Groundwater Modeling

239

For example, in a one-dimensional (1D) groundwater flow equation for a hetero-

geneous, isotropic, and confined aquifer, Equation 6.12 can be written as

⎛ h n +1 − hin ⎞ ⎛ hin−1 − hin ⎞ ⎛ hin+1 − hin ⎞

S⎜ i ⎟ = Kb ⎜ 2 ⎟ + Kb ⎜ 2 ⎟ (6.13)
⎝ Δt ⎠ ⎝ (Δx ) ⎠ ⎝ (Δx ) ⎠

or

T Δt ⎛ 2T Δt ⎞
hin +1 = 2
(hin−1 + hin+1 ) + hin ⎜ 1 − 2 ⎟ (6.14)
S (Δx ) ⎝ S (Δx ) ⎠

Explicit finite-difference equations are simple to solve but when time increments are
too large, small numerical errors can propagate into larger errors in the next compu-
tational stages. A stable solution is ensured in 1D heterogeneous case if

T Δt 1
< (6.15)
S (Δx )2 2

Consequently, the time increment cannot be selected independently of the space

increment.

Example 6.1

Consider a nonsteady, 1D flow in a confined aquifer shown in Figure 6.4. Let

Δx = 3 m, b = 3 m, h1 = 5 m, h5 = 1 m for t > 0, K = 0.5 m/day, S = 0.03. The initial
conditions are h1 = h2 = h3 = h4 = h5 = 5 m. Determine the spatial variation of
piezometric head.

h1
h5
b

1 2 3 4 5
∆x

FIGURE 6.4  The confined aquifer in Example 6.1.

240 Groundwater Hydrology: Engineering, Planning, and Management

Solution
To satisfy the stability requirement of Equation 6.15, the maximum time step Δt is
computed as

T Δt ⎛ 2T Δt ⎞
hin +1 = (hin−1 + hin+1) + hin ⎜ 1−
S( Δx )2 ⎝ S( Δx )2 ⎟⎠

S( Δx )2 1 (0.03)(3)2
Δt < = = 0.09 day
2T 2 0 .5 × 3

Therefore, the time increment is selected as 0.08 days. With assumption of h1 =
h2 = h3 = h4 = h5 = 5 m, h10.08 = h20.08 = h30.08 = h40.08 = 5 m, h50.08 = 1m. For the first
time, step grid (4) is affected and Equation 6.14 becomes

T Δt ⎛ 2T Δt ⎞
h42× 0.08 = (h30.08 + h50.08 ) + h40.08 ⎜ 1−
S( Δx )2 ⎝ S( Δx )2 ⎟⎠

1.5 × 0.08 ⎛ 2 × 1.5 × 0.08 ⎞

= (5 + 1) + 5 × ⎜ 1− = 3.22m
0.03(3)2 ⎝ 0.03(3)2 ⎟⎠

h22× 0.08 = h32× 0.08 = 5 m

For the second time step t = 3 × 0.08 for grid (4):

h23× 0.08 = 5 m,  h33× 0.08 = 4.21m,  h43× 0.08 = 3.02m

The above process is repeated until the head at each grid is calculated at the
desired time. To illustrate the stability problem, a set of calculations was made in
which Δt was selected to be 0.12 days so that the expression for stability results in

T Δt 1.5 × 0.12 1
= >
S( Δx )2 0.03(3)2 2

The calculated head in grid (4) as a function of time is shown in Figure 6.5.
The computed values fluctuate with each time step for Δt = 0.12, giving com-
pletely erroneous results. Also, the amplitude of the fluctuation increases with
increasing time.

6.4.2  Backward Difference Equation

Figure 6.3b shows the time derivative as a backward difference from the heads at
time level, n − 1, which are the known heads. Therefore, the difference equation
of each node will have five unknown variables. For a grid, which has N nodes,
there is a system of N equations containing N unknown variables. This system
Groundwater Modeling 241

8
dt = 0.08 days
6 dt = 0.12 days
Head (m) 4

0
1 3 5 7 9 11 13 15
–2

–4

–6
Time (days)

FIGURE 6.5  Calculated piezometric head in grid (4), using forward difference equation.

of equations can be solved simultaneously considering the boundary conditions.

This method is called forward difference or implicit method. The implicit finite
differential form of 2D groundwater equation, (Equation 6.1) can be expressed as
follows:

⎛ h n − h n −1 ⎞ ⎛ hn − hn ⎞ ⎛ hn − hn ⎞
ss ⎜ i, j i, j ⎟ = k x (i −1/ 2, j ) ⎜ i −1, j 2 i, j ⎟ + k x (i +1/ 2, j ) ⎜ i +1, j 2 i, j ⎟
⎝ Δt ⎠ ⎝ ( Δx ) ⎠ ⎝ ( Δx ) ⎠

⎛ hn − hn ⎞ ⎛ hn − hn ⎞
+ k y (i, j −1/ 2 ) ⎜ i, j −1 2 i, j ⎟ + k y (i, j +1/ 2 ) ⎜ i, j +1 2 i, j ⎟ (6.16)
⎝ ( Δy) ⎠ ⎝ ( Δy) ⎠

For example, in a 1D groundwater flow equation for a heterogeneous, isotropic, and

confined aquifer, Equation 6.16 can be written as

⎛ h n − hin −1 ⎞ ⎛ hn − hn ⎞ ⎛ hn − hn ⎞
S⎜ i ⎟ = Kb ⎜ i −1 2 i ⎟ + Kb ⎜ i +1 2 i ⎟ (6.17)
⎝ Δt ⎠ ⎝ ( Δx ) ⎠ ⎝ ( Δx ) ⎠

Rearranging Equation 6.17 so that all of the known values are on the right-hand side
of the equal sign results in

⎛ S ( Δx )2 ⎞ S ( Δx )2 n −1
hin−1 − hin ⎜ 2 + ⎟ + hin+1 = − hi (6.18)
⎝ T Δt ⎠ T Δt

The head in grid (i) depends upon the value of head at time n in the adjacent grids,
(i + 1) and (i − 1). Thus, Equation 6.18 represents a set of algebraic equations that
must be solved simultaneously.
242 Groundwater Hydrology: Engineering, Planning, and Management

Example 6.2

Solve Example 6.1 using the backward difference equation.

Solution
Equation 6.18 is used for determining the three interior grids (2), (3), and (4). Grids
(1) and (5) are boundary grids and values of head at these grids are specified as
5 and 1 m, respectively. With assumption of h1 = h2 = h3 = h4 = h5 = 5 m, and
h10.08 = h20.08 = h30.08 = h40.08 = 5 m and h50.08 = 1 m, for the second time step Δt = 0.08
days, the following equations for grids (2), (3), (4) are obtained:

5 − 4.25h20.16 + h30.16 = −11.25

h20.16 − 4.25h30.16 + h40.16 = −11.25

h30.16 − 4.25h40.16 + 1 = −11.25

Rearranging the above equations so that all known values are placed on the right-
hand side and summing them up, we get

h20.16 = 4.9451, h30.16 = 4.7665, h40.16 = 4.0627 m

Example 6.3

Develop the finite-difference approximation (FDA) of flow equation for a 2D semi-

confined, inhomogeneous, and anisotropic aquifer.

Solution
The FDA of flow equation in semi-confined aquifer equation in an inhomoge-
neous and isotropic aquifer is as follows:

1 1
Δx 2
{ }
Txi +1/ 2, j (hi +1, j − hi , j ) − Txi −1/ 2, j (hi , j − hi −1, j ) +
Δy 2
{
Tyi , j +1/ 2(hi , j +1 − hi , j )

Qw ,i , j K a,i , j (Ha,i , j − hi , j )
}
−Tyi , j −1/ 2(hi , j − hi , j −1) ±
Δx Δy
+
ba
= Si , j hi , j (6.19)

By rewriting Equation 6.19 for all internal nodes (i, j; i = 1, …, m and j = 1, …, m)

of the domain, a set of linear equations can be expressed in the following form:

Ah• + Bh + g = 0 (6.20)

In these equations, which are called dynamic response equations, h is a col-

umn vector of unknown heads, h = (h11, …, h1m, …, hn,1, …, hnm)T. Vector g con-
tains the rate of pumping and injections that can be the decision variables in the