Journal of Water Resource and Protection, 2011, 3, 192-198

doi:10.4236/jwarp.2011.33025 Published Online March 2011 (http://www.scirp.org/journal/jwarp)

Finite Difference Method of Modelling Groundwater Flow

Magnus. U. Igboekwe*, N. J. Achi
Department of Physics, Michael Okpara University of Agriculture,
Umuahia, Nigeria
E-mail: [email protected]
Received Janaury 1, 2011; revised February 8, 2011; accepted March 2, 2011

Abstract

In this study, finite difference method is used to solve the equations that govern groundwater flow to obtain
flow rates, flow direction and hydraulic heads through an aquifer. The aim therefore is to discuss the princi-
ples of Finite Difference Method and its applications in groundwater modelling. To achieve this, a rectangu-
lar grid is overlain an aquifer in order to obtain an exact solution. Initial and boundary conditions are then
determined. By discretizing the system into grids and cells that are small compared to the entire aquifer, ex-
act solutions are obtained. A flow chart of the computational algorithm for particle tracking is also developed.
Results show that under a steady-state flow with no recharge, pathlines coincide with streamlines. It is also
found that the accuracy of the numerical solution by Finite Difference Method is largely dependent on initial
particle distribution and number of particles assigned to a cell. It is therefore concluded that Finite Difference
Method can be used to predict the future direction of flow and particle location within a simulation domain.

Keywords: Finite Difference Method, Groundwater Modelling, Particle Tracking, Algorithm, Discretization,
Flow Rates, Hydraulic Heads

1. Introduction speed computers (digital computers). This method is an

approach to computational fluid dynamics (CFD) and
Groundwater models are mathematical models derived very effective in groundwater flow modelling. Ground-
from Darcys law which is used to calculate the rate and water is an important resource in so many areas for its
movement of groundwater through the aquifer and con- use as a source of drinking water and irrigation water. In
fining units in the sub-surface [1]. Ground-water models many areas, groundwater is threatened by leaching of
which can also be used to evaluate impact assessment pesticides and other agricultural chemicals and the
required for water in a regulated aquifer system has been leaching of industrial chemicals from hazardous-waste
of importance to agriculturists, environmentalists, hydro- sites. Because of the importance of this resource, and
logists etc. It is necessary to study the groundwater re- because the degradation of groundwater cannot be easily
source potentials of a site because the simulation of reversed, the assessment of threat to groundwater quality
groundwater flow requires a thorough knowledge and from human activities is often required. So groundwater
understanding of hydrogeologic characteristics of the models are increasingly used as part of this assessment.
site. The major aim of this research work is to discuss the
Groundwater models are used as tools for decision principles of Finite Difference Method and its applica-
making in the management of a water resource system. tion in groundwater modeling
They may also be used to predict some future ground- The word model is simply a representation of a real
water flow. Some of the established solution techniques system or process. But model has variety of definitions
available for solving the governing equations of the that it is often difficult to define [2].
model are Finite Difference and Finite Element ap- A model is a hypothesis for how a system or process
proximation or a combination of both provided that operates. These models in use today are deterministic
model parameters and initial and boundary conditions are mathematical models that are based on conservation of
properly specified. The numerical solution applied in this mass, momentum and energy. Modelling is the process
research work is the finite difference method. This is an by which a physical system is simplified to obtain a ma-
old method made more useful with the advent of high thematically tractable situation. The resulting simplified

Copyright 2011 SciRes. JWARP

 M. U. IGBOEKWE ET AL. 193

version of the real system is called model of the system tion is the change in flow condition from one steady-state
or a mathematical model [3]. to another. The steady-state flow is described by a form
Numerical modelling of groundwater is a relatively of Laplace equation. It is a flow in which all conditions
new field. It was not extensively pursued until the (velocity, pressure, etc) remain constant with respect to
mid-1960s when digital computers with adequate capac- time. The groundwater flow equation is often derived for
ity became generally available. Since then, significant a small representative elemental volume where the prop-
progress has been made in the development and applica- erties of the medium are assumed to be effectively con-
tion of such techniques to groundwater flow. stant. A mass balance is obtained on the water flowing in
Numerical models are used in groundwater modelling and out of this small volume along with Darcys law to
because it yields approximate solutions to the governing arrive at the transient groundwater flow equation.
equations through the discretization of space and time. It The flow equation is based on the continuity equation
helps in assessing the impact of pollution on an aquifer. [10]:
Groundwater models generally require the solution of Inflow Outflow = change of storage. (1)
partial differential equations. The equations describing For a small portion of an aquifer it can be restated as:
the groundwater flows are second order partial differen- Subsurface sum + net flow= change in storage (2)
tial equations which can be classified on the basis of Combining Darcys law with this continuity equation
their mathematical properties. There are basically three yields the general form of the equation describing the
types of second order partial differential equations: pa- transient flow
rabolic, hyperbolic and elliptic equations [4,5]. h
Ss q G (3)
The two main types of numerical models that are ac- t
cepted for solving the groundwater equations are the where Ss = specific storage
Finite Difference Method and the Finite Element Method q = flux
presented by [6,7]. Both of these numerical approaches h = Hydraulic head
require that the aquifer be sub-divided into a grid and t = time
analyzing the flows associated within a single zone of the The statement indicates that the change in hydraulic
aquifer or nodal grid. head with time equals the negative divergence of the flux
The equation describing the groundwater flow is a (q) and the source (G).
Partial Differential Equation. It can be solved mathe- Here the head and flux are unknown but Darcys law
matically by analytic or numerical solutions. But analytic relates the flux to the head by substituting it in the flux,
solutions are very difficult to apply because it requires that is
that parameter and boundaries should be highly idealized. h
Ss K h G (4)
The advantages of analytical solution, if it is possible to t
apply, are that it yields an exact solution to the equation
where G = external flow
and it is simple and efficient to obtain. Many of them
K = hydraulic conductivity
have been developed for the flow equations, but most of
Simplifying this we have
them are limited to well hydraulics problems [8]. The
h
continuous variable is replaced by discrete variables that Ss K 2 h G (5)
are defined at grid blocks. Also the continuous differen- t
tial equations which define the hydraulic head in the if K is replaced by its equivalent T, Equation (5) can be
system, is replaced by a finite number of head at differ- written as
ent grids [9]. h h h

t x x y y
Ss Txx Tyy
A common method for solution of this equation in civ-
il engineering and soil mechanics is to use the graphical (6)
techniques of drawing flow nets, where contours of hy- h
Tzz G
draulic head and the stream function make a curvilinear z z
grid, allowing complex geometries to be solved ap- where: T = transmissivity
proximately. A steady-state may be reached if the aquifer has re-
The groundwater flow equation is the mathematical charging boundary conditions (or it may be used as an
relationship which is used to describe the flow of approximation in many cases). The equation describing
groundwater through an aquifer [6]. In the study of this flow is a form of the Laplacian equation given as
groundwater flow equation, one may discuss about tran-
2 h 0 (7)
sient flows and steady state flows. The transient flow
which is described by a form of diffusion equation simi- This equation states that hydraulic head is a harmonic
lar to that used in heat transfer to describe heat conduc- function and has many analogs in other fields. The equa-

Copyright 2011 SciRes. JWARP

194 M. U. IGBOEKWE ET AL.

tion above can be rewritten as zone of the aquifer [12]. It utilizes a time distance grid of
nodes and a truncated Taylor series approach to deter-
h h h
2 2 2
0 (8) mine the condition of flow at any particular node. A brief
x2 y 2 z 2 coverage of the application of Taylor Series and nodal
The above groundwater flow equations are valid for grid will illustrate several points fundamental to flow
three dimensional flow. simulation.
We also have two dimensional groundwater flow and The flow at time t, the profile of variable y with x may
the general governing equation is given by be described by a truncated Taylor Series given as
h h h Y x0 x y x0 y1 x0 x
K x b K y b G Ss (9)
x x y y t x 2 x 3 (15)
y11 x0 y111 x0
where k = hydraulic conductivity 2! 3!
b = saturated thickness The value at x of node i is x0 and the + notation in eq-
But, Kb = T, therefore: uation 15 allows it to be used in forward or backward
h h h difference approach.
Tx Ty G S s (10) The forward difference first order is
x x y y t
Y 1 x0 y x0 x y x0 (16)
In finite difference form, eqn.10 can be expressed as:
W AB S B J +I The backward difference first order is
LiB TiB hBJ +I hBJ +I AB GBJ +I
t
hB hBJ (11)
Y 1 x0 y x0 x y x0 (17)
iB

where Wi and Li are boundary width and flow path The summation of forward or backward difference
length. also yields a central difference expression for the first
A = Area of a single zone. order derivative of variable y at x0.
Groundwater velocity is based on hydraulic conduc- The central difference first order
tivity (k), as well as the hydraulic gradient (I). Therefore x

the equation determined by Darcy to describe the basic Y 1 x0 y x0 i, j y x0 x (18)
2 x
relationship between sub-surface material and the
movement of water through them is In computer notation, these equations (16-18) can be
Q = KIA (12) re-written as
where Q = volumetric flow rate (Discharge) u u
u1 xi i+1 i
K = Hydraulic conductivity h
A = Area that the groundwater is flowing through u 1 ui
This relationship is known as Darcys law [11]. u1 xi i
h
Rearranging Equation (12) we obtain the flux (V)
u u 1
which is known as the apparent velocity. u1 xi i 1
That is: 2h
Q
KI V (13)
A
where V = Apparent velocity, m / sec
The actual groundwater velocity which is called the
Darcys flux (Vx) is given by
Q KI
Vx (14)
A n
where Vx = Actual velocity m / sec
n = Porosity
This is the actual velocity of groundwater and does
account for tortuosity of flow paths by including porosity
in its calculation.
The Finite Difference Method is a computational pro-
cedure based on dividing an aquifer into a grid (see Fig-
ure 1) and analysing the flows associated within a single Figure 1. Computer notation for finite difference grid.

Copyright 2011 SciRes. JWARP

 M. U. IGBOEKWE ET AL. 195

If there is recharge to the aquifer, a more useful result

is the summation of the second order forward or back-
ward difference forms of the truncated Taylor series to
determine the second differential of variable y at x0 i.e.
ui +1 2ui
u11 xi ui -1 (19)
h2
A special notation is used to describe the positions of the
node in the finite difference grids.
In order to solve the groundwater flow equation we
must be able to specify the initial and boundary condi-
tions. There are two basic types of boundary conditions. Figure 2. Boundary conditions for a cross section of a re-
If the head is known at the boundary of the flow region, gional aquifer.
this is known as a Drichlet condition. Then if the flux
across a boundary to the flow region is known, this is grid. Several different hydraulic heads and flux values
known as Neumann condition. For a steady-state are specified along the boundaries. A river enters the
problem, boundary conditions are required; whereas for a aquifer at the northern boundary and exits at the south.
transient problem, boundary and initial conditions must The production wells near the river withdraw water from
be specified. But in some cases the boundary conditions the aquifer. Groundwater flow occurs mainly from the
will be mixed with some portions having known head north to the south of the domain. The aquifer is hetero-
and some portions having known flux [13].Initial condi-
tions and boundary conditions can be related to levels,
pressure, and hydraulic head on the one hand (head con-
ditions), or to groundwater inflow and outflow on the
other hand (flow conditions).As an example of boundary
conditions, let us consider Figure 2 below. This is a sand
and gravel aquifer overlying an impermeable basement
rock. There are several flow regions present, each lead-
ing away from a groundwater divide towards a stream.
The plane of W to X is a groundwater divide. There is no
flow across the divide and the boundary condition along
the divide is
dh
0 (20)
dx
The plane of Z to Y is the centre of the stream and it is
also a groundwater divide, across which no flow takes
place. The boundary condition here is also:
dh
0 (21)
dy
No flow occurs from the sand deposit into the imperme-
able bedrock, so along the X to Y plane there is no flow
and the boundary condition is
dh
0 (22)
dz
These three boundaries are Neumann boundaries where
the flow conditions are specified.

2. Methods

A hypothetical site such as the one shown below (Figure Figure 3. Discretized domain of a hypothetical site showing
3) is chosen. The domain is discretized into an irregular wells and river source.

Copyright 2011 SciRes. JWARP

196 M. U. IGBOEKWE ET AL.

geneous and anisotropic. nodal components of the Darcys flux qx and qy are ob-
A finite difference grid is laid over an aquifer. It is a tained by differentiation of the calculated hydraulic
block centred finite difference grid where the node points heads.
fall in the centre of the grid. The grid parameters are:
Number of the grid columns 43 qxi 1 2, j K xxi 1 2, j
hi 1, j hij
Number of grid rows 65 xi, j 1 2
(26)
Minimum i coordinate 0
q yi 1 2, j K yyi 1 2, j
hi 1, j hij
yi , j 1 2
Minimum j coordinate 0
The basic grid is regular, with the rows and columns
being normal to each other. The rows and columns may where, xi+ ,j, yi,j + = distance between nodes i, j and i +1,
be varied so that there are more node points in certain j and nodes i, j and i, j +1 respectively Kxxi + ,j, Kyyi + , j =
parts of the aquifer than others. But this is desirable in Hydraulic conductivities between i + i, j at x direction
the area around a well field. and nodes i, j and i, j +1 at y direction respectively These
For unconfined aquifers the saturated thickness bij is a values are taken as the weighted harmonic mean between
function of the hydraulic heads. A priori both the hy- the hydraulic conductivities.
draulic heads and the saturated thickness are unknown.
Mathematically, this leads to non-linear behaviour. An K xxi ,+ j +1 2, =
x + xi +1, j K xxi , j K xxi +1, j
i, j

initial guess for the saturated thickness is required in K xxi , j xi , j + K xxi + I , j xi +1, j
(27)
order to estimate the aquifer transmissivities Txxi,j and
K yyi , j +1 2 =
yi , j + yi +1, j K yyi , j K yyi , j +1
Tyyi,j. An iterative scheme must be used to calculate the
K yyi , j yi , j + K yyi , j +1yi , j +1
hydraulic head distribution, update the transmissivities
and check if the discrepancy between the previously es- From the Darcys fluxes the pore velocities are calcu-
timated saturated thickness and the updated one is great- lated as:
er than a specific tolerance i.e.
qxi 1 2, j
New old Vxxi 1 2, j
ni 1 2, j
hi , j h j ,i (28)
max min
> toleranc (23) qyi 1 2, j
h h Vyyi , j 1 2
ni , j 1 2
The directional transmissivities between adjacent
nodes: Txxi + j, Tyyi, j + are calculated by multiplying where
harmonic mean of hydraulic conductivities and geomet- ni + ,j, ni,j + = effective porosities between nodes i, j and
ric mean of saturated thickness [14]. i+1, j; and nodes i, j and i, j + 1 respectively. These
values are taken as the weighted arithmetic mean be-
Txxi 1 2 j K xxi 1 2 j h i 1, j BTi 1, j hi, j BTijj tween the porosities of adjacent blocks.
(24) nij xi, j ni 1, j xi+1, j
Tyy 1 2 j K yyi 1 2 j h i, j 1 BTi, j 1 hi, j BTijj
ni+1 2 , j
x i, j xi+1, j


(29)
where BTi, j = aquifer bottom elevation at node i, j nij yi, j ni+1, j yi+1, j
Kxxi + j, Kyyi + j = Hydraulic conductivities between ni, j+1 2
yi, j yi+1, j
blocks i, j and i+1, j .
Particle tracking provides a clearer description of
K xxi 1
x i, j xi 1, j K xxi i , j K xxi , j groundwater flow within an aquifer. In steady-state flow
2
j
K xxi 1, j xi 1, j K xxi , j xi , j field with no recharge, pathline (particle trace) coincides
(25) with streamlines. The two dimensional equation of path-
K yyi 1
y i, j yi 1, j K yyi 1, j K yyi , j lines is given by:
2
j
K yyi 1, j yi 1, j K yyi , j yi , j P x, y P x0 , y0 vdt (30)
There are two different types of velocity vectors where
(a) Nodal velocities P = Vector containing the x, y coordinates of the path-
(b) Boundary velocities line.
Boundary velocities are calculated directly from Darcys P(xo,yo) = The starting point of the pathline (initial con-
law using the hydraulic head difference and material dition).
properties between two nodes, whereas the nodal veloci- V = Average linear velocity
ties are interpolated values at the grid nodes. The inter- t = Time

Copyright 2011 SciRes. JWARP

 M. U. IGBOEKWE ET AL. 197

Equation 30 is written in discretized form as an explicit velocity component at particle location is obtained from
time stepping scheme iterative schemes.
X t + t = X t Vxt t It is also found that the accuracy of the numerical so-
lution by Finite Difference Method is largely dependent
Yt + t = Yt Vyt t on initial particle distribution and number of particles
where assigned to a cell. This scheme provides improved accu-
t = time increment, Vxt, Vyt, are the upstream compo- racy over interpolation scheme especially for block het-
nents of the average linear velocity of the current particle erogeneous aquifer but with increased computational
locations at time t. efforts.
In this scheme, a set of uniformly distributed particles
is assigned to each calculation cell. With each time step, 4. Conclusions
every particle is moved to a new location based on ve- Numerical modelling has found interesting application in
locity of the particle in the x and y directions. The veloc- groundwater flow and transport since the mid-1960 s,
ity component at particle location is to be obtained from when digital computers with adequate capacity became
iterative scheme. In this scheme the particle representa- generally available. The need for a numerical model
tive area is assigned to each particle by dividing the cell cannot be over emphasized since analytical models as-
into a set of equal number of squared sub cells with each sume a homogenous aquifer.
particle at its centre.
The temporal weighted velocity is to be obtained by
fourth order classical Runge Kutta method [15].
A weighted velocity is based on its values evaluated at
four points in time, and then it is used to move the parti-
cle to a new direction.
Figure 4 shows the particle tracking algorithm. The
aquifer is discretized using equally sized calculation cells.
In the particle tracking scheme, one particle per cell is
used.

3. Results and Discussion

The hydrodynamics of groundwater flow in a hypotheti-
cal aquifer chosen for this study has been found to de-
pend mainly on the surface water flow within the site.
The surface water flow is from north to south. So also is
the groundwater. The aquifer is heterogeneous and ani-
sotropic.
In the limiting case of a model calculation of an infi-
nitely small grid space, the solution approaches the exact
solution. Nodal and boundary velocities are calculated as
the weighted harmonic mean between hydraulic conduc-
tivities and the arithmetic mean of the porosities of adja-
cent blocks. Initial and boundary conditions are used to
specify the flow condition.
A numerical problem had been taken which allows the
testing of numerical modelling technique called Finite
Difference Method for the simulation of groundwater
flow.
The region is bounded by two no flow boundaries as
show in Figure 2. The aquifer is discretized using equa-
lly sized 2 x 2 calculation cells. In the scheme, one parti-
cle per cell is used.
Results from the flow chart of a particle tracking algo-
rithm show that in a steady-state field with no recharge, Figure 4. Flowchart of the computational algorithm for
pathlines coincide with streamlines. This means that the particle tracking.

Copyright 2011 SciRes. JWARP

198 M. U. IGBOEKWE ET AL.

[5] T. Narasimha Reddy and V. V. S. Gurunadha Rao, Wa-

The numerical method applied in this study is the Fi- ter Balance Model and Groundwater Flow Model of Du-
nite Difference Method, which in its simplicity provides lapally Basin, Granitic Terrain, A.P., Research Series
No. 9, 1991.
necessary aids in finding solution to groundwater prob-
lems. [6] M. P. Anderson and W. W. Woessner, Applied Gro-
undwater Modeling, Simulation of Flow and Advective
The finite difference grid overlain over the aquifer
Transport, Academic Press, San Diego, C. A, 1992.
improves the accuracy in the calculation of flow rate and
direction. It is also an advantage in particle tracking [7] M. U. Igboekwe, V. V. S. Gurunadha Rao and E. E. Ok-
wueze, Groundwater Flow Modelling of Kwa Ibo River
within the aquifer domain. The modelling studies have
Watershed Southeastern Nigeria, Hydrological Proc-
been carried out in a hypothetical site with the aim of esses, Vol. 22, No. 10, 2008, pp. 1523-1531.
determining rate and direction of flow of groundwater doi:10.1002/hyp.6530
through an aquifer domain. The domain is discretized
[8] M. K. Hubbert, Darcys Law and Field Equations of
with boundary conditions considered for easy calcula- Flow of Underground Fluids Transaction, American In-
tions. stitute of Mining and Metallurgical engineers, 1956, pp.
In the flow model it was observed that water flows 22-239.
from region of high hydraulic head to region of low hy-
[9] T. A. Prickette and C. G. Longuist, Selected Digital Co-
draulic head and that an exact solution could be obtained mputer Techniques for Groundwater Resource Evalua-
if the grid spacing is small enough say 5 cm. Results tion, Bull.55, Illinois State Water Survey, Urbana, 1981.
show that in a steady-state flow field with no recharge, p. 62.
pathlines coincide with streamlines. It is therefore con-
[10] W. J. Gray and J. L. Hoffman, A Numerical Model Stu-
cluded that Finite Difference Method can be used to pre- dy of Ground-Water Contamination from Price's Landfill,
dict the future direction of flow and particle location New Jersey I. Data Base and Flow Simulation,
within a simulation domain. Groundwater, Vol. 21, No. 1, 1983, pp. 7-14.
doi:10.1111/j.1745-6584.1983.tb00699.x
5. References [11] D. K. Todd, Groundwater Hydrology, 2nd Edition, John
Wiley and Sons, New York, 2001.
[1] R. A. Freeze, Three Dimensional Transient Saturated [12] Peter. C. Trescott and S. P. Larson, Finite Difference
Unsaturated Flow in a Groundwater Basin, Water re- Model for Aquifer Simulation in Two-Dimensions with
sources Research, 1979, pp. 347-366. Results of Numerical Experiments, U.S. Geological
[2] L. F. Konikow and J. D. Bredehoeft, Computer Model Survey Techniques at Water-Resources Investigations,
of Two Dimensional Solute Transport and Dispersion in Book 7, Chapter C1, 1976, p. 116.
Groundwater, U.S. Geological Survey, Techniques of [13] C. W. Fetter, Applied Hydrogeology: Finite Difference
Water Resources Investigation Book 7, Chapter C2, 1992, Model, CBS Publishers and Distributors, New Delhi,
p. 90. India, 2000, p. 592.
[3] V. Vu.Hung and Ramin S. Estandiari, Dynamic Systems: [14] M. U. Igboekwe, Geoelectrical Exploration for Ground-
Modelling and Analysis, McGraw-Hill, New York, 1997 Water Potentials in Abia State, Nigeria, An Unpublished
p. 612. PhD dissertation. Michael Okpara University of Agricul-
[4] J. M. McDonald and A. W. Harbaugh, A Modular Thr- ture, Umudike, 2005, p. 131.
ee-Dimensional Finite-Difference Groundwater Flow M- [15] R. T. Peter, Numerical Analysis: RungerKutta Meth-
ode, Techniques of Water Resources Investigations of od, ISBN 0. 333 -58665- 4, 1994, pp. 214-217.
the U. S. Geological Survey Book.6, 1988, p. 586.

Copyright 2011 SciRes. JWARP