Journal of Hydrology 533 (2016) 541–556

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Combining groundwater quality analysis and a numerical flow

simulation for spatially establishing utilization strategies for
groundwater and surface water in the Pingtung Plain
Cheng-Shin Jang a, Ching-Fang Chen b, Ching-Ping Liang c, Jui-Sheng Chen b,⇑
a
Department of Leisure and Recreation Management, Kainan University, Luzhu, Taoyuan 338, Taiwan
b
Graduate Institute of Applied Geology, National Central University, Jhongli, Taoyuan 320, Taiwan
c
Department of Environmental Engineering and Science, Fooyin University, Kaohsiung 831, Taiwan

a r t i c l e i n f o s u m m a r y

Article history: Overexploitation of groundwater is a common problem in the Pingtung Plain area of Taiwan, resulting in
Received 4 October 2015 substantial drawdown of groundwater levels as well as the occurrence of severe seawater intrusion and
Received in revised form 8 December 2015 land subsidence. Measures need to be taken to preserve these valuable groundwater resources. This study
Accepted 14 December 2015
seeks to spatially determine the most suitable locations for the use of surface water on this plain instead
Available online 24 December 2015
This manuscript was handled by Geoff
of extracting groundwater for drinking, irrigation, and aquaculture purposes based on information
Syme, Editor-in-Chief obtained by combining groundwater quality analysis and a numerical flow simulation assuming the plan-
ning of manmade lakes and reservoirs to the increase of water supply. The multivariate indicator kriging
Keywords:
method is first used to estimate occurrence probabilities, and to rank townships as suitable or unsuitable
Groundwater for groundwater utilization according to water quality standards for drinking, irrigation, and aquaculture.
Multivariate indicator kriging A numerical model of groundwater flow (MODFLOW) is adopted to quantify the recovery of groundwater
MODFLOW levels in townships after model calibration when groundwater for drinking and agricultural demands has
Drinking been replaced by surface water. Finally, townships with poor groundwater quality and significant
Irrigation increases in groundwater levels in the Pingtung Plain are prioritized for the groundwater conservation
Aquaculture planning based on the combined assessment of groundwater quality and quantity. The results of this
study indicate that the integration of groundwater quality analysis and the numerical flow simulation
is capable of establishing sound strategies for joint groundwater and surface water use. Six southeastern
townships are found to be suitable locations for replacing groundwater with surface water from
manmade lakes or reservoirs to meet drinking, irrigation, and aquaculture demands.
Ó 2015 Elsevier B.V. All rights reserved.

1. Introduction overexploitation of groundwater and the low ratio of tap water

Owing to limited surface water, groundwater is a critical water
resource in the western and southwestern coastal areas of Taiwan.
The Pingtung Plain has the lowest ratio of tap water use, approxi-
mately 45.8%, significantly less than the average in Taiwan, 92.7%
(Taiwan Water Resources Agency [WRA], 2012), because of the
abundant and inexpensive groundwater. However, a substantial
amount of groundwater has already been extracted to meet drink-
ing, irrigation, and aquaculture demands, leading to significant
drawdown in overall groundwater levels and the occurrence of
severe seawater intrusion and land subsidence over the past 5
decades (Ting et al., 1998). To solve the problem of the
⇑ Corresponding author. Tel.: +886 3 2807427; fax: +886 3 4263127.
E-mail address: jschen@geo.ncu.edu.tw (J.-S. Chen).
usage, the Taiwanese government plans to establish manmade
lakes and reservoirs to increase the supply of surface water. Hence,
the planning for utilization and management of groundwater
resources is critical in the Pingtung Plain. Sustainable and reason-
able groundwater utilization plan minimizing the impact of extrac-
tion on the groundwater levels that still meets the quality
conditions of specified groundwater uses is proposed. This study
integrates the concept of groundwater quality and quantity
towards spatially exploring sustainable and reasonable groundwa-
ter use in the Pingtung Plain.
Groundwater typically contains a variety of natural and anthro-
pogenic contaminants, so its quality is one of the most crucial fac-
tors affecting groundwater utilization for various purposes. For
example, the use of the water with high concentrations of man-
ganese (Mn) and iron (Fe) for purposes of agriculture or aquacul-
ture may lead to retardation of growth or even the death of

http://dx.doi.org/10.1016/j.jhydrol.2015.12.023
0022-1694/Ó 2015 Elsevier B.V. All rights reserved.
542 C.-S. Jang et al. / Journal of Hydrology 533 (2016) 541–556

cultivated plants or fish (Tsai et al., 2006; Liao et al., 2008; Ye et al.,
2009). The use of groundwater with high level of arsenic (As) is
harmful to human health. Arsenic can also accumulate in crops
and fish, posing an indirect threat to human health through inges-
tion (Huq and Naidu, 2005; Jang et al., 2006). It has been observed
that employing water with high chloride (Cl
) and sodium (Na+)
contents in irrigation can lead to toxicity and growth retardation
in many crops (Grieve et al., 2006). Moreover, water with high
levels of Cl
and Na+ is not suitable for drinking (D’Alessandro
et al., 2008). Although nitrogen (N) compounds assist plant growth,
irrigation water with high amounts of N can cause quality
problems in some crops and excessive vegetative growth in certain
vegetables (Bauder et al., 2007). Epidemiological evidence has indi-
cated that long-term exposure to nitrate-N through the drinking
water pathway is strongly associated with numerous diseases
(Yang et al., 1998; Knobeloch et al., 2000). Clearly, the quality of
groundwater extracted for use in the water supply needs to be
carefully observed.
Typically there is considerable spatial variability in the param-
eters indicative of water quality. However, only a finite amount of
field data has been surveyed because of constraints on funding,
time, and resources. The spatial uncertainty of variables frequently
results from limited observations; therefore, geostatistics has been
widely adopted for spatially characterizing the variability and

Fig. 1. Map of the study area.

 C.-S. Jang et al. / Journal of Hydrology 533 (2016) 541–556 543

distribution of finite surveyed data with uncertainty (Goovaerts,
1997). Indicator kriging (IK) is the most widely used nonparamet-
ric geostatistical approach without assumption on the distribu-
tions of spatial variables. In IK, binary transformation of the data
is used to make the predictor robust for outliers. The estimated
value made by IK represents a probability less than or equal to a
certain threshold at an unknown location. The spatial uncertainty
of variables can be characterized according to probability estima-
tion. Furthermore, the water quality standards for drinking, irriga-
tion, and aquaculture generally regulate the thresholds of
numerous hydrochemical parameters. The quality assessment of
groundwater for various water resource demands need to be
addressed using a multivariate analysis approach. Smith et al.
(1993), Oyedele et al. (1996), Halvorson et al. (1996), and
Diodato and Ceccarelli (2004) developed the multivariate indicator
kriging (MVIK) method for probabilistically evaluating soil quality
in farmland through the integration of numerous chemical param-
eters. Lee et al. (2008), Jang and Chen (2009), and Jang et al. (2012a,
2012b) used MVIK for spatially locating suitable groundwater to
meet irrigation, aquaculture, and drinking demands. In addition,
stochastic simulation techniques, such as sequential Gaussian sim-
ulation and sequential indicator simulation, which reproduce
alternative and equally probable realizations for spatial variables
(Goovaerts, 1997), can also be used to characterize the spatial
uncertainty of variables. However, because more computation
efforts of the multivariate integration require to be conducted for
the stochastic simulation techniques, this study only considered
MVIK to characterize the spatial uncertainty of variables according
to the probability estimation.
Overexploitation of groundwater may result in a substantial
decrease in groundwater levels and an increase in the potential of
seawater intrusion and land subsidence. The numerical model of
groundwater flow, MODFLOW, developed by Harbaugh and
McDonald (1996), has become a widely used tool for quantifying
changes in groundwater flow and levels caused by natural recharge,
artificial withdrawal, and specified hydrological events (Rejani
et al., 2008; Liu et al., 2010; Mittelstet et al., 2011; Al-Salamah
et al., 2011; Jang et al., 2012b). This numerical model of groundwa-
ter flow is applied to predict the effects of recovery on groundwater
levels for specified hydrological cases in this study.
To reach the goal of sustainable and reasonable groundwater
utilization, it is necessary to integrate groundwater quality and
quantity, and a variety of factors that should be considered in the
regional development of water resources. Jang et al. (2012a) and
levels are prioritized when planning groundwater conservation in
the Pingtung Plain.
2. Materials and methods
2.1. Hydrogeology of the study area
The Pingtung Plain is situated in the southwestern part of Tai-
wan (Fig. 1), and has an area of 1210 km2. It is bounded to the
north by foothills and river valleys, to the west by the Fengshan
Fault, to the south by the Taiwan Strait, and to the east by the
Chaozhou Fault. The Kaoping River, Donggang River, and Linbian
River are the three major rivers crossing this plain. According to
climate statistics for the period between 2005 and 2012, the
annual precipitation in the area ranges from 1928 to 3827 mm,
with an average of 2840 mm. Rainfall is concentrated between
May and September. Unconsolidated Quaternary deposits underly-
ing the plain are rich in groundwater resources. This plain is
divided primarily into a proximal-fan and a distal-fan. The
proximal-fan and the river valleys at the eastern and northern foot-
hills are the principal zones for natural aquifer replenishment.
Farmlands, paddy fields and dry farmlands receive the most use,
and make up of approximately 45.7% of the study area with fish-
ponds covering approximately 5.1% of the study area. Approxi-
mately 60% of farmland and fishponds and 54% of households
have used groundwater to meet water resource demands over
the last several decades (Taiwan Directorate-General of Budget,
Accounting and Statistics [DGBAS], 2005; Taiwan WRA, 2012),
resulting in the occurrence of severe land subsidence and seawater
intrusion in this plain (Ting et al., 1998).
A hydrogeological survey within a depth of 250-m was com-
pleted between 1995 and 1998 to characterize the spatial pattern
Valley
recharge
zone
Rainfall
recharge 100
zone
50
Over-pumping
zone
Fault

0
Jang et al. (2013) integrated the spatial variability of groundwater Sea Aquifer 1

Elevation (m)
quality parameters and aquifer transmissivity (an indicator of
groundwater quantity) to determine sustainable groundwater uti-
-50
lization for purposes of irrigation and aquaculture, respectively.
Jang et al. (2012b) combined the spatial variability of groundwater Aquifer 2
quality parameters and a groundwater model simulation when
planning emergency utilization in an urban region during -100
droughts.
The objective of this study is to make a spatial determination of Aquifer 3
the most suitable locations for surface water (after the construc- -150
tion of man-made lakes and reservoirs) for drinking, irrigation,
and aquaculture purposes in the Pingtung Plain through a combi-
Aquifer 4
nation of groundwater quality analysis and a numerical flow sim-
ulation. First, according to the water quality standards for drinking, -200
irrigation, and aquaculture, MVIK, which integrates several
groundwater quality parameters, is used to model occurrence S N
probabilities, and rank suitable and unsuitable townships for -250
groundwater use. Second, MODFLOW is adopted to evaluate the 60 50 40 30 20 10 0
effects of recovery of groundwater levels after model calibration Distance (Km)
showing the replacement of groundwater for drinking and agricul- Aquifer Aquitard Rock
tural usages by surface water. Finally, townships with poor
groundwater quality and significant increases in groundwater Fig. 2. A conceptual hydrogeological profile of the Pingtung Plain.
544 C.-S. Jang et al. / Journal of Hydrology 533 (2016) 541–556

Aquifer 1 Aquifer 2 Aquifer 3 Aquifer 4

(a) (b) (c) (d)

Aquifer 1 Aquifer 2 Aquifer 3 Aquifer 4

(e) (f) (g) (h)

Aquifer 1 Aquifer 2 Aquifer 3 Aquifer 4

(i) (j) (k) (l)

Fig. 3. Observation data on groundwater quality parameters meeting (d) or exceeding (s) water quality standards for (a–d) drinking, (e–h) irrigation and (i–l) aquaculture.
 C.-S. Jang et al. / Journal of Hydrology 533 (2016) 541–556 545

The water quality standards for drinking, irrigation, and aquaculture in Taiwan are available at http://w3.epa.gov.tw/epalaw/docfile/090040.doc, http://www.coa.gov.tw/files/web_articles_files/8655/1087.doc, and http://law.
of aquifers and aquitards in this plain. A network of groundwater

D: 0.01

A: 0.05
(mg/L)
monitoring was also implemented, including 51 hydrogeological
I: 0.05

0.35
0.02
0.04
ND
investigation stations and 126 monitoring wells (Taiwan Central

29

11

11
As

Geology Survey [CGS], 2002). The hydrogeological investigation

stations were equipped with one to four monitoring wells located
D: 0.05

A: 0.05
(mg/L)

I: 0.2

11.0 in various aquifers. The plain sediments in the distal-fan can be

1.10
0.40
Mn

ND

93

56

93
further divided into eight overlapping sequences, including four
marine sequences and four nonmarine sequences (Taiwan CGS,
(mg/L)

2002). Nonmarine sequences with highly permeable coarse

D: 0.3

18.7

2.32
1.20
I: 5

ND

sediments, ranging from medium sand to gravel, are considered

74
Fe

–
aquifers, whereas marine sequences with lowly permeable fine
sediments, ranging from clay to fine sand, are regarded as aqui-
(meq/L)

6.68

tards (Liu et al., 2006) (Fig. 2). Aquitards are primarily present in

105
I: 2.5

22.1
8.10
RSC

11

the distal-fan, but absent in the proximal-fan. The four aquifers

–

labeled ‘‘Aquifer 1,” ‘‘Aquifer 2,” ‘‘Aquifer 3,” and ‘‘Aquifer 4” from
top to bottom; are at depths ranging from 0 to 70 m, 40 to 130 m,
(meq/L)0.5

90 to 180 m, and 160 to 250 m, respectively.

47.6
3.78
0.12

9.03
I: 6
SAR

16
–

2.2. Hydrochemical data for the groundwater

35,505
D: 500
(mg/L)

2520
7190

The Agricultural Engineering Research Center (AERC) which is

TDS

120

30

financially supported by the Taiwan WRA implemented a long-

–

term groundwater quality survey beginning in 2009 (AERC, 2009,

Hardnessb

2010, 2011, 2012), monitoring 31 groundwater quality parameters.

Statistics concerning hydrochemical parameters in groundwater, and the water quality standards of drinking, irrigation, and aquaculture in Taiwan.

D: 300
(mg/L)

Temperature, dissolved oxygen, electrical potential, pH value, and

7256

1263
58.0

592

electrical conductivity (EC) were measured in situ, and the remain-

32

ing hydrochemical parameters were analyzed in the laboratory.

This study used data on the groundwater quality parameters
Total Na
(mg/L)

reported by the AERC (2009, 2010, 2011, 2012) to assess suitable

45.8
3.35
6.45
0.12
I: 3

29

and unsuitable locations for groundwater use. The quality of water

–

for drinking and irrigation in Taiwan is gauged from ten and thir-
teen hydrochemical parameters, respectively. The hydrochemical
(mg/L)
NH+4-N

D: 0.1
A: 0.3

45.4
2.14

parameters used for the assessment of drinking water included

ND

6.5
79

48
–

Cl
, SO2


+
4 , NO2 -N, NO3 -N, NH4-N, hardness, total dissolved solid
(TDS), Fe, Mn, and As and those for the assessment of irrigation
3 -N
(mg/L)

water contained EC, Cl
, SO2

D: 10

4 , total N (the sum of NO3 -N, NO2 -N

NO

1.17
1.76
10.1
ND

and NH+4-N), Ca2+, Mg2+, Na+, alkalinity, Fe, Mn, and As. In practice,
1

the sodium adsorption ratio (SAR) and the residual sodium carbon-
ate (RSC), which affect the structural stability of soil and usage
2 -N
(mg/L)
D: 0.1
NO

0.36
0.01
0.04

restrictions for irrigation (USSL, 1954; Eaton, 1950), were used to

ND

gauge the quality of irrigation water rather than Ca2+, Mg2+, Na+,
and alkalinity. Three hydrochemical parameters were used to
D: 250
(mg/L)

I: 200

gauge the quality of water used in aquaculture, namely NH+4-N,

2377
SO2

148
402
4

2.1

15

Mn, and As.

9

Fig. 3 maps observation data for groundwater quality parame-

17,981
D: 250
(mg/L)

ters that meet or exceed water quality standards. Table 1 lists

I: 175

1163
3788
Cl

3.2

the water quality standards, as well as the average, standard devi-

17

17

ation, and maximum and minimum of the groundwater quality

parameters. The statistical results illustrate a wide distribution of
EC (lS/cm

the hydrochemical parameters with several outliers. A

50,078
25 °C)
I: 750

9893
3530

epa.gov.tw/zh-tw/laws/309417667.html, respectively.

Kolmogorov–Smirnov normality test adopted to examine ground-

148

29

3 -N, NO2 -N and NH4-N.

–

water quality parameters reveals that they were not normally or

+

log-normally distributed. Therefore, nonparametric geostatistics

Water quality standards for drinking (D),

are suitable for modeling the spatial distributions and uncertainty.

Number of wells with water exceeding

Number of wells with water exceeding

Number of wells with water exceeding

irrigation (I) and aquaculture (A)c

the standard for drinking water

2.3. Numerical model of groundwater flow

A numerical model of groundwater flow in the Pingtung Plain

Total N: the sum of NO

the aquaculture water

was developed using MODFLOW-96 Version 4.2 (Harbaugh and

the irrigation water

Expressed by CaCO3.

McDonald, 1996). MODFLOW is a modular three-dimensional

Standard deviation

finite difference groundwater flow model, which can simulate

Statistical data

groundwater flow and levels under complex hydrogeological con-

Maximum
Minimum

ditions with various hydrological processes. In the conceptual

Average

model, this study area was divided vertically into four aquifers
Table 1

and three aquitards according to various hydrogeological

b
a

situations (Fig. 2), and each aquifer and aquitard was discretized
546 C.-S. Jang et al. / Journal of Hydrology 533 (2016) 541–556

: Aquifer 1
: Aquifer 2
: Aquifer 3
: Aquifer 4
: Aquitard

Fig. 4. Cross sections of the three-dimensional grid of groundwater flow model.

horizontally into a grid of 34  72 cells with a spacing of 1000 m. model, cðhÞ, such as the spherical, exponential or Gaussian model.
Fig. 4 displays cross sections of the three-dimensional grid of the The nugget effect (c0), the sill (c), and the range (a) were acquired
groundwater flow model based on the hydrogeological situations. during the fitting process. The theoretical models are mathemati-
According to the hydrogeological judgment of the Taiwan CGS cally expressed as follows (Goovaerts, 1997):
(2002), constant head boundaries were assigned to the northern, 8 h    3 i
northeastern, and eastern river valleys, and the southern coast >
< cðhÞ ¼ c0 þ c 1:5 ha
0:5 ha ; h 6 a
(Fig. 1) and no flow boundaries were set for the others regions. Spherical model : ;
>
:
The hydraulic conductivity and storage coefficient of the aquifers cðhÞ ¼ c0 þ c; h>a
were obtained from pumping test data reported by the Taiwan ð2Þ
CGS (2002) (Fig. 5). The parameter estimation program, UCODE
(Poeter and Hill, 1998), was used to inversely determine recharge   
3h
and pumping rates in townships, minimizing errors between Exponential model : cðhÞ ¼ c0 þ c 1
exp
; ð3Þ
a
observed and simulated heads. Data on monthly groundwater
levels of monitoring wells in the Pingtung Plain, which were "  2 !#
3h
recorded by the WRA (2012), from January 2007 to January 2010 Gaussian model : cðhÞ ¼ c0 þ c 1
exp
: ð4Þ
were used for model calibration. The root mean square error a
(RMSE) was employed to gauge the performance of the model cal-
ibration, defined as follows (Hill and Tiedeman, 2007).
2.4.2. Indicator kriging
" #0:5
1 X
SP X
NW
sim obs 2
IK is a nonparametric geostatistical method for characterizing
RMSE ¼ ðhi;t
hi;t Þ ð1Þ the probability that the attribute value does not exceed a certain
NW  SP t¼1 i¼1
threshold, zk, at a known site u. In IK, the spatial variable, Z(u), is
where NW represents the number of the wells; SP denotes the sim- converted into an indicator variable with a binary datum, Iðu; zk Þ,
ulated period; hsim and hobs are the simulated and observed ground- and given as follows (Goovaerts, 1997).
water levels, respectively; and t is the duration of the model 8
< 1; if ZðuÞ 6 zk ; k ¼ 1; 2; . . . l
simulation. Additionally, the coefficient of determination between Iðu; zk Þ ¼ ð5Þ
observed and simulated groundwater levels was also adopted to :
0; otherwise
measure the performance of the model calibration (Hill and
Tiedeman, 2007). The water budgets of the model calibration and Detailed analysis procedures of IK at an unknown site can be found
prediction processes were analyzed to explore the temporal in Goovaerts (1997).
changes of subsurface hydrology (Hill and Tiedeman, 2007).
2.4.3. Multivariate indicator kriging
2.4. Multivariate indicator-based geostatistical approaches The water quality standards for drinking, irrigation, and aqua-
culture contain different thresholds for various hydrochemical
2.4.1. Variogram analysis parameters which in the process of the multivariate integration
An experimental variogram was used to quantify the spatial are determined by multiplying indicator variables, and defined as
variability of random variables, and fitted using a theoretical follows (Smith et al., 1993):
 C.-S. Jang et al. / Journal of Hydrology 533 (2016) 541–556 547

(a) Aquifer 1 (b) Aquifer 2

(c) Aquifer 3 (d) Aquifer 4

Fig. 5. Distributions of hydraulic conductivity (m/day) used in the numerical model of groundwater flow.

Y
m
3. Results and discussion
Iðu; zd=i=a Þ ¼ Ij ðu; zk Þ; ð6Þ
j¼1
3.1. Multivariate integration of hydrochemical parameters based on
where zd=i=a is the water quality standard for drinking, irrigation or water quality standards for drinking, irrigation and aquaculture
aquaculture, and m is the number of indicator variables combined
for the water quality standard. The GAMV and KT3D codes in GSLIB Each hydrochemical parameter was first transformed into an
(Deutsch and Journel, 1998) were employed to execute the indicator variable by turns according to Taiwan’s water quality
experimental variogram and MVIK, respectively. standards for drinking, irrigation and aquaculture. Table 1 reports
548 C.-S. Jang et al. / Journal of Hydrology 533 (2016) 541–556

the number of wells exceeding the standards for each hydrochem- Table 2
ical parameter. The component exceeding water quality standards The fitted parameters of the variograms of the multivariate integration for ground-
water use purposes.
in the greatest number of wells was Mn, and those exceeding the
water quality standards for drinking, irrigation, and aquaculture Use purposes Aquifer Model type Nugget Sill Range (m)
in the smallest number of wells were NO
3 -N, Fe, and As, respec- Drinking 1 Spherical 0.000 0.110 40,000
tively. According to the results of multivariate integration, water 2 Exponential 0.060 0.060 35,000
quality standards for drinking, irrigation and aquaculture were 3 Exponential 0.070 0.045 40,000
4 Spherical 0.000 0.200 40,000
exceeded in 115, 93, and 98 wells, respectively. Fig. 3 displays spa-
tial data from multivariate integration of the hydrochemical Irrigation 1 Exponential 0.100 0.080 40,000
2 Exponential 0.140 0.130 40,000
parameters in various aquifers. Wells where the water quality 3 Exponential 0.150 0.080 15,000
standards for drinking, irrigation, and aquaculture were exceeded 4 Exponential 0.020 0.077 40,000
were located at varied aquifer depths and primarily in the distal Aquaculture 1 Exponential 0.060 0.040 13,000
fan as opposed to the proximal fan. 2 Exponential 0.120 0.140 35,000
3 Spherical 0.060 0.105 18,000
4 Exponential 0.135 0.030 10,000

(a) Aquifer 1 (b) Aquifer 2 (c) Aquifer 3

(d) Aquifer 4 (e) Aquifer 1 (f) Aquifer 2

(g) Aquifer 3 (h) Aquifer 4 (i) Aquifer 1

(j) Aquifer 2 (k) Aquifer 3 (l) Aquifer 4

Fig. 6. Experimental and fitted theoretical variograms of multivariate integration variables for (a–d) drinking, (e–h) irrigation and (i–l) aquaculture use purposes.
 C.-S. Jang et al. / Journal of Hydrology 533 (2016) 541–556 549

Aquifer 1 Aquifer 2 Aquifer 3 Aquifer 4

F(u; zd/i/a|(n))

(a) (b) (c) (d)

Aquifer 1 Aquifer 2 Aquifer 3 Aquifer 4

(e) (f) (g) (h)

Aquifer 1 Aquifer 2 Aquifer 3 Aquifer 4

(i) (j) (k) (l)

Fig. 7. Estimated probabilities of multivariate integration variables for (a–d) drinking, (e–h) irrigation and (i–l) aquaculture use purposes.

3.2. Spatial variability and probability of the multivariate integration
for drinking, irrigation and aquaculture purposes
A lag of 5000 m was used to analyze variograms for the multi-
variate integration of hydrochemical parameters. To fit the vari-
ograms, the least squares method was employed to generate a
relatively compatible set of best fitting models with the smallest
sum of squares of errors. The exponential and spherical models
showed the most optimal fitting in the variograms (Fig. 6 and
Table 2). The fitted nugget, sill and range were 0–0.07 m, 0.045–
0.2 m, and 35,000–40,000 m, respectively, for the standard for
drinking water; 0.02–0.15 m, 0.077–0.13 m, and 15,000–40,000 m,
550 C.-S. Jang et al. / Journal of Hydrology 533 (2016) 541–556

observed

simulated

(a) (b)

(c) (d)
Fig. 8. Contoured comparison of observed and simulated groundwater levels in July 2008.

respectively, for the standard for irrigation water; and
0.06–0.135 m, 0.03–0.14 m, and 10,000–35,000 m, respectively,
for the standard for aquaculture water.
The MVIK method is used in this study to estimate the probabil-
ities of the occurrence through the multivariate integration of
hydrochemical parameters for the various aquifers and for the
quality standards for drinking, irrigation, and aquaculture water.
Each aquifer is discretized into a grid of 34  72 cells, with a
spacing of 1000 m. Fig. 7 shows the probabilities estimated
through multivariate integration of the hydrochemical parameters:
F(u; zd/i/a|(n)). A greater F(u; zd/i/a|(n)) represents more suitable
groundwater for drinking, irrigation, or aquaculture. The high esti-
mated probabilities were mainly located in the northern, north-
eastern, and eastern aquifers, and the low estimated probabilities
were primarily situated in the southern coastal, southwestern,
and western aquifers. The estimated probabilities for drinking
 C.-S. Jang et al. / Journal of Hydrology 533 (2016) 541–556 551

Table 3
Average estimated probabilities and ranks of townships for drinking, irrigation, and aquaculture use purposes.

Townships Drinking Irrigation Aquaculture

Aquifer Aquifer Aquifer
1 2 3 4 Average Ranka 1 2 3 4 Average Ranka 1 2 3 4 Average Ranka
Qishan (QS) 0.20 0.07 0.09 0.13 0.12 7 0.32 0.82 0.62 0.43 0.55 25 0.33 0.24 0.44 0.24 0.31 17
Maynong (MN) 0.03 0.14 0.14 0.24 0.14 8 0.44 0.65 0.66 0.19 0.49 24 0.24 0.30 0.54 0.35 0.35 18
Wanluan (WL) 0.12 0.03 0.03 0.06 0.06 5 0.21 0.27 0.46 0.09 0.26 12 0.11 0.41 0.16 0.11 0.20 13
Likang (LK) 0.56 0.16 0.17 0.22 0.28 12 0.31 0.70 0.67 0.23 0.48 23 0.50 0.38 0.41 0.51 0.45 22
Kaoshu (KS) 0.05 0.40 0.39 0.41 0.31 13 0.47 0.55 0.65 0.10 0.44 22 0.39 0.60 0.66 0.61 0.57 24
Dashu (DS) 0.06 0.04 0.04 0.01 0.04 3 0.31 0.37 0.45 0.04 0.29 15 0.31 0.04 0.08 0.15 0.14 11
Jiuru (JR) 0.23 0.13 0.13 0.11 0.15 9 0.44 0.54 0.58 0.03 0.40 20 0.51 0.24 0.26 0.51 0.38 21
Pingtung (PT) 0.03 0.06 0.06 0.06 0.05 4 0.42 0.27 0.46 0.02 0.29 16 0.23 0.15 0.22 0.20 0.20 14
Yanpu (YP) 0.20 0.40 0.38 0.40 0.34 14 0.50 0.44 0.56 0.13 0.41 21 0.60 0.78 0.63 0.38 0.60 25
Zhangzhi (ZZ) 0.21 0.26 0.26 0.27 0.25 11 0.54 0.35 0.56 0.12 0.40 19 0.40 0.62 0.54 0.48 0.51 23
Daliao (DL) 0.00 0.00 0.00 0.00 0.00 1 0.04 0.04 0.18 0.00 0.06 2 0.01 0.00 0.00 0.01 0.00 2
Wandan (WD) 0.00 0.00 0.00 0.00 0.00 1 0.13 0.13 0.30 0.00 0.14 8 0.08 0.01 0.13 0.20 0.10 8
Linluo (LL) 0.14 0.04 0.04 0.12 0.08 6 0.46 0.20 0.62 0.03 0.33 17 0.30 0.27 0.36 0.60 0.38 20
Zhutian (ZT) 0.03 0.00 0.00 0.02 0.01 2 0.26 0.24 0.57 0.00 0.27 14 0.19 0.16 0.24 0.36 0.24 15
Chiadong (CD) 0.00 0.00 0.00 0.00 0.00 1 0.07 0.21 0.23 0.00 0.13 7 0.00 0.34 0.20 0.00 0.14 10
Neipu (NP) 0.51 0.15 0.15 0.18 0.25 10 0.37 0.23 0.60 0.21 0.35 18 0.26 0.50 0.45 0.31 0.38 19
Shinyuan (SY) 0.00 0.00 0.00 0.00 0.00 1 0.08 0.10 0.13 0.00 0.08 4 0.00 0.00 0.01 0.01 0.01 3
Kanding (KD) 0.00 0.00 0.00 0.00 0.00 1 0.18 0.28 0.21 0.00 0.17 9 0.00 0.07 0.05 0.01 0.03 6
Chaozhou (CZ) 0.00 0.00 0.00 0.00 0.00 1 0.18 0.45 0.43 0.00 0.26 13 0.03 0.30 0.12 0.08 0.14 9
Nanzhou (NZ) 0.00 0.00 0.00 0.00 0.00 1 0.14 0.29 0.26 0.00 0.17 10 0.00 0.12 0.00 0.00 0.03 5
Xinpi (XP) 0.00 0.00 0.00 0.00 0.00 1 0.09 0.38 0.41 0.00 0.22 11 0.00 0.44 0.22 0.08 0.19 12
Fangliao (FL) 0.00 0.00 0.00 0.00 0.00 1 0.05 0.16 0.25 0.00 0.12 6 0.00 0.58 0.36 0.02 0.24 16
Linbian (LB) 0.00 0.00 0.00 0.00 0.00 1 0.08 0.16 0.13 0.00 0.09 5 0.00 0.17 0.06 0.00 0.06 7
Dongkang (DK) 0.00 0.00 0.00 0.00 0.00 1 0.08 0.12 0.10 0.00 0.08 3 0.00 0.06 0.01 0.00 0.02 4
Linyuan (LY) 0.00 0.00 0.00 0.00 0.00 1 0.00 0.03 0.14 0.00 0.04 1 0.00 0.00 0.00 0.00 0.00 1
a
The ranks indicate low to high probabilities.

water were higher in Aquifers 1 and 4 than in Aquifers 2 and 3. The
estimated probabilities for irrigation and aquaculture were smaller
in Aquifer 4 than in Aquifers 1, 2, and 3.
The goal of this study was to evaluate groundwater utilization
in townships, so an average probability within a cell-shaped town-
ship with four aquifers (i.e., a composite conditional cumulative
distribution function (ccdf) [Goovaerts, 1997]) was used to gauge
groundwater quality for the purposes of drinking, irrigation, and
aquaculture. Townships were ranked from low to high depending
on probabilities (Table 3). The poorest groundwater quality for
drinking was found in the townships DL, WD, CD, SY, KD, CZ, NZ,
XP, FL, LB, DK, and LY, and that for irrigation and aquaculture pur-
poses was found in the township LY. The rank denoting optimal
conditions for groundwater quality for drinking, irrigation, and
aquaculture purposes occurred in townships YP, QS, and YP.
3.3. Numerical simulation of groundwater flow and levels
This study adopted MODFLOW to simulate groundwater flow
and water levels in the Pingtung Plain. Because of the absence of
observation data, parameters for monthly pumping and recharging
between January 2007 and January 2010 were inversely deter-
mined through minimizing the difference between observed and
simulated groundwater levels. The average RMSE in the calibration
process was 0.757 m. The coefficient of determination between
observed and simulated groundwater levels was 0.998. Fig. 8
shows the contours of the observed and simulated groundwater
levels. The simulated groundwater levels were quite close to the
observed levels. Fig. 9(a) displays the simulation results for the
fluxes in water balance components between January 2007 and
January 2010. In the simulation, the recharge rate rose substan-
tially during the rainy seasons and the pumping was reduced but
then increased in the dry seasons, which matches the actual situa-
tion of groundwater utilization in this study area. Moreover, the
flux of constant head boundaries was minor compared with the
pumping and recharge rates. Fig. 9(b) shows the simulated fluxes
of constant head boundaries in detail. The influx from the seawater
and outflux to the seawater were approximately 52 and 98 thou-
sand m3/day, respectively. The recharge from river valleys
increases in the rainy seasons and decreases in the dry seasons.
Although data on actual groundwater withdrawal were lacking,
information on groundwater rights, which represents legal
groundwater extraction, was used to explore the regional utiliza-
tion of groundwater resources. The Taiwan Water Corporation, irri-
gation associations, and agricultural and aquacultural corporations
must apply for groundwater rights before pumping groundwater.
According to application data on groundwater rights in 2013 pub-
lished by the Taiwan WRA, the groundwater use amounts were
0.26 billion m3/year (15.9%) and 1.23 billion m3/year (74.5%) for
drinking and agricultural purposes, respectively, and Table 4
reported the data in townships for drinking and agricultural pur-
poses. This study predicts the recovery of groundwater levels for
each township based on the calibrated groundwater model when
the amount of groundwater used is completely replaced by surface
water. Table 5 reports the average responses of simulated ground-
water levels in townships, and ranks them in categories from high
to low. The largest increase in groundwater levels for agricultural
and drinking purposes would occur in townships KS and ZZ,
respectively. The lowest increase for agricultural purposes would
be assigned to townships LB, DK, and LY, and for drinking purposes
to townships JR, LL, ZT, CD, SY, KD, NZ, FL, LB, DK, and LY because
groundwater in these townships was not pumped for agricultural
or drinking purposes. Fig. 9(c) also shows the flux of water balance
components for the six-month prediction processes in six town-
ships with large recovery responses. The simulation results reveal
that a decrease in pumping dominates the groundwater flow and
levels.
3.4. Assessment combining groundwater quality and quantity
A comprehensive plan for protecting groundwater resources
should account for an assessment of combining groundwater
552 C.-S. Jang et al. / Journal of Hydrology 533 (2016) 541–556

150

Fluxes for water balance (million m3/day)

Constant
Head
Recharge
100
Pumping

50

-50

-100
Jan-07 Apr-07 Aug-07 Nov-07 Feb-08 Jun-08 Sep-08 Dec-08 Mar-09 Jul-09 Oct-09 Jan-10
(a) Time

1.4
Influx from the seawater Outflux to the seawater Recharge from river valleys
Fluxes for constant head boundaries

1.2

1.0
(million m3/day)

0.8

0.6

0.4

0.2

0.0

-0.2
Jan-07 Apr-07 Aug-07 Nov-07 Feb-08 Jun-08 Sep-08 Dec-08 Mar-09 Jul-09 Oct-09 Jan-10
(b) Time

200
Constant head (CD)
Fluxes for water balance (million m /day)

Pumping and recharge rates (CD)

Constant head (FL)
3

150 Pumping and recharge rates (FL)

Constant head (NZ)
Pumping and recharge rates (NZ)
Constant head (XP)
100 Pumping and recharge rates (XP)
Constant head (WL)
Pumping and recharge rates (WL)
50 Constant head (CZ)
Pumping and recharge rates (CZ)

-50
1 2 3 4 5 6
(c) Months

Fig. 9. Simulated fluxes of (a) water balance components and (b) constant head boundaries between January 2007 and January 2010 and (c) for the six-month recovery
model.

quality and quantity (Jang et al., 2012a, 2012b). This study evalu- assessment of groundwater quality and quantity was adopted to
ated groundwater quality using MVIK taking into account various rate townships where groundwater is replaced by surface water.
demands for water resources, and analyzed the responses of Furthermore, the former half of the ranks in the assessment of
groundwater levels after halting the use of groundwater, using a groundwater quality or quantity was considered for rating
numerical model of groundwater flow. For convenient manage- groundwater replaced by surface water in townships. Agricultural
ment of groundwater resources, the sum of the ranking of the groundwater rights were not further divided into irrigation and
 C.-S. Jang et al. / Journal of Hydrology 533 (2016) 541–556 553

Table 4
Groundwater right amounts (10
3 m/day) for townships used in the numerical model of groundwater flow according to groundwater rights published by the Taiwan WRA.

Townships Agriculture Drinking

Aquifer Aquifer
1 2 3 4 1 2 3 4
Qishan (QS) 1.25 93.74 59.81 93.74 2.36 4.69 3.55 5.94
Maynong (MN) 1.25 1.51 2.49 1.51 2.36 28.72 20.87 17.70
Wanluan (WL) 6.12 6.12 6.12 2.68 12.04 12.04 12.04 32.73
Likang (LK) 4.41 4.35 2.49 4.35 1.19 11.71 20.87 11.71
Kaoshu (KS) 4.41 4.35 4.38 4.35 1.19 11.71 14.36 11.71
Dashu (DS) 9.25 9.25 59.81 9.25 1.85 18.54 3.55 18.54
Jiuru (JR) 4.58 9.07 9.07 9.07 0.00 0.00 0.00 0.00
Pingtung (PT) 4.58 1.47 2.33 2.33 96.58 1.67 72.15 86.12
Yanpu (YP) 6.41 9.07 9.07 9.07 18.36 2.13 2.13 2.13
Zhangzhi (ZZ) 6.41 3.16 5.88 3.16 18.36 33.90 25.43 33.90
Daliao (DL) 39.01 29.10 39.01 29.10 72.15 53.83 72.15 53.83
Wandan (WD) 3.25 3.25 2.33 2.33 0.02 0.02 72.15 86.12
Linluo (LL) 6.12 6.12 5.88 6.12 0.00 0.00 0.00 0.00
Zhutian (ZT) 6.12 6.12 1.74 6.12 0.00 0.00 0.00 0.00
Chiadong (CD) 1.31 1.90 2.34 1.90 0.00 0.00 0.00 0.00
Neipu (NP) 46.12 2.58 4.61 2.68 0.05 0.05 0.05 32.73
Shinyuan (SY) 2.30 2.58 2.30 1.37 0.00 0.00 0.00 0.00
Kanding (KD) 1.60 56.98 1.60 88.96 0.00 0.00 0.00 0.00
Chaozhou (CZ) 11.11 20.18 40.36 88.96 11.11 20.18 40.36 23.96
Nanzhou (NZ) 4.75 9.61 66.25 66.25 0.00 0.00 0.00 0.00
Xinpi (XP) 4.75 20.18 66.25 66.25 3.42 20.18 4.77 4.77
Fangliao (FL) 2.66 1.90 2.34 1.90 0.00 0.00 0.00 0.00
Linbian (LB) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Dongkang (DK) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Linyuan (LY) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

Table 5
Average responses of simulated groundwater levels (m) in townships due to stopping utilization of groundwater.

Townships Agriculture Drinking

Aquifer Aquifer
1 2 3 4 Average Ranka 1 2 3 4 Average Ranka
Qishan (QS) 3.34 0.25 0.75 0.13 1.12 21 3.51 0.57 0.75 0.65 1.37 11
Maynong (MN) 3.34 0.17 0.71 3.61 1.96 14 3.51 0.17 0.71 3.61 2.00 10
Wanluan (WL) 3.55 6.51 4.83 2.04 4.23 7 4.43 4.99 4.51 1.35 3.82 5
Likang (LK) 3.21 4.02 0.71 4.84 3.20 11 0.91 3.68 0.71 4.68 2.50 7
Kaoshu (KS) 3.21 4.02 14.33 4.84 6.60 1 0.91 3.68 14.17 4.68 5.86 2
Dashu (DS) 3.65 3.18 0.75 0.95 2.13 13 4.80 2.33 0.75 0.95 2.21 8
Jiuru (JR) 3.34 4.29 7.69 0.09 3.85 8 0.00 0.00 0.00 0.00 0.00 15
Pingtung (PT) 3.34 2.06 0.22 0.12 1.44 19 0.11 2.14 0.20 0.12 0.64 14
Yanpu (YP) 9.07 4.29 7.69 0.09 5.28 5 8.10 3.48 6.52 0.09 4.54 4
Zhangzhi (ZZ) 9.07 7.34 4.30 4.76 6.37 2 8.10 7.23 4.30 4.71 6.08 1
Daliao (DL) 1.99 1.04 0.21 0.21 0.86 22 2.20 1.09 0.21 0.21 0.93 12
Wandan (WD) 3.12 1.89 0.22 0.12 1.34 20 1.05 1.89 0.20 0.12 0.82 13
Linluo (LL) 3.64 2.35 4.30 0.13 2.61 12 0.00 0.00 0.00 0.00 0.00 15
Zhutian (ZT) 3.64 2.35 0.40 0.13 1.63 16 0.00 0.00 0.00 0.00 0.00 15
Chiadong (CD) 0.80 8.47 8.13 2.85 5.07 6 0.00 0.00 0.00 0.00 0.00 15
Neipu (NP) 0.42 0.04 3.43 2.04 1.48 18 3.42 0.04 3.34 1.35 2.04 9
Shinyuan (SY) 3.34 1.45 1.29 0.22 1.58 17 0.00 0.00 0.00 0.00 0.00 15
Kanding (KD) 2.00 1.45 1.64 1.79 1.72 15 0.00 0.00 0.00 0.00 0.00 15
Chaozhou (CZ) 0.66 9.69 2.98 1.79 3.78 9 0.66 9.69 2.98 1.14 3.62 6
Nanzhou (NZ) 3.90 0.37 6.90 3.59 3.69 10 0.00 0.00 0.00 0.00 0.00 15
Xinpi (XP) 1.24 9.69 6.90 3.59 5.36 4 1.24 9.69 6.90 3.59 5.36 3
Fangliao (FL) 2.13 8.47 8.13 2.85 5.40 3 0.00 0.00 0.00 0.00 0.00 15
Linbian (LB) 0.00 0.00 0.00 0.00 0.00 23 0.00 0.00 0.00 0.00 0.00 15
Dongkang (DK) 0.00 0.00 0.00 0.00 0.00 23 0.00 0.00 0.00 0.00 0.00 15
Linyuan (LY) 0.00 0.00 0.00 0.00 0.00 23 0.00 0.00 0.00 0.00 0.00 15
a
The ranks are from high to low levels.

aquaculture uses, therefore, the assessment of groundwater quan-
tity was used for both irrigation and aquaculture demands for
water resources. Table 6 lists the sums of the ranks obtained in
the assessment of groundwater quality and quantity. A smaller
sum represents more suitable replacement of groundwater by sur-
face water in a township. This study suggested that there are three
townships suitable for groundwater replacement for drinking pur-
poses with sums ranging from 4 to 10, five townships suitable for
groundwater replacement for irrigation purposes with sums rang-
ing from 9 to 20, and four townships of groundwater replacement
or aquaculture purposes with sums ranging from 15 to 18. Fig. 10
shows the suggested townships of groundwater replacement for
554 C.-S. Jang et al. / Journal of Hydrology 533 (2016) 541–556

Fig. 10. Townships recommended for replacement of groundwater for drinking, irrigation and aquaculture demands. The names of the townships are listed in Tables 3–6.

drinking, irrigation, and aquaculture purposes. The recommended 4. Conclusion

townships are mainly located in the southeastern study area.
Groundwater is widely extracted to meet various water
resources demands in the Pingtung Plain. To conserve water
resources, and to enhance the supply of surface water, the
Taiwanese government plans to construct several man-made lakes
or reservoirs rather than continue to withdraw groundwater. This
study establishes assessment criteria and prioritizes townships
where groundwater should be replaced with surface water for
drinking, irrigation, and aquaculture purposes, where surface
water in planned man-made lakes and reservoirs is still relatively
limited. The research results reveal that six southeastern town-
ships with poor groundwater quality and fast recovery effects are
the most suitable locations where surface water should be
supplied.
Overexploitation of groundwater is common in the Pingtung
Plain, and can easily lead to the occurrence of severe seawater
intrusion and land subsidence. To assist in the planning of a sustain-
able and reasonable scheme for groundwater and surface water use,
this study seeks to determine the most suitable locations where
surface water could be used instead of groundwater after the con-
struction of man-made lakes or reservoirs in terms of the integra-
tion of groundwater quality and quantity. The spatial variability
of groundwater quality is analyzed using MVIK. The responses in
groundwater quantity due to the cessation of groundwater pump-
ing for drinking, irrigation, and aquaculture are rated using the
numerical model of groundwater flow. The findings indicated that
the ranking obtained from assessments through combining analysis
 C.-S. Jang et al. / Journal of Hydrology 533 (2016) 541–556 555

Table 6
The sums of the ranks and townships recommended for replacement of groundwater for various demands of water resources based on the assessment of combining groundwater
quality and quantity.

Townships Drinking Irrigation Aquaculture Demands of water resources

Qishan (QS) 18 46 38 –
Maynong (MN) 18 38 32 –
Wanluan (WL) 10 19 20 Drinking and irrigation
Likang (LK) 19 34 33 –
Kaoshu (KS) 15 23 25 –
Dashu (DS) 11 28 24 –
Jiuru (JR) 24 28 29 –
Pingtung (PT) 18 35 33 –
Yanpu (YP) 18 26 30 –
Zhangzhi (ZZ) 12 21 25 –
Daliao (DL) 13 24 24 –
Wandan (WD) 14 28 28 –
Linluo (LL) 21 29 32 –
Zhutian (ZT) 17 30 31 –
Chiadong (CD) 16 13 16 Irrigation and aquaculture
Neipu (NP) 19 36 37 –
Shinyuan (SY) 16 21 20 –
Kanding (KD) 16 24 21 –
Chaozhou (CZ) 7 22 18 Drinking and aquaculture
Nanzhou (NZ) 16 20 15 Irrigation and aquaculture
Xinpi (XP) 4 15 16 Drinking, irrigation and aquaculture
Fangliao (FL) 16 9 19 Irrigation
Linbian (LB) 16 28 30 –
Dongkang (DK) 16 26 27 –
Linyuan (LY) 16 24 24 –

of the groundwater quality and numerical flow simulation is useful
for planning the regional development of water resources. Six
southeastern townships with poor groundwater quality and signif-
icant increases in groundwater levels are deemed optimal locations
for supplying surface water from man-made lakes and reservoirs
instead of from groundwater abstraction. The research findings
can also provide Taiwanses government administrators with
detailed information on how to sustainably use and conserve
groundwater in overexploited areas.
Acknowledgements
The authors would like to thank the Taiwan Water Resources
Agency for generously providing the data on groundwater quality
parameters, groundwater levels and groundwater rights in the
Pingtung Plain, and the Ministry of Science and Technology, Tai-
wan for financially supporting this research under Contract No.
NSC 102-2116-M-242-001.
References
AERC (Agriculture Engineering Research Center), 2009. Survey, Analysis and
Assessment of Groundwater Quality in Taiwan Areas in 2009. Water
Resources Agency, Ministry of Economic Affairs, Executive Yuan, Taiwan, pp.
100–102.
AERC (Agriculture Engineering Research Center), 2010. Survey, Analysis and
Assessment of Groundwater Quality of Taiwan in 2010. Water Resources
Agency, Ministry of Economic Affairs, Executive Yuan, Taiwan, pp. 151–158.
AERC (Agriculture Engineering Research Center), 2011. Survey, Analysis and
Assessment of Groundwater Quality of Taiwan in 2011. Water Resources
Agency, Ministry of Economic Affairs, Executive Yuan, Taiwan, pp. 191–203.
AERC (Agriculture Engineering Research Center), 2012. Survey, Analysis and
Assessment of Groundwater Quality of Taiwan in 2012–2014 (1/3). Water
Resources Agency, Ministry of Economic Affairs, Executive Yuan, Taiwan, pp.
85–101.
Al-Salamah, L.S., Ghazaw, Y.M., Ghumman, A.R., 2011. Groundwater modeling of
Saq Aquifer Buraydah Al Qassim for better water management strategies.
Environ. Monitor. Assess. 173 (1–4), 851–860.
Bauder, T.A., Waskom, R.M., Davis, J.G., 2007. Irrigation Water Quality Criteria.
<http://www.ext.colostate.edu/pubs/crops/00506.html>.
D’Alessandro, W., Bellomo, S., Parello, F., Brusca, L., Longo, M., 2008. Survey on
fluoride, bromide and chloride contents in public drinking water supplies in
Sicily (Italy). Environ. Monitor. Assess. 145 (1–3), 303–313.
Deutsch, C.V., Journel, A.G., 1998. GSLIB: Geostatistical Software Library and User’s
Guide, second ed. Oxford University Press, New York.
Diodato, N., Ceccarelli, M., 2004. Multivariate indicator kriging approach using a GIS
to classify soil degradation for Mediterranean agricultural lands. Ecol. Indicators
4 (3), 177–187.
Eaton, F.M., 1950. Significance of carbonates in irrigation water. Soil Sci. 69,
123–133.
Goovaerts, P., 1997. Geostatistics for Natural Resources Evaluation. Oxford
University Press, New York, pp. 259–368.
Grieve, C.M., Poss, J.A., Amrhein, C., 2006. Response of Matthiola incana to irrigation
with salinity wastewaters. Hortscience 41 (1), 119–123.
Halvorson, J.J., Smith, J.L., Papendick, R.I., 1996. Integration of multiple soil
parameters to evaluate soil quality: a field example. Biol. Fertility Soils 21 (3),
207–214.
Harbaugh, A.W., McDonald, M.G., 1996. Programmer documentation for
MODFLOW-96, an update to the U.S. Geological Survey modular finite-
difference groundwater flow model. USGS Open-File Report, pp. 96–486.
Hill, M.C., Tiedeman, C.R., 2007. Effective Groundwater Model Calibration: With
Analysis of Data, Sensitivities, Predictions, and Uncertainty. John Wiley & Sons
Inc., USA, p. 229.
Huq, S.M.I., Naidu, R., 2005. Arsenic in groundwater and contamination of the food
chain: Bangladesh scenario. Natural Arsenic in Groundwater Occurrence,
Remediation and Management. In: The 32nd International Geological
Congress, Florence, Italy, pp. 95–101.
Jang, C.S., Liu, C.W., Lin, K.H., Huang, F.M., Wang, S.W., 2006. Spatial analysis of
potential carcinogenic risks associated with ingesting arsenic in aquacultural
tilapia (Oreochromis mossambicus) in blackfoot disease hyperendemic areas.
Environ. Sci. Technol. 40, 1707–1713.
Jang, C.S., Chen, J.S., 2009. Probabilistic assessment of groundwater mixing with
surface water for agricultural utilization. J. Hydrol. 376 (1–2), 188–199.
Jang, C.S., Chen, S.K., Kuo, Y.M., 2012a. Establishing an irrigation management plan
of sustainable groundwater based on spatial variability of water quality and
quantity. J. Hydrol. 414–415, 201–210.
Jang, C.S., Liu, C.W., Chou, Y.L., 2012b. Assessment of groundwater emergency
utilization in Taipei basin during drought. J. Hydrol. 414–415, 405–412.
Jang, C.S., Liang, C.P., Wang, S.W., 2013. Integrating the spatial variability of
water quality and quantity to probabilistically assess groundwater
sustainability for use in aquaculture. Stochastic Environ. Res. Risk Assess. 27
(6), 1281–1291.
Knobeloch, L., Salna, B., Hogan, A., Postle, J., Anderson, H., 2000. Blue babies and
nitrate-contaminated well water. Environ. Health Perspect. 108 (7), 675–678.
Lee, J.J., Liu, C.W., Jang, C.S., Liang, C.P., 2008. Zonal management of multi-purpose
use of water from arsenic-affected aquifers by using a multi-variable indicator
kriging approach. J. Hydrol. 359 (3–4), 260–273.
Liao, C.M., Jau, S.F., Chen, W.Y., Lin, C.M., Jou, L.J., Liu, C.W., Liao, V.H.C., Chang, F.J.,
2008. Acute toxicity and bioaccumulation of arsenic in freshwater clam
Corbicula fluminea. Environ. Toxicol. 23 (6), 702–711.
Liu, C.W., Wang, S.W., Jang, C.S., Lin, K.H., 2006. Occurrence of arsenic in
groundwater of the Choshui river alluvial fan, Taiwan. J. Environ. Quality 35,
68–75.
556 C.-S. Jang et al. / Journal of Hydrology 533 (2016) 541–556
Liu, C.W., Chou, Y.L., Lin, S.T., Lin, G.J., Jang, C.S., 2010. Management of high
groundwater level aquifer in the Taipei Basin. Water Resour. Manage. 24 (13),
3513–3525.
Mittelstet, A.R., Smolen, M.D., Fox, G.A., Adams, D.C., 2011. Comparison of aquifer
sustainability under groundwater administrations in Oklahoma and Texas. J.
Am. Water Resour. Assoc. 47 (2), 424–431.
Oyedele, D.J., Amusan, A.A., Obi, A.O., 1996. The use of multiple-variable indicator
kriging technique for assessment of the suitability of an acid soil for maize.
Tropical Agric. 73 (4), 259–263.
Poeter, E.P., Hill, M.C., 1998. Documentation of UCODE: A computer code for
universal inverse modeling. U.S. Geological Survey, Water Resources
Investigations Report 98-4080, USA.
Rejani, R., Jha, M.K., Panda, S.N., Mull, R., 2008. Simulation modeling for efficient
groundwater management in Balasore coastal basin, India. Water Resour.
Manage. 22 (1), 23–50.
Smith, J.L., Halvorson, J.J., Papendick, R.I., 1993. Using multiple-variable indicator
kriging for evaluating soil quality. Soil Sci. Soc. Am. J. 57, 743–749.
Taiwan CGS (Central Geological Survey), 2002. Hydrogeological Survey Report of
Pingtung Plain, Taiwan. Central Geological Survey, Ministry of Economic Affairs,
Executive Yuan, Taiwan, pp. 97–142.
Taiwan DGBAS (Directorate-General of Budget, Accounting and Statistics), 2005.
The Inquiry System of Agriculture, Forestry, Fishery and Pasturage Data in
Taiwan Townships in 2005. Directorate-General of Budget, Accounting and
Statistics, Executive Yuan, Taiwan <http://win.dgbas.gov.tw/agr/ics_main.asp>.
Taiwan WRA (Water Resources Agency), 2012. Application Network of Hydrological
Data. Water Resources Agency, Ministry of Economic Affairs, Executive Yuan,
Taiwan <http://gweb.wra.gov.tw/HydroApplication/index.aspx>.
Ting, C.S., Zhou, Y., de Vries, J.J., Simmers, I., 1998. Development of a preliminary
ground water flow model for water resources management in the Pingtung
Plain, Taiwan. Ground Water 35 (6), 20–36.
Tsai, J.W., Liao, C.M., Liao, V.H.C., 2006. A biologically based damage assessment
model to enhance aquacultural water quality management. Aquaculture 251
(2–4), 280–294.
USSL, 1954. Diagnosis and Improvement of Saline and Alkali Soils. USDA Hand Book,
60, Washington DC, pp. 147.
Yang, C.Y., Cheng, M.F., Tsai, S.S., Hsieh, Y.L., 1998. Calcium, magnesium, and nitrate
in drinking water and gastric cancer mortality. Cancer Sci. 89, 124–130.
Ye, C.X., Tian, L.X., Yang, H.J., Liang, J.J., Niu, J., Liu, Y.J., 2009. Growth performance and
tissue mineral content of juvenile grouper (Epinephelus coioides) fed diets
supplemented with various levels of manganese. Aquacult. Nutr. 15 (6), 608–614.