Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Journal of Hydrology 399 (2011) 158–172

Contents lists available at ScienceDirect

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

GIS mapping of regional probabilistic groundwater potential in the area


of Pohang City, Korea
Hyun-Joo Oh a, Yong-Sung Kim b,c, Jong-Kuk Choi d, Eungyu Park b, Saro Lee a,⇑
a
Geoscience Information Center, Korea Institute of Geoscience & Mineral Resources (KIGAM), 92 Gwahang-no, Yuseong-gu, Daejeon 305-350, Republic of Korea
b
Geology Dept., Kyungpook National University, 1370 Sangyeok-dong, Buk-gu, Daegu 702-701, Republic of Korea
c
Yooshin Engineering Corporation, 832-40 Yeoksam-dong, Gangnam-gu, Seoul 135-936, Republic of Korea
d
Korea Ocean Satellite Centre, Korea Ocean Research & Development Institute, 454 Haean-no, Sangrok-gu, Ansan, Gyeonggi 426-744, Republic of Korea

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

Article history: This study analyzed the relationships between groundwater specific capacity (SPC) and its related
Received 16 July 2010 hydrological factors to assess the sensitivity of each factor and map the regional groundwater potential
Received in revised form 14 December 2010 for the area of Pohang City, Korea, using a geographic information system (GIS) and a probability model.
Accepted 24 December 2010
All related factors including topography, geology, lineament, and soil data were collected and entered
Available online 30 December 2010
into a spatial database. SPC data were collected from well locations, and SPC values of P6.25 m3/d/m,
This manuscript was handled by G. Syme, corresponding to a yield of 500 m3/d, were input to a spatial database. SPC data were then randomly
Editor-in-Chief, with the assistance of Craig selected in a 66/34 ratio to train and validate the model. A frequency-ratio model and sensitivity analysis
T. Simmons, Associate Editor were used to determine the relationships between SPC and its related factors and the importance of SPC-
related factors. Sensitivity analysis allows for comparison of the combined effects of all factors except for
Keywords: one. The validation of the groundwater potential map overlain by all factors showed 77.78% accuracy. In
Groundwater potential the sensitivity analysis, the best accuracy was obtained by omitting ground elevation data (78.64%), and
Groundwater exploration the worst accuracy resulted when soil texture was not included (76.64%). The results show that soil tex-
Frequency ratio ture had the greatest effect on the groundwater potential and ground elevation had the least effect. Such
Sensitivity analysis information and the maps generated from it can be applied to groundwater management and groundwa-
GIS
ter resource exploration.
Korea
Ó 2010 Elsevier B.V. All rights reserved.

1. Introduction conditions, and the interrelationships among these factors


(Greenbaum, 1992; Mukherjee, 1996; Roy, 1996). However, con-
In Korea, domestic water consumption has increased rapidly ventional exploration methods such as field-based hydrogeological
over the last decade. The increased demand for high-quality water, and geophysical resistivity surveys do not always account for the
combined with the expected pressures of global climatic change, diverse factors that control the occurrence and movement of
create an urgent need for quantitative methodologies by which groundwater. To clarify the relationships between hydrogeological
to evaluate the groundwater production of aquifers. However, factors and groundwater productivity, such as transmissivity (T),
sound tools for assessing aquifer productivity have not been well specific capacity (SPC) and yield, regional groundwater potential
established because of the relatively short history of investigative should be assessed and mapped.
research in this field. The development of a reasonable model for Remote sensing and geographic information system (GIS) tech-
groundwater potential is therefore crucial for future systematic nologies have great potential for use in groundwater hydrology.
developments, efficient management, and sustainable use of GIS is a powerful tool for handling spatial data and decision making
groundwater resources. in several areas, including geological and environmental fields
In general, the occurrence and movement of groundwater, espe- (Stafford, 1991; Goodchild, 1993). Remote sensing is one of the
cially in fractured bedrock aquifers, in a given area is governed by main sources of information about surface features related to
factors such as topography, lithology, geological structures, frac- groundwater such as lineament, land use, and landforms. Such
ture density, aperture and connectivity, secondary porosity, information can be easily input to a GIS environment for integra-
groundwater table distribution, groundwater recharge, slope, tion with other types of data, followed by analysis (Faust et al.,
drainage pattern, landforms, land use and land cover, climatic 1991; Hinton, 1996; Jha et al., 2007).
Many studies have used GIS and remote sensing techniques
⇑ Corresponding author. Tel.: +82 42 868 3057; fax: +82 42 868 3413. along with thematic layers such as geomorphology, drainage pat-
E-mail address: leesaro@kigam.re.kr (S. Lee). tern, lineament, lithology, and soil (Chowdhury et al., 2009; Jaiswal

0022-1694/$ - see front matter Ó 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhydrol.2010.12.027
H.-J. Oh et al. / Journal of Hydrology 399 (2011) 158–172 159

et al., 2003; Prasad et al., 2008; Saha et al., 2010; Saraf and dependent variable and independent variables, including multi-
Choudhury, 1998; Solomon and Quiel, 2006). Some studies have classified maps. The frequency-ratio model is easy to apply in a
applied probabilistic models such as multi-criteria decision analy- GIS. The relationships revealed between specific capacity and the
sis and weights-of-evidence modeling for groundwater potential reclassified hydrogeological factors make it easier to quantitatively
mapping (Chenini et al., 2010; Corsini et al., 2009; Gupta and calculate and understand the groundwater productivity potential
Srivastava, 2010; Murthy and Mamo, 2009). Other studies have for the class of each factor. Although the frequency-ratio model
used personal judgments or local information to assign weight to has been applied to examine landslide susceptibility (Lee and Choi,
different thematic layers and their features (Kumar et al., 2009; 2003; Lee and Dan, 2005; Oh et al., 2009, 2010; Pradhan and Lee,
Madrucci et al., 2008; Mondal et al., 2008; Nagarajan and Singh, 2010; Pradhan et al., 2010) and ground subsidence hazard map-
2009; Pradhan, 2009; Yeh et al., 2009). More sophisticated assess- ping using GIS and remote sensing (Kim et al., 2006; Lee et al.,
ments have conducted numerical modeling, decision trees, fuzzy 2010), this approach has not yet been used to delineate groundwa-
logic, and analytic hierarchy process analysis (Chenini and Ben ter potential. Moreover, most previous reports on frequency-ratio
Mammou, 2010; Chowdhury et al., 2009; Ghayoumian et al., modeling have provided little or no information about the sensitiv-
2007; Srivastava and Bhattacharya, 2006). Some researchers have ity or quality of this approach. Technique-related factors could af-
also integrated GIS, remote sensing, and geophysical surveys to fect model output with respect to: (1) output changes with change
derive additional thematic layers of surface parameters such as in model inputs and (2) parameters that significantly influence the
resistivity, aquifer thickness, or fault maps (Israil et al., 2006; model predictions.
Kumar et al., 2009; Ranganai and Ebinger, 2008; Srivastava and
Bhattacharya, 2006). However, such studies had limitations be-
cause they used indirect indicators such as yield, groundwater 2. Study area and hydrogeological setting
depth, resistivity, and spring location, rather than hydraulic con-
stants such as transmissivity (T) and specific capacity (SPC). Fur- The study focuses on the Pohang City area of Korea. This area
thermore, past studies generally produced small datasets and did has experienced rapid population growth and increased demand
not validate the results by comparison with other datasets. for groundwater reserves and is thus appropriate for evaluating
In this study, the groundwater potential of fractured bedrock groundwater potential. The site lies between 35°500 0700 N and
was analyzed and validated using a GIS that was based on a fre- 36°160 3400 N latitude and 129°050 3100 E and 129°340 5700 E longitude
quency-ratio model with training and validation datasets of spe- (Fig. 1) and covers 891.44 km2 on a digital topographic map with
cific capacity. Groundwater potential maps were used to assess a 1:5000 scale (Table 1). Elevation in the area ranges from 0 to
the sensitivity of each factor. 919.3 m above sea level (a.s.l.), with an average of 144.2 m (stan-
A frequency-ratio model can provide a simple geospatial assess- dard deviation = 143.64 m). The terrain gradient computed from
ment tool to calculate the probabilistic relationship between the a 30 m  30 m digital elevation model (DEM) extracted from the

KOREA

Fig. 1. Study area.


160 H.-J. Oh et al. / Journal of Hydrology 399 (2011) 158–172

Table 1 mountain system in the study area trends mostly N10°–20°E par-
Geographic information of study area. allel to Yangsan fault zone; the other small systems trend N60°–
Geographic information Study area N70° W parallel to faults that deviate from the Yangsan fault.
Topography Coastal hills in Northeastern area Precipitation is most abundant from June to September. The mean
Rugged mountain range in northwestern area annual rainfall between 1973 and 2009 was 1119 mm.
Location 35°500 0700 N–36°160 3400 N The groundwater table was recorded monthly at 431 wells from
129°050 3100 E–129°340 5700 E October 2002 to November 2003, and the annual mean groundwa-
Area 891.44 km2 ter table was in April 2003 according to time-series analysis. Fig. 2
Elevation shows the distribution of groundwater features in April 2003 such
Range 0–919.3 m as groundwater table elevation and depth. The groundwater table
Mean 144.2 m map (Fig. 2a) was created by universal cokriging analysis. The
Std 143.64 m
groundwater table depth (Fig. 2b) was mapped by subtraction of
Gradient the groundwater elevation from the DEM.
Range 0–50.4°
Mean 14.4°
The bedrock geology of the study area consists mainly of Creta-
Std 10.70° ceous sedimentary, granitic, and volcanic rocks; Tertiary granitic,
volcanic, and sedimentary rocks; and Quaternary basalt (Hwang
et al., 1996; Kim et al., 1998). The study area has two major faults:
1:5000-scale topographic map ranges from 0° to 50.4° with a mean the Yangsan and Ulsan faults that trend N10°–20°E and N10°–
value of 14.4° and a standard deviation of 10.70°. The study area 20°W. In the study area, Cretaceous volcanic and granitic rocks
was divided into two areas. The northeastern area consists of are distributed mainly in the west, Tertiary sedimentary in the
coastal hills produced by Tertiary deposition, and the northwestern east, and Tertiary volcanic and granitic rocks in the south.
area is a rugged mountain range consisting of volcanic rock and To analyze relationships between groundwater productivity
granite. The Taebaek mountain range (Mt. Bade, 645.0 m a.s.l.; and hydrogeological factors such as land cover, topology, geology,
Hyangnobong, 929.9 m a.s.l.; Satgatbong, 718.0 m a.s.l.; and Mt. groundwater table distribution, and groundwater recharge, the
Bihak, 762.0 m a.s.l.) lies along the Yangsan fault zone. The main first stage was data collection and construction of the spatial

Fig. 2. Groundwater table distribution of April 2003: (a) is groundwater table elevation by universal cokriging analysis and (b) is groundwater table depth by subtraction GW
table elevation from DEM.
H.-J. Oh et al. / Journal of Hydrology 399 (2011) 158–172 161

Fig. 2 (continued)

database from which relevant factors were extracted (Table 2). geospatial dataset (hydrogeological factors, see Table 2). In this pa-
Groundwater productivity data such as transmissivity, specific per, the specific capacity was extracted from well location; thus, the
capacity, yield, well depth, well diameter, and groundwater table probability was represented by specific capacity potential values or
were collected from the national groundwater survey in the Po- the groundwater potential value. To use a probabilistic model,
hang area (MLTM, 2003), the national groundwater monitoring groundwater productivity data were converted into binary form,
network construction report (MLTM, 1995, 1998, 2001), and the where success was represented by a 1 and failure was represented
rural groundwater survey report (MIFAFF, 1985–2005). by a 0. The split criteria was the median value of the transmissivity,
3.79 m2/d, and its corresponding specific capacity value, 6.25 m3/d/
3. Methodology and spatial data m, was calculated from the relationship between transmissivity (T)
and specific capacity (SPC) T = 0.9948  SPC0.7293. The mean specific
3.1. Methodology capacity was 16.54 m3/d/m (median 6.73 m3/d/m) with a maximum
of 111.14 m3/d/m and a minimum of 1.42 m3/d/m.
The methodology is summarized in Fig. 3. After the study area Groundwater productivity data were randomized, but their sta-
was selected, groundwater productivity data, such as specific capac- tistical properties were retained and divided into a training dataset
ity as derived from pumping tests, were collected and outliers were and a validation dataset based on a 70/30 ratio as well as expert
rejected (Fig. 4). Eighty-three measurements of specific capacity knowledge due to the limited data available (Carranza and Hale,
were collected. Pumping tests are expensive and time-consuming, 2002). The training dataset contained 55 (64%) of the well points,
and thus the specific capacity data are somewhat lacking; however, and the validation dataset contained 28 (34%) of the well points.
this limitation can be overcome by using a probabilistic model with To represent the groundwater potential areas and the sensitiv-
hydrogeological factors at high resolution. These groundwater pro- ity of each hydrogeological factor to the groundwater potential
ductivity data are set as dependent variables, and various hydrogeo- map quantitatively, this study used the probability–frequency ra-
logical factors, which are known to influence groundwater tio and sensitivity analysis. The frequency ratio is the probability
productivity, are set as independent valuables. Using the fre- of occurrence of a certain attribute (Bonham-Carter, 1994). If we
quency-ratio model, the probability was calculated from the create an event E and certain factors attributed to F, the probabil-
groundwater productivity data located in each range or class of each ity-frequency ratio of F is the ratio of the conditional probability:
162 H.-J. Oh et al. / Journal of Hydrology 399 (2011) 158–172

Table 2
Data layer related to groundwater productivity of study area.

Category Hydrogeological factors (unit) Description Data


type
Topologya Ground elevation (m) Ground elevation obtained from DEM (Res. 30 m) for local ground elevation property. Negative Grid
relationship, in general
Ground elevation within 300 m (m) Mean ground elevation within R = 300 m from DEM for regional ground elevation property.
Negative relationship, in general
Mean ground elevation within Mean ground elevation within watershed area from DEM for hydrological regional ground
watershed area (m) elevation property. Negative relationship, in general
Ground slope (%) Ground slope obtained from DEM (Res. 30 m) for local ground slope property. Negative
relationship, in general
Mean ground slope within watershed Mean ground slope within watershed area from DEM for hydrological regional ground slope
area (%) property. Negative relationship, in general
Curvature (unit less) Ground curvature from DEM (Res. 30 m) for local ground curvature property. Negative
relationship, in general
TWI (unit less) Topographic wetness index from DEM (Res. 30 m) for local ground wetness property. Positive
relationship, in general
River density (km/km2) Density of river line feature for surface drainage property. Positive relationship, in general
Distance from river (m) Distance from river line feature for water supply from river. Negative relationship, in general
Cumulative watershed area (km2) Watershed area with accumulation through flow direction for water body scale. Positive
relationship, in general
Landsat-TMb Lineament length density (km/km2) Density of lineament length. Positive relationship, in general Line
Lineament length density weighted by its Density of lineament length weighted by its length. Positive relationship, in general
length (km/km2)
Lineament frequency weighted by its Density of lineament frequency weighted by its length. Positive relationship, in general
length (count  km/km2)
Geologyc Hydrogeological units (unit less) Hydrogeological units classified from bedrock lithology by similar hydrological property Polygon
Soild Soil texture (unit less) Soil texture property for groundwater infiltration Polygon
a
Topological factor was extracted from digital topographic map (1:5,000 scale) by National Geographic Information Institute (NGII; http://www.ngii.go.kr).
b
The Landsat TM image (Path 144/Row 25) of study area was acquired on 3–2003.
c
The geological map (1:50,000 scale) produced by the Korea Institute of Geoscience & Mineral Resource (KIGAM; http://www.kigam.re.kr).
d
The detailed soil map (1:25,000 scale) produced by Rural Development Administration (RDA; http://www.rda.go.kr).

PfEjFg ¼ ðPfE \ FgÞ=PfFg ð1Þ total number of input factors. The groundwater potential index
(GPI) was used for mapping the groundwater potential.
Fig. 5 presents the steps for calculating the frequency ratio,
Finally, the groundwater potential was mapped with the fre-
namely: (a) finding well locations, (b) representing cells of class
quency ratio, and the sensitivity analysis was validated using exist-
1 for factor A, and (c) describing the area of spatial overlap for well
ing specific capacity locations that were not used to train the
areas and areas of class 1 for factor A. In the present example, in
model. The effect of each factor in the sensitive analysis was eval-
which the study area comprises 25 cells, 5 well cells and 13 cells
uated by validation of the groundwater potential maps that ex-
are of class 1 for factor A. Three well cells are also cells of class 1
cluded each factor.
for factor A. The percentages for cell areas with respect to class 1
for factor A and the entire domain are 60% and 52%, respectively.
Therefore, the frequency ratio of class 1 for factor A is 60%/ 3.2. Spatial database
52% = 1.15.
In determining the probability–frequency ratio, the area ratio In general, groundwater productivity is governed by many
for specific capacity occurrence was calculated for the range or hydrogeological factors, such as surface and bedrock lithology,
type of each factor, and the area ratio for the range or type of each structure, slope steepness and morphology, stream evolution, cli-
factor to the total area was calculated. Finally, probability–fre- mate, soil, vegetation cover, land use, and human activity, but
quency ratios for the range or type of each factor were calculated these relationships have not been verified statistically and quanti-
by dividing the specific capacity–occurrence ratio by the area ratio. tatively. In this study, the 27 hydrogeological factors expected to
The frequency ratio was assigned to each factor’s class. The fre- be related to groundwater potential were reviewed with transmis-
quency ratio of the groundwater potential was created using the sivity, specific capacity, and yield. Finally, 15 hydrogeological fac-
overlay functions in the GIS, which were used to merge different tors (Table 2) were selected and applied to the probabilistic model.
factors that were assigned the ratio. The spatial database was constructed with a resolution of 30 m
A sensitivity analysis shows how a solution changes when the by 30 m on the basis of Landsat TM. The 15 hydrogeological factors
input factors are changed. If the selected factor results in a rela- (Table 2) were converted to ArcGIS grid format, and the GRID set
tively large change in the outcome, then the outcome is said to comprised 1822 rows by 1787 columns. In the study area, the total
be affected by that factor. The factors that have the greatest impact number of cells was 990,495 and specific capacity in 83 cells (55
on the calculated groundwater potential map can therefore be cells for training and 28 cells for validation).
identified using sensitivity analysis. In this study, the sensitivity
analysis was conducted by excluding each factor in turn during
3.3. Topographic data
the frequency ratio summation stage:

GPI SEN i ¼ ðFr all Þ  Fri ði ¼ 1; 2; . . . ; nÞ ð2Þ A Triangulated Irregular Network (TIN) was made using the ele-
vation value, and a DEM was made with 30 m  30 m resolution
where GPI_SENi is the groundwater potential index of an omitted after elimination of internal drainage (‘‘sink area’’) in the elevation
factor using sensitivity analysis, Frall is the sum of frequency ratios grid using the FILL command in ArcGIS 9.0. One input factor, the
of all factors, and Fri is the frequency ratio of a factor. Here n is the ground elevation within 300 m, was given a mean value at each
H.-J. Oh et al. / Journal of Hydrology 399 (2011) 158–172 163

Construction of database

Dependent var.
Specific capacity (SPC) corresponding yield 500 m3/d

Training set (55 well locations) Random Validation set


Partition
Indep. var. 0 : 22 well of SPC < 6.25 m3/d/m 1 : 32 well of SPC ≥ 6.25 m3/d/m (28 well locations)

15 spatial factors relevant to SPC


Geographic Information System (GIS)

Topographic, geological, lineament and soil dataset

Groundwater potential mapping Validation

Frequency ratio (FR) Sensitive analysis (SA) Area under curve (AUC)

Overlay analysis of 15 Overlay analysis of 14


factors assigned FR value factors assigned FR value x-axis : Groundwater
potential index rank (%) AUC
(Groundwater potential map Artificial neural networks(76.09%)

minus each factor)


y-axis : Cumulative % of well location
15 groundwater potential maps for SPC ≥ 6.25 m3/d/m

Comparison of groundwater
potential map

1 groundwater potential map


by FR (used 15 factors)
Groundwater potential map Vs.

15 groundwater potential maps


by FR (used 14 factors)

Fig. 3. Flow chart of the study.

grid cell based on the value of neighboring cells within a 300-m ra- and intersections of lineaments reflect rock structures through
dius. Using the DEM, the slope, curvature, and Topographic Wet- which water can percolate and travel up to several kilometers
ness Index (TWI) were all calculated. The mean ground elevation within. Depending upon the terrain, the lineament may be a zone
and slope within watershed area were obtained after watershed of influence. To predict ground control problems in underground
delineation. The river density and distance from the river were structures (Kane et al., 1996), lineament is strongly related to dis-
considered as the prime indicators for selection of groundwater continuities such as joints, faults, and folds. For these reasons, lin-
potential sites because they indirectly indicate the permeability eament was used for structural analysis, analysis of the
and porosity of the terrain. relationship with lithology, and assessment of groundwater pro-
ductivity. Lineament identification from satellite images could be
3.4. Hydrogeology and lineament data considered a subjective process (Tam et al., 2004; Mabee et al.,
1994). However, a totally objective system that incorporates image
The study area has 24 geological units of bedrock. With the processing analysis to distinguish natural linear features in remo-
exception of alluvial units, these units were classified into six geo- tely sensed data without involvement of human judgment is not
logical groups which have similar hydrogeological properties, such possible (Teeuw, 1995). In this study, lineament was detected
as porous volcanics (P-Volc), semi-consolidated sediments 1 (SC- through interpretation of Landsat TM imagery and hillshade maps
Sed 1) and 2 (SC-Sed 2), nonporous volcanics (NP-Volc), intrusives from a DEM made by a structural geologist with much experience
(Intrus1), and clastic sediments (Figs. 6 and 7a). These hydrogeo- in such interpretations (Fig. 7b). To eliminate errors resulting from
logical units are identified by groundwater productivity statistics, scale deviation and sun illumination, the DEM data were used. The
spatial distribution, and stratigraphic history. Fig. 6 shows a box hillshade maps from the DEM were used for detecting the linea-
plot of transmissivity values for each hydrological unit. Transmis- ment. In the hillshade maps, the sun altitude is 45°, and the sun
sivity data were not available in the area of clastic sediments. azimuths are 0°, 45°, 90°, 135°, 180°, 225°, 270°, and 315°. We then
The semi-consolidated sediments 1, consisting predominantly of overlaid the Landsat TM image onto the hillshade maps, and the
sandstone and conglomerate, are overlain by the semi-consoli- lineament was detected. Finally, we used the digital topographic
dated sediment 2, consisting mainly of mudstone and shale. map with a scale of 1:5000 to exclude drainage and roads, both
In an image, a lineament is defined as a straight or slightly of which could be detected as lineament. There are 268 final linea-
curved surface feature of natural origin, interpreted directly from ments, which are shown in Fig. 7b. The total length of the linea-
the imagery (O’Leary et al., 1976; Koike et al., 1998). Lineaments ments is 877 km, the shortest lineament is 0.1 km, the longest
164 H.-J. Oh et al. / Journal of Hydrology 399 (2011) 158–172

Fig. 4. Eighty-three well locations and value range of specific capacity (SPC) in the study area.

(a) Well location (b) Class 1 of factor A (c) Well for class 1 of factor A

1. Total number of well cells in the study area : 5


2. Percentage of well area for class 1 of a factor : (3/5) x 100 = 60 %
3. Total number of cells of factor A : 25
4. Number of cells for class 1 of factor A : 13
5. Percentage of domain : 13/25 x 100 = 52 %
6. Frequency ratio for class 1 of factor A : 60 % / 52% = 1.15

Fig. 5. Diagram showing the processes for calculation of frequency ratio values.
H.-J. Oh et al. / Journal of Hydrology 399 (2011) 158–172 165

watershed, ground slope, mean ground slope within watershed


and curvature) were excluded for the topographic wetness index
(TWI) to explore the possibility of a negative correlation between
the factors and specific capacity. Such a negative correlation indi-
cates a higher groundwater potential with a lower elevation, gentler
slope, and flatter curvature. In general, flow follows the groundwater
slope from areas of high potential to areas of low potential. The for-
mer occurs where water is recharged to a groundwater system, and
the opposite occurs when water is taken from a groundwater sys-
tem. Slope plays a significant role in infiltration versus runoff. Infil-
tration is inversely related to slope, such that the gentler the slope,
the more infiltration and the less runoff and vice versa when the
slope is steeper (Sarkar et al., 2001). Mathematically, curvature is de-
fined as the change in slope angle along a very small arc of the curve
(Thomas, 1968). All the low slope areas almost have plan-curvature
values near zero. As a result, a higher elevation, steeper slope, and
convex–concave curvatures could produce a higher rainfall–runoff
rate and lower infiltration, thus possibly producing a lower ground-
water potential. A lower elevation, gentler slope, and plan slope
could produce the opposite result. TWI (Beven and Kirkby, 1979) re-
lates upslope areas as a measure of water flowing towards a certain
point, to the local slope, which is a measure of the subsurface lateral
transmissivity. A higher TWI value indicates a lower slope and a lar-
ger slope area. With higher TWI, groundwater potential could there-
Fig. 6. Box plot of transmissivity by classifying hydrogeological units from bedrock
lithology, stratigraphic history and spatial distribution. fore also be higher because the conditions are favorable for
groundwater recharge. There is a positive correlation between TWI
and SPC as shown in Table 3.
Rivers may be categorized according to discharge, for which
lineament is 25.2 km, and the average lineament length is 3.1 km. width, depth, and velocity are the major variables. These variables
Lineament density is useful for understanding the local distribu- form the basis of the hydraulic geometry of a single channel river
tion of lineaments. Lineament density analysis was considered to (Leopold et al., 1964). Thus, river density and distance from river
be the lineament frequency and length, and the density of linea- were considered factors related to the groundwater potential anal-
ment length and frequency were also analyzed using the weight ysis in this study (Table 3). As a result, in the case of river density
of lineament length in the study area. without agricultural irrigation, the relationship between the den-
sity and SPC could be positive. However, distance from river
3.5. Soil data showed a negative correlation. These results indicate that the clo-
ser and denser the rivers, the greater the groundwater potential.
As land surface data, the soil was used as a factor related to The presence of a watershed is related to topographic forms
groundwater potential. Soil texture invariably controls the penetra- that concentrate runoff. Surface runoff from the watershed is pro-
tion of surface water into an aquifer system. The soil texture of the duced by groundwater seepage and by water falling onto the land
study area was generated by a 1:25000 scale soil map published by surface (Reid, 1993). Regarding the relationship between cumula-
National Institute of Agricultural Science and Technology. Fourteen tive watershed area and SPC, the cumulative watershed area could
different categories of soil were extracted from the soil map: forest, reveal a positive correlation between the factor and SPC, which
grassland, gravel, gravelly sandy loam, fine gravelly sandy loam, indicates a higher groundwater potential over an increasing value
sandy loam, loamy fine sand, fine sandy, loam, gravelly loam, silt of cumulative area.
loam, gravelly silt loam, clay silty loam, and silty clay loam. With respect to the relationship between SPC and lineament fac-
tors (Table 3), the lineament length density, the lineament length
4. Results density weighted by its length, and the lineament frequency
weighted by its length all show a positive correlation. In other words,
4.1. Relationships between the SPC and related factors as the values of these factors increase, then the groundwater potential
could also increase. Therefore, the lineament could be associated with
Topographic, lineament, geological, and soil data from the study a higher SPC. In the case of hydrogeological units, there are five classes
area were collected, processed, and digitized for use within a GIS. of hydrogeology: porous volcanics, semi-consolidated sediments 2,
Maps relevant to specific capacity were assembled in a spatial semi-consolidated sediments 1, nonporous volcanics, and intrusives.
database. The spatial data were evaluated using the probability– The frequency ratio is higher in porous volcanics and semi-consoli-
frequency-ratio model to reveal the correlation between specific dated sediments 2 classes and is lower in the other classes.
capacity and the hydrogeological factors (Table 3) in the study The relationship between SPC and soil texture is as follows. The
area. A positive correlation means higher groundwater potential; groundwater potential value is higher in gravel, gravely sandy
a negative correlation means lower groundwater potential. loam, fine sandy, loam, gravely silt loam, clay silty loam, and silty
The relationship between specific capacity P6.25 m3/d/m and clay loam showing a >1.0 ratio, and the groundwater potential va-
topography-related factors is as follows (Table 3). Topography plays lue is lower in textures showing a <1.0 ratio (Table 3).
a very important role in hydrogeological systems (Saraf et al., 2004).
Some derivatives of the DEM such as topography-related factors, 4.2. Groundwater potential mapping and its validation
which are valuable in groundwater potential analysis, are shown
in Table 3. All the topography-related factors (ground elevation, The probability–frequency model was used to derive and calcu-
ground elevation within 300 m, mean ground elevation within late correlation ratings between the SPC and each groundwa-
166 H.-J. Oh et al. / Journal of Hydrology 399 (2011) 158–172

Fig. 7. Hydrogeological units extracted from geological map (a) and lineament distribution from Landsat TM image (b) of study area.

ter-related factor (Table 2). Each factor’s rating was assigned as the specific capacity P6.25 m3/d/m values were detected from
relationship between SPC and each factor’s type or range (Table 3). groundwater survey data. The specific capacity locations were ran-
The ratio of the number of cells where SPC had not occurred to the domly divided into a training set (66%; 55 specific capacity loca-
number of cells where SPC had occurred more than 6.25 m3/d/m is tions) to analyze groundwater potential using the frequency-ratio
shown in Table 3. The groundwater potential index (GPI) (Eq. (3)) is model and a validation set (34%; 28 specific capacity locations)
calculated by summation of each factor’s ratio value. to validate the predicted groundwater potential map (see footnote
in Table 3). The groundwater potential analysis result was vali-
GPI FR ¼ RFr ðFr : Rating of each factors’ type or rangeÞ ð3Þ dated using known specific capacity locations that were not used
The relation analysis is the ratio of the area in which specific for training the model. Validation was performed by comparing
capacity occurred to the total area. A value of 1 is an average value. the known specific capacity P6.25 m3/d/m location data with the
If the value is greater than 1, then it shows a higher correlation; if groundwater potential map. A rate curve was created, and the area
the value is less than 1, then it shows a lower correlation. under the curve was calculated. The rate explains how well the
The groundwater potential map was quantitatively developed model and factors predict the specific capacity. The area under
using the groundwater potential index value for the interpretation the curve line can qualitatively assess the accuracy of the predic-
shown in Fig. 8. The minimum and maximum values obtained were tion. To obtain the relative ranks for each prediction pattern, the
4.17 and 19.05, respectively. The mean value was 12.89, and the calculated index values of all cells in the study area were sorted
standard deviation was 2.53. The index was composed of four clas- in descending order. The ordered cell values were then divided into
ses based on area for easy and visual interpretation with index 100 classes with accumulated 1% intervals. The rate validation re-
ranges of very high, high, moderate, and low in 5%, 10%, 15%, and sult appears as a line in Fig. 9. For example, an index rank above
70% of the study area, respectively. The classification was useful 10% of the groundwater potential index could explain 33% of all
to visually delineate the predicted groundwater potential areas. the SPCs. To obtain the results quantitatively, the areas under the
The groundwater potential map should effectively predict fu- curve were recalculated as if the total area were 1, which would
ture groundwater potential areas. This could be validated using mean perfect prediction accuracy. By this method, the area under
new exploration locations as they occur. In the study, many a curve can be used to assess the prediction accuracy qualitatively.
H.-J. Oh et al. / Journal of Hydrology 399 (2011) 158–172 167

Fig. 7 (continued)

Here, this area is 0.7778, as shown in Fig. 9. Therefore, the ground- length, and soil texture all had a positive influence on the ground-
water potential showed a prediction accuracy of 77.78%, represent- water potential maps using all factors. In contrast, ground eleva-
ing an overall satisfactory agreement of more than 70%. tion, ground elevation within 300 m, ground slope, ground
curvature, river density without agricultural irrigation, and hydro-
geological units all had a very small negative influence on the
4.3. Sensitivity analysis groundwater potential maps using all factors. This is because the
lower the value of the area below the curve, the greater the effect
The sensitivity analyses were conducted by excluding each fac- that the factor has on the groundwater potential maps. Conversely,
tor (Eq. (2)) in turn during the summation stage using Eq. (3); the a larger area below the curve means that the factor has a more neg-
effect of each factor was then evaluated. That is, the groundwater ative effect on the groundwater potential maps.
potential maps using Eq. (2) were validated using existing specific
capacity locations not used for training the model. For the valida-
tion, the method was subjected to tests to determine whether its 5. Discussion and conclusion
predictions matched the expected results based on knowledge of
the factors. In other words, we conducted a sensitivity analysis in The variables related to the hydrologic data are based on geo-
which the model system was subjected to various selections of fac- graphic data for the study area. The hydrologic data from the well
tors, and the outputs were compared with expected changes in the location represents a value at only one point in the study area and
outputs. cannot represent the overall study area. To address this use, the
To perform the sensitivity analysis, we again used the rate curve probability method was employed to delineate the groundwater
and also the methods for calculating the area below the curve that potential in the study area. If there are many hydrological data
were used to validate the result. According to the validation of the available, this method may be appropriate for the analysis. How-
groundwater potential maps by sensitivity analysis (Table 4), in se- ever, it is not possible to obtain a large number of well data points
quence, mean ground elevation within watershed, mean ground in the study area because of limitations such as high cost and inac-
slope within watershed, TWI, distance from river, cumulative wa- cessibility in mountainous areas. Thus, in this study, accessible
tershed area, lineament length density, lineament length density well locations and widely available hydrological factors (Table 2)
weighted by its length, lineament frequency weighted by its were used to predict groundwater potential in the study area.
Table 3

168
Frequency ratio between specific capacity (SPC) and related factors.

Factor Class No. of pixels % of domain SPC < 6.25 m3/d/m SPC P 6.25 m3/d/m
in domain
No. of data 0a % of data 0 Ratio of data 0 No. of data 1b % of data 1 Ratio of data 1
r s t u v w x y

Ground elevation 32 21,600 43.64 8 34.78 0.80 16 50.00 1.15


75 19,800 40.00 9 39.13 0.98 13 40.63 1.02
128 4500 9.09 3 13.04 1.43 2 6.25 0.69
239 3600 7.27 3 13.04 1.79 1 3.13 0.43
240 0 0.00 0 0.00 0.00 0 0.00 0.00
Ground elevation within 300 m 35 22,500 45.45 8 34.78 0.77 17 53.13 1.17
76 17,100 34.55 9 39.13 1.13 10 31.25 0.90
127 6300 12.73 3 13.04 1.02 4 12.50 0.98
233 3600 7.27 3 13.04 1.79 1 3.13 0.43
245 0 0.00 0 0.00 0.00 0 0.00 0.00
Mean ground elevation within 46 18,000 36.36 7 30.43 0.84 13 40.63 1.12
watershed area
84 15,300 30.91 8 34.78 1.13 9 28.13 0.91

H.-J. Oh et al. / Journal of Hydrology 399 (2011) 158–172


128 9000 18.18 4 17.39 0.96 6 18.75 1.03
235 4500 9.09 2 8.70 0.96 3 9.38 1.03
236 2700 5.45 2 8.70 1.59 1 3.13 0.57
Ground slope 5.0 19,800 40.00 8 34.78 0.87 14 43.75 1.09
17.7 15,300 30.91 5 21.74 0.70 12 37.50 1.21
30.3 8100 16.36 6 26.09 1.59 3 9.38 0.57
45.4 1800 3.64 2 8.70 2.39 0 0.00 0.00
45.5 4500 9.09 2 8.70 0.96 3 9.38 1.03
Mean ground slope within 14.5 14,400 29.09 5 21.74 0.75 11 34.38 1.18
watershed area
21.6 15,300 30.91 8 34.78 1.13 9 28.13 0.91
27.9 11,700 23.64 4 17.39 0.74 9 28.13 1.19
37.8 6300 12.73 4 17.39 1.37 3 9.38 0.74
38.0 1800 3.64 2 8.70 2.39 0 0.00 0.00
Curvature 0.782 6300 12.73 4 17.39 1.37 3 9.38 0.74
0.081 13,500 27.27 4 17.39 0.64 11 34.38 1.26
0.070 18,000 36.36 8 34.78 0.96 12 37.50 1.03
0.862 9900 20.00 5 21.74 1.09 6 18.75 0.94
0.863 1800 3.64 2 8.70 2.39 0 0.00 0.00
TWI 5.418 4500 9.09 2 8.70 0.96 3 9.38 1.03
5.842 1800 3.64 2 8.70 2.39 0 0.00 0.00
6.441 7200 14.55 6 26.09 1.79 2 6.25 0.43
7.840 17,100 34.55 5 21.74 0.63 14 43.75 1.27
7.841 18,900 38.18 8 34.78 0.91 13 40.63 1.06
River density 3.110 5400 10.91 5 21.74 1.99 1 3.13 0.29
3.917 9000 18.18 7 30.43 1.67 3 9.38 0.52
4.599 9000 18.18 3 13.04 0.72 7 21.88 1.20
5.306 10,800 21.82 3 13.04 0.60 9 28.13 1.29
5.307 15,300 30.91 5 21.74 0.70 12 37.50 1.21
Distance from river 0.026 20,700 41.82 8 34.78 0.83 15 46.88 1.12
0.057 10,800 21.82 5 21.74 1.00 7 21.88 1.00
0.096 9000 18.18 5 21.74 1.20 5 15.63 0.86
0.159 2700 5.45 0 0.00 0.00 3 9.38 1.72
0.160 6300 12.73 5 21.74 1.71 2 6.25 0.49
Cumulative watershed area 17 9900 20.00 6 26.09 1.30 5 15.63 0.78
29 10,800 21.82 6 26.09 1.20 6 18.75 0.86
52 8100 16.36 4 17.39 1.06 5 15.63 0.95
170 6300 12.73 2 8.70 0.68 5 15.63 1.23
176 14,400 29.09 5 21.74 0.75 11 34.38 1.18
Lineament length density 0.507 6300 12.73 3 13.04 1.02 4 12.50 0.98
0.778 6300 12.73 3 13.04 1.02 4 12.50 0.98
1.003 8100 16.36 7 30.43 1.86 2 6.25 0.38
1.284 13,500 27.27 5 21.74 0.80 10 31.25 1.15
1.285 15,300 30.91 5 21.74 0.70 12 37.50 1.21
Lineament length density weighted by its length 2.432 7200 14.55 4 17.39 1.20 4 12.50 0.86
4.370 7200 14.55 6 26.09 1.79 2 6.25 0.43
6.683 11,700 23.64 4 17.39 0.74 9 28.13 1.19
9.822 14,400 29.09 7 30.43 1.05 9 28.13 0.97
9.823 9000 18.18 2 8.70 0.48 8 25.00 1.38
Lineament frequency weighted by its length 8.663 7200 14.55 4 17.39 1.20 4 12.50 0.86
13.715 7200 14.55 4 17.39 1.20 4 12.50 0.86
20.263 13,500 27.27 7 30.43 1.12 8 25.00 0.92
29.397 12,600 25.45 5 21.74 0.85 9 28.13 1.10
29.398 9000 18.18 3 13.04 0.72 7 21.88 1.20

H.-J. Oh et al. / Journal of Hydrology 399 (2011) 158–172


Hydrogeological units P-Volc. 5400 10.91 2 8.70 0.80 4 12.50 1.15
SC-Sed.2 14,400 29.09 2 8.70 0.30 14 43.75 1.50
SC-Sed.1 5400 10.91 3 13.04 1.20 3 9.38 0.86
NP-Volc. 18,900 38.18 11 47.83 1.25 10 31.25 0.82
Intrus. 5400 10.91 5 21.74 1.99 1 3.13 0.29
Soil texture Forest 6300 12.73 4 17.39 1.37 3 9.38 0.74
Grassland 2700 5.45 2 8.70 1.59 1 3.13 0.57
Gravel 1800 3.64 0 0.00 0.00 2 6.25 1.72
Gravelly sandy L. 2700 5.45 1 4.35 0.80 2 6.25 1.15
Fine gravelly sandy 900 1.82 1 4.35 2.39 0 0.00 0.00
L. Sandy L. 4500 9.09 3 13.04 1.43 2 6.25 0.69
Loamy fine sand 1800 3.64 2 8.70 2.39 0 0.00 0.00
Fine sand 3600 7.27 0 0.00 0.00 4 12.50 1.72
Loam 900 1.82 0 0.00 0.00 1 3.13 1.72
Gravelly L. 7200 14.55 4 17.39 1.20 4 12.50 0.86
Silt L. 4500 9.09 4 17.39 1.91 1 3.13 0.34
Gravelly silt L. 9900 20.00 2 8.70 0.43 9 28.13 1.41
Clay silty L. 900 1.82 0 0.00 0.00 1 3.13 1.72
Silty clay L. 1800 3.64 0 0.00 0.00 2 6.25 1.72

Total number of well location: 83.


Number of training and validation location: 55 and 28.
% of domain (s) = No. of pixels in each domain (r)/Total number pixels domain  100.
% of data 0 (u) = No. of data 0 in each domain (t)/23  100.
Ratio of data 0 (v) = % of data 0 (u)/% of domain (s).
% of data 1 (x) = No. of data 1 in each domain (w)/32  100.
Ratio of data 1(y) = % of data 1 (x)/% of domain (s).
a
Number of SPC < 6.25 m3/d/m location assigned ‘‘0’’ in training data: 23.
b
Number of SPC  6.25 m3/d/m location assigned ‘‘1’’ in training data: 32.

169
170 H.-J. Oh et al. / Journal of Hydrology 399 (2011) 158–172

Fig. 8. Groundwater potential map for specific capacity (SPC) P6.25 m3/d/m based on frequency ratio; very high (5% of study area: 16.71–19.05, range of frequency ratio in
the class), high (10%: 15.70–16.70), moderate (15%: 14.55–15.69), and low (the remaining 70%: 4.17–14.54).

Table 4
Area of under the curve and accuracy of groundwater potential map for specific
capacity (SPC) P6.25 m3/d/m by sensitivity analysis.

Excepted factor Area of below the Accuracy


curve (%)
Using all factors 0.7778 77.78
Ground elevation 0.7864 78.64
Ground elevation within 300 m 0.7816 78.16
Mean ground elevation within watershed 0.7762 77.62
area
Ground slope 0.7781 77.81
Mean ground slope watershed area 0.7723 77.23
Curvature 0.7839 78.39
TWI 0.7772 77.72
River density 0.7780 77.80
Distance from river 0.7697 76.97
Cumulative watershed area 0.7695 76.95
Lineament length density 0.7757 77.57
Lineament length density weighted by its 0.7729 77.29
length
Lineament frequency weighted by its 0.7805 78.05
length
Hydrogeological units 0.7774 77.74
Fig. 9. Success rate curve of groundwater potential index for specific capacity (SPC) Soil texture 0.7664 76.64
P6.25 m3/d/m by frequency ratio.

With regard to the ratio between training and validation sets, the amount of data available. If more data are used for the estima-
this ratio could be changed for training and validation based on tion, the data could be split 50–50 between training and validation
H.-J. Oh et al. / Journal of Hydrology 399 (2011) 158–172 171

sets. However, for datasets that are too small to evenly split be- selected by relationship analysis among all 27 variables were em-
tween training and validation data (i.e., well location), the ground- ployed for sensitivity analysis. Therefore, the change in accuracy of
water potential could not be employed for a sufficient assessment. each factor could be small because all of these variables are impor-
Carranza and Hale (2002) noted that expert knowledge is required tant in the analysis. In particular, 9 of the 15 factors showed a more
to divide the dataset into training and validation data. In the study positive influence than the remaining factors on groundwater po-
area, on the basis of the 70/30 ratio, the 66/34 ratio was selected to tential mapping. In increasing order of positive influence, these
randomly divide the training and validation sets because of limited nine factors were soil texture, cumulative watershed area, distance
data. from river, mean ground slope within watershed area, lineament
GIS and remote sensing technology provide a means of intro- length density weighted by its length, lineament length density,
ducing information and knowledge from other data sources into mean ground elevation within watershed area, TWI, and hydrolog-
the decision making process and aid in the handling and manipu- ical units (accuracies of 77.62%, 77.23%, 77.72%, 76.97%, 77.74%,
lation of classified remote sensing data (Adinarayana and Krishna, 76.95%, 77.57%, 77.29%, and 76.64%, respectively). Soil texture
1996). Use of a GIS enables quantitative assessment of the conse- had the most positive effect in the sensitivity analysis.
quences of heterogeneity in ecological systems over a broad range
of spatial and temporal scales. Integration of several surface fea- Acknowledgements
tures that indicate groundwater potentialities systematically is
an important aspect in water-management studies. A database de- This research was supported by the Basic Research Project of
signed to support water-resource decisions must contain various the Korea Institute of Geoscience and Mineral Resources (KIGAM)
thematic information because of the interdisciplinary nature of funded by the Ministry of Knowledge Economy of Korea.
water problems. Conventional methods of exploration do not al-
ways account for the diverse factors affecting the presence of Appendix A. Supplementary material
groundwater (Murthy, 2000).
In this study, only a regional groundwater potential analysis Supplementary data associated with this article can be found, in
was performed because the small area studied did not let us deter- the online version, at doi:10.1016/j.jhydrol.2010.12.027.
mine the distribution of rainfall. However, if data on factors caus-
ing groundwater productivity exist, then a more detailed analysis References
could also be done.
Groundwater potential maps are useful to planners and engi- Adinarayana, J., Krishna, N.R., 1996. Integration of multi-seasonal remotely-sensed
images for improved landuse classification of a hilly watershed using
neers seeking suitable locations at which to implement resource
geographical information systems. Int. J. Remote Sens. 17, 1679–1688.
explorations. The results presented here can be used as basic data Beven, K.J., Kirkby, M.J., 1979. A physically based, variable contributing area model
to assist in slope management and land-use planning. Moreover, of basin hydrology. Hydrol. Sci. Bull. 24, 43–69.
the methods used in this study are also valid for generalized plan- Bonham-Carter, G.F., 1994. Geographic Information Systems for Geoscientists:
Modeling with GIS. Pergamon Press, Ontario.
ning and assessment purposes. Note, however, that these methods Carranza, E.J.M., Hale, M., 2002. Evidential belief functions for data-driven
might be less useful at the site-specific scale, where local geologi- geologically-constrained predictive mapping of gold potential, Baguio district,
cal and geographic heterogeneities may prevail. For the method to Philippines. Ore Geol. Rev. 22, 117–132.
Chenini, I., Ben Mammou, A., 2010. Groundwater recharge study in arid region: an
be more generally applied, more specific capacity data are needed, approach using GIS techniques and numerical modeling. Comput. Geosci. 36,
and the method must be applied to more regions. 801–817.
Chenini, I., Mammou, A.B., May, M.E., 2010. Groundwater recharge zone mapping
using GIS-based multi-criteria analysis: a case study in Central Tunisia
(Maknassy Basin). Water Resour. Manage. 24, 921–939.
6. Conclusion Chowdhury, A., Jha, M.K., Chowdary, V.M., Mal, B.C., 2009. Integrated remote
sensing and GIS-based approach for assessing groundwater potential in West
Medinipur district, West Bengal, India. Int. J. Remote Sens. 30, 231–250.
Groundwater is among the most valuable natural resources. Corsini, A., Cervi, F., Ronchetti, F., 2009. Weight of evidence and artificial neural
Government and research institutions worldwide have tried for networks for potential groundwater spring mapping: an application to the Mt.
years to assess groundwater potential and predict its spatial distri- Modino area (Northern Apennines, Italy). Geomorphology 111, 79–87.
Faust, N.L., Anderson, W.H., Star, J.L., 1991. Geographic information systems and
bution. This study presented a probabilistic approach that used
remote sensing future computing environment. Photogramm. Eng. Remote
both satellite imagery and GIS to estimate an area’s potential Sens. 57, 655–668.
groundwater resources. Ghayoumian, J., Mohseni Saravi, M., Feiznia, S., Nouri, B., Malekian, A., 2007.
To map groundwater potential, the first step was selecting the Application of GIS techniques to determine areas most suitable for artificial
groundwater recharge in a coastal aquifer in southern Iran. J. Asian Earth Sci. 30,
15 most important variables that affect the groundwater potential. 364–374.
The groundwater potential was then mapped using a frequency-ra- Goodchild, M.F., 1993. The state of GIS for environmental problem-solving. In:
tio model, which represents the relationship between hydrologic Goodchild, M.F., Parks, B.O., Steyaert, L.T. (Eds.), Environmental Modeling with
GIS. Oxford University Press, pp. 8–15.
data (specific capacity) and the variables. Greenbaum, D., 1992. Structural Influences on the Occurrence of Groundwater in SE
The frequency ratio results indicate that the elevation and Zimbabwe, vol. 66. Geological Society, London, Special Publications, pp. 77–85.
slope-related factors show a negative correlation with specific Gupta, M., Srivastava, P.K., 2010. Integrating GIS and remote sensing for
identification of groundwater potential zones in the hilly terrain of Pavagarh,
capacities P6.25 m3/d/m, whereas other factors (TWI, river den- Gujarat, India. Water Int. 35, 233–245.
sity, and lineament-related factors) show a positive correlation. Hinton, J.C., 1996. GIS and remote sensing integration for environmental
The curvature shows a high frequency ratio (ratios over 1.1) in applications. Int. J. Geogr. Inform. Syst. 10, 877–890.
Hwang, J.H., Kim, D.H., Cho, D.L., Song, K.-Y., 1996. Explanatory Note of the Andong
nearly flat areas, and hydrological units appear in the porous vol- Sheet. Korea Institute of Geoscience and Mineral Resources, Daejeon.
canic class and soil textures appear in gravel. The validation of a Israil, M., Al-hadithi, M., Singhal, D.C., 2006. Application of a resistivity survey and
map that used all of the factors and the frequency ratios showed geographical information system (GIS) analysis for hydrogeological zoning of a
piedmont area, Himalayan foothill region, India. Hydrogeol. J. 14, 753–759.
a satisfactory agreement of 77.78%. To identify the order of relative
Jaiswal, R.K., Mukherjee, S., Krishnamurthy, J., Saxena, R., 2003. Role of remote
influence of each factor within the 15 important factors, a sensitiv- sensing and GIS techniques for generation of groundwater prospect zones
ity analysis was performed. The change of percentage accuracy of towards rural development – an approach. Int. J. Remote Sens. 24, 993–1008.
each factor in the sensitivity analysis is less than 2% in comparison Jha, M.K., Chowdhury, A., Chowdary, V.M., Peiffer, S., 2007. Groundwater
management and development by integrated remote sensing and geographic
to a map based on all of the factors (accuracy of 77.87%). The rea- information systems: prospects and constraints. Water Resour. Manage. 21,
son for this small change could be that 15 related variables 427–467.
172 H.-J. Oh et al. / Journal of Hydrology 399 (2011) 158–172

Kane, W.F., Peters, D.C., Speirer, R.A., 1996. Remote sensing in investigation of Oh, H.-J., Lee, S., Soedradjat, G., 2010. Quantitative landslide susceptibility mapping
engineered underground structures. J. Geotech. Eng. 122, 674–681. at Pemalang area, Indonesia. Environ. Earth Sci. 60, 1317–1328.
Kim, D.H., Hwang, J.H., Park, K.H., Song, K.-Y., 1998. Explanatory Note of the Busan O’Leary, D.W., Friedman, J.D., Pohn, H.A., 1976. Lineament, linear, lineation; some
Sheet. Korea Institute of Geoscience and Mineral Resources, Daejeon. proposed new standards for old terms. Geol. Soc. Am. Bull. 87, 1463–1469.
Kim, K.-D., Lee, S., Oh, H.-J., Choi, J.-K., Won, J.-S., 2006. Assessment of ground Pradhan, B., 2009. Groundwater potential zonation for basaltic watersheds using
subsidence hazard near an abandoned underground coal mine using GIS. satellite remote sensing data and GIS techniques. Cent. Eur. J. Geosci. 1, 120–
Environ. Geol. 50, 1183–1191. 129.
Koike, K., Nagano, S., Kawaba, K., 1998. Construction and analysis of interpreted Pradhan, B., Lee, S., 2010. Delineation of landslide hazard areas on Penang Island,
fracture planes through combination of satellite-image derived lineaments and Malaysia, by using frequency ratio, logistic regression, and artificial neural
digital elevation model data. Comput. Geosci. 24, 573–583. network models. Environ. Earth Sci. 60, 1037–1054.
Kumar, M.G., Bali, R., Agarwal, A.K., 2009. Integration of remote sensing and Pradhan, B., Lee, S., Buchroithner, M.F., 2010. Remote sensing and GIS-based
electrical sounding data for hydrogeological exploration – a case study of landslide susceptibility analysis and its cross-validation in three test areas using
Bakhar watershed, India. Hydrolog. Sci. J. 54, 949–960. a frequency ratio model. Photogrammetrie Fernerkundung Geoinformation 1,
Lee, S., Choi, U., 2003. Development of GIS-based geological hazard information 17–32.
system and its application for landslide analysis in Korea. Geosci. J. 7, 243–252. Prasad, R.K., Mondal, N.C., Banerjee, P., Nandakumar, M.V., Singh, V.S., 2008.
Lee, S., Dan, N., 2005. Probabilistic landslide susceptibility mapping in the Lai Chau Deciphering potential groundwater zone in hard rock through the application of
province of Vietnam: focus on the relationship between tectonic fractures and GIS. Environ. Geol. 55, 467–475.
landslides. Environ. Geol. 48, 778–787. Ranganai, R.T., Ebinger, C.J., 2008. Aeromagnetic and landsat TM structural
Lee, S., Oh, H.J., Kim, K.D., 2010. Statistical spatial modeling of ground subsidence interpretation for identifying regional groundwater exploration targets,
hazard near an abandoned underground coal mine. Disaster Adv. 3, 11–23. south-central Zimbabwe Craton. J. Appl. Geophys. 65, 73–83.
Leopold, L.B., Wolman, M.G., Miller, J.P., 1964. Fluvial Processes in Geomorphology. Reid, L.M., 1993. Research and cumulative watershed effects. U.S.D.A. Forest Service
W.H. Freeman and Company, San Francisco, California. General Technical Report PSW-GTR-141, Albany, California.
Mabee, S.B., Hardcastle, K.C., Wise, D.U., 1994. A method of collecting and analyzing Roy, A.K., 1996. Hydromorphogeological Mapping for Ground Water Targeting and
lineaments for regional-scale fractured-bedrock aquifer studies. Ground Water Development in Dehradun valley. Surya Publications, Dehradun.
32, 884–894. Saha, D., Dhar, Y.R., Vittala, S.S., 2010. Delineation of groundwater development
Madrucci, V., Taioli, F., de Araújo, C.C., 2008. Groundwater favorability map using potential zones in parts of marginal Ganga Alluvial Plain in South Bihar, Eastern
GIS multicriteria data analysis on crystalline terrain, São Paulo State, Brazil. J. India. Environ. Monit. Assess. 165, 179–191.
Hydrol. 357, 153–173. Saraf, A.K., Choudhury, P.R., 1998. Integrated remote sensing and GIS for
Ministry for Food, Agriculture, Forestry and Fisheries (MFAFF), 1985–2005. Rural groundwater exploration and identification of artificial recharge sites. Int. J.
Groundwater Survey Report. MIFAFF. Remote Sens. 19, 1825–1841.
Ministry of Land, Transport and Maritime Affairs (MLTM), 1995. National Saraf, A.K., Choudhury, P.R., Roy, B., Sarma, B., Vijay, S., Choudhury, S., 2004. GIS
Groundwater Monitoring Network Construction Report. MLTM. based surface hydrological modelling in identification of groundwater recharge
Ministry of Land, Transport and Maritime Affairs (MLTM), 1998. National zones. Int. J. Remote Sens. 25, 5759–5770.
Groundwater Monitoring Network Construction Report. MLTM. Sarkar, B., Deota, B., Raju, P., Jugran, D., 2001. A geographic information system
Ministry of Land, Transport and Maritime Affairs (MLTM), 2001. National approach to evaluation of groundwater potentiality of Shamri micro-
Groundwater Monitoring Network Construction Report. MLTM. watershed in the Shimla Taluk, Himachal Pradesh. J. Indian Soc. Remote
Ministry of Land, Transport and Maritime Affairs (MLTM), 2003. Report of Sens. 29, 151–164.
Groundwater in Pohang Area. MLTM. Solomon, S., Quiel, F., 2006. Groundwater study using remote sensing and
Mondal, M.S., Pandey, A.C., Garg, R.D., 2008. Groundwater prospects evaluation geographic information systems (GIS) in the central highlands of Eritrea.
based on hydrogeomorphological mapping using high resolution satellite Hydrogeol. J. 14, 729–741.
images: a case study in Uttarakhand. J. Indian Soc. Remote Sens. 36, 69–76. Srivastava, P.K., Bhattacharya, A.K., 2006. Groundwater assessment through an
Mukherjee, S., 1996. Targetting saline aquifer by remote sensing and geophysical integrated approach using remote sensing, GIS and resistivity techniques: a
methods in a part of Hamirpur–Kanpur, India. Hydrol. J. 19, 1867–1884. case study from a hard rock terrain. Int. J. Remote Sens. 27, 4599–4620.
Murthy, K.S.R., 2000. Ground water potential in a semi-arid region of Andhra Stafford, D.B., 1991. Civil Engineering Applications of Remote Sensing and
Pradesh – a geographical information system approach. Int. J. Remote Sens. 21, Geographic Information Systems. ASCE, New York.
1867–1884. Tam, V.T., De Smedt, F., Batelaan, O., Dassargues, A., 2004. Study on the relationship
Murthy, K.S.R., Mamo, A.G., 2009. Multi-criteria decision evaluation in groundwater between lineaments and borehole specific capacity in a fractured and karstified
zones identification in Moyale–Teltele subbasin, South Ethiopia. Int. J. Remote limestone area in Vietnam. Hydrogeol. J. 12, 662–673.
Sens. 30, 2729–2740. Teeuw, R.M., 1995. Groundwater exploration using remote sensing and a low-cost
Nagarajan, M., Singh, S., 2009. Assessment of groundwater potential zones using GIS geographical information system. Hydrogeol. J. 3, 21–30.
technique. J. Indian Soc. Remote Sens. 37, 69–77. Thomas Jr., G.B., 1968. Calculus and Analytic Geometry: Part Two Vectors and
Oh, H.J., Lee, S., Chotikasathien, W., Kim, C.H., Kwon, J.H., 2009. Predictive landslide Functions of Several Variables. Addison-Wesley Publishing, Reading, MA.
susceptibility mapping using spatial information in the Pechabun area of Yeh, H.F., Lee, C.H., Hsu, K.C., Chang, P.H., 2009. GIS for the assessment of the
Thailand. Environ. Geol. 57, 641–651. groundwater recharge potential zone. Environ. Geol. 58, 185–195.

You might also like