Advances in Engineering Software 36 (2005) 374–384

www.elsevier.com/locate/advengsoft

Transient groundwater modeling using spreadsheets

Halil Karahan*, M. Tamer Ayvaz
Department of Civil Engineering, Faculty of Engineering, Pamukkale University, 20017 Denizli, Turkey
Received 16 July 2003; received in revised form 20 December 2004; accepted 4 January 2005

Abstract

This study proposes a transient groundwater modeling using spreadsheet simulation (TGMSS) model for solving groundwater problems.
TGMSS may be considered as a practical method and introduction to groundwater modeling that uses spreadsheets instead of conventional
groundwater model codes. Irregular aquifer geometry, variable boundary conditions and sinks and/or source values, heterogeneous aquifer
parameters (conductivity, storage capacities), may be evaluated in the TGMSS. Different management scenarios may be analyzed by
obtaining the groundwater level of different times. Two hypothetical examples are tested with TGMSS and verified with MODFLOW.
Results showed that the TGMSS and MODFLOW results were in good agreement in terms of resulting values of hydraulic heads in all cases.
q 2005 Elsevier Ltd. All rights reserved.

Keywords: Transient; Groundwater modeling; Spreadsheet; Hydraulic head; Modflow

1. Introduction easily be plotted at the same worksheet. Any changes in the

Groundwater modeling is one of the most important
topics in engineering and geosciences. Groundwater systems
can be modeled using partial differential equations (PDEs).
Governing equations of groundwater modeling can be solved
by analytical and numerical solution methods. Analytical
solutions of these equations are possible in simple and ideal
cases with regularly shaped aquifers and homogeneous
hydraulic properties. Aquifer systems in the groundwater
modeling are a complex task since its heterogeneous
structure. Many numerical solution algorithms have been
developed to solve PDEs such as finite difference method
(FDM) and finite element method (FEM), which have used
since 1960s. Several studies have been carried out by many
researchers using FDM [1–4] and FEM [5–8].
The development of computer technology may ease
solving the PDE in groundwater modeling. One of the best
tools for solving the PDEs is spreadsheet. There are many
advantages of spreadsheets such as having numerical and
visual feedback and fast calculating capabilities. One of the
most important advantages of spreadsheets is its graphical
interface. The solution obtained through the spreadsheet can
* Corresponding author. Tel.: C90 258 2134030; fax: C90 258 2125548.
E-mail address: hkarahan@pamukkale.edu.tr (H. Karahan).
0965-9978/$ - see front matter q 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.advengsoft.2005.01.002
input parameters of the solution domain will be directly
reflected to the graphical representation of the solutions.
Spreadsheets are user-friendly and easy to program.
Recently, popularity of spreadsheets in the solution of
engineering problems has been increasing. Several studies
have been carried out using spreadsheets for the last 10
years. The application of them have been carried out in
different fields of engineering problems such as in solution
of PDEs [9], one-dimensional transient heat-conduction
problems [10], and free-surface seepage problems [11].
There are some studies in the literature [12–15] for
groundwater applications, although there are many appli-
cations of spreadsheet in engineering fields. Olsthoorn [16]
showed that spreadsheet is useful tool for two and three-
dimensional steady-state and transient groundwater pro-
blems with homogeneous aquifer parameters and constant
sinks and/or source terms.
Anderson and Bair [17] showed that spreadsheets provide
an easy way for understanding groundwater problems prior
to using MODFLOW. They solved one-dimensional transi-
ent and two-dimensional steady-state groundwater problems
for homogeneous aquifer parameters and constant sinks and/
or source terms. In addition they suggested as advanced
topics that the block-centered flow (BCF) package in
MODFLOW may be simulated on the spreadsheets.
 H. Karahan, M.T. Ayvaz / Advances in Engineering Software 36 (2005) 374–384 375

In groundwater modeling on a spreadsheet, it is not Uk vector of boundary conditions which is a function of

necessary to write out an equation in all cells to carry out U(x, y) (kZ1, 2, 3,.) [L]
iterative calculations. Using copy and paste features of the
spreadsheets, FDM equation can be copied to other cells
without writing the equations to all cells individually. When 3. Mathematical model
the equation pasted to all cells of the solution domain,
iterative calculation is started. It is carried out until the The governing equation of a vertically integrated Darcy’s
given number of iteration is finished or maximum flow in a two-dimensional confined, compressible, isotropic,
convergence criterion has been met. heterogeneous aquifer is:
While solving the groundwater modeling system in the


v vh v vh vh
steady-state case is relatively easy task, but transient K C K GW Z S (1)
solutions are quite difficult in terms of time dimension in vx vx vy vy vt
the governing equations. Inclusion of time dimension in One of the well-known solution methods to solve Eq. (1)
PDEs may lead to increase CPU time if conventional is the FDM. FDM solution technique can be classified into
methods are used. However, transient groundwater model- two groups: implicit and explicit methods. Implicit methods
ing problems may be solved based on iterative spreadsheet are interpolative and require a solution of simultaneous
calculation since spreadsheets eliminates the matrix algebra equations for each time interval using matrix algebra and
to vector form. require finding the inverse of the matrix for each time
The main objective of this study is to demonstrate a user- interval. Thus, computations are complicated and time
friendly and flexible groundwater modeling simulation consuming. Explicit methods are extrapolative and require a
algorithm using the FDM. Therefore, transient groundwater simple solution of an algebraic equation for one unknown
modeling using spreadsheet simulation (TGMSS) has been for each point and for each time interval. The CPU time is
proposed. Variable grid sizes, aquifer parameters, sinks usually much smaller than the implicit technique.
and/or source terms may be applied to groundwater FDM solution technique can be further classified into two
modeling problems having ideal or complex geometries in groups: alternating direction explicit method (ADEM) [18]
the model. The results of the TGMSS are composed with and alternating direction implicit method (ADIM) [19].
MODFLOW results, which is a well-known model code for ADEM equations can be written as:
groundwater modeling. mC1 m m m
Hi;j K Hi;j ðHiC1;j K Hi;j Þ
Si;j Z KiC12 ;j 2
Dt Dx
2. Notation mC1 mC1 m m
ðHiK1;j K Hi;j Þ ðHi;jC1 K Hi;j Þ
C KiK12 ;j 2
C K i;jC1
2
N number of grid spacing in the x direction (iZ1, 2, Dx 2 Dy
3,., N) mC1 mC1
ðHi;jK1 K Hi;j Þ
M number of grid spacing in the y direction (jZ1, 2, C Ki;jK12 2
C Wi;j (2)
3,., M) Dy
Dx grid spacing in the x direction [L] and
Dy grid spacing in the y direction [L] mC1 m m m
Hi;j K Hi;j ðHiK1;j K Hi;j Þ
Dt time step [T] Si;j Z KiK12 ;j
Dt 2
t time [T] Dx
n number of iteration mC1
ðHiC1;j mC1
K Hi;j Þ m
ðHi;jK1 m
K Hi;j Þ
m time indicator C KiC12 ;j 2
C K 1
i;jK2 2
Dx Dy
Dm increment of time indicator
h hydraulic head, h2H [L] mC1
ðHi;jC1 mC1
K Hi;j Þ
H matrix of hydraulic head, ci, j (iZ1, 2, 3,., N and C Ki;jC12 2
C Wi;j (3)
Dy
jZ1, 2, 3,., M) [L]
K matrix of hydraulic conductivity, ci, j (iZ1, 2, 3,., N where
and jZ1, 2, 3,., M) [LTK1] mC1
S matrix of specific storage, ci, j (iZ1, 2, 3,., N and Qi;j 2
Wi;j ZG (4)
jZ1, 2, 3,., M) [LK1] DxDyHi;j
Q matrix of pumping rate, ci, j (iZ1, 2, 3,., N and jZ1,
In order to obtain a stable solution for
2, 3,., M) [L3TK1]
explicit
h groundwater
i modeling, the stability criterion of
R matrix of recharge, ci, j (iZ1, 2, 3,., N and jZ1, 2, KH Dt Dt 1
3,., M) [LTK1] S ðDxÞ2
C ðDyÞ2
% 2 must be satisfied [20].
W matrix of sinks and/or sources, ci, j (iZ1, 2, 3,., N The ADIM is a cycle of calculation and consists of two
and jZ1, 2, 3,., M) [TK1] stages in Eqs. (2) and (3). In each stage, weight is placed on
376 H. Karahan, M.T. Ayvaz / Advances in Engineering Software 36 (2005) 374–384

two flux terms at tZ(mC1)Dt and two flux terms at tZmDt. 4. Model development
In the first stage, the flux terms in the x and y directions are
taken tZ(mC1)Dt and tZmDt, respectively. In the second The general structure of problem domain for the
stage, the flux terms in the x and y directions are taken TGMSS, which uses FDM, can be seen in Fig. 1. The
tZmDt and tZ(mC1)Dt, respectively. There is no FDM equations in the cell is easily generated as much as it is
instability problem in this method. required depending on the Dx and Dy, which are the size of
By arranging Eqs. (2) and (3), Eq. (5) which consists of grids in the solution domain.
one-stage solution algorithm can be obtained as:
 
m CCðHi;j Þ C CEðHiC1;j Þ C CWðHiK1;j Þ C CSðHi;jC1 Þ C CNðHi;jK1 Þ C Wi;j
Hi;j Z
ðCE C CW C CS C CN C CCÞ
(5) The TGMSS model is divided into rectangular grid
where intervals both in x and y directions in order to carry out the
iterative spreadsheet calculation. The TGMSS model takes
Si;j Dx, Dy, Dt, aquifer parameters (e.g. hydraulic conductivity,
CC Z (5a)
Dt storage capacity, initial hydraulic heads), iteration number

 and/or maximum change (3), time (mDt) as input par-
1
CE Z ½K i;j C KiC1;j  (5b) ameters. The flowchart of TGMSS can be seen in Fig. 2.
2Dx2 The TGMSS consist of two loops as in Fig. 2; inner loop

 and outer loop. The inner loop computes the hydraulic heads
1 for given time interval until the maximum convergence
CW Z ½Ki;j C KiK1;j  (5c)
2Dx2 criterion (3) is met, then, the outer loop controls the time

 dimensions for subsequent use in inner loop. Maximum
1 convergence criteria (3) for inner and outer loops can be
CS Z ½Ki;j C Ki;jC1  (5d)
2Dy2 given as follows:
m;nC1 m;n

 Emax;1 Z MaxjHi;j K Hi;j j (6)
1 mC1;n m;n
CN Z ½Ki;j C Ki;jK1  (5e) Emax;2 Z MaxjHi;j K Hi;j j (7)
2Dy2

Qi;j START
Wi;j ZG (5f)
DxDyHi;j
- Definition of Geometry
TGMSS uses the iterative alternating direction implicit - Aquifer Parameters
- Initial Time and Time Steps
method (IADIM). So, it does not need any stability criteria - Tolerance (e)
since Eq. (5) is iteratively solved with neighboring cells
simultaneously. t = mnew∆t

i Hj,i

n=n+1
1 5 10 15 20 m,n+1 m,n+1
Emax,1= Max Hi,j − Hi,j
1
mnew = mold + ∆m

j NO
Emax,1 < ε
5
YES
m
Hj,i
10
m+1,n
Emax,2= Max Hi,j − Hi,jm,n

15 NO Emax,2 < ε

YES

20 RESULTS

END

Fig. 1. General structure of problem domain. Fig. 2. The TGMSS flowchart.

 H. Karahan, M.T. Ayvaz / Advances in Engineering Software 36 (2005) 374–384 377

Fig. 3. Solution domain of the TGMSS model.

Fig. 3 shows the solution domain represented in criteria (3) as can be seen in Fig. 4. When both criteria are
spreadsheet. This sheet is called ‘Solution’. Six additional met, the iterative calculation is stopped.
sheets are indicated by the ‘tabs’ in Fig. 3. Solution of TGMSS model is carried out based on Eq. (5)
Storage describes the inputs for the storage capacity as a in the following spreadsheet format:
separate sheet, conductivity describes the inputs for the
hydraulic conductivity as a separate sheet, and source and C4 Z ððStorage!C4=Solution! $ AA $ 11Þ * Solution!C4
recharge_sources describe the inputs for the sinks and/or C $ AA $ 9 * ðConductivity!C4
sources, which computed using the pumping rate, and
recharge values in the pumping and recharge sheets. Note C Conductivity!D4Þ * Solution!D4
that the sinks are computed using negative (K) pumping rates,
and sources are computed using positive (C) pumping rates. C $ AA $ 9 * ðConductivity!B4
Iteration number and/or maximum convergence criteria
C Conductivity!C4Þ * Solution!B4
(3) are given in the TGMSS as follows:
Select ‘Tools’ pull-down menu of the spreadsheet and C $ AA $ 10 * ðConductivity!C4
select ‘Options’ in it. Then, select the calculation tab. The
iteration option is not selected when the spreadsheet is C Conductivity!C3Þ * Solution!C3
loaded by default. There are two important options given in
this window: one is maximum iterations and the other is C Solution! $ AA $ 10 * ðConductivity!C4
maximum allowable convergence (maximum change) C Conductivity!C5Þ * Solution!C5

C Source!C4Þ=ððStorage!C4=Solution! $ AA $ 11Þ

C $ AA $ 9 * ðConductivity!C4

C Conductivity!D4Þ

C $ AA $ 9 * ðConductivity!B4

C Conductivity!C4Þ

C $ AA $ 10 * ðConductivity!C4

C Conductivity!C3Þ

C $ AA $ 10 * ðConductivity!C4

C Conductivity!C5ÞÞ
Fig. 4. Maximum iterations and maximum allowable convergence criteria. (8)
378 H. Karahan, M.T. Ayvaz / Advances in Engineering Software 36 (2005) 374–384

Fig. 5. Pumping sheet.

where C4, which is an intersection of the third column (C) 5.1. First hypothetical example
and fourth row (4), represents the hydraulic head in cell.
Related sheets are used in solving Eq. (5). Cell names Grid spacing, DxZDyZ100 m
consist of column and row numbers and follow the sheet Number of grid spacing, NZMZ23
names ending with the ‘!’ indicator (e.g. Conductivity!C5 or Hydraulic conductivity, KZ15.0 m/day (fix in all cells)
Solution!B5). Storage capacity, SZ1.0 mK1 (fix in all cells)
The output of TGMSS can be seen visually using Time step, DtZ1 day
graphical interface of spreadsheet simultaneously.

5. Numerical applications
The application of the TGMSS model for groundwater
systems is tested using two hypothetical examples. The first
hypothetical example has homogeneous aquifer parameters
and is further divided into two cases. In addition, the second
hypothetical example has heterogeneous aquifer parameters
and is examined under two scenarios.
5.1.1. Case I
Three hypothetical wells having constant pumping rates
are located as sinks and/or sources. The wells are located
symmetrically in the solution domain. Pumping rates,
Q 1 ZQ 3Z432.0 m 3/day (sinks), Q 2Z864.0 m3 /day
(source).
The pumping rates in Fig. 5 are converted to sink and/or
source values using the procedure in Fig. 6 to avoid the
problem of zero division. Note that the value of H in Fig. 8
(i.e. in Eq. (5f)) is equal to initial hydraulic head before

Fig. 6. Source sheet.

 H. Karahan, M.T. Ayvaz / Advances in Engineering Software 36 (2005) 374–384 379

Fig. 7. Values of Hi,j after computations (tZ3600 days).

the iteration is started. For this example, the initial hydraulic
head is given 20 m in each cell.
To start the iterative calculation, cell of C4 is copied and
pasted from C4 to W24. Column and row numbers of Eq. (8)
will be changed in each cell among the range of C4:W24.
Results of the TGMSS can be seen in Fig. 7. The
boldface cells indicate the well locations. Note that the
values of hydraulic heads in two symmetrically located
wells are equal.
Three-dimensional representation of hydraulic heads can
be seen in Fig. 8. This figure shows the level of hydraulic
heads over the solution domain. Each contour of Fig. 8
indicates the various levels of hydraulic heads.
The TGMSS model has been verified with MODFLOW.
The results of TGMSS model and MODFLOW showed good
agreement. The contour of TGMSS is given in Fig. 9(a), and
the contour of MODFLOW is given in Fig. 9(b).
The contours of the TGMSS model can be seen in
Fig. 10(a) for tZ1 day, Fig. 10(b) for tZ5 days and
Fig. 10(c) for tZ10 days. Variation of groundwater table
(GWT) at any node becomes steady-state as the time
increases. Fig. 11 shows the variation of GWT where

20,90
20,75
20,60
20,45
20,30
20,15
20,00
19,85
19,70
19,55
19,40
H (m) 19,25
19,10
18,95
18,80
18,65
18,50
18,35
18,20
18,05
17,90
17,75
17,60
17,45 S21
17,30
17,15 S17
17,00
S13
1
6 S9
y (grid numbers)
11 S5
x (grid numbers) 16
S1
21

Fig. 8. Three-dimensional variation of groundwater level (tZ3600 days).

380 H. Karahan, M.T. Ayvaz / Advances in Engineering Software 36 (2005) 374–384

(a) 2200 (b) 2200

2000 2000
21.00 21.00
1800 1800
20.80 20.80
20.60 20.60
1600 20.40
1600 20.40
20.20 20.20
1400 20.00 1400 20.00

y (m)
19.80 19.80
y (m)

1200 19.60 1200 19.60

19.40 19.40
1000 19.20 1000 19.20
19.00 19.00
800 18.80 800 18.80
18.60 18.60
600 18.40 600 18.40
18.20 18.20
400 18.00 400 18.00

200 200

200

400

600

800

1000

1200

1400

1600

1800

2000

2200
200

400

600

800

1000

1200

1400

1600

1800

2000

2200
x (m) x (m)

Fig. 9. Contours of the TGMSS and MODFLOW for tZ3600 days. (a) Graphical output of the TGMSS. (b) Graphical output of the MODFLOW.

(a) 2200 ( b ) 2200

2000 2000
1800 21.00 21.00
20.80
1800 20.80
20.60 20.60
1600 20.40 1600 20.40
20.20 20.20
1400 20.00 1400 20.00
y (m)

y (m)

19.80 19.80
1200 19.60 1200 19.60
19.40 19.40
1000 19.20 1000 19.20
19.00 19.00
800 18.80 800 18.80
18.60 18.60
600 18.40 600 18.40
18.20 18.20
400 18.00 400 18.00

200 200
200
400
600
800
1000
1200
1400
1600
1800
2000
2200
1000
1200
1400
1600
1800
2000
2200
200
400
600
800

x (m) x (m)

(c ) 2 2 0 0
2000
21.00
1800 20.80
20.60
1600 20.40
20.20
1400 20.00
y (m )

19.80
1200 19.60
19.40
1000 19.20
19.00
800 18.80
18.60
600 18.40
18.20
400 18.00

200
200
400
600
800
1000
1200
1400
1600
1800
2000
2200

x (m)

Fig. 10. Variation of groundwater level for (a) tZ1 day, (b) tZ5 days and (c) tZ10 days.
 H. Karahan, M.T. Ayvaz / Advances in Engineering Software 36 (2005) 374–384 381

20.00 Table 1
19.80 Total mass errors and CPU times for Case 1 and 2
Hydraulic Head (m)

19.60 Time (day) Case 1 Case 2

19.40
19.20
Total mass CPU time Total mass CPU time
error (%) (s) error (%) (s)
19.00
18.80 1 0.0420 37 0.0044 1
18.60 3 0.0890 48 0.0036 5
18.40 5 0.1641 58 0.0033 9
18.20 10 0.2369 66 0.0141 17
18.00 30 0.3358 75 0.1339 48
0.01 0.1 1 10 100 1000 10000 50 0.3411 76 0.2463 70
Log Time (day) 100 0.3559 78 0.3654 95
1000 0.3698 117 0.1689 473
Fig. 11. Variation of the GWT against time at the middle point (M14). 3600 0.3698 231 0.0029 1393

the time indicator (m) is changed from 1 to 3600 in the

The results of the total mass errors and CPU times for the
middle well. For this example, after 100 days GWT
Case 1 and 2 can be seen in Table 1. Note that this example
becomes steady-state.
has been solved on a personal PC (P4 2.80 GHz processor
and 512 MB RAM) and screen updating feature of the
5.1.2. Case 2 spreadsheet was closed during the calculation.
Three hypothetical wells having variable pumping rates
are considered in this case. The following pumping rate 5.2. Second hypothetical example
function may be given as sinks and/or sources.


2pmDt In this example, application of TGMSS model has been
Qi;j Z Qmax sin (9)
360 carried out using heterogeneous aquifer parameters and
where m is the time indicator and Qmax is 432.0 m3/day variable boundary conditions in the complex solution
(sinks) for Well 1 and 3 and 864.0 m3/day (source) for Well 2. domain. The data used in this example are given as follows:
After describing the model, iterative calculations can be
started using the procedure previously described. Fig. 12 Grid spacing, DxZDyZ100 m
Number of grid spacing, NZ23, MZ39
shows the variation of hydraulic heads for the well locations
while the time indicator (m) is changing from 1 to 3600. Hydraulic conductivities, K1Z15, K2Z10, K3Z35,
In the TGMSS model, volumetric rates of sinks and/or K4Z30, K5Z20, K6Z25 m/day
sources need to be checked in each time interval to control Storage capacity, SZ1.0 mK1 (fix in all cells)
entered- and exit-water to the solution domain. Total mass Six hypothetical wells are defined as sinks. Their rates
error can be written as: are:
Q 1ZQ4 ZQ 5Z864.0, Q 2ZQ 6Z432.0, Q3Z
Total mass error ð%Þ 1728.0 m3/day
P P  Recharge, RmaxZ35.265 mm/month (fix in all cells as
 Mass before calculation K Mass after calculation 
Z  P  source),
Mass before calculation 


 pmDt 
Ri;j Z Rmax sin (11)
360 
!100
(10)
The input parameters of hydraulic conductivities and
23.00
Well 1&Well 3
pumping rates are shown in Fig. 13. This example has been
22.50
Well 2 examined under two scenarios. Details of both scenarios are
Hydraulic Head (m)

22.00
21.50 given in Table 2.
21.00 Iterative calculations for Scenario A and B can be started
20.50
20.00
using the procedure previously described. The initial
19.50 hydraulic heads are given 15 m. Before the calculation is
19.00 started, this value needs to be pasted to all cells in the
18.50
18.00
solution domain. The results of the total mass errors and
17.50 CPU times for Scenario A and B is given in Table 3.
17.00 As can be seen in Table 3, total mass errors of Scenario A
0 360 720 1080 1440 1800 2160 2520 2880 3240 3600
Time (day)
and B are in good agreement with accuracy of result, but
CPU times are very different. It may be concluded that
Fig. 12. Variation of hydraulic heads against time for the well locations. Scenario A is better for obtaining the hydraulic heads at
382 H. Karahan, M.T. Ayvaz / Advances in Engineering Software 36 (2005) 374–384

Fig. 13. Hydraulic conductivities and pumping rates.

a desired time than Scenario B. However, Scenario B is

better for obtaining hydraulic heads at sequential times than
Scenario A. Fig. 14 shows the variation of hydraulic heads 19.50
19.00
for the wells locations while the time indicator (m) is varied
18.50
Hydraulic Head (m)

between 1 and 3600. The hydraulic heads at the well 18.00

locations vary depending on recharge values although 17.50
pumping rates of them are constant as in Fig. 14. Moreover, 17.00
total mass conversation corresponding to the Scenario A and 16.50
16.00
B is given in Fig. 15. 15.50
15.00 Well 1 Well 3 Well 5
Table 2
14.50 Well 2 Well 4 Well 6
Scenario A and B used in Example 2
14.00
m Dt (day) 0 360 720 1080 1440 1800 2160 2520 2880 3240 3600
Time(day)
Scenario A 1 1, 3, 5, 10, 30, 50, 100, 1000, 3600
Scenario B 1, 2, 3,., 3600 1 Fig. 14. Variation of hydraulic heads against time for the well locations.

Table 3
Total mass errors and CPU times for Scenario A and B
1.088E+008
Scenario A
Time (day) Scenario A Scenario B Scenario B

Total mass CPU time Total mass CPU time 1.084E+008

Total Mass (m3)

error (%) (s) error (%) (s)

1.080E+008
1 0.0123 172 0.0123 172
3 0.0143 83 0.0143 194 1.076E+008
5 0.0146 57 0.0143 195
10 0.0146 36 0.0147 201 1.072E+008
30 0.0128 17 0.0135 221
50 0.0109 11 0.0112 239 1.068E+008
100 0.0069 7 0.0071 281 0 360 720 1080 1440 1800 2160 2520 2880 3240 3600
1000 0.0084 3 0.0084 793 Time (day)
3600 0.0161 2 0.0161 2364
Fig. 15. Total mass conversation for Scenario A and B.
 H. Karahan, M.T. Ayvaz / Advances in Engineering Software 36 (2005) 374–384 383

Fig. 16. Values of Hi,j after computations (tZ3600 days).

Outputs of the TGMSS model for both Scenario A and B

can be seen in Fig. 16 for tZ3600 days.
Fig. 17 shows contours of hydraulic heads. While the
effect of pumping can be seen around the wells of 1, 2, and
3, and inclination of water surface can be seen around the
wells of 4, 5, and 6.
500

20.00
6. Conclusions and limitations 1000 19.90
19.75
This study proposes the TGMSS model to solve 19.50
groundwater-modeling problems. Each cell in the spread- 1500 19.35
19.25
sheet was matched with middle point of the each element with
19.00
FDM. Copying–pasting, iterative calculation and graphical
y(m)

2000 18.50
representation of spreadsheets has been illustrated. The
18.00
TGMSS model was tested with two hypothetical examples.
17.50
The following conclusions can be drawn from study.
2500 17.00
The TGMSS model may be adapted to learn for basic
16.50
groundwater problems, which may be considered as a
16.00
flexible structure. There are no difference between the
3000 15.50
numerical solutions of homogeneous-isotropic systems and 15.00
heterogeneous-anisotropic systems in terms of solution
techniques and data processing for this problem. 3500
There may be no need to use macros and subroutines in
the TGMSS model. The writing of the input values to the
related sheets and the generations are easy in terms of 500 1000 1500 2000
solution procedure. x(m)
The present method may not be applied for solving large
systems of equations due to the limitation in the size of Fig. 17. Variation of groundwater level for tZ3600 days.
384 H. Karahan, M.T. Ayvaz / Advances in Engineering Software 36 (2005) 374–384
spreadsheets. Another limitation of spreadsheet is that finite
difference grids, which are mapped into spreadsheet cells,
can only be rectangular or triangular. Hence, some geometry
problems may be occurred at the excluding curved
boundaries in complex geometries. Future study should be
on this issue.
Acknowledgements
The authors say a special thanks to anonymous referees
and Dr Halim Ceylan from the Department of Civil
Engineering at Pamukkale University for meticulous proof-
reading and suggestions.
References
[1] Trescott PC, Larson SP. Solution of three-dimensional groundwater
flow equations using the strongly implicit procedure. J Hydrol 1977;
35:49–60.
[2] McDonald MG, Harbough AW. A modular three-dimensional finite
difference groundwater flow model. US Geol Surv Tech Water Resour
Invest 1988;Book 6(Chapter A1):586.
[3] Wang PP, Chunmaio Z. An efficient approach for successively
perturbated groundwater models. Adv Water Resour 1998;21(6):
499–508.
[4] Bakker M. Simulating groundwater flow in multi-aquifer systems with
analytical and numerical Dupuit-models. J Hydrol 1999;222(1–4):
55–64.
[5] France PW. Finite element analysis of three-dimensional groundwater
flow problems. J Hydrol 1974;21:381–98.
[6] Grupta SK, Cole CR, Pinder GF. A finite element three-dimensional
groundwater (FE3DGW) model for a multiaquifer system. Water
Resour Res 1984;20(5):553–63.
[7] Li H, Yang QA. Least-squares penalty method algorithm for inverse
problems of steady-state aquifer models. Adv Water Res 2000;23(8):
867–80.
[8] Mazzia A, Putti M. Mixed-finite element and finite volume
discretization for heavy brine simulations in groundwater. J Comp
Appl Math 2002;147(1):191–213.
[9] Hagler M. Spreadsheet solution of partial differential equations. IEEE
Trans Educ 1987;E-30:130–4.
[10] Gvirtzman Z, Garfunkel Z. Numerical solutions for one-dimensional
heat-conduction equation using a spreadsheet. Comput Geosci 1996;
22(10):1147–58.
[11] Bardet JP, Tobita T. A practical method for solving free-surface
seepage problems. Comput Geotech 2002;29:45–475.
[12] Anderson MP, Woessner WW. Applied groundwater modelling
simulation of flow and advective transport. New York: Academic
Press; 1992.
[13] Freeze A, Cherry JA. Groundwater. Englewood Cliffs, NJ: Prentice-
Hall; 1979.
[14] Fetter CW. Applied hydrogeology. 3rd ed. New York: Macmillan
Company; 1994.
[15] Kinzelbach W. Groundwater modelling. Amsterdam: Elsevier; 1986.
[16] Olsthoorn TN. The power of electronic worksheet: modeling without
special programs. Ground Water 1985;23(3):381–90.
[17] Anderson MP, Bair ES. The power of spreadsheet models. MOD-
FLOW 2001 and other modeling odysseys proceedings, international
ground water modeling center, Colorado School of Mines 2001
p. 815–22.
[18] Saulyev FK. Integration of equations of parabolic type by the method
of nets. New York: Macmillan Company; 1964.
[19] Bobba AG. Numerical model of contaminant transport through
conduit-porous matrix system. Math Geol 1989;21:861–90.
[20] Bear J. Hydraulics of groundwater. New York: McGraw-Hill,
Inc; 1979.