Mineralogical Magazine, October 2005, Vol. 69(5), pp. 719±736 Aqueous exposure and uptake of arsenic by riverside communities affected by mining contamination in the R|¨o Pilcomayo basin, Bolivia J. ARCHER1, K. A. HUDSON-EDWARDS1,*, D. A. PRESTON2, R. J. HOWARTH3 AND K. LINGE4 School of Earth Sciences, Birkbeck, University of London, Malet St., London WC1E 7HX, UK School of Geography, University of Leeds, Leeds LS2 9JT, UK Department of Earth Sciences, University College London, Gower St., London WC1E 6BT, UK NERC ICP Facility, School of Earth Sciences and Geography, Kingston University, Penrhyn Road, Surrey KT1 2EE, UK 1 2 3 4 ABS TR AC T The headwaters of the RõÂo Pilcomayo drain the Cerro Rico de PotosõÂ precious metal-polymetallic tin deposits of southern Bolivia. Mining of these deposits has taken place for around 500 years, leading to severe contamination of the Pilcomayo's waters and sediments for at least 200 km downstream. Communities living downstream of the mines and processing mills rely on the river water for irrigation, washing and occasionally, cooking and drinking, although most communities take their drinking water from springs located in the mountains above their village. This investigation focuses on arsenic exposure in people living in riverside communities up to 150 km downstream of the source. Sampling took place in April±May 2003 (dry season) and was repeated in January±March 2004 (wet season) in five communities: El Molino, Tasapampa, Tuero Chico, Sotomayor and Cota. Cota was the control in 2003 and again in 2004; a nearby city, Sucre, and several locations in the UK were also used as controls in 2004. Drinking, irrigation and river waters, hair and urine samples were collected in each community, digested where appropriate andÿanalysed for As using ICP-MS. Arsenicÿ1concentrations in drinking watersÿ1 ranged 0.2 ÿ112 mgÿ1l 1, irrigation water 0.6 ÿ329 m g l , river waters 0.9ÿ12,800 mg l , hair 37ÿ2110 mg kg and urine 11ÿ891 mg lÿ1. All but one drinking water sample ÿwas found to contain As below the World Health Organization recommended guideline of 10 mg l 1, although a number of irrigation and river water concentrations were above Canadian and Bolivian guidelines. Many As concentrations in the hair and urine samples from this study exceeded published values for non-occupationally exposed subjects. Analysis of mean concentration values for all media types showed that there were no statistically significant differences between the control locations and the communities exposed to known As contamination, suggesting that the source of As may not be mining-related. Arsenic concentration appears to increase as a function of age in hair samples from males and females older than 30 years. Male volunteers over the age of 35 showed increasing urine-As concentrations as a function of age, whereas the opposite was true for the females. K EY WORDS : arsenic, mining, water, hair, urine, human exposure, Bolivia, Pilcomayo. Introduction ARSENIC (As) is a human poison. Excessive uptake via the skin, lungs and gastrointestinal tract can lead to a wide range of pathologies, including * E-mail: k.hudson-edwards@geology.bbk.ac.uk DOI: 10.1180/0026461056950283 # 2005 The Mineralogical Society gastrointestinal discomfort, vomiting, convulsions, skin lesions, blackfoot disease, cancer and death (Hughes, 2002; Wright and Welbourn, 2002). According to the World Health Organization (WHO) (2001), pigmentation changes in the skin are the ®rst manifestations of chronic exposure to As. These are followed by hyperkeratosis, which in turn may be followed by J. ARCHER ET AL. cancer, although this usually takes more than 10 years to develop. Water is one of the major pathways for human exposure to As. High groundwater concentrations of As (190ÿ740 mg lÿ1) in West Bengal have been linked to high concentrations in human hair and urine, which in turn have been related to skin lesions and cancers (Chatterjee ., 1995; Das ., 1995, 1996; Samanta ., 2004). In Wisconsin, by contrast, individuals drinking water for over 20 years with much lower concentrations of As (2 mg lÿ1 or more) were more likely to report a history of depression, high blood pressure, circulatory problems and bypass surgery compared ÿto1 others drinking water containing <2 mg l of As (Zierold ., 2004). Similarly, Knobeloch and Anderson (2003) have reported that ÿlong-term ingestion of water containing >5 mg l 1 of As signi®cantly increases the risk of skin cancer. Human populations exposed to As will not necessarily react in the same manner: health effects will depend on the As bioavailability and exposure, and on nutritional status, diet, age and sex. Present-day and historic mining activities have released large quantities of As-bearing waters and sediments to ¯uvial environments (Miller, 1997; Foster ., 1998; Hudson-Edwards ., 1999). These contaminate river and ground water either directly or indirectly, through release of As from contaminated sediments, for tens to hundreds of km downstream of source areas. This contamination has the potential to severely affect plants, micro-organisms, water supply and soils, and as a consequence, the food chain. Humans are susceptible to As through drinking contaminated water or eating crops grown on contaminated soils using contaminated irrigation water. Studies of human As exposure and uptake in rivers affected by mining-related As contamination are rare. In some non-¯uvial areas affected by mining, Ascontaminated wastes, soils and waters have been linked to elevated human hair and urine concentrations (Farago and Kavanagh, 1999; Matshullat . 2000), but in others, few cases of As toxicity have been reported (Hamilton, 2000). In order to determine the impacts of miningcontaminated water on riverside communities, and ultimately, to develop management and remediation strategies, it is important to evaluate the level of exposure and uptake of As by human resident populations, and their current contaminant and water management strategies. This paper et al al et Study area et al et et al presents the results of an investigation of the aqueous pathways, exposure and uptake of As in residents living in communities along a miningcontaminated river system in Bolivia. al et al et al 720 The RõÂo Pilcomayo (20ëS 65ëW) rises at an elevation of ~5200 m in the central Andes and ¯ows in a southeasterly direction for ~670 km until reaching the semi-arid/semi-tropical Chaco Plains, where it forms Argentina's northern boundary with Paraguay. The Pilcomayo basin covers an2 area of 272,000 km2, of which 98,100 km lies in Bolivia. In the lower reaches, the Pilcomayo enters the Chaco2 Plains where it has constructed a 210,000 km , low-declivity, fan-shaped body of sediment (Iriondo, 1993; Wilkinson and Mohler, 1995). The RõÂo Pilcomayo traverses Ordovician, Silurian, Cretaceous, Tertiary and Quaternary volcanic and sedimentary rocks that are intruded by dacites, quartz porphyries and adamellites. The climate is dominated by distinct wet (November to April) and dry seasons (May to October). Average monthly rainfall is ~100 mm during peak wet season and 5 mm during dry season (PROVISA, 1989). Discharges are 3equally diverse, with average discharges of 80 m sÿ1 at 3 ÿ 1 low stage and 3600 m s in ¯ood (Wilkinson and Mohler, 1995). The chemical composition of Pilcomayo river water is variable throughout the year, and is controlled by variations in discharge during the wet and dry seasons and the weathering of evaporitic minerals (halite, gypsum), carbonates (calcite, dolomite) and pyrite from mine tailings that are ejected into the river (see below) (Smolders ., 2004). The headwaters of the RõÂo Pilcomayo in Bolivia drain precious metal-polymetallic ore deposits at PotosõÂ (Cunningham ., 1991; Fig. 1). Mining of these deposits since 1540 and ejection of mine tailings into the river has resulted in As contamination of water, sediment and soil of the basin for at least 200 km downstream of the mines (JICA, 1999; Schollaert, 2000; HudsonEdwards ., 2001; Smolders ., 2002; Miller ., 2004). The banks of the RõÂo Pilcomayo are populated by many Andean communities who rely on the river water for irrigation and washing, and on alluvial soils for growing crops. It is therefore highly likely that these communities are being directly exposed to As. Preliminary studies carried out in 2001 to et al et al et al et al et al AS UPTAKE BY RIVERSIDE COMMUNITIES, BOLIVIA FIG. 1. Location of the RõÂo Pilcomayo basin, Bolivia, showing the mining area at PotosõÂ and the villages sampled for this study (adapted from Miller ., 2004). et al investigate the risks of this exposure revealed that many drinking, irrigation and river waters were enriched in As above recommended guidelines, and thus, that water may be a pathway of human As exposure, but no samples were taken to evaluate the level of human As uptake at this time (Miller ., 2004). et al Methods and materials Sampling took place during both dry (AprilÿMay 2003) and wet seasons (JanuaryÿMarch 2004) in the uppermost 150 km of the RõÂo Pilcomayo that was shown previously to be the most highly contaminated of the whole river system (HudsonEdwards ., 2001; Miller ., 2004). Sample collection took place in ®ve villages: El Molino, et al et al 721 Tasapampa, Tuero Chico, Sotomayor and Cota (Fig. 1). Four of these are located on the banks of the RõÂo Pilcomayo and one, Cota, is a control located adjacent to a tributary known not to ¯ow through any mining districts or to be affected by mining-related dust. During the wet season, additional samples were taken at Sucre, a city located ~40 miles north of Tuero Chico, and three locations in the UK (Aberdeen, London and Surrey) to provide extra controls. A research ethics proposal was submitted and approved by the Research Ethics Committee of Birkbeck College, University of London, prior to sample collection. Samples of drinking, irrigation and river water, hair and urine were collected in all the villages (Table 1). Drinking waters were taken from taps J. ARCHER ET AL. )13.0( )92( 84 ,01 )71( 16 ,01 91.1 ,4 )132( )70.0( )8( 22 ,6 38 1 , 6 03.0 ,8 )60.0( )20.0( )40.0( )4( 82 ,4 )6( ,111 ,4 09.0 ,2 56.0 ,2 87.0 ,5 )991( )38.0( )5( 93 ,5 )89( 971 ,5 124 ,2 24.1 ,6 )235( )44.9( )31.0( )41( 92 ,11 716 ,11 991 ,2 20.1 ,4 )1.91( )92.0( )21( 33 ,4 )44( 011 ,4 311 ,5 39.0 ,3 )064( )932( )46.61( )55.0( )02( 04,5 225 ,5 0472 ,2 1.81 ,3 00.2 ,8 retaw retaw enirU riaH retaw reviR noitagirrI gniknirD ÐÐÐÐÐÐÐÐÐÐ 4002 ÐÐÐÐÐÐÐÐÐÐ )lortnoc( KU )lortnoc( ercuS )lortnoc( atoC royamotoS ocihC oreuT apmapasaT oniloM lE .)noitaived dradnats( ,naem ,selpmas fo .oN :yeK )511( )01( 63 ,5 161 ,5 )3( 4 ,2 )0.1( 8.1 ,3 )271( )971( )971( )51( 84 ,7 203 ,7 3121 ,1 001 ,1 )9.1( 1.3 ,3 )611( )971( )6( 32 ,11 352 ,11 451 ,1 )98( 372 ,3 )4.1( 5.1 ,2 )283( )31( 33 ,6 )46( 911 ,6 272 ,2 )31(11 ,2 )1.1( 3.1 ,2 )311( )971( )0.55( )72( 46 ,5 563 ,5 008,21 ,1 )11( 7 ,3 1.92 ,4 retaw retaw enirU riaH retaw reviR noitagirrI gniknirD ÐÐÐÐÐÐÐÐÐÐ 3002 ÐÐÐÐÐÐÐÐÐÐ .)riah rof 1ÿgk gm ,eniru dna sretaw rof 1ÿl gm( stluser noitartnecnoc sA .1 ELBAT 722 AS UPTAKE BY RIVERSIDE COMMUNITIES, BOLIVIA and storage tanks. The taps chosen for sampling were all located in public places (e.g. medical centres, schools and/or the main square of each village) and storage tanks were the principal tanks used by each village for storage of all potable water. Taps were allowed to run freely for 3 min prior to sample collection. Irrigation waters were collected from canals serving cultivated land, and river water was collected from the banks of the RõÂo Pilcomayo, adjacent to each village. Water samples collected in 2004 were taken from the same taps, tanks and locations as those sampled during the 2003 ®eldtrip. However, collection of irrigation water samples was problematic during both ®eldtrips, but particularly during the wet season, as many villagers relied on rainfall to irrigate their crops. As a result, only two samples from El Molino and two from Cota were collected in 2004. UK water samples comprised potable tap water only and were treated in an identical fashion to the Bolivian water samples. All water samples were 30 ml in volume, and were collected un®ltered so as to most closely represent those consumed and used by inhabitants. They were acidi®ed with 0.6 ml of concentrated nitric acid and stored in acid-washed HDPE bottles below 4ëC. Field blanks were opened and acidi®ed in the same manner as samples. Field duplicates and blanks form 14% and 13%, respectively, of the total number of sample sets collected. Hair and urine samples were collected within two to three days of the water samples from between 4 and 11 individuals in each village (see Table 2 for volunteer details). Bolivian volunteer selection was conducted on a ®rst come-®rst served basis, although only volunteers who were non-smokers, apparently healthy and permanent residents of the villages were used in the study. Children below the age of 11 and adults over the age of 56 were not included. Volunteers of both sexes were included, though females were often more willing volunteers; thus for some villages the number of female volunteers sometimes substantially outweighs male (Table 2). This is particularly true for Tuero Chico. Volunteers used the same drinking water sources and ate crops and livestock grown and reared in the same area. Where possible the sample volunteers were the same for both ®eldtrips, although in 2004 some of the original volunteers were unavailable. Volunteers from the UK were selected from non-smoking, healthy individuals who used the same water source for drinking, cooking and washing. They were requested to refrain from drinking alcoholic beverages and consuming ®sh or sea vegetables for one week prior to sample collection. Hair from the occipital region of the head was tied with cotton before being cut with stainless steel scissors. Samples were then placed into zip-lock bags and stored in a cool, dark container. Grey hairs were discarded from the samples. Volunteers providing hair samples were also asked to provide a 30 ml urine sample. The urine samples comprised the ®rst void of the day and were acidi®ed and stored in acid-washed HDPE bottles below 4ëC until digestion/analysis ~3 months later. Creatinine (C4H7N3O) measurements were taken prior to digestion or analysis in order to standardize all urine samples (Matschullat ., 2000). et al TABLE 2. Hair and urine volunteer information. 2003 No. of samples Mean age (years) Range No. of males No. of females 2004 No. of samples Mean age (years) Range No. of males No. of females El Molino Tasapampa Tuero Chico Sotomayor 5 33.2 12ÿ44 3 2 6 26.7 16ÿ42 3 3 11 39.5 21ÿ55 1 10 7 23.4 16ÿ34 2 5 5 29.4 11ÿ50 1 4 5 34.0 13ÿ45 3 2 4 25.5 17ÿ33 2 2 11 40.1 22ÿ56 1 10 5 25.8 19ÿ35 2 3 4 34.8 16ÿ51 1 3 723 Cota Sucre 6 33.0 24ÿ45 4 2 UK 10 28.9 25ÿ34 7 4 J. ARCHER ET AL. Questionnaires requesting information about lifestyle, occupation, diet, health and water management techniques were also undertaken in each village and in the UK. Health status information for the community was obtained during interviews with medical personnel in each village. In Tasapampa, Tuero Chico and Cota, the health workers interviewed were residential nurses working alone, whereas in El Molino and Sotomayor the interviews were carried out with a residential doctor and nurse. A total of 23 families was questioned regarding water use. These families included the hair and urine volunteers. Sample volunteers were also questioned as to what they had consumed during the two days prior to urine sample collection. In the laboratory, hair samples were washed according to the method described in IAEA (1977) and 50 mg of the youngest 3ÿ4 cm of hair separated for digestion. Hair and un®ltered waters collected during the 2003 dry season visit were all digested in open acid-washed glass boiling tubes. Hair samples were digested using a reagent mixture of concentrated HNO3 and H2O2, whereas the un®ltered waters were digested using only concentrated HNO3. Urine samples were ®ltered. Hair, urine and un®ltered waters collected during the 2004 wet season were digested using a newly acquired microwave digestion instrument. The reagents used to digest the hair and un®ltered waters were the same as above, but smaller volumes were used. The urine was digested using the same reagents as those used for the hair. The digests were ®ltered and made up to 25 ml volume with deionized ultrapure water. All samples were analysed for As using quadrupole ICP-MS (a Plasmaquad PQ2+ in 2003 and an Agilent 7500c in 2004), with external calibration using As solutions diluted from a 10 mg lÿ1 stock standard (SPEX CertiPrep, USA). Suppression tests were carried out before the analysis to determine the correct dilution factor required for each sample media. To determine the accuracy of the digestion and analysis, reference materials were included in every analysis batch. The reference materials used were TMDA-52.2 (Lake Ontario water), Seronorm201205 (urine) and GBW09101 human hair (LGC Promochem, Teddington, UK). Table 3 contains the measured As concentration in reference materials, blanks and duplicates. This table also shows average detection limits for each sampling medium for each year. Detection limits improved by approximately an order of magnitude in 2004 compared to 2003, due to improved sample preparation procedures and the use of the more sensitive Agilent 7500c ICP-MS. Statistical methods Mean analyte concentration and standard deviation values for all groups were calculated using Cohen's (1959, 1961) technique to correct for the presence of any determinations below analytical method detection limits. Pairwise and multiple comparisons between means for the various groups (e.g. differences between towns or seasons) were made using Analysis of Means (ANOM; Ott, 1967; Nelson, 1974, 1983, 1993). This is a very effective graphical technique. Group means and standard deviations are ®rst calculated for each of the groups to be compared. From these, the grand mean and pooled standard deviation over all the groups are then obtained. Upper and lower decision limits are then calculated for each group, taking into account group sizes and the statistical signi®cance level of the required result. If one or more group means fall outside the upper or lower decision limits (e.g. Fig. 2), then the mean(s) for that group(s) is (are) statistically signi®cantly different from the rest at the chosen level of signi®cance (`signi®cantly' FIG. 2. ANOM spatial analysis for log-transformed As concentrations in 2004 river water. Thin black bar, group mean; thick dark grey bar, lower 95% decision limit; thick light grey bar, upper 95% decision limit (95% decision limits are bounds such that if one or more group means fall beyond the limits there is a 95% probability that it is different from the groups, that are not statistically different from each other, which fall between the lower and upper bounds). 724 AS UPTAKE BY RIVERSIDE COMMUNITIES, BOLIVIA TABLE 3. ( ) Reference, ( ) blank and ( ) duplicate As data and ( ) detection limit values for all sampling media for 2003 and 2004. All As concentrations are in mg lÿ1 for waters and urine and mg kgÿ1 for hair. a b 2003 a mean % rsd certified value tolerance limits range % bias c Waters d 2004 2003 30.7 25.4 18.9% 1.0% 25.5 25.5 25.6ÿ35.8 22.3ÿ28.4 20.4 ÿ0.6 Hair 2004 2003 549 20.8% 590 0ÿ1990 520ÿ660 ÿ7.0 615 2.2% 590 451ÿ778 520ÿ660 4.2 164 4.6% 184 68ÿ260 ÿ10.7 2004 0.075 0.005 0.077 0.005 0.072 0.005 2003 0.001 0.007 0.017 0.008 0.014 0.008 Urine 2004 180 0.8% 184 175ÿ185 ÿ2.3 b Sample blank 1 sd blank 2 sd blank 3 sd Waters 2003 2004 0.032 0.096 0.016 0.002 0.022 0.087 0.045 0.003 0.030 0.081 0.060 0.006 2003 0.015 0.006 0.011 0.003 0.008 0.004 Hair Urine 2004 0.064 0.004 0.064 0.006 0.068 0.003 c Sample no. 3D/I 3D/I/D % bias 4E/R 4E/R/D % bias 2003 320 324 ÿ1.1 543 579 ÿ6.4 % rsd 1.2 2.6 2.3 0.9 Sample no. 2004 6B/D 0.231 6B/D/D 0.335 % bias ÿ31 6D/D 0.323 6D/D/D 0.316 % bias 2.2 % rsd 19.1 3.9 10.4 16.1 d Media Waters Hair Urine 2003 3.18 141 3.00 2004 0.111 22.3 0.698 implies `statistically signi®cant with 95% con®dence' hereafter). If they all lie within the decision limits, then there are no statistically signi®cant differences between the groups. Robust regression, which down-weights outliers, was used when ®tting standard linear regression models (Rousseeuw and Leroy, 2003). 725 Mahon's (1996) revision of the York (1967) errors-in-variables regression method was used where it was necessary to incorporate errors in both and when ®tting a functional regression model to data drawn from two different populations (As concentrations in hair or urine as a function of As concentrations in water). Nonx y J. ARCHER ET AL. linear trends were identi®ed using Cleveland's (1979) LOESS technique. This applies a locallyweighted smooth to the data: a window, centred on each successive coordinate in turn, is passed across the data and a locally-weighted regression is used to predict the value within the window. The ensemble of these points produces the desired smooth curve. This and the 95% con®dence limits were obtained using (Insightful Corporation). xy x y S-Plus Verif|cation of in-house reference materials It can be assumed that in the case of replicate analysis of homogeneous reference materials, analytical error conforms to the normal distribution (Thompson and Howarth, 1980). With regard to Table 3, statistical tolerance limits (Hahn and Meeker, 1991) can be calculated such that one can be 95% certain that 99% of all the possible values of individual observations drawn from the sampled distributions will fall between their lower and upper bounds. These limits (Table 3) all enclose the certi®ed values for As (although the 2003 water at 25.5 mÿg1lÿ1 is just below the lower limit of 25.6 mg l ). This difference is small and has been deemed reasonable for the purposes of this investigation. If the argument is based on a con®dence interval for the mean of the observed distribution, a similar conclusion is reached. Arsenic concentrations found in the blanks are all low, and precisions are also acceptable for the purposes of this study. Results Lifestyle, health and water management Most of the inhabitants of the study area depend on farming and livestock for their livelihood. The methods used to work the land are largely traditional, with most agricultural tasks performed by hand using simple tools and livestock for ploughing and some transport. Vegetables grown in the ®elds are eaten by the farmers and their extended families or sold to other inhabitants of the community and in nearby cities (Sucre and PotosõÂ). The communities are relatively poor and their mostly vegetarian diet re¯ects this: meat consumption is generally reserved for special occasions. Dairy products are seldom consumed and most protein intake comes from eggs and pulses. Fish is not consumed. Consumption of potable water increases slightly from an average of two litres per day during the dry season to 2.25 726 litres per day during the wet season when the annual temperatures are at their highest. Information on community health, gleaned from medical staff, revealed that the commonest complaints are diarrhoea, nausea and vomiting in all ®ve villages. Health workers also reported stunted growth in children from Sotomayor and Tuero Chico, and headaches and tiredness in Tasapampa and Tuero Chico. Amongst these reports there were no obvious symptoms that could be expressly related to arsenic toxicity alone, i.e. no mention of skin thickening, ulceration or changes in pigmentation. The long dry season means that water management is very important in the communities. Each village has an elected water committee that manages water distribution and the maintenance of the system. Drinking water is taken from what are believed to be uncontaminated springs or streams located in the mountains surrounding each of the ®ve villages. This water is transferred by plastic pipe to the main concrete supply tank above each village. All villages rely on a single spring, except Sotomayor, which has access to three. The tanks are covered, and regional government agencies sometimes provide chlorine and disinfectants to be mixed with water prior to consumption, although this is rarely done. Most households have access to a private tap supplied by water from the main tank. Those without a private tap collect water in plastic containers from communal taps located in either local schools, health centres or the village main square. Some families collect rainwater to provide potable water when, on occasion, none is available from the taps during the dry season. No family reported taking water from the RõÂo Pilcomayo for cooking or drinking. Irrigation water is collected from either an uncontaminated tributary or a spring, except in Tuero Chico and Sotomayor, where water from the RõÂo Pilcomayo is used during the dry season. During the wet season the communities frequently rely on rainfall alone to irrigate their ®elds. Almost all families questioned about water quality stated that drinking water is often very turbid during the wet season and therefore regarded as `cleaner' during the dry season. Families in Tasapampa, Tuero Chico, Sotomayor and Cota regard the RõÂo Pilcomayo and its tributaries as `dirtiest' during the wet season, but those in El Molino (closest to the mines) believed the opposite, claiming that the river smells bad and the water turns to `sludge' during the dry season. Community members in AS UPTAKE BY RIVERSIDE COMMUNITIES, BOLIVIA was determined in water from one tap. This very high As concentrations was found in a 2003 water sample from a tap which, when sampled again in 2004, was foundÿ1 to be much lower in concentration (2.21 mg l ).ÿ1A Bolivian agricultural water standard (50 mg l ; Bolivian agricultural waters standard, 1995) andÿ1 a Canadian irrigation water standard (100 mg l ; Peterson, 1999) were used for the irrigation and river water assessment. Almost all river waters displayed As concentrations in excess of these standards; the exception is Cota, the control village, where concentrations were less than all other village concentrations. Irrigation waters collected in 2003 show that As concentrations for Tuero Chico (average 273 mg lÿ1) are double that of the Canadian standard, and that the As concentration for Sotomayor match the standards. Arsenic concentration in irrigation waters collected in 2004 are all well below the Bolivian and Canadian Tasapampa, Tuero Chico and Sotomayor reported ¯ooding to be an occasional problem (once in ®ve years on average), affecting a small number of ®elds closest to the river. Overall, these ¯oods are regarded as positive phenomena by the communities as they supply fertile silt to the ®elds. Many families surveyed did not perceive the use of contaminated river water for irrigation as a threat to health. Water analysis The WHO recommended guideline for ÿAs1 concentrations in drinking water (10 mg l ; WHO, 1996, 1998) was used to assess the quality of the drinking waters sampled (Fig. 3). Arsenic concentrations in the drinking waters of the communities in this study all fall below this concentration, with the exception of El Molino, where a very high As concentration (112 mg lÿ1) FIG. 3. As concentrations in: ( ) drinking water; ( ) irrigation water; and ( ) river water samples. a b 727 c J. ARCHER ET AL. standards (Fig. 3). Overall, the river waters contain the greatest concentrations of As followed by the irrigation waters and the drinking water samples. Analysis of means (ANOM) was used to assess the magnitudes of apparent differences between As concentrations in the different groups of water data. Because there was evidence that, in general, analyte concentrations were lognormally distributed and that group standard deviation tended to increase as a function of group mean, all data were log-transformed prior to ANOM statistical analysis. The ANOM results for 2004 river water (Fig. 2) illustrate the use of this test. For the purposes of display, the back-transformed logarithmic group means and 95% decision limits are shown plotted on a logarithmic-scaled ordinate. ANOM was used to compare the group means for the water samples for all locations. As stated above, only group mean concentration values which fall beyond their corresponding upper or lower decision limits can be regarded (with 95% con®dence) as having a statistically signi®cant different mean concentration compared to those for other groups which lie within the ANOM limits. There are no signi®cant differences between the mean As village concentrations in the 2003 river waters. River waters collected in 2004 show Cota as possessing mean As concentrations that are signi®cantly low compared to those from the other four sample villages (Fig. 2). Conversely, the 2004 river water samples collected in El Molino and Sotomayor contain mean As concentrations that are signi®cantly high compared to the other three villages. Because of the small numbers of samples taken from each location, the standard deviations are relatively large, which in turn results in a wide decision limit range. Despite this, there is a similarity between the trends seen in the 2003 and the 2004 river water data, with the average As concentrations decreasing in the order: El Molino > Sotomayor > Tuero Chico > Tasapampa > Cota. In the drinking waters the average As concentration for El Molino is signi®cantly higher than the other villages in both 2003 and 2004. The highest to lowest trend in average As concentrations for the other villages is similar to that seen in the river waters. Interestingly, Sucre is the only one of the three controls to show an average drinking water-As concentration that is signi®cantly lower than the sample villages. There are no signi®cant differences between Cota, the UK and the remaining villages. Mean As concentrations in irrigation water samples show a different pattern to that seen in the river and drinking waters. In the 2003 data, only Tuero Chico shows As concentrations that are signi®cantly higher than the others and El Molino shows As concentrations that are signi®cantly lower. El Molino and Cota were the only villages from which irrigation waters were collected in 2004. ANOM on the data shows that El Molino has a signi®cantly higher mean As concentration in irrigation water than Cota. ANOM was also used for comparisons between As concentration means in the waters for 2003 and 2004. There were no signi®cant differences between mean As concentration for irrigation and river for the dry (2003) and wet (2004) seasons. The mean As drinking water concentration is signi®cantly higher in the dry season than the wet season. There appears to be a pattern, repeated consistently in many of the individual villages, in which dry season concentrations are higher than wet season concentrations, although not necessarily with a statistically-signi®cant difference in magnitude (recall that the ANOM decision limits tend to be relatively broad, mainly as a result of the small number of observations in each group). Hair and urine 728 Hair grows at an average rate of approximately 1.1ÿ1.5 cm per month (Akagi ., 1995; Boischio and Cerichiari, 1998). Because the hair lengths sampled for this study were between 3 and 4 cm, the concentrations in Fig. 4 represent As expelled via the body from up to four months previously. The 2003 samples therefore re¯ect As exposure during the wet season, and the 2004 samples, the middle to end ofIIIthe dry season/early wet season. The 1/2 for As and its metabolites (including AsV) to leave the body via the kidneys as urine are reported by Crecelius (1977) as 10 and 30 h, respectively. The urine-As concentrations presented herein therefore represent the expulsion of As consumed during the preceding 24 h. Arsenic concentration in 2003 urine samples represent dry season As exposure and urine 2004 data represent wet season exposure. To determine if the As concentrations in hair and urine collected during this investigation are relatively high or low compared with the rest of the world, they have been compared to `reference concentrations' derived from studies of nonoccupationally exposed reference populations in Sweden, the UK, Japan and Italy (Minoia ., et al T et al AS UPTAKE BY RIVERSIDE COMMUNITIES, BOLIVIA FIG. 4. As concentrations in all human hair and urine samples. Reference lines represent published reference concentrations derived from studies of non-occupationally exposed populations: ( ) urine; ( ) hair. a 1990; White and Sabbioni, 1998; Rodushkin and Axelsson, 2000; Morton ., 2002; Sera ., 2002). Figure 4 shows the As concentration in each sample for all 2003 and 2004 data, together with the reference concentrations. Many samples from each village lie above at least one of the reference concentrations. Villages where As concentrations exceed all the reference values are found in the 2003 hair data for El Molino, Tuero Chico, Sotomayor and Cota; in the 2004 hair data for El Molino, Tuero Chico, Sotomayor and Sucre; in the 2003 urine data for El Molino, Sotomayor and Cota; and in the 2004 urine data for El Molino, Tasapampa, Tuero Chico, Sotomayor, Sucre and the UK. The villages of Tuero Chico, Sotomayor and particularly, El Molino, show consistently high As concentrations compared to the reference values in most, if not all, of the hair and urine data. In the 2003 hair data, two As concentrations in El Molino and one in Sotomayor exceed 400 mg kgÿ1, which is almost double the highest reference value of 230 mg kgÿ1 (Sera ., 2002). Arsenic concentrations in some of the 2004 hair samples show the greatest differences between Bolivian concentrations and the other studies. Individual As concentrations in the 2004 hair samples from El Molino and Tuero Chico reached concentrations of 1300 and 2110 mg kgÿ1, respectively, which is considerably higher than the aforementioned highest published reference value (230 mg kgÿ1; Sera ., 2002). In the 2003 urine data, two Asÿ1 concentrations from El Molino (89 and 95 mg l ) exceed the highest reference concentration of 31.1 mg lÿ1 (Minoia ., 1990) by approximately three times. ANOM was also applied to log-transformed hair- and urine-As concentrations. No signi®cant et al et et al al et al et al 729 b differences are exhibited between mean As concentrations in 2003 hair samples from each of the ®ve villages, including the control village Cota. In the 2004 hair data, only El Molino and Tuero Chico are signi®cantly higher. Mean As concentrations in 2004 hair from two of the three controls, Cota and Sucre, are not signi®cantly different to the other villages. However, the third control, the UK, has a mean As concentration that is signi®cantly lower than all the other villages. Interestingly, the downstream pattern seen in the hair samples for 2003 is similar to the trend exhibited by both the river and drinking water data, although the 2004 hair data do not follow this trend, as the Tuero Chico mean ÿAs concentration is very high (617 mg kg 1). Arsenic concentrations in the 2003 urine samples from El Molino and Sotomayor are signi®cantly higher than those from Tasapampa, Tuero Chico and Cota. The mean As concentration in 2003 urine from Tuero Chico is signi®cantly lower than the other villages. Arsenic concentrations for the UK and the other two controls, Sucre and Cota, show no signi®cant difference from the other villages in the 2004 data. The downstream pattern of As concentration in the 2004 urine data is repeated in the 2003 urine data, but this differs from that seen in the previous ANOM data for hair and waters. There are no signi®cant differences between the pooled mean urine-As concentrations for all villages, in either 2003 or 2004. The 2003 urineAs concentration does appear to be higher than the 2004 concentration and this pattern is repeated in most of temporal pairs for individual villages. The same mean As concentrations in hair shows that As concentrations are signi®cantly lower in the 2003 samples. There is no correlation between J. ARCHER ET AL. hair and urine data from 2003 and 2004; nor is there between 2003 and 2004 hair; nor 2003 and 2004 urine. Errors-in-variables functional regression between log-transformed As concentrations in the hair and urine and those in the drinking waters (Fig. 5) shows increasing trends in all cases, although only that for the 2003 urine is statistically signi®cant ( 2 = 0.87 at 90% con®dence). The UK has been omitted from this analysis since it is the As pathways operating in the Bolivian communities which are under investigation here. Robust linear regression analysis of As concentration in the hair and urine as a function of age did not show any statistically signi®cant trends. However, LOESS suggests the presence of non-linear trends in the data (Fig. 6). In the hair data, females show an increase in As concentration with age and males show a small initial r decrease followed by an increase in As concentration in males >26. Male and female As concentrations in urine remain relatively static up to the age of around 35 after which concentrations in the males appear to increase, but decrease in females. We believe that, while relatively low-magnitude, these are real phenomena since the method is non-parametric and the independent ®ts to male and female data indicate correlated and inverse patterns of change in the older population in the case of hair and urine-As concentrations. When As concentration as a function of sex is analysed using ANOM, there appears to be no signi®cant difference between males and females in the 2003 data. In the 2004 data, however, the hair-As concentrations are signi®cantly higher for females compared to males. By contrast, the 2004 urine-As concentrations are signi®cantly lower for females compared to males. FIG. 5. As concentration in human hair and urine as a function of its concentration in drinking water. Data points represent mean values with and error bars representing 1 standard deviation. All data have been transformed to logs to base 10: ( ) hair 2003; ( ) hair 2004; ( ) urine 2003; ( urine 2004. x y a b c 730 d) AS UPTAKE BY RIVERSIDE COMMUNITIES, BOLIVIA excessive and likely to lead to poisoning (Arnold ., 1990). Hair samples taken in 2004 from El Molino (1300ÿ1 mg kgÿ1) and Tuero Chico (907 and 2110 mg kg ) have concentrations approaching, or in excess of, the 1000 mg kgÿ1 criterion. Similarly, in Utah, USA, some inhabitants showing hyperkeratosis exhibited highest average urine-As concentrations of 0.175 to 0.211 mg lÿ1, and highest average hair-As concentrations of 1.09 Discussion et al Although the Pilcomayo village hair- and urine-As concentrations exceed those in non-occupationally exposed populations, it is also important to verify if they approach concentrations in persons exhibiting signs of As toxicity. Arsenic concentrations in hair that are ÿregarded as `normal' are 80 to 250 mg kg 1, with 1000 mg kgÿ1 considered FIG. 6. As concentration in human hair and urine as a function of age. LOESS regression lines and the respective 95% con®dence limits are shown: ( ) female hair; ( ) male hair; ( ) female urine; ( ) male urine. a b 731 c d J. ARCHER ET AL. to 1.21 mg kgÿ1 (Southwick ., 1983). Urine As concentrations in this study all lie below this range. The As range in the American hair samples is exceeded by one individual in El Molino and another in Tuero Chico. Average As concentrations in biological media from individuals living in West Bengal, some of whom have As-related skin lesions, are 3.43ÿ1mg kgÿ1 (Samanta .,ÿ2004) and 6.75 mg kg (rangeÿ11.18ÿ31.0 mg kg 1) ÿfor1 hair and 0.43 mg l (0.03 ÿ2.00 mg l ÿ1; Chatterjee ., 1995) and 259.5 ng mg creatinine (range 20.5ÿ2890 ng mgÿ1 of creatinine; Tokunaga ., 2005) for urine. In total, ~10% of the individuals in the current investigation showed hair or urine concentrations in excess of published averages from studies where people are known to show symptoms of As poisoning. A lack of correlation between the hair and urine data can be explained by the different time periods for As expulsion that these media represent (Crecelius, 1977; Akagi ., 1995; Boischio and Cerichiari, 1998). The lack of correlation for the 2003 and 2004 hair and urine data may be due to different amounts of As uptake in the wet and dry seasons. Similar seasonal effects on urinary inorganic As concentrations were reported by Hinwood . (2004) for areas affected by gold mining in Australia. The broad spread of results for hair and urine concentrations is not unexpected when analysing samples from human volunteers who vary considerably in such characteristics as age, sex and health status. Differences in As uptake based on age and sex and the associated differences in lifestyle and occupation could be considered as confounding factors in this study, but strenuous attempts have been made to limit these. For example, the age range of 11 to 56 years old is broad, but most of the male (79%) and female (81%) volunteers for this study were involved in the same farming activities using the same methods. The crops grown and consumed, and their preparation and cooking methods, are also virtually identical amongst community members. Some of the female members are housewives caring for babies and small children, but they usually return to farming once the children begin to attend school. Because the Bolivian communities are relatively remote and poor, they do not have access to a wide range of different food and beverage products, so members of these communities eat and drink the same limited number of products. Volunteers in this study were chosen based on their non-smoking habits so as to et al et al et al et al et al et al 732 eliminate this potentially confounding factor from the study. Average weekly alcohol consumption was not recorded for individuals, but apart from special occasions, moderate amounts of alcohol only are consumed in the communities. Some community members in Tasapampa, Tuero Chico and Sotomayor described crossing the RõÂo Pilcomayo for various purposes and some of the teenage volunteers also described swimming in the river during very hot weather, but accurate measurement of these activities was not possible. In a relatively small sample population such as this, these confounding factors can bias the results and so any conclusions should be treated with some caution. It is unclear whether drinking water is a pathway for human As uptake in the Pilcomayo study villages. The consistent mutually increasing relationship between drinking water-As and hairand urine-As concentrations (Fig. 5) suggests that it may be so, but the drinking water-impacted As hair, and to a lesser extent, urine concentrations, are likely to be older than the actual drinking water As concentrations used for this plot. El Molino is the only village to have As concentrations in drinking water above WHO recommended guidelines, with all the other study villages showing concentrations well below such guidelines. Whether this will lead to health effects is uncertain. The recorded health problems of stunted growth, widespread anaemia, headaches and tiredness have been attributed by local health workers to Pb poisoning, sunstroke and the relatively common chagas disease, rather than As accumulation. Because As poisoning can take up to 10 years to be expressed, some of these health problems could still be related to As accumulation. Some believe that drinking water with As concentrations much lower than the WHO and US EPA recommended guideline (10 mg lÿ1) can have an adverse effect on human health. A study in Wisconsin, USA, found that individuals drinking water for over 20 years containing 2 mg lÿ1 of As or greater were more likely to report a history of depression, high blood pressure, circulatory problems and bypass surgery compared to others drinking water containing <2 mg lÿ1 of As (Zierold ., 2004). Similarly, Knobeloch and Anderson (2003) reported that long-term ingestion of water containing >5 mg lÿ1 of As signi®cantly increases the risk of skin cancer. The Pilcomayo communities have, however, been using the same sources et al AS UPTAKE BY RIVERSIDE COMMUNITIES, BOLIVIA of drinking water for decades, suggesting that if As toxicity were present, it would have revealed itself before now (although it is possible that the medical staff in the area may not be fully trained in the recognition of As toxicity symptoms). The source of As in the drinking water in the Pilcomayo Basin is probably not directly miningrelated, given that communities use tributary waters unaffected by mining activity, and that the drinking water-As concentrations in most of the study villages are not statistically different from those of the control village Cota. The only exception to this is El Molino, where one of the four tap waters sampled in 2003 contained an anomalously high As concentration, despite the fact that another of the other taps receiving water from the same tank, and sampled on the same day, showed a far lower concentration. In 2004 the water sampled from the tap which had previously shown a high As concentration containedÿ1 far less As (below the recommended 10 mg l guideline). The same 2003 tap water was analysed ®ltered (0.2 mm) and found toÿ1 contain far lower As concentrations (2.09 mg l ), suggesting that the As was associated with particulate matter. Further sampling from this tap should allow con®rmation of whether this tap water is highly contaminated or had simply discharged an anomalous particulate, in¯ating the As concentration. Smolders . (2004) showed that there is a geologically signi®cant chemical weathering control on the major element chemistry of the RõÂo Pilcomayo, suggesting that it is possible that the As is derived from natural weathering of bedrock in the catchment. In the irrigation waters, Tuero Chico and Sotomayor have average As concentrations that are well above guidelines. This is due to the fact that both villages use Pilcomayo waters to irrigate their crops during the dry season, whereas the other villages use water from uncontaminated tributaries. Villagers claim that they do not drink from the irrigation canals, but they do eat vegetables grown in the irrigated ®elds. Consequently, vegetables grown in Tuero Chico and Sotomayor may represent another exposure pathway by which As reaches the human body. This may explain why the results from the ANOM tests show that Tuero Chico 2004 hair-As concentrations deviate from the trend noted in the drinking and river water and 2003 hair data. The increase of hair-As concentration as a function of age (in individuals over 30; Fig. 6) suggests that either adults of >30 y are exposed to and/or consume greater quantities of As, or that As may be accumulating in the bodies of those who are >30 y old, resulting in greater excretion via hair. This trend is repeated in the male urine data. The counterpart downward trend seen in the female urine data can not be explained at present. Conclusions The lack of statistically signi®cant differences between the controls and the other communities suggest that the mining-contaminated RõÂo Pilcomayo should not be held solely responsible for As contamination in this area of Southern Bolivia. Natural chemical weathering of bedrock may play a signi®cant role in exposing local communities to As. The results of the study have not clearly shown that drinking water is an As exposure pathway. Even if it is, the fact that all but one drinking water sample were found to contain As below the WHO recommended guideline of 10 mg lÿ1 suggests that there are other pathways through which As enters the human body in the Pilcomayo communities, including ingestion of vegetables grown using contaminated irrigation water, ingestion of meat from livestock reared drinking contaminated water, and inhalation of contaminated dust, as have been shown for other areas (e.g. Lee and Chon, 2003). Analysis of these media was outside of the scope of this study, but should certainly be attempted in the future. Around 10% of the sample population possessed hair- or urine-As concentrations equal to concentrations from other studies where symptoms of As toxicity have been observed, but the long-term effects of this apparent exposure are currently unknown. The human populations sampled in this study were relatively small compared to the total population in the river catchment area; therefore any future investigations into As exposure in the riverside communities should include a larger sampling program encompassing a greater proportion of the population. et al Acknowledgements 733 This work was funded by NERC through a studentship to J. Archer (number NER/S/A/ 2002/11010) and ICP support (OSS/253/0204), and through a travel grant awarded by the Society for Latin American Studies. Special thanks go to Teresa Bohorquez, Rosmary Herobas, David Gomez, Evelyn Rossa and Mary Rossa for J. ARCHER ET AL. Chakraborti, D. (1995) Arsenic in groundwater in six assisting with health surveys and arranging districts of West Bengal, India: the biggest arsenic interviews with community members, and to calamity in the world. Part 2: arsenic concentration Dario Urquizo, Mamerto Ortiz, Juan Vedia in drinking water, hair, nails, urine, skin-scale and Partes, Luis Salazar, Riberto Vilasques and liver tissue (biopsy) of the affected people. , Pastor Quispe for giving permission to work in 120, 917ÿ924. their communities. The authors are grateful to D., Samanta, G., Mandal, B.K., Chowdhury, T.R., Carlos Aliaga Arriola and Elizabeth Romero for Das,Chanda, C.R., Chowdhury, P.P., Basu, G.K. and advice and help with water sampling in Sucre. We Chakraborti, D. (1996) Arsenic in groundwater in six thank A. Osborn and S. Houghton for analytical districts of West Bengal, India. assistance using Wolfson Laboratory for 5ÿ16. Environmental Geochemistry facilities (UCL- Farago, M.E. and Kavanagh,, 18P., (1999) High arsenicBirkbeck, University of London), and F. Barry containing soils in SW England and human exposure at the NERC ICP-MS Facility at Kingston assessment. Pp. 181 ÿ 184: University. (H. Armannsson, editor). Balkema, Analyst Environmental Geochemistry and Health Geochemistry of the Earth's Surface References Akagi, H., Malm, O., Kinjo, Y., Harada, M., Branches, F.J.P., Pfeiffer, W.C. and Kato, H. (1995) Methylmercury pollution in the Amazon, Brazil. , 175, 85ÿ95. Arnold, H.L., Odam, R.B. and James, W.D. (1990)th , 8 edition. Saunders W.B. Company, Philadelphia, USA. Boischio, A.A.P and Cernichiari, E. (1998) Longitudinal hair mercury concentration in riverside mothers along the Upper Madeira River (Brazil). , 77, 79ÿ83. Bolivian Agricultural Waters Standard (1995) Water Polluting Materials. , Regulation no. 24176, Bolivia. Chatterjee, A., Das, D., Mandal, B.K., Chowdhury, T.R., Samanta, G. and Chakraborti, D. (1995) Arsenic in groundwater in six districts of West Bengal, India: the biggest arsenic calamity in the world. Part 1: arsenic species in drinking water and urine of affected people. , 120, 643ÿ650. Cleveland, W.S. (1979) Robust locally weighted regression and smoothing scatterplots. , 74, 829ÿ836. Cohen, A.C. (1959) Simpli®ed estimators for the normal distribution when samples are singly censored or truncated. , 1, 217ÿ237. Cohen, A.C. (1961) Tables for maximum likelihood estimates: singly truncated and singly censored samples. , 3, 535ÿ541. Crecelius, E.A. (1977) Changes in the chemical speci®cation of arsenic following ingestion by man. , 19, 147ÿ150. Cunningham, C.C., McNamee, J., Vasquez, J.P. and Ericksen, G.E. (1991) A model of volcanic domehosted precious metal deposits in Bolivia. , 86, 415ÿ421. Das, D., Chatterjee, A., Mandal, B.K., Samanta, G. and Science of the Total Environment Disease of the skin: Clinical Dermatology Environmental Research Section A Regulation for the Environmental Law Analyst Journal of the American Statistical Association Technometrics Technometrics Environmental Health Perspectives Economic Geology Rotterdam.. Foster, A.L., Brown Jr., G.E., Tingle, T.N. and Parks, G.A. (1998) Quantitative arsenic speciation in mine tailings using X-ray absorption spectroscopy. , 83, 553ÿ568. Hahn, G.J. and Meeker, W.Q. (1991) . Wiley, New York. Hamilton, E.I. (2000) Environmental variables in a holistic evaluation of land contaminated by historic mine wastes: a study of multi-element mine wastes in West Devon, England using arsenic as an element of potential concern to human health. , 249, 171ÿ221. Hinwood, A.L., Sim, M.R., Jolley, D, de Klerk, N., Bastone, E.B., Gerostamoulos, J. and Drummer, O.H. (2004) Exposure to inorganic arsenic in soil increases urinary inorganic arsenic concentrations of residents living in old mining areas. , 26, 27ÿ36. Hudson-Edwards, K.A., Schell, C. and Macklin, M.G. (1999) Mineralogy and geochemistry of alluvium contaminated by metal mining in the Rio Tinto area, southwest Spain. , 14, 1015ÿ1030. Hudson-Edwards, K.A., Macklin, M.G., Miller, J.R. and Lechler, P.J. (2001) Sources, distribution and storage of heavy metals in the Rio Pilcomayo, Bolivia. , 72, 229ÿ250. Hughes, M.F. (2002) Arsenic toxicity and potential mechanisms of action. , 133, 1ÿ16. IAEA (1977) in Subramanian, K.S. (1996) Determination of metals in bio¯uids and tissues: sample preparation methods for atomic spectroscopic techniques. , 51, 291ÿ319. Iriondo, M. (1993) Geomorphology and late Quaternary of the Chaco (South America). , 7, 289ÿ303. JICA (1999) 734 American Mineralogist Statistical Intervals. A Guide for Practitioners Science of the Total Environment Environmental Geochemistry and Health Applied Geochemistry Journal of Geochemical Exploration Toxicology Spectrochimica Acta Letters Part B Geomorphology Estudio de evaluacõÂon del impacto AS UPTAKE BY RIVERSIDE COMMUNITIES, BOLIVIA ambiental producido por las actividades mineras en . Unpublished report for Pilcomayo Master Plan. Knobeloch, L. and Anderson, H. (2003) Effect of arsenic-contaminated drinking water on skin cancer prevalence in Wisconsin's Fox River Valley. Pp. 155ÿ163: (W.R. Chappell, C.O. Abernathy, C.L. Calderon and D.J. Thomas (editors). Elsevier B.V., Oxford, UK. Lee, J.-S. and Chon, H.-T. (2003) Exposure assessment of heavy metals on abandoned metal mine areas by ingestion of soil, crop, plant and groundwater. , 107, 757ÿ760. Mahon, K.I. (1996) The New `York' regression: application of an improved statistical method to geochemistry. , 38, 293ÿ303 Matschullat, J., Perobelli Borba, R., Deschamps, E., Ribeiro Figueiredo, B., Gabrio, T. and Schwenk, M. (2000) Human and environmental contamination in the Iron Quadrangle, Brazil. 15, 193ÿ202. Miller, J.R. (1997) The role of ¯uvial geomorphic processes in the transport and storage of heavy metals from mine sites. , 58, 101ÿ118. Miller, J.R., Hudson-Edwards, K.A., Lechler, P.J., Preston, D. and Macklin, M.G. (2004) Heavy metal contamination of water, soil and produce within riverine communities of the RõÂo Pilcomayo basin, Bolivia. , 320, 189ÿ209. Minoia, C., Sabbioni, E., Apostoli, P., Pietra, R., Pozzoli, L., Gallorini, M., Nicolaou, G., Alessio, L. and Capodaglio, E. (1990) Trace element reference values in tissues from inhabitants of the European community I. A study of 46 elements in urine, blood and serum of Italian subjects. , 95, 89ÿ105. Morton, J., Carolan, V.A. and Gardiner, P.H.E. (2002) Removal of exogenously bound elements from human hair by various washing procedures and determination by inductively coupled plasma mass spectrometry. , 455, 23ÿ34. Nelson, L.S. (1974) Factors for the Analysis of Means. , 6, 175ÿ181. Nelson, L.S. (1983) Exact critical values for use with the Analysis of Means. , 15, 40ÿ44. Nelson, L.S. (1993) Additional uses for the Analysis of Means and extended tables of critical values. , 35, 61ÿ71. Ott, E.R. (1967) Analysis of means ÿ a graphical procedure. , 24, 101ÿ109. Peterson, H.G. (1999) . http://www.agr.ca/pfra/water/®elPotosõÂ Arsenic Exposure and Health Effects V Journal de Physique IV France International Geology Review Applied Geochemistry Journal of Geochemical Exploration Science of the Total Environment Science of the Total Environment Analytica Chimica Acta Journal of Quality Technology Journal of Quality Technology Technometrics Industrial Quality Control dirr.pdf (accessed 20 April 2005). PROVISA (1989) Unpublished report for Ambio Chaco, Villa Montes, Bolivia. Rodushkin, I. and Axelsson, M.D. (2000) Application of double focusing sector ®eld ICP-MS for multielemental characterization of human hair and nails. Part II. A study of the inhabitants of northern Sweden. , 262, 21ÿ36. Rousseeuw, P.J. and Leroy, A.M. (2003) , Wiley, New York. Samanta, G., Sharma, R., Roychowdhury, T. and Chakraborti, D. (2004) Arsenic and other elements in hair, nails, and skin-scales of arsenic victims in West Bengal, India. , 326, 33ÿ47. Schollaert, A. (2000) . Unpublished report for AsociacõÂon Sucrense de Ecologia, Bolivia. Sera, K., Futatsugawa, S. and Murao, S. (2002) Quantitative analysis of untreated hair samples for monitoring human exposure to heavy metals. , 189, 174ÿ179. Smolders, A.J.P., Guerrero Hiza, M.A., Van Der Velde, G. and Roelofs, J.G.M. (2002) Dynamics of discharge, sediment transport, heavy metal pollution and Sabalo ( ) catches in the lower Pilcomayo River (Bolivia). , 18, 415ÿ427. Smolders, A.J.P., Hudson-Edwards, K.A., Van der Velde, G. and Roelofs, J.G.M. (2004) Controls on water chemistry of the Pilcomayo River (Bolivia, South-America). , 19 , 1745ÿ1758. Southwick, J.W., Western, A.E., Beck, M.M., Whitely, T., Isaacs, R., Petajan, J. and Hansen, C.D. (1983) An epidemiological study of arsenic in drinking water in Millard County, Utah. Pp. 210ÿ225 in: (W.H. Leadere and R.J. Fensterheim, editors). Van Nostrand Reinhold Company, New York. Thompson, M. and Howarth, R.J. (1980) The frequency distribution of analytical error. , 105, 1188ÿ1195. Tokunaga, H., Roychowdhury, T., Uchino, T. and Ando, M. (2005) Urinary arsenic species in an arsenicaffected area of West Bengal, India (part III). , 19, 246ÿ253. US EPA (2001) Quality Agriculture and Agri- Food Canada 735 Estudio de Factibilidad. Volumen 4, Anexo II: Hidrologia. Science of the Total Environment Robust regression and outlier detection Science of the Total Environment Monitoreo Ân contaminacio R Âõ o de voluntario Pilcomayo de en la el Departamento de Chuquisaca Nuclear Instruments and Methods in Physics Research B Prochilodus lineatus River Research and Applications Applied Arsenic: Industrial, Geochemistry Biomedical, Environmental Perspectives Analyst Applied Organometallic Chemistry National R e g u l a t io n s ; Compliance Monitoring Arsenic and New Primary Drinking and Water Clari®cations Source to Contaminants . http://www.epa.gov/safewater/ars/ J. ARCHER ET AL. arsenic_®nalrule.html (accessed 16 December 2004). White, M.A. and Sabbioni, E. (1998) Trace element reference values in tissues from inhabitants of the European Union. X. A study of 13 elements in blood and urine of a United Kingdom population. , 216, 253ÿ270. WHO (World Health Organization) (1996) Health criteria and other supporting information. Pp. 940 ÿ949 ndin: , 2 edition, Vol. 2. WHO, Geneva. WHO (World Health Organization) (1998) Addendum to Vol. 2. Pp. 281 ÿ283 in: , 2nd edition. WHO, Geneva. WHO (World Health Organization) (2001) , 2nd edition. WHO, Geneva. Wilkinson, M. and Mohler, R.R.J. (1995) Colmatation Science of the Total Environment Guidelines for Drinking-water Quality Guidelines for drinking- water quality Arsenic compounds, Environmental Health Criteria 224 of the Pilcomayo river as observed from space shuttle photography. , 27, 116. Wright, D.A. and Welbourn, P. (2002) Environmental toxicology. , 11 . Cambridge University Press, Cambridge, UK. York, D. (1967) The best isochron. , 2, 479ÿ482. Zierold, K.M., Knobeloch, L. and Anderson, H. (2004) Prevalence of chronic diseases in adults exposed to arsenic-contaminated drinking water. , 94, 1936ÿ1937. [ ] GSA Program with Abstracts Cambridge Environmental Chemistry Series Earth and Planetary Science Letters American Journal of Public Health Manuscript received 2 February 2005: revised 7 July 2005 736