using indicator kriging for the evaluation of arsenic potential contamination in an abandoned mining...

Science of the Total Environment 442 (2013) 545–552

Contents lists available at SciVerse ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r .com/ locate /sc i totenv

Using indicator kriging for the evaluation of arsenic potential contamination in anabandoned mining area (Portugal)

I.M.H.R. Antunes ⁎, M.T.D. AlbuquerqueCVRM-Geossystems Centre, ISTUL, Lisbon and Polytechnic Institute of Castelo Branco, 6001-909 Castelo Branco, Portugal

H I G H L I G H T S

► Arsenic is associated with sulfide mineralization and is toxic in the environment at low levels.► Intrinsic and specific vulnerabilities quantify anthropogenic activities.► Arsenic anomalies are mainly associated with the water drainage from abandoned mining activities.► The waters are not fit for human consumption.

⁎ Corresponding author. Tel.: +351 272339900; fax:E-mail address: [email protected] (I.M.H.R. Antune

0048-9697/$ – see front matter © 2012 Elsevier B.V. Allhttp://dx.doi.org/10.1016/j.scitotenv.2012.10.010

a b s t r a c t
a r t i c l e i n f o
Article history:Received 10 August 2012Received in revised form 1 October 2012Accepted 1 October 2012Available online 5 December 2012

Keywords:Sulfide minesArsenicWatersContaminationIndicator krigingSegura

Mining andmineral-processing activities canmodify the environment in a variety ofways. Sulfidemineralizationis notorious for producing waters with highmetal contents. Arsenic is commonly associated with sulfideminer-alization and is considered to be toxic in the environment at low levels. The studied abandoned mining area islocated in central Portugal and the resulting tailings and rejected materials were deposited and exposed to theair and water for the last 50 years. Sixteen water sample-points were collected. One of these was collectedoutside the mining influence, with the aim of obtaining a reference background.The risk assessment, concerning the proximity to abandoned mineralized deposits, needs the evaluation ofintrinsic and specific vulnerabilities aiming the quantification of the anthropogenic activities. In this study,two indicator variables were constructed. The first one (I1), a specific vulnerability, considers the arsenicwater supply standard value (0.05 mg/L), and the probability of it being exceeded is dependent on thegeologic and hydrological characteristics of the studied area and also on the anthropogenic activities. Thesecond one (I2), an intrinsic vulnerability, considers arsenic background limit as cut-off value, and dependsonly on the geologic and hydro-geological characteristics of the studied area.At Segura, the arsenic water content found during December 2006 (1.190 mg/L) was higher than the arsenicwater content detected in October 2006 (0.636 mg/L) which could be associated to the arsenic released fromFe oxy-hydroxide. At Segura abandoned mining area, the iso-probability maps of October 2006 and December2006, show strong anomalies associated with the water drainage from abandoned mining activities. Near thevillage, the probability of exceeding the arsenic background value is high but lower than the probability ofexceeding the arsenic water supply value. The arsenic anomalies indicate a high probability for water arseniccontamination and those waters should not be used for human consumption.

© 2012 Elsevier B.V. All rights reserved.

1. Introduction

Portuguese polymetallicmining activities were very important for theeconomy and were actively developed until the early 1970s. Since then,metal production has declined and most of the mines are now closedand/or have been abandoned. Mining activities extract and process valu-able pit materials, which have large surfaces and are very susceptible toboth erosion and chemical weathering, causing a potential danger to theenvironment. The abandoned mining sites are frequently located close

+351 272339901.s).

rights reserved.

to occupied rural areas (Allen et al., 1996) and some of the waters areused for agriculture or human consumption without any assessmentof environmental and human health risks (e.g., Antunes et al.,2002; Abreu et al., 2008; Carvalho et al., 2009; Gomes et al., 2010).

In general, high concentrations of metals and metalloids in tailingsare due to sulfide oxidation and the subsequent redistribution of tracemetals and metalloids by secondary Fe precipitates and phase for-mation, as well as adsorption on clay fraction (Carlsson et al., 2002;Heikkinen and Räisänen, 2009). Elevated arsenic levels in surfacewatersare commonly associated with sulfide mineralizations (e.g., Antunes etal., 2002; Lee et al., 2007; Carvalho et al., 2009) and are considered tobe toxic in the environment at low levels. In addition, arsenic

http://dx.doi.org/10.1016/j.scitotenv.2012.10.010

mailto:[email protected]

http://dx.doi.org/10.1016/j.scitotenv.2012.10.010

http://www.sciencedirect.com/science/journal/00489697

3

4

56

7

8

9

1011

12

13

14

1615

2

1

C. Branco

a)b)

Fig. 1. a) Location of the Segura mine area on the map of Portugal; b) Collected water samples (▲). C.Branco — Castelo Branco; Village — Segura; abandoned mines.

Table 1Arsenic (mg/L) contents of waters from Segura.

Sample points October 2006 December 2006

1 0.006 0.0032 0.636 1.1903 0.002 0.0014 0.409 0.4905 0.017 0.0146 0.082 0.0157 0.039 0.0018 0.017 0.0019 0.008 0.01210 0.033 0.04111 0.046 0.06312 0.006 0.05613 0.012 –

14 0.005 0.00115 0.004 0.00216 0.001 –

– below detection limit.

546 I.M.H.R. Antunes, M.T.D. Albuquerque / Science of the Total Environment 442 (2013) 545–552

contamination by mining activities tends to be confined in space andarsenic contents can reach levels thousands of times higher than thenatural background (e.g., Smedley and Kinniburgh, 2002).

The main purpose of this research is a detailed study of arsenicgeochemical behavior and the evaluation of potential water contam-ination related to an abandoned mining area. The spatial distributionof water properties commonly exhibits some heterogeneity whichmay be difficult to capture when few in-situ data are available dueto time and budget constraints. The uncertainty associated with pre-dicted values is one of the most sensitive issues in spatial and tempo-ral mapping, which requires the use of probabilistic methods.

Geostatistics iswidely used tomodel the spatial variability of environ-mental attributes and map the distribution of both the attribute valuesand the probability of exceeding specific thresholds. In particular, indica-tor kriging is the most commonly used non-parametric geostatisticalmethod (e.g., Liu et al., 2007; Lee et al., 2008; Ungaro et al., 2008;Hassan and Atkins, 2011) as no assumptions on the underlying sampledistribution are made, and the coding of data into 0 and 1 (indicatortransform) makes the predictor robust to outliers (Cressie, 1993). At anunsampled location, the values estimated by indicator kriging representa probability of the attribute to exceed a specified threshold (Goovaerts,1997, 1999). Indicator kriging has been frequently used to soil pollutionmapping. Smith et al. (1993) and Oyedele et al. (1996) used multipleindicator kriging to analyze soil's characteristics. Liu et al. (2004) appliedindicator kriging to evaluate the arsenic contamination potential in agri-cultural land. Liu et al. (2010) studied the arsenic contamination and itspotential health risk implications at an abandoned tungsten mine whileGoovaerts et al. (2005) modeled geostatistically the spatial variability ofarsenic in groundwater of southeast Michigan.

Analyzing the spatial and temporal dissemination of arsenic inwaters allows a better understanding of its mobility and a more accu-rate risk assessment. This work uses indicator kriging to assess thepotential contamination of arsenic associated to the old Sn–W andBa–Pb abandoned mines from Segura (Central Portugal).

2. Study area

The Sn–W and Ba–Pb Segura abandoned mines are located about60 km east of Castelo Branco, central Portugal, close to the Portuguese

Spanish borders (Fig. 1a). The area is approximately 22 km2 and in-cludes the village of Segura (Fig. 1b). Agriculture is the primary activ-ity requiring a large amount of superficial and sub superficial waterfor irrigation. The other water use is human consumption. The regionis characterized by a dry climate and only a few streams are not dryduring the summer (Antunes et al., 2002).

Nowadays it is difficult to locate the mineralized veins. They aregenerally found through the old mine workings, which were carriedout from 1942 to 1953 (Antunes et al., 2002). In this region,100 tons of cassiterite and 12 tons of wolframite were extractedfrom Sn–W quartz veins, while 525 tons of barite and 211 tons of ga-lena were extracted from Ba–Pb quartz veins. Arsenopyrite in themineralized veins was oxidized and dissolved from these veins,mine tailings and waste heaps (Antunes et al., 2002). Since themine closure 50 years ago, large amounts of tailings and rejected ma-terials have been exposed to the air and water without any develop-ment or revegetation.

Fig. 2. Box-plot representation of As from waters of Segura (October 2006). Point 2:Outlier.

Fig. 3. Box-plot representation of As from waters of Segura (December 2006). Point 2:Outlier.

547I.M.H.R. Antunes, M.T.D. Albuquerque / Science of the Total Environment 442 (2013) 545–552

3. Material and methods

3.1. Sampling and datasets

To assess the contamination of superficial and sub superficial wa-ters, sixteen water samples were collected (Fig. 1b), according to thelocation of probable input sources which are abandoned mines. Thelocal geochemical background is identified with the concentrationmeasured in the water sample collected outside the contaminationsource influence (sample number 1; Fig. 1b). The other fifteen sam-ples were collected downstream of the abandoned mining sites(Fig. 1b), thus inside the immediate influence zone of the contamina-tion sources. The same 16 locations were sampled twice during thefall of 2006 (October and December), leading to a total of 32 watersamples.

Natural water sampling points were selected and localized by GPSand geo-referenced to UTM coordinates. Water samples were collect-ed using polyethylene bottles and some physical parameters were de-termined in situ. After sampling, the waters were acidified and kept at4 ºC. Arsenic was determined by Spectrometer Perkin Elmer 303flame atomic absorption at the Department of Earth Sciences, Univer-sity of Coimbra, with a precision of 5% (Antunes et al., 2002).

To assess arsenic environmental toxicity and potential impact onhuman health two indicator variables were constructed. The first indi-cator (I1) used the arsenic water supply standard contamination value(0.05 mg/L; Portuguese Law, 2007) as cut-off. The probability of ex-ceeding this cut-off depends not only on the geological and hydrologicalcharacteristics of the studied area, but also on the anthropogenic activ-ities, and can be interpreted as a specific vulnerability probability. Thesecond indicator (I2) used the arsenic background value (i.e. value ofsample #1) as cut-off andwas computed in order to quantify the intrin-sic vulnerability probability since the background value depends onlyon the geologic and hydro-geological characteristics of the studiedarea. The evaluation of potential arsenic contamination caused by theproximity to abandoned mineralized deposits, needs the evaluation of

Table 2Descriptive statistics from As contents (mg/L) of waters from Segura (October 2006).

Minimum 0.002Maximum 0.6361° Quartil 0.006 Background

Median 0.0173° Quartil 0.048 Water supply standardAverage 0.086Variance (n) 0.029

both intrinsic and specific vulnerabilities to quantify the impact of an-thropogenic activities (Albuquerque and Antunes, 2010).

3.2. Geostatistical methods

3.2.1. Variogram analysisGeostatisticalmethodologies are based on the theory of regionalized

variables (Matheron, 1971) which states that attributes within an areaexhibit both random and spatially structured properties (Journel andHuijbregts, 1978). Sample variograms should first be estimated andmodeled in order to quantify the spatial variability of random variablesas a function of their separation distance. In practice, an experimentalvariogram, is computed as:

γ̂ hð Þ ¼ 12N hð Þ

XN hð Þ

i¼1

z ui þ hð Þ−z uið Þ½ �2( )

where γ̂ hð Þ denotes the variogram for a distance lag h; N(h) representsthe number of data pairs for that lag h, and z(ui) and z(ui+h) are thevalues of the regionalized variable of interest (e.g. arsenic concentra-tion) at locations ui and ui+h respectively. The experimental variogramγ̂ hð Þ is then fitted by a theoretical model, γ(h) (Isaaks and Srivastava,1989).

3.2.2. Indicator krigingIndicator kriging is a non-parametric geostatistical method for es-

timating the probability of exceeding a specific threshold value, zk, ata given location. In indicator kriging, the stochastic variable, Z(u), istransformed into an indicator variable with a binary distribution, asfollows:

I u; zkð Þ ¼ 1;0;

if Z uð Þ≤zk;K ¼ 1;2;…;mOtherwise

:

�

Table 3Descriptive statistics from As contents (mg/L) of waters from Segura (December 2006).

Minimum 0.000Maximum 1.1901° Quartil 0.001

0.003 Background

Median 0.0123° Quartil 0.041 Water supply standardAverage 0.113Variance (n) 0.085

image of Fig.�3

I2 - October

I2 - December

I1 - October

I1 - December

Nugget 0.0988841637132209Number of basic models

1

Model 1Type SphericalSill 0.20017545

Ellipse parametersAngle: 0.0Min range: 1529.46839Max range: 1529.46839


1




2




1



1.76

0.0

0.0 4000.0hl

0.0 4000.0hl

0.0 4000.0hl

0.0 4400.0hl

γγ

2.2

0.0

1.73

3333

0.0

γ0.

550.

0γ

Fig. 4. Experimental indicator variogram with cut-off values of the water supply standard (indicator I1) and the arsenic background limit (indicator I2) measured in October 2006and December 2006 and the variogram fitted model (SpaceStat software).


The expected value of I(u;zk), conditional to n surrounding data,can be expressed as:

E I u; zk nð Þj Þð � ¼ Prob Z uð Þ≤zk nð Þj g ¼ F u; zk nð Þj Þðf½

P u; zk nð Þj Þ ¼ 1−F u; zk nð Þj Þðð

where F(u; zk|(n)) is the value of the conditional cumulative distribu-tion function of Z(u) for a threshold zk, and P(u; zk|(n)) is the proba-bility that Z(u)>zk. At an unsampled location, u0, the indicatorkriging estimator is written as:

I� u0; zkð Þ ¼Xnj¼1

λj zkð ÞI uj; zk� �

where I(uj;zk) represents the values of the indicator at sampled loca-tions, uj, j=1,2,3,…,n, and λj is the weight assigned to I(uj;zk) in the

estimation of I*(u0;zk). The estimator must be unbiased and withminimum estimation error variance; that is:

E I� u0; zkð Þ−I u0; zkð Þ� � ¼ 0

Var I� u0; zkð Þ−I u0; zkð Þ� �is minimum:

Both conditions are fulfilled by computing the weights, λj, as asolution of the following system of linear equations (Goovaerts, 1997),

Xnj¼1

λj zkð ÞγI ui−uj; zk� �

−μ zkð Þ ¼ γI ui−u0; zkð Þ

i ¼ 1;…;nXnj¼1

λj zkð Þ ¼ 1

8>>>>>><>>>>>>:where μ(zk) is the Lagrange multiplier, уI(ui−uj; zk) specifies thevariogram value between the indicator variables at the uith and ujth

image of Fig.�4

Fig. 5. Results of cross validation conducted for the two indicators (I1, I2) and the two sampling times (October, December 2006). The performance criteria are the kriging meanerror (KME) and the mean square standard error (MSSE).


sampling points; уI(ui−u0; zk) is the variogramvalue between the indi-cator variables at the ui-th sampling point and u0. The indicatorvariogram уI(h; zk) is obtained by applying the procedure described inSection 3.2.1 to indicator values i(u;zk) instead of attribute values z(u).

3.2.3. Cross validationIn a cross-validation or “leave-one-out” procedure, measured data

are dropped one at a time and re-estimated from the remainingneighboring data. Observed and estimated indicators are then com-pared using the following two statistics (Isaaks and Srivastava,1989; Deutsch and Journel, 1998):

KME ¼ 1N

XNα¼1

i� uα; zkð Þ−i uα; zkð Þ� �

MSSE ¼ 1N

XNα¼1

i� uα; zkð Þ−i uα zkð Þ½ �2σ2 uα; zkð Þ

where i*(uα;zk) and i(uα;zk) are the estimated and measured indica-tor values at the uαth location, respectively, N is the number of mea-sured points, and σ2(uα;zk) is the kriging variance. The kriging meanerror (KME) and mean square standard error (MSSE) should be closeto zero and one, respectively.

4. Results and discussion

4.1. Statistical analysis

The arsenic content of waters collected during October andDecember 2006 in the Segura abandoned mining area is listed in

Table 1. A classical statistical analysis was first performed in orderto evaluate the behavior of water arsenic concentrations in two dis-tinct hydrological seasons: October 2006, the dry season (Fig. 2;Table 2) and December 2006, the wet season (Fig. 3; Table 3).

Summary statistics (average, median and variance) indicates thatthe distribution of this variable display a strong positive asymmetry,due to the presence of a few very high concentrations (outliers) inboth sampled months (Figs. 2 and 3). These outliers however shouldnot be discarded since the aim of this work is risk assessment. The 1stquartile obtained during October corresponds to the local geochemi-cal background (Table 2). For October and December, the 3rd quartileis similar and corresponds to the arsenic water supply content(Tables 2 and 3). December presents the highest arsenic water con-tent (1.190 mg/L; Tables 1 and 3). The higher variance values(Table 3) can be associated to water samples with arsenic contentbelow the detection limit, considering the analytical method. The me-dian values for October and December are similar (Tables 2 and 3).The lower arsenic values recorded in October (Table 2) were unex-pected according to the concentration effect. A possible explanationis the arsenic retention by oxy-hydroxide materials. In December,the highest arsenic contents observed (Table 3), could be associatedwith oxy-hydroxide Fe dilution, which retains this metal.

4.2. Geostatistical analysis

The use of indicator geostatistics allows the mapping of the prob-ability for attribute values to exceed a certain cut-off value. Theseprobability maps are very useful for decision makers, because oftheir easy interpretation and the ability to produce as many maps asthresholds of interest (e.g., Ribeiro et al., 1997; Stigter et al., 2006).In the present application, the two cut-off values were 0.05 mg/L

N

1

Estimated values:

Observed points

Village

1 – Background observation

m

I1 - October

1

Estimated values:

Observed points

Village


N

mI1 – December

Fig. 6. Maps of the probability that the arsenic water content exceeds the water supply standard cut-off value (indicator I1) or the arsenic background limit (indicator I2) for theOctober 2006 and December 2006 sampling events.


(Indicator I1) and the monthly corresponding background (IndicatorI2) (Tables 2 and 3). The analysis was conducted using SpaceStat(BioMedware, 2011) to compute the experimental variograms andperform indicators kriging whereas the geostatistical analyst (ESRI,2004) was used for the cross validation and KME and MSSEcomputation.

Fig. 4 shows the experimental omnidirectional variograms withthe spherical model fitted for the indicator variables, I1 and I2,recorded in October 2006 and December 2006, respectively. Forboth indicators, the spatial structure (e.g. range, nugget effect ofvariogram) is greater in October compared to December. The studywatershed was mostly dry during October 2006 and the metals' dis-persion was controlled mostly by random sub superficial water flowwhich imposed a weaker spatial structure for the indicator variables.Cross validation led to acceptable results for both variables althoughthe two statistics KME and MSSE are closer to their ideal value forthe first indicator relative to I2 (Fig. 5).

The probability maps generated by indicator kriging are displayedin Fig. 6. The I1 probability map, for October 2006, shows two stronganomalies, located in the north of the area and corresponding to

water drainage from abandoned mines (Figs. 1 and 6). The higher I2probabilities, for October 2006, are similar to the ones observed forI1 (Fig. 6). Near Segura, the probability of exceeding the backgroundvalues is high but lower than the probability to exceed the water sup-ply value (Fig. 6). During December 2006, two other strong anomaliesare identified (Fig. 6). One of them coincides with the October northarsenic anomaly while the other one is found near Segura (Fig. 6).These anomalies indicate a high probability for arsenic water contam-ination and those waters should not be used for human consumption.Almost all arsenic anomalies are associated with the proximity to themineralization and the old mining activities. However, the irregulararsenic probability pattern can be supported by the evidence that ar-senopyrite (arsenic sulfide mineral) is not the principal mineral ofthese mineralized veins, occurring associated with other sulfide min-erals. Therefore some arsenic amount can be hypothesized to be acontribution due to precipitation processes.

The comparison of I1 and I2 probability maps reveals different spa-tial pattern (Fig. 6). The probability of exceeding the local geochemi-cal arsenic background is high across the entire study area. However,the highest probabilities tend to be collocated with the highest

image of Fig.�6

1

Estimated values:

Observed points

Village

1– Background observation

N

m

I2 - October

1

Estimated values:

Observed points

Village


N

m

I2 – December

Fig. 6 (continued).


probabilities of exceeding the water supply standard. The lower prob-ability values for I2 (exceeding seasonal/local geochemical back-ground) are located in the SW part of the study area where thelower arsenic probability of exceeding the water supply standardwere also found.

5. Conclusions

Arsenic is commonly associated with sulfide mineralizations andis toxic at low levels. At Segura area, the highest arsenic water contentfound during December could be associated to the dilution of Feoxy-hydroxide, which retains toxic metals.

The risk assessment of arsenic associated to abandoned mines canbe evaluated with the probability of exceeding a specific (I1 — arsenicwater supply value) and intrinsic (I2 — arsenic background) vulnerabil-ity. During October and December, the iso-probability maps show

strong anomalies associated with the water drainage from abandonedmining activities.

At Segura, the probability of exceeding the arsenic backgroundvalue is high but lower than the probability of exceeding the arsenicwater supply value. However, the arsenic anomalies indicate a highprobability for water arsenic contamination and those waters shouldnot be used for human consumption.

Acknowledgments

Thanks are due to Prof. Ana Neiva for providing laboratory facili-ties in the Department of Earth Sciences, University of Coimbra,Portugal. This research was carried out under the CVRM GeosystemsCentre, ISTUL, Portugal. The reviewers' comments helped to improvethis paper. English grammar and spelling of the manuscript have beenrevised by a native person.


References

Abreu MM, Matias MJ, Magalhães MCF, Basto MJ. Impacts on water, soil and plantsfrom the abandoned Miguel Vacas copper mine, Portugal. J Geochem Explor2008;96:161–70.

Albuquerque MTD, Antunes IMHR. Probability mapping of arsenic vulnerabilities— riskassessment, a Portuguese study case. Hungary; Budapest: IAMG — InternationalAssociation for Mathematical Geology; 2010.

Allen SK, Allen JM, Lucas S. Concentrations of contaminants in surface water samplescollected in west-central Indiana impacted by acidic mine drainage. Environ Geol1996;27:34–7.

Antunes IMHR, Neiva AMR, Silva MMVG. The mineralized veins and the impact of oldmine workings on the environment at Segura, central Portugal. Chem Geol2002;190:417–31.

BioMedware. SpaceStat user manual, version 2.2. New York: BioMedware, Inc.; 2011.Carlsson E, Thunberg J, Öhlander B, Holmström H. Sequential extraction of

sulphide-rich tailings remediated by the application of till cover, Kristinebergmine, northern Sweden. Sci Total Environ 2002;299:207–26.

Carvalho PCS, Neiva AMR, Silva MMVG. Geochemistry of soils, stream sediments andwaters close to abandoned W–Au–Sb mines at Sarzedas, Castelo Branco, centralPortugal. Geochem Explor Environ Anal 2009;9:341–52.

Cressie N. Statistics for spatial data. New York: Wiley; 1993.Deutsch CV, Journel AG. GSLIB: geostatistical software library and user's guide. 2nd ed.

New York, USA: Oxford University Press; 1998.ESRI. ArcGIS desktop, version 9.3. Washington, Reedlands: Environmental Systems

Research Institute, Inc.; 2004.Gomes MEP, Antunes IMHR, Neiva AMR, Silva PB, Rodrigues AM. Geochemistry of

waters associated with the old mine workings at Fonte Santa (NE of Portugal).J Geochem Explor 2010:153–65.

Goovaerts P. Geostatistics for natural resources evaluation. New York: Oxford UniversityPress; 1997.

Goovaerts P. Geostatistics in soil science: state-of-the-art and perspectives. Geoderma1999;89:1-45.

Goovaerts P, AvRuskin G, Meliker J, Slotnick M, Jacquez GM, Nriagu J. Geostatisticalmodeling of the spatial variability of arsenic in groundwater of Southeast Michigan.Water Resour Res 2005;41(7). [W07013 10.1029].

Hassan MM, Atkins PJ. Application of geostatistics with indicator kriging for analyzingspatial variability of groundwater arsenic concentrations in Southwest Bangladesh.J Environ Sci Health A 2011;46(11):1185–96.

Heikkinen PM, Räisänen ML. Trace metal and As solid-phase speciation in sulphidemine tailings — indicators of spatial distribution of sulphide oxidation in activetailings impoundments. Appl Geochem 2009;24:1224–37.

Isaaks EH, Srivastava RM. An introduction to applied geostatistics. New York: OxfordUniversity Press; 1989. p. 278–322.

Journel AG, Huijbregts CJ. Mining geostatistics. San Diego: Academic Press; 1978.Law Portuguese. Decree 306/2007 — Portuguese legislation on water quality. Diário da

República 2007;I-A:5747–65. [Portugal].Lee JJ, Jang CS, Wang SW, Liu CW. Evaluation of potential health risk of arsenic-affected

groundwater using indicator kriging and dose response model. Sci Total Environ2007;384:151–62.

Lee JJ, Liu CW, Jang CS, Liang CP. Zonal management of multi-purpose use of water fromarsenic-affected aquifers by using a multi-variable indicator kriging approach.J Hydrol 2008;359(3–4):260–73.

Liu CW, Jang CS, Liao CM. Evaluation of arsenic contamination potential using indicatorkriging in the Yun-Lin aquifer (Taiwan). Sci Total Environ 2004;321:173–88.

Liu JCS, Lu CW, Lin KL. Delimitation of arsenic-contaminated groundwater usingrisk-based indicator approaches around blackfoot disease hyperendemic areas ofsouthern Taiwan. Environ Monit Assess 2007;134(1–3):293–304.

Liu CP, Luo CL, Gao Y, Li FB, Lin LW, Wu C, et al. Arsenic contamination and potentialhealth risk implications at an abandoned tungsten mine, southern China. EnvironPollut 2010:820–6.

Matheron G. The theory of regionalized variables and its applications. Cahiers duCentre de Morphologie Mathématique de Fountoinebleau; 1971 [v.05].

Oyedele DJ, Amusan AA, Obi A. The use of multiple-variable indicator kriging techniquefor assessment of the suitability of an acid soil for maize. Trop Agric 1996;73(4):259–63.

Ribeiro L, Pina P, Muge F. Contribution of indicator geostatistics and mathematical mor-phology to the characterization of aquifer heterogeneities in the vicinities of wastedisposal sites. In: Marinos PG, Koukis GC, Tsiambaos GC, Stournaras GC, editors.Eng Geol Environ 1997:2127–32. [A.A. Balkema: Rotterdam].

Smedley PL, Kinniburgh DG. A review of the source, behaviour and distribution of arsenicin natural waters. Appl Geochem 2002;17:517–68.

Smith JL, Halvorson JJ, Papendick RI. Using multiple-variable indicator kriging forevaluating soil quality. Soil Sci Soc Am J 1993;57:743–9.

Stigter TY, Ribeiro L, Carvalho DAMM. Application of a groundwater quality index as anassessment and communication tool in agro-environmental policies — two Portu-guese case studies. J Hydrol 2006;327:578–91.

Ungaro F, Ragazzi F, Cappellin R, Giandon P. Arsenic concentration in the soils of theBrenta Plain (Northern Italy): mapping the probability of exceeding contaminationthresholds. J Geochem Explor 2008;96(2–3):117–31.

using indicator kriging for the evaluation of arsenic potential contamination in an abandoned mining...

Documents