the kinematic wave model of using the kinematic wave model. Smith et al. 共1995兲 and Borah et
overland flow over a sloping plane for constant rainfall in time al. 共2002兲 succeeded in estimation of surface and subsurface
and space. Prasad 共1967兲 numerically predicted basin response by storm water runoff, propagation of flood waves, soil erosion, and
a nonlinear ordinary differential equation including the loop effect transport of sediment, nutrients, and pesticides in watersheds.
of a rating curve. Smith and Woolheiser 共1971兲 and Beven 共1982兲 Woolhiser 共1996兲 also suggested using physically based runoff
compared overland flow on an infiltrating surface with observa- model to predict flood, although it might contain uncertainties.
tions. The computations were conducted using Richards equation Singh 共1996兲 summarized all the kinematic wave modeling in
and kinematic wave model. Muzik 共1974兲 conducted the experi- analytical and numerical methods. It included all techniques to
treat or compute overland flow using the kinematic wave model.
mental work of the overland flow with unsteady uniform rainfall
Onizuka and Odai 共1998兲 compared the numerical solution of St.
on the slope and found the property of discharge hydrograph.
Venant equations with that of Burgers’ equation. Odai 共1999兲 nu-
Ponce and Simons 共1977兲 used an analytical solution of the lin-
merically approximated St. Venant equation using the nonlinear
earized equation set to develop criteria for the applicability of the
kinematic wave equation. Garg and Sen 共2001兲 presented the
kinematic wave model to open channel flow. Hjelmfelt 共1981兲
finite-element method 共FEM兲 computation of the kinematic wave
obtained the analytical solution of the kinematic wave model for
model. Kavvas 共2003兲 developed probabilistic description of non-
constant excess rainfall in time and space. Parlange et al. 共1981兲
linear hydrologic processes by the conservation equations with
found an exact analytical solution of the kinematic wave model
the stochastic approach. Moramarco and Singh 共2002兲 gave the
when the excess rainfall is a function of time. Cundy and Tento
accuracy condition for the kinematic wave method for spatial-
共1985兲 analyzed the overland flow on the porous layer using the
varying rainfall excess as KF2 艌 20, where K⫽kinematic wave
kinematic wave equation. Eggert 共1987兲 investigated upstream
number and F⫽Froude number at plane end. This is the same as
the accuracy condition for the kinematic wave method for time-
1
Professor, Dept. of Civil Engineering, Kanazawa Institute of varying rainfall excess. The differences of this study from the
Technology, 7-1, Ogigaoka, Nonoichimachi, Ishikawa Prefecture previous ones 共Mizumura 1989, 1992兲 are the comparison of the
921-8812, Japan. E-mail: mizu@neptune.kanazawa-it.ac.jp analytical solution with the numerical ones, the comparison of the
Note. Discussion open until April 1, 2007. Separate discussions must
analytical one with the observations, and the application of the
be submitted for individual papers. To extend the closing date by one
month, a written request must be filed with the ASCE Managing Editor. analytical solution to given hyetographs in time. This research
The manuscript for this paper was submitted for review and possible supplies the deficiency in the above two papers 共Mizumura 1989,
publication on March 15, 2004; approved on January 31, 2006. This 1992兲. The analytical solution might be suitable in a context of
paper is part of the Journal of Hydrologic Engineering, Vol. 11, No. 6, lumped modeling addressed to estimate the design flood at the
November 1, 2006. ©ASCE, ISSN 1084-0699/2006/6-539–546/$25.00. outlet of gauged basins. In general, the basin could be schema-

JOURNAL OF HYDROLOGIC ENGINEERING © ASCE / NOVEMBER/DECEMBER 2006 / 539

J. Hydrol. Eng., 2006, 11(6): 539-546


tized as one single plane and the calibration procedure can be
based on observed flood events. A similar consideration can be
done for the case of basins represented through an ensemble of
planes and channels. Thus, the analytical solution forms a funda-
mental tool in estimation of the design flood in basins. This re-
search is the first step to rainfall and runoff analysis using the
analytical solution.

Kinematic Wave Model

The kinematic wave model for flow over a sloped plane is given
Downloaded from ascelibrary.org by University of Technology, Sydney on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

by 共Lighthill and Whitham 1955兲


⳵h ⳵Q
+ = q共x,t兲 共1兲
⳵t ⳵x
in which h⫽flow depth; t⫽time; Q = uh; u⫽velocity; x⫽distance
downslope 共measured from upstream end on the plane兲; and
q共x , t兲⫽lateral inflow rate per unit width or rainfall excess 共rain-
fall minus infiltration兲. According to Manning’s formula, the dis-
charge per unit width Q is assumed to be
Fig. 1. Approximation of discharge per unit width of Manning’s
冑S0 formula by polynomial
Q= h5/3 ⬵ ah2 + bh + c 共2兲
n
in which S0⫽bottom slope; n⫽Manning’s roughness coefficient;
⳵h ⳵h
and a , b , c⫽constant coefficients to approximate discharge per + 共2ah + b兲 = q共x,t兲 共10兲
unit width. The coefficients a, b, and c are determined by the ⳵t ⳵x
condition such as This is the fundamental equation of the kinematic wave model in

冕冉 冑S0
冊 this study. This holds in the domain for x 艌 0 and t 艌 0 when
hd 2
S= ah2 + bh + c − h5/3 dh → min 共3兲 accuracy of the kinematic wave approximation is shown.
0 n

This is obtained by
Analytical Solutions for Time-Varying Excess
⳵S ⳵S ⳵S Rainfall
= = =0 共4兲
⳵a ⳵b ⳵c
When the excess rainfall is time varying such as q共x , t兲 = q共t兲, the
Thus, solving the three equations in Eqs. 共4兲, we get
characteristic differential equation of Eq. 共10兲 is given by
冑S0
a = 0.730 h−1/3
d 共5兲 dt dx dh
n = = = d␴ 共11兲
1 2ah + b q共t兲
冑S0 From the first and third terms, we have
b = 0.292 h2/3
d 共6兲
n


t

冑S0 h= q共␴兲d␴ + h␶ 共12兲



c = 0.0146 h5/3
d 共7兲
n
in which ␶⫽time that characteristic starts; h␶⫽water depth at
in which hd⫽maximum 共downstream兲 water depth during the t = ␶; and ␴⫽dummy variable. From the first and second terms,
computation. The comparison of Eqs. 共5兲–共7兲 with Q · n / 冑S0 is we get
shown in Fig. 1 in the case of hd = 1 m. This indicates that the


t
applicability of Eq. 共2兲 is reasonable. The boundary and initial
conditions are given as follows: x= 共2ah + b兲d␴ + ␰ 共13兲

h共x = 0,t兲 = 0 共8兲
␰ = x coordinate that characteristic starts at t = 0 共␶ = 0兲. If ␶ is not
zero, ␰ = 0. Fig. 2 indicates that ␶ 艌 0 for ␰ = 0 and ␰ 艌 0 for ␶ = 0.
h共x,t = 0兲 = h0共x兲: steady flow condition 共9兲 The substitution of Eq. 共12兲 into Eq. 共13兲 gives
When a = 0 in Eq. 共2兲, Eq. 共1兲 becomes linear and supports the

冕冕
t ␩
assumption of the unit hydrograph method. The solution of this
linear partial differential equation gives the result that the up- x−␰= 2aq共␴兲d␴d␩ + 共2ah␶ + b兲共t − ␶兲 共14兲
␶ ␶
stream hydrograph propagates downstream without any change of
its shape. With the use of Eq. 共2兲 for the nonlinear form 共a ⫽ 0兲, in which ␩⫽dummy variable. Using Eqs. 共12兲 and 共14兲, the gen-
Eq. 共1兲 is written as eral solution of Eq. 共11兲 is obtained as follows:

540 / JOURNAL OF HYDROLOGIC ENGINEERING © ASCE / NOVEMBER/DECEMBER 2006

J. Hydrol. Eng., 2006, 11(6): 539-546


t共s,␶兲 = s + ␶ 共20兲


t
h共t,␶兲 = q共␴兲d␴ 共21兲


t
x共t,␶兲 = 关2ah共␴,␶兲 + b兴d␴ 共22兲

in which the range of the integration is from ␶ to t. The variable


s is the difference between t and ␶.
Downloaded from ascelibrary.org by University of Technology, Sydney on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

Analytical Solutions for Excess Rainfall


of Sinusoidal Function

To investigate the temporal properties of the nonlinear solution,


the following excess rainfall of sinusoidal type as the system
input is given as
q共t兲 = q0 for t ⬍ 0 共23兲

Fig. 2. Explanation of characteristics and x − t plane =q0 + A sin ␻t for t 艌 0 共24兲


in which q0; A; and ␻⫽constants to define excess rainfall of sinu-

冕 冕冕
t t ␩ soidal function and q0 艌 A is assumed. For t ⱕ 0, the steady solu-
h − h␶ − q共␴兲d␴ = f x − ␰ − 2aq共␴兲d␴ d␩ − 2a共t − ␶兲 tion h0共x兲 is easily obtained. By using the steady solution as the


␶ ␶ ␶
initial condition and Eq. 共19兲, the analytical solution after the


⫻ h− 冕

t


q共␴兲d␴ − b共t − ␶兲 共15兲
limiting characteristic arrives is given by

h=
A
共cos ␻␶ − cos ␻t兲 −
b
+
b2 q0x 2q0A
+ − 冋
共t − ␶兲cos ␻␶
␻ 2a 4a2 a ␻


in which f共·兲⫽arbitrary function to be determined by initial con-
dition of water depth. Since h共x , t = 0兲 = h0共x兲 = 共冑b2 + 4aq0x − b兲 / 2Aq0 1/2
共sin ␻t − sin ␻␶兲 共25兲
共2a兲 and h␶ = 共冑b2 + 4aq0␰ − b兲 / 共2a兲 are considered, Eq. 共15兲 at
+
␻2
t = ␶ = 0 is reduced to
To obtain ␶ for given t and x, we proceed as follows: From Eq.
冑b2 + 4aq0x − 冑b2 + 4aq0␰ 共20兲
= f共x − ␰兲 共16兲
2a s=t−␶ 共26兲
Therefore, the arbitrary function f共·兲 is determined as follows: From Eq. 共21兲, h共t , ␶兲 is obtained as
冑b2 + 4aq0共x + ␰兲 − 冑b2 + 4aq0␰

t
f共x兲 = 共17兲 A
2a h共t,␶兲 = 共q0 + A sin ␻␴兲d␴ = q0共t − ␶兲 − 共cos ␻t − cos ␻␶兲
␶ ␻
Substituting the above equation into Eq. 共15兲, we have 共27兲


t
1 2 b The substitution of Eq. 共27兲 into Eq. 共22兲 gives
兵b + 4aq0关W1 − 2a共t − ␶兲h兴 − 4ac其1/2 −

冕再 冋 册 冎
h= q共␴兲d␴ +
␶ 2a 2a t
A
共18兲 x共t,␶兲 = 2a 共␴ − ␶兲q0 − 共cos ␻␴ − cos ␻␶兲 + b d␴
␶ ␻

冋 册
in which W1 = x − 兰t␶兰␩t 2aq共␴兲d␴d␩ + 共t − ␶兲关兰t␶2aq共␴兲d␴ − b兴. Solv-
ing the above equation with respect to h, we obtain the following: 2aA sin ␻t − sin ␻␶
= aq0共t − ␶兲2 − − 共t − ␶兲cos ␻␶

再 1
h = − W2 + W22 − 关W2 − 共t − ␶兲q0兴2 + 2 共b2 + 4aq0W1兲
4a
冎 1/2
+ b共t − ␶兲
␻ ␻
共28兲

共19兲 From Eq. 共28兲 we obtain the following nonlinear equation with
respect to ␶:

冋 册
in which W2 = 共b / 2a兲 − 兰t␶q共␴兲d␴ + 共t − ␶兲q0. In Fig. 2, the solution
in Region 1 is dependent on the initial condition and the solution 2aA sin ␻t − sin ␻␶
x − aq0共t − ␶兲2 + − 共t − ␶兲cos ␻␶ − b共t − ␶兲
in Region 2 is dependent on the initial and upstream boundary ␻ ␻
conditions. The two regions are classified by the limiting charac-
=0 共29兲
teristic 共Eagleson 1970兲. To obtain the solution in Region 2 we
may proceed as follows: By solving Eq. 共11兲, we have the follow- The characteristic that passes the given point 共x , t兲 in the x − t
ing equation for time ␶, which is the time a characteristic starts at plane starts at x = 0 and t = ␶. The value of ␶ is obtained by nu-
x = 0: merically solving Eq. 共29兲. When ␶ = 0, Eq. 共29兲 coincides the

JOURNAL OF HYDROLOGIC ENGINEERING © ASCE / NOVEMBER/DECEMBER 2006 / 541

J. Hydrol. Eng., 2006, 11(6): 539-546


limiting characteristic that passes through the origin in the x − t it, we consider the stability of the solutions 关Eqs. 共25兲 and 共35兲兴.
plane. It is expressed as After the substitution of Eq. 共29兲 into Eq. 共25兲, the inside of the
root of Eq. 共25兲 is given by
x − aq0t2 +
2aA

+


sin ␻t

− t − bt = 0 共30兲
D= 冋 b
+ 共t − ␶兲q0 册 2
⬎0 共36兲
2a
Defining the left side of Eq. 共29兲 by G共␶兲, we have
Eq. 共36兲 is always positive or zero. Thus, Eq. 共25兲 is always stable
G共␶兲 = x − aq0共t − ␶兲 +2


2aA sin ␻t − sin ␻␶

− 共t − ␶兲cos ␻␶ 册 and does not break. Eq. 共35兲 is the solution for Eq. 共30兲, which is
positive. This shows that Eq. 共35兲 is the solution before the lim-
iting characteristic arrives. Using Eq. 共30兲, the inside of the root
− b共t − ␶兲 共31兲 of Eq. 共35兲, D, is written by
Downloaded from ascelibrary.org by University of Technology, Sydney on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

冉 冊
There exists only a solution of G共␶兲 = 0 for given x ⬎ 0 and t ⬎ 0 2
b
because there is a characteristic that passes through the point D⬎ + q 0t ⬎0 共37兲
共x , t兲. This indicates a point ␶ where a characteristic crosses the t 2a
axis as shown in Fig. 2. Taking the derivative of Eq. 共31兲 with Thus, D is always greater than zero. Therefore, the solutions 关Eqs.
respect to ␶ 共25兲 and 共35兲兴 are always stable for any q0 and A.
dG共␶兲
= 2a共t − ␶兲共q0 + A sin ␻␶兲 + b 共32兲
d␶ Comparison with Numerical Solutions
Since dG共␶兲 / d␶ ⬎ 0 for 0 ⬍ ␶ ⬍ t , G共0兲 ⬍ 0, and G共t兲 ⬎ 0, only a
solution ␶ exists and the solution ␶ in Eq. 共31兲 is numerically To check the verification of the analytical solutions, the numerical
obtained as follows: If G共t / 2兲 ⬎ 0, the solution ␶ is 0 ⬍ ␶ ⬍ t / 2. solution is used. To compare the analytical solution with the nu-
Otherwise, t / 2 ⬍ ␶ ⬍ t. The interval 关0 , t兴 is divided into two sub- merical solution, Eq. 共10兲 is transformed to the Lax–Wendroff
intervals 关0 , t / 2兴 and 关t / 2 , t兴. The subinterval including ␶ is also second-order scheme 共abbreviated by FDM兲 共Liggett and Cunge
divided into two again. This process is continued N times until the 1975兲 as
length of the subinterval t / 2N including the point ␶, which be-
⌬t i
comes much smaller than 1. At the final stage, the coordinate of hi+1 i
j = hj − 共F − Fij−1兲
the center point of the subinterval is approximately considered to 2⌬x j+1
equal ␶.
In Eq. 共25兲, when ␻ goes to infinity or zero, the solution ap-
proaches the steady solution
+ 冉 冊
1 ⌬t
2 ⌬x
2
关Hij+1/2共Fij+1 − Fij兲 − Hij−1/2共Fij − Fij−1兲兴

1 i
冑b2 + 4aq0x − b + 共qi+1
2
+ qij兲 共38兲
lim h = = h0共x兲 共33兲
␻→⬁ 2a
in which hij⫽water depth at x = 共j − 1兲⌬x and
t = 共i − 1兲⌬t ; q j⫽lateral inflow at x = 共j − 1兲⌬x and t = 共i − 1兲⌬t;
i

冑b2 + 4aq0x − b ⌬t⫽time increment; ⌬x⫽space increment; Hij+1/2 = H关共hij+1


lim h = = h0共x兲 共34兲 + hij兲 / 2兴; Hij−1/2 = H关共hij + hij−1兲 / 2兴; and H = ⳵F / ⳵h = 2ah + b. The su-
␻→0 2a
perscript i and subscript j indicate time and space steps, respec-
The above two equations indicate the decrease of the influence of tively. To obtain stable numerical solutions, the following condi-
time variation for small or large ␻. The solution before the lim- tion for the Lax–Wendroff second-order scheme must be satisfied
iting characteristic arrives is derived setting ␶ = 0 in Eq. 共25兲. It is 共Liggett and Cunge 1975兲
given by
⌬t 艋 ⌬x/H 艋 ⌬x/Hmin = ⌬x/b 共39兲
A
h = 共1 − cos ␻t兲 −

b
2a
+
4a2
+冉
b2 q0x 2q0At 2Aq0 sin ␻t
a


+
␻2
冊 1/2
In this study, ⌬x = 1.0 m and ⌬t = ⌬x / 共2b兲 s were selected.
For illustrative numerical computation, q0 = 100 mm/ h,
共35兲 A = 0.9⫻ q0 mm/ h, ␻ = 2␲ / 共10⫻ 60兲 s−1, S0 = 0.1, and n = 0.03 are
used. Fig. 3 represents the water depth changes in time at
This solution also has the property such that Eqs. 共33兲 and 共34兲
x = 200 and 1,000 m. Two kinds of water depth computations, the
have. Thus, for x → ⬁, ␻ → ⬁, or ␻ → 0, the solution approaches
nonlinear and numerical solutions, are compared. The analytical
the steady solution, since the steady part becomes relatively pre-
solution is strictly the same as the numerical ones. For at least
dominant to the unsteady part.
95% accuracy of the kinematic wave approximation after one
period, the dimensionless period has to be greater than 171 共Singh
1996兲. Thus, the wave period T p follows
Stability Analysis
171hd
Since Eq. 共10兲 is nonlinear, the rise and fall of the hydrograph do Tp 艌 ⬵ 24.43 s 共40兲
S 0u d
not take place symmetrically, the fall occupying a larger time than
the rise. The face and back of the hydrograph become steeper and in which hd⫽downstream water depth for the overland flow and
wider, respectively, as it travels. It causes shock phenomenon and ud⫽downstream velocity for the overland flow. The illustrative
finally breaks down if the right side of Eq. 共10兲 has a specified example satisfies Eq. 共40兲. In order to satisfy the accuracy condi-
function such as the nonlinear shallow water equations. To avoid tion of kinematic wave method, the following equation holds:

542 / JOURNAL OF HYDROLOGIC ENGINEERING © ASCE / NOVEMBER/DECEMBER 2006

J. Hydrol. Eng., 2006, 11(6): 539-546



A b
h= 兺 m 共cos ␻m␶ − cos ␻mt兲 − 2a
m=1 ␻m

+ 冋 b 2 q 0x
4a 2 +
a

2q0Am
m=1 ␻m


共t − ␶兲cos ␻m␶ 册
+ 冋兺

m=1
2Amq0
␻m2
共sin ␻mt − sin ␻m␶兲 册 1/2

共45兲

and

Downloaded from ascelibrary.org by University of Technology, Sydney on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

A b
h= 兺 ␻
m

m=1 m
共1 − cos ␻mt兲 −
2a

+ 冉 b 2 q 0x
4a2
+
a

m=1
2q0Amt
␻ m

+
m=1

2Amq0 sin ␻mt

␻m2
冊 1/2

共46兲
in which ␻m = m␻. The limiting characteristic Eq. 共30兲 is given by

冉 冊

2aAm sin ␻mt
x − aq0t + 2

m=1 ␻m ␻m
− t − bt = 0 共47兲

Applying Eq. 共29兲, ␶ is computed by

冋 册
Fig. 3. Comparison of nonlinear solution with FDM ⬁
2aAm sin ␻mt − sin ␻m␶
x − aq0共t − ␶兲2 + 兺
m=1 ␻m ␻m
− 共t − ␶兲cos ␻m␶

冉 冊 3/4 − b共t − ␶兲 = 0 共48兲


n2Lg
hd ⬍ 共41兲 The analytical solutions are found to be applicable to given hy-
20
etographs. To verify the theoretical results, they are compared
in which L⫽plane length and g⫽gravitational acceleration. Using with two experimental data sets. The experiments of overland
Eq. 共41兲 and Manning’s formula, we derive flow were conducted by sprinklers on an experimental slope of
10° 共Kishii 1981兲. Since the experimental slope is permeable, the
205q40 overland flow is Hortonian. The experimental slope 共13.8 m long,
L⬎ = 2.89 ⫻ 10−6 共m兲 共42兲 4 m wide, and 0.50 m deep兲 consisted of uniform soil, where soil
S20n6g5
was Kanto loam, the average hydraulic conductivity was 4.6
Since L is usually greater than the right side of Eq. 共42兲, it indi- ⫻ 10−2 cm/ s, and a layer beneath the Kanto loam in the slope was
cates that the analysis of the kinematic wave model holds in this impermeable. The rainfall intensity was 104.7 mm/ h and the re-
case. sulting runoff ratio to rainfall was 0.615 that determined excess
rainfall. The rainfall duration was 64.2 min and the runoff was
observed from the beginning of the experiment for 5 h and
Applications to Observations 50 min. The runoff discharge at the downstream end was plotted
in Fig. 4. The measurement and collection of the runoff discharge
The given hyetograph is expressed by the following Fourier sine continued from 0 to 5 h and 50 min to determine the runoff ratio
series to rainfall. The deficiency of water 共difference between excess
rainfall and runoff兲 in Fig. 4 was supplied after 100 min by the
⬁ hydrograph recession. But since the analytical solution of the ki-
q共t兲 = q0 + 兺 Am sin m␻t
m=1
共43兲 nematic wave model becomes small early in time, there is a dif-
ference between the theory and the observation. In the same fig-
ure, the computed runoff discharges are also shown for different
in which m⫽integer. The coefficient Am is computed by values of Manning’s roughness coefficient n. Due to the initial


loss and excess rainfall, runoff started after 5 min from the start
T/2
4 of the rain. The number of the terms used of Fourier sine series
Am = 关q共t兲 − q0兴sin m␻t dt 共44兲
T 0
for this computation is 30. In these computations, the numbers 30
of Fourier series terms are suitable, because the input hyetograph
in which T / 2⫽time base length of water depth hyetograph and is simple in time. But when the input hyetograph is not simple,
T = 2␲ / ␻. The analytical solution for Eq. 共43兲 is expressed by the the number must be increased to more than 50. The steady flow q0
summation of the analytical solution to Am sin m␻t. For Eq. 共43兲, is selected to be 1 / 100 of the excess rainfall intensity. The com-
although a nonlinear solution is not applicable to the principle of putation started at 5 min from the beginning of the rainfall. For a
superposition, the analytical solutions of Eqs. 共25兲 and 共35兲 be- constant rainfall, the runoff discharge at the downstream end in-
come creases almost linearly and becomes constant. At this stage the

JOURNAL OF HYDROLOGIC ENGINEERING © ASCE / NOVEMBER/DECEMBER 2006 / 543

J. Hydrol. Eng., 2006, 11(6): 539-546


Downloaded from ascelibrary.org by University of Technology, Sydney on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Comparison of nonlinear solution with observed data

Fig. 4. Comparison of nonlinear solution with observed data though the flow is caused on the permeable slope and not fully
turbulent, the analytical solutions are coincident with the obser-
vations by selecting the appropriate value of n. These experimen-
tal data satisfy the accuracy condition of the kinematic wave
flow condition is steady. This condition leads to the relation be- method.
tween water depth at the downstream end and Manning’s rough-
ness coefficient by Manning’s formula. The application of a linear
relation at the downstream end between observed time to peak t p Conclusion
and water depth hd to the rising rimb for a constant excess rainfall
q0 indicates hd = q0t p. The solution for these two relations gives The discharge per unit width in Manning’s formula is approxi-
the approximate numerical value of Manning’s roughness coeffi- mated by polynomials of water depth. Thus, the analytical solu-
cient. Since Manning’s roughness coefficient is large, the subsur-
face flow plays an important role in these data. The hydrograph of
the analytical solution for n = 4.04 rises faster but falls very well.
The observed hydrograph continued to flow by the influence of
the subsurface flow. The hydrograph of the analytical solution for
n = 12.45 rises as the observations do and falls slower than the
observations. It is considered that the above-mentioned difference
between the analytical solution and observations is caused by the
subsurface flow in the experimental slope and the friction law.
Fig. 5 shows a comparison of the analytical solution with the
observed data of the overland flow on a permeable slope of 8.9°
共Hatano and Inokuma 1979兲. The hydrograph was observed at the
downstream end. In these experiments, the overland flow was
produced on a permeable slope that was 520 cm long, 22.5 cm
wide, and consisted of artificial gravels whose diameter is 0.9 cm
in the slope surface. A layer beneath the artificial gravels in the
slope was impermeable. The rainfall intensity was 759 mm/ h.
The flow was not fully turbulent. The numerical value of
n = 36.3 is obtained from the time that the flow becomes steady.
The artificial gravel gives great resistance. Fig. 6 compares the
analytical solution with the observation of the overland flow on a
slope of 0.04 共Yamada and Kikkawa 1979兲. The hydrograph was
observed at the downstream end. In their experiments, the over-
land flow is also produced in a permeable slope that was 440 cm
long, 13 cm wide, and consisted of glass spheres whose diameter
was 1.6 cm in the slope surface. A layer beneath the glass spheres
in the slope was impermeable. The rainfall intensity was
87 cm/ min. The numerical value of n = 0.27 is also obtained. Al- Fig. 6. Comparison of nonlinear solution with observed data

544 / JOURNAL OF HYDROLOGIC ENGINEERING © ASCE / NOVEMBER/DECEMBER 2006

J. Hydrol. Eng., 2006, 11(6): 539-546


tion of the nonlinear kinematic wave model for time-varying ex- ␶ ⫽ time that characteristic starts at x = 0;
cess rainfalls of sinusoidal functions is derived in the closed form. ␻ ⫽ angular frequency of time-varying function of
It is compared with the numerical solutions of the Lax–Wendroff excess rainfall; and
second-order scheme and found to be in good agreement with ␻m ⫽ mth component of angular frequency of time-varying
them. The application of Fourier series theory to the analytical function.
solution computes the water depth hydrograph for the given hy-
etograph. As the verification of this theory, the analytical solution
is compared with the observations. Their agreements are satisfac- References
tory if Manning’s roughness coefficient is appropriately selected.
Beven, K. 共1982兲. “On subsurface stormflow: Prediction with simple ki-
nematic theory for saturated and unsaturated flows.” Water Resour.
Acknowledgments Res., 18共6兲, 1627–1633.
Downloaded from ascelibrary.org by University of Technology, Sydney on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

Borah, D. K., Xia, R., and Bera, M. 共2002兲. “A dynamic watershed simu-
The writer wishes to thank T. Kishii, a professor at Kanazawa lation model.” Mathematical models of small watershed hydrology
Institute of Technology, for providing the observed data of and applications, V. P. Singh and D. K. Frevert, eds., Chap. 5, Water
overland flow experiments. He also wishes to express his sincere Resources Publications, LLC, Highland Ranch, Colo., 113–166.
appreciation to anonymous referees who gave comments and sug- Cadavid, L., Obeysekera, J. T. B., and Shen, H. W. 共1991兲. “Flood-
gestions to strengthen the paper. frequency derivation from kinematic wave.” J. Hydraul. Eng., 117共4兲,
489–510.
Cundy, T. W., and Tento, S. W. 共1985兲. “Solution to the kinematic wave
approach to overland flow routing with rainfall excess given by Phil-
Notation ip’s equation.” Water Resour. Res., 21共8兲, 1132–1141.
Eagleson, P. 共1970兲. Dynamic hydrology, McGraw-Hill, New York, 337–
The following symbols are used in this paper: 344.
A ⫽ amplitude of sinusoidal function of time-varying Eggert, K. G. 共1987兲. “Upstream calculation of characteristics for kine-
excess rainfall; matic wave routing.” J. Hydraul. Eng., 113共6兲, 743–752.
Am ⫽ amplitude of mth component in Fourier sine series; Garg, N. K., and Sen, D. J. 共2001兲. “Integrated physically based rainfall-
a , b , c ⫽ coefficients to approximate Manning’s formula; runoff model using FEM.” J. Hydrol. Eng., 6共3兲, 179–188.
D ⫽ function in root of analytical solution; Govindaraju, R. S., Jones, S. E., and Kavvas, M. L. 共1988兲. “On the
diffusion model for over-land flow. I: Solution for steep slope.” Water
F ⫽ Froude number at plane end;
Resour. Res., 24共5兲, 734–744.
f共·兲 ⫽ function to be determined by initial condition for Hatano, M., and Inokuma, H. 共1979兲. “Experiments of overland flow on
time-varying excess rainfall; steep rough surfaces and estimation of resistance formula.” Proc.,
G共␶兲 ⫽ equation to determine ␶; 23rd Japanese Conf. on Hydraulics, Vol. 23, JSCE, Tokyo, 331–338
g ⫽ gravitational acceleration; 共in Japanese兲.
H ⫽ ⳵Q / ⳵h = 2ah + b; Henderson, F. M., and Wooding, R. A. 共1964兲. “Overland flow and
h ⫽ water depth; groundwater flow from a steady rainfall of finite duration.” J. Geo-
hd ⫽ water depth at downstream end; phys. Res., 69共8兲, 1531–1540.
h0 ⫽ steady water depth; Hjelmfelt, A. T. 共1981兲. “Overland flow from time-distributed rainfall.” J.
h␶ ⫽ water depth at t = ␶; Hydr. Div., 107共2兲, 227–238.
K ⫽ kinematic wave number; Izzard, C. 共1946兲. “Hydraulics of runoff from developed surfaces.” Proc.,
L ⫽ length of catchment; Highway Research Board, Vol. 26, New York, 129–150.
m ⫽ integer; Kavvas, M. L. 共2003兲. “Nonlinear hydrologic processes: Conservation
n ⫽ Manning’s roughness coefficient; equations for determining their means and probability distributions.”
J. Hydrol. Eng., 8共2兲, 44–53.
Q ⫽ discharge per unit width;
Kishii, T. 共1981兲. “On the experiment of surface runoff from experimental
q共t兲 ⫽ time-varying excess rainfall;
slope.” Rep. of National Research Center for Disaster Prevention, No.
q共x , t兲 ⫽ general excess rainfall; 25, Tsukuba, Japan, 19–27 共in Japanese兲.
q0 ⫽ constant part of excess rainfall in time; Liggett, J. A., and Cunge, J. A. 共1975兲. “Numerical methods of solution
S ⫽ summation of errors due to polynomial of the unsteady flow equations.” Unsteady flow in open channels, K.
approximation of Manning’s formula; Mahmood and V. Yevjevich, eds., Water Resources Publications, Fort
S0 ⫽ bottom slope; Collins, Colo., 126–128.
s ⫽ t − ␶; Lighthill, M. J., and Whitham, G. B. 共1955兲. “On kinematic waves. I:
T ⫽ twice time base of given hyetograph; Flood movement in long rivers.” Proc. R. Soc. London, Ser. A, 229,
T p ⫽ wave period of excess rainfall; 281–316.
t ⫽ time; Mizumura, K. 共1989兲. “Nonlinear analysis of rainfall and runoff process.”
t p ⫽ time to peak; Proc., Int. Conf. Channel Flow and Catchment Runoff, B. C. Yen, ed.,
u ⫽ mean velocity of overland flow; Charlottesville, Va., 196–205.
ud ⫽ velocity at downstream end; Mizumura, K. 共1992兲. “Nonlinear analysis of rainfall and runoff process.”
W1 , W2 ⫽ dummy functions; Catchment runoff and rational formula, B. C. Yen, ed., Water Re-
sources Publications, Fort Collins, Colo., 128–144.
x ⫽ coordinate along flow direction;
Moramarco, T., and Singh, V. P. 共2002兲. “Accuracy of kinematic wave
⌬t ⫽ time increment; and diffusion wave for spatial-varying rainfall excess over a plane.”
⌬x ⫽ space increment; Hydrolog. Process., 16, 3419–3435.
␩ ⫽ dummy integration variable; Muzik, I. 共1974兲. “Laboratory experiments with surface runoff.” J. Hydr.
␰ ⫽ x coordinate that characteristic starts at t = 0; Div., 100共4兲, 501–515.
␴ ⫽ dummy integration variable; Odai, S. N. 共1999兲. “Nonlinear kinematic-wave model for predicting

JOURNAL OF HYDROLOGIC ENGINEERING © ASCE / NOVEMBER/DECEMBER 2006 / 545

J. Hydrol. Eng., 2006, 11(6): 539-546


open-channel flow rate.” J. Hydraul. Eng., 125共8兲, 886–889. Smith, R. E., Goodrich, D. C., Woolheiser, D. A., and Unkrich, C. L.
Onizuka, K., and Odai, S. N. 共1998兲. “Burgers’ equation model for un- 共1995兲. “KINEROS—A kinematic runoff and erosion model.” Com-
steady flow in open channels.” J. Hydraul. Eng., 124共5兲, 509–512. puter models of watershed hydrology, V. P. Singh, ed., Chap. 20,
Parlange, J. Y., Rose, C. W., and Sanda, G. 共1981兲. “Kinematic flow Water Resources Publications, Highland Ranch, Colo., 697–732.
approximation of runoff on a plane: An exact analytical solution.” J. Smith, R. E., and Woolheiser, D. A. 共1971兲. “Overland flow on an infil-
Hydrol., 52, 171–176. trating surface.” Water Resour. Res., 7共4兲, 899–913.
Ponce, V. M. 共1991兲. “Kinematic wave controversy.” J. Hydraul. Eng., Tayfur, G., and Kavvas, M. L. 共1994兲. “Spatially averaged conservation
117共4兲, 511–525. equations for interacting rill–interrill area overland flows.” J. Hy-
Ponce, V. M., and Simons, D. B. 共1977兲. “Shallow wave propagation in draul. Eng., 120共12兲, 1426–1448.
open channel flow.” J. Hydr. Div., 103共12兲, 1461–1476. Woolhiser, D. A. 共1996兲. “Search for physically based runoff model—A
Prasad, R. 共1967兲. “A nonlinear hydrologic system response model.” J. hydrologic El Dorado.” J. Hydraul. Eng., 122共3兲, 122–129.
Hydr. Div., 93共4兲, 201–221. Yamada, T., and Kikkawa, H. 共1979兲. “The hydraulic study on the flow in
Singh, V. P. 共1996兲. Kinematic wave modeling in water resources: Sur- porous media 共IV兲.” Technical Rep., No. 25, Dept. of Civil Engineer-
Downloaded from ascelibrary.org by University of Technology, Sydney on 10/13/22. Copyright ASCE. For personal use only; all rights reserved.

face hydrology, Wiley, New York. ing, Tokyo Inst. of Technology, Tokyo, 55–66 共in Japanese兲.

546 / JOURNAL OF HYDROLOGIC ENGINEERING © ASCE / NOVEMBER/DECEMBER 2006

J. Hydrol. Eng., 2006, 11(6): 539-546

You might also like