Sei sulla pagina 1di 11

1

2-D Groundwater Flow Through A Confined Aquifer


Gordon Whyburn and Ajoy Vase
Pomona College
May 5
th
, 2006
Abstract
We attempted to model the groundwater flow in a 2-D confined aquifer under different
conditions using the finite difference method. We used Matlab for all our computations.
We solved the Laplace equation numerically to model the steady state flow through an
aquifer with separate regions of different conductivities. We solved the time dependent
flow case to predict the behavior of the aquifer under certain initial flows when the flow
was turned off. We also attempted to solve the time independent flow case for an aquifer
that has a conductivity that varies continuously.
Introduction
An aquifer is a layer of permeable material that can transmit water underground.
Aquifers channel groundwater to wells. The conductivity of an aquifer is a measure of
how fast water can flow through it. A single aquifer is comprised of different materials,
enabling the local water flow rate to vary in different regions of the aquifer. Common
aquifer materials include sand, gravel, sandstone, limestone and dolomite.
Aquifers fall into two main categories unconfined and confined. An unconfined
aquifer has permeable material extending from the land surface down to the bottom of the
aquifer. It can be recharged by water seeping down from the surface or by lateral
groundwater flow. An ideal confined aquifer is one that is bounded above and below by a
completely impermeable layer of material. Thus the only way in which the aquifer can be
recharged is by lateral groundwater flow. The water in a confined aquifer is under
pressure.
Aquifers are a vital source of freshwater to human beings in rural areas and even
entire cities. The Edwards aquifer in Texas provides enough water to sustain 1.5 million
people. Groundwater flow models for confined aquifers can predict the pressure in
different regions of the aquifer. This information can be used to select the most
appropriate spot for digging a well.
Overexploitation of aquifers leads to depletion. Aquifer depletion is a growing problem.
Methods of artificial recharge are being developed in order to address the issue.
Groundwater flow models that predict water flow in an aquifer for different entry flow
conditions will be useful for implementing artificial recharge methods.
2
Equations used:
We wanted to model groundwater flow through the following 2-D aquifer:
Figure 1. Diagram of aquifer.
The PDE equation that describes 2-D groundwater flow in a confined aquifer is:
(1)
Here h is the hydraulic head and T is the constant conductivity of the aquifer medium.
This equation assumes that the aquifer is 100% confined, i.e. there are no leaks from the
top and the bottom. This equation is derived using mass and energy conservation for a
control volume in the aquifer.
The hydraulic head is defined to be the total mechanical energy per unit weight of the
water. In groundwater flow, the hydraulic head equals the sum of the elevation head and
the pressure head. This is describes mathematically by the following equation:
(2)
Equation (1) can be solved for the time independent case, which is tantamount to solving
Laplaces equation. For the time-independent case where the conductivity varies
continuously, the flow is described by the following equation:
0 ) , (
2
2
2
2
=
c
c
c
c
+
c
c
c
c
+
|
|
.
|

\
|
c
c
+
c
c
x
h
x
T
y
h
y
T
y
h
x
h
y x T (3)
x
y
Permeable region
Impermeable region
Impermeable region
Inflow
t
h
T y
h
x
h
c
c
=
c
c
+
c
c 1
2
2
2
2
g
P
z h

+ =
3
Once the head has been computed, the horizontal and vertical components of
groundwater flow can be found using Darcys law:
(4a)
(4b)
Boundary conditions:
Since there is no flow across the top or the bottom of the aquifer, q
y
= 0. From (3b),
therefore,
y
h
c
c
= 0 at all points along the ceiling and the floor of the aquifer. We assume
that the incoming flow to the aquifer and the head at the beginning of the aquifer are
known. The outgoing flow depends on the interaction of the water with the local regions
in the aquifer, so we leave this boundary condition unspecified.
Solution strategy:
We decided to use the finite difference method to look at the kind of flow models that
could be created with this relatively easy method.
We approached this model in the following steps:
1. Solve Laplaces equation for the steady state flow. The solution of this equation is
independent of the local conductivity of different regions in the aquifer. The
conductivity is considered when computing the velocity field of the groundwater
in the aquifer.
2. Solve equation (1) for different initial flows to observe how the flow varies in
time.
3. Solve equation (2) to observe the flow when the conductivity of the material in
the aquifer varies continuously, i.e. a case where the packing of the gravel in the
aquifer varied linearly along its length.
Assumptions made:
1. The aquifer is rectangular in shape
2. The aquifer is ideally confined, which means that the floor and the ceiling are
absolutely impermeable. In normal conditions, there is some leakage of water into
the aquifer from the top and out through the bottom
3. The initial head and velocity of the water are known.
4. In the steady state case, we assume that the flow does not change significantly
with time. This is a good assumption in cases where the slope or position of the
water table does not change.
x
h
T q
x
c
c
=
y
h
T q
y
c
c
=
4
Finite difference grid
In order to compute the head at all points in the aquifer, the first step was to discretize the
domain into a grid of total height N and length M.
Figure 2. Finite difference grid representing the aquifer domain.
The distance between each point is h, such that the location of a point on the grid is
described by
N n nh y
M m mh x
n
m
.. 0 ;
.. 0 ;
= =
= =
The next step was to discretize the PDE and organize the resulting set of equations into a
linear system. The discretization proceeded according to the location of the points in the
aquifer. Table 1 summarizes the discretization formulas used for the solution to the
Laplace equation and the final appearance of the linear system in each case. We used
second order difference formulas to discretize this PDE.
h =h
0
dh/dx = q
0
dh/dy = 0
dh/dy = 0
5
Table 1. Discretization formulas used in the finite difference grid set up to solve
Laplaces equation.
Note: ) , (
n m
n
m
y x h f = We relabeled the head f to avoid confusion with the distance
between each point, h. The superscript is the index for the y component of the flow, and
the subscript is the index for the x-component.
We made a vector of the unknowns
n
m
f and a matrix of the coefficients. It was not
possible to specify both the initial head and the initial velocity in our matrix. We worked
around this problem by specifying the head for points in the first three columns of the
grid. This automatically sets an initial slope for the head.
A similar general procedure was followed in order to discretize the time dependent PDE
in equation (1). An explicit Euler first order forward difference method was used to
project the PDE forward every time step. This formula is of the form:
n
n
m
n
m
tf v v A + =
+1
The solutions were not expected to be oscillatory in nature, which made the Euler method
suitable since there was no danger of overshooting the correct answer.
x-
component
y-component Stencil
in
Figure
2
Linear system for Laplace equation
All points
excluding
boundary
conditions
Second
order
central
difference
Second order
central
difference
Blue
0 ) (
4
2
2
1 1
1 1
= +
+ + +
+
+
h O
h
f f f f f
n
m
n
m
n
m
n
m
n
m
Aquifer
floor and
ceiling
Second
order
central
difference
Second order
forward
difference
(discretization
of the
boundary
condition)
Green
0 ) (
2
4 3 2
2
2 1
2
1 1
= +
+
+
+

+
h O
h
f f f
h
f f f
n
m
n
m
n
m
n
m
n
m
n
m
Right
boundary
of the
aquifer
Second
order
backwards
difference
Second order
central
difference
Red
0 ) (
2 4 5 2
2
2
1 1
2
3 2 1
= +
+
+
+
+

h O
h
f f f
h
f f f f
n
m
n
m
n
m
n
m
n
m
n
m
n
m
Top and
bottom
right
corner of
the
aquifer
Second
order
backwards
difference
Second order
backwards
difference
Yellow
0 ) (
4 5 2
4 5 2
2
2
3 2 1
2
3 2 1
= +
+
+
+


h O
h
f f f f
h
f f f f
n
m
n
m
n
m
n
m
n
m
n
m
n
m
n
m
6
For the PDE in equation (3), we used second order difference formulas to represent the
first derivatives in the equation. We also picked functions of T with simple derivatives
that we calculated and specified in our matrix.
Analysis:
Laplaces Equation:
The Matlab code worked qualitatively for this steady state case with local regions of
varying conductivity. We plotted h as a function of x and y in order to observe how it
varied in different regions of the aquifer. We solved this steady state equation for
different cases.
Figure 3. Aquifer with Constant Input Head.
Figure 4. Plot of flow as a function of position
Case 1 (simplest case):
Constant head at the
entrance of the aquifer
with pockets of different
conductivity.
Our results indicate that the
head decreases linearly in
the horizontal direction.
This proves that the code
works because decreasing
head is a requirement for the
water to flow from left to
right (see equation 4a).
Furthermore, the slope of h
with respect to y is zero at
the floor and ceiling of the
aquifer. The plot of the flow
as a function of the position
(see Figure 4) also verifies
that our code is qualitatively
working. As we would
expect, in regions of a
higher conductivity, the
flow speeds up (indicated by
larger arrows in the red
portion of the grid).
7
Our results in the test case allow us to test the code for more complex situations.
Figure 5. Aquifer with larger head in the middle portion
of the entrance.
Figure 6. Plot of flow for the above case.
0 5 10 15 20 25
0
2
4
6
8
10
12
14
16
18
20
x
y
Case 2: Water is artificially
injected into the middle
region of the aquifer (perhaps
through a large pipe): These
conditions mean that there is
higher head in the middle of the
aquifer than at the top or the
bottom. Once again, the results
match what we would naturally
expect. The water spreads out
from the region of higher head
to lower head, so it flows away
from the central region of the
aquifer (see Figure 6). The
higher pressure in the center
forces some of the water to flow
backwards in the aquifer. In
case this is an undesirable side
effect, geologists can use this
result to decide that this
particular recharge method
should not be employed.
8
Figure 7. Aquifer with Increasing Head
Case 3: The initial head
increases linearly from top
to bottom at the entrance
of the aquifer. This
simulates the entry of water
from a sloping region of an
aquifer into our horizontally
oriented region. As
expected, the water flows
towards the bottom of the
aquifer and curves around to
flow horizontally through it.
Our flow plot also indicates
that the water would flow
faster through the upper
region of the aquifer than
through the lower region. It
would be interesting to
check if this was true in an
actual aquifer.
9
Time-dependent solutions (equation 1):
Solving this equation enabled us to observe how the head would change with time in
cases where the water flow was suddenly switched off. We checked this phenomenon in
the event that the initial head was constant. When the flow is turned off quickly, the water
retained in the aquifer flows out of both ends until all the water is drained out. The same
is true for an artificial recharge case where the water in the middle region of the aquifer is
at a higher pressure than the water at the top or the bottom. Again, this behavior is what
we would expect.
Figure 8. Movie of Aquifer Losing the Artificial Injection
10
Continuously varying conductivity:
For the case with a continuously varying conductivity (we tried T = x + y) , we
had problems due to the boundary conditions at the floor and ceiling of the aquifer. The
head at the floor and
the ceiling of the aquifer goes to zero very quickly. This forces the head at other points in
the aquifer to zero very quickly as well. We thought that the problem lay in the difference
formula that we were using. We tried
to check this by setting the head for Figure 9. Aquifer with T= x+y.
the two layers surrounding the
ceiling and the floor so that the
derivative was automatically set to
zero, in much the same way as we
set the initial entry flow velocity.
This did not have any substantial
effect on the outcome. We
rechecked the math numerous
times and have come to the
conclusion that it is either some
error in the code or a fundamental
problem with our theory of how
this should work.
Problems with our code in general:
Since we are attempting to solve a boundary value problem, we would typically
set the initial head or its derivative on all four boundaries. However, we have no
knowledge of the outgoing groundwater flow and we expect the head to change across
the aquifer, so we cannot set the final boundary condition in x. The way we tried to work
around this was to specify the head and its derivative for the initial boundary condition in
x. However, it is not possible to specify both in the same matrix. Specifying the head at
the first three initial points allowed us to indirectly specify the initial derivative as well.
We tried to implement a method where we solved for the first set of unknown
points by using a backwards difference method that took into account all three initial
points. Our solution for this looked really unstable, and so we resorted to using the
central difference formula.
We left in the first three known points because it makes our matrix non singular
but the initial derivative of the head (i.e. the initial velocity) does not affect the
calculation of the head.
We also observed that whatever the initial head is set to, it will always decrease to
zero at the end. This is incorrect because head is a measure of energy and should not
necessarily go to zero at the exit point of the aquifer, since the water should still possess
some energy. The velocity is somehow set so that the final head will go to zero, which is
11
why the initial velocity that we set never affects the velocity and the head calculated by
the program. We think that there is a problem with implementing the math in our method
or that there exists an underlying part of the math that we have not taken into account.
Conclusion
We conclude that the finite difference method may not be the best way to model
groundwater flow in an aquifer since we had trouble specifying the boundary conditions
in our computation. Qualitatively, our solutions are correct for the simple cases of
Laplaces equation and simple versions of the time dependent flow. However, when we
tried to adapt our code for more complex problems where the conductivity of the aquifer
varied continuously, the method breaks down. There are also parts of our solution that we
dont really have an explanation for, such as the manner in which our program disregards
the initial flow velocity. Sorting out these errors might result in a working version of the
program, but we doubt that it could be upgraded to solve very complex cases of the flow.
References
1. C.W. Fetter, Applied Hydrogeology, Third Ed., Merill Publishing Company,1994
2. Private Communication:
a. Dr. Richard Elderkin, Mathematics Dept., Pomona College
b. Dr. Linda Reinen, Geology Dept., Pomona College

Potrebbero piacerti anche