A GIS-based DRASTIC model for assessing aquifer vulnerability in Kakamigahara Heights, Gifu Prefecture, central Japan

Transcription

1 Science of the Total Environment 345 (2005) A GIS-based DRASTIC model for assessing aquifer vulnerability in Kakamigahara Heights, Gifu Prefecture, central Japan Insaf S. Babiker*, Mohamed A.A. Mohamed, Tetsuya Hiyama, Kikuo Kato Hydrospheric Atmospheric Research Center, Nagoya University, Furo-cho, Chikusa-ku, Nagoya , Japan Received 6 July 2004; accepted 5 November 2004 Available online 21 January 2005 Abstract Vulnerability assessment to delineate areas that are more susceptible to contamination from anthropogenic sources has become an important element for sensible resource management and land use planning. This contribution aims at estimating aquifer vulnerability by applying the DRASTIC model as well as utilizing sensitivity analyses to evaluate the relative importance of the model parameters for aquifer vulnerability in Kakamigahara Heights, Gifu Prefecture central Japan. An additional objective is to demonstrate the combined use of the DRASTIC and geographical information system (GIS) as an effective method for groundwater pollution risk assessment. The DRASTIC model uses seven environmental parameters (Depth to water, net Recharge, Aquifer media, Soil media, Topography, Impact of vadose zone, and hydraulic Conductivity) to characterize the hydrogeological setting and evaluate aquifer vulnerability. The western part of the Kakamigahara aquifer was dominated by bhighq vulnerability classes while the eastern part was characterized by bmoderateq vulnerability classes. The elevated north-eastern part of the study area displayed blowq aquifer vulnerability. The integrated vulnerability map shows the high risk imposed on the eastern part of the Kakamigahara aquifer due to the high pollution potential of intensive vegetable cultivation. The more vulnerable western part of the aquifer is, however, under a lower contamination risk. In Kakamigahara Heights, land use seems to be a better predictor of groundwater contamination by nitrate. Net recharge parameter inflicted the largest impact on the intrinsic vulnerability of the aquifer followed by soil media, topography, vadose zone media, and hydraulic conductivity. Sensitivity analyses indicated that the removal of net recharge, soil media and topography causes large variation in vulnerability index. Moreover, net recharge and hydraulic conductivity were found to be more effective in assessing aquifer vulnerability than assumed by the DRASTIC model. The GIS technique has provided efficient environment for analyses and high capabilities of handling large spatial data. D 2004 Elsevier B.V. All rights reserved. Keywords: Aquifer vulnerability; DRASTIC; GIS; Model sensitivity; Kakamigahara; Japan * Corresponding author. Tel.: ; fax: address: s040116dd@mbox.nagoya-u.ac.jp (I.S. Babiker) /$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi: /j.scitotenv

2 128 I.S. Babiker et al. / Science of the Total Environment 345 (2005) Introduction Groundwater has been considered as an important source of water supply due to its relatively low susceptibility to pollution in comparison to surface water, and its large storage capacity (US EPA, 1985). However, there are significant sources of diffuse and point pollution of groundwater from land use activities, particularly agricultural practices. The intrusion of these pollutants to groundwater alters the water quality and reduces its value to the consumer (Melloul and Collin, 1994). Prevention of contamination is therefore critical for effective groundwater management. Spatial variability and data constraints preclude monitoring all groundwater and make remediation activities expensive and often impractical. Vulnerability assessment has been recognized for its ability to delineate areas that are more likely than others to become contaminated as a result of anthropogenic activities at/or near the earth s surface. Once identified, these areas can be targeted by careful land-use planning, intensive monitoring, and by contamination prevention of the underlying groundwater. The Kakamigahara aquifer, which is located in the southern part of Gifu Prefecture, central Japan, provides a source of water for domestic, industrial and agricultural use. High nitrate levels exceeding the 13 mg/l concentration (the human affected value) have been encountered in groundwater of the Kakamigahara aquifer (Terao et al., 1985, 1993; Mohamed et al., 2003) while some concentrations have already exceeded the maximum acceptance level (44 mg/l NO 3 ) according to Japan regulations (Babiker et al., 2004). The dominant source of nitrate contamination has been identified as agricultural land use, particularly vegetable cultivation with some possibility of urban sources (residential, commercial, and industrial; Babiker et al., 2004). To ensure that this aquifer can remain as a source of water for the Kakamigahara area, it is necessary to estimate whether certain locations in this groundwater basin are more susceptible to receive and transmit pollution. Therefore, the first objective of this study is to evaluate the Kakamigahara aquifer vulnerability using Depth to water, net Recharge, Aquifer media, Soil media, Topography, Impact of vadose zone, and hydraulic Conductivity (DRASTIC), the empirical model of the U.S. Environmental Protection Agency (EPA; US EPA, 1985). The second objective is to evaluate the relative importance of the DRASTIC model parameters for assessing aquifer vulnerability in Kakamigahara Heights through sensitivity analysis. An additional objective is to demonstrate the combined use of DRASTIC and geographical information system (GIS) as an effective method for groundwater pollution risk assessment and water resource management. 2. Background The concept of groundwater vulnerability was first introduced in France by the end of the 1960s to create awareness of groundwater contamination (Vrba and Zoporozec, 1994). It can be defined as the possibility of percolation and diffusion of contaminants from the ground surface into the groundwater system. Vulnerability is usually considered as an bintrinsicq property of a groundwater system that depends on its sensitivity to human and/ or natural impacts. bspecificq or bintegratedq vulnerability, on the other hand, combines intrinsic vulnerability with the risk of the groundwater being exposed to the loading of pollutants from certain sources (Vrba and Zoporozec, 1994). Groundwater vulnerability deals only with the hydrogeological setting and does not include pollutant attenuation. The natural hydrogeologic factors affect the different pollutants in different ways depending on their interactions and chemical properties. Many approaches have been developed to evaluate aquifer vulnerability. They include processbased methods, statistical methods, and overlay and index methods (Tesoriero et al., 1998). The processbased methods use simulation models to estimate the contaminant migration but they are constrained by data shortage and computational difficulties (Barbash and Resek, 1996). Statistical methods use statistics to determine associations between spatial variables and actual occurrence of pollutants in the groundwater. Their limitations include insufficient water quality observations, data accuracy and careful selection of spatial variables. Overlay and index methods combine factors controlling the movement of pollutants from the ground surface into the saturated zone

3 I.S. Babiker et al. / Science of the Total Environment 345 (2005) resulting in vulnerability indices at different locations. Their main advantage is that some of the factors such as rainfall and depth to groundwater can be available over large areas, which makes them suitable for regional scale assessments (Thapinta and Hudak, 2003). However, their major drawback is the subjectivity in assigning numerical values to the descriptive entities and relative weights for the different attributes. DRASTIC is an index model designed to produce vulnerability scores for different locations by combining several thematic layers. It was originally developed for manual overlay of semiquantitative data layers however the simple definition of its vulnerability index as a linear combination of factors shows the feasibility of the computation using GIS (Fabbri and Napolitano, 1995). GIS are designed to collect diverse spatial data to represent spatially variable phenomena by applying a series of overlay analysis of data layers that are in spatial register (Bonham-Carter, 1996). 3. Study area Kakamigahara Heights is composed of low hills 20 to 60 m above sea level and located on the northern border of the Nohbi plains in central Japan (Fig. 1). The Kiso River intersects the southern edge of the area, running from east to west towards the plains. Kakamigahara City is the third largest city in Gifu Prefecture having a population over 135,000 inhabitants. The dominant industries are aircraft, transport equipment and industrial metal manufacturing. Intensive vegetable cultivation, mainly carrots, is practiced in the eastern side of the Heights, sustained by nitrogen fertilizer applications. The area is characterized by a warm and mild climate with an average annual temperature of C and a rainfall of 1915 mm, (mean of 30-year records from Gifu City rainfall station, Japan Meteorological Agency). Kuroboku soils, Kisogawa mudflow deposits, and the Kakamigahara formation constitute the Kakamigahara Terrace, which is underlain by the Nohbi Second Gravel Formation, the Tokai Group and the basement (Yokoyama and Makinouchi, 1991). The Kakamigahara Formation consists of sands with pumice ranging in age from ca to 0.06 Ma. The Nohbi Second Gravel Formation is composed of boulder gravels with high permeability while the Tokai Group consists of weathered clayey gravels with low permeability. The Kakamigahara groundwater basin consists of the highly permeable Nohbi Second Gravel formation and the overlying coarse sediments. The aquifer forms an east west elongated valley-like topography deepening westward. The aquifer thickness ranges from 15 meters in the east to more than 90 m in the west. Due to the lack of confining clay layers, the aquifer is considered typically unconfined. The groundwater flow is from east to west and the water velocity is high on average (210 3 m/s; Babiker et al., 2004). The aquifer is characterized by a high recharge rate from direct infiltration (Mohamed, 2003). 4. Methods Fig. 1. Location of Kakamigahara Heights. A DRASTIC model applied in a GIS environment was used to evaluate the vulnerability of the Kakamigahara aquifer. The DRASTIC model was developed by the U.S. Environmental Protection Agency (EPA) to evaluate groundwater pollution potential for the entire United States (Aller et al., 1987). It was based on the concept of the hydro-

4 130 I.S. Babiker et al. / Science of the Total Environment 345 (2005) geological setting that is defined as ba composite description of all the major geologic and hydrologic factors that affect and control the groundwater movement into, through and out of an areaq (Aller et al., 1987). The acronym DRASTIC stands for the seven parameters used in the model which are: Depth to water, net Recharge, Aquifer media, Soil media, Topography, Impact of vadose zone and hydraulic Conductivity (Table 1). The model yields a numerical index that is derived from ratings and weights assigned to the seven model parameters. The significant media types or classes of each parameter represent the ranges, which are rated from 1 to 10 based on their relative effect on the aquifer vulnerability. The seven parameters are then assigned weights ranging from 1 to 5 reflecting their relative importance. The DRASTIC Index is then computed Table 1 The DRASTIC model parameters Factor Description Relative weight Depth to water Net Recharge Aquifer media Soil media Topography Impact of vadose zone Hydraulic Conductivity Represents the depth from the ground surface to the water table, deeper water table levels imply lesser chance for contamination to occur. Represents the amount of water which penetrates the ground surface and reaches the water table, recharge water represents the vehicle for transporting pollutants. Refers to the saturated zone material properties, which controls the pollutant attenuation processes. Represents the uppermost weathered portion of the unsaturated zone and controls the amount of recharge that can infiltrate downward. Refers to the slope of the land surface, it dictates whether the runoff will remain on the surface to allow contaminant percolation to the saturated zone. Is defined as the unsaturated zone material, it controls the passage and attenuation of the contaminated material to the saturated zone. Indicates the ability of the aquifer to transmit water, hence determines the rate of flow of contaminant material within the groundwater system applying a linear combination of all factors according to the following equation: DRASTIC Index ¼ D r D w þ R r R w þ A r A w þ S r S w þ T r T w þ I r I w þ C r C w ð1þ where D, R, A, S, T, I, and C are the seven parameters and the subscripts r and w are the corresponding rating and weights, respectively. This model was selected based on the following considerations. DRASTIC uses a relatively large number of parameters (seven parameters) to compute the vulnerability index, which ensures the best representation of the hydrogeological setting. The numerical ratings and weights, which were established using the Delphi technique (Aller et al., 1987), are well defined and are used worldwide. This makes the model suitable for producing comparable vulnerability maps on a regional scale. The necessary information needed to build up the several model parameters was available in the study area or could easily be inferred. Data analyses and model implementation were performed using the GIS software of the International Institute for Geo-Information Science and Earth Observation (ITC), Netherlands, bintegrated Land and Water Information SystemQ (ILWIS 3.1) Preparation of the parameter maps Several types of data were used to construct thematic layers of the seven model parameters. A summary of the data types, sources and usages is available in Table 2. The location of the 134 water wells was digitized from the accompanying topographic map and was linked to an attribute table containing the depth to groundwater table. Essentially, the elevation of the well and the mean water level table were provided for the last 6 years ( ). The depth to water table was obtained by subtracting the water table level from the elevation of the well and averaging over a six-year period. In a relatively small area and a fairly isotropic aquifer, the static water table level exhibits a smooth and gradual change of heads. Therefore, an exact interpolation scheme is appropriate for generating a smooth surface representation for the high degree of spatial continuity of the groundwater surface in an aquifer. The inverse distance moving average interpolation technique was

5 I.S. Babiker et al. / Science of the Total Environment 345 (2005) Table 2 Data used for constructing the seven parameter layers Data type Source Format Scale Date Used to produce Borehole data Kakamigahara City Hall Table 1998~2002 D (water table level) Environmental Office, Division Location map 1:15, of Human Health and Environment Annual rainfall Japan Meteorological Agency website: Table 1971~2000 R (mean) ( Geology map Ministry of Land Infrastructure and Transport website: ( Digital 1:50, A I Geological profiles Yokoyama and Makinouchi 1991 A and I Soil map Ministry of Land Infrastructure Digital 1:50, S and Transport website: ( Land use and topographic map Japan Geographical Survey Institute Hard copy 1:50, T and R Pollution sources Hydraulic conductivity Kakamigahara City Hall Table C Environmental Office, Division of Human Health and Environment Map ~1:70,000 therefore performed on the point data using a limiting search distance of 3000 m to ensure including a high number of input points. In the moving average algorithm, every pixel in the output map is assigned a value based on the weighted average of all points at a distance smaller than the limiting distance. The weight of an input point is proportional to its distance from the output pixel. The inverse distance is the weight function which ensures that relatively larger weights are assigned to points close to an output pixel. This better satisfies the interpolation scheme and helps generate the smooth surface representation suitable for such type of data. The depth to water table map was then classified into ranges defined by the DRASTIC model and assigned rates ranging from 1 (minimum impact on vulnerability) to 10 (maximum impact on vulnerability). The deeper the groundwater, the smaller the rating value. Because the Kakamigahara aquifer is recharged mainly by direct infiltration from precipitation (Mohamed, 2003, Master thesis), the recharge map was constructed from the rainfall data according to the following formula: Net recharge ¼ðrainfall evapotranspirationþ recharge rate ð2þ The rainfall map was obtained by interpolating a thirty years mean of annual precipitation (mm/year) from seven representative rainfall stations in the south of Gifu Prefecture (Gifu, Mino Kamo, Ōgaki, Mino, and Inigawa) and the north of Aichi Prefecture (Inabu and Nagoya). The inverse distance moving average interpolation technique described above was used to construct the rainfall map. Evapotranspiration data from the Kakamigahara was not available, so instead the evapotranspiration value of 900 mm/year from a representative area in central Japan (Kondoh and Nishiyama, 2000) was used. The recharge rate was assumed to be 20% for the urbanized area and 85% for the rest of the study area (Babiker, 1998). The recharge map was then classified into ranges and assigned ratings from 1 to 10. High recharge rates were assigned high numerical rates. The Aquifer media and the Impact of vadose zone were obtained using a subsurface geology map, geological sections, and drilling profiles of the Kakamigahara aquifer (Figs. 4 and 7 in Yokoyama and Makinouchi, 1991). The subsurface geology map was imported in digital format from the website of the Japan Ministry of Land, Infrastructure and Transport ( geo-referenced and on-screen digitized to create a representation of the different geological units. The geological sections were then

6 132 I.S. Babiker et al. / Science of the Total Environment 345 (2005) used to encode the geological units according to the DRASTIC model rating system. The coarse (saturated or unsaturated) media was assigned a high rating value compared to the fine media types. The soil map was also imported in digital format into the ILWIS program, referenced to area coordinates and on-screen digitized to delineate the different soil units. The texture of the various soil types in Kakamigahara was obtained from a representative description of World soil types provided by the United States Department of Agriculture (USDA, 1999). The soil media types were then assigned ratings from 1 to 10 according to their permeability. Coarse soil media have high rates in comparison to fine soil media. The topography layer was constructed from the topography and land use map by interpolation. The topography and land use map was first scanned and registered. The elevation contour lines were digitized together with the elevation points to be used in the interpolation. The contour interpolation process involves two steps (ITC-ILWIS, 2001): rasterization and interpolation. Both the segment contour map and the point elevation map were rasterized and combined. The inclusion of the elevation points in the contour interpolation assist to avoid the creation of flat areas on hilltops or bottom of depressions as a result of being enclosed by the same contour line. The interpolation process was then performed on the combined map using the Borgefors distance method (ITC-ILWIS, 2001) and yielded a digital elevation map (DEM). The height value for each undefined pixel between two contours was found by calculating the shortest distance towards the two isolines. The vertical and horizontal filters bdyq and bdxq were applied on the DEM in order to calculate the horizontal and vertical gradients. The two filter maps were used to obtain the slope percentage according to the following formula: Slope ð% Þ ¼ ðhypðdx; DyÞ=pixel sizeþ100 ð3þ where HYP is a map calculation function of ILWIS to calculate the positive root of the sum of square Dx plus square Dy (Pythagoras rule) where Dx and Dy are the horizontal and vertical gradients, respectively. The slope map was then sliced into ranges and assigned ratings ranging from 1 to 10. Flat areas obtain high rates because they slow down the runoff allowing more time for the contaminant to percolate down to reach the groundwater, while steep areas increase the runoff washing out the contaminant hence are assigned low rates. The hydraulic conductivity map was also scanned and spatially registered. The different hydraulic conductivity zones in the area were defined and assigned ratings according to DRASTIC Aquifer vulnerability assessment The DRASTIC vulnerability index was computed according to Eq. (1). In order to understand the vulnerability index, it is necessary to choose a representation method which can expose the aquifer vulnerability in an appropriate fashion and simultaneously allows comparability between different areas. Basically, Aller et al. (1987) introduced a national colour coding of the vulnerability maps. The DRAS- TIC indices are first classified into ranges by imposing arbitrary thresholds. Those ranges are then assigned the colours of the pseudo colour look-up table (ranging from violet to red). Here, the vulnerability scores are presented based on the classification scheme introduced by Chung and Fabbri (2001). In this method, the vulnerability indices are classified based on a fixed interval of area percentage in the study area. The vulnerability index values were first sorted in a descending form and then the vulnerability indices corresponding to each 5% of the total number of pixels in the study area were taken as thresholds for the classification. Colours were then assigned to the ranges of the subsequent percentages of pixels. The cool colours (shades of blue) indicate blowq vulnerability, the shades of green indicates bmoderateq vulnerability while the warm colours (shades of red) indicate bhighq vulnerability. Because this representation demonstrates the results without imposing arbitrary thresholds, it is considered free of subjectivity and useful in comparing results from different areas or vulnerability models. According to Civita (1994), the integrated vulnerability can be obtained by overlying a representation of the actual pollution sources, which are subdivided on the basis of their pollution potential (e.g., urban areas, cultivated areas, waste dumps, industrial complexes, and the like), on the intrinsic vulnerability map. The integrated vulnerability map was obtained

7 I.S. Babiker et al. / Science of the Total Environment 345 (2005) by combining the vulnerability map and the potential pollution sources extracted from the land use map. In Kakamigahara Heights, nitrate contamination of groundwater was found to closely associate mainly with vegetable cultivation and to a lesser extent with urban land (Babiker et al., 2004). Therefore, the two land use types were considered as potential pollution sources of groundwater of different degrees. To obtain the integrated or specific vulnerability, we assumed that there is no risk of contamination from land use/ cover types rather than vegetable fields and urban land. Vegetable fields are expected to impose a high risk of contaminant loading onto the ground surface (e.g., nitrogen fertilizers) compared to urban land. It has been suggested that the vulnerability classes should be critically considered only where a groundwater contamination risk exists. The vulnerability classes are not considered where there is no risk of contamination Sensitivity analysis One of the major advantages of the DRASTIC model is the implementation of assessment using a high number of input data layers (Evans and Myers, 1990) which is believed to limit the impacts of errors or uncertainties of the individual parameters on the final output (Rosen, 1994). Some authors (e.g., Barber et al., 1993; Merchant, 1994) however, have argued that a DRASTIC-equivalent result can be obtained using a lower number of input parameters and can achieve a better accuracy at less cost. Some studies employed the DRASTIC model for aquifer vulnerability using a lower number of parameters (McLay et al., 2001) or assuming the constancy of the missing parameter (Secunda et al., 1998). Moreover, the unavoidable subjectivity associated with the selection of the seven parameters, the ratings, and the weights used to compute the vulnerability index has also been criticized (Napolitano and Fabbri, 1996). Here, we attempt to evaluate whether it was really necessary to use all of the seven DRASTIC parameters to assess the Kakamigahara aquifer vulnerability by performing model sensitivity analysis. The rated DRASTIC parameters were first evaluated for interdependence and variability. According to Rosen (1994), the independency of DRASTIC parameters decreases the probability of misjudgment. In fact, most of the DRASTIC parameters are naturally closely related. Two sensitivity tests were performed; the map removal sensitivity analyses introduced by Lodwick et al. (1990) and the single-parameter sensitivity analysis introduced by Napolitano and Fabbri (1996). The map removal sensitivity measure identifies the sensitivity of the suitability map (vulnerability map) towards removing one or more maps from the suitability analysis and is computed in the following way: S ¼ðjV=N VV=nj=VÞ100 ð4þ where S is the sensitivity measure expressed in terms of variation index, V and VV are the unperturbed and the perturbed vulnerability indices respectively, and N and n are the number of data layers used to compute V and V V. The actual vulnerability index obtained using all seven parameters was considered as an unperturbed vulnerability while the vulnerability computed using a lower number of data layers was considered as a perturbed one. The single-parameter sensitivity measure was developed to evaluate the impact of each of the DRASTIC parameters on the vulnerability index. It has been made to compare the beffectiveq or brealq weight of each input parameter in each polygon with the btheoreticalq weight assigned by the analytical model. The beffectiveq weight of each polygon is obtained using the following formula: W ¼ðP r P w =VÞ100 ð5þ where W refers to the beffectiveq weight of each parameter, P r and P w are the rating value and weight of each parameter, and V is the overall vulnerability index. The implementation of the sensitivity analysis requires a well-structured database and a GIS capable of manipulating large tables. To avoid analysing the large number of individual pixels in the study area (639,794), the idea of bunique condition subareasq introduced by Napolitano and Fabbri (1996) was used. It is defined as one or more polygons (area) consisting of pixels with a unique combination of D i, R i, A i, S i, T i, I i, and C i where D i, R i, A i, S i, T i, I i, and C i are the rating values of the seven parameters used to compute the vulnerability index and 1ViV10. The crossoperation of ILWIS 3.1 was used to obtain the

8 134 I.S. Babiker et al. / Science of the Total Environment 345 (2005) subareas. The cross-operation performs an overlay of two raster maps by combining pixels at the same locations in both maps and tracking all the combinations that occur between the different values or classes in both maps (ITC-ILWIS, 2001). All the unique combinations between the different parameters were traced through multiple crossing. A total of 818 unique condition subareas have been found in the study area; however, only 663 subareas that are larger than 10 pixels in size (1000 m 2 ) were considered in the statistical analysis of the results. 5. Results and discussion 5.1. DRASTIC parameters and aquifer vulnerability Fig. 2 shows examples of the rated parameter maps used to obtain the DRASTIC aquifer vulnerability index. The depth to water table in Kakamigahara Heights is shallow (average is 26 m), decreasing gradually from east to west. This makes the western part of the Heights more susceptible to contamination according to DRASTIC assumptions. The rating scores ranges between 1 and 7 where the highest scores were assigned to the far eastern part of the area (depth to water is b9.1 m). Although the study area is characterized by a high annual rainfall (N1900 mm/year), the net recharge to the groundwater aquifer is mainly controlled by the type of land use/cover on the surface. The lowest recharge rate (mean, 203 mm/year) was associated with the urban land use because roofs and pavements prevent the penetration of rainwater downward. The rest of the study area had relatively higher recharge rates (mean, 860 mm/year). The Kakamigahara aquifer has generally high net recharge (N200 mm/ year) which was assigned high rating scores (8 and 9). The gravel-rich alluvium of the Second Gravel Formation mostly constitutes the Kakamigahara aquifer and was assigned a high rating score (8). The Mesozoic consolidated sediments and the tuff and volcanic ash deposits were assigned low rating scores ranging from 2 to 4. Fig. 2. Examples of the rated maps used to compute the DRASTIC vulnerability index. The rating score 1 implies a minimum impact on vulnerability while the score 10 indicates the maximum impact.

9 I.S. Babiker et al. / Science of the Total Environment 345 (2005) The soil media is generally variable. The major soil types are the Andosols, the Gleysols and the Fluvisols. The different soil types were assigned rates according to their permeability (depending on the texture). A high score (10) was assigned to the immature soil and the lowest score (1) was assigned to the soil in the built-up areas. The loamy Andosols and Gleysols were assigned moderate rating scores (3~6). The topography layer displayed a gentle slope (0~6%) over most of the study area which has been assigned the DRASTIC rating scores of 9 and 10. The slope percentage increases northeast and northwest of the study area associated with the mountain range. These areas with steep slopes (N18%) were typically assigned a low rating score (1) indicating their minimal effect on the aquifer vulnerability. In the impact of vadose zone layer, the gravel/sandrich alluvial deposits were assigned a high rating value (8), the clayey gravel deposits and the Mesozoic sandstone were assigned the moderate score 6, while the low rating values 2, 3, and 4 were assigned to the Mesozoic chert, mud-rich alluvium, and the tuff and volcanic ash deposits, respectively. Generally, the Kakamigahara aquifer is characterized by high hydraulic conductivity (N m/ s), therefore, was assigned the maximum rating score of 10. The northern part of the aquifer has a relatively lower conductivity ( ~ m/s) hence was assigned low rating values ranging from 1 to 6. The DRASTIC aquifer vulnerability map clearly shows the dominance of bhighq vulnerability classes (shades of red) in the western part of Kakamigahara Heights while the eastern part is characterized by bmoderateq vulnerability (shades of green; Fig. 3). This pattern is mainly dictated by the variation in depth to water from east to west. The elevated northeastern part of the study area displays blowq aquifer vulnerability. This is due to the combination of deep water table, less-porous vadose and aquifer media and steep topography. Ten vulnerability classes are identified in the study area at a 5% interval. The classification scale reflects the same level of detail for the nine highest classes corresponding to 45% of the total number of pixels in the study area. The least vulnerable 55% of the study area is considered less important and is assigned a one-class interval (N45%). The integrated vulnerability map shows that the eastern part of the Kakamigahara aquifer is under a higher risk of contamination despite its moderate intrinsic vulnerability (Fig. 4). The more vulnerable western part of the aquifer is, however, under a lower contamination risk. This is mainly due to the high pollution risk associated with vegetable cultivation. Recently, the analysis of nitrate concentration in 57 water samples from private, farm, monitoring, and public water supply wells in Kakamigahara Heights indicated that 90% of the water samples have shown concentrations above the human affected value (13 mg/l NO 3 ) while 30% exceeded the permissible concentration (44 mg/l NO 3 ) according to Japan Fig. 3. Aquifer vulnerability map of Kakamigahara Heights.

10 136 I.S. Babiker et al. / Science of the Total Environment 345 (2005) Fig. 4. Integrated vulnerability map combining the aquifer vulnerability with the potential pollution sources in Kakamigahara Heights. regulation (Babiker et al., 2004). More than 70% of those wells were associated with intensive vegetable cultivation fields mainly on the eastern side of the Heights Sensitivity of the DRASTIC model The statistical summary of the seven rated parameter maps used to compute the DRASTIC index is provided in Table 3. The highest risk of contamination (mean value is 9) of groundwater in Kakamigahara Heights originates from the net recharge parameter. The soil media, topography, impact of vadose zone, and hydraulic conductivity imply moderate risks of contamination (5 and 6) while depth to water and aquifer media impose a low risk of aquifer contamination (4). Aquifer media, topography, soil media, and hydraulic conductivity are highly variable (CV% are 75, 66.7, 60 and 60, respectively) while depth to water table and impact of vadose zone are moderately variable (CV% are 50 and 40, respectively). Net recharge is the least variable parameter (CV% is 11.1). Table 3 A statistical summary of the DRASTIC parameter maps D R A S T I C Minimum Maximum Mean SD CV (%) S.D. stands for standard deviation and CV for coefficient of variation. The low variability of the parameter implies a smaller contribution to the variation of the vulnerability index across the study area. The rank-order correlation analysis (a summary of the result is provided in Table 4) between the seven DRASTIC parameters indicated that a relatively strong relationship exists between aquifer media and impact of vadose zone (r=0.81), aquifer media and topography (r=0.73), and impact of vadose zone and topography (r=0.56). The former relationship can be attributed to the procedure followed in order to construct the two layers from the subsurface geology map, and the uniform and unconfined nature of the Kakamigahara aquifer. The latter two relationships can be explained by the fact that the Kakamigahara aquifer occupies the alluvial flood plain which has a typically low slope in comparison to the mountain range in the north and northeastern part of the area. Moderate correlation was found between net recharge and soil media Table 4 Summary of rank-order correlation analysis result between seven DRASTIC parameters Correlated parameters Correlation coefficient, r Significance level, p Aquifer media and impact of vadose 0.81 b Aquifer media and topography 0.73 b Impact of vadose and topography 0.56 b Net recharge and soil media 0.46 b Aquifer media and hydraulic 0.30 b0.001 conductivity Depth to water and topography 0.29 b0.001 Only statistically significant (confidence level at/or more than 95%) intercorrelations are tabulated.

11 I.S. Babiker et al. / Science of the Total Environment 345 (2005) (r=0.46) mainly due to the different recharge rates assigned to the soil of the urban land cover and the rest of the area (see Section 4.1). This agrees well with the basic assumptions of DRASTIC model (see Table 1). As expected, depth to water table was slightly correlated to the hydraulic conductivity (r=0.3). Because of the relatively few significant correlations at 95% confidence level (Table 4), the DRASTIC parameters in Kakamigahara Heights were generally considered independent Map removal sensitivity analysis The results of the map removal sensitivity analysis computed by removing one or more data layers at a time are presented in Tables 5 and 6. The statistical analysis of the variation index (sensitivity measure) was applied only to those subareas which are 10 pixels or more in size (633 subareas). Table 4 displays the variation of the vulnerability index as a result of removing only one layer at a time. It is clear that high variation of the vulnerability index is expected upon the removal of the net recharge parameter from the computation (mean variation index=15.1%). This could mainly be attributed to the relatively high theoretical weight assigned to this layer (Table 1) and the high recharge rate characteristic of the Kakamigahara aquifer (rating score 9). The vulnerability index seems to be sensitive to the removal of soil media and topography layers although the two parameters are considered theoretically less important (weights 2 and 1, respectively). This also because of their relatively high contamination risk (mean rating scores 5 and 6, respectively). In Table 6, the variation Table 5 Statistics of the map removal sensitivity analysis Parameter Variation index (%) removed Mean Minimum Maximum S.D. D R A S T I C One parameter is removed at a time. Total of 633 subareas (z10 pixels in size) were considered. S.D. refers to the standard deviation. Table 6 Statistics of the map removal sensitivity analysis Parameters used Variation index (%) Mean Minimum Maximum S.D. D, R, S, T, I and C D, R, S, T, and I R, S, T and I R, S and T R and S R One or more parameters are removed at a time. Total of 633 subareas (z10 pixels in size) were considered. S.D. refers to the standard deviation. of the vulnerability index due to the removal of one or more layers at a time from the model computation is presented. The removal of the map(s) was based on the previous map removal sensitivity measure (Table 5). The layers, which compel less variation of the final vulnerability index, were preferentially removed. The least average variation index resulted after removing only the aquifer media layer (4.5%). As more data layers are excluded from the computation of the vulnerability index, the average variation index increases. Although the layers assumed to be the most effective were considered each time, the interpretation of the growing average is not clear. It might partly be attributed to one or more of the following reasons: the internal variation of the individual parameter, the weights assigned to them, and their weak representation of the real world. Conclusively, considerable variation in the vulnerability assessment is expected if a lower number of data layers have been used Single-parameter sensitivity analysis While the map removal sensitivity analysis presented in the previous section has confirmed the significance of the seven parameters in the assessment of the DRASTIC vulnerability index in the study area, the single parameter sensitivity analysis compares their beffectiveq weights with their btheoreticalq weights. The beffectiveq weight is a function of the value of the single parameter with regard to the other six parameters as well as the weight assigned to it by the DRASTIC model. The beffectiveq weights of the DRASTIC parameters exhibited some deviation from their btheoreticalq weights (Table 7). The net recharge

12 138 I.S. Babiker et al. / Science of the Total Environment 345 (2005) Table 7 Statistics of the single parameter sensitivity analysis Parameter Theoretical weight Theoretical weight (%) Effective weight (%) Mean Minimum Maximum S.D. D R A S T I C S.D. refers to the standard deviation. Total of 633 subareas (z10 pixels in size) were considered. tends to be the most effective parameter in the vulnerability assessment (mean effective wt.% is 27.2%) in agreement with the result from the map removal sensitivity analysis. Its beffectiveq weight exceeds the theoretical weight assigned to it by DRASTIC (17.4%). The hydraulic conductivity also shows high beffectiveq weight (19.7) that exceeds the btheoreticalq weight assigned by DRASTIC (13%). The rest of the parameters, excluding topography, exhibit lower beffectiveq weights compared to the btheoreticalq weights. The significance of net recharge and hydraulic conductivity layers highlights the importance of obtaining accurate, detailed, and representative information about these factors. 6. Conclusion In this paper, we have attempted to assess the aquifer vulnerability of the Kakamigahara groundwater basin employing the empirical index DRASTIC model of the U.S. Environmental Protection Agency (EPA). Seven environmental parameters were used to represent the natural hydrogeological setting of the Kakamigahara aquifer; Depth to water, net Recharge, Aquifer media, Soil media, Topography, Impact of vadose zone, and hydraulic Conductivity. The DRAS- TIC aquifer vulnerability map indicated that the western part of the Kakamigahara aquifer is dominated by bhighq vulnerability classes while the eastern part was characterized by bmoderateq vulnerability classes. The elevated northeastern part of the study area displayed blowq aquifer vulnerability. The integrated vulnerability map shows the high risk imposed on the eastern part of the Kakamigahara aquifer due to the high pollution potential of intensive vegetable cultivation. The more vulnerable western part of the aquifer is, however, under a lower contamination risk. No contamination risk is expected from land use/cover types other than vegetable fields and urban lands in the Kakamigahara. In the particular case of Kakamigahara Heights, land-use seems to be a better predictor of groundwater contamination by nitrate, although assessing natural aquifer vulnerability is still necessary to avoid augmentation of the problem and introduction of other pollutants. The net recharge parameter inflicts the largest impact (9) on the intrinsic vulnerability of the Kakamigahara aquifer. Soil media, topography, impact of vadose zone, and hydraulic conductivity have moderate impacts (5 and 6) on aquifer vulnerability while depth to water table and aquifer media have low impacts (4). The rank-order correlation analysis between the seven DRASTIC parameters indicated that a relatively strong relationship exists between aquifer media and impact of vadose zone (r=0.81), aquifer media and topography (r=0.73), and impact of vadose zone and topography (r=0.56). However, the relatively few significant correlations at 95% confidence level between the DRASTIC parameters in Kakamigahara Heights reveal their independence. The map removal sensitivity analysis indicated that the vulnerability index is highly sensitive to the removal of net recharge, soil media, and topography layers but is least sensitive to the removal of the aquifer media layer. As more data layers are excluded from the vulnerability index computation, the average variation index increases. Therefore, considerable variation in the vulnerability assessment is expected if a lower number of data

13 I.S. Babiker et al. / Science of the Total Environment 345 (2005) layers have been used. The single-parameter sensitivity analysis has shown that net recharge and hydraulic conductivity are the most significant environmental factors which dictate the high vulnerability of the Kakamigahara aquifer. This highlights the importance of obtaining accurate, detailed, and representative information about these factors. The GIS technique has provided an efficient environment for analyses and high capabilities in handling a large quantity of spatial data. The seven model parameters were constructed, classified and encoded employing various map and attribute GIS functions. The DRASTIC vulnerability index, which is defined as a linear combination of factors, was easily computed. Nevertheless, GIS greatly facilitated the implementation of the sensitivity analysis applied on the DRASTIC vulnerability index which otherwise could have been impractical. Acknowledgment We thankfully acknowledge the assistance of the Kakamigahara City Hall officials who have provided the necessary materials to carry on this study. Thanks are also due to Japan Meteorological Agency and Ministry of Land Infrastructure and Transport, Japan for making data available online to support different kinds of research. References Aller L, Bennet T, Leher JH, Petty RJ, Hackett G, DRASTIC: a standardized system for evaluating ground water pollution potential using hydrogeological settings. EPA 600/ ; Babiker IS, GIS for aquifer vulnerability assessment applying DRASTIC and SINTACS (A case study in the Ostrava Karvina region, Czech Republic). MSc thesis, International Institute for Aerospace Survey and Earth Sciences (ITC), Enschede, The Netherlands, 124 pp. Babiker IS, Mohamed AAM, Terao H, Kato K, Ohta K. Assessment of groundwater contamination by nitrate leaching from intensive vegetable cultivation using geographical information system. Environ Int 2004;29(8): Barbash JE, Resek EA. Pesticides in ground water: distribution, trends, and governing factors. Chelsea, MI7 Ann Arbor Press; Barber C, Bates LE, Barron R, Allison H. Assessment of the relative vulnerability of groundwater to pollution: a review and background paper for the conference workshop on vulnerability assessment. J Austr Geol Geophys 1993;14(2/3): Bonham-Carter GF. Geographic information systems for geoscientists: modelling with GIS. Computer Methods in the Geosciences, vol. 13. Pergamon/Elsevier Sci Pub; p. 98. Chung C.F., Fabbri A.G. Prediction models for landslide hazard using a Fuzzy set Approach. In: Marchetti M., Rivas V., editors. Geomorphology and environmental impact assessment. A.A. Balkema; 2001, p Civita M. Le carte della vulnerabilita ddegli acquiferi allt inquinamento (in Italian). In: Pitagora, editor. Teoria and Practica, Bologna; pp. Evans BM, Myers WL. A GIS-based approach to evaluating regional groundwater pollution potential with DRASTIC. J Soil Water Conserv 1990;45: Fabbri AG, Napolitano P. The use of database management and geographical information systems for aquifer vulnerability analysis. Contribution to the International Scientific Conference on the occasion of the 50th Anniversary of the founding of the Vysoka Skola Banska, Ostrava, Czech Republic; ITC-ILWIS. Ilwis 30 academic user s guide. The Netherlands7 International Institute for Aerospace Survey and Earth Sciences (ITC); Kondoh A, Nishiyama J. Changes in hydrogeological cycle due to urbanization in the suburb of Tokyo metropolitan area, Japan. Adv Space Res 2000;26(7): Lodwick WA, Monson W, Svoboda L. Attribute error and sensitivity analysis of map operations in geographical information systems: suitability analysis. Int J Geogr Inf Syst 1990;4(4): McLay CDA, Dragten R, Sparling G, Selvarajah N. Predicting groundwater nitrate concentrations in a region of mixed agricultural land use: a comparison of three approaches. Environ Pollut 2001;115: Melloul A, Collin M. Water quality factor identification by the dprincipal ComponentsT statistical method. Water Sci Technol 1994;34: Merchant JW. GIS-based groundwater pollution hazard assessment: a critical review of the DRASTIC model. Photogramm Eng Remote Sensing 1994;60(9): Mohamed MA. Dynamics of nitrate contamination in the groundwater system (A case study in the Kakamigahara area, Gifu Prefecture, central Japan). MSc thesis, Division of Earth and Environmental Sciences, Graduate School of Environmental Studies, Nagoya University, Nagoya, Japan; pp. Mohamed MA, Terao H, Suzuki R, Babiker IS, Ohta K, Kaori K, et al. Natural denitrification in the Kakamigahara groundwater basin, Gifu Prefecture, central Japan. Sci Total Environ 2003; 307(1 3): Napolitano P, Fabbri AG. Single-parameter sensitivity analysis for aquifer vulnerability assessment using DRASTIC and SINTACS HydroGIS 96: application of geographical information systems in hydrology and water resources management. Proceedings of Vienna Conference. IAHS Pub, vol. 235, p Rosen L. A study of the DRASTIC methodology with emphasis on Swedish conditions. Ground Water 1994;32(2):

14 140 I.S. Babiker et al. / Science of the Total Environment 345 (2005) Secunda S, Collin ML, Melloul AJ. Groundwater vulnerability assessment using a composite model combining DRASTIC with extensive agricultural land use in Israel s Sharon region. J Environ Manage 1998;54: Terao H, Kajikawa M, Morishita Y, Kato K. Influence of pesticides and fertilizers on the groundwater in vegetable field zone. Geochemistry 1985;19: Terao H, Yoshioka R, Kato K. Groundwater pollution by nitrate originating from fertilizer in Kakamigahara Heights, central Japan. IAH Hydr, vol. 4. Hanover7 Verlag Heinz Heise; p Tesoriero AJ, Inkpen E.L, Voss FD. Assessing ground-water vulnerability using logistic regression. Proceedings for the Source Water Assessment and Protection 98 Conference, Dallas, TX; p Thapinta A, Hudak PF. Use of geographic information systems for assessing groundwater pollution potential by pesticides in Central Thailand. Environ Int 2003;29(1): USDA, (United States Department of Agriculture). Natural Resource Conservation Service Soil taxonomy, a basic system of soil classification for making and interpreting soil surveys. Agriculture Handbook, vol Washington DC7 U.S. Government Printing Office; US EPA (Environmental Protection Agency). DRASTIC: a standard system for evaluating groundwater potential using hydrogeological settings, Ada, Oklahoma WA/EPA Series; Vrba J, Zoporozec A. Guidebook on mapping groundwater vulnerability. IAH International Contribution for Hydrogeology, vol. 16. Hannover7 Heise; p Yokoyama T, Makinouchi T. Geology of the Kakamigahara Terrace and the underlying groundwater basin, Gifu Prefecture, central Japan. J Geol Soc Jpn 1991;97(11):