Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/269775984

RiverTransportofMercuryfromArtisanalandSmall-ScaleGoldMiningandRisksforDietaryMercuryExposureinMadredeDios,Peru

ArticleinEnvironmentalSciences:ProcessesandImpacts·December2014

ImpactFactor:2.17·DOI:10.1039/C4EM00567H

CITATION

1

READS

345

8authors,including:

BethFeingold

DukeUniversity

4PUBLICATIONS1CITATION

SEEPROFILE

ErnestoJOrtiz

DukeUniversity

11PUBLICATIONS147CITATIONS

SEEPROFILE

AxelBerky

DukeUniversity

1PUBLICATION1CITATION

SEEPROFILE

WilliamKYPan

DukeUniversity

85PUBLICATIONS1,567CITATIONS

SEEPROFILE

Allin-textreferencesunderlinedinbluearelinkedtopublicationsonResearchGate,

lettingyouaccessandreadthemimmediately.

Availablefrom:ErnestoJOrtiz

Retrievedon:24June2016

EnvironmentalScienceProcesses & Impacts

PAPER

Publ

ishe

d on

19

Dec

embe

r 20

14. D

ownl

oade

d by

Duk

e U

nive

rsity

on

10/0

1/20

15 0

4:17

:18.

View Article OnlineView Journal

River transport o

aDepartment of Civil and Environmental E

Duke University, 121 Hudson Hall, Box 90

hsukim@duke.edu; Tel: +1(919)-660-5109bNicholas School of the Environment, Duke U

27710, USA. E-mail: william.pan@duke.educDuke Global Health Institute, Duke UniversidDepartment of Biostatistics and Bioinform

DUMC Box 2721, Durham, NC 27710, USAeDepartment of Environmental Biology an

Campus El Carmen 21071, Huelva, Spain

† Electronic supplementary informa10.1039/c4em00567h

‡ Current affiliation: Department of EnvPublic Health, University at Albany, GeoUniversity Place, Rensselaer, NY 12144, U

Cite this: DOI: 10.1039/c4em00567h

Received 21st October 2014Accepted 19th December 2014

DOI: 10.1039/c4em00567h

rsc.li/process-impacts

This journal is © The Royal Society of

f mercury from artisanal andsmall-scale gold mining and risks for dietarymercury exposure in Madre de Dios, Peru†

Sarah E. Diringer,a Beth J. Feingold,‡bc Ernesto J. Ortiz,c John A. Gallis,cd

Julio M. Araujo-Flores,e Axel Berky,b William K. Y. Pan*bc and Heileen Hsu-Kim*a

Artisanal and small-scale gold mining (ASGM) is a major contributor to deforestation and the largest

anthropogenic source of atmospheric mercury worldwide. Despite significant information on the direct

health impacts of mercury to ASGM miners, the impact of mercury contamination on downstream

communities has not been well characterized, particularly in Peru's Madre de Dios region. In this area,

ASGM has increased significantly since 2000 and has led to substantial political and social controversy.

This research examined the spatial distribution and transport of mercury through the Madre de Dios

River with distance from ASGM activity. This study also characterized risks for dietary mercury exposure

to local residents who depend on fish from the river. River sediment, suspended solids from the water

column, and fish samples were collected in 2013 at 62 sites near 17 communities over a 560 km stretch

of the Madre de Dios River and its major tributaries. In areas downstream of known ASGM activity,

mercury concentrations in sediment, suspended solids, and fish within the Madre de Dios River were

elevated relative to locations upstream of mining. Fish tissue mercury concentrations were observed at

levels representing a public health threat, with greater than one-third of carnivorous fish exceeding the

international health standard of 0.5 mg kg�1. This study demonstrates that communities located

hundreds of kilometers downstream of ASGM activity, including children and indigenous populations

who may not be involved in mining, are at risk of dietary mercury exposure that exceed acceptable body

burdens. This report represents the first systematic study of the region to aid policy decision-making

related to ASGM activities in Peru.

Environmental impact

Artisanal and small-scale gold mining (ASGM) is the largest anthropogenic source of mercury to the atmosphere. While the impacts of ASGM to the health ofminers have been studied in great detail, the implications for water quality in local watersheds have not. This study is the rst to describe the extent of mercuryrelease in the Madre de Dios watershed, a biodiversity hotspot within the Peruvian Amazon where ASGM activity is prevalent. The results demonstrate a gradientof increasing mercury contamination in the river downstream of mining areas and signicant risk of dietary mercury exposure for community members who eatsh hundreds of kilometers from the mining activity.

ngineering, Pratt School of Engineering,

287, Durham, NC 27710, USA. E-mail:

niversity, 450 Research Dr, Durham, NC

; Tel: +1(919)-684-4108

ty, 310 Trent Dr, Durham, NC 27710, USA

atics, Duke University Medical Center,

d Public Health, University of Huelva,

tion (ESI) available. See DOI:

ironmental Health Sciences, School ofrge Education Center, Room 145, OneSA.

Chemistry 2015

Introduction

Artisanal and small-scale gold mining (ASGM) is a largelyunregulated sector of the global economy in more than 70countries and is rising at a dramatic rate.1,2 During the miningprocess, elemental mercury Hg(0) is added to large quantities ofsediment and soils that have been exhumed from riverbanksand forested areas. Mercury forms a strong bond with gold toseparate it from the river sediments or solids. The gold–mercuryamalgam is then heated in the eld and in gold shops toseparate the metals. Miners involved in amalgam burning areexposed to extremely high levels of Hg(0) and have greaterincidence of neurological disorders and kidney dysfunction, acommon side effect of Hg(0) inhalation.3 Along with

Environ. Sci.: Processes Impacts

Environmental Science: Processes & Impacts Paper

Publ

ishe

d on

19

Dec

embe

r 20

14. D

ownl

oade

d by

Duk

e U

nive

rsity

on

10/0

1/20

15 0

4:17

:18.

View Article Online

contributing signicantly to deforestation, ASGM has surpassedfossil fuel combustion as the largest anthropogenic source ofmercury (Hg) to the global atmosphere.2 Up to 60% of the Hgused in ASGM is released directly to the environment and canaccumulate in sediment on site, in the atmosphere, and indownstream rivers.4

In river sediments, anaerobic microorganisms methylate Hgto produce monomethylmercury (MeHg), a highly neurotoxicform of Hg that biomagnies in aquatic food webs.5 In humans,excess MeHg intake through sh consumption can decreasecognitive and kidney function. Moreover, MeHg is capable ofcrossing the placental barrier to affect fetuses in pregnantwomen, leading to impaired neurodevelopment, cognitivefunction, and motor skills. Fish is an important food source forboth nutrition and culture in communities throughout theworld and is an especially important source of nutrients forchildren.6–8 People who are not involved directly in ASGM butlive near mining sites oen have their greatest Hg exposurefrom sh consumption.9,10 Thus, the bioaccumulation of MeHgin sh poses signicant human health risks for these commu-nities, especially for children and women of childbearing age.

The Madre de Dios (MDD) region of Peru is located in theheadwaters of the tropical Amazon. It is one of the world's mostbiodiverse ecosystems and a prioritized biodiversity hotspot forconservation.11 ASGM has occurred in the MDD region since the1970's. However, activity has grown immensely since the early2000's, corresponding with the signicant increases in theinternational gold price.1 The extent of land use for gold miningin the MDD region increased by 400% between 1999 and 2012.12

With this increase in ASGM activity, rapid deforestation hasbeen recorded, including nearly 2000 ha per year of newlydeforested area between 2006 and 2009.1 ASGM provides animportant source of income in the MDD region. Unconrmedreports have estimated that ASGM employs between 10 000 and30 000 active miners in the MDD region and makes up nearly50% of the local economy.9,13,14 While Hg use for ASGM iswidespread in the region, the release of this metal to thesurrounding watershed and the implications for human expo-sure have not been studied in a systematic manner. Neverthe-less, the Peruvian government has recently begun to enforceregulatory requirements for ASGM with military action andembargos on gasoline, resulting in riots and signicant socialunrest.15,16 With these economic and social implications atstake, it is imperative that the environmental and healthimpacts of Hg from ASGM are quantied.

The specic objectives of this study were to: (1) examine thedistribution of Hg in the MDD River environment at pointsupstream and downstream of ASGM activities; (2) examinetransport of Hg from concentrated areas of mining throughdirect release and runoff; and (3) evaluate potential risks ofdietary Hg exposure via sh for communities along the river.

Study region

The study region included the portion of theMDD River startingat Atalaya (located at km 1) to Puerto Pardo (located at km 560downstream) at the Peru-Bolivia border (Fig. 1). For analysis and

Environ. Sci.: Processes Impacts

discussion purposes, the MDD River was divided into threesections to determine the inuence of ASGM on Hg in the river.Section 1 (km 1 to 180) represents the upstream area with littleor no active ASGM activity. Beginning near km 180, ASGMoccurred with the greatest density along the MDD River up tokm 400.12 Moreover, the most concentrated areas of miningactivity have been focused along two major tributaries that feedinto the MDD River: the Colorado and Inambari Rivers withoutlets at km 225 and 356 respectively.1,12 These areas (known asHuepetuhe, Guacamayo, and Delta-1) represent approximately50% of the total gold mining in the region.1,12 Section 3 (km 401to km 560) contains fewer active ASGM areas and the largepopulation center of Puerto Maldonado where a large numberof gold shops are located. Additional samples were collected atLago Valencia, an oxbow lake in Section 3 at km 545 that hasseasonal sediment exchange during the wet season (roughlyNovember–March) and limited exchange during the dry season(roughly April–October). Fishing is common in Lago Valencia,and most of the sh caught within the lake are sold in areasbetween Puerto Maldonado and Puerto Pardo.

Materials and methodsField methods

Two sampling events were performed in 2013: once during thewet season (March–April) and once during the dry season (June–July). Sediment and whole water samples were collected at 62sites along a 560 km reach of the MDD River from Atalaya (km 1)through Puerto Pardo (km 560 downstream) at the Peru-BoliviaBorder. Sites on the MDD River included locations within 3 kmupstream and 3 km downstream of conuences with majortributaries (Manu, Colorado, Inambari, and Tambopata Rivers).Additional samples were collected on each of the majortributaries approximately 1 km upstream of the conuence withthe MDD River. At each site, a water quality data sonde(Professional Plus Pro, YSI, Incorporated, Yellowsprings, Ohio)was deployed in the water at mid-channel for quantication ofpH, dissolved oxygen, conductivity, and temperature.

Surface water samples were collected in duplicate or tripli-cate at mid-channel and <0.5 m depth using 1 L pre-cleanedpolyethylene containers (VWR). Whole water samples wereltered in the eld using pre-weighed 0.22 mm glass ber lters(Whatman QM-A grade) placed on a polyethylene vacuumltration apparatus for total suspended solids (TSS) and totalsuspended particulate Hg (HgP) concentrations. Filters for TSSwere placed in plastic bags for storage. Filters for HgP wereplaced in glass vials with uoropolymer-lined screw caps (pre-cleaned vials for trace mercury analysis from Brooks Rand).Sediment samples were obtained in triplicate at locationswithin 2 m of the bank edge on both sides of the river. Sedimentfrom the top 5 cm were scooped by gloved hand and placed intotrace metal-cleaned polyethylene containers (VWR). All sampleswere placed on dry ice and frozen at �20 �C in the eld usingCredo Cube Series 20M (Minnesota Thermal Science, Plymouth,MN). Samples were then shipped and stored at �20 �C uponreturn to the laboratory at Duke University.

This journal is © The Royal Society of Chemistry 2015

Fig. 1 Environmental and human health study sites in the Madre de Dios (MDD) watershed in Peru. Communities (red circles) on the MDD Rivernear sampling sites for sediment, water, and fish. Known regions of mining layers are courtesy of the Asociacion para la Conservacion de laCuenca Amazonica (ACCA) and denoted in pink; concentrated areas of mining include the Huepetuhe/Delta-1 mining area (red) and theGuacamayo mining area (orange).

Paper Environmental Science: Processes & Impacts

Publ

ishe

d on

19

Dec

embe

r 20

14. D

ownl

oade

d by

Duk

e U

nive

rsity

on

10/0

1/20

15 0

4:17

:18.

View Article Online

Analysis of Hg concentration in ltered water samples,which typically includes dissolved and colloidal Hg phases, wasnot performed due to the lack of immediate access to trace-metal clean facilities to preserve these samples within 2–3 daysof collection. Mercury in surface waters, however, tends to bepredominantly associated with large particles (>0.22 mm).17–19

Thus, with the study's aim to assess the extent of mercurycontamination over the entire reach of the river, we focusedresearch efforts on the quantication of suspended particulateHg concentration (HgP) in the river water column.

Surface water velocity was measured in the MDD River andthe four tributaries during the June/July 2013 sampling event.For these measurements, a buoyant surface oat was releasedfrom the boat and allowed to reach steady speed in the river.Travel time of the oat for a dened distance was observed fromshore and recorded as the surface water velocity. Cross-sectionalarea at each site was estimated using river width and depth atthree equally spaced points across a transect of the river. Riverdischarge rates were calculated for each location by multiplyingthe cross-sectional area with the surface water velocity.

Whole sh were obtained by local shermen at 37 of thesediment/water sampling sites along the MDD River usingtraditional shing techniques, including cast net, overnight

This journal is © The Royal Society of Chemistry 2015

netting, and hook-and-line. A total of 200 sh were collectedduring both seasons, including 123 carnivorous sh, 74 non-carnivorous sh, and 3 unknown. Fish were weighed,measured, and identied in the eld, with identicationconrmed through photos. Carnivorous sh included piscivo-rous, insectivorous, and omnivorous sh species. Fish scaleswere removed in the eld and skin was removed upon return tothe laboratory at Duke University. Fish muscle-tissue sampleswere obtained in triplicate from the anterior portion of each shabove the lateral line. Samples were frozen at�20 �C in the eldand then shipped and stored at �20 �C upon return to thelaboratory at Duke University.

Chemical analyses

Total Hg concentration in sediments and sh muscle tissuewere determined by direct thermal decomposition, amalgam-ation, and atomic absorption spectrometry (Milestone DMA-80).20 Total Hg was quantied as a proxy for MeHg in sh tissuesince generally >80% of Hg in sh tissue is in the form ofMeHg.21 Analysis of standard reference materials for Hg insediment (NIST 2709a) and sh (DORM-3 and DORM-4) by thismethod resulted in recoveries of 98%� 5% SD (n¼ 52), 103%�4% SD (n ¼ 28), and 101% � 5% SD (n ¼ 33), respectively.

Environ. Sci.: Processes Impacts

Environmental Science: Processes & Impacts Paper

Publ

ishe

d on

19

Dec

embe

r 20

14. D

ownl

oade

d by

Duk

e U

nive

rsity

on

10/0

1/20

15 0

4:17

:18.

View Article Online

Methylmercury concentration in sediments was determinedby dichloromethane extraction, aqueous phase ethylation,purge-trap on Tenax resin and gas chromatographic separation,and inductively coupled plasma mass spectrometry (Tekran2600, Agilent 7700).22 A stable MeHg isotope (Me201Hg) wasadded to each sample as an internal standard prior to theextraction step. MeHg concentrations were corrected for eachindividual sample based on the recovery of its respective stableisotope spike. Extractions that achieved stable isotope recov-eries between 75% and 125% were accepted in the data set. Theaverage recovery was 87%.

Filters analyzed for TSS were dried at 104 �C for 4 h andweighed to determine total suspended particulate mass. Filtersfor HgP determination were dissolved in 3 : 7 sulfuric acid–nitric acid and heated at 60 �C for 16 h. The extracts werepreserved with 1% v/v bromine monochloride, stored at roomtemperature for at least 12 h, and analyzed by stannous chloride(SnCl2) reduction, gold amalgamation, and cold vapor atomicuorescence spectroscopy (MERX-T, Brooks Rand).17,22 Tracemetal grade acids (VWR) were used for digestion reagents, andmethod blanks were performed to ensure that background Hgconcentration was below instrument detection limits. The totalHg quantied on each 0.22 mm glass-ber lter was divided bythe volume of water passed through the lter to yield theparticulate mercury concentration (HgP, ng L�1). The concen-tration of Hg per mass of suspended solid (HgSS, ng g�1) wascalculated from the HgP value divided by the TSS concentration(mg L�1).

Statistical methods

Descriptive statistics stratied by season were produced for theconcentrations mentioned above using means, standard devi-ations (SD), and 95% condence intervals (C.I.) (Table 1).

Comparisons of river sections in regards to total Hg insediment, MeHg in sediment, Hgp and TSS were performed withordinary least squares (OLS) regression analysis of the rawaverages of duplicate or triplicate eld samples at each site.Sites with only a single measurement were excluded from theanalysis. Total Hg in sediment, MeHg in sediment, Hgp and TSSconcentrations in the three river sections were estimated fromthe OLS models. Linear splines were created for each of the

Table 1 Average concentrations of total mercury (HgT), methylmercurySolids (TSS) in each section with 95% confidence intervals based on ord

Total Hg(mg kg�1)n ¼ 59

MeHg(pg g�1)n ¼ 20

Suspended(ng L�1)

Dry seasonseason 2

River Section 1(km 1 to 180)

9.0 (1.4, 16.6) 32.5 (�268.4, 333.4) 0.4 (�3.2,

River Section 2(km 181 to 400)

22.6 (17.4, 27.7) 277.7 (159.7, 395.8) 10.4 (7.6, 1

River Section 3(km 401 to 560)

31.0 (23.7, 38.3) 102.5 (�85.9, 292.8) 15.0 (11.6,

Environ. Sci.: Processes Impacts

three sections in order to compare spatial changes in Hgconcentrations between each section (Table S1†). Each sectionwas also compared using a simple categorical variable forstream section (Tables S2 and S3†). Suspended particle (TSS andHgP) models were adjusted for season.

Fish tissue Hg concentrations were also compared betweensections for carnivorous and non-carnivorous sh (Table 2).Correlations between sh parameters (e.g., Hg concentration,sh diet, sh length, sh weight, and river section) wereexamined using Pearson correlations (denoted PC). Statisticalsignicance was dened as p < 0.05. All statistical analyses wereperformed using SAS version 9.4 (SAS Institute, Cary, NC).

Results and discussionMercury distribution in the Madre de Dios River

Total mercury concentrations in sediment increased withdistance downstream along the MDD River (Fig. 2, Tables 1 andS3†). In locations upstream of known mining inuences (km 1to 180; Section 1), Hg concentrations averaged 9.0 ng g�1 (95%C.I. 1.5, 16.6). Total Hg concentrations in sediment in Section 2near active mining (km 181 to 400) were signicantly greaterthan in Section 1 by approximately 13.5 ng g�1 (95% C.I. 4.4,22.6; p¼ 0.004). Sediment concentrations continued to increasesignicantly in Section 3, downstream of the dense miningactivity with an average of 21.9 ng g�1 (95% C.I. 11.5, 32.4;p < 0.001) greater sediment Hg in Section 3 than in Section 1. Thegreatest observedHg concentration in sediment (up to 95.3 ng g�1)was measured in Section 3 at Palma Real (km 538) (Fig. 2a). Theincrease in the average concentration from Section 2 to Section3 was nearly signicant (p ¼ 0.064) and indicates continuedinput of mercury to Section 3. Further, sediment Hg concen-trations within Inambari River, which drains the Huepetuheand Guacamayo mining areas, were twice as high on averagecompared to the MDD River upstream of the tributary (19.2 mgkg�1 compared to 39.6 mg kg�1). This suggests new inputs of Hgand transport of contaminated sediments to areas downstreamof large mining areas. Sediment concentrations did not varysignicantly between seasons (p ¼ 0.66), so the model was notadjusted for season.

(MeHg), suspended particulate mercury (HgP), and Total Suspendedinary least squares regression analysis

particulate Hg (HgP),Total suspended solids (mg kg�1)

, Wet season,season 1

Dry season,season 2

Wet season,season 1

3.9) 7.0 (3.5, 10.6) 124.8 (30.2, 29.5) 526.5 (437.3, 615.8)

3.1) 17.0 (14.2, 19.8) 215.0 (140.5, 290.3) 617.1 (545.8, 688.4)

18.3) 21.6 (18.3, 25.0) 112.4 (21.7, 203.0) 514.1 (426.0, 602.1)

This journal is © The Royal Society of Chemistry 2015

Table 2 Total Hg concentrations in fish muscle tissue (mg kg�1) from the Madre de Dios River and the percentage of fish exceeding 0.5 mg kg�1,the World Health Organization (WHO) standard

Non-carnivorous sh Carnivorous/omnivorous sh

Location in Madre de DiosRiver: communities affected

Average Hg conc. in mg kg�1

(95% C.I.)% Exceeding WHOthreshold

Average Hg conc. in mg kg�1

(95% C.I.)% Exceeding WHOthreshold

Section 1 (km 1 to 180):Atalaya through TamboBlanquillo

0.08 (0.00, 0.17) 0% (0/13) 0.20 (0.01, 0.39) 6% (1/18)

Section 2 (km 181 to 400):Tambo Blanquillo throughLaberinto

0.29 (0.24, 0.34) 12% (5/41) 0.77 (0.64, 0.90) 55% (26/47)

Section 3 (km 401 to 560):Tres Islas to Puerto Pardoa

0.18 (0.11, 0.26) 9% (1/20) 0.34 (0.22, 0.45) 24% (14/58)

Lago Valencia (km 545) 0.14 (0.07, 0.21) 0% (0/9) 0.45 (0.20, 0.70) 50% (4/8)

a Section 3 includes Lago Valencia (km 545).

Paper Environmental Science: Processes & Impacts

Publ

ishe

d on

19

Dec

embe

r 20

14. D

ownl

oade

d by

Duk

e U

nive

rsity

on

10/0

1/20

15 0

4:17

:18.

View Article Online

The concentration of methylmercury (MeHg) in the sedi-ments and the percent of the total Hg as MeHg (%MeHg)increased with proximity to mining areas (Fig. 2b and c). At sitesin Section 2, proximal to the greatest density of active mining

Fig. 2 (a) Total mercury (Hg) concentrations and (b) methylmercury(MeHg) concentrations (dry weight basis) in bottom river sediment ofthe MDD River and major tributaries to the river; (c) the percentage ofthe total Hg as MeHg in sediments. Each data point represents theaverage of all samples at one field site collected during the wet season(March–April) and dry season (June–July) of 2013. Error bars arestandard deviations of the average field sample (n ¼ 2–3).

This journal is © The Royal Society of Chemistry 2015

activities, MeHg concentrations reached concentrations thatwere more than ten times the maximum concentrations foundin Section 1 (670 pg g�1 compared to 43.8 pg g�1; Fig. 2b). Thelargest MeHg concentration (837 ng g�1) for the entire study wasobserved in the Inambari River. The average concentrations inSection 2 were 254.8 pg g�1 (95% C.I. 22.8, 486.9; p ¼ 0.03) andgreater than in Section 1. The average MeHg concentration wasalso greater in Section 2 than in Section 3, though not statisti-cally signicant for alpha ¼ 0.05 (p ¼ 0.11; Table S3†). Theincrease in MeHg from Section 1 to Sections 2 and 3 also cor-responded with increased temperature and decreased dissolvedoxygen (Fig. S1†), which may indicate conditions suitable forMeHg production. However, there were no apparent correla-tions between sediment or water quality parameters thatdirectly impacted MeHg concentration or percent. Average Hgand MeHg in sediment as well as condence intervals withineach river section can be found in Table S2.†

These results collectively indicate that portions of the MDDRiver downstream of mining activity not only had signicantlygreater total Hg content in sediments, but also appeared to hostlocations with large net methylation potential for mercury, asindicated by the greater %MeHg values in downstream sitesrelative to upstream sites. While MeHg generally represented asmall proportion of the total Hg in the MDD River sediments(<3% as MeHg), this type of Hg can biomagnify within the foodchain and is the major form of mercury in sh muscletissue.23–25 Areas with relatively greater %MeHg values maysignify regions of the river with high bioaccumulation potential.

Mercury and particle transport in the Madre de Dios River

In river systems, suspended particles play an important role inHg transport, as the predominant fraction of Hg is typicallyassociated with particulate matter in the water column.17–19

Total suspended solids (TSS) concentrations were generallygreater during the March–April wet season than in the June–Julydry season (p < 0.01) (Fig. 3a). The large scatter in the data islikely due to rain events, which were not removed from the dataseries. During each sampling season, Section 2 contained thegreatest average TSS concentrations relative to the other

Environ. Sci.: Processes Impacts

Environmental Science: Processes & Impacts Paper

Publ

ishe

d on

19

Dec

embe

r 20

14. D

ownl

oade

d by

Duk

e U

nive

rsity

on

10/0

1/20

15 0

4:17

:18.

View Article Online

sections of the river for the same season. The average TSS inSection 2 was 228 mg L�1 (95% C.I. 161, 295) during the dryseason and 642 mg L�1 (95% C.I. 574, 711) during the wetseason. TSS concentrations in Section 2 were greater than inSection 1 (p ¼ 0.03) and in Section 3 (p ¼ 0.016) (Table S2†).

Deforestation oen leads to increased sediment loading towaterways and downstream rivers by increasing surfacerunoff.26,27 In the MDD region, ASGM activities have led to morethan 50 000 hectares of deforestation since 1999, which exceedsforest loss from ranching, agriculture, and loggingcombined.12,28 In the catchment area that feeds into the riversampling sites of this study, much of the mining activity anddeforestation is localized within Section 2 and its majortributaries as shown through satellite images from Asner et al.2013 and mapping efforts by Asociacion para la Conservacionde la Cuenca Amazonica. The increased TSS in Section 2 wascoincident with increased mining activity and likely indicatesmining in the region enhances sediment transport to down-stream areas.

Fig. 3 (a) Total suspended solids (TSS) and (b) suspended particulateHg (HgP) from the surface water of theMDD River andmajor tributariesduring the two sampling events in 2013. Error bars indicate standarddeviation among triplicate field samples. The solid lines and dashedlines represent linear splines for each section during the March/Apriland June/July seasons, respectively. Regression analysis and splineinformation available in Tables S1 and S2.†

Environ. Sci.: Processes Impacts

Suspended particulate mercury (HgP) generally represents themajor Hg transportmechanism throughout the river system.17–19 Inthe MDD River surface water, HgP concentrations increased withdistance downstream through Section 2 (Fig. 3b, Table S3†),similar to trends observed for Hg concentrations in the bottomsediments. River Section 2 had an estimated average HgPconcentration that was 9.97 ng L�1 greater than in Section 1 (95%C.I. 6.1, 13.8; p < 0.001). Concentrations in Section 3 stabilized withan estimated increase of 3.5 ng L�1 relative to Section 2 (95% C.I.�0.2, 7.3; p¼ 0.063; Table S2†). HgP concentrations were greater inthe March–April wet season and lower in the June–July dry season(p < 0.01) due to greater TSS concentration during the wet seasonthan the dry season (Fig. 3).

The increase of HgP with distance downstream, however, wasnot simply due to increased TSS concentration in the watercolumn. In both seasons TSS decreased in Section 3 while HgPremained relatively constant (the trends calculated for HgP inSection 3 were not signicant, Fig. 3 and Table S1†). Theseresults suggest the existence of additional sources of Hg to theregion. Section 3 includes the capital city of the MDD region,Puerto Maldonado, where TSS and HgP are likely inuenced byurbanization, sewage discharge, and gold amalgam burning.Further work should examine the inuence of urban releases toTSS and HgP dynamics.

Other surface water quality parameters such as pH,conductivity, temperature and dissolved oxygen did not varydramatically between sections or seasons, though some trendswere identied (Table S4†). From the upper watershed (Section1) to mid-region and lower watershed (Sections 2 and 3),temperature increased and dissolved oxygen concentrationsdecreased during both seasons. This pattern was consistentwith conditions of atmospheric O2 saturation in the surfacewater (p < 0.01; Fig. S1†).

Collectively, these results of the surface water analysesindicated that particulate Hg concentrations in the watercolumn were closely linked to suspended solids and Hg in theriver sediment, probably via particle settling and reentrain-ment. However, differences in HgP concentrations betweenSections 2 and 3 were not explained simply by changes in TSSconcentrations in the river. In both the water column andsediment, increased amounts of Hg concentrations coincidedwith increased anthropogenic activities such as ASGM, defor-estation, and agricultural expansion.

Hydrologic transport of mercury from large areas of mining

The Colorado and Inambari Rivers drain the largest areas ofactive mining in the watershed and are major tributaries thatfeed into the MDD River at km 221 and km 355, respectively(Fig. 1).1,12 The surface water in these tributaries containedsignicantly higher total concentrations of particulate mercurythan in the MDD River, expressed as concentration in the water(HgP) (Fig. S2†). Furthermore, the concentration of Hg stan-dardized by total suspended solids mass (HgSS, mg Hg per kgsolids) was greater in the tributaries as compared to the MDDRiver (Fig. S3†). The TSS in the major tributaries was generallyless than in the MDD River (Fig. S4†).

This journal is © The Royal Society of Chemistry 2015

Paper Environmental Science: Processes & Impacts

Publ

ishe

d on

19

Dec

embe

r 20

14. D

ownl

oade

d by

Duk

e U

nive

rsity

on

10/0

1/20

15 0

4:17

:18.

View Article Online

While HgP and HgSS values were greater in the tributaries,HgP in the MDD River directly downstream of conuences withColorado and Inambari Rivers did not increase as compared tothe upstream (Fig. S2†). This was likely due to dilution of thetributaries by the main river. For example during the June–Julysampling season, the ow rate in the MDD River was approxi-mately 80 m3 s�1 at sites within 40 km upstream of theconuence with the Colorado River (Fig. S5†). In contrast, theColorado River ow rate was approximately 10 m3 s�1, indi-cating that the Colorado was diluted by a factor of nine aerjoining the MDD River. At the Inambari River, the ow rate ofthe tributary (120 m3 s�1) was approximately equal to the MDDRiver ow (140m3 s�1), representing a dilution of approximatelytwo upon entering the MDD River.

A mass balance of Hg inputs from the tributaries for theJune/July season was performed by comparing expected andcalculated HgP concentrations downstream of each conuence.In all four tributary conuences, the observed and calculatedHgP values were consistent with each other (Fig. S6†), indicatingthat the major input at this point in the MDD River wascaptured by the tributary input.

While HgP concentrations did not appear to signicantlychange downstream of conuences, Hg loading rates fromthese tributaries were relatively high, indicating a source of totalHg to the river. Hg loading rates (calculated by multiplying owrate and HgP concentration) increased from 1.3 mg s�1

upstream of the conuence with the Inambari River (km 356) to2.4 mg s�1 directly downstream (Fig. S7†). Likewise at theconuence with the Colorado River (km 224), Hg loading rateswere 0.18 mg s�1 upstream and 0.37 mg s�1 downstream, adoubling of Hg transport downstream. Additional monitoringof the Inambari River is needed in order to identify specicareas of direct Hg release from mining areas.

The Tambopata River drains from the Tambopata Reserve,which until recently was a relatively protected region. In the pastve years, mining has increased in the Malinowski River area,which drains into the Tambopata River.12 While the TambopataRiver does not appear to contribute signicantly to Hg loading(Fig. S7†), there was a dramatic increase of Hg loading down-stream of the conuence. Puerto Maldonado is an urban centerlocated at the conuence of the Tambopata and MDD Rivers.While mining occurs near Puerto Maldonado with lessfrequency, Hg is actively used in gold shops to further purify orestaken from the mining areas. Release of Hg and urban runofffrom both mining in the Malinowski River and Puerto Maldo-nado may strongly impact Hg and suspended solids concentra-tions in the MDD River in Section 3 near km 484. Further, asdiscussed above, the factors controlling HgP and TSS in urban-ized areas within the Amazon should be examined further.Urban runoff, including direct release of sewage and erosion,may have dramatic effects on local Hg dynamics in the rivers.

Fig. 4 Average Hg concentrations in fish tissue (wet weight basis) for18 commonly eaten fish collected from the Madre de Dios River. Errorbars represent the 95% confidence interval around the meanconcentration throughout the region. The green dashed line repre-sents the WHO and Peru's Ministry of Health recommendation for safehuman consumption. The solid and bold dashed lines representweekly and monthly consumption guidelines from the US EPA.

Mercury in sh tissue

Fish obtained from the sampling events included a wide varietyof species (Table S5†) and represented sh that are typicallyeaten by local residents. In this study, 47 of 200 (24%) sh

This journal is © The Royal Society of Chemistry 2015

samples, and 39 of 123 (32%) carnivorous sh samples excee-ded the mercury guideline for human consumption set by theWorld Health Organization (0.5 mg kg�1). Moreover, 79 of 200(40%) total sh samples and 59 of 123 (48%) in carnivorous shsamples exceeded the US Environmental Protection Agency(EPA) sh tissue based water quality limit of 0.3 mg Hg kg�1 ofsh, indicating severe water quality impairment.

Carnivorous sh species contained greater concentrations ofHg than non-carnivorous sh (p < 0.01). The higher trophic levelspecies demonstrated greater frequency for exceeding the WHOstandard for Hg. For example, the average Hg concentrations ofsabalo and chambira species were greater than 1 mg kg�1 of Hg(Fig. 4), a threshold in which sh consumption is recom-mended to be no more than once per month.29 For other speciesincluding dorado, doncella, toa, and corvina, the average Hgconcentrations exceeded the WHO concentration guideline of0.5 mg kg�1. All of these sh species were omnivores or strictcarnivores and are expected to contain greater levels of Hg dueto food web biomagnication.

Mercury concentrations in the muscle tissue of the shincreased with distance from the headwaters at Atalaya (km 1)to sites further downstream (Fig. 5), and were signicantlygreater in Section 2 (km 181 to 400) than in Section 1 (p < 0.001)or Section 3 (p < 0.001). Changes in mercury concentrations inmuscle tissue of the sh cannot fully be explained by shlength, which might be a proxy for sh age. Mercury concen-trations did not strongly correlate with sh length for carnivo-rous or non-carnivorous species (PC < 0.40; Fig. S8a†). For thesix species of sh with sample numbers equal to or greater than10, Hg concentrations were not strongly correlated with length(PC < 0.5 in all cases; Fig. S8b and c†). Instead, for the non-carnivorous species bocachico and yahuarachi, Hg concentra-tions exceededWHO standards only at sites near Boca Inambari(km 355), a community directly downstream of dense miningactivity. The highest Hg concentration in a single sh was 2.8mg kg�1 from a chambira caught within the river downstream ofmining at San Juan Grande (km 265).

Environ. Sci.: Processes Impacts

Environmental Science: Processes & Impacts Paper

Publ

ishe

d on

19

Dec

embe

r 20

14. D

ownl

oade

d by

Duk

e U

nive

rsity

on

10/0

1/20

15 0

4:17

:18.

View Article Online

Fish mobility in the region is not well documented and maycontribute to the high variability of Hg concentrations amongsh tissue samples within sections. As indicated in Table S5,†most of the sh species in this study are migratory (i.e. capableof traveling more than 100 km during its lifetime). Non-migratory sh accounted for 48 of 200 (41 of 123 carnivorous)sh samples. An analysis of Hg concentration in sh was per-formed for the non-migratory sh, and the outcome alsodemonstrated increased concentrations in Sections 2 and 3relative to Section 1 (Fig. S9†). However, the condence intervalswere relatively large due to small sample size.

Mercury release in the MDD and implications forbioaccumulation

Mercury concentrations in the MDD River sediments wereapproximately one order of magnitude lower than Hg concen-trations reported for other regions with ASGM activities, such asin Brazil (41 to 346 ng g�1),30 Guyana (29 to 1200 ng g�1),31

Suriname (130 to 220 ng g�1),32 and the La Rinconada MineComplex in Peru (up to 232 000 ng g�1).33 Concentrations weremore similar to regions not directly impacted by ASGM,including the Tambopata River in the MDD region in 2009 (13ng g�1),34 and Veracruz State, Mexico (27.5 to 90.5 ng g�1).35

Despite relatively low concentrations of Hg in sediment andsuspended solids within the MDD River, sh tissue concentra-tions in the central and lower region were in the same range orhigher than regions with greater total Hg in sediment.30,32,33

The benthic ecosystem in the central and lower region of theMDD River may be areas of high net MeHg production relativeto portions near the headwaters of the river (as indicated bygeographical trends in %MeHg values, Fig. 2c). Studies haveshown that Hg aging in sediments over days and weeks result inless reactive and bioavailable Hg forms such as HgS mineralswhen compared to newer sources such as Hg freshly depositedfrom the atmosphere or released into the river as dissolved

Fig. 5 Total Hg concentrations in carnivorous and non-carnivorousfish muscle tissue (mg kg�1, wet weight) collected from the main stemof the Madre de Dios River in 2013. 95% confidence intervals aroundthe mean fish concentration within each section are represented byred, solid boxes for carnivorous fish and blue, dashed boxes for non-carnivorous fish. The WHO limit (solid line) for human consumption(0.5 mg per kg Hg) is based on human health threshold for weeklyconsumption. The US EPA Criterion (dashed line) is based on a fishtissue based water quality criterion (0.3 mg per kg MeHg).

Environ. Sci.: Processes Impacts

Hg.5,36 In regions with high rainfall, Hg inputs from landscaperunoff can contribute to increasing ‘new’ sources of Hg inwaterways before it is able to be sequestered in soil.37 While theresults of this study indicated that Hg was transported from themining sites to the MDD River via the Colorado and InambariRivers, the reactivity and methylation potential of this mercuryneeds to be further explored, particularly in relation to naturallyoccurring Hg, and Hg that deposited onto the landscape viaatmospheric deposition and reached the river via landscapedrainage. If Hg directly released to waterways from ASGM ishighly bioavailable for methylating bacteria, then this source ofHg may have a greater risk for methylation and bio-accumulation than Hg from native or natural sources.

Risks for dietary exposure to mercury

Mercury concentrations in sh from the MDD River were usedto calculate probabilities of dietary Hg exposure exceedingpublic health guidelines. For communities located in Section 2,the risk of consuming sh that exceed the 0.5 mg kg�1 WHOthreshold is greater than in other two sections of the MDDRiver. Between Tambo Blanquillo and Laberinto (km 181 to400), 55% of the carnivorous sh samples contained Hgconcentrations exceeding 0.5 mg kg�1 as compared 6%exceeding in Section 1 and 24% in Section 3 (Table 2). In LagoValencia (a large oxbow lake that receives water from and drainsinto the MDD River at km 545), concentrations of Hg in shtissue were also high relative to Sections 1 and 3 of the MDDRiver. Average carnivorous sh tissue concentrations from LagoValencia did not differ signicantly from Section 2 (p ¼ 0.39,Fig. S10†).

The observed sh Hg concentrations in the MDD Riverindicated that consumption of carnivorous sh must be limitedin order to avoid excessive Hg body burdens. Residents living inSection 2 of the MDD River who consume two carnivorous shmeals weekly would likely exceed the United Nations Environ-ment Program (UNEP) provisional tolerable weekly intake(PTWI) of 1.6 mg per kg body weight (bw) for pregnant womenand the PTWI of 3.2 mg per kg bw for children and women ofchildbearing age (Fig. 6). In order to avoid exceeding the 1.6 mgper kg bw PTWI, pregnant women should consume fewer than 2carnivorous sh meals per week in Sections 2 and 3 (Fig. S11†).Children living in the region who consume two carnivorous shper week would have an approximate body burden of 5 mg per kgbw, which exceeds the 3.2 mg per kg bw. In contrast, childrenliving in communities near the headwaters (Section 1) canconsume nearly 5 carnivorous sh meals per week withoutexceeding health guidelines. With the exception of Section 2,children throughout the watershed can consume at least twonon-carnivorous sh meals.

The concentrations of Hg in carnivorous sh tissue may be apublic health risk in the region, especially in the downstreamregions occupied by Section 2 of the MDD River. Given the levelof Hg bioaccumulation in sh in MDD, it is imperative toquantify the Hg exposure for residents. The risk of dietaryHg exposure must also be balanced with the healthbenets of sh consumption, such as improved childhood

This journal is © The Royal Society of Chemistry 2015

Fig. 6 Weekly Hg body burdens based on consumption of fish fromthe MDD River and Lago Valencia, a lake that drains into the MDD Riverat km 545. The calculations assumed consumption rates of two fishper week for adults (70 kg) and children (30 kg). Thresholds representthe UNEP provisional tolerable weekly intake (PTWI) of 1.6 mg per kgbody weight (green dot-dashed line) for the protection of fetuses andembryos, two times the PTWI (orange-dashed line) for protection ofchildren and women of child-bearing age, and four times the PTWI(red line) to protect the general population from neurotoxicity effects.

Paper Environmental Science: Processes & Impacts

Publ

ishe

d on

19

Dec

embe

r 20

14. D

ownl

oade

d by

Duk

e U

nive

rsity

on

10/0

1/20

15 0

4:17

:18.

View Article Online

neurodevelopment.38 The US EPA recommends approximatelytwo sh meals per week for omega-3 fatty acids. In the MDDregion, consumption of non-carnivorous sh may providehealth benets while also minimize Hg exposure risk. However,risk from mercury consumption may still be present for chil-dren in communities located in Section 2 of the MDD River(Fig. 6). Future work must assess Hg exposure and healthoutcomes in the MDD region so that the health risks of dietaryHg exposure can be appropriately identied.

Conclusions

This study is the rst, to the best of our knowledge, to identify agradient of increasing Hg contamination and food web accu-mulation over a relatively broad geographical range whereASGM is prevalent. Rivers draining major ASGM mining areas,Huepetuhe and Guacamayo, are an important source ofmercury loading to the MDD River. Future research shouldspecically address mercury reactivity and transport fromASGM sites to determine how to best curb environmental andhuman health impacts.

While the direct human health risks of Hg from ASGMactivity are relatively well known, many studies focus closely ondirect Hg exposure to miners without examining exposure tonon-miners with distance or density of mining activity.2,39 It isclear from other studies that exposure during amalgamationand burning are extremely hazardous.3,9,10 This study is one ofthe rst to show signicant health risks to communities notdirectly involved in mining through sh consumption andmercury accumulation in downstream regions. Children livingwithin the central portion of the watershed cannot safelyconsume carnivorous sh without exceeding recommendedinternational Hg body burdens. The health risks of consumingcertain sh in the central and lower portions of the MDD

This journal is © The Royal Society of Chemistry 2015

watershed must be communicated to communities. Theregional health directorate, Direction Regional de Salud (DIR-ESA), is uniquely positioned to disseminate this information tolocal residents.

In the Madre de Dios region, deforestation and mercuryrelease are an immediate threat to both local and distantdownstream communities, many of which do not beneteconomically from ASGM. Policies are needed to simulta-neously decrease atmospheric Hg releases from gold amalgamburning, sediment mobilization from deforestation, and directreleases of Hg to the river from large areas of mining activity.Demand for gold has existed for millennia and is not likely todecrease in the near future. Realistic approaches are needed tobalance the need for wealth accumulation among some of thepoorest communities in the region and environmental stew-ardship, rather than the aggressive military efforts that attackindividual mining operations. Governments need to facilitatethe establishment of legal mining operations that integrate Hgcapture systems, environmental remediation, and healthmonitoring that enhances community-based cooperation andprovides residents in Madre de Dios a viable pathway towardsustainable human and environmental health.

Acknowledgements

This research was funded by the Duke Global Health Institute,the Pratt School of Engineering, the Bass Connections Program,and the Center for Latin American and Caribbean Studies atDuke University. Additional support was provided by the InterAmerican Institute for Global Change Research Grant#CRN3036. The local health directorate, Direccion Regional deSalud (DIRESA), Peruvian Navy, United States Naval MedicalResearch Unit (NAMRU-6), and Asociacion para la Conservacionde la Cuenca Amazonica (ACCA) provided invaluable logisticalsupport for this project. The Duke Global Health BassConnections Team in Peru assisted in eld collection of dataduring June/July 2013. We thank Crissel Vargas, Jaime Villamarand Cecilio Huamantupa for their assistance in eld datacollection, and Lauren Riedle, Laura Rogers, and Kaitlyn Porterfor their assistance with mercury analyses.

References

1 J. J. Swenson, C. E. Carter, J. C. Domec and C. I. Delgado,PLoS One, 2011, 6, e18875.

2 UNEP, Global Mercury Assessment 2013: Sources, Emissions,Releases and Environmental Transport, UNEP ChemicalsBranch, Geneva, Switzerland, 2013.

3 E. E. Yard, J. Horton, J. G. Schier, K. Caldwell, C. Sanchez,L. Lewis and C. Gastanaga, J. Med. Toxicol., 2012, 8, 441–448.

4 L. Maurice-Bourgoin, I. Quiroga, J. L. Guyot and O. Malm,Ambio, 1999, 28, 302–306.

5 H. Hsu-Kim, K. H. Kucharzyk, T. Zhang andM. A. Deshusses,Environ. Sci. Technol., 2013, 47, 2441–2456.

6 D. Mergler, H. A. Anderson, L. H. Chan, K. R. Mahaffey,M. Murray, M. Sakamoto and A. H. Stern, Ambio, 2007, 36,3–11.

Environ. Sci.: Processes Impacts

Environmental Science: Processes & Impacts Paper

Publ

ishe

d on

19

Dec

embe

r 20

14. D

ownl

oade

d by

Duk

e U

nive

rsity

on

10/0

1/20

15 0

4:17

:18.

View Article Online

7 A. Parlesak, D. Geelhoed and A. Robertson, Food Nutr. Bull.Suppl., 2014, 35, 191–199.

8 H. Ghattas, J. M. Barbour, M. Nord, R. Zurayk andN. R. Sahyoun, J. Nutr., 2013, 143, 1666–1671.

9 K. Ashe, PLoS One, 2012, 7, e33305.10 F. L. Barbieri, A. Cournil and J. Gardon, Int. J. Environ. Health

Res., 2009, 19, 267–277.11 N. Myers, R. A. Mittermeier, C. G. Mittermeier, G. A. da

Fonseca and J. Kent, Nature, 2000, 403, 853–858.12 G. P. Asner, W. Llactayo, R. Tupayachi and E. R. Luna, Proc.

Natl. Acad. Sci. U. S. A., 2013, 110, 18454–18459.13 B. Fraser, Environ. Sci. Technol., 2009, 43, 7162–7164.14 P. R. Mallen, in International Business Times, 2013, vol. 2014.15 S. Boyd, in The New Yorker, New York, 2013, vol. 2014.16 N. Miroff, in The Washington Post, 2014.17 A. L. Riscassi, K. J. Hokanson and T. M. Scanlon, Water, Air,

Soil Pollut., 2011, 220, 23–36.18 P. F. Schuster, J. B. Shanley, M. Marvin-Dipasquale,

M. M. Reddy, G. R. Aiken, D. A. Roth, H. E. Taylor,D. P. Krabbenho and J. F. DeWild, Water, Air, Soil Pollut.,2008, 187, 89–108.

19 J. P. Hurley, S. E. Cowell, M. M. Shafer and P. E. Hughes,Environ. Sci. Technol., 1998, 32, 1424–1432.

20 U. S. EPA, Method 7473: Mercury in Solids and Solutions byThermal Decomposition, Amalgamation, and AtomicAbsorption Spectrophotometry, Washington, DC, 1998.

21 J. W. Huckabee, J. W. Elwood and S. G. Hildebrand,Accumulation of mercury in freshwater biota, Elsevier/NorthHolland Biomedical Press, Amsterdam, 1979.

22 U. S. EPA, Method 1630: Methyl Mercury in water byDistillation, Aqueous Ethylation, Purge and Trap, and CVAFS,Washington, DC, 2001.

23 S. C. Hight and J. Cheng, Anal. Chim. Acta, 2006, 567, 160–172.

24 P. Krystek and R. Ritsema, Anal. Bioanal. Chem., 2005, 381,354–359.

Environ. Sci.: Processes Impacts

25 Y. Gong, D. P. Krabbenho, L. Ren, B. Egelandsdal andM. P. Richards, J. Agric. Food Chem., 2011, 59, 11050–11057.

26 M. T. Coe, E. M. Latrubesse, M. E. Ferreira and M. L. Amsler,Biogeochemistry, 2011, 105, 119–131.

27 T. Dunne, Water in environmental planning, W. H. Freeman,San Francisco, 1978.

28 G. P. Asner, G. V. Powell, J. Mascaro, D. E. Knapp, J. K. Clark,J. Jacobson, T. Kennedy-Bowdoin, A. Balaji, G. Paez-Acosta,E. Victoria, L. Secada, M. Valqui and R. F. Hughes, Proc.Natl. Acad. Sci. U. S. A., 2010, 107, 16738–16742.

29 U. S. EPA, Risk Assessment and Fish Consumption Limits, Officeof Science and Technology, Office of Water, Washington,DC, 2000.

30 P. J. Lechler, J. R. Miller, L. D. Lacerda, D. Vinson,J. C. Bonzongo, W. B. Lyons and J. J. Warwick, Sci. TotalEnviron., 2000, 260, 87–96.

31 J. Howard, M. A. Trotz, K. Thomas, E. Omisca, H. T. Chiu,T. Halide, F. Akiwumi, R. Michael and A. L. Stuart,Environ. Monit. Assess., 2011, 179, 555–573.

32 P. E. Ouboter, G. A. Landburg, J. H. Quik, J. H. Mol and F. vander Lugt, Ambio, 2012, 41, 873–882.

33 C. H. Gammons, D. G. Slotton, B. Gerbrandt, W. Weight,C. A. Young, R. L. McNearny, E. Camac, R. Calderon andH. Tapia, Sci. Total Environ., 2006, 368, 637–648.

34 K. A. Roach, N. F. Jacobsen, C. V. Fiorello, A. Stronza andK. O. Winemiller, J. Environ. Prot., 2013, 04, 51–60.

35 J. L. Guentzel, E. Portilla, K. M. Keith and E. O. Keith, Sci.Total Environ., 2007, 388, 316–324.

36 T. Zhang, K. H. Kucharzyk, B. Kim, M. A. Deshusses andH. Hsu-Kim, Environ. Sci. Technol., 2014, 48, 9133–9141.

37 H. Hintelmann, R. Harris, A. Heyes, J. P. Hurley, C. A. Kelly,D. P. Krabbenho, S. Lindberg, J. W. M. Rudd, K. J. Scott andV. L. St. Louis, Environ. Sci. Technol., 2002, 36, 5034–5040.

38 C. M. Klemens, E. L. Mozurkewich and K. Salari, in ClinicalLipidology, 2012, vol. 7, pp. 93–109.

39 H. Gibb and K. G. O'Leary, Environ. Health Perspect., 2014,122, 667–672.

This journal is © The Royal Society of Chemistry 2015