Sei sulla pagina 1di 11

A delineation of regional hydraulic conductivity based on water table uctuation

Eungyu Park
a,
, Kue-Young Kim
b
, Guoping Ding
c
, Kangjoo Kim
d
, Weon Shik Han
e
,
Yeongkyoo Kim
a
, Namjin Kim
a
a
Department of Geology, Kyungpook National University, Daegu, South Korea
b
Geologic Environment Div., Korea Institute of Geoscience and Mineral Resources, Daejeon, South Korea
c
Department of Water Resources and Hydrogeology, School of Environmental Studies, China University of Geosciences, Wuhan 430074, PR China
d
School of Civil and Environmental Engineering, Kunsan National University, Kunsan, South Korea
e
Energy and Geoscience Institute, University of Utah, Salt Lake City, UT, USA
a r t i c l e i n f o
Article history:
Received 1 June 2010
Received in revised form 5 December 2010
Accepted 4 January 2011
Available online 8 January 2011
This manuscript was handled by P. Baveye
Editor-in-Chief
Keywords:
Hydraulic conductivity
Water table uctuation
Regional scale
Recharge
Jeju Island
s u m m a r y
The estimation of regional hydraulic conductivity is essential for the sustainable development and ef-
cient management of groundwater. This paper presents a new physical method of estimating hydraulic
conductivity based on the water level uctuation model developed by Park and Parker (Park, E., Parker,
J.C., 2008. A simple model for water table uctuations in response to precipitation. J. Hydrol. 356, 344
349). The developed procedure was validated by comparison with the hypothetical numerical model
for a variety of hydraulic conductivities, aquifer inclinations, and the heterogeneities. The model was
found to be highly reliable when using water level data far from the hydraulic boundaries. The results
show that the monitoring data obtained over a range of relative distances of an observation location
(x) to the ow domain (W), x/W > 0.2 avoided the boundary effects. A correction factor was also proposed
based on the deviation analysis of the calibrated K and f/n (ratio of recharge to precipitation divided by
porosity). The results from the basal inclinations and the heterogeneities cases suggest that the estimated
hydraulic conductivity is representing regional property rather than the local one. To conrm the validity
of the proposed procedure, it was applied to the coastal aquifer of Jeju Island, Korea. The calibrated
hydraulic conductivities were consistent with the results from previous studies. These results show that
the proposed method is an effective and economical means to estimate hydraulic conductivity of large
supporting volumes of subsurface aquifers with the convenient use of recharge and groundwater level
data.
2011 Elsevier B.V. All rights reserved.
1. Introduction
Regional scale hydrogeologic characterization of the subsurface
is essential for the sustainable development and efcient manage-
ment of groundwater resources. Conventional hydraulic tests and
their subsidiary modications have served as useful tools for sub-
surface hydraulic characterizations for several decades (e.g., Theis,
1935; Hantush, 1964; Boulton and Streltsova, 1975; Bear, 1979;
Butler, 1990; Ramey, 1992). Among the existing methods, pumping
tests (e.g., Theis, 1935; Cooper and Jacob, 1946; Neuman, 1972,
1974; Schad and Teutsch, 1994; Moench, 1995; Sa nchez-Vila
et al., 1999) and slug tests (e.g., Hvorslev, 1951; Cooper et al.,
1967; Bouwer and Rice, 1976; Zlotnik and McGuire, 1998; Audouin
and Bodin, 2007) are most frequently applied for these purposes.
The goals of those methods are to delineate localized values of
hydrogeologic parameters through a model calibration based on
stress-response time series.
In general, the supporting volume of the conventional hydraulic
tests is largely variable to the properties of the surrounding aquifer
media and is limited by hydrologic boundaries. For this reason, the
hydraulic conductivity values acquired from the conventional
methods are often not regionally representative for inhomoge-
neous aquifers in terms of its properties and structures. Also, for
an aquifer with extremely high hydraulic conductivity or with a
large body of surface water, the imposed stress by pumping or
injection is not enough to induce adequate response of the aquifer
and the application of conventional tests is not appropriate for re-
gional scale characterizations.
Aquifer hydraulic conductivities in extreme conditions have
been previously reported. The range of estimates for the horizontal
hydraulic conductivity of volcanic rocks in Oahu is about
4 10
3
2 10
2
ms
1
(%3451728 md
1
) (Mink, 1980; Mink
and Lau, 1980; Souza and Voss, 1987). Jussel et al. (1994) described
openwork gravels in Switzerland as having a geometric mean
hydraulic conductivity of 8640 md
1
. Pang et al. (1998) reported
hydraulic conductivity values ranging from 5064 to 14,112 md
1
in an alluvial gravel aquifer. As the conventional pumping test
0022-1694/$ - see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhydrol.2011.01.002

Corresponding author. Tel.: +82 53 950 5356; fax: +82 53 950 5362.
E-mail address: egpark@knu.ac.kr (E. Park).
Journal of Hydrology 399 (2011) 235245
Contents lists available at ScienceDirect
Journal of Hydrology
j our nal homepage: www. el sevi er. com/ l ocat e/ j hydrol
usually fails to assess the hydraulic properties in extremely perme-
able aquifers, the demand for alternative devices is increasing. On
the other hand, in coastal aquifers where persisting extraction of
groundwater with heavy rates may cause salt-water intrusion and
endanger freshwater resources, the applicability of conventional
methods for delineating regional hydraulic properties is, therefore,
limited.
Information on subsurface hydraulics may be inferred from nat-
urally occurring stresses such as precipitation, evapotranspiration,
and tidal effects. Camp Dresser and McKee Inc. (1993) employed ti-
dal analyses for the estimation of the extreme hydraulic conductiv-
ity in a coastal area of Oahu, Hawaii. Hamm et al. (2005) delineated
hydraulic conductivities based on the localized specic capacity.
Dann et al. (2008) combined pumping and tracer test data to derive
equivalent average hydraulic conductivities for each test in a het-
erogeneous channelized aquifer system.
Recently, Park and Parker (2008) proposed a differential equa-
tion for water table uctuation that converges to an equation for
unconned groundwater ow through a nite difference approxi-
mation. The model is physically based, and the effect of recharge
and groundwater ow through the aquifer can be addressed in a
simple semi-analytical solution. However, delayed drainage from
an unsaturated zone, sink or source and recharge due to precipita-
tion are not considered as potential causes of groundwater level
uctuation. The application of the model, therefore, is limited to
several favorable conditions such as relatively thin or a very high
unsaturated hydraulic conductivity of the aquifers.
In order to use naturally induced stresses, the amount of the
stress must be clearly dened, or the range of the stresses needs
to be delineated with relatively small uncertainty. Furthermore,
to evaluate regional hydraulic properties, the areas affected by
the stresses have to be large enough to cover the region of concern.
In this respect, the best candidate parameter might be the recharge
from precipitation.
The effect of recharge is generally represented by water table
uctuations, conditioned on the hydraulic properties of saturated
and unsaturated subsurface media. Therefore, careful observation
of the resultant groundwater level uctuation due to precipitation
provides information on the hydraulic system. In this regard,
numerous researchers have investigated the water table uctua-
tions in response to the rate of groundwater recharge since the
1950s (Rasmussen and Andreasen, 1959; Sophocleous, 1991; Rai
and Singh, 1995; Bierkens, 1998; Rai and Manglik, 1999; Knotters
and Bierkens, 2000; Coulibaly et al., 2001; Healy and Cook, 2002;
Rai et al., 2006). In order to extract meaningful information from
water table uctuations, a physically based water table uctuation
model needs to be developed. Among the existing models, the early
prototype of Rasmussen and Andreasen (1959) has been frequently
applied. However, this model is unable to address the discharge
through the aquifers during the recharge event.
In this study, the water table uctuation model developed by
Park and Parker (2008) was modied to t the purpose of the anal-
ysis (i.e., a large scale model), and a procedure was proposed for
the regional hydraulic conductivity delineation based on the mod-
el. The developed procedure was rst applied to a numerical case
study to validate the methodology. The procedure was also applied
to the eastern coastal aquifer of Jeju Island, Korea, which is well
equipped for validation in terms of hydrologic facilities and struc-
tural characterizations. Jeju Island, composed of volcanic rocks, is
reported to have particularly high permeable layers (Hahn et al.,
1997; Hamm et al., 2005). The location for the application of the
model was carefully chosen from the hydrograph of a series of
wells located at the aquifer, in which relatively quick recharge
and groundwater uctuations have been observed. Jeju Island is
a volcanic island in which aquifers are composed of lithologies that
are highly permeable. A few pumping tests carried out in this area
showed less than 10 cm water level declination with the large
pumping rate of 1000 m
3
d
1
. These observations imply that con-
ventional pumping tests are not well-suited for many of the loca-
tions due to extreme hydraulic conductivity. The time from
recharge to precipitation generally takes less than a day, judging
from the observation data. The delineated mean hydraulic conduc-
tivities and their range were compared to the statistics, which
were directly measured from the vicinity or inferred by other
means.
2. Model development
The groundwater storage above minimum groundwater level
can be dened as discharge potential (Park and Parker, 2008), as
this volume of water has the potential to be discharged to the near-
by discharge area. As dened by Park and Parker (2008), the dis-
charge potential is given by the following equation of
v nAh 1
where v is the discharge potential (L
3
), n is the porosity () above
the minimum water level, A is the representative area (L
2
), and h
is the discharge head (L), which is dened as HH
min
, where H is
the actual groundwater elevation (L) and H
min
is the minimum
groundwater level (L) relative to a reference elevation (e.g. mean
sea level, MSL).
The groundwater ow direction is assumed to be from left-to-
right, as shown in Fig. 1, and the ow is assumed to be horizontally
parallel and symmetric. Therefore, the problem can be reduced to a
two-dimensional problem. In this gure, the x-axis is in the right-
to-left direction and the origin of the coordinate system corre-
sponds to the RHS boundary. The z-axis is along the vertical direc-
tion from bottom to top, and the origin is at the bottom of the
submerged portion of the aquifer. Within a certain interval with
a length D (L) along the direction of ow, the temporal change in
the discharge potential can be written as the following water bud-
get equation (Park and Parker, 2008) as
Fig. 1. (a) Conceptual model of the problem addressed in this study, and (b)
schematic diagram of an inclined basement aquifer.
236 E. Park et al. / Journal of Hydrology 399 (2011) 235245
dv
dt
Q
OUT
Q
IN
DG 2
where Q
IN
is the inbound and Q
OUT
is the outbound groundwater
ow rate (L
3
T
1
) through the saturated portion of the aquifer,
and G is the unit recharge rate (L T
1
) imposed on the top boundary
that is given as the product of recharge to precipitation ratio, f(),
and precipitation rate, P (L T
1
).
As in Park and Parker (2008), the storage change due to ground-
water ow is assumed to be proportional to the discharge potential
as
Q
OUT
Q
IN
kv 3
where k is a rate coefcient (T
1
) related to the decay characteristics
of the discharge potential or discharge head with time. The theoret-
ical relationship of this parameter to hydraulic parameters has been
derived in previous reports (i.e., Eq. (8) of Park and Parker, 2008);
however, the relationship cannot be directly applied to delineate
hydraulic parameters inversely due to the ambiguity of the param-
eters used in the development of the relationship. A practical rela-
tionship of this rate coefcient to the actual aquifer hydraulic
parameter will be discussed in the following context of this study.
The groundwater inow and outow rates entering and exiting
the domain can be expressed as the following equations by apply-
ing Darcys law as
Q
OUT
Kb h
1

dh
dx

1
4
and
Q
IN
Kb h
3

dh
dx

3
5
where (dh/dx)
1
and (dh/dx)
3
are the local hydraulic gradients at
locations x
1
and x
3
, respectively, which may be given as a function
of the relative location in the domain; h
1
and h
3
are the head at loca-
tions x
1
and x
3
, respectively, which may be given as a function of the
relative location; b is the saturated thickness of the aquifer, mea-
sured from the bottom of the aquifer to the minimum water level
of the domain; and K is the hydraulic conductivity of the domain.
For an ideally homogeneous and isotropic case, the steady-state
solution of Eqs. (4), (5), with mean recharge rate C (L T
1
), is de-
rived as
h b W

b
W

2

C
K
x
W

2
2
C
K
x
W

6
with boundary conditions of h(0) = 0 and (dh/dx|
x=W
) = 0, where Q
IN
or Q
OUT
is equated to the integrated recharge rate, such that
Q =

Cdx. A few detail on the derivation of Eq. (6) is summarized


in Appendix A. W is the length of the entire domain (L) (Fig. 1). In
the development of Eq. (6), the bottom elevation of b and the
mean recharge rate C are assumed to be constant. Combining Eq.
(3) with Eqs. (4), (5) becomes
K b h
3

dh
dx

3
b h
1

dh
dx

1

kv 7
In the above equation, k is a representative head decay coef-
cient that is inversely acquired from water table uctuation model
(Park and Parker, 2008) (PP model hereafter) at the center of x
1
and
x
3
(i.e., x
2
). The model equation used for the estimation of k by Park
and Parker (2008) is
h
i1
h
i
expkDt
i

fP
i
expkDt
i
1
kn
8
where h
i
indicates the head at the current time-step and h
i+1
is the
head at the next time-step, which is to be computed through the
equation; P
i
is the precipitation rate at the next time-step from
piece-wise-constant function; and Dt
i
is the length of the current
time-step. Eq. (8) is a piece-wise solution and requires numerical
computation. More detailed explanations of the model and the der-
ivation procedure for the equation are given in Park and Parker
(2008).
It is also assumed that under high hydraulic conductivity of the
aquifer, the hydraulic head change with distance is close to linear,
and the head at x
2
represents the average head of h
1
and h
3
. With
this assumption, we can compose an equation for discharge poten-
tial as
v n

x
3
x
1
hndn % nx
3
x
1
h
2
9
where h
2
is the head at a point centered between h
1
and h
3
. In the
case of low hydraulic conductivity, therefore, the interval between
x
1
and x
3
needs to be small to have a linear head. This means that
when the hydraulic conductivity is small, the support volume of
the head is also small, and the estimated hydraulic properties rep-
resent local characteristics of the aquifer. Based on Eq. (7) and Eq.
(9), one can compose
K
knx
3
x
1
h
2
b h
1

dh
dx

1
b h
3

dh
dx

3
10
In Eq. (10), the denominator of the right-hand-side can be re-
duced using Eq. (6) and its derivative as
b h
1

dh
dx

1
b h
3

dh
dx

3

C
K
x
3
x
1
11
The numerator of the right-hand-side in Eq. (10) can be approxi-
mated based on the identity of
lim
b!0
a

a
2
b

% b=2a;
as knx
3
x
1
h
2
%
knC
bK
x
3
x
1
Wx
2

x
2
2
2

12
by considering a relatively small value of C/K. Plugging Eqs. (11),
(12) into Eq. (10), the relationship between the decay coefcient,
k, and the hydraulic conductivity, K, is given by a simple equation,
K
knx
b
W
x
2

13
where x(W x/2) has a gradual increasing pattern within the model
domain as the saturated thickness of the aquifer. As can be found
from Eq. (13), the head decay parameter is not affected by an exter-
nal source or sink such as recharge rate, and only depends on the
transmissivity of the aquifer such as the saturated thickness and
hydraulic conductivity. A similar relationship has been given by
Park and Parker (2008), Eq. (8)), but the equation is rather imprac-
tical for application to the estimation of hydraulic conductivity from
the decay coefcient, due to ambiguities in the equation. It can be
also found from Eq. (13) that the uncertainty in the hydraulic con-
ductivity estimation is proportional to the recharge to precipitation
ratio, f. The correct information on f is crucial for the accuracy of the
estimation from Eq. (13). The estimation of f for a particular eld
site is highly variable. In general, this variability is site and scale
dependent, and the uncertainty is higher for the small scale domain
where the lump scale mass balance approach is unavailable. There-
fore, the domain scale and the uncertainty in the estimated f need to
be seriously considered when one would apply Eq. (13) to a site to
delineate the regional hydraulic conductivity.
E. Park et al. / Journal of Hydrology 399 (2011) 235245 237
3. Numerical conrmation
3.1. Estimation procedure
To conrm the relationship between hydraulic conductivity, K,
and the head decay coefcient, k, as in Eq. (10), the estimation abil-
ity of the equation was inversely tested based on numerical simu-
lations. In this procedure, three different processes were used to
generate data for water level uctuation, calibration, and estima-
tion of hydraulic conductivity or transmissivity.
A numerical model was developed for the transient water level
generating process using nite element (FE) method in COMSOL
Multiphysics

(COMSOL AB, 2005a,b). All the modeling parameters


were assumed to be fully-known in this process. The groundwater
uctuations at different locations and at various times were
acquired from the modeling results, and they were assumed to
be actual observations made from real aquifers.
In the calibration process, the head decay coefcient (k hereaf-
ter) and the recharge to precipitation ratio with porosity as a single
parameter (f/n hereafter) in Eq. (8), were calibrated for each uctu-
ation dataset, based on the model and numerical code previously
developed by Park and Parker (2008). Only the daily precipitation
rate was assumed to be known as a real situation in the calibration
process. The FORTRAN code coupled with PP model and the non-
linear least squares optimization algorithm (Hanson, 1986) was
developed to enable unknown model parameters to be estimated
from modeled water levels and precipitation rates.
In the estimation process, the hydraulic conductivity of the
aquifer was estimated on the basis of the calibrated k. In this pro-
cess, the aquifer scale along the direction of ow and the relative
locations of the observation points were assumed to be known.
These scale parameters are easily delineated fromgeographic maps
in real applications. The porosity of the aquifer was also assumed
to be known; this could have served as a source of uncertainty in
the estimation of hydraulic conductivity. The aquifer thickness be-
low the minimum water level was also assumed to be uniform and
known a priori. The thickness of the groundwater conducting aqui-
fer could also serve as a source of uncertainty, thus detailed char-
acterization of the aquifer geometry is necessary for the actual
application. However, this is not necessary if the aim is to estimate
transmissivity. To further improve the development, cases of basal
inclination and heterogeneities of the aquifer are also developed
and tested.
3.2. Estimation from a synthetic case
3.2.1. Effect of the magnitude of hydraulic conductivity
The conceptual model is identical to that of the analytical devel-
opment as shown in Fig. 1. In the model, the left-hand-side is no
ow boundary, and the right-hand-side is specied head boundary
with value set to 0 m. The lower boundary has no ow condition
and the upper boundary has a specied ux boundary, for which
the recharge rate was 40% of the precipitation rate (Fig. 1a).
Table 1 shows the parameters used in the development of the
numerical model. In the model, the length of the aquifer along
the direction of ow (W) was 5000 m, the minimum water level as-
signed as a specied head at the RHS boundary was 0 m, and the
thickness of the aquifer below the minimum water level was
100 m. The number of elements used was 6544 where the average
element length was 20 m. To acquire the aquifer response to a tem-
porally varying recharge rate, the daily precipitation rate of the Jeju
area of Korea was used (Fig. 2a). In the Jeju area, more than 70% of
precipitation happens during MayAugust, with about 40% of the
rain falling during Aug. In the modeling, the steady-state head dis-
tribution was generated using the average daily precipitations, and
the transient modeling was followed based on the daily precipita-
tion data.
As a result of the numerical simulations, three different tran-
sient water level records over time (for K = 100, 500, and
1000 md
1
) were acquired, and a portion of the transient proles
at specic locations were plotted in Fig. 2b. The water levels shown
in the records indicate a gradual increase from the left-hand-side
(LHS hereafter) to the right-hand-side (RHS hereafter) boundary,
since the RHS boundary forces the head to remain constant, at zero
elevation. On the other hand, the interference of the LHS boundary
was much weaker compared to that of the RHS boundary. As a no
ow boundary, forcing a horizontal water level at the boundary
might reduce the maximum water level that could be simulated
at the LHS boundary. The effective range, however, was conned
around the boundary. The water level distribution generated by
the simulation can be roughly estimated by Eq. (6), which means
Table 1
Parameters used for the development of model cases.
Common parameters Homogeneous and at basement
Case 1 Case 2 Case 3
W = 5000 m K (md
1
) 100 500 1000
b = 100 m Homogeneous and inclined basement
n = 0.25 Case 1 Case 2
h(0) = 0 m Db/W () 0.01 0.02
@h
@x

xW
0 md
1
Heterogeneous and at basement
Case 1 Case 2 Case 3 Case 4
r
lnK
() 0.5 1 2 4
Fig. 2. (a) Daily precipitation rate used in both models, and (b) Water level
uctuation obtained from a numerical modeling (Case 2, solid line) and calibrated
water level obtained through the simple water level uctuation model of Park and
Parker (2008) (circles) at x = 4000, 3000, 2500, 2000, and 1000 m (from top to
bottom) from the RHS boundary.
238 E. Park et al. / Journal of Hydrology 399 (2011) 235245
that the hydraulic conductivity of the aquifer can be inversely com-
puted from Eq. (6) once the recharge rate of the area is known.
However, this estimation is only feasible when large numbers of
monitoring wells are well deployed along the direction of ground-
water ow. Fig. 2b shows the simulated transient heads for
365 days at observation points x = 1000, 2000, 2500, 3000, and
4000 m. The peak pattern generated by numerical modeling (cir-
cles) was generally higher and wider with the distance from the
RHS boundary, and became shorter and narrower near the RHS
boundary, which is also attributed to the effect of the specied
head at the boundary.
The model tness (numerical modeling results and PP model)
was generally good for all the studied cases, however the RMS val-
ues in Table 2 exceed 0.066 m at K = 100 md
1
and x/W = 0.2. The
PP model showed excellent tness to the simulated water table
uctuations near the center of the domain, and showed slight
untness near the boundaries (especially at the specied head
boundary). Table 2 shows the calibrated model parameters of the
PP model. The general pattern of the head decay coefcient, k, in-
creased from the LHS to the RHS boundary, and f/n decreased from
the LHS to the RHS boundary. Both curves showed a non-linear
pattern. Higher hydraulic conductivity generally has a higher cali-
brated k value and a larger calibrated f/n value for the same obser-
vation locations.
As expected from a good model t, the estimated hydraulic con-
ductivity from Eq. (10) had less than 10% error for the observation
locations at x/W > 0.2 from the RHS boundary. The small errors
indicate that the used assumptions in the development process
of Eq. (13) are more or less admissible. However, the estimated
hydraulic conductivity below the range, which was relatively close
to the RHS boundary, had a larger deviation of more than 10% of
the actual value. This larger estimation error near boundaries can
be attributed to the boundary effect, which forces both the level
of the water table and the amplitude of the uctuation to be small
and the resultant hydraulic conductivity from Eq. (13) may be
overestimated. The larger range for the higher hydraulic conduc-
tivity case indicates that the effect of the boundary condition has
a larger inuence on the domain when the hydraulic conductivity
increases, and the estimated hydraulic conductivity farther inside
domain is a more reliable choice for representing the aquifer
hydraulic property. Near the constant head boundary the head de-
cay coefcient k, and hence the hydraulic conductivity obtained by
the PP model, was low compared to the numerical modeling re-
sults, indicating severe model deviation near x/W$ 0. Also, the
boundary effect transferred easily into the inland area with higher
hydraulic conductivity, and the zone with an estimation error lar-
ger than 10% became wider. The estimated f/n had a similar devi-
ation pattern to that of the estimated hydraulic conductivity, in
which the calibrated values at x/W > 0.5 were the closest to the ac-
tual value used in numerical modeling, 1.6 (=0.4/0.25), while the
estimated value at x/W < 0.2 had a larger deviation of more than
40%. To compare the deviation patterns of the estimated K and
f/n, the normalized deviations of the values were plotted in
Fig. 3, which includes the deviation of calibrated f/n from the input
f/n used in the numerical modeling versus the deviation of the esti-
mated K based on Eq. (13) from the input K used in the numerical
modeling. The patterns showed a clear linear relationship between
the two deviations. The clear linear relationship suggests that the
deviation of f/n can be used as a correction factor in K estimations
to improve the quality of the estimation. This study proposes the
following correction factor of
a
f
n

x
W
$0:5

f
n

cal
14
where (f/n)
x/L$0.5
is the ratio calibrated by the PP model around the
center of the domain, and (f/n)
cal
is the ratio calibrated at each
observation point. For the simulated cases, (f/n)
x/W$0.5
was close
to the actual input f/n, and this value can be used to further improve
the estimation by Eq. (13). Based on the multiplication of Eq. (14) to
Eq. (13), the estimations at each observation location were cor-
rected and tabulated in Table 3, and compared to the uncorrected
estimations. The use of Eq. (14) requires caution, because deviations
may arise from the interference of the strong aquifer heterogeneity
or local variation of actual recharge rate induced by changing thick-
ness of unsaturated zone and surface condition.
The RMS error normalized by the input hydraulic conductivity
for the rst, second, and third cases were 0.17, 0.23, and 0.26 under
the uncorrected estimations, respectively, which had errors of
about 20%. This suggests that the estimation error increases when
the hydraulic conductivity becomes larger if the correction factor is
not applied. These error increases can be attributed to the lack of
Table 2
Calibrated model parameters of the PP model for the base case and the estimations
where K is uncorrected and K
0
is corrected estimation from Eq. (14).
Basecase (K = 100 md
1
, Db/W = 0, r
lnK
= 0)
x (m) b (m) k (d
1
) f/n () K
est
(md
1
) K
0
est
(md
1
) RMSE (m)
4000 100 3.65E03 1.74 109.4 103.1 0.0207
3000 100 3.97E03 1.66 104.3 103.4 0.0127
2500 100 4.43E03 1.64 103.9 103.9 0.0175
2000 100 4.65E03 1.47 93 104 0.0241
1000 100 9.41E03 1.6 105.8 108.1 0.0665
Fig. 3. Percent deviation of estimated hydraulic conductivity versus percent
deviation of calibrated f/n.
Table 3
Calibrated model parameters of the PP model for different K values and the
estimations where K is uncorrected and K
0
is corrected estimation from Eq. (14).
x (m) b (m) k (d
1
) f/n () K
est
(md
1
)
K
0
est
(md
1
)
RMSE
(m)
K = 500 md
1
4000 100 1.84E02 1.77 553.4 498.9 0.0097
3000 100 1.94E02 1.63 508.5 500.0 0.0064
2500 100 2.14E02 1.60 500.9 500.9 0.0079
2000 100 2.12E02 1.35 423.4 501.5 0.008
1000 100 2.48E02 0.89 279.3 503.6 0.0097
K = 1000 md
1
4000 100 3.61E02 1.74 1081.7 997.2 0.0082
3000 100 3.75E02 1.58 984.2 996.9 0.006
2500 100 4.26E02 1.6 999.4 999.4 0.0061
2000 100 5.01E02 1.6 1001.4 1001.4 0.01
1000 100 4.91E02 0.88 551.9 1007.5 0.0052
E. Park et al. / Journal of Hydrology 399 (2011) 235245 239
sensitivity to recharge as well as boundary effect, as described pre-
viously. In cases when the estimation was made for locations
where x/W > 0.2, the normalized RMS error became 0.02, 0.07,
and 0.09, respectively, for the rst, second, and third cases and it
seems that the error is slightly amplied with increased hydraulic
conductivity. These results indicate that when applying Eq. (13) for
the estimation of the aquifer hydraulic conductivity, the water le-
vel uctuation data in locations where x/W = 0.2 or higher gives a
more reliable estimation. The corrected estimations had lower er-
rors, 0.01, 0.02, and 0.02 for the rst, second, and third cases,
respectively.
3.2.2. Effect of inclined basement
To delineate the effect of the inclined aquifer basement, two
new cases are developed with different linear inclinations of 1%
(Db/W = 0.01) and 2% (Db/W = 0.02). As seen in Figs. 1b, the satu-
rated aquifer thicknesses at RHS boundaries of the two different
inclination cases are kept b which is 100 m while the thicknesses
are gradually decreasing from the right to the left with different
rate for each case. The input hydraulic conductivities of the aquifer
are both 100 md
1
for the cases. All the other modeling parameters
are identical to the at basement case as in Table 1.
The numerical modeling results at given monitoring locations
(x = 1000, 2000, 2500, 3000, and 4,000 m) for each inclination case
are tabulated in Table 4. The calibrated k shows gradual increase
whereas the calibrated f/n shows decreasing pattern from the
LHS to RHS boundary for both 1% and 2% inclination cases. In the
inclined basement cases, the downgradient averaged saturated
thicknesses from the observation locations are used as input aqui-
fer thickness of Eq. (13) (Table 4). The usage of the downgradient
averaged thickness of the saturated aquifer for the estimation of
the hydraulic conductivity is reasonable because the downgradient
portion of the aquifer is more responsible to the drainage of the
elevated head at the observation locations. The estimated hydrau-
lic conductivity using Eq. (13) for 1% inclination case is within the
range of 84.7113.6 md
1
where the estimation errors are the
largest at both the observation locations near the RHS and LHS
boundary. The estimation made at the central observation point
(x/W = 0.5) shows the lowest deviation from the input hydraulic
conductivity of 100 md
1
. Using the correction factor in Eq. (14),
the estimated hydraulic conductivities are also improved and the
values are within the range of 104.2 and 107.2 md
1
which estima-
tion errors are all less than 10% of the input hydraulic conductivity.
The estimations from the 2% inclination case shows almost sim-
ilar to those of the 1% inclination case. The estimated hydraulic
conductivities using Eq. (13) are within the range of 72 and
111.6 md
1
which is a bit wider than the 1% inclination case. The
corrected estimations show improvement and the deviations from
the input hydraulic conductivity is at most 4.2% of the input value.
As benchmark cases, blindness of the basal inclination informa-
tion is assumed and the estimations are made at the same loca-
tions with uniform thickness of 100 m. The estimations are also
tabulated in Table 4. For 1% inclination case, the range of the esti-
mations without the correction are from 80.5 to 91.5 md
1
in
which the value nearest to RHS boundary is the smallest and that
in the center observation point is the largest. The estimated
hydraulic conductivity near LHS boundary (x/W = 0.8) is
90.9 md
1
and the value is slightly lower than the central estima-
tion. For 2% inclination case, the range of the estimated hydraulic
conductivities is from 64.8 to 76.3 md
1
. The pattern of the estima-
tions is close to that of 1% inclination case and the highest value is
estimated at the center of the domain (x/W = 0.5) and the lowest is
estimated from the observation point near RHS boundary
(x/W = 0.2). The value at the observation point near LHS boundary
(x/W = 0.8) is 66.9 md
1
. With the correction factor in Eq. (14), the
lowest estimations are made at the observation point near LHS
boundary (x/W = 0.8) for both inclination cases and they are 84
and 62.5 md
1
for 1% and 2% inclination cases, respectively. The
highest estimated hydraulic conductivities are estimated from
the observation point near RHS boundary (x/W = 0.2) and they
are 99.1 and 92.9 md
1
for 1% and 2% inclination cases, respec-
tively. The estimations with the correction factor are generally
increasing from LHS to RHS boundaries (Table 4). The overall esti-
mations indicate that the estimated hydraulic conductivities are
increasing with the degree of basal inclination. Also the estimated
deviation from the input hydraulic conductivity is the lowest at the
central observation locations and the largest at the most downgra-
dient observation point without correction factor. With correction
factor, the deviation is the highest near LHS boundary and the low-
est near RHS boundary.
The results generally indicate that the model proposed in Eq.
(13) is valid for the inclined aquifers with the information of the
aquifer structures. The cases also indicate that the correction factor
in Eq. (14) is useful and generally results in improved estimations.
By estimating the true hydraulic conductivity with acceptable er-
ror even for the inclined basement cases, it can be suggested that
the estimation from the proposed method is more representing
the regional hydraulic conductivity of the aquifer, which could be
severely over- or under-estimated otherwise due to the thinning
or thickening saturated thickness of the aquifers.
3.2.3. Effect of aquifer heterogeneity
The effect of the aquifer heterogeneities are tested using a
single realization of random hydraulic conductivity eld from
unconditional sequential Gaussian simulation (SGS) which uses
ordinary kriging as an interpolator. The number of cells along x-
and y-directions are 250 and 20, respectively, and, therefore, the
size of each cell is 20 20 m. In the simulation, it is assumed that
Table 4
Calibrated model parameters of the PP model for the inclination cases and the estimations where K is uncorrected and K
0
is corrected estimation from Eq. (14). K
00
is the
estimation made from uninformed aquifer thickness (assumed as b = 100 m).
x (m) b (m) k (d
1
) f/n () K
est
(md
1
) K
0
est
(md
1
) K
00
est
(md
1
) RMSE (m)
Db/W = 1%
4000 60 3.03E03 1.73 113.6 107.2 90.9 0.0217
3000 70 3.44E03 1.65 106.1 104.9 90.2 0.0132
2500 75 3.91E03 1.63 104.6 104.6 91.5 0.0207
2000 80 4.19E03 1.46 93.2 104.2 83.9 0.0294
1000 90 7.15E03 1.3 84.7 106.5 80.5 0.0503
Db/W = 2%
4000 20 2.23E03 1.71 111.6 104.2 66.9 0.021
3000 40 2.85E03 1.67 106.7 102.1 74.7 0.0157
2500 50 3.25E03 1.6 101.7 101.7 76.3 0.0253
2000 60 3.62E03 1.43 90.6 101.4 72.5 0.0366
1000 80 5.76E03 1.12 72 103.2 64.8 0.0486
240 E. Park et al. / Journal of Hydrology 399 (2011) 235245
the correlation scales along x- and y-directions are 200 and 10 m,
respectively, and no nugget effect is assumed. From the simulation
data, the arithmetic mean of the hydraulic conductivity is steer to
be 100 md
1
while the variance of the log transformed hydraulic
conductivity (r
lnK
) is controlled to be 0.5, 1, 2, and 4 to become het-
erogeneous cases 14, respectively. Fig. 4a shows the log trans-
formed hydraulic conductivity at r
lnK
= 1 in which the similar
hydraulic conductivity parcels take horizontal layering due to the
input correlation scales. All the other modeling parameters are
identical to that in Table 1.
Using the hydraulic conductivity random eld as model input
parameter, the steady-state numerical modeling is performed with
recharge rate of 0.002 md
1
and the resulting height of the water
table at r
lnK
= 1 is shown in Fig. 4b. In Fig. 4b, the increment of
the head shows undulating pattern due to changing local hydraulic
conductivity until it reaches to a certain height. Using the same
method stated in the previous section, transient observations at gi-
ven locations for each log transformed variance cases are prepared
and calibrated (Table 5). As the previous homogeneous cases, the
calibrated k values show gradual increment while f/n is gradually
decreasing with the distance from RHS boundary. By applying
Eqs. (13) and (14), the hydraulic conductivity of the aquifer is esti-
mated (Table 5). The estimated hydraulic conductivities are close
to the arithmetic mean hydraulic conductivity for the case of
r
lnK
= 0.5, however, the deviation becomes higher when r
lnK
be-
comes larger. For the case with r
lnK
= 0.5, the hydraulic conductiv-
ity estimated from the central observation location is around
85 md
1
. For the cases with r
lnK
= 1, 2, and 4, the estimated
hydraulic conductivities at the central point are 71.3, 50.8, and
25 md
1
, respectively. These values seems that the estimated
hydraulic conductivities becomes farther from the arithmetic
mean of the model domain hydraulic conductivity and closer to
geometric means which are 78.6, 62.4, 40, and 17.3 md
1
for
r
lnK
= 0.5, 1, 2, 4, respectively. Note that the estimated hydraulic
conductivities more close to horizontal harmonic mean of verti-
cally averaged hydraulic conductivity of each cell as in following
equation of
K
est

n
x

nx
i1
1
Kv i
15
where the vertical arithmetic mean of the domain hydraulic con-
ductivity is given by
K
v
i
1
n
y

ny
j1
Ki; j 16
The mixed mean hydraulic conductivities are 87.7, 76.5, 57.4,
and 31.4 md
1
for r
lnK
= 0.5, 1, 2, and 4, respectively. If the mixed
mean hydraulic conductivities are assumed as the desirable esti-
mating value, the maximum deviation of the estimated hydraulic
conductivities are at most 20% except the estimation made at the
observation locations closest to RHS boundary. In the cases, the
maximum deviation happens at r
lnK
= 4 that has the most severe
heterogeneity. The estimated hydraulic conductivities of all cases
at all observation locations but ones near to RHS boundary show
very consistent values. These consistent values indicate that the
estimation based on Eq. (13) is more or less representing regional
Fig. 4. (a) Log transformed heterogeneity eld with variance of 1, and (b) the modeled water level of the domain from a steady-state numerical simulation.
Table 5
Calibrated model parameters of the PP model for the heterogeneous cases and the
estimations where K is uncorrected and K
0
is corrected estimation from Eq. (14).
x (m) b (m) k (d
1
) f/n () K
est
(md
1
)
K
0
est
(md
1
)
RMSE
(m)
r
lnK
= 0.5
4000 100 3.13E03 1.73 94.0 87.3 0.0206
3000 100 3.36E03 1.67 88.3 85.0 0.013
2500 100 3.63E03 1.61 85.0 85.0 0.0152
2000 100 4.13E03 1.45 82.6 91.6 0.0287
1000 100 1.01E02 1.61 113.2 113.2 0.0737
r
lnK
= 1
4000 100 2.64E03 1.72 79.3 74.0 0.0209
3000 100 2.83E03 1.67 74.3 71.3 0.0133
2500 100 3.04E03 1.61 71.3 71.3 0.0148
2000 100 3.57E03 1.45 71.5 79.0 0.0316
1000 100 9.32E03 1.61 104.9 104.7 0.0774
r
lnK
= 2
4000 100 1.85E03 1.70 55.6 53.5 0.0211
3000 100 1.98E03 1.67 52.0 50.8 0.014
2500 100 2.17E03 1.63 50.8 50.8 0.014
2000 100 2.64E03 1.49 52.8 58.0 0.0347
1000 100 4.49E03 1.00 50.5 82.2 0.0566
r
lnK
= 4
4000 100 8.92E04 1.64 26.8 26.9 0.0178
3000 100 9.58E04 1.64 25.1 25.1 0.0153
2500 100 1.07E03 1.64 25.0 25.0 0.0121
2000 100 1.42E03 1.59 28.4 29.3 0.0274
1000 100 4.19E03 1.60 47.1 48.4 0.0852
E. Park et al. / Journal of Hydrology 399 (2011) 235245 241
hydraulic conductivity rather than local value. Therefore, the pro-
posed method may serve as a useful tool of large scale hydraulic
conductivity estimator with that the conventional well hydraulics
tools which usually support small scale.
4. Application to Jeju Island
Jeju Island is located 140 km south of the Korean Peninsula
(Fig. 5). The island was formed through multiple volcanic activities,
and mainly consists of volcanic rocks, including permeable basalts.
The basement underlying the island consists of Cretaceous welded
tuff and granite covered by an unconsolidated formation (the U
Formation), which can be found at a uniform depth of 150 m be-
low MSL throughout the eastern coast. The major aquifers are
formed in hyaloclastite, clinker layers, and lava tubes (Won et al.,
2006).
The annual water budget of Jeju Island is 3.61 billion m
3
year
1
,
of which 1.22 billion m
3
year
1
(33.7%) is estimated to be lost in
the form of evapotranspiration, 0.74 billion m
3
year
1
(20.5%) is
lost to direct runoff, and 1.65 billion m
3
year
1
(45.8%) recharges
the groundwater (Jejudo and Korea Water Resources Corporation,
2003).
Due to the high permeability, especially in the eastern coastal
aquifers of Jeju Island, conventional hydraulic tests such as pump-
ing or slug tests often fail to delineate the meaningful hydraulic
parameters of the aquifers. According to previous studies, the
hydraulic conductivity has been reported to have a wide range of
values and a large amount of uncertainty. Hahn et al. (1997) re-
ported that transmissivity values range from 1.09 m
2
d
1
to
768,000 m
2
d
1
with an average of 29,300 m
2
d
1
1, and K values
range from 1.1 10
3
md
1
to 2120 md
1
with an average of
84.6 md
1
. Won et al. (2005) performed statistical analyses for
the data from 27 groundwater wells located on the eastern coast
of Jeju Island monitoring salt-water intrusion, and suggested trans-
missivity values ranging from731 m
2
d
1
to 44,100 m
2
d
1
, with an
average of 17,726 m
2
d
1
. Hamm et al. (2005) estimated transmis-
sivity based on specic capacities from117 sets of timedrawdown
data. The geometric mean of transmissivity was 88.23 m
2
d
1
, with
a maximum of 6212 m
2
d
1
. The statistics, however, were calcu-
lated frompreviously reported values obtained in wide areas of Jeju
Island, and could be biased because the extreme hydraulic conduc-
tivities cannot be successfully delineated by conventional methods.
Eq. (13), developed for the delineation of a regional hydraulic
conductivity, was applied to delineate the transmissivity of the
eastern coastal aquifers of Jeju Island. It is well known that
the unconsolidated formation (U-formation), composed of
uncemented sand and clay, has a hydraulic characteristic of low
permeability and is considered as the reservoir boundary in Jeju
(Koh, 1997). According to the geologic log of the coastal area in
eastern Jeju, the U-formation was found at a almost uniform depth
of around 150 m below sea level, and extended consistently to
the inland area (Korea Institute of Geosciences and Mineral Re-
sources and Jejudo, 2004). This relatively constant depth of the
impervious layer allows the developed model to be applied using
zero basal inclination assumption of a uniform aquifer thickness.
Four different sets of well systems, the HC, HD, SS, and JD series,
Fig. 5. Model application area and location of monitoring well systems.
Table 6
Statistics of groundwater level obtained from monitoring wells on Jeju Island.
Well name Daily groundwater level (m)
Mean Max Min Range Md STD
HC-1 0.69 1.39 0.17 1.22 0.67 0.21
HC-2 1.19 2.85 0.76 2.09 1.09 0.35
HC-3 2.56 6.33 1.40 4.93 2.33 0.93
HD-1 2.02 4.11 1.45 2.66 1.90 0.40
HD-2 2.08 3.79 1.60 2.19 1.98 0.32
HD-3 2.75 5.97 2.07 3.90 2.52 0.65
SS-1 0.87 1.73 0.52 1.21 0.85 0.20
SS-2 1.26 2.85 0.82 2.03 1.13 0.36
SS-3 1.95 4.54 1.14 3.40 1.72 0.68
JD-1 0.91 1.71 0.64 1.07 0.89 0.15
JD-2 1.10 2.30 0.75 1.55 1.04 0.23
JD-3 1.76 3.89 1.31 2.58 1.58 0.50
Groundwater level: meters in above mean sea level; Range = maximum water level
minimum water level; Md: median; STD: standard deviation.
242 E. Park et al. / Journal of Hydrology 399 (2011) 235245
were used for the applications, and were linearly deployed consid-
ering the direction of the regional groundwater ow (Fig. 5). The
original purpose of the wells is to monitor seawater intrusion
and the wells have uniform diameters of 150 mm with installation
depths of approximately 150 m below sea level. The basic informa-
tion on the groundwater uctuation statistics of the series are tab-
ulated in Table 6. Table 7 presents the distance of each well from
the coast. The daily precipitation data recorded at the Seongsanpo
rainfall station was acquired from the Korea Meteorological
Administration (2008), and the water level records of four series
of boreholes were obtained from the Institute of Environmental
Resource Research (2008). The calibration period for the HC
series was from January 1, 2004 to July 31, 2005; for the HD series,
from January 1, 2004 to December 31, 2004; for the SS series, from
March 1, 2003 to February 29, 2004; and for the JD series,
from September 23, 2005 to June 30, 2006.
By applying the PP model, the parameters k and f/n were cali-
brated by tting the computed values to the observed water table
uctuation within the calibration period. Fig. 6 shows both the ob-
served and the model-calibrated (PP model) water levels at one of
the sampled locations for each of the four series. The average nor-
malized RMS errors for the optimized model parameters were
0.11 m for HC, 0.09 m for HD, 0.09 m for SS and 0.06 m for JD series.
Table 7 shows the calibrated model parameters for the four differ-
ent series of boreholes. In the calibration processes, the porosity
could not be uniquely decided due to the linkage of the model
parameters between f and n. However, it has been consistently re-
ported that the regional recharge to precipitation ratio of the area
is more than 40% (Hahn et al., 1997; Hamm et al., 2005; Won et al.,
2006), and the unique value of the porosity was assumed by
employing the reported value (Table 7). The model application
areas generally have low elevation for the most of the parts and
the topographic gradients are at most 3%, and a spatially uniform
precipitation is assumed. By assuming a 40% regional recharge to
precipitation ratio, the porosity of the HC series was 0.051, the
HD series was 0.042, the SS series was 0.067, and the JD series
was decided to be 0.014. The small values of the delineated poros-
ity do not indicate low regional hydraulic conductivity. The geo-
logic logs of the monitoring boreholes show that the conductive
portion with high effective porosity takes up relatively small vol-
ume while the surrounding basalt with almost inappreciable effec-
tive porosity occupies the majority. Therefore, the conductive parts
may have capabilities of conducing large amount of water while
the volume averaged effective porosity of the whole aquifer is rel-
atively small. For these estimations, the calibrated f/n values for
Table 7
Estimated transmissivities from each observation well in eastern coastal aquifers of
Jeju Island (the shaded values are the estimations from x/W > 0.2).
k (d
1
) f/n
()
n () x (m) W (m) T
(m
2
d
1
)
HC
Series
HC1 0.0040 0.618 0.051 943.2 16,700 3119
HC2 0.0098 2.481 0.051 2727.2 16,700 21,197
HC3 0.0109 6.043 0.051 4299.2 16,700 35,024
HC4 0.0075 7.772 0.051 6329.9 16,700 32,873
HD
Series
HD1 0.0018 0.682 0.042 802.7 16,200 990
HD2 0.0035 1.218 0.042 2441.1 16,200 5353
HD3 0.0104 5.126 0.042 5223.2 16,200 31,283
HD4 0.0075 9.437 0.042 8844.2 16,200 33,088
SS Series SS1 0.0088 1.064 0.067 1429.8 10,300 8065
SS2 0.0126 2.427 0.067 3225.7 10,300 23,559
SS3 0.0138 4.447 0.067 5484.5 10,300 38,234
SS4 0.0128 6.003 0.067 8135.4 10,300 43,407
JD Series JD1 0.0051 1.203 0.014 1700 19,000 2229
JD2 0.0050 1.175 0.014 2916.7 19,000 3621
JD3 0.0041 1.616 0.014 5521.1 19,000 5161
JD4 0.0049 2.126 0.014 8704.1 19,000 8851
JD5 0.0047 2.251 0.014 11065.7 19,000 9825
Fig. 6. Examples of measured water table uctuations and the results of the calibrated PP model during the observation period from four monitoring wells: (a) HC2, (b) HD2,
(c) JD2, (d) SS2.
E. Park et al. / Journal of Hydrology 399 (2011) 235245 243
the most inland observation wells of the series were selected to
minimize the estimation error as in synthetic cases in the previous
sections. For the estimation of the porosities, it is also assumed
that the heterogeneity of the aquifer is not that severe and the cal-
ibrated value of f/n near x/W$ 0.5 is close to the actual value of the
aquifers. The delineated porosity was for the upper portion of the
aquifer, where the water table uctuates.
Table 7 shows the calculations and results for the estimations of
transmissivity values for the domains. Throughout the estimation
process, the transmissivity of the aquifer for the HC series was esti-
mated to be 3.5 10
4
m
2
d
1
at HC3 and 3.3 10
4
m
2
d
1
at HC4.
For the HD series, the values were 3.1 10
4
m
2
d
1
at HD3 and
3.3 10
4
m
2
d
1
at HD4. For the SS series, the estimated value
from SS3 was 3.8 10
4
m
2
d
1
and for SS4 was 4.3 10
4
m
2
d
1
.
For the JD series, the values were somewhat lower than the others,
5.1 10
3
m
2
d
1
for JD3, 8.7 10
3
m
2
d
1
for JD4, and
9.8 10
3
m
2
d
1
for JD5. The estimated transmissivities were all
within the range of the values commonly encountered in the east-
ern coastal aquifers of Jeju Island (Hahn et al., 1997; Hamm et al.,
2005; Won et al., 2006). Also, almost consistent values were esti-
mated from the HC, HD, SS, and JD series, and the range was rela-
tively narrow (0.5 10
4
4 10
4
m
2
d
1
). By considering 150 m as
the thickness of the aquifer below MSL, the range of hydraulic con-
ductivity was 30300 md
1
, which is also within the range of val-
ues measured from the aquifers of Jeju Island (Hahn et al., 1997;
Hamm et al., 2005; Won et al., 2006). The range of values are com-
monly observed from permeable basalt (Freeze and Cherry, 1979).
In this study, hydraulic heterogeneities in the aquifer such as
aquifer media, bottom layer elevation, recharge rate, and the asso-
ciated uncertainty of the estimation were not considered. These is-
sues, however, need to be resolved to increase the practicality of
the proposed methodology for the estimation of hydraulic proper-
ties, and will be covered in further studies.
5. Conclusions
This study proposes a new physically based method of estimat-
ing regional scale hydraulic conductivity from water level uctua-
tion. The method is based on the water level uctuation model
developed by Park and Parker (2008). The analytical approaches
used in this study take into account the boundary conditions, spec-
ied as head for one side and no ow for the other side.
To conrm the validity of the developed equation, numerical
models were developed and the water level uctuation data were
generated using hydraulic parameters and varying precipitation
rates. The hydraulic conductivities of the aquifer were estimated
from the synthesized transient water level data. The largest esti-
mation error of hydraulic conductivity was found in the case with
the highest hydraulic conductivity (i.e., K = 1000 md
1
), and the
minimum estimation error was from the case with the lowest
(i.e., K = 100 md
1
). The estimation deviations were attributed to
the boundary effect, and the effect was easily transferred into
the domain with higher hydraulic conductivity. On the other
hand, the minimum estimation errors were from the observation
points around the center of the domain (i.e., x/W$ 0.5). This
observation implies that the proposed method works best when
observation data is available around the center of the ow
regime.
To further validate the method, the inuences of the basal
inclinations and the degree of the heterogeneities of the aquifer
are considered. In the basal inclination cases, the estimations
made from Eqs. (13), (14) were consistent as the at basement
cases and close to the input hydraulic conductivity with less than
10% error. The estimations made from the heterogeneous cases
with different magnitude of log transformed variance of the
hydraulic conductivity show that the estimated hydraulic conduc-
tivities are consistent for the observation points located at
x/W > 0.2 and the values are within the range of the geometric
and the arithmetic mean of the modeled aquifer. When the esti-
mated values are compared with mixed mean value of the
hydraulic conductivity elds, the deviations are at most 5% which
happens in the largest heterogeneity case. The overall results
from the inclined basement and heterogeneity cases indicate that
the developed method is robust way of delineating regional
hydraulic conductivity once the observation well is apart
(x/W > 0.2) from the constant head boundary. Also the results
indicate that the local variation is not too much inuential to
the estimation.
The developed equation was also applied to the real case of the
eastern coastal aquifers of Jeju Island. From the estimations,
the transmissivities of the aquifers were around 3.4 10
4
m
2
d
1
for the HC series, 3.2 10
4
m
2
d
1
for the HD series,
4.1 10
4
m
2
d
1
for the SS series, and 7.5 10
3
m
2
d
1
for the JD
series. The values were in the range of those commonly encoun-
tered in the areas where the estimations were made.
The proposed method is well-suited for the aquifer which re-
charge rate is characterized in detail as in the presented synthetic
and real cases. Therefore, the application of the proposed method
to the aquifers with large uncertainty in the recharge rate is not
desirable. However, considering the magnitude of the averaged re-
charge ratio which is commonly between 0.1 and 0.5, the proposed
method may serve to reduce the uncertainty in the hydraulic con-
ductivity of the aquifer, which is a few order magnitudes in many
cases.
The developed methodology could be conveniently applied to
aquifers where the boundary conditions are consistent with this
study, especially the coastal aquifers, once long-term continuous
observation of the water level uctuation is available. The heter-
ogeneities of the model parameters, however, were not consid-
ered in depth. Also, the conguration of the conceptual model
was very limited, because characteristics such as at or linearly
changing bottom elevation, parallel ow along the direction of
the observations, or no sources or sinks other than areal recharge
from precipitation, which are commonly assumed in conventional
estimations of permeability, were assumed for the development
of the model.
Acknowledgements
This research was supported by Basic Science Research Program
through the National Research Foundation of Korea (NRF) funded
by the Ministry of Education, Science and Technology (2010-
0009900).
Appendix A
A.1. Derivation of Eq. (6)
By Darcys law, the specic discharge pass through a cross sec-
tion of an aquifer at a certain location can be written as
Kb h
@h
@x

Cdx A:1
and which is more conveniently expressed as
Kb h
@h
@x
Cx C
1
A:2
By the separation of the variables, Eq. (A2) can be rewritten as
Kh
2
2Kbh Cx
2
C
1
x C
2
0 A:3
244 E. Park et al. / Journal of Hydrology 399 (2011) 235245
and the general solution can be given by
h b

b
2

Cx
2
K
C
1
x C
2

A:4
constrained h P0.
From the boundary condition of constant head of 0 m at x = 0,
one can get
h0 b

b
2
C
2

0 A:5
and C
2
= 0. From the boundary condition of no ow at x = W, one can
get
dh
dx

xW


2CW
K
C
1
2

b
2

CW
2
K
C
1
W C
2
0 A:6
and C
1
= 2CW/K.
Therefore, the solution of Eq. (A1) becomes
h b W

b
W

2

C
K
x
W

2
2
C
K
x
W

A:7
References
Audouin, O., Bodin, J., 2007. Analysis of slug tests with high frequency oscillations. J.
Hydrol. 334, 282289.
Bear, J., 1979. Hydraulics of Groundwater. McGraw-Hill Series in Water Resources
and Environmental Engineering. McGraw-Hill, New York.
Bierkens, M.F.P., 1998. Modeling water table uctuations by means of a stochastic
differential equation. Water Resour. Res. 34, 24852499.
Boulton, N.S., Streltsova, T.D., 1975. New equations for determining the formation
constants of an aquifer frompumping test data. Water Resour. Res. 11, 148153.
Bouwer, J., Rice, R.C., 1976. A slug test for determining hydraulic conductivity of
unconned aquifers with completely or partially penetrating wells. Water
Resour. Res. 12, 423428.
Butler Jr., J.J., 1990. The role of pumping tests in site characterization: Some
theoretical considerations. Ground Water 28, 394402.
Camp Dresser and McKee Inc., 1993. Groundwater Modeling Study: Review of
Existing Data, Ewa Marina Project: Report to Haseko (Ewa), Inc, Honolulu,
Hawaii, September 1993, variously paged.
COMSOL AB, 2005a. COMSOL Multiphysics Earth Science Module Users Guide
(Version 3.2), 120p.
COMSOL AB, 2005b. COMSOL Multiphysics Modeling Guide (Version 3.2), 336p.
Cooper, H.H., Bredehoeft, J.D., Papadopulos, I.S., 1967. Response of a nite-diameter
well to an instantaneous charge of water. Water Resour. Res. 3, 263269.
Cooper, H.H., Jacob, C.E., 1946. A generalized graphical method for evaluating
formation constants and summarizing well eld history. Trans. Am. Geophys.
Union 27, 526534.
Coulibaly, P., Anctil, F., Aravena, R., Bobe e, B., 2001. Articial neural network
modeling of water table depth uctuations. Water Resour. Res. 37, 885896.
Dann, R.L., Close, M.E., Pang, L., Flintoft, M.J., Hector, R.P., 2008. Complementary use
of tracer and pumping tests to characterize a heterogeneous channelized
aquifer system in New Zealand. Hydrogeol. J. 16, 11771191.
Freeze, R.A., Cherry, J.A., 1979. Groundwater. Prentice-Hall Inc., Englewood Cliffs, NJ.
604 p.
Hahn, J., Lee, Y., Kim, N., Hahn, C., Lee, S., 1997. The groundwater resources and
sustainable yield of Cheju volcanic island. Korea Environ. Geol. 33, 4353.
Hanson, R.J., 1986. Least squares with bounds and linear constraints. SIAM J. Sci.
Stat. Comput. 7, 826834.
Hamm, S.-Y., Cheong, J.Y., Jang, S., Jung, C.Y., Kim, B.S., 2005. Relationship between
transmissivity and specic capacity in the volcanic aquifers of Jeju Island. Korea
J. Hydrol. 310, 111121.
Hantush, M.S., 1964. Hydraulics of Wells. In Advances in Hydrosciences, vol. 1.
Academic Press, New York. pp. 281432.
Healy, R.W., Cook, P.G., 2002. Using groundwater levels to estimate recharge.
Hydrogeol. J. 10, 91109.
Hvorslev, M.J., 1951. Time lag and soil permeability in groundwater observations.
Waterways Experiment Station Bulletin, vol. 36. USACOE, Vicksburg, MS.
Institute of Environmental Resource Research, 2008. <http://eri.jeju.go.kr/>.
Jejudo, Korea Water Resources Corporation, 2003. Report on hydrogeology and
coordinated investigation of groundwater resources in Jeju Island: 425 (in
Korean, title translated).
Jussel, P., Stauffer, F., Dracos, T., 1994. Transport modeling in heterogeneous
aquifers, 1. statistical description and numerical generation of gravel deposits.
Water Resour. Res. 30, 18031817.
Knotters, M., Bierkens, M.F.P., 2000. Physical basis of time series models for water
table depths. Water Resour. Res. 36, 181188.
Koh, G.-W., 1997. Characteristics of the groundwater and hydrogeologic
implications of the Seoguipo formation in Cheju Island. PhD thesis, Busan
National University, Busan, Korea, 325 p.
Korea Institute of Geoscience and Mineral Resources, Jejudo, 2004. Establishment of
sustainable groundwater supply system in Jeju Island. 214 p. (in Korean).
Korea Meteorological Administration, 2008. <http://www.kma.go.kr/index.html>.
Mink, J.F., 1980. State of the groundwater resources of southern Oahu. Honolulu
Board of Water Supply, 83 p.
Mink, J.F., Lau, L.S., 1980. Hawaiian groundwater geology and hydrology, and early
mathematical models. University of Hawaii Water Resources Research Center
Technical Memorandum Report, 62, 74 p.
Moench, A.F., 1995. Combining the Neuman and Boulton models for ow to a well
in an unconned aquifer. Ground Water 33, 378384.
Neuman, S.P., 1972. Theory of ow in unconned aquifers considering delayed
response of the water table. Water Resour. Res. 8, 10311045.
Neuman, S.P., 1974. Effects of partial penetration on ow in unconned aquifers
considering delayed gravity response. Water Resour. Res. 10, 303312.
Pang, L., Cose, M.E., Noonan, M., 1998. Rhodamine WT and Bacillus subtilis transport
through an alluvial gravel aquifer. Ground Water 36, 112122.
Park, E., Parker, J.C., 2008. A simple model for water table uctuations in response to
precipitation. J. Hydrol. 356, 344349. doi:10.1016/j.jhydrol.2008.04.22.
Ramey, H.J., 1992. Advances in practical well-test analysis. J. Petrol. Technol. 44,
650689.
Rai, S.N., Manglik, A., 1999. Modelling of water table variation in response to time-
varying recharge from multiple basins using the linearised Boussinesq
equation. J. Hydrol. 220, 141148.
Rai, S.N., Singh, R.N., 1995. Two-dimensional modelling of water table uctuation in
response to localized transient recharge. J. Hydrol. 167, 167174.
Rai, S.N., Manglik, A., Singh, V.S., 2006. Water table uctuation owing to time-
varying recharge, pumping and leakage. J. Hydrol. 324, 350358.
Rasmussen, W.C., Andreasen, G.E., 1959. Hydrologic Budget of the Beaverdam Creek
Basin, Maryland: US Geological Survey Water-Supply Paper 1472, 106 p.
Sa nchez-Vila, X., Meier, P.M., Carrera, J., 1999. Pumping tests in heterogeneous
aquifers: an analytical study of what can be obtained from their interpretation
using Jacobs method. Water Resour. Res. 35, 943952.
Schad, H., Teutsch, G., 1994. Effects of the investigation scale on pumping test
results in heterogeneous porous aquifers. J. Hydrol. 159, 6177.
Sophocleous, M.A., 1991. Combining the soil water balance and water-level
uctuation methods to estimate natural groundwater recharge: practical
aspects. J. Hydrol. 124, 229241.
Souza, W.R., Voss, C.I., 1987. Analysis of an anisotropic coastal aquifer system
using variable-density ow and solute transport simulation. J. Hydrol. 92,
1741.
Theis, C.V., 1935. The relation between the lowering of the piezometric surface and
the rate and duration of discharge of a well using groundwater storage. Trans.
Am. Geophys. Union 16, 519524.
Won, J.-H., Kim, J.-W., Koh, G.-W., Lee, J.-Y., 2005. Evaluation of hydrogeological
characteristics in Jeju Island, Korea. Geosciences J. 9, 3346.
Won, J.-H., Lee, J.-Y., Kim, J.-W., Koh, G.-W., 2006. Groundwater occurrence on Jeju
Island, Korea. Hydrogeol. J. 14, 532547.
Zlotnik, V.A., McGuire, V.L., 1998. Multi-level slug tests in highly permeable
formations: 1 modication of the SpringerGelhar (SG) model. J. Hydrol. 204,
271282.
E. Park et al. / Journal of Hydrology 399 (2011) 235245 245

Potrebbero piacerti anche