Published online 31 August 2007
Published in J Environ Qual 36:1479-1487 (2007)
DOI: 10.2134/jeq2006.0361
© 2007 American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
TECHNICAL REPORTS
Ground Water Quality
Factors Affecting the Spatial Pattern of Nitrate Contamination in Shallow Groundwater
Dugin Kaown,
Yunjung Hyun,
Gwang-Ok Bae and
Kang-Kun Lee*
School of Earth and Environmental Sciences (BK21 SEES), Seoul National Univ., Seoul 151-747, Korea
* Corresponding author (kklee{at}snu.ac.kr).
Received for publication September 8, 2006.
 |
ABSTRACT
|
|---|
The elevated level of nitrate in groundwater is a serious problem in Korean agricultural areas. To control and manage groundwater quality, the characterization of groundwater contamination and identification of the factors affecting the nitrate concentration of groundwater are significant. The characterization of groundwater contamination at a hydrologically complex agricultural site in Yupori, Chuncheon (Korea) was undertaken by analyzing the hydrochemical data of groundwater within a statistical framework. Multivariate statistical tools such as cluster analyses and Tobit regression were applied to investigate the spatial variation of nitrate contamination and to analyze the factors affecting the NO3–N concentration in a shallow groundwater system. The groundwater groups from the cluster analysis were consistent with the land use pattern of the study area. The clustered group of a gentle-slope area with lower elevations showed higher NO3–N contamination of groundwater than groups on a hillside with higher elevations. Tobit regression results indicated that the agricultural activity in the vegetable fields and barns were the major factors affecting the elevated NO3–N concentration while the land slopes and elevations were negatively correlated with the NO3–N concentration. This shows that topographic characteristics such as land slopes and elevations should be considered to evaluate the land use impact on shallow groundwater quality.
|
INTRODUCTION
|
|---|
AGRICULTURAL activities have been shown to yield high nitrate concentrations in groundwater at many sites (Spalding and Exner, 1993; Choi et al., 2000; McLay et al., 2001; Nolan, 2001; Vinten and Dunn, 2001; Harter et al., 2002; Almasri and Kaluarachchi, 2004; Jun et al., 2005; Liu et al., 2005; Benson et al., 2006). The elevated level of nitrate in groundwater is also a serious problem in Korean agricultural areas. Yupori, a small (1 km2) agricultural basin in Chuncheon (Korea), shows a rising level of nitrate in shallow aquifer groundwater. In 2004, more than 63% of the samples showed nitrate concentrations above the human affected value (3 mg L–1 of NO3–N), while about 23% have exceeded the maximum acceptable level (10 mg L–1 of NO3–N) according to USEPA regulations. Because groundwater is the major source of potable water in Yupori, the groundwater quality must be protected from nitrate contamination. To control and manage groundwater quality, the characterization of groundwater contamination and identification of the factors affecting the nitrate concentration of groundwater at this basin are significant.
The regional characterization of nitrate-contaminated groundwater is of great concern to the agricultural community because the identification of the groundwater zones receiving heavy nitrogen loads from the surface is important to land use planners and environmental regulators (Almasri and Kaluarachchi, 2004). The characterization of groundwater contamination is subject to many uncertainties about potential contamination sources and about the hydrogeological conditions and heterogeneity. To characterize these uncertainties in groundwater contamination, multivariate statistical methods can be usefully applied (Ferguson, 1998; Carlon et al., 2001).
Cluster analysis is a multivariate statistical method that groups samples having similar properties. In groundwater analyses, cluster analysis can be used to classify zones of groundwaters with similar physical and geochemical characteristics. The classified groups can be effectively used to characterize groundwater contamination. The combined use of chemical and physical properties in cluster analysis provides consistent and reliable information to delineate the spatial extent of groundwater contamination (Suk and Lee, 1999; Lee et al., 2001; Swanson et al., 2001; Kim et al., 2005; Zeng and Rasmussen, 2005).
Regression analysis is a commonly used method to identify the factors affecting groundwater quality. However, the ordinary least squares regression method involving the substitution of an arbitrary value for censored data (i.e., data with a specific lower threshold) produces biased and inconsistent parameter estimates when the data are censored or below the threshold limit. Tobit regression can often be used to correct the censoring in this instance (Maddala, 1983; Yen et al., 1996; Lichtenberg and Shapiro, 1997; Liu et al., 1997; Kroll and Stedinger, 1999; Gardner and Vogel, 2005).
When the nitrate concentration has a censored threshold, the factors affecting the groundwater quality can be effectively analyzed by Tobit regression (Yen et al., 1996; Lichtenberg and Shapiro, 1997; Gardner and Vogel, 2005). Yen et al. (1996) used Tobit regression to determine the primary factors that affect nitrate concentration in near-surface aquifers, considering the land use and aquifer properties. Lichtenberg and Shapiro (1997) showed that hydrological characteristics played an important role in the nitrate contamination of groundwater. Gardner and Vogel (2005) considered land use as a major factor in the control of groundwater quality and analyzed its effect on groundwater by Tobit regression. They also suggested that site-specific factors such as the water table and flow path should be included to analyze the land use impacts on groundwater quality.
Recently, several studies have examined site topography as a possible factor influencing the nitrate contamination in groundwater in agricultural areas. Devito et al. (2000) showed that topography affected the hydrologic functioning of riparian zones and had an impact on their nitrate removal efficiency. Vidon and Hill (2004) examined how landscape hydrologic characteristics influenced groundwater nitrate input and removal in riparian zones with different slopes. Rashid and Voroney (2005) observed that N-fertilizer application rate was affected by the position of the slope in the landscape. These topography-related studies considered the impact of land slope on nitrate input and removal at different slope position. However, the impacts of both topography and land use on groundwater nitrate contamination were hardly considered. Thus, in this study, land use and topography (land slope and elevation) were simultaneously evaluated to determine the factors affecting the elevated concentrations of NO3–N in groundwater.
The objectives of this study are (i) to statistically characterize groundwater contamination by cluster analyses and (ii) to determine the major factors affecting the concentrations of NO3–N by means of Tobit regression, considering topographic properties such as land slope and elevation as well as land use.
|
Materials and Methods
|
|---|
Study Area
The Yupori, located at Chuncheon in the Kangwon province of Korea, is surrounded by hills, and part of the area is located on the hillside. Land use patterns of this site seem to be related to the topography. Therefore, the topography should be considered when evaluating the land use impact on the shallow groundwater quality.
Site Description
The Yupori site is a small agricultural basin surrounded by low hills at the northern and eastern borders (Fig. 1
). Most residents of Yupori practice agriculture and livestock farming. To a large extent, farmland and orchards occupy this site along with small barns and pigsties. Accordingly, chemical fertilizers and manure are frequently applied for cultivation purposes.
View larger version (52K):
[in this window]
[in a new window]
|
Fig. 1. Location map of the Yupori site (Chuncheon, Korea) and the layout of the groundwater sampling wells.
|
|
The groundwater systems in Yupori have been monitored for 3 yr since 2002, and high levels of nitrate concentrations have been observed. The data reveals that the observed NO3–N concentrations range from 0.2 to 29.0 mg L–1. The extent of the observed NO3–N contamination is shown in Fig. 2
. The western part shows an elevated level of NO3–N and the eastern part shows a low level of NO3–N.
View larger version (36K):
[in this window]
[in a new window]
|
Fig. 2. Spatial distributions of NO3–N concentrations at each sampling period (April 2002; February 2003; June 2003; August 2003; February 2004; and May 2004). Dashed lines indicate the contour lines of the measured groundwater table in meters for August 2004.
|
|
The aquifer of this area consists of weathered Chuncheon granite overlain by alluvial deposits of 10 to 20 m thickness (Yu et al., 1994). The alluvial deposits consist primarily of two sediment types: silty sand and coarse gravel. The hydraulic conductivity was measured by pumping tests and is in the range of 8.67 x 10–4 to 1.04 x 10–3 cm sec–1. Steep slopes occur in the eastern part of the study area while gentle slopes occur in the western part of the study area. The water table was measured in summer and winter. In the summer, the measured depth to water table ranged from 5.5 m in the western Yupori to 21.7 m in the eastern Yupori. Figure 2 shows the observed water table in the summer. The regional groundwater flow direction of this site is from the northeast to the south. The measured water table shows a seasonal variation, which fluctuates with a vertical range of approximately 3 m over the year. This can be attributed to the fact that more than 50% of the annual precipitation (about 1270 mm yr–1) is concentrated in the wet season (June to August) and less than 10% in the dry season (November to January).
Land Use
A land use map was constructed based on the present land use pattern. The vegetable fields are abundant in the western part of the study area (Fig. 3
). The major vegetables grown in this area include tomatoes, cucumbers, beans, and squashes. The estimated total amount of nitrogen fertilizer applied in the vegetable fields is 200 to 250 kg N ha–1 yr–1. Chemical fertilization practices are multiple applications at the vegetable field. In the eastern part of the study area, orchards cover most of the area at the hillside and the vegetable fields are scattered. Peaches, apples, pears, and grapes are grown in this field. Cow manure is used abundantly in the orchards and the estimated total usage is about 50 kg N ha–1 yr–1. The quantity of chemical fertilizer used in the orchards is much less than that of the vegetable fields and the estimated total amount is less than 50 kg N ha–1 yr–1. Chemical fertilizer was mostly applied in spring and cow manure was applied in spring and autumn at the orchards. Cows were held in barns and cow manure was removed regularly and stored at the outside of the barn for the usage of organic fertilizer.
Hydrochemical Analysis
The groundwater hydrochemistry was analyzed for characterizing the groundwater contamination at the Yupori site. The groundwater samples were collected for chemical analyses from 39 domestic and four monitoring wells. The sampled locations are shown in Fig. 1. The number of collected samples was 20 and 41 for February and August 2004, respectively. The groundwater samples were collected directly from the taps at the domestic wells and by using a submersible pump at the monitoring wells. The sampled groundwater was filtered through a 0.45-µm filter and collected in 60-mL bottles for chemical analysis. The samples for cation analysis were preserved using ultra-pure HNO3. The DO (dissolved oxygen), pH, EC (electrical conductivity), and temperature were measured with standard probes in the field. The DOC (dissolved organic carbon) samples were collected in autoclaved 40-mL amber vials with Teflon-lined septa and no headspace, containing 30 µg of the preservative phosphoric acid.
Cluster Analysis
Cluster analysis is a method for categorizing by grouping similar variables. Hierarchical tree clustering is known to be useful when abundant data are available and clear hydrological models have not yet been developed of the subsurface (Suk and Lee, 1999; Lee et al., 2001; Swanson et al., 2001). Each clustered group shows a specific and similar hydrogeochemical state of groundwater. For clustering analysis, the data were standardized to equalize the influence of parameters.
Cluster analysis was done using 20 samples in February and 41 samples in August. Calculations were performed with the SPSS statistical package 12.0 (SPSS Inc., 2003).
Tobit Regression
Tobit regression is a maximum likelihood estimation technique for censored observations such as an analytical detection limit or a regulatory limit. When data is censored, simple linear regression by substituting an arbitrary value for censored data produces biased and inconsistent parameter estimates (Yen et al., 1996; Liu et al., 1997). Censored regression analysis like Tobit regression is a method to apply censoring in the response variables. Tobit regression model can be expressed in terms of the underlying latent dependent variables yi*
where Ni is the nitrate concentration (mg L–1), 
is a constant, 
is a vector of parameter slope estimates, ln(
) is a vector of independent explanatory variables, and the error term 
i is assumed to be an independently and normally distributed residual error with a mean of zero and variance 
2. The observed dependent variables yi can be expressed in terms of the underlying latent dependent variables yi*, if the nitrate concentration, Ni, is above the censoring threshold value c and yi are set as lnc otherwise (Kroll and Stedinger, 1999; Gardner and Vogel, 2005).
If the error terms in the Tobit model are assumed to be homoscedastic and independent, the model parameters (, , and i) may be efficiently estimated using the maximum likelihood estimation method (Amemiya, 1985).
The significance of each model parameter in the resulting model is determined by the Wald chi-square statistic, which is the ratio of the maximum likelihood estimate of the slope coefficient to its standard error. The maximum likelihood estimators for the parameters were computed from the LIFEREG procedure in SAS by using a Newton-Raphson algorithm (SAS Institute, 2003).
|
Results and Discussion
|
|---|
Hydrochemical Characteristics
The statistics of the geochemical data of the groundwater sampled in February and August 2004 are summarized in Table 1. The data show that the groundwater has high concentrations of NO3–N, K+, Cl–, Ca2+, Mg2+, and SO42– at both sampling dates. This is related to the chemical fertilizers and manure applications ((NH4)2SO4, (Ca, Mg)CO3, and KCl) (Babiker et al., 2004). The enriched species (NO3–N, K+, Ca2+, Mg2+, Cl–, and SO42–) have large standard deviation values. The coefficients of variation are slightly increased in the August data due to the increase in contaminant flux.
View this table:
[in this window]
[in a new window]
|
Table 1. Summarized statistics of the hydrochemical data of groundwater for February and August 2004 (SD: standard deviation; CV: coefficient of variation).
|
|
The analyzed hydrochemical data were plotted on a Piper diagram (Fig. 4
). For the hydrochemical data obtained in both February and August, the groundwater samples are spread out over the wide range of Ca-HCO3 and Ca-Cl(NO3+SO4) water types. The Ca-HCO3 water type shows a low NO3–N concentration (NO3–N < 3 mg L–1), and the Ca-Cl(NO3+SO4) water type shows a high NO3–N concentration (NO3–N > 10 mg L–1). The Ca-HCO3 water type is mostly located in the eastern side of the study area, and Ca-Cl(NO3+SO4) is generally concentrated in the western side of the study area.
View larger version (24K):
[in this window]
[in a new window]
|
Fig. 4. Piper plots showing the various chemical compositions of the groundwater sampled in (a) February and (b) August 2004 (•: Group A, +: Groups B and D, and  : Group C).
|
|
Cluster Analysis
The groundwater sampling points were divided into groups with similar geochemical composition through cluster analysis. The results showed that three groups were clustered for February 2004, and four groups were selected for August 2004 (Fig. 5
). These groups were designated on the map with symbols in Fig. 6
. The wells in Group A were clustered in the western part of the study area, and those in Group C were displayed in the eastern part of the sites. Groups B and D were located between Groups A and C.
View larger version (21K):
[in this window]
[in a new window]
|
Fig. 6. Distribution of each clustered group for the data collected in (a) February and (b) August 2004 (•: Group A, +: Groups B and D, and : Group C).
|
|
Group A was composed of wells that were highly contaminated with nitrate (NO3–N > 10 mg L–1). Groups B and D consisted of sampling points with low NO3–N concentrations (3 mg L–1 < NO3–N < 10 mg L–1). Group C indicated well points showing the background concentration of NO3–N (<3 mg L–1). The clustered groups were consistent with the classification results of water types by piper plots (Fig. 4). Group A comprised sampling wells of Ca-Cl(NO3+SO4) type, Group C comprised sampling wells of Ca-HCO3 type, and Groups B and D comprised sampling wells of the intermediate type.
Box-and-whisker plots of the major ions were plotted to see the chemical variations in each group (Fig. 7
). Each group showed distinct hydrochemical variations. In Group A, NO3–N, Ca2+, Cl–, SO42–, DOC, and EC showed high concentrations, whereas HCO3–, pH, and DO had lower values. In Group C, HCO3–, pH, and DO had high values.
View larger version (36K):
[in this window]
[in a new window]
|
Fig. 7. Box plots of the major chemical components in each clustered group (A, B, C, and D indicate Groups A, B, C, and D, respectively).
|
|
The wells in Group A had nitrate concentrations exceeding the drinking water standard level (10 mg L–1). This result can be explained by the fact that the use of chemical fertilizer was mostly concentrated on the vegetable fields in the western part of the study area where the land slope was lower than that of the eastern part (Fig. 1 and 3). The sampled groundwater in Group C on the eastern parts of the study area showed low concentrations of NO3–N. The input amount of NO3–N on the eastern part of the area was less than that of the western part because it had a lower number of barns and vegetable fields on land with high slopes, as seen from the topography (Fig. 1 and 3). Group B and D, located between Group A and C, were intermediate type wells, which showed NO3–N concentrations exceeding 3 mg L–1 and below 10 mg L–1.
The resulting clustered groups were supported by the piper plot and land use pattern of the site (Fig. 3 and 4). The results were consistent with the land-use pattern of the study area. They also seemed to be associated with the topographic characteristics of the study area.
Tobit Regression Model
Multivariate regression by the Tobit model was used to identify the factors influencing the NO3–N concentration in shallow groundwater. Independent explanatory variables affecting NO3–N concentration in groundwater included the percentage of barns, orchards, and vegetable fields and surface elevation and land slope in this study. The background concentration of NO3–N in the study area is estimated to be approximately 2.6 mg L–1. Thus, NO3–N with a concentration of 2.6 mg L–1 was chosen as a threshold value to evaluate the factors that influence the elevated NO3–N concentrations resulting from anthropogenic activities. The use of the circular areas around wells is an effective method for the correlation of land use and groundwater quality (Barringer et al., 1990). The boundary in the y direction of a hypothetical capture zone of infinite area in the x direction can be defined by
where Q is the well discharge rate (m3 sec–1), B is the aquifer thickness (m), U is the Darcian velocity of groundwater (m sec–1), and x and y are Cartesian coordinates (Javandel and Tsang, 1986; Fetter, 1999). The time-related capture zones can be calculated from the equations by Shafer (1987) and can be simulated using MODPATH. The simulated results obtained from the MODPATH indicated that the time-related capture zone can be approximated to the circular shape. Thus, the maximum width (w) of y direction can be used to approximate the diameter (D) of a circular buffer zone in the finite area. In this study, D of the buffer zone was approximated by
The Darcian velocity of groundwater is estimated to be 1.04 x 10–5 – 3.12 x 10–5 cm sec–1 using the hydraulic conductivity (1.04 x 10–3 cm sec–1) from the pumping test and hydraulic gradient of 0.01–0.03. When the total amount of groundwater for domestic and agricultural usage is about 20000 to 40000 L d–1 and the aquifer thickness is about 20 to 30 m, the estimated D of the buffer zone will be about 74 to 223 m. Thereby, land use within a 100-m radius of each well was taken for a buffer area.
The percentage of barns, orchards, and vegetable fields were used to analyze the effects of land use on NO3–N concentration. Results of Tobit regression are shown in Table 2. The major independent model variables were selected based on their associated p values, which are highly significant at the 0.05 level (95% confidence). It was found that two independent variables—percentage of vegetable fields (VE) and percentage of barns (BA)—showed significant influence on NO3–N concentration. The other variable, percentage of orchards (OR), did not show significant effect (95% confidence) on NO3–N concentration in groundwater. The resulting Tobit model including percentage of vegetable fields (VE) and barns (BA) is shown in Table 3.
The Tobit regression results in Table 2 revealed that percentage of orchards was not significant and showed a negative effect, even though orchards received abundant NO3–N from chemical fertilizer and manure (100 kg N ha–1 yr–1). This motivated the correlation of a Tobit model including topographic factors and land use because orchards are located at a steep-slope area, indicating correlation between orchards and topography. The regression results showed that the percentage of vegetable fields (VE), percentage of barns (BA), and the land slope (SL) were the dominant variables that affected nitrate concentration. Another Tobit model also showed that the percentage of vegetable fields (VE), percentage of barns (BA), and the elevations (EL) were significant variables. The model parameter estimates along with their standard errors, Wald chi-square statistics, and associated p values are shown in Table 4.
View this table:
[in this window]
[in a new window]
|
Table 4. Model parameter estimates for the Tobit model using variables including land use, and land slope and elevation.
|
|
The statistical models strongly support a relationship between the vegetable fields and groundwater quality. The higher value of the Wald chi-square statistic for vegetable fields indicates a greater explanatory power. It may imply that most of the contamination in the study area originates from these nonpoint sources. Model parameters of the land slope and elevation showed negative values. The likelihood of NO3–N contamination in groundwater decreased with the increase in the land slopes and elevations. This is true because the regional groundwater flow direction is from east to west, which corresponds to flow from areas of steeper to shallower slopes in the study area. Nitrate will tend to accumulate in the down gradient area at low elevations since both up gradient and locally applied sources of NO3–N will contribute to observed concentrations down gradient. The observed versus predicted NO3–N concentrations were plotted in Fig. 8
. The Tobit models showed correlations of 0.64 for land slope and 0.65 for elevation as a function of three variables.
View larger version (13K):
[in this window]
[in a new window]
|
Fig. 8. Observed NO3–N concentration vs. the predicted NO3–N concentration based on Tobit regression using variables including land use and topography [(a) land slope and (b) elevation].
|
|
|
Summary and Conclusion
|
|---|
The characterization of shallow groundwater contamination at a basin-type agricultural site was undertaken by analyzing the hydrochemical data of groundwater within a multivariate statistical framework. A cluster analysis indicated that the groundwater at this site was classified into four groups on the basis of the NO3–N concentration, and the resulting clustered groups were consistent with the land use pattern and the associated topographic characteristics of the study area.
The Tobit regression revealed that the agricultural activities in the vegetable fields and the storage and disposal of animal waste are the major contributors to the rising NO3–N contamination in the Yupori groundwater, and that the wells in areas with steep land slopes and high elevations were less susceptible to NO3–N contamination. At the basin-type site, land use and topographic properties were found to influence nitrate concentrations in groundwater. It is concluded that the land use and topography should be considered to evaluate the land use impact on the shallow groundwater quality. The results of the cluster analysis and Tobit regression can provide an effective management guide for future agricultural practices to protect the groundwater quality at the site.
|
ACKNOWLEDGMENTS
|
|---|
This study was supported by the Advanced Environmental Biotechnology Research Center (AEBRC) at POSTECH and the Sustainable Water Resources Research Center of the 21st Frontier Project (Project #3-4-2).
|
NOTES
|
|---|
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher.
|
REFERENCES
|
|---|
- Almasri, M.N., and J.J. Kaluarachchi. 2004. Assessment and management of long-term nitrate pollution of ground water in agriculture-dominated watersheds. J. Hydrol. 295:225–245.[CrossRef]
- Amemiya, T. 1985. Advanced econometrics. Harvard Univ. Press, Cambridge, MA.
- Babiker, I.S., A.A. Mohamed, H. Terao, K. Kato, and K. Ohta. 2004. Assessment of groundwater contamination by nitrate leaching from intensive vegetable cultivation using geographical information system. Environ. Int. 29:1009–1017.[CrossRef][Web of Science][Medline]
- Barringer, T.H., D. Dunn, W.A. Battaglin, and E.F. Vowinkel. 1990. Problems and methods involved in relating land use to groundwater quality. Water Resour. Bull. 226(1):1–9.
- Benson, V.S., J.A. VanLeeuwen, J. Sanchez, I.R. Dohoo, and G.H. Somers. 2006. Spatial analysis of land use impact on ground water nitrate concentrations. J. Environ. Qual. 35:421–432.[Abstract/Free Full Text]
- Carlon, C., A. Critto, A. Marcomini, and P. Nathanail. 2001. Risk based characterization of contaminated industrial site using multivariative and geostatistical tools. Environ. Pollut. 111:417–427.[CrossRef]
- Choi, J.D., S.O. Jang, B.Y. Choi, and S.H. Yoo. 2000. Monitoring study on groundwater quality of an alluvial plain in the north Han River basin. J. Korean Soc. Water Qual. 16:283–293.
- Devito, K.J., I.F. Creed, R.L. Rothwell, and E.E. Prepas. 2000. Landscape controls on phosphorus loading to boreal lakes: Implications for the potential impacts of forest harvesting. Can. J. Fish. Aquat. Sci. 57:1977–1984.[CrossRef]
- Ferguson, C.C. 1998. Techniques for handling uncertainty and variability in risk assessment models. Umweltbundesamt, Berlin.
- Fetter, C.W. 1999. Contaminant hydrogeology. Prentice Hall, Inc., Upper Saddle River, NJ.
- Gardner, K.K., and R.M. Vogel. 2005. Predicting groundwater nitrate concentration from land use. Ground Water 43:343–352.[CrossRef][Web of Science][Medline]
- Harter, T., H. Davis, M.C. Mathews, and R.D. Meyer. 2002. Shallow groundwater quality on diary farms with irrigated forage crops. J. Contam. Hydrol. 55:287–315.[CrossRef][Web of Science][Medline]
- Javandel, I., and C.-F. Tsang. 1986. Capture-Zone type curves: A tool for aquifer cleanup. Ground Water 24:616–625.[CrossRef][Web of Science]
- Jun, S.C., G.O. Bae, K.K. Lee, and H.J. Chung. 2005. Identification of the sources of nitrate contamination in ground water below an agricultural site, Jeungpyeong. Korean J. Environ. Qual. 34:804–815.
- Kim, J.H., R.H. Kim, J.H. Lee, T.J. Cheong, B.W. Yum, and H.W. Chang. 2005. Multivariate statistical analysis to identify the major factors governing groundwater quality in the coastal area of Kimje, South Korea. Hydrol. Processes 19:1261–1276.[CrossRef]
- Kroll, C.N., and J.R. Stedinger. 1999. Development of regional regression relationships with censored data. Water Resour. Res. 35(3):775–784.
- Lee, J.Y., J.Y. Cheon, K.K. Lee, S.Y. Lee, and M.H. Lee. 2001. Statistical evaluation of geochemical parameter distribution in a ground water system contaminated with petroleum hydrocarbon. J. Environ. Qual. 30:1548–1563.[Web of Science][Medline]
- Lichtenberg, E., and L.K. Shapiro. 1997. Agriculture and nitrate concentrations in Maryland community water system wells. J. Environ. Qual. 26:145–153.[Web of Science]
- Liu, S., J. Lu, D.W. Kolpin, and W.Q. Meeker. 1997. Analysis of environmental data with censored observations. Environ. Sci. Technol. 31:3358–3362.
- Liu, A., J. Ming, and R.O. Ankumah. 2005. Nitrate contamination in private wells in rural Alabama, United States. Sci. Total Environ. 346:112–120.[CrossRef][Medline]
- Maddala, G.S. 1983. Limited dependent and qualitative variables in econometrics. Cambridge Univ. Press, Cambridge, MA.
- McLay, C.D.A., R. Dragten, G. Sparling, and N. Selvarajah. 2001. Predicting groundwater nitrate concentrations in a region of mixed agricultural land use: A comparison of three approaches. Environ. Pollut. 115:191–204.[CrossRef][Medline]
- Nolan, B.T. 2001. Relating nitrogen sources and aquifer susceptibility to nitrate in shallow ground waters of the Unites States. Ground Water 39:290–299.[CrossRef][Web of Science][Medline]
- Rashid, M.T., and R.P. Voroney. 2005. Field-scale application of oily food waste and nitrogen fertilizer requirements of corn at different landscape positions. J. Environ. Qual. 34:963–969.[Abstract/Free Full Text]
- SAS Institute. 2003. SAS for Windows. Version 9.1. SAS Inst., Cary, NC.
- Shafer, M.J. 1987. Reverse pathline calculation of time-related capture zones in nonuniform flow. Ground Water 25:283–289.[CrossRef][Web of Science]
- Spalding, R.F., and M.E. Exner. 1993. Occurrence of nitrate in groundwater-A review. J. Environ. Qual. 22:392–402.[Web of Science]
- SPSS Inc. 2003. SPSS for Windows. Version 12.0. SPSS Inc., Chicago, IL.
- Suk, H.J., and K.K. Lee. 1999. Characterization of a ground water hydrochemical system through multivariate analysis: Clustering into ground water zones. Ground Water 37:358–366.[CrossRef][Web of Science]
- Swanson, S.K., J.M. Bahr, M.T. Schwar, and K.W. Potter. 2001. Two-way cluster analysis of geochemical data to constrain spring source waters. Chem. Geol. 179:73–91.[CrossRef][Web of Science]
- Vidon, P.G., and A.R. Hill. 2004. Landscape controls on nitrate removal in stream riparian zones. Water Resour. Res. 40:W03201 doi:10.1029/2003WR002473.
- Vinten, A.J.A., and S.M. Dunn. 2001. Assessing the effects of land use on temporal change in well water quality in a designated nitrate vulnerable zone. Sci. Total Environ. 2001:253–268.
- Yen, S.T., S. Liu, and D.W. Kolpin. 1996. Analysis of nitrate in near-surface aquifer in the midcontinental United States: An application of the inverse hyperbolic sine Tobit model. Water Resour. Res. 32:3003–3011.[CrossRef]
- Yu, J.Y., I.K. Choi, and H.S. Kim. 1994. Geochemical characteristics of the surface water depending on the bedrock type in the Chuncheon area. J. Geol. Soc. Korea 30(3):307–324.
- Zeng, X., and T.C. Rasmussen. 2005. Multivariate statistical characterization of water quality in Lake Lanier, Georgia, USA. J. Environ. Qual. 34:1980–1991.[Abstract/Free Full Text]
TOP
ABSTRACT
INTRODUCTION
