You are on page 1of 15

This article was downloaded by: [78.93.44.

206]
On: 21 June 2015, At: 01:05
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Hydrological Sciences Journal


Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/thsj20

Optimization of groundwater
abstraction from a coastal aquifer
a b
E. H. MOREL & A. B. BIRTLES
a
Institute of Hydrology , Wallingford , Oxfordshire , OX10 8BB ,
UK
b
Central Water Planning Unit , Reading , Berkshire , RG1 8PS ,
UK
Published online: 24 Dec 2009.

To cite this article: E. H. MOREL & A. B. BIRTLES (1983) Optimization of groundwater


abstraction from a coastal aquifer, Hydrological Sciences Journal, 28:1, 169-182, DOI:
10.1080/02626668309491150

To link to this article: http://dx.doi.org/10.1080/02626668309491150

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms
& Conditions of access and use can be found at http://www.tandfonline.com/page/
terms-and-conditions
Hydrological Sciences -••• Journal — des Sciences Hydrologiques, 28, 1, 3/1983

Optimization of groundwater abstraction from


a coastal aquifer

E.H. MOREL*
Institute of Hydrology, Wallingford,
Oxfordshire OXlO 8BB, UK
A.B. BIRTLES
Formerly at: Central Water Planning Unit,
Reading, Berkshire RG1 8PS, UK

ABSTRACT An analytical mathematical model, based on


Jacobian elliptic functions, has been used to identify
feasible wellfield locations and pumping rates for large-
Downloaded by [78.93.44.206] at 01:06 21 June 2015

scale abstraction from an unconfined coastal aquifer.


The choice of optimum feasible wellfield strategy has
been made using a simple economic model which calculates
the cost of the pipelines required to transport the
abstracted groundwater to a large coastal city which
forms the demand centre. Results indicate that the
cheapest wellfield design would be a single large well-
field. However, because of the need to maintain at least
a minimum supply in the city until a new surface water
source is developed, a better solution may well be to
develop two smaller wellfields pumping a greater total
abstraction.

Optimisation du prélèvement d'eau souterraine à partir


d'un aquifère cotier
RESUME Un modèle mathématique analytique, basé sur les
fonctions elliptiques de Jacob a été utilisé pour ident-
ifier des zones favorables d'implantation de puits et les
taux de pompage pour des prélèvements très importants
dans un aquifère cotier libre. Le choix d'une stratégie
pour l'implantation optimale de la zone des puits a été
fait en utilisant un modèle économique simple qui calcule
la prix des conduites nécessaires pour amener l'eau
souterraine provenant des puits à une grande ville litt-
orale qui constitue le centre de la demande. Les
résultats montrent que les dispositions les plus écono-
miques pour l'ensemble des puits serait une seule zone
étendue d'implantation. Cependant par suite de la
nécessité de maintenir au moins un apport d'eau minimal
dans la ville jusqu'à ce que une nouvelle source d'eau
de surface soit aménagée, une solution meilleure pouvait
bien consister à aménager deux zones de puits moins
étendues pompant un débit total plus élevé.

*Now at: Burmah Oil Exploration Ltd, Pipers Way, Swindon,


Wiltshire SN3 IRE, UK.
169
170 E.H. Morel & A.B. Birtles

INTRODUCTION
Economic resources for water supply development are often limited
in developing countries. The optimum choice of groundwater
pumping regimes depends not only on the hydrogeological restrict-
ions on drawdown and water quality, but also on the cost of
providing the water at the demand centre. This means that the
best development scheme in arid areas is not necessarily the one
which maximizes the safe yield of the aquifer.
The present study aims to identify possible locations for new
wellfields which would minimize the cost of water supplied to a
large city located on an unconfined coastal aquifer. Demand for
water is expected to grow very rapidly in the next 15 years, and
far exceed the capacity of the small existing wellfield. The new
wellfields will be required to meet the extremely rapid growth in
demand for water in the city, and in this study potential wellfield
locations have been optimized for steady state abstractions of at
Downloaded by [78.93.44.206] at 01:06 21 June 2015

least 75% of the safe yield of the aquifer.


The hydrogeological data available indicate that the aquifer is
largely homogeneous and is in hydraulic contact with the sea.
Natural recharge to the aquifer is derived solely from a large
river which runs approximately parallel to the shoreline. Demand
for water in the study area is entirely from the coastal city, and
will be met by groundwater abstraction from the proposed wellfields
until a new source (probably river water piped directly to the city)
can be developed. This is unlikely to be available until 1990, and
the requirement to provide water of acceptable quality implies a
need to limit sea water intrusion into the aquifer by maintaining
a sufficient outward flow of fresh groundwater at the coast. The
development area lies between the river and the sea, and is shown
schematically in Fig.l.
Groundwater flow in homogeneous aquifers with simple geometry

. array of wells
roh
— o ...
Specified

Sea

h-0

Fig. 1 Idealized model of groundwater system.


Optimization of groundwater extraction from a coastal aquifer 171

may be analysed using analytical methods. The model proposed in


this study contains relatively few parameters. These include the
bulk transmissivity, the groundwater recharge, the proposed
wellfield abstractions and the aquifer geometry. The actual values
of some of these parameters are not accurately known, and the model
has been used to determine amongst other things the sensitivity
of the computed optimum pumping regime to variations in input para-
meters .

CONSTRAINTS ON GROUNDWATER DEVELOPMENT


The safe yield of an aquifer is here defined as the average rate
of groundwater abstraction which can be sustained without causing
an undesirable deterioration of water quality. In the study area
the safe yield of the aquifer depends primarily on the long term
groundwater recharge from the river. However, the need to protect
Downloaded by [78.93.44.206] at 01:06 21 June 2015

the groundwater supply from contamination by non-potable water


means that the optimum location of wellfields is a compromise
between sites located close to the coast, where saline intrusion
is a potential problem, and sites near the river where groundwater
drawdowns are the limiting factor for reasons discussed below.
The quality of groundwater in the sand aquifer is generally
good although the quality deteriorates at the coast where sea water
has intruded into the lower part of the aquifer. If fresh water,
overlying the salt water wedge, is pumped excessively, saline water
will contaminate the supply. Potential sites for the additional
wellfields have therefore been restricted to locations inland of
the computed position of the saline wedge. A relatively simple
mathematical description has been used to calculate the position
of the boundary between the fresh and salt water. This is describ-
ed by Todd (1959), and gives the steady state position of the
wedge :

fp s _ p f Kb 2 ]
A= 2 (1)
h ^ QTJ
where A is the length of the wedge; Qf is the freshwater flow per
unit length of interface; b is the saturated aquifer thickness and
K .is the permeability; p s and pf are the densities of saline and
fresh water respectively.
This simple description is justified because there is a lack of
detailed field information in the study area which would be
necessary to calibrate more sophisticated models of the saline
front. The maximum acceptable intrusion of the saline wedge has
been assumed to be 5 km and this figure yields a value of 1.75 m
for the minimum allowable fresh water head at the toe of the
wedge. Management schemes which would lead to lower heads have
been rejected because they would probably cause excessive sea water
intrusion into the aquifer.
The river is not actually in direct hydraulic contact with the
main groundwater body in the aquifer. Water leaks through the
river bed and seeps downwards through unsaturated aquifer before
reaching the water table. The quality of the groundwater deter-
172 E.H. Morel & A . B . Birtles

iorates to the north of the river where the permeability of the


aquifer is considerably reduced. It is possible that over-pumping
in the study area could eventually lead to contamination of the
aquifer if the groundwater gradient were reversed to the north of
the recharge mound caused by the river. The optimization proced-
ure used to determine optimum wellfield locations has therefore
been constrained to limit drawdowns in this area.

ANALYTICAL GROUNDWATER MODEL


Natural groundwater level variations in the study area and likely
drawdowns are small relative to the saturated thickness of the
aquifer. The equation describing steady state flow in the aquifer
is therefore given by

T V2 h (2)
Downloaded by [78.93.44.206] at 01:06 21 June 2015

where q = QJJJ - Q; the transmissivity of the aquifer (T) is assumed


to be uniform over the area; QJJJ is the rate of infiltration and Q
is the rate of abstraction. Solutions to equation (2) depend on
the boundary conditions specified at the edges of the model shown
in Fig.l. No flow boundaries have been imposed at each end of the
model because little is known of the aquifer potential away from
the study area, and it would be imprudent to depend on an, as yet,
unproven resource. Any subsequent development of these unexplored
areas would justify the approximate no flow boundary condition on
an image basis.
The steady state groundwater head within the model aquifer may
be easily computed in the absence of pumping as groundwater levels
decline linearly towards the coast.

TN V (3)
h(x) - x
T 12
where QJJ^ is the infiltration/unit length along the river. The
model shown in Fig.l may be simplified by drawing axes of symmetry
parallel to the x-axis through each wellfield, and halfway between
wellfields. The basic element of the groundwater model is then as
shown in Fig.2.
Full abstraction from each wellfield is Q 0 . Therefore abstract-
ion from the area shown in Fig.2 is Q 0 /2. If we abstract a prop-
ortion f of the input from the river distributed equally between
n wellfields, then

Coast
h=0
bz T
half abstraction image
from each wellfield wellfield
y=o

Fig. 2 Model element.


Optimization of groundwater extraction from a coastal aquifer 173

Q0 = 2u f Q I N (4)

where u = L/2n, and L is the width of the model shown in Fig.l.


The remaining fraction (1 - f) of the input from the river
flows to the sea, and is required to limit saline intrusion into
the aquifer at the coast. The maximum value of f depends on the
length of the saline wedge calculated from equation (1), which may
now be written

(ps - p f ) Tb
(5)
2 " ~~ "
Pf (1 - f)QiN

Substitution of local data in equation (5) leads to a maximum value


of f = 0.74 for the study area.
The fixed head condition at the coast in the basic element shown
Downloaded by [78.93.44.206] at 01:06 21 June 2015

in Fig.2 can be replaced mathematically using the theory of images


by an equivalent recharge located at a distance (v/2 - z) away from
the coast. An analytical solution to the steady state groundwater
equations can then be obtained as follows using conformai trans-
formation.
The solution for the complex drawdown X is

X = s + i ijJ (6)
2 2
' sn (w.k) - sn (w-, ,k) )
log,
°
2 ITT » csn (w,k)
- - sn^Cwg.k)
- (7)

where sn(w,k) is a Jacobian elliptic function. C is an arbitary


constant which will be chosen to make the drawdown at the coast
equal to zero:

w = z/p and z = x + iy (8)

where p is a scaling factor given by

v u
P = K = K- <9>

and

K/K' = v/u (10)

Similarly

w
2 = Z
2//P and w
l = Z
i/P (I1)

where

On the x-axis both z, and z„ are purely real so


174 E.H. Morel & A.B. Birtles

and Zo = x 2 (12)

K and K' are the quarter periods of sn(w,k) in the x and y directions
respectively. k is called the modulus, and is defined by the ratio
K/K'. This ratio can be expressed in the form

q = exp (-TFK'/K) (13)

where the parameter q is known as the nome. The value of k can be


calculated from product series. For example, Copson (1935) shows
that

k = 4q 2 H U + q Zn )/(l + q 2 n )} (14)

This formula is difficult to calculate numerically when the value of


Downloaded by [78.93.44.206] at 01:06 21 June 2015

k is nearly equal to unity; then the following formula is more


suitable for use in digital computers:

k' = n^ = {(1 - q 2 n - 1 )/(l + q 2 n _ 1 ) } (15)

where

k' = 1 - k = m-L (16)

The value of the elliptic integral K is most easily calculated using


polynomial approximations. The following approximate formula, given
by Abramowitz & Stegun (1965), may be used to evaluate K(m) to an
accuracy of eight decimal places.

K(m) = (a Q + aj^m-^ + . . . a 4 mj) + (b Q + b-jm-^ + . . . b 4 m 1 )

* log e (l/ m i ) (17)

where
2
m = k (18)

and

a Q = 1.386 294 361 12 b 0 = 0.5


a x = 0.096 634 425 90 bj = 0.124 985 935 97
a 2 = 0.035 900 923 83 b 2 = 0.068 802 485 76
(19)
a 3 = 0.037 425 637 13 b 3 = 0.033 283 553 46

a4 = 0.014 511 962 12 b 4 = 0.004 417 870 12

To find C in equation (7), we note that the groundwater drawdown is


zero at the coast (z = v/2) so that
Optimization of groundwater extraction from a coastal aquifer 175

2 2
sn (w 3 ,k) - sn (w 2 ,k) (20)
c = 2 2
sn (w3,k) - sn (w^,k)
where

v/2 k (21)
w _ 2
3 p
Equation (7) may now be simplified using various identities given,
for example, by Abramowitz & Stegun (1965), so

cn(w - w 2 ) cn(w + w 2 ) s n ^ - w2J sn(j + Woj


27TT Y (22)
— X = - loge cn + W
sn(w - w 2 ) sn(w + w 2 ) en [— - w 2 j ["ô 2^J
Downloaded by [78.93.44.206] at 01:06 21 June 2015

where the notation sn(w,k) = sn(w) etc. has been used for simplicity.
On any line parallel to the coast greatest drawdowns in the
aquifer will always occur in the x-axis since this is the shortest
distance to the source of recharge at the river. We can therefore
confine our investigation to the real (x) axis, so

JL3 cn&Jl3 sn£


2TTTS P 2 p (23)
log- p p 2
Qo rx - x2i rx ^ c n ( | Î2)cn(| + ^
sn * sn — P J

The value of the drawdown at the river (s ) is calculated from


equation (23) using x = 0, and the value of the drawdown near the
coast is given by using x = v/2 - A, where A is the length of the
saline wedge.
We are interested in the drawdowns at two places in the elemental
model shown in Fig.2, (a) at the river, and (b) adjacent to (some
specified distance from) the coast.

Drawdown at the river


Drawdowns at the river (s 0 ) are obtained from equation (23). Figure 3
show the computed drawdown at the river for a single line of well-
fields pumping 60% of the total recharge (f = 0.6). The predicted
drawdown depends mainly on the distance of the wellfields from the
river ( x 2 ) , and decreases towards the coast. The predicted draw-
down is very similar over a wide range of wellfield size (for n > 2) ,
although a single large wellfield (n = 1) may be expected to draw
down water levels about 10% more than a line of smaller wellfields.
The maximum acceptable drawdown at the river is indicated on Fig.3,
and permissible wellfield locations are those in the unshaded region.
Permissible values of x 2 are x 2 > 0.28v for two or more wellfields,
and X2 > 0.30v for a single large wellfield.
176 E.H. Morel & A.B. Birtles
6,4 3
10-^U 2 1 : Numbers of wellfields

Maximum
«_ acceptable
drawdown

0.4 0.5
RIVER X^/V" CBAST

Wellfield position
Fig. 3 Drawdown at river (x = 0) for wellfields pumping a total of 60% of the
Downloaded by [78.93.44.206] at 01:06 21 June 2015

groundwater recharge.

Groundwater head near the coast


The groundwater head (h ) at a distance d from the coast may be
computed by superposition of the head calculated in the absence of
pumping from equation (4), and the drawdown obtained by putting
x = v/2 - d in equation (23).
Figure 4 shows the computed groundwater head inland from the
coast (x = 0.41v) again assuming 60% of the input is abstracted
(f = 0.6). The predicted water levels vary considerably for
different values of n, and a single large wellfield can be expected
to lower groundwater levels near the coast much more than a line
of several smaller wellfields pumping the same total yield.
The minimum acceptable groundwater head at this point is indicated
in Fig.4. Lower levels would lead to excessive saline intrusion.
The constraint on the drawdown at the river has also been transposed
on to Fig.4, and permissible wellfield locations are those in the
unshaded area of Fig.4, for 0.30v < 0.41v. This shows that,
at a total abstraction of f = 0.6, a single large wellfield is not

Minimum
«— acceptable
head

.1 0.2 0.3 0.4' 0.5


RIVER >L/V C8RST

Wellfield position
Fig. 4 Groundwater head near the coast (x : 0.41 v) for wellfields pumping a total
of 60% of the groundwater recharge.
Optimization of groundwater extraction from a coastal aquifer 177

a feasible management strategy since either one or both hydrogeolog-


ical constraints would be exceeded.

ECONOMIC MODEL
The optimum wellfield location is that which provides the cheapest
water subject to the hydrogeological constraints concerning saline
intrusion at the coast and the groundwater drawdown at the river.
These constraints limit the possible locations to a relatively
narrow strip parallel to the coast, as shown in Fig.4.
For economic and engineering reasons it is desirable to have a
uniform rate of abstraction from each production borehole, and the
adoption of a uniform pumping rate is encouraged by experience of
an existing wellfield. Regional variations in borehole construction
costs and running costs are likely to be small in the feasible area
of groundwater development, and have therefore been assumed to be
Downloaded by [78.93.44.206] at 01:06 21 June 2015

the same for all boreholes in this study.


Pipelines are required to connect the production boreholes to a
distribution centre in each wellfield, and to transport the water
from each distribution centre to the city on the coast which is the
centre of demand. The costs of pipelines within each wellfield
are assumed to be similar and variations small in comparison with
overall pipeline costs. In fact it turns out that the major diff-
erences in the capital costs of groundwater development schemes in
the study area are due to the costs of the pipelines which transport
the water from each wellfield to the city. Schemes involving
several small wellfields require either longer total length of
small-diameter pipelines or more complicated distribution systems
than schemes involving only one or two large wellfields. The costs
of these pipelines have been estimated from local data and are
given below:

Pipe internal diameter (mm) 250 300 400 500 600


Cost per metre 1.0 1.1 1.5 2.0 2.3
Extra for improving access 0.2 0.2 0.2 0.2 0.2

Savings in pipeline costs are possible if the pipeline route follows


a road. The data shown above indicate that this saving is large for
small-diameter pipelines, but is insignificant compared to the cost
of large-diameter pipelines.
The diameter of each pipeline has been calculated on the basis
of head loss, and the capacities given below are based on a velocity
of 1 m s .

Pipe internal diameter (mm) 250 300 400 500 600


1 3 _1
Flow rate at 1 m s " (m h ) 177 254 452 707 1018

Three different pipeline strategies have been investigated in this


study. These are illustrated in Fig.5 for a value of n = 4.
Figure 5(a) shows individual wellfields independently connected
directly to the city. This scheme would be relatively expensive,
178 E.H. Morel & A.B. Birtles

Fig. 5 Pipeline strategies: (a) independent pipelines; (b) wellfield output piped to a
single distribution centre; (c) pipelines along existing roads.

particularly at large values of n, but has the advantage that each


wellfield would supply the city independently of the state of repair
of the others. In Fig.5(b) the output of each wellfield is piped
to a single distribution centre and thence to the city. This would
involve the building of fewer access roads and would enable larger
pipe diameters to be used in the central portion of the area. In
Fig.5(c) use is made of the two existing roads in the area.
Downloaded by [78.93.44.206] at 01:06 21 June 2015

Figure 6 shows the computed cheapest pipeline combinations for


wellfields pumping a total of 60% of the recharge (f = 0.6). The
two hydrogeological constraint lines have been transposed from
Figs 3 and 4, and the region of permissible solutions is shown
unshaded. The optimum wellfield location (for a value of f = 0.6)
is the cheapest feasible solution, and is for two wellfields each
producing f/2 at a distance of X2/V = 0.34 from the river.

Excessive drawdown
at coast

° 40.0

.1 0.2 0.3 0.4 0.5


C8BST

Wellfield position
Fig. 6 Computed cost of pipelines for wellfield pumping a total of 60% of the
groundwater recharge.

The cheapest pipeline system would be for two independent well-


fields, each with their own access roads to the city. The total
cost of pipelines (including access roads) is calculated to be 5.0
units. This figure is only slightly less than the computed cost of
constructing pipelines via the existing roads (5.1 units). This
small difference (2%) is certainly less than the accuracy with which
the economic model is able to predict the actual pipeline costs in
the study area. A severe cost penalty would, however, be incurred
if a single distribution centre were to be adopted as in Fig.5(b)
since the total computed cost of this scheme would be 12.2 units.
The computed optimum wellfield locations (x2/v) o p ^ are shown in
Optimization of groundwater extraction from a coastal aquifer 179

Fig.7 for wellfields pumping different proportions (f) of the re-


charge to the aquifer. At low abstraction rates only one wellfield
would be necessary, and the optimum position of the wellfield is
determined solely by the need to limit sea water intrusion at the
coast. The optimum position moves inland from the coast (x2/v = 0.5)
as the abstraction rate is increased up to a maximum of f = 0.54
when a single wellfield located at X2/V = 0.23 would provide the
cheapest water to the city.
More than one wellfield would be necessary for requirements
greater than f = 0.54 since the drawdown at the river would also
become a limiting factor. Two wellfields, each located at x„/v =
0.31, would provide the cheapest water for total abstraction in
the range of f = 0.54 to f = 0.66. These wellfields would be located
L/2 km apart and at x2/v = 0.26 from the city so the total pipeline
costs would be approximately double the costs of the pipeline
required for a single wellfield pumping f = 0.54.
Downloaded by [78.93.44.206] at 01:06 21 June 2015

oH—1—1—1—1—1 1 1 1 1 1
0 -1 2 3 4 -5 -55 -6 -65 7 74
Abstraction _ , ï
Recharge Maximum safe yield I

Fig. 7 Cost of pipelines.

For any particular value of n, the position of the constraint


concerning sea water intrusion (the right hand constraint line in
Fig.6) does not vary very rapidly with changes in f, so the computed
optimum wellfield location ( x 2 / v ) o p t m °ves only gradually away from
the coast as the abstraction is increased. However, for the case of
two wellfields, the resulting increased pipeline costs are more
than compensated by the increases in the available abstraction, and
the unit cost of water decreases up to f = 0.66. Fairly large
increases in cost, with only small increases in available yield,
may be expected for three or more wellfields, so the optimum well-
field design is for one or two wellfields (i.e. f < 0.66).
Although a single wellfield would be able to provide about f =
0.54 with low unit costs, the best solution may well be for two
wellfields pumping a total of f = 0 . 6 6 even though this would be
more expensive. This is because of the extremely high growth rate
that has been projected for the demand for water in the city and
the need to provide at least a minimum supply until a new supply can
be installed.
180 E.H. Morel & A.B. Birtles

COMPARISON WITH NUMERICAL MODEL


The computed optimum wellfield location may be compared with that
obtained using a distributed parameter finite difference ground-
water model similar to that described by Larson et al. (1977). The
latter model was calibrated from borehole water level measurements
using an inverse method similar to that proposed by Smith & Piper
(1978). The model used technological functions (Maddock, 1972) and
linear programming (see e.g. Dantzig, 1963) to compute the optimum
locations for wellfield development along the existing roads, subject
to the same hydrogeological constraints as this study. The results
of the numerical model indicated that the maximum abstraction that
could be sustained was f = 0.63, and that this should be obtained
from two new wellfields, one on the western road at X2/V = 0.27,
and another on the eastern road at x2/v = 0.18, supplemented by
reduced pumping from the existing wellfield (x2/v = 0.36).
The optimum wellfield parameters calculated using the analytical
Downloaded by [78.93.44.206] at 01:06 21 June 2015

and numerical methods are as follows:

Analytical solution Numerical solution


f 0.66 0.63
x2/v 0.31 0.27

It is interesting to note that the resolution of the finite diff-


erence model is limited by the size of the model grid blocks, and
the extent of the wellfield grid block chosen by the optimization
procedure is actually from x2/v = 0.23 to x2/v = 0.32. The analy-
tical solution x2/v = 0.31 lies in this range, thus giving confidence
in the analytical method proposed in this study.
The numerical and analytical models described in this study comp-
lement each other. The numerical model calculates the areal distri-
bution of groundwater levels as well as the hydrogeologically opti-
mum wellfield positions and pumping rates. The analytical model
calculates groundwater levels only at a single specified position
due to pumping at a predetermined rate from a chosen potential well-
field location. However, the speed at which the analytical comput-
ations can be done is very fast, and all the calculations required
to draw Figs 3 and 4 took less than 1 min CPU time and 5 k words
storage for execution on a Univac 1108 computer. By contrast a
single run on the numerical model required approximately 10 min
CPU time and 35 k words storage, and repeated simulation runs to
determine optimum pumping rates from different wellfield locations
would have been prohibitively expensive. The calculations of the
Jacobian elliptic functions were done using computer codes given
by Bulirsch (1965).

SENSITIVITY OF ANALYTICAL MODEL


The sensitivity of hydrogeologically feasible solutions to uncert-
ainty in the bulk transmissivity (T), input recharge (QJN)» and
the abstraction from each wellfield (Q 0 ) may be easily established
since the drawdowns calculated using the equation (23) are in
Optimization of groundwater extraction from a coastal aquifer 181

normalized units. The y-value of the horizontal line in Fig.3


representing the maximum permissible drawdown at the river is given
in normalized units by

2TTT s„ 2TTT S
o m a x = - — — omax (24)
fLQ
nQ 0 iN
The effects of changes in these parameters, either separately or
combined, are really to move the positions of the constraint lines
in Figs 3 and 4 relative to the other graphs.
Although there is considerable uncertainty in the actual values
of T and Q j ^ , the ratio of QJJJ/T is equal to the regional water
level gradient which is known from the measurements made in 30
boreholes in the study area. These observations were made in a
single season, and do not necessarily coincide with the long term
average values. Subsequent well records suggest that the uncert-
ainty in the ratio of Q J N / T is about 10%.
Downloaded by [78.93.44.206] at 01:06 21 June 2015

The major uncertainty in equation (24) is the value of the re-


charge QJN- The results of this study are based on a particular
design recharge value and indicate that the optimum wellfield design
is for two wellfields each pumping f = 0.66 at location X2/V = 0.31.
It is, however, possible that the average recharge is as much as
20% larger than the design recharge (in which case f would equal
0.53), or as little as 25% smaller than the design recharge (f = 0.88)
Fortunately, the variation of the optimum wellfield location
(x 2 /v) 0 p t with changes in f is small within these limits as is shown
in Fig.7, so that the maximum possible cost penalty incurred through
the incorrect choice of f is likely to be relatively small. This
is largely due to the fact that the maximum available pipeline
diameter (600 mm) would be required under all three recharge cond-
it ions.
In practice, the response of the water table to abstraction from
the proposed wellfields would give the best indication of the actual
value of QIJJ and f, and additional boreholes may be constructed if
the recharge is found to exceed the design recharge. If the re-
charge is less then abstraction from the wellfields will have to be
reduced to keep the water levels within the hydrogeological constr-
aints, and the additional water supplies which the city will un-
doubtedly need will have to be made available sooner than at present
anticipated.

CONCLUSIONS
An analytical model, based on Jacobian elliptic functions, has been
used to identify feasible wellfield locations and pumping rates for
large-scale groundwater abstraction from an unconfined coastal
aquifer. The requirement to provide water of an acceptable quality
implies a need to limit the extent of saline intrusion at the coast,
and to limit drawdowns in the vicinity of'the river. Permissible
wellfield locations are restricted to bands which run parallel to
the coast.
The close similarity between the optimum solutions obtained from
the simple groundwater model based on Jacobian elliptic functions
182 E.H. Morel & A.B. Birtles

described in this study and the results of the more complicated


numerical model used in a previous study gives confidence in the
use of the Jacobian elliptic functions to simulate groundwater
conditions in the aquifer. However, a major advantage of the
method described in this paper is that the Jacobian elliptic
functions may be readily computed using codes described in the
literature (Bulirsch, 1965) which make only small demands on
computer facilities.
The choice of the optimum feasible wellfield development
strategy has been made using a simple economic model which cal-
culates the cost of the pipelines required to transport the abstrac-
ted groundwater to the demand centre. The cheapest wellfield design
would be for a single wellfield pumping 54% of the aquifer's re-
charge at a distance of x2/v = 0 . 2 3 from the coast. However,
because of the need to provide at least a minimum supply to the
city until a new supply can be installed, a better solution may
well be for two wellfields supplying a total of 66% of the recharge
Downloaded by [78.93.44.206] at 01:06 21 June 2015

even though this would be more expensive. These wellfields would


be located closer to the coast at X2/v = 0.31. The optimum wellfield
location has been found to be relatively insensitive to the actual
amount of the annual average recharge to the aquifer.

ACKNOWLEDGEMENTS The work described in this paper was undertaken


at the Institute of Hydrology. The views expressed are not neces-
sarily those of the Institute.

REFERENCES
Abramowitz, M. & Stegun, I.A. (1965) Handbook of Mathematical
Functions with Formulas, Graphs and Mathematical Tables. Dover
Publications.
Bulirsch, R. (1965) Numerical calculation of elliptic integrals
and elliptic functions. Numerische Math.ema.tik 7, 75-90.
Copson, E.T. (1935) An Introduction to the.Theory of Functions
of a Complex Variable, 381-419. Clarendon Press.
Dantzig, G.B. (1963) Linear Programming and Extensions. Princeton
University Press, Princeton, New Jersey, USA.
Larson, S.P., Maddock, T. & Papadopulos, S.S. (1977) Optimisation
techniques applied to groundwater development. Mémoires IAH
Birmingham Congress, UK, XIII, Part 1 Papers, E57-66.
Maddock, T. III. (1972) Algebraic technological function from a
simulation model. Water Resour. Res. 8, 129-134.
Smith, P.J. & Piper, B.S. (1978) A non-linear optimisation method
for the estimation of aquifer parameters. J. Hydrol. 39, 255-271.
Todd, D.K. (1959) Ground Water Hydrology, p.281. John Wiley, New
York.

Received 7 July 1982.

You might also like