643

Ser.B, 2007,19(5):643-652
NUMERICAL STUDY ON THE PURIFICATION PERFORMANCE OF
RIVERBANK
*
DAI Yu, WANG De-guan
College of Environmental Science and Engineering, Hohai University, Nanjing 210098, China,
E-mail: daiyu_hhu@yahoo.com.cn
(Received April 9, 2007; Revised September 1, 2007)
ABSTRACTDuring the rain time, the runoff infiltrates into
the riverbank through the collecting gutter and slope surface.
The city runoff is generally polluted by organic, oil, heavy
metal particulates, etc. The pollutants moving with the water
through the riverbank experience advection, dispersion,
diffusion, adsorption, biochemical reaction and plant uptaking
processes. In this article, a mathematical model was developed
to simulate the performance of pollutant removal of the
riverbank. The model took those main mechanisms into
account. The modified Richards equation was used in
simulating flow field. The mass balance law was employed in
deriving the equation for pollutant transport, where the
diffusion and dispersion were described with the Fick-type law,
the adsorption was macroscopically expressed as form isotherm,
and the bio-chemical degradation process was assumed to
follow the Monod kinetics. The NH
3
-N and TP were
considered in the present model. The mathematical model was
descritized with a finite element numerical model, which was
applied to two types of model riverbanks. In the model test, the
hydraulic loading was assumed to have the intermittent pattern
simulating the storm runoff of certain return period, and the
values of the rainfall runoff and concentrations of the NH
3
-N
and TP were taken from the model test. The computed overall
removal rates for the NH
3
-N and TP in 6 cases are in the range
of 88%-98%, 87%-97%, respectively. The differences between
the computed and tested overall removal rates for the NH
3
-N
and TP are less than 5%. The time-varying oscillation pattern
of the concentrations of the NH
3
-N and TP were rationally
simulated, which shows that the model presented in this article
can be used to assess the purification performance of the
riverbank constructed with sand or soil.
KEY WORDSriverbank, numerical modelling, purification,
pollutant, experiments
1. INTRODUCTION
Riverbank plays an important role in the
ecosystem of the river, which is the buffer area
between the river and the surrounding terrain. The
riverbank should not only be able to achieve the
goal of preventing floods and erosion, but should
also be able to support the development of a
balanced river ecosystem. The riverbank is
normally constructed with four parts: (1) the main
body mostly composed of sand or sand and clay
mixture, (2) the top part consisting of road, side
walks, planted surface and collecting gutter, (3)
slope face with a flat platform at the slope toe
normally planted with grass, flowers and shrub
zone on the slope and emergent plants on the flat
platform, (4) the erosion prevention part
constructed with rock riprap or cement rubber
masonry. The ground water goes through the
riverbank and connects to the river flow. When the
polluted rainfall runoff infiltrates into riverbank
through the collecting gutter, it flows inside the
riverbank, raises the ground water table, moves
with the groundwater and interacts with the soil.
The main mechanisms affecting the transport of the
solute (pollutants) in porous media are convection,
mechanical dispersion, molecular diffusion,
solid-solute interactions, various chemical and
biological reactions and decay phenomena.
* Biography: DAI Yu(1979-), Female, Ph. D. Student
644
There are several kinds of mathematical
models established for investigating the fates of
pollutants moving with the water in the ground or in
the wetlands. The wetland models are usually
conceptual models
[1-4]
. Among them a sophisticated
model includes hydrological, primary production,
sedimentation and nutrient sub-models
[5]
. The
hydrological sub-model is a regression equation
describing water quantity balance. The primary
production sub-model describes the multiply,
respiration, growth and dead cycling processes of
phytoplankton, macrophytes and periphton. The
sedimentation model describes the functions of
standing dead macrophytes, bottom detritus,
suspended solids, active sediment layers, and deep
sediments in nutrient cycle, especially in the
phosphorus cycle. The phosphorus sub-model
describes the balance and cycling process of total
phosphorus among water column, bottom detritus,
active sediment layers, and deep sediments. The
ground water pollution model basically describes
all the transportation and transformation processes
of the pollutants in the ground water and predicts
the fates and spatial and temporal distribution of the
pollutants in the soil. Those models consider the
advection-dispersion- diffusion process, assume the
pollutant being equilibrium partitioning between
water and solid phases, and biochemical reaction
being the first-order Monod and double Monod
kinetics
[6-8]
. In order to make the model more
universal, several models have been proposed based
on the concept of the terminal electron acceptor
process, and the so-called one-step model, two-step
model and partial equilibrium approach model have
been developed for predicting the translation and
biochemical degradation of the pollutants in the soil
[9]
.
According to the problem considered and
principles mentioned above, a mathematical model
is established to simulate the performance of
pollutant removal of the riverbank. The model takes
those main mechanisms into account. And various
rational assumptions are made in the formation of
mathematical model. The mathematical model is
discretized by the finite element numerical method.
The numerical model is applied to two types of
model riverbanks. The results from the
computations agree well with the results from
physical experiments, which shows that the model
presented in this article can be used to assess the
purification performance of the riverbank
constructed with sand or soil.
2. MATHEMATICAL MODEL
2.1 Mathematical model for water movement
2.1.1 Governing equation
The flow through the riverbank constructed
with sand follows the Darcy law. As the soil in the
riverbank consists of two zones i.e., the unsaturated
zone and saturated zone, water content is the
important variable under the consideration. Under
the assumption that interaction between air and
water is negligible, the flow in the riverbank is
governed by the modified Richards equation. In the
vertical planar flow
[10]
,
A A
xx xz
h h
K K K
t x x z z
u c c c c ( c | |
= +
| (
+
c c c c
\ . ¸ ¸

c
A A
zz zx r
h h
K K K S
z x
c c ( | |
+ ÷

c c
\ . ¸ ¸
|(
(1)
where, u is the volumetric water content,
3 3
L L
÷
(
¸ ¸
, is the pressure head, | , is
sink term,
h | L
r
S
1
T
÷
(
¸ ¸
, x, z are the spatial coordinates,
| | L , t is the time, | | T ,
A
xx
K ,
A
zz
K ,
A A
xz zx
K K =
are the components of a dimensionless tensor ,
A
K
K is the unsaturated hydraulic conductivity,
1
LT
÷
(
¸ ¸
, which is given by
( , , ) ( , ) ( , , )
s r
K h x z K x z K h x z = (2)
where is the saturated hydraulic
conductivity, and is the relative
unsaturated hydraulic conductivity, which is
defined as ratio of the unsaturated hydraulic
conductivity and the saturated hydraulic
conductivity. In order to take the evapotranspiration
of the plants growing on the riverbank into account,
a sink term, , is added, which represents the
volume of water uptaken by the plants per unit time
from a unit volume of soil. is defined as
( , )
s
K x z
( , , )
r
K h x z
r
S
r
S
( ) ( )
r
S h h S
p
o = (3)
where the water response function ( ) h o is
645
assumed as a dimensionless function of the soil
water pressure head, 0 1 o < < , and is the
potential water uptake rate, .
p
S
1
T
÷
(
¸ ¸
( ) h o is
assumed to be zero when the soil is close to
saturation (
s
h h = ), and is assumed to be zero when
the moisture in the soil is at wilting point
( ).There is a rang of soil water pressure head,
bounded by and , within which the
has the value of , i.e.,
w
h h =
u
h
d
h ( )
r
S h
p
S ( ) h o takes the value of
one, and its value of is linearly varies from 0.0 to
1.0 as the value of h varies from h
w
to
s
h and from
1.0 to 0.0 as the value of h varies from h
u
to
s
h .
The potential water uptake rate can be expressed as
( )
x z
p
d
L L
S h T
L
=
p
(4)
where
x
L is the width of the root area, | | L ,
is the depth of the root area, , is the soil
surface width related to each plant root,
z
L
| | L
d
L
| | L , and
is the potential transpiration rate,
p
T
1
LT
÷

¸
(
¸
[11]
.
2.1.2 Unsaturated zone hydraulic parameters
In order to make the Eq. (1) a closed form, a
set of the hydraulic parameter relations, i.e., the
water content ( ) h u and relative hydraulic
conductivity, for the unsaturated zone are
given as
r
K
[12-14]
,
( )
( )
[1 ]
s
s
sa r
r n
m
h
h
u u
u u
o
÷
= +
+
(5)
and
( ) ( )
{ }
( )
2
1
2
1 1
( , , )
1
s
s s
s
s
m
n n
r m
n
h h
K h x z
h
o o
o
÷
÷
(
÷ +
¸ ¸
=
(
+
¸ ¸
(6)
respectively, where
1
1 ,
s
s
m n
n
= ÷ >1
s
(7)
in which
sa
u and
s
u denote the residual and
saturated water contents, respectively. o ,
1
L
÷
(
¸ ¸
,
s
m ,
s
n are the nonlinear regression coefficients
for the soil water retention curve.
2.2 Mathematical model for pollutant transport
As was explained above, the pollutants
moving with the water experience several physical,
chemical and biological processes. Those processes
make the pollutants degraded on the course of
transportation. The mathematical model has taken
those processes into account. The related
assumptions are
[6, 10]
:
(1) The dispersive flux can be expressed by the
Fick law.
(2) The biochemical degradation process
follows the Monod kinetics.
(3) The transfer of pollutants between the solid
and water by diffusion, adsorption, desorption,
ion-exchange, etc., is expressed macroscopically as
form isotherm.
(4) The densities of the water and soil particles
are constant.
(5) The rate of removal of pollutant by the
plants is assumed to be expressed by the first
reaction law.
Under those assumptions, the mathematical
model may have the form
div(
w
w s s w
C F
C
t t
w
u
µ µ u µ
c c
+ = ÷ ÷
c c
q
grad )
w w h s s s
C F µ u µ u ÷ ÷ D ì
w w m r
C S C µ u ì ÷ (8)
For the reversible system,
(
s s
F k
C mF
t µ u
)
c
= ÷
c
(9)
and for the irreversible system,
s s
F k
C
t µ u
c
=
c
(10)
where
w
µ is the density of the water,
3
ML
÷
(
¸ ¸
,
s
µ the density of the soil,
3
ML
÷
(
¸ ¸
,
w
u the
volumetric fraction of the water phase,
s
u the
646
volumetric fraction of the solid phase, C the
concentration of the pollutants,
3
ML
÷
(
¸ ¸
, F the
pollutant concentration on the solid (the mass of
pollutants per unit mass of solid)
s
s
C
F
µ
= ,
s
C
the mass of pollutants per unit volume of solid, m
and k the constants of an isotherm, ,
the flow velocity of the water,
3 1 1
L M T
÷ ÷
(
¸ ¸
w
q
1
LT
÷
(
¸ ¸
,
h
*
d
D = D+ D the coefficient of hydrodynamic
dispersion. D mechanical dispersion coefficient,
d
*
D the molecular diffusion coefficient, ,
2 1
L T
÷
(
¸ ¸
s
ì the degradation coefficients for the soil,
1
T
÷
(
¸ ¸
,
and
m
ì chemical, biological or plant root uptake
process, respectively, .
1
T
÷
(
¸ ¸
Substituting Eq. (10) into Eq. (8), expanding
the divergence in the right hand side of Eq. (8), and
using the continuity equation
w
r
u v
S
t x z
u c c c
= ÷ ÷ ÷
c c c
turn the Eq. (8) into
hxx
C C C C
u v D
t x z x x
u u
c c c c c | |
= ÷ ÷ + +
|
c c c c c
\ .
hxz hzx
C C
D D
x z z x
u u
c c c c | | |
+
|
c c c c
\ . \
|
+
|
.
hzz m
C k
D C
z z
u uì
µ
c c | |
÷ ÷
|
c c
\ .
C ÷
s
s r
s
C S C
ì
µµ
÷ (11)
where
w
µ is replaced by µ
w
u is replaced by u , u,
v are the velocity components in the x- and
z-direction,
1
LT
÷

¸
(
¸
. The dispersion
coefficients , , are as
follows:
hxx
D
hzz
D
hxz hzx
D D =
2 2
*
( )
L T
hxx d
D u D v
D D
U
u u t
+
= + (12)
2 2
*
( )
L T
hzz d
D v D u
D D
U
u u t
+
= + (13)
( )
L T
hxz
D D uv
D
U
u
÷
= (14)
where is the longitudinal dispersivity,
L
D | | L ,
is the transverse dispersivity, | , and
T
D | L t is
the tortuosity factor, which is estimated by the
relationship
[15]
7/ 3
2
d
u
t
u
= (15)
3. FINITE ELEMENT DISCRETIZATION
The Galerkin residual method is used to
discretize the governing equations for both water
movement and pollutant transport
[16, 17]
.
3.1 Finite element discretization for water
movement
The pressure head is
approximated by the function
( , , ) h x z t
( , , ) h x z T in terms
of the base function ( , ) x z | ,
1
( , , ) ( , , ) ( ) ( , )
N
j j
j
h x z t h x z t h t x z |
=
~ =
_
(16)
where ( , )
j
x z | is the piecewise linear base
function, which satisfies the essential boundary
conditions,
j
h is the nodal value of the
to be solved, N is the total number of nodes of the
computational domain. On the basis of Eq.(16) and
the principle of the Galerkin residual method, the
Galerkin integral can be formulated as
( , , ) h x z t
{ [ ( )]
A A
xx xz
h h
K K K
t x x z z
O
u c c c c
÷ +
c
÷
c c c c
)

c
[ ( )] } d 0
A A
zz zx r j
h h
K K K S
z x
| O
c c
+ ÷
c c
= (17)
647
After replacing by h h , applying principle
of partial integration, incorporating Neumann type
boundary condition and performing the integration
over an element, Eq. (17) can be reduced to
(
j j A A
j xx xz
h h
KK KK
t x x z x
| |
u
|
c c
c c c
+ +
c c c c
)

+
c
)d
j j A A
zx zz r j
h h
KK KK S
x z z z
| |
| O =
c c
c c
+ ÷
c c c c
(
e
A A A
xx xz zx
S
h h h
KK KK KK
x z z
c c c
+ +
c c c
)
+
)d
A
z
h
KK S
z
c
c
(18)
where is the area of an element, is the
boundary of the element. Substituting Eq.(16) into
Eq. (18) and carrying out the integration lead to a
set of ordinary differential equations in matrix form
e
O
e
S
| | | | | |
c
+ + =
c
A K M
t

h h P
u
(19)
where , h

u are the vector of pressure head at the
nodal points. The matrices | | A , | | K , | | M and
the vector P are
| | { }
d
)
e
j

A = (20)
{ } { }
[ ] {
e
j j A A
xx j xz j

K = KK h +KK h g
x x
c c
c c
)
{ }
| | | c c c c
+
c c c c
j j j j A
zx j
KK h
z x x z
|
+
{ }
}d
| |
O
c c
c c
j j A
zz j
KK h
z z
(21)
{ } { }
[ ] {
e
j A A
xx j xz j
S
M KK h KK h
j
x z
| | c c
= +
c
)
+
c
{ } { }
}d
j j A A
zx j zz j
KK h KK h S
x z
| | c c
+
c c
(22)
{ }d
)
e
r j

P= S (23)
3.2 Finite element discretization for pollutant
transport
The concentration of pollutantis is
approximated by the function
( , , ) C x z t
( , , ) C x z t in terms
of the base function ( , ) x z | ,
1
( , , ) ( , , ) ( ) ( , )
N
j j
j
C x z t C x z t C t x z |
=
~ =
_
(24)
where ( , )
j
x z | is the piecewise linear base
function, which satisfies the essential boundary
conditions,
j
C the nodal value of the
to be solved, and N the total number of nodes of
the computational domain. On the basis of Eq.(24)
and the principle of the Galerkin residual method,
the Galerkin integral can be formulated as
( , , ) C x z t
[
e
hxx
C C C C
u v D
t x z x x
O
u u
c c c c c | |
+ + +
|
c c c c c
\ .
)
+
hxz hxx
C C
D D
x z z x
u u
c c c c | | | |
+ +
|
c c c c
\ . \
|
.
hzz m
C k
D C
z z
u uì
µ
c c | |
C ÷ ÷ ÷
|
c c
\ .
] d 0
s
s r j
s
C S C
ì
| O
µµ
÷ = (25)
In a similar way, Eq. (25) can be transformed
into the time-dependant ordinary differential
equation in the matrix form
| | | | | |
d
d
A K M
t
' ' ' ' + + =

C
C C P (26)
where is the vector of pressure head at the
nodal point. The matrices| |

C
A' , | | K' , | | M' and
the vector ' P are
| | { }
d
e
i j
A
O
u | O ' =
)
(27)
648
Fig.1 Cross-section of suburb type (upper) and cross-section of urban type (down)
{ } { }
[ ] {
e
j j A A
hxxi j hxzi j
K D h D h
x x
O
| | c c
' = +
c c
)


{ } { }
j j j j A A
hzxi j hzzi j
D h D h
z x x z
| | | | c c c c
+ +
c c c c

}d
j j
z z
| |
O
c c
c c
(28)
{ } { }
[ ] {
e
j j A A
hxxi j hxzi j
S
M D h D h
x z
| | c c
' = +
c c
)
+
{ } { }
}
j j A A
hzxi j hzzi j
D h D h
x z
| | c c
+
c c

{ }
d
i j
n | S (29)
{ } { }
{
e
i mi j j
k
C C
O
u ì
µ
' = + +
)
P
{ }
}
ì
µµ
+
s
sj r
d
C S
{ }
d
j
| O (30)
where
A
hxxi
D ,
A
hxzi
D ,
A
hzxi
D ,
A
hzzi
D ,
i
u and
mi
ì
are the values in an element i.
4. APPLICATION
The presented model was applied to two types
of model riverbanks. One was the most possibly
used in the urban area, and the other was the most
possibly used in the suburban area. Those
experiments were especially designed for testing
the removal performance of the pollutants of the
riverbank. The cross-sections of the two types of
the riverbanks were shown in Fig.1. The
information was obtained from the experiments.
649
4.1 Computation of water movement and pollutant
transport
4.1.1 Computation for suburban model riverbank
Basic properties of material:
The model riverbank was constructed with
natural fine sand. The body of riverbank was
assumed as an isotropic material. The basic
properties of the sand are given in the Table 1. The
data in the table were from the experiments carried
out in this study and from literature
[18-21]
.
Computational mesh and condition:
The cross-section was discretized with the
triangular element (see Fig.2). The computational
parameters used in calculating the water movement
and pollutant transport are summarized in Table 1.
in the table, D
50
is the median size of the sand.
Fig. 2 Computational mesh of suburb type of riverbank
The boundary conditions were given as
0 q = at upstream vertical face and at bottom,
h = 0.26 m, constant water level at downstream
vertical face and part of the downstream slope, q =
0.240 m
3
(dm)
-1
with intermittent pattern, i.e.,
loading for one day and stopping for 6 d, at top flat
face, C
Tpin
= 0.00167 kg/m
3
, at top flat face, C
NH3-Nin
= 0.0089 kg/m
3
, at top flat face.
At downstream slope, the seepage boundary
condition was applied. The Neumann condition was
applied at the downstream.
The initial conditions were given as
0.260 ( ) h z = ÷ n for the pressure head at each
node n, where was the elevation of the node n,
and the datum of the z was at the bottom of the
model riverbank and upwards, 0 at each node,
, , and
the background concentrations of the
pollutants.
( ) z n
= q
TP TPO
= C C
3 3
NH -N (NH -N)O
= C C
TPO
C
3
(NH -N)O
C
The hydraulic loading had the intermittent
pattern to simulate the rainfall duration and the
occurrence of frequency.
Computational results:
The results for time-varying graphs of the
concentration of water quality index NH
3
-N at two
sampling points are shown in Fig.3. The results for
time-varying graphs of the concentration of water
quality index TP at two sampling points are shown
in Fig.4.
Fig.3 The time-varying graphs of the concentration of NH
3
-N
at two sampling Points 1,3 of Suburb type
4.1.2 Computation for urban model riverbank
Basic properties of material:
The model urban riverbank was constructed
with two kinds of materials. One material is natural
fine sand and the other is the macadam of D
50
=30
mm. The basic properties of the sand are given in
Table 1 and the basic properties macadam are given
in Table 2.
Computational mesh and condition:
The cross-section was discretized with the
triangular element (see Fig.5). The domain was
divided into two sub-domains which represent the
two kinds of materials used in constructing the
model urban riverbank. The computational
parameters used in the water movement and
pollutant transport are summarized in Table 2.
The boundary conditions were given as
0 q = at upstream vertical face and at bottom,
h = 0.385 m, constant water level at downstream
vertical face and part of the downstream slope, q =
0.240 m
3
(dm)
-1
with intermittent pattern, i.e. loading
for one day and stopping for 6 days at top flat face,
= 0.00167 kg/m
TPin
C
3
, at top flat face,
3
NH -Nin
C =
0.0089 kg/m
3
, at top flat face.
650
Table 1 The properties of sand and the computational parameters
D
50
(mm)
A
xx
K (md
-1
)
A
zz
K (md
-1
)
s
µ (kgm
-3
)
r
u
sa
u
0.15 11.8 11.8 1718.0 0.056 0.350
o (m
-1
) n
s hxx
D (m
2
d
-1
)
hzz
D (m
2
d
-1
) k(m
-3
kg
-1
d
-1
)(NH
3
-N) k(m
-3
kg
-1
d
-1
) (TP)
2.25 2.50 0.01 0.001 14.05 2.846
m
ì (d
-1
)
s
ì (d
-1
)
0.34 0.14

Table 2 The properties of macadam and the computational parameters
D
50
(mm)
A
xx
K (md
-1
)
A
zz
K ( md
-1
)
s
µ (kgm
-3
)
r
u
sa
u
30.0 150.0 150.0 1650.0 0.056 0.280
o (1/m)
s
n
hxx
D (m
2
d
-1
)
hzz
D (m
2
d
-1
) k(m
3
kg
-1
d
-1
)(NH
3
-N) k(m
3
kg
-1
d
-1
)(TP)
2.0 15.0 0.01 0.001 14.05 2.846
m
ì (d
-1
)
s
ì ( d
-1
)
0.34 0.14

Fig.4 The time-varying graphs of the concentration of TP at
two sampling Points 1,3 of suburb type
Fig.5 Computational mesh of urban type of riverbank
At downstream slope the seepage boundary
condition was applied. The Neumann condition was
applied at the downstream.
The initial conditions were given as
0.385 ( ) h z n = ÷ for the pressure head at each
node n, where was the elevation of the node n,
and the datum of the z was at the bottom of the
model riverbank and upwards.
( ) z n
The other initial conditions were the same as
used in the simulation of suburban riverbank.
651
Computational results:
The results for time-varying graphs of the
concentration of water quality index NH
3
-N at two
sampling points are shown in Fig.6. The results for
time-varying graphs of the concentration of water
quality index TP at two sampling points are shown
in Fig.7.
Fig. 6 The time evolution of the concentration of NH
3
-N at two
sampling Points 1,3 of urban type
Fig.7 The time evolution of the concentration of TP at two
sampling Points 1,3 of urban type
4.2 Discussion
Some features can be found from the results.
(1) The computational results show properly
the time varying pattern of the pollutant
concentrations at the sampling points, and well
agree with the main feature of the experimental data.
The time varying pattern is caused by the
intermittent pattern of the hydraulic loading. The
discrepancy between the computed results and the
experimental data in the amplitude of the oscillation
is mainly due to following reasons. There was time
delay between sampling and analyzing the water
samples. The whole processing time took normally
14 h. The natural degradation would occur during
the process. There was error in analyzing the
samples, especially when the concentration of the
pollutant is relatively small. The parameters used in
the simulation were not perfect.
(2)The computed and tested overall removal
rates, which is defined as the ratio of time-mean
value of the pollutant concentration at the effluent
and the time-mean pollutant concentration at the
influent, for three runs of experiments are shown in
Table 3. It can be seen from the values in the table
that the largest difference between the computed
value and tested value is 5%. This error is
acceptable in practice.
5. CONCLUSION
The mathematical model established in this
study attempts to simulate the performance of the
riverbank in removal of the pollutants from the
rainfall runoff that is main source for the water
pollution of the canals in the urban and suburban
area. The results from the computation have been
compared with the data from the specially
performed model tests. The computed overall
removal rates for the NH
3
-N and TP in 6 cases are
in the range of 88%-98% and 87%-97%,
respectively. The differences between the computed
and tested overall removal rates for the NH
3
-N and
TP are less than 5%. The time-varying oscillation
pattern of the concentrations of the NH
3
-N and TP
has been rationally simulated. The predication
capacity of the model is well assessed, but further
study on model validation with more experimental
data and field measurements are needed.
ACKNOWLEDGEMENT
This study was partially supported by the
Nanjing Research Institute of Water Resources in
Nanjing of China and the UNESCO-IHE in Delft of
the Netherlands.
652
Table 3 Comparison of computed and tested overall removal rate (%)
Suburban riverbank model Urban riverbank model
TP NH
3
-N TP NH
3
-N Runs
Computed Tested Computed Tested Computed Tested Computed Tested
1 97.1 94.9 98.4 95.9 92.3 96.1 95.3 97.0
2 97.3 96.4 98.2 98.8 96.3 95.4 95.2 93.8
3 93.0 93.5 94.3 94.4 88.2 91.1 87.9 92.9
REFERENCES
[1] MITSCH W. J. Wetland modelling[M]. Amsterdam:
Elsevier Science Publishers, 1988, 115-132.
[2] ASAEDA T., TRUNG V. K. and MANATUNGE J. et al.
Modelling of macrophyte–nutrient–phytoplankton inter-
actions in shallow eutrophic lakes and the evaluation of
environmental impacts [J]. Ecological Engineering, 2001,
16(3): 341-357.
[3] Van der PEIJL M. J., VERHOEVEN J. T. A. A model of
carbon, nitrogen, and phosphorus dynamics and their
interactions in river marginal wetlands [J]. Ecological
Modelling, 1999, 118(2): 95-130.
[4] SPIELES D. J., MITSCH W. J. A model of
macroinvertebrate trophic structure and oxygen demand in
freshwater wetlands [J]. Ecological Modelling, 2003,
161(3): 181-192.
[5] WANG N., MITSCH W. J. A detailed ecosystem model
of phosphorus dynamics in created riparian wetlands [J].
Ecological Modelling, 2000, 126(2): 101-130.
[6] STURMAN P. J., STEWART P. S. and CUNNINGHAM
A. B. et al. Engineering scale up of in situ bioremediation
processes: a review[J]. Journal of Contaminant
Hydrology, 1995, 19(3): 171-203.
[7] ADIL Elkrail, SHU Long-cang and HAO Zhen-chun.
Numerical simulation of groundwater dynamics for
Songhuajiang River valley in China [J]. Journal of
Hydrodynamics, Ser. B, 2004, 16(3): 332-335.
[8] CHEN Jing, ZHOU Zhi-fang and JIA Suo-bao. Analytical
assessment of the influence of soil-infiltrated solution of
groundwater fluctuations in estuarine aquifer wastewater
on groundwater[J]. Journal of Hydrodynamics, Ser. B,
2005, 17(5): 641-646.
[9] BRUM A., ENGESGAARD P. Modelling of transport
and biogeochemical processes in pollution plumes:
literature review and model development[J]. Journal of
Hydrology, 2002, 256(3): 211-227.
[10] BEAR J. Hydraulics of goundwater[M]. New York:
McGraw-Hill, 1979.
[11] FEDDES R. A., KOWALIK P. J. and ZARADNY H.
Simulation monographs[M]. The Netherlands: Pudoc.
Wageningen, 1978.
[12] Th. Van GENUCHTEN M. A closed form equation for
predicting the hydraulic conductivity of unsaturated
soils[J]. Soil Science Society, 1980, 44: 892-898.
[13] LUCKNER L. M., Th. Van GENUCHTEN M. and
NIELSEN D. R. A consistent set of parametric models
for the two-phase flow of immiscible fluids in the
subsurface[J]. Water Resources Research, 1989,
25(10): 2187-2193.
[14] VOGEL T., GISLENOVA M. C. On the reliability of
unsaturated hydraulic conductivity calculated from the
moisture retention curve[J]. Transport in porous
media, 1988, 3(1): 1-15.
[15] MILLINGTON R. J., Quirk J. M. Permeability of porous
solids[J]. Transactions Faraday Society, 1961, 57:
1200-1207.
[16] BAKER A. J. Finite element computational fluid
mechanics[M]. New York: Hemisphere Publishing,
1983.
[17] YIN Hai-long, XU Zu-xin and LI Huai-zheng et al.
Numarical modeling of wastewater transport and
degradation in soil aquifer [J]. Journal of
Hydrodynamics, Ser. B, 2006, 18(5): 597-605.
[18] TOOD D. K. Groundwater hydrology[M]. New York:
John Wiley and Sons, 1980.
[19] KASHEF A. I. Groundwater engineering[M]. New
York: McGraw-Hill, 1986.
[20] YAN Fei, ZHAN Mei-li and SU Bao-yu. Analysis on
unsaturated seepage calculation parameter[J]. Journal
of Yangtze River Scientific Research Institute, 2004,
21(5): 28-31 (in Chinese).
[21] HEYMANS J. J., ULANOWICZ R. E. and
BONDAVALLI C. Network analysis of the south
Florida Everglades graminoid marshes and comparison
with nearby cypress ecosystems[J]. Ecological
Modelling, 2002, 149(1): 5-23.