You are on page 1of 12

Brackish Karstic Springs Model: Application

to Almiros Spring in Crete


by Athanasios Maramathas1, Zacharias Maroulis2, and Dimitrios Marinos-Kouris3

Abstract
A mathematical model is proposed to simulate brackish karstic springs. Rainfall data constitutes model input
information while output information is the discharge and the chloride concentration of the water versus time. The
model was constructed by considering the mass and mechanical energy balance on the hydrodynamic analog, which
includes three reservoirs outflowing in a tube that lies adjacent to the spring. Two reservoirs emulate the karstic system, and the third one emulates the sea. The discharge of the spring is given by the sum of the discharge of the reservoirs, and the chloride concentration by the solution of the mixing problem between the fresh and the salty water,
which exists in the tube leading to the spring. The model is applied to the spring of Almiros at Heraklion, Crete,
Greece. The agreement between model values and field measurements is very good for depletion periods and satisfactory for recharge periods.

Introduction
The term karstic spring denotes the point of water
outflow of karstic hydrogeologic reservoirs. The latter are
systems of ground water storage and transfer that comprise
a great number of tubes of various shapes and sizes. Karstic
reservoirs occur in water-soluble rocks, such as limestone.
The tubes are formed by the solvent action of water, usually
on pre-existing discontinuities. In the preceding years,
many studies with computer models have been conducted
on the evolution of a karstic system during the geologic
time (Dreybrodt 1996; Gabrovsek and Dreybrodt 2000;
Palmer 2000; Clemens et al. 1999; Ford et al. 2000). From
the hydraulic point of view, the characteristic of these reservoirs is preferential flow through the tubes and, in many
cases, turbulent flow (Dreybrodt 1988), which makes
Darcys law invalid.

1National Technical University of Athens, School of Chemical


Engineering, Zografos Campus GR-15780, Athens, Greece;
+30.1.7723290; thamar@chemeng.ntua.gr
2National Technical University of Athens, School of Chemical
Engineering, Zografos Campus GR-15780, Athens, Greece;
+30.1.7723151; maroulis@chemeng.ntua.gr
3National Technical University of Athens, School of Chemical
Engineering, Zografos Campus GR-15780, Athens, Greece;
+30.1.7723148, marinos@chemeng.ntua.gr
Received July 2001, accepted November 2002.

608

For many years, efforts have been made toward the


mathematical simulation of karstic springs. These efforts
evaluated the water reserves of such springs during the
period of irrigation, so the simulations had to do with the
depletion period, during which there is no recharge by rainfall. The first proposed formula was exponential (Maille
1905), giving the depletion period discharge as a function
of initial discharge and time. Ever since, many other
researchers (Shoeller 1967; Castany 1967; Mangin 1970;
Forkasiewicz and Paloc 1967) have suggested similar formulas. Bezes (1976) proposed the BEMER model to simulate the entire hydrograph of a karstic spring. After the
widespread use of computers, ground-water flow solution
models, such as MODFLOW have been used (Angelini and
Dragoni 1997). However, these models have to be supplied
with a great number of geographically spread data, which
are difficult to collect, in order to offer reliable results. Furthermore, they are based on Darcys law, which is not valid
in such reservoirs due to turbulent flow. All of these refer to
karstic springs without brackish water. In the case of brackish or periodically brackish springs, the seawater intrusion
discharge should be taken into account. Some researchers
have used transport models, such as SUTRA and HST3D,
to simulate the sea intrusion to karstic aquifers (Ghassemi
et al. 1996).
The objective of this paper is to present a model that
simulates the entire hydrograph of a karstic spring and, in

Vol. 41, No. 5GROUND WATERSeptember-October 2003 (pages 608619)

case it gets brackish, the variation of the chloride concentration of its water against time. The model, which is fed by the
rainfall measures of the recharge area, uses parameters that
represent the properties of the karstic system, which are estimated by fitting the model to the available field measurements. Actually, the model transforms the rainfall data of the
recharge area to the hydrographs of the spring (Figure 1). It
has been applied to the Almiros spring at Heraklion, Crete,
Greece.

Hydrodynamic Model of
Brackish Karstic Springs
During the initial phase of karstic aquifer depletion,
two emptying rates can be identified: a fast one corresponding to a section of karst that is emptied quickly, and a slow
one corresponding to another section of karst that is emptied rather slowly. The first part of karst consists of a few
tubes of large diameter, and the second consists of many
tubes of small diameter. Shoeller (1967) was the first to
introduce such an idea. Recently, Worthington (1999) and
Worthington and Ford (1997) postulated for a mature karst
aquifer a low hydraulic conductivity related to the rock
matrix and a high one related to large conduits draining the
system. Eisenlohr et al. (1997), on the other hand, speak of
conduits that are immersed in a low hydraulic conductivity
medium with slow flow velocities and discharge at one or
more discrete springs. Thus, it can be argued that two different karstic subsystems, according to hydrogeologic
behavior, feed the karstic springs. If the later happen to be
brackish, they are also fed by the sea. In Figure 2, a hydrodynamic model is presented that is constructed according to
these ideas. The reservoirs karst 1 and karst 2 represent
the karstic subsystems. while the other one represents the
sea. The sea reservoir is of infinite size. The discharge of
the spring at any time is given by the sum of the discharge
of the reservoirs. The chloride concentration of the spring
water is a result of the mixing of fresh water of the karstic
reservoirs with the salty water of the sea. This happens in
the tube, which brings the water to the spring.

Mathematical Model
Spring Discharge Calculation
In order to develop a mathematical model based on the
previously mentioned karstic spring hydrodynamic model,
the mass and energy balances were considered in the reservoirs, which emulate the two karstic subsystems and the
sea. Each reservoir is considered as a whole in which the
mass and energy conservation principle is applied. The pro-

Figure 1. Schematic of the suggested model.

Figure 2. Hydrodynamic model of the karstic springs.

posed model does not estimate the flow characteristics at


any point of the karstic aquifer, but it simulates only what
happens at the water outlet point, i.e., the discharge and the
chloride concentration of the spring water.
The macroscopic mass-balance equation (continuity
equation) and the macroscopic mechanical energy balance
equation in a control volume and the corresponding control
surface have been used. The flow has been taken to be transient in the reservoirs of the karstic subsystems and steady
in the sea. That is because the water levels in the reservoirs
of the karstic subsystems vary with time, whereas the level
in the sea reservoir is stable as this is of infinite size.
Thus, a system of nonlinear differential and algebraic
equations has been obtained from the mathematical manipulation of the mass-balance and energy-balance equations
(Appendix A). The equations of this system and their symbol notation are presented in Table 1. All the parameters are
effective and can be estimated by fitting the proposed
model to available field measurements. Both karstic subsystems have similar equations, but with different parameter values.
From the solution of this system of equations, the
karstic subsystems discharge and the discharge of the sea
reservoir are estimated. The spring discharge is given by the
sum of these discharges.
Estimation of Chloride Concentration of the Spring Water
The chloride concentration of the spring water is given
by the solution of the problem of mixing fresh water of the

Figure 3. Reservoir that emulates the karstic subsystems. EE


is the control surface.

A. Maramathas et al. GROUND WATER 41, no. 5: 608619

609

Table 1
System Equations
Karstic Subsystems Equations

sS1

sS1H

.
dH
1 Q2 2 PS1 5 0
dt

.
dH
Q 32
Q2 2
1
2
HP
S
1
L
c
d Q2 5 0
1
dt
2gS22
sS1

Equations for the Sea


Figure 4. Reservoir that emulates the sea.

P
Q 23
1 3 2 H 5 0
g
2gS23
s
p3 5 wgaH 1 h 1 Hb 2 La Q 22

Variables

Q2
Q3
H
P3

Karstic system discharge (L3/T)


Seawater discharge (L3/T)
Water level over discharge point (L)
Pressure on the sea tube outlet (M/LT2)

Parameters
P
s
S1

g
S2
L
S3
H

La

Rainfall rate (L/T)


Coefficient of storage
Recharge area (L2)
Infiltration coefficient
Gravity (L/T2)
Karstic system outflow tube cross section (L2)
Energy loss coefficient (L1T2)
Seawater ouflow tube cross section (L2)
Depth of seawater outflow tube outlet point (L)
Seawater density (M/L3)
Freshwater density (M/L3)
or w, if the sea feeds the karstic reservoir or fed by
the karstic reservoir respectively
Spring elevation (L)
Specified yield
Pressure loss coefficient of the sea tube outlet region
(ML7)

karstic subsystems with salty water of the sea. This mixing


happens in the tube, which brings the fresh water from the
reservoirs of the karstic subsystems to the spring, beyond
the point where this tube meets the tube from the reservoir
of the sea (Figure 2). The following equation is used to calculate the chloride concentration of the spring water as a
function of the discharge of fresh water of the two karstic
subsystems and the discharge of the sea:
C5

ruCuQu 1 rwCw 1 QK1 1 QK2 2


ruQu 1 rw 1 QK1 1 QK2 2

In this mixing rule type equation, C is the chloride


concentration of the spring water in ppm (parts per million
by weight), the seawater density, C the chloride concentration of the seawater in ppm, Q the discharge of the sea,
610

A. Maramathas et al. GROUND WATER 41, no. 5: 608619

Cw the chloride concentration of fresh water in ppm, QK1


the discharge of the reservoir karst 1 (Figure 2), and QK2
the discharge of the reservoir karst 2 (Figure 2). The variables Q, QK1, and QK2 are estimated by the solution of the
system of equations of the model, while all the other parameters have to be taken from the related bibliography.
Almiros Spring in Heraklion, Crete
The Almiros spring is located 10 km west of the city of
Heraklion at an elevation of ~4 m and at a distance of 1 km
from the sea. It is a periodically brackish karstic spring. Its
discharge fluctuates between 4 m3/sec in the summer and
70 to 80 m3/sec in the winter following heavy rainfall on the
Psiloritis Mountain. The annual water spring quantity
approaches 250  106 m3. At low discharge, the spring
water is brackish because of seawater entering the spring
reservoir.
The hydrogeologic basin of the spring is structured by
three geologic formations (Fitrolakis 1980; Vidakis 1983).
These formations, from top to bottom, are:
1.
2.

3.

4.

Marl (Neogene; Vidakis 1983)an impermeable rock.


Limestone of the geotectonic zone of Tripolis (upper
Triassicupper Jurassic; Vidakis 1983)has preferential permeability and small porosity.
Phyllites (Permianupper Triassic; Vidakis 1983)an
impermeable rock that is not impeded everywhere but
at certain points between the two other rocks.
Platen limestone of the Ionian (Vidakis 1983) geotectonic zone (middle JurassicEocene; Vidakis 1983)
rock with less preferential permeability and larger
porosity than the Tripolis zone limestone (Figure 5).

Phyllites is overthrust on the Ionian limestone and the


Tripolis limestone is overthrust on the Phyllites. However,
in some places, the Tripolis limestone is overthrusted
directly on the Ionian limestone (Fitrolakis 1980, Vidakis
1983). In the area there are more or less vertical faults of
north-south and east-west general direction that cut across
all the mentioned formations. The spring is on the mixing
zone of two faults, one on the direction of north-south and
the other on the direction of east-west (Vidakis 1983).
The springs recharge area is almost the entire Psiloritis Mountain area with the exception of a rather small
region in the south, the water of which feeds the Zaros

Figure 5. Conceptual representation of the Almiros karstic spring reservoir.

spring next to Zaros village. Because of contact of the


spring reservoir with the sea, its water gets brackish in the
summer when fresh water pressure is low as there is no rainfall.
With regard to the problem of the spring water becoming brackish, it has been assumed by most geologists, following Breznic (1973), that the water becomes brackish
where the fresh water tube meets another tube that comes
from the sea. If the pressure at this mixing zone is smaller
than rugH , where ru is the seawater density, g the gravity,
and H the depth of the mixing zone under the sea level, seawater enters into the fresh water tube through the sea tube
and causes the spring water to become brackish. (Figure 5,
inset). From a simplified point of view, the pressure at the
mixing zone is proportional to the depth of this zone below
the water level of the karstic reservoir and to the fresh water
density. Following that, in some cases the pressure becomes
smaller than rugH because the seawater density is larger
than the fresh water one.

Table 2
Almiros Spring Model Equations
Tripolis Limestone
SS

.
1 Q 2 (1 2 f)PS 5 0

t
sSHt

Ht 1t 2 Ht
t

Q 3
2gS2

Q 2
.
2 Ht 1 t (1 2 f)PS 1 L a
b Q 5 0
sS
Ionian Zone Limestone
sS

Ht 1 t 2 Ht
t

.
.
1 Q 2 ftPS 2 PS 5 0

sSHt

Almiros Spring Simulation


with the Proposed Model
The proposed model has been used to simulate the
Almiros spring. We believe that in this spring the two
karstic subsystems, which the model has taken into account,
are geographically discrete. Karst 1 was identified with the
Tripolis zone limestone while karst 2 with the limestone of
the Ionian zone. In the case of that spring, we introduced
one more parameter to the model, to make it more realistic.
This is the percentage of water (estimated by the fitting of
the model) that feeds the Ionian zone limestone directly
from the Tripolis reservoir. This is a possible process as the
Tripolis zone limestone lies over the Ionian zone limestone

Ht 1 t 2 Ht

Ht 1 t 2 Ht
t

Q 3
2gS2

Q 2
.
.
2 Ht 1 t afPS 1 PS 1 L a
b Q 5 0
sS
Sea
Q2
2gS2

P3
g

2 H 5 0

P3 5 wgaH 1 h 1

Ht 1 t b 2 LaQ2

A. Maramathas et al. GROUND WATER 41, no. 5: 608619

611

stratigraphically (Figure 5). The period of simulation covers


the years 199596 and 199697.
Following replacement of the differentials by finite differences, the nonlinear differential equations of Table 1
become nonlinear algebraic equations (Table 2). Thus, a
system of algebraic equations must be solved, whose symbol notation is shown in Table 3.
The solution of the system of equations has been
obtained using a commercially available spreadsheet. The
rainfall data has been taken from the rain gauge of the
Anogia area. The governmental bureau for the water
resources of Crete carried out all the field measurements
(Tables 4, 5, and 6). Rainfall during the depletion period
that had not affected the spring discharge was not taken into
account. Such rainfall was detected from time series of the
discharge of the spring (field measurements) and the rainfall for the covered period. Precipitation from melting snow

was not taken into account either. Time interval for the
model calculations was one day (t = 1 day).
The model fitting has been done with the least squares
method using commercially available optimization software. According to this method, the optimization program
was tasked to estimate the parameter values so that the
mathematical expression
2
2
a 3 1 QMt 2 QFt 2 4 1 a 3 1 CMt 2 CFt 2 4
D

t51

t51

becomes least. In this expression, QMt, CMt are the model


calculated values for the discharge and chloride concentration; QFt, CFt the field measurements, respectively; and D
the time period for the optimization (number of days).
Detailed analysis showed that there is not a strong correlation between the fitted parameters. The most sensitive
parameters (coefficient of storage, specific yield) were

Table 3
Parameters, Variables, and Output Information of Almiros Spring Model
Parameters Symbol Notation

Indices Notation

f
S,

S, ,
H
h
s

L
L

w
g

Daily rainfall
Infiltation coefficient
Percentage of Tripolis water that feeds Ionian limestone directly
Recharge area

Tube cross section


Depth of seawater outflow tube outlet point
Spring elevation
Coefficient of storage
Specific yield
Energy loss coefficient
Pressure loss coefficient of the sea tube outlet region
or w, if the sea feeds the karstic reservoir or fed by the karstic
reservoir respectively
Seawater density
Freshwater density
Gravity

Tripolis zone limestone


Ionian zone limestone
Sea
Tripolis reservoir outlet
Ionian reservoir outlet
Sea reservoir outlet

Variables Symbol Notation

Time Interval

Qt+t
Qt+t
Q
Ht+t
Ht+t
p3

t = 1 day

Tripolis limestone discharge


Ionian limestone discharge
Seawater discharge
Water level over discharge point of Tripolis limestone
Water level over discharge point of Ionian limestone
Pressure on the seawater tube outlet region

Model Output Information

Q 5 Qt 1 t 1 Qt 1 t 1 Q*

ppmCl 2 5

CQ 1 w Cw (Qt 1 t 1 Qt 1 t )
Q 1 w (Qt 1 t 1 Qt 1 t )

*Seawater discharge is accepted positive if the sea feeds spring reservoir or negative if the sea is fed by spring reservoir.

612

A. Maramathas et al. GROUND WATER 41, no. 5: 608619

Table 4
Rainfall 19951997 (mm)
Sep Oct Nov Dec Jan Feb Mar Apr May June July Aug
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

0
1.4
0
3.4
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2
0
0
0
0
0
3
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1 12.6
0
11
0
0
0
0
8.4
0
4
0.5
0
0
0

2.7
1.4
0
6.4
4
3
7
0
1
0
0
0
0
0
0
0
0
0
0
1
43.6
26
4.2
0
0
0
0
0
0
0

0
0
0
21
0
0
0.5
20
0
9
9.4
0
2
0
0
3
0
1.4
2
0.4
0
0
0
0
0
0
0
0
2
3.6
0

15.6 38
4.2
0
20.7 0
47.2 0
12
2
0
0
3
15
4
30
14.4 22
5.4 13
0
3
0
0
18.2 9
0
0
35.4 1
5.3
0
20.4 1
21.1 0
0
3.4
0
0
0
0
1
0
16 1.6
0
0
0
0
0
0
0
0
0
10
0
21
0
20

34.5
0
0
0
12.3
41.4
0
0
8.3
5
0
0
0
25.8
0
1
0
0
6.4
6.2
4
5
1
0
0
0
0
0
0
0
1

0
0
0
0
0
0
6.6
0
0
0
4.5
17
8.4
0
0
0
0
1
0
11.6
0
0
0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
2
5
6

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

estimated precisely while the others with less accuracy, but in


either case the parameter estimates lie within an acceptable
range.

Results and Discussion


The proposed model is not a black box model. It is a
deterministic model as each of its parameters has an exact
geometric or physical meaning. Nevertheless, all the param-

Sep Oct Nov Dec Jan Feb Mar Apr May June July Aug
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
43.2
0
0
0
0
0
0
0
0
0
0
0
0
17.3

13.3 12
0
0
0
0
5.7 14 3.6
0
0 63.3
2
0
11
0
5
16
0
0 60.4 0
5.6
0
0
0 31.6 0
4
0
0
0
0
0
0
0
0
0
0
9
0
0
0
0
0
0
44
0
0
0
0
0
6.4
0
0
0
0
0
7
3
12.4 0
8
0
0 79.7
0
0
10
3
0 77.7
0
0
1
50
0
24
0
0
0
0
0
0
0
0
0
38
0
0
0
0
0
27
0
0
0
0
0
0
0
0
0
00
0
0
14
0
0
0
0
0
43
0
0
0
0
0
14
0
3.8
0
0
0 16.4 6
1.5
0
0
0
0
19
4.8
0
0
0
0
3
0
0
0
0
0
3
20.6 0
0
0
0
2
58.4 0
0
0
0
0
9
6
8.6
1
0
11
0
13
2
9.1
0
0
0
0
0
0
0
0
6.8
0
0
0
0
0
0
7

0
0
16
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

0
0
0
0
0
0
3
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0

eters are effective, i.e., through model fitting, they receive an


average value. However, in our opinion, we can arrive at
some conclusions about the physical system (karstic aquifer)
from the order of magnitude of some of these parameters,
such as the storage coefficient.
In Figure 6 the model calculated hydrograph against
real hydrograph and rainfall are indicated for the years
199597. In Figure 7 the model calculated chloride concentration values are indicated against the real ones. It is impor-

Figure 6. Comparison between Model calculated discharge values and field measurements. AB, DE: Recharge periods, CD,
EF: Depletion periods.

A. Maramathas et al. GROUND WATER 41, no. 5: 608619

613

tant to note that the model calculated chloride concentration


values result from the model on a daily basis, but the field
values from the governmental service for ground water are
not on a daily basis. In Table 8, the standard deviation is
calculated between model calculated values and field measurements for recharge and depletion periods, respectively.
From the comparison between values calculated by the
model and field measurements (Figures 6 and 7) and from
the standard deviation values for recharge and depletion
periods (Table 8), it follows that the agreement between
model values and field measurements is very good for
depletion periods and satisfactory for recharge periods. The
difference in agreement between recharge and depletion
periods is due to:
1.

2.

The existence of only one rain gauge in the recharge


area, the Anogia rain gauge, for the rainfall estimation.
Perhaps the solution to this problem is to take into
account rain gauges that lie on the periphery of the
recharge area.
The use of an invariable infiltration coefficient to calculate the rainfall percentage that recharges the spring
reservoir. As a result, basic natural processes have not
been simulated. Such processes are the subsurface layers humidity saturation before rainfall water begins to
recharge the ground water reservoir, and siphon phenomena in the unsaturated karstic zone. These
processes will be included in future simulations.

In Table 7 the parameter values characterizing the


karstic system, which have been estimated by model fitting,
are shown. All these parameter values from the physical
point of view are acceptable.

Conclusions
With the proposed model it has been possible to simulate a karstic spring even if it gets brackish. The advantages
of the model are:
1.

The simulation of the entire hydrograph and quality


curve against time, not just the corresponding quantities for the depletion period.

2.

3.

4.
5.

It does not require data from fieldwork that is difficult


or impossible to collect. Input of the model is only rainfall.
The results (discharge, water chloride concentration)
are directly comparable to the field measurements that
are usually taken.
It offers the ability to make inferences about the nature
of a simulated karstic system.
It provides simplicity.

Regarding the Almiros spring that has been simulated


by the proposed model, the conclusions at this stage are:
1.
2.

It seems that the aforementioned mechanism causing


the water to become brackish has been confirmed.
The aquifer of the Ionian zone limestone is probably
confined. This was inferred from the value of the storage coefficient that was extracted by the model fitting.

In future work, after spring simulation for even more


years, additional conclusions will be presented about the
nature of the karstic system of Psiloritis, and proposals will
be suggested for the improvement of the water quality of
this significant spring.

Acknowledgments
The authors would like to thank Editor Mary P. Anderson, who handled the review, Ken Hardcastle, and two
anonymous reviewers for their valuable suggestions.

References
Angelini, P., and W. Dragoni. 1997. Problem of modeling limestone springs: The case of Bagnara (North Apennines, Italy).
Ground Water 35, no. 5: 612618.
Bezes, C. 1976. Contribution a la modelisation des systemes
aquiferes karstiques; establissement du modele BEMER;
son application a 4 systemes karstiques du Midi de la
France. These 3e cycle. Universite de Montpellier, Memoires du CERGH.
Breznic, M. 1973. The origin of brackish karstic springs and their
development. Col. Razpr. In Por. 16 Knjig, 83186.
Castany, G. 1967. Traite Pratique des Eaux Souterraines. Paris:
Dunod.

Figure 7. Comparison between Model calculated water chloride concentration values and field measurements. AB, EF:
Recharge periods, CD, GH: Depletion periods.

614

A. Maramathas et al. GROUND WATER 41, no. 5: 608619

Table 5
Discharge 19951997 (m3/sec)
Sep Oct Nov Dec Jan Feb Mar Apr May June July Aug
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

4.0

3.7

3.5

4.2
4.5
3.7
3.5
5.6
5.3
4.7
5.0
5.0
5.7
7.6
7.1
6.4
5.8
7.8
12.8
10.0
7.6
6.1
5.0
4.5
4.2
4.0
3.7
3.6
3.5

5.4
15.1
14.0
17.4
22.4
20.9
15.9
14.0
11.9
14.8
13.7
12.9
12.3
14.4
16.8
19.9
21.2
21.1
20.6
18.0
15.5
14.6
15.7
16.6
15.1
16.6
20.7
20.5
14.6
18.0
18.4

24.4
25.0
28.3
23.4
20.9
20.7
19.7
20.3
22.1
24.2
24.8
23.0
19.9
19.3
17.2
16.1
15.1
14.2
13.5
12.7
11.9
11.7
11.2
10.8
10.3
8.8
8.3
8.1
8.6

22.5
21.1
17.2
17.1
17.0
23.4
22.3
18.4
16.8
16.5
15.9
14.6
16.1
18.0
20.3
19.9
17.8
15.9
14.0
13.1
12.9
12.9
12.6
12.3
11.7
11.0
10.6
10.6
10.5
9.8
9.3

9.1
8.8
8.7
8.5
8.3
8.1
7.8
7.6
7.4
7.4
7.4
7.5
7.6
7.8
7.8
7.8
7.8
7.6
7.4
7.4
7.4
7.4
7.2
7.2
7.2
7.1
7.1
7.1
7.1
7.1

7.1
7.1
7.0
6.9
6.9
6.9
6.8
6.8
6.8
6.8
6.8
6.8
6.8
6.8
6.7
6.6
6.5
6.4
6.4
6.4
6.4
6.4
6.4
6.4
6.4
6.4
6.3
6.2
6.1
6.0
5.6

5.6

5.0

4.5

Sep Oct Nov Dec Jan Feb Mar Apr May June July Aug
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

4.0

5.2
7.9
5.6
5.0
4.5
4.1
4.0

4.5
6.6
6.1
5.3
4.8
4.5
4.2
4.0
4.0
4.0
3.8
3.8
3.8
3.8
3.8
3.8
3.8
3.8
3.8
3.8
3.8
3.8
3.8
3.8
3.8
3.8
3.8
3.8
3.8
3.8

3.8
7.8
15.7
18.5
27.8
21.7
20.9
14.8
10.3
9.5
8.8
8.6
8.6
8.3
8.2
7.6
6.9
6.6
6.6
6.1
5.8
5.6
5.3
5.0
5.0
5.0
5.0
19.5
32.4
24.4
17.6

15.1
11.6
11.3
10.0
9.3
8.8
7.9
7.9
7.9
7.8
7.6
8.8
14.0
15.1
25.6
32.4
25.9
21.7
15.4
14.0
12.7
11.7
10.5
9.5
9.0
8.6
8.3
7.9
7.6
7.4
7.1

6.9
6.9
6.9
6.9
6.9
6.8
6.4
13.5
16.5
18.4
16.3
13.7
12.3
11.7
11.5
11.3
11.0
13.3
19.5
19.5
20.7
18.4
16.1
14.4
13.5
13.1
12.8
12.8

Jan

Feb Mar APR May June July Aug

13.1
23.2
40.3
34.3
30.6
22.4
18.4
16.1
14.2
13.7
25.6
50.2
52.1
34.4
35.6
30.4
31.6
28.0
23.0
20.3
18.6
17.6
16.6
15.8
15.0
13.8
13.6
13.5
13.5
13.3
13.3

13.3
12.9
12.2
11.7
11.7
12.3
13.2
13.0
12.7
12.5
12.5
12.4
12.2
12.0
11.7
11.0
10.5
9.9
9.7
9.4
9.2
9.2
9.1
8.9
8.6
8.4
8.2
8.0
7.8
7.7

7.6
7.6
7.5
7.3
7.1
6.9
6.8
6.7
6.6
6.6
6.6
6.5
6.5
6.4
6.4
6.3
6.3
6.3
6.2
6.2
6.2
6.2
6.2
6.2
6.2
6.2
6.2
6.2
6.2
6.2
6.2

6.1

5.9

5.3

Table 6
Chloride Concentration 19951997 (mm)
Sep
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

Oct

Nov Dec

Jan

Feb Mar Apr May June July Aug

791
159
4792

852

4940

2130
5076

4757
71

3603
2378
142
4573

71
1100
4704

3248
1917
5083

722
4792

603
2875

Sep
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31

Oct

Nov Dec

3585
3658
2502

576

3674
1331

284
3639
4250

124

4045

781

4508

35

3816

3514

2511
2982
2236
497
4440

142
923

1065

A. Maramathas et al. GROUND WATER 41, no. 5: 608619

615

Ghassemi, F., A.J. Jakeman, G. Jacobson, and K.W.F. Howard.


1996. Simulation of seawater intrusion with 2D and 3D
models: Nauru Island case study. Hydrogeology Journal 4,
no. 3, 422.
Maille, E. 1905. Essais dHydraulique Souterraine et Fluviale.
Paris: Hermann.
Mangin, A. 1970. Contribution a letude des aquiferes karstiques
a partir de lanalyse des courbes de decrue et tarissement.
Annales de Speleologie 25, no. 3: 581610.
Palmer, A.N. 2000. Digital modeling of individual solution conduits. In Speleogenesis: Evolution of Karst Aquifers, ed. A.
Klimchouk, D.C. Ford, A.N. Palmer, and W. Dreybrodt.
Huntsville, Alabama: National Speleological Society.
Schoeller, H. 1967. Hydrodynamique dans le karst (ecoulement
et emmagasinement). Chronique d Hydrogeologie 10,
721.
Vidakis, M. 1983. Geological map of Greece 1:50000, Heraklion
sheet. Athens, Greece: Institute of Geology and Mineral
Exploration.
White, F. 1994. Fluid Mechanics, 3rd ed. New York: McGrawHill.
Worthington, S.R.H. 1999. A comprehensive strategy for understanding flow in carbonate aquifers. In Karst Modeling, ed.
A.N. Palmer, M. Palmer, and I.D. Sasovsky, 1729. Charles
Town, West Virginia: Karst Waters Institute.
Worthington, S.R.H., and D.C. Ford. 1997. Borehole tests for
megascale channeling in carbonate aquifers. In Proceedings
of the 12th International Congress of Speleology 11,
195198. La Chaux- de-Fonds, Switzerland.

Table 7
Parameter Estimates by Model Fitting
Tripolis Zone Limestone
Recharge area St (m2)
Percentage of water that feeds
Ionian limestone directly f
Cross section of reservoir outlet tube S (m2)
Energy loss coefficient Lt (m-1  sec-2)
Infiltration coefficient of recharge area
Coefficient of storage s

2.2 3 108
0.1
1
1
0.6
0.0075

Ionian Zone Limestone


Recharge area Sp (m2)
Cross section of reservoir outlet tube S (m2)
Energy loss coefficient Lp (m-1  sec-2)
Infiltration coefficient of recharge area
Coefficient of storage s
Specific yield

3.1 3 108
1
1
0.6
1.25 3 105
0.0195

Sea
Cross section of reservoir outlet tube S (m2)
Pressure loss coefficient of the sea tube
outlet region La (kgr  m-7)
Depth of seawater outflow tube outlet point
H (m)

0.06
10
850

Clemens, T., D. Huckinghaus, R. Liedl, and M. Sauter. 1999.


Simulation of the development of karst aquifers: Role of
epikarst. International Journal of Earth Sciences 88,
157162.
Dreybrodt, W. 1988. Processes in Karst Systems: Physics, Chemistry, and Geology. Berlin: Springer-Verlag.
Dreybrodt, W. 1996. Principles of early development of karst
conduits under natural and man-made conditions revealed
by mathematical analysis of numerical models. Water
Resources Research 32, 29232935.
Eisenlohr, L., M. Bouzelboudjen, L Kiraly, and Y. Rossier. 1997.
Numerical versus statistical modeling of natural response of
a karst hydrogeological system. Journal of Hydrology 202,
244262.
Fitrolakis, S. 1980. The geologic structure of Crete (in Greek).
National Technical Department of Engineering and Metallurgy. University of Athens, Athens, Greece.
Ford, D.C., S.E. Lauritzen, and R.O. Ewers. 2000. Hardware and
software modeling of initial conduit development in karst
rocks. In Speleogenesis: Evolution of Karst Aquifers, ed. A.
Klimchouk, D.C. Ford, A.N. Palmer, and W. Dreybrodt.
Huntsville, Alabama: National Speleological Society.
Forkasiewicz, J., and M. Paloc. 1967. Le regime de tarissement
de la Foux de la Vis. Chronique d Hydrogeologie 10,
5973.
Gabrovsek, F., and W. Dreybrodt. 2000. A model of early evolution of karst aquifers in limestone in the dimensions of
length and depth. Journal of Hydrology 240, 206224.

Appendix A: Mathematical Analysis of the Model


Karstic Subsystems Equations (Figure 3)
Continuity equation (White 1994):
d
dV 1 " (un)dS 5 0
dt 9
OE

(A1)

EE

where is the density of the fluid, V is the fluid volume in


the reservoir, u is the velocity vector of fluid on the inlet or
outlet surfaces of the reservoir, n is the unit vector normal
to the inlet or outlet surfaces of the reservoir, S is the inlet or
outlet surface actually part of the control surface through
which mass and mechanical energy are exchanged between
the control volume and the environment outside, OE
denotes the control volume, and EE the control surface. The
first term expresses the rate of change of mass in the control
volume. The second expresses mass flux from the control

Table 8
Standard Deviation Between Model Calculated Values and Field Measurements
Recharge
Period
19951996

Depletion
Period
19961997

Depletion
Period
19951996

Depletion
Period
19961997

Total
19951997

Discharge (m3/sec)

3.37

6.24

0.38

0.51

3.57

Chloride
concentration (ppm)

1035

1356

200

309

1160

616

A. Maramathas et al. GROUND WATER 41, no. 5: 608619

volume through the control surface because of the flow. In


the case of unconfined aquifers, two fluids (water and air)
are contained in the control volume, but only the water was
taken into account in the calculations because air mass is
negligible compared to water mass. In view of the aforementioned, the first term of Equation A1 becomes

sS1
or

sS1

d
d
d
rdV 5 9 rd 1 sVs 2 5 3 srwS1dH (A2)
dt 9
dt
dt
OE

OE

where s is the storage coefficient of the aquifer, Vs the saturated volume of the rock, w the fresh water density, S1 the
transverse inlet or outlet surface of the reservoir, and H the
level of water in the reservoir (Figure 3). If the aquifer is
unconfined, level (H) corresponds to the mean level of
water in the entire aquifer. In the case of confined aquifers,
the variable H does not have a direct physical meaning, but
if this variable is multiplied with the coefficient s/, where
is the specific yield, it corresponds to the mean water
level of an equivalent unconfined aquifer, which has the
same volume of water.
The second term of Equation A1 becomes

.
dH
1 Q2 2 gPS1 5 0
dt

(A4)

In this last equation, Q2 is the outflow discharge of the


reservoir, is the infiltration coefficient, and P the rate of
rainfall. P is measured by rain gauges, the parameters s, S1,
and are estimated by the model fitting, while H and Q2 are
variables.
Mechanical energy equation (White 1994):

u2
u2
p
d
a
a
1
gzbrdV
1
" 2 1 gz 1 b
dt 9 2
EE

OE

.
(u  n)dS 1 Sw L 5 0

" u  n)dS 5 6 w (u1  n1 )dS1

(A5)

S1

EE

1 6 w (u2 ? n 2 )dS2
S2

For the meaning of indices 1 and 2, see Figure 3. The


direction of water movement is from the outside to the
inside through surface S1 and from the inside to the outside
through surface S2. Thus u  n 1 5 2u1 and u2  n 2 5 u2,
where u1, u2 are the water velocity components vertical to
surfaces S1 and S2. In this way, the previous equation can be
written as
"

dH
1 u2S2 2 u1S1 5 0
dt

u2
d
a 1 gzbdV
9
dt
2

(u  n) dS 52 w U1dS1 1 rwu2dS2
6
6
S1

EE

where g is gravity, z the distance from the reference level,


and p the pressure on the inlet or the outlet surface. The first
term expresses the rate of change of mechanical energy in
the control volume, the second represents the flux of
mechanical energy and pressure work through the control
surface, and the third denotes the rate of energy loss in the
interior of the control volume because of friction.
Assuming that the change of kinetic energy, which is
contained within the control volume, is negligible compared to the change in the potential energy, the first term of
Equation A5 becomes

OE

S2

It has been assumed that the water density is


unchanged and that the distribution of the water velocity on
surfaces S1 and S2 is uniform, so that this last equation is
converted to
" (u  n)dS 5 w u2 S2 2 w u1 S1
EE

From Equations A1, A2, and A3 it follows that


H

d
d
5
gzdV 5
sgHwS1dH
9
dt
dt 3
0

OE
H

d
csgwS1 3 HdHd
dt
0

(A3)

5 sgwS1

1 dH
dH
5 sgwS1H
2 dt
dt

(A6)

The second term of Equation A5, on the other hand,


becomes (Figure 3)

d
sr S dH 1 rwu2S2 2 rwu1S1 5 0
dt 3 w 1
0

And upon division by rw:

u2
p
a
" 2 1 gz 1 b(u  n)dS

EE

A. Maramathas et al. GROUND WATER 41, no. 5: 608619

617

u21
p
5 6a
gz1 1 1 b w (u1  n1 )dS1

2
w

sS1H

dH
u2
w
1 2 Q2 2 HQ1 1 L Q2 5 0
g
dt
2g

s1

u2
P2
1 6 a 2 1 gz2 1 bw (u2  n2 )dS2
w
2

Considering turbulent flow, energy loss because of friction


has been taken proportional to water velocity squared so
that

S2

u2
P
u2
5 6 a 2 1 gz2 1 2 u2 dS2 26 a 1 1 gz1
w
2
2
S2

S1

It has been assumed that:


1. Kinetic energy and pressure work of rainfall water are
negligible compared to its potential energy.
2. z2 = 0.
3. Pressure work of the water in the reservoir is negligible.
On the basis of these assumptions, the previous formula becomes

sS1H

dH
u2
1 2 Q2 2 HQ1 1 Lu2Q2 5 0
dt
2g

where L is a coefficient, which depends on geometry and


asperity of rock tubes, and uis the mean water velocity in
the reservoir.
Q
It has been assumed that approximately u 5 2 and
sS1
thus

p
u22
u2
" a 2 1 gz 1 b(un)dA 5 6 2 w u2dS2
s2

sS1H

dH
u2
Q 2
1 2 Q2 2 HQ1 1 L c 2 d Q2 5 0
dt
2g
sS1

EE

or
2 6 gz1wu1dS1

sS1H

s2

Since uniform flow characteristics on the inlet or outlet


surfaces have been assumed,
p
u2
" a 2 1 gz 1 b(un)dA

#
dH
Q 32
Q2 2
1
d Q2 5 0 (A9)
2 2 HPS1 1 L c
dt
2gS2
sS1

#
Q2
,Q1 5 PS1. The parameters S2 and L are
S2
estimated by the fitting of the model.
where u2 5

Equations for the Sea (Figure 4)

EE

u2
u2
5 w 2 u2S2 2 gz1wu1S1 5 w 2 Q2 2 gz1wQ1
2
2

Mechanical energy equation:


d
u2
a 1 gzbdV
9
dt
2

(A7)

OE

And from Equations A4, A5, and A6 it follows that

p
u2
.
1 " a 1 gz 1 b(u  n)dS 1 wL 5 0

dH u2
#
#
sgwS1H +w 2 Q2 2 gz1wQ1+wL 1 wS 5 0 (A8)
dt
2
#
z1 5 H,wL 5 wLQ2

EE

Assuming steady flow,


d
u2
a 1 gzbdV 5 0
9
dt
2

where wL is the energy loss per moving water mass unit


because of friction.
Thus, Equation 7 becomes
sgwS1H

dH u22
+ Q 2 gHwQ1+wLwQ2 = 0
dt w 2 2

After suitable manipulations,

A. Maramathas et al. GROUND WATER 41, no. 5: 608619

(A10)

OE

The second term of Equation A4 becomes (Figure 4)


u2
p
u23
a
1
gz
1
b(u

n)dS
5
a
" 2
6 2 1 gz3

EE

618

(A5)

S3

After suitable manipulations,

P
u2
P
1 3 b(u3  n3 )dS3 1 6 a 4 1 gz4 1 4 b
a
S4 2

Q 23
p
1 3 2 H 5 0
g
2gS23

(A12)

a (u4  n4 )dS4

where a is the density of rainfall water and is the density


of seawater, if the sea feeds the spring, or the density of the
reservoir water if the sea is fed by the spring.
Assumptions:
1.
2.
3.

u4, p4 negligible
z3 = 0
Uniform flow distributions on the inlet-outlet surfaces.

On the basis of these assumptions, the previous equation becomes

p
u2
" a 2 1 gz 1 b(u  n)dA
EE

5a

u23 p3
1 bQ3 1 gHaQ4

2

(A11)

where u3S3 5 Q3, u4S4 5 Q4 and z4 5 H (Figure 4).


From Equations A4, A8, and A9 it follows that
a

u23
p
#
#
1 3 bQ3 1 gHaQ4 1 wL 1 wS 5 0

2

#
It has been assumed that wL 5 0. Therefore,
a

u23
p
1 3 bQ3 1 gHaQ4 5 0

2

Since the flow is steady aQ4 + Q3 = 0 and thus


a

u23
p
1 3 b 2 gH 5 0

2

The parameters S3 and H are estimated by the model


fitting, is taken from related references, while Q3 is a variable.
Pressure p3 is calculated by the formula
p3 5 wg(H 1 h 1 Heq ) 2 LQ22

(A13)

where w is the density of the water of the karstic system


reservoir (fresh water density), h the elevation of the spring,
and L coefficient that depends on the geometric characteristics of surface 3 region (Figure 4).
The first term expresses the water pressure on surface 3,
while the second one expresses the pressure loss, which
caused the water velocity and water energy loss on surface 3.
Considering turbulent flow, energy loss is taken proportional to the square at the water velocity or, in other words,
proportional to the square of the water discharge. The equivalent water level Heq over the discharge point is given by the
formula sH 5 Heq. In the case of unconfined aquifers
where s 5 , it will be Heq 5 H, whereas in the case of
s
confined aquifers it will be Heq 5 H. Actually, confined

aquifers are emulated by unconfined aquifers with the same
volume of water. On the basis of the aforementioned, Equation A13 becomes
s
p3 5 wg aHq 1 h 1 Hb 2 LaQ 22

(A14)

p3 is a variable. The parameter L is estimated by the model


fitting while w and h are taken from references related to
the spring.
Equations A4, A9, A12, and A14 constitute a system
that has to be solved. This system is presented in Table 1.
Actually, the equations are six since the spring, as previously stated, is fed by two karstic subsystems, which are
represented by the reservoirs karst 1 and karst 2 (Figure 2). Thus, the karstic system equations are four indeed,
two for the reservoir karst 1 and two for the reservoir
karst 2. These two pairs of equations are similar in form.
The only differences between them are the different values
of the parameters.

p
u23
1 3 b Q3 1 gH aQ4 5 0

Since the flow is steady aQ4 + Q3 = 0 and thus:


a

u23
p
1 3 b 2 gHq 5 0

where according to the water movement. If seawater goes


to the reservoir, then = 1, otherwise = +1.
A. Maramathas et al. GROUND WATER 41, no. 5: 608619

619

You might also like