Fd

Ma

b

a

ARRAA

KAOSMA

1

smhaittgs

diwtaoq

ST

0d

Journal of Hazardous Materials 177 (2010) 1102–1111

Contents lists available at ScienceDirect

Journal of Hazardous Materials

journa l homepage: www.e lsev ier .com/ locate / jhazmat

ield rates for natural attenuation of arsenic in Tinto Santa Rosa acid minerainage (SW Spain)

aria P. Astaa,∗, Carlos Ayoraa, Patricia Acerob, Jordi Camaa

Institute of Environmental Assessment and Water Research (IDAEA), CSIC, C/Jordi Girona, 18-26, 08034 Barcelona, SpainPetrology and Geochemistry Area, Earth Sciences Department, C/Pedro Cerbuna 12, 50009 Zaragoza, Spain

r t i c l e i n f o

rticle history:eceived 30 November 2009eceived in revised form 4 January 2010ccepted 5 January 2010vailable online 13 January 2010

a b s t r a c t

Reactive transport modelling of the main processes related to the arsenic natural attenuation observed inthe acid mine drainage (AMD) impacted stream of Tinto Santa Rosa (SW Spain) was performed. Despitethe simplicity of the kinetic expressions used to deal with arsenic attenuation processes, the modelreproduced succesfully the major chemical trends observed along the acid discharge. Results indicatedthat the rate of ferrous iron oxidation was similar to the one obtained in earlier field studies in which

eywords:rsenicxidationorptionodelling

cid mine drainage

microbial catalysis is reported to occur. With regard to the scaled arsenic oxidation rate, it is one orderof magnitude faster than the values obtained under laboratory conditions suggesting the existence of acatalytic agent in the natural system. Schwertmannite precipitation rate, which was represented by asimple kinetic expression relying on Fe(III) and pH, was in the range calculated for other AMD impactedsites. Finally, the obtained distribution coefficients used for representing arsenic sorption onto Fe(III)precipitates were lower than those deduced from reported laboratory data. This discrepancy is attributed

ertm

to a decrease in the schw

. Introduction

Acid mine drainage (AMD), generated by sulphide oxidative dis-olution, may lead to high concentrations of dissolved metals andetalloids, including arsenic. Given its importance, much research

as been aimed at characterizing the arsenic sources in miningreas, the mechanisms controlling arsenic release and mobility andts natural attenuation [1]. These studies have led to the identifica-ion and description of the main geochemical processes that controlhe mobility and fate of arsenic in surface and ground waters. Aroup of studies have also focused on the complex interrelation-hips and couplings among these processes [2–6].

Since many of the cited processes are, moreover, strongly depen-ent on the hydrogeological processes present in each system, it

s not surprising that many recent works have approached them

ith the assistance of reactive transport modelling [7–9]. Reactive

ransport modelling is an essential tool for the quantitative evalu-tion of arsenic in the environment because the location and ratef reactions are controlled by both chemistry and transport. Suchuantitative models are indispensable to plan efficient remediation

∗ Corresponding author. Present address: Petrology and Geochemistry Area, Earthciences Department, C/Pedro Cerbuna 12, 50009 Zaragoza, Spain.el.: +34 976 761000x3159; fax: +34 976 761106.

E-mail address: mpasta@unizar.es (M.P. Asta).

304-3894/$ – see front matter © 2010 Elsevier B.V. All rights reserved.oi:10.1016/j.jhazmat.2010.01.034

annite arsenate sorption capacity as sulphate increases in the solution.© 2010 Elsevier B.V. All rights reserved.

strategies and to predict the evolution of water quality in rivers andreservoirs.

In line with this, one of the main problems encountered dur-ing the implementation of reactive transport models is related tothe choice of the rates for the involved geochemical processes.One possible approach to this problem consists in using laboratoryexperimental reaction rates and rate laws. However, the poorly-known scaling of such rates to adapt them to field sites representsa serious drawback to this approach. Therefore, it is of paramountimportance to obtain quantitative sets of field data to determinethe reaction rates and rate laws for the processes involved in themetal(loid)s transport and attenuation.

For these purposes, the waters and mineral precipitates from theacid discharge of the Tinto Santa Rosa abandoned mine (IberianPyritic Belt (IPB), SW Spain) were thoroughly studied and a 1D-reactive transport model, coupling advective flux and the maingeochemical processes controlling arsenic mobility, was devel-oped. This stream was chosen because of the simplicity of its flowsystem, and because its hydrogeochemical and mineralogical fea-tures are generally similar to many acid mine drainage streamsnot only in the Iberian Pyritic Belt [10–12] but also in other aciddrainages world-wide [13–15]. The implementation of the reactive

transport model applied to this thoroughly-characterized natu-ral system allowed evaluating the relative influence of the manygeochemical processes on the arsenic fate along a representativeAMD-discharge and exploring the scaling of reaction rates to fieldconditions.

dous M

2

2

cSca

dtofisodfmttapa

2

twwSs(bfr

bA0

M.P. Asta et al. / Journal of Hazar

. Materials and methods

.1. Field site and sampling description

All the water and sediment samples described in this study wereollected in March of 2007 and February of 2008 along the Tintoanta Rosa acid discharge (Fig. 1). The stream flows in a narrowhannel (1.5–2 m wide) over different terrace levels of yellowishnd reddish loose and crusty precipitates.

At each of the sampling points, pH, temperature, Eh and con-uctivity were measured in situ and three water subsamples wereaken in acid-pre-washed polyethylene bottles, after rinsing thor-ughly with filtered local water. Two of the subsamples wereltered through a 0.1-�m pore membrane filter. One of these sub-amples was acidified with HNO3 for elemental analyses and thether was acidified with HCl, adjusting its pH to less than 1, forissolved Fe(II)/Fe(III) determination. The third water subsample,or arsenic speciation determination, was preserved following the

ethod of Oliveira et al. [16] for AMD samples. It consists of elutinghe sample through a column filled with a cationic exchange resinhat reduces the load of metals that can act as potential oxidizinggents but without affecting the arsenic species. All the water sam-les were preserved in the dark at 4 ◦C until analyzed. Fe(II)/Fe(III)nd As(III)/As(V) were determined in less than 48 h since sampling.

.2. Analytical methods

Water pH was measured in the field using a Crison® glass elec-rode with automated temperature compensation after calibrationith standard buffer solutions of pH 2 and 7. Redox potentialas measured using a Pt combination electrode (ThermoOrion

ureFlow®) checked against solutions of 220 and 468 mV. The mea-urements were corrected to the Standard Hydrogen ElectrodeSHE). Electrical conductivity was measured with a Pt cell cali-rated with KCl 0.1 and 0.01 m solutions. The measurement errorsor pH, Eh and conductivity were ≤0.05 pH units, ≤5 mV and ±1%,

espectively.

Concentrations of major elements in solution were measuredy inductively coupled plasma atomic emission spectrometry (ICP-ES) using a PerkinElmer® Optima 3200 RL. Detection limits were.1 mg L−1 for Al, Si and S; 0.05 mg L−1 for Ca, Mg, 0.5 mg L−1 for

Fig. 1. Site map and sampling points in the acid

aterials 177 (2010) 1102–1111 1103

K, 0.025 mg L−1 for Fe, Zn, Cu and Mn and 2 mg L−1 for Na. Theerror was estimated to be below 3%. Concentrations of trace met-als (Ni, Cd, Co, As, Pb, Sb, Ti and V) were analyzed by inductivelycoupled plasma mass spectroscopy (ICP-MS) using a PerkinElmer®

Sciex Elan 6000 instrument. Detection limits were on the order of1 �g L−1 and the error was estimated to be below 5%.

Ferrous and total dissolved iron concentrations were deter-mined by colourimetry using the ferrozine method [17] in a UV–VISHP spectrophotometer. Fe(III) was taken as the difference betweenFe(tot) and Fe(II). The quality of the results was assured by measur-ing several standards, blanks and duplicates. Fe(tot) concentrationsmatched ICP-AES results within 10%.

Arsenic speciation from the AMD water samples was obtainedby high performance liquid chromatography-inductively coupledplasma mass spectroscopy (HPLC-ICP–MS). The microHPLC sys-tem consisted of an Agilent 1100 Series (Agilent, Waldbronn,Germany) binary pump and auto injector with a programmablesample loop (20 �L maximum). The separations were performedon a PRP-X100 (Hamilton, Reno, NV, USA) anion exchange col-umn (100 mm × 1 mm, 7 �m i.d.) and a phosphate buffer (H3PO412 mM) as a mobile phase (pH 3) at a flow rate of 80 �L min−1. Themicrobore column was connected directly to a Micromist nebu-lizer and a high-efficiency nebulizer (HEN) using its own capillary(480 mm × 0.25 mm and 480 mm × 0.10 mm, respectively). An HP4500 ICP-MS instrument (Yokogawa Analytical Systems, Tokyo,Japan) was used for the determinations. Detection limit for thearsenic species was <0.05 �g L−1. Concentrations matched ICP-MSresults within 15%.

3. Geochemical characterization of the Tinto Santa RosaAMD system

The water flow measured in the 2007 and 2008 campaignswas 0.7 and 1.1 L s−1, respectively. About 2/3 of the water comesfrom the mouth of an aerated adit and has large concentrationsof Fe and sulphate, with a noticeable content of minor elements

(TS1 in Table 1). The water Eh and conductivity values were inthe range of 510–645 mV and of 2480–4050 �S cm−1, respectively,and pH showed a variability ranging from 2.6 to 3.5. Seven metresdownstream from the adit mouth, the stream receives a ground-water discharge probably connected with flooded galleries (TS2 in

discharge of the Tinto Santa Rosa mine.

1104 M.P. Asta et al. / Journal of Hazardous MTa

ble

1Ph

ysic

alp

aram

eter

san

dh

ydro

geoc

hem

istr

yof

the

Tin

toSa

nta

Ros

ast

ream

wat

ers

du

rin

gth

esa

mp

lin

gca

mp

aign

s.

Sam

ple

(filt

ered

)T

(◦ C)

pH

mg

L−1�

gL−1

Ca

Mg

Na

KA

lSO

4Fe

TOT

Fe(I

II)

Fe(I

I)M

nSi

ZnC

uA

s TO

TA

s(V

)A

s(II

I)Pb

Ni

SbTi

VC

dC

o

Febr

uar

y20

08TS

118

.93.

4916

713

327

283

2,89

374

916

558

438

3769

171,

956

1,34

860

81

490

bdl

455

8285

8TS

218

.73.

5316

913

527

281

2,85

372

616

256

439

3768

152,

382

1,36

21,

020

387

493

44

5582

855

TS3

18.1

3.41

171

136

271

842,

861

722

188

534

3938

6917

1,35

196

238

912

949

2bd

l3

3084

855

TS4

15.3

3.37

171

136

271

842,

885

708

185

523

3938

6917

1,41

21,

064

348

127

486

bdl

333

8486

2TS

519

.13.

2617

213

827

186

2,87

668

319

548

839

3970

171,

145

855

290

127

512

bdl

325

8688

6TS

615

.23.

2717

113

526

184

2,86

666

420

446

038

3868

171,

064

783

261

118

490

bdl

324

8384

7TS

712

.53.

2517

213

626

285

2,89

966

621

345

339

3970

1799

170

628

511

850

2bd

l3

2384

863

TS8

14.9

3.16

170

136

272

862,

865

645

238

407

3938

6917

863

656

207

113

487

bdl

320

8283

6

Mar

ch20

07a

TS1

15.1

2.93

187

153

265

116

3,62

384

518

765

845

3783

461,

365

1,34

916

190

44

5410

214

61,

306

TS2

18.5

2.97

172

138

264

973,

024

694

151

543

3934

6833

2,73

11,

763

968

654

754

2442

7411

81,

078

TS3

12.0

2.95

194

157

274

118

3,56

377

924

753

245

4080

431,

401

1,19

620

521

183

67

4670

135

1,18

4TS

413

.42.

8118

314

926

411

23,

353

719

313

406

4339

7641

1,38

41,

271

113

214

895

749

7014

61,

277

TS5

12.9

2.78

195

159

274

119

3,56

375

942

733

246

4182

441,

149

1,11

929

196

827

542

5313

01,

170

TS6

13.3

2.76

188

154

263

115

3,50

371

734

037

744

4180

421,

155

1,12

627

202

869

547

5413

91,

243

TS7

12.6

2.79

190

155

273

116

3,44

371

134

736

444

4179

4294

894

8bd

l17

577

73

4042

123

1,10

4TS

813

.22.

6719

315

927

211

83,

563

708

372

336

4542

8143

953

953

bdl

191

856

348

4313

71,

214

bdl:

belo

wd

etec

tion

lim

it.

aD

ata

from

Ast

aet

al.[

20].

aterials 177 (2010) 1102–1111

Table 1) and that contains higher concentrations of arsenic and leadand larger proportions of As(III) than the water coming from theadit mouth. No water source other than the mine drainage (TS1 andTS2) was observed: the bed-stream is made up of low permeabilityschists, and no lateral tributary is present. Downstream from thesepoints, the pH decreases, which is accompanied by a systematicdecrease of ferrous and total iron concentrations. These changescan be reasonably attributed to the Fe(II) oxidation to Fe(III) and itssubsequent precipitation as ferric oxy-hydroxy-sulphates.

The evolution along the stream of some major dissolved ele-ments (Na, Mg, Ca, Al, Zn, Cu, Si, Mn and SO4) shows a constantconcentration (Table 1) due to their low reactivity during transportin aquatic systems at the measured pH range. Moreover, the con-stant concentration of these elements downstream points out thatno important evaporation takes place, and confirms the absence ofother water sources that could cause dilution.

Whereas the concentrations of other dissolved minor elementsdo not change significantly downstream (Cd and Co) or decreaseonly slightly (Ni and Pb), the concentrations of total arsenicdecreased sharply downstream (Table 1). This clear decrease indissolved arsenic concentrations is accompanied by a decrease ofAs(V) concentrations and, especially, of As(III) contents (Table 1).Since no other inflows were detected, this behaviour can beattributed to the oxidation of As(III) to As(V), which presents astronger affinity than As(III) for Fe(III)-precipitates [18,19] withinthe measured pH range and is partially sorbed onto them. The quan-tification and assessment of all these processes will be performedin the following sections.

Iron precipitates are extensively described in Asta et al. [20].Mineralogically, the solid phases collected in the stream bed consistof schwertmannite, goethite and jarosite. Schwertmannite, withideal formula Fe8O8(OH)5.5(SO4)1.25, is the dominant secondaryphase although its proportion decreases downstream, accompa-nied by an increase in goethite (FeOOH) and jarosite ((K, Na,H3O)Fe3(SO4)2(OH)6). The most concentrated element amongst theminor elements analyzed in the solids is arsenic (Table 2), whichdecreases downstream mirroring its behaviour in the stream waterconcentration. Whereas upstream this element is mainly linkedto schwertmannite, downstream it appears also associated withgoethite and jarosite [20]. The contents of the rest of elements (Zn,Cd, Ni, Sb, V, Co, Mn, Cu and Pb) in the iron precipitates are alsosignificant.

4. Model description

4.1. Conceptual model

The conceptual model of arsenic mobilization and attenuationincludes the following processes: (1) Fe(II) oxidation to Fe(III);(2) oxidation of As(III) to As(V); (3) Fe(III) precipitation as oxy-hydroxy-sulphates, causing a pH decrease; and (4) sorption of As(V)onto the solid Fe(III)-phases. The global overall reactions represent-ing these processes are shown in Table 3 (Eqs. (1)–(4)).

The rate of Fe(II) oxidation was implemented in the model byusing the expression of Singer and Stumm [21] for pH below 3.5:

rFe(II) = −d[Fe(II)]dt

= kFe[Fe(II)]fO2 (1)

where rFe(II) is the oxidation rate expressed in mol L−1 s−1, [Fe(II)]is the ferrous iron concentration (mol L−1) at a time t (s), fO2 is theoxygen fugacity and kFe (s−1) is the rate constant of the kinetic

expression. This reaction is known to be very slow at low pH [21]except in the presence of microorganisms, which increase the rateof Fe(III) production by up to six orders of magnitude.

As reported in many earlier works [22,23], the oxidation ofAs(III) to As(V) is very slow. However, the rate of oxidation of As(III)

M.P.A

staet

al./JournalofHazardous

Materials

177 (2010) 1102–11111105

Table 2Concentrations of major and minor constituents in the precipitates collected from the Tinto Santa Rosa stream bed (data from Asta et al. [20]).

Sample Water sample Major oxides (wt.%) Trace metals (mg kg−1)

Na2O MgO Al2O3 SO3 K2O CaO Fe2O3 As2O3 Cu Mn Zn As Cd Ni Sb V Pb Co Sr

TSR-M-1 TS1 0.03 0.02 9.55 8.92 0.01 0.05 83.76 4.51 192 70 118 34,200 bdl bdl 14 397 bdl 2 2TSR-M-2 TS7 bdl 0.38 12.28 12.94 0.02 0.58 84.68 0.16 840 1,120 1,496 1,176 72 64 2 84 92 26 5TSR-M 3 TS8 bdl 0.09 bdl 9.96 0.01 0.14 76.70 0.10 2,816 420 400 756 72 36 2 64 100 4 2

TSR-M-4 TS3 0.26 0.16 7.26 6.59 2.30 0.09 77.29 0.64 618 127 129 4,829 bdl 11 136 184 5,743 3 27TSR-M-5 TS6 0.25 0.11 7.74 8.16 1.39 0.07 82.57 0.48 1,176 115 602 3,616 15 5 27 69 6,535 3 23TSR-M-6 TS7 0.26 0.16 1.26 7.82 1.63 0.14 81.55 0.19 332 400 180 1,436 52 16 39 136 1,204 4 27

TSR-M-7-(0–3.5 cm)TS1

0.02 0.03 0.20 9.03 0.08 0.33 88.42 1.49 332 584 248 11,280 100 44 17 640 196 28 1TSR-M-7-(3.5–5.5 cm) bdl bdl bdl 8.40 0.10 0.05 87.49 2.91 140 460 128 22,040 72 bdl 26 820 128 2 2

TSRM-8-(0–0.5 cm)

TS3

bdl 0.01 1.40 8.25 0.17 0.22 88.03 1.88 168 476 148 14,240 70 bdl 9 220 372 1 4TSR-M-8-(0.5–4 cm) bdl bdl 0.72 9.73 0.31 0.05 86.12 1.30 168 360 132 9,880 72 12 71 320 352 2 6TSR-M-8-(4–4.5 cm) bdl 0.01 0.38 8.32 0.07 0.08 88.85 1.89 220 572 220 14,320 100 16 21 320 256 7 3TSR-M-8-(4.5–7 cm) bdl 0.05 0.00 8.50 0.10 0.13 89.10 1.10 340 528 368 8,320 112 bdl 7 220 264 5 3

TSR-M-9-(0–5 cm)TS6

bdl 0.03 1.18 8.82 0.26 0.09 88.97 0.37 200 456 160 2,784 88 16 15 180 480 2 5TSR M-9-(5–10 cm) bdl 0.02 bdl 7.79 0.25 0.12 89.98 0.48 280 472 180 3,628 80 bdl 8 148 448 2 5

TSR-M-10-(0–2 cm)TS8

bdl 0.01 0.27 8.93 0.17 0.10 89.42 0.43 140 48 180 3,236 92 bdl 8 136 332 2 4TSR-M 10-(2–4.5 cm) 0.61 bdl 2.26 7.43 0.09 0.11 90.90 0.31 180 404 168 2,340 64 bdl 3 96 28 2 3TSR-M-10-(4.5–7 cm) bdl 0.01 bdl 4.57 0.08 0.04 93.86 0.44 220 1,056 1,500 3,340 200 bdl 10 160 352 14 7

bdl: below detection limit. Numbers in brackets “( )” indicate the depth below the precipitate–water interface.

Table 3Main reactions and rate laws used in the model and rate constant values obtained for the Tinto Santa Rosa stream.

Eq. Reaction Rate expression (mol L−1 s−1) Model rate constant Reported and calculatedrate constant values

Mar 2007 February 2008

Fe(II) oxidation(1)Fe2+ + 0.25O2(aq) + H+ = Fe3+ + 0.5H2O rFe(II) = −d[Fe(II)]

dt= kFe[Fe(II)]fO2 10−3.55 10−3.75

10−3.44 (1)10−3.47 (2)10−3.9–10−5.2 (3)

As(III) oxidationa

(2) 2Fe3+ + H3AsO3 + H2O + h� = 2Fe2+ + H3AsO4 + 2H+ rAs(III) = −d[As(III)]dt

= kAs[As(III)][Fe(III)] 10−1.20 10−1.20 10−2.2–10−2.8 (3)

Schwertmannite precipitationb

(3) 8Fe3+ + 2.07SO42− + 11.86H2O = Fe8O8(SO4)2.07(OH)3.86 + 19.86H+ rSch = kSch[Fe(III)][H+]−1 10−8.0 10−8.5 10−8.10 (4)

As (V) sorptionc

(4) Schwertmannite + H3AsO4 + 0.24OH− = As-Schw + 0.62SO42− + 0.24H2O + H+ rAs(V) = rSchkd[As(V)] 103.42 103.90 104.5–105.0 (5)

“h�” is photons. (1) fO2 of 10−0.67 was included in the rate constant reported by McKnight and Bencala [36] but substracted in the rate constant value of the table for comparison purposes. (2) fO2 of 10−0.67 was included in therate constant reported by McKnight et al. [35] but substracted in the rate constant value of the table for comparison purposes. (3) Rate constant values obtained under laboratory conditions (unpublished data). (4) Calculatedfrom modelling the data of Acero et al. [25]. (5) Calculated from the data of Fukushi et al. [30].

a Reaction proposed by Emett and Khoe [24].b Calculated schwertmannite composition of Tinto Santo Rosa acid discharge.c Reaction proposed by Fukushi et al. [30]. kd = distribution coefficient.

1 dous M

ttTttmsf

r

waf

abcoittftsti[ofs

itp

r

wtrhttniui[i

r

wmc(ii

dhptASt

106 M.P. Asta et al. / Journal of Hazar

o As(V) by Fe(III) is increased by several orders of magnitude byhe presence of near-ultraviolet light because the reaction (Eq. (2);able 3) takes place in conjunction with the photochemical reduc-ion of Fe(III) [24]. Moreover, As(III) oxidation to As(V) is assumedo be catalyzed by the biotically-oxidized Fe(III) and, therefore,

icrobially-mediated redox processes affecting iron and arseniceem to be coupled. In order to take into account this coupling, theollowing equation is proposed:

As(III) = −d[As(III)]dt

= kAs[As(III)][Fe(III)] (2)

here kAs is an oxidation coefficient (L mol−1 s−1), calibratedgainst the field data; and [As(III)] and [Fe(III)] are the arsenite anderric iron concentrations, respectively (in mol L−1).

The oxidation of As(III) to As(V) plays a central role in arsenicttenuation, since the ferric precipitates that cover the streamed (mainly schwertmannite, goethite and jarosite) show a higherapacity to sorb As(V) than As(III) under the acidic pH conditionsbserved in the system. The usual ferric phase directly precipitatedn many sites of the IPB is schwertmannite [25,26]. The precipita-ion of goethite and jarosite in low-temperature systems is knowno be very slow [27] (weeks to months at room temperature). There-ore, although goethite and jarosite are present in the studied site,hese phases are considered to have formed by recrystallization ofchwertmannite. The schwertmannite ageing and transformationo goethite and/or jarosite may also last some months [25], and its still delayed by the presence of large amounts of sorbed arsenate13]. Given that the water residence time estimated in the stream isnly of a few hours, we reasonably assume that ferric iron removalrom water between the mixing point and TS8 (Fig. 1) is due tochwertmannite precipitation.

As regards the rate of schwertmannite precipitation, no exper-mental rates of schwertmannite precipitation have been found inhe literature, although it is reasonable to assume that this rate isroportional to supersaturation (see for example, Molson et al. [9]):

Sch = kSch(˝ − 1) (3)

here ˝ is the schwertmannite saturation, defined as the ratio ofhe ion activity product and the solubility product, and kSch is theate constant (mol L−1 s−1). However, this expression introduces aigh variability in the deduced rate constant values even thoughhis process was apparently constant in the field. This is due tohe high stoichiometric coefficients of the ions in the schwertman-ite precipitation reaction (Table 3). Thus, small variations in an

on concentration cause high variations in the ion activity prod-ct. In addition to that, several solubility products with differences

n orders of magnitude, have been proposed for schwertmannite28,29]. Therefore, a simple kinetic expression that relies on empir-cal concentrations was used:

Sch = kSch[Fe(III)][H+]−1 (4)

here rSch is the rate of schwertmannite precipitation (inol L−1 s−1); [Fe(III)] and [H+] are the ferric iron and proton con-

entrations (in mol L−1), respectively; and kSch is the rate constantin mol L−1 s−1). It is worth noting that the used expression is qual-tatively similar to that of Eq. (3), since the sulphate concentrations almost conservative (it can be integrated in the kSch value).

The last process included in the model was the arsenic decreaseue to As(V) sorption onto schwertmannite. This Fe(III)-oxy-ydroxy-sulphate is a reputed sink for As(V) in AMD due to sorption

rocesses [13,30]. Sorption of As(V) can occur through true adsorp-ion (or surface complexation) or through co-precipitation. Boths(V) uptake mechanisms seem to consist of the replacement ofO4 groups by AsO4 [13,30,31]. In the Tinto Santa Rosa model,he As(V) sorption rate has been assumed to depend mainly on

aterials 177 (2010) 1102–1111

the following three factors: (1) the availability of schwertman-nite (i.e. schwertmannite precipitation rate); (2) the concentrationof As(V); and (3) a distribution coefficient characteristic of thesorption/co-precipitation process. Therefore, the following simplekinetic expression, combining the effects of these three processes,was included in the model:

rAs(V) = rSchkd[As(V)] (5)

where rAs(V) is the As(V) sorption rate (mol L−1 s−1); rSch is the schw-ertmannite precipitation rate (mol L−1 s−1), as calculated by Eq. (4);[As(V)] is the arsenate concentration (in mol L−1) and kd representsthe partition coefficient for As(V) into schwertmannite.

4.2. Reactive transport model implementation

The one-dimensional (1D) reaction-transport model was car-ried out with the assistance of the PHREEQC code [32] and usingthe WATEQ4F thermodynamic database [33]. In the simulations, a1D path 300 m long and divided into 20 equal cells was used. Aflow rate of 0.7 and 1.1 L s−1 was considered, according to the ratemeasurements carried out in March 2007 and in February 2008,respectively. The total length of 300 m and the distances betweenthe different sampling points, located by GPS in the field, were cal-culated by means of ArcGIS 9 over the aerial orthoimage of theTinto Santa Rosa stream. The travelling time between points wasestimated with the flow rate and the stream section measured inthe field.

Model inputs included measured temperature, pH, pE and ele-mental concentrations. The water composition that resulted fromthe mixing water of the two contaminant sources (TS1 and TS2 fromTable 1 in a mixing ratio of 0.7TS1/0.3TS2) was used as the initialwater composition.

5. Results and discussion

Figs. 2–4 display the dissolved concentrations of total iron, Fe(II),Fe(III), total arsenic, As(III) and As(V) and the measured pH alongthe Tinto Santa Rosa stream, together with the results obtained inthe simulations for the same type of data.

The rate of Fe(II) oxidation was calculated from the decreasein the Fe(II) concentration between the TS1–TS2 mixing point andpoint TS8 (Fig. 1). The residence time of water estimated betweenthese points was 3.2 and 2.1 h, depending on the flow rate of thecorresponding campaign. The resulting Fe(II)-oxidation field rateranged from 3.4 × 10−7 to 4.3 × 10−7 mol L−1 s−1. These rates areroughly 5 orders of magnitude higher than the abiotic laboratoryvalue of 3.0 × 10−12 mol L−1 s−1, proposed by Singer and Stumm[34], which is clearly attributable to the catalyzing effect of bacte-ria under field conditions. The corresponding rate constant valuesfor Fe(II) oxidation are of 10−3.55 and 10−3.75 s−1. These kinetic con-stant values are only slightly higher than those obtained by usingthe same first-order kinetic expression (Eq. (1)) for biotic experi-ments carried out under laboratory conditions (10−3.9 to 10−5.2 s−1;unpublished data). Moreover, the rate constants for the Tinto SantaRosa system are very similar to the values obtained in earlier fieldworks [35,36], which adds robustness to the model and parametersused for Fe(II) oxidation in streams. After fixing the oxidation ratesof Fe(II), As(III) oxidation and Fe(III) precipitation rates were fitted.As done for Fe(II), the As(III) oxidation rate at the field scale wascalculated taking into account the decrease in As(III) concentra-tion from TS1–TS2 to TS8 and the residence time required (Fig. 4).

The obtained oxidation rate constant is 10−1.2 L mol−1 s−1, whichis around one order of magnitude faster than the one obtainedin the laboratory under atmospheric conditions (from 10−2.2 to10−2.8 L mol−1 s−1, unpublished data). As explained above, As(III)may be oxidized by Fe(III) in the presence of light (Eq. (2) in Table 3)

M.P. Asta et al. / Journal of Hazardous Materials 177 (2010) 1102–1111 1107

Fig. 2. Comparison between simulated and measured data of concentration of Fe(II),FFFt

aorbcd

FSF

e(III) and total iron from the Tinto Santa Rosa stream in the March of 2007 (a) andebruary 2008 (b) campaigns. Error bars correspond to the analytical error (10% fore(II) and Fe(III), 3% for total ifor total iron). First data at 0 m distance correspond tohe TS1–TS2 mixing point.

nd this oxidation path is considerably faster than that promoted

nly by dissolved oxygen. Therefore, the higher As(III) oxidationate observed in the field is attributed to the catalytic effect exertedy ferric iron and solar light, although some biocatalytic influenceannot be ruled out. Nonetheless, a marked decrease in the As(III)issolved concentration on February 2008 is observed in the upper

ig. 3. Comparison between simulated and measured data of pH from the Tintoanta Rosa stream. Error bars correspond to the analytical error of 0.05 pH units.irst data at 0 m distance correspond to the TS1–TS2 mixing point.

Fig. 4. Comparison between simulated and measured data for As(III) and As(V) con-centrations in the Tinto Santa Rosa stream in the March of 2007 (a) and February2008 (b) campaigns. Error bars correspond to the analytical error (15%). The firstdata at 0 m distance correspond to the TS1–TS2 mixing point.

reaches of the stream (between TS1–TS2 mixing point and TS3)and it is not totally reproduced by the model (Fig. 4b). This dis-crepancy might be attributed to diel differences in solar irradiationduring sampling, since ultraviolet light induces rapid photooxida-tion of As(III) to As(V) [37]. Another possible explanation is thathigh Fe(II)/Fe(III) ratios (as observed upstream) increase the pro-duction of hydroxyl radicals, therefore increasing the oxidationrate of As(III). However, the obtained data do not allow confirm-ing or discarding any of these hypotheses, which should certainlybe addressed in the future.

The Fe(III)-removal rate owing to schwertmannite precipitationwas calculated from the differences in the Fe(III)-concentrationalong the stream (Eq. (3); Table 3). The estimated schwertmanniteprecipitation rate was of 3.1 × 10−7 and 1.7 × 10−7 mol L−1 s−1 forthe data of March 2007 and February 2008, respectively. These ratevalues are in the range of 1.7 × 10−6 to 10−7 mol L−1 s−1 reported bySánchez-Espana et al. [26] using the differences of total dissolvediron for other AMD impacted sites in the Iberian Pyritic Belt (IPB).

The model reproduces reasonably the field data from the twosampling campaigns with precipitation rate constant values of10−8.0 and 10−8.5 mol L−1 s−1 (kSch in Eq. (4)). As a test of the signifi-cance of these values and given the absence of similar precipitationexpressions in the literature, our kinetic expression was used tomodel the data reported by Acero et al. [25] for the Cueva de laMora stream, which is located also in the Iberian Pyritic Belt. A pre-

−8.1 −1 −1

cipitation constant of 10 mol L s was obtained by couplingFe(II)-oxidation (with an oxidation rate constant of 10−3.65 s−1, inthe range proposed here for the Tinto Santa Rosa system) and schw-ertmannite precipitation, which is in good agreement with theprecipitation rate constant obtained in the Tinto Santa Rosa AMD.

1 dous M

udRmrrahhTgArfti

ptactA(qTfiatc3(o

sanw

Fpas

attenuation due to the precipitation, kFe(II) values higher than 10led to an anomalous pH increase in the upper stream reaches.This pH increase, not observed under field conditions, occurredas a result of the predominance of the oxidation processes (whichconsume protons) over precipitation (which generates protons).

108 M.P. Asta et al. / Journal of Hazar

The fitting between the simulated and measured pH values wassed as an independent check of the model implemented to repro-uce the observed iron concentration evolution in the Tinto Santaosa system. The observed evolution trend of pH is known to beainly controlled by the iron geochemistry in AMD systems. Iron

emoval from the dissolved phase demands the oxidation of fer-ous iron followed by hydrolysis of ferric iron and precipitations ferric oxy-hydroxy-sulphate mineral phases. All these processesave a clear influence on pH evolution; Fe(II) oxidation consumesydrogen ions whereas ferric iron precipitation releases protons.hus, the balance between the consumption and release of hydro-en ions is the main pH-determining control in this type of systems.s displayed in Fig. 3, the measured pH was suitably fitted by theeactive transport model without any additional calibration. There-ore, this is an independent evidence supporting the significance ofhe obtained rates for Fe(II) oxidation and schwertmannite precip-tation and the general soundness of the developed model.

Finally, the arsenic uptake processes by precipitating ferrichases were incorporated to the model by means of distribu-ion coefficients representing the partitioning of arsenic betweenqueous and solid phases owing to both potential adsorption ando-precipitation in association with schwertmannite. Kinetic sorp-ion of As(V) onto schwertmannite was modelled by incorporatings(V) as a minor constituent into precipitated schwertmannite (Eq.

4); Table 3). Distribution coefficients of 103.42 and 103.90 ade-uately predict the evolution of the measured As(V) concentration.hese values are lower than the coefficients that can be deducedrom the laboratory data obtained by Fukushi et al. [30] using sim-lar As(V) concentrations (distribution coefficients between 104.5

nd 105.0). Since arsenic sorption onto schwertmannite is thougho occur via sulphate–arsenate exchange [31], the large sulphateoncentrations in the Tinto Santa Rosa stream (from 2800 to500 mg L−1), much higher than in their laboratory experiments10–70 mg L−1), could be the cause of the lower uptake of arsenicbserved in our case.

As can also be observed in Fig. 4, the predicted As(V) evolution

lightly overestimates the field values in the lower reaches of thecid stream. This is possibly due to the fact that only schwertman-ite has been included as a precipitating ferric phase in the modelhereas also goethite and jarosite, with notably different arsenic

ig. 5. Simulated pH evolution for different rate constants of Fe(II) oxidation com-ared with measured data from the March 2008 campaign (except the first datumt 0 m distance, which corresponds to the TS1–TS2 mixing point). Error bars corre-pond to the pH measurement error, which is ±0.05 pH units.

aterials 177 (2010) 1102–1111

uptake capacities [38], are present (especially in those lower streamreaches).

6. Sensitivity analysis

In the following section, the sensitivity of model results as wellas the uncertainties associated with the major processes and thevalues of key parameters in the calibrated model are tested withthe 2008 data set.

The impact of uncertainty in the Fe(II)-oxidation rate constant(kFe) was tested by increasing and decreasing the rate constantby one order of magnitude. The increase of the kFe(II) leads to apH increase since the oxidation process consumes protons (Fig. 5).Nevertheless, the expected increase in pH is mitigated to an extentby increasing the schwertmannite precipitation rate. Despite this

−3.4

Fig. 6. Simulated evolution of Fe(II) (a) and Fe(III) (b) concentrations (mol L−1) fordifferent rate constants of Fe(II) oxidation compared with measured data from theMarch 2008 campaign (except the first data at 0 m distance, which correspond tothe TS1–TS2 mixing point). Error bars in concentrations correspond to the analyticalerror of 10%.

M.P. Asta et al. / Journal of Hazardous Materials 177 (2010) 1102–1111 1109

Fig. 7. Simulated evolution of As(III) (a) and As(V) (b) concentrations (mol L−1) fordifferent rate constants of arsenite oxidation compared with measured data fromtte

Ahfii1

tcuostOtdvaptc

Fig. 8. Simulated evolution of pH (a) and Fe(III) concentration (mol L−1) (b) for dif-ferent rate constants of schwertmannite precipitation compared with measured

he March 2008 campaign (except the first data at 0 m distance, which correspond tohe TS1–TS2 mixing point) Error bars in concentrations correspond to the analyticalrror of 15%.

s expected, high oxidation rates also lead to lower Fe(II) andigher Fe(III) contents in the simulations than those observed in theeld (Fig. 6a and b). According to the sensitivity results, an exper-

mentally supported range for the Fe(II) oxidation rate constant is0−3.7±0.2 for the Tinto Santa Rosa stream.

The influence of the arsenite oxidation rate constant (kAs) inhe calculations was tested by increasing and decreasing the rateonstant by one order of magnitude (Fig. 7). The results of the sim-lations showed that very high arsenite oxidation rates led notnly to As(III) concentrations much lower than measured in thetream but also to an anomalous increase in the As(V) contents inhe upper part of the stream, which was not observed in the field.n the other hand, as expected, lower oxidation rate constants led

o higher As(III) concentrations in the simulations than in the fieldata. According to the results, As(III) is more sensible than As(V) toariations in the oxidation rate constants (Fig. 7). Despite the poor

greement between the modelled and experimental points of thisarticular data set, the model adequately predicted the experimen-al points of 2007 data set. Acceptable values for the oxidation rateonstants are those ranged from 10−1.2±0.2.

data from the March 2008 campaign (except the first data at 0 m distance, whichcorrespond to the TS1–TS2 mixing point). Error bars correspond to the analyticalerror (10% for Fe(III) and 0.05 for pH).

As stated above, the changes in the pH value due to Fe(II) oxi-dation were mitigated by the Fe(III) precipitation that producesprotons. According to the simulation results, the main process thataffects the pH value is the schwertmannite precipitation, as it wasexpected since the process generates around 20 mol of protons foreach mol of schwertmannite precipitated (Fig. 8a). Experimentallysupported values for the precipitation constant are ranged within10−8.4±0.2 (Fig. 8a and b).

The sensitivity of the model to the arsenic distribution coef-ficient (kd in Eq. (5)) was checked by increasing and decreasingits value between 103.6 and 104.5 (104.5 corresponds to the distri-bution coefficient calculated from the Fukushi et al. [30] data). Asexpected, the use of larger distribution coefficients results in higher

sorption capacities of the bed-stream solids and, therefore, leads tolower As(V) concentrations than observed under field conditions(Fig. 9). On the contrary, the use of smaller distribution coefficientsproduces As(V) concentrations higher in the simulations than in

1110 M.P. Asta et al. / Journal of Hazardous M

Fig. 9. Simulated evolution of As(V) concentrations (mol L−1) using different distri-bdct

tc

7

egtsao

ooip

fwifirfi

aoowsbmfdtso

ta

[

[

[

[

[

[

ution coefficients (including that of Fukushi et al. [30]) compared with measuredata from the March 2008 campaign (except the first datum at 0 m distance, whichorresponds to the TS1–TS2 mixing point). Error bars in concentrations correspondo the analytical error of 15% for As(V).

he field. The experimentally supported range for this distributionoefficient is 103.95±0.1.

. Conclusions

A 1D reactive transport model coupling advective flux and sev-ral chemical reactions has proved to be useful in predicting theeochemical behaviour of an AMD stream. A remarkable innova-ion in this model is the incorporation of key chemical reactionsuch as iron and arsenic oxidation, precipitation of schwertmannitend arsenic sorption onto the newly precipitated schwertmanniten the stream bed.

Our geochemical modelling highlights the importance of ironxidation, which ultimately controls the geochemical changesbserved in the system. On the one hand, Fe(III) is capable of oxidiz-ng As(III) and, on the other hand, the precipitation of Fe(III) phaseslays a key role as a sink for As(V) due to sorption.

Despite the simplicity of the model, the simulations satis-actorily reproduced the main chemical features of the streamater, such as the pH decrease downstream and the accompany-

ng changes in As(III), As(V), Fe(II), and Fe(III) concentrations. Thetted kinetic constants for iron oxidation are comparable to thoseeported in earlier studies using similar rate law expressions undereld and laboratory biotic conditions.

The model calibration demonstrated the need for scaling thersenic oxidation rate obtained under laboratory conditions inrder to reproduce real field rates, which have shown to be up tone order of magnitude faster. The reason is not clear from ourork, but it could be due to the effect of biocatalysts, Fe(III) and/or

olar light in As(III) oxidation. The same need for field scaling haseen proved for the arsenate distribution coefficient onto schwert-annite, which have been observed to be lower than calculated

rom data obtained under laboratory conditions. This is probablyue to the fact that the effectiveness of the schwertmannite sorp-ion mechanism, which was suggested in earlier studies to occur viaulphate–arsenate exchange, is likely diminished by the presence

f high sulphate concentrations.

Finally, a schwertmannite precipitation rate law, calibratedo reproduce the iron evolution, has been used in the model andpplied to successfully reproduce not only the data from the two

[

[

aterials 177 (2010) 1102–1111

sampling campaigns presented in this study but also the trendsobserved at another independent AMD site in the Iberian PyriticBelt.

In light of the sensitivity analysis results, it can be concludedthat the calculations are very sensitive to the used rate constantsand, therefore, the values obtained for these factors are meaningful.

Acknowledgements

This research was funded by the project CTM2007-66724-C02-01/TECNO from the Spanish Government. MPA was financiallysupported by the Spanish Government with a PhD fellowship.PA was financially supported by the Spanish Government with aresearch contract from the “Juan de la Cierva” program. We arealso indebted to M. Caraballo, A.M. Sarmiento, C. Torrentó, R. Pérez-Lopez and J.M. Nieto for their help during the field campaign andto V. Ouro and A. Roig for analytical assistance. We are gratefulto Josep M. Soler, two anonymous reviewers and associate editorGerasimos Lyberatos for their insightful comments.

References

[1] H. Cheng, Y. Hu, J. Luo, B. Xu, J. Zhao, Geochemical processes controlling fateand transport of arsenic in acid mine drainage (AMD) and natural systems, J.Hazard. Mater. 165 (2008) 13–26.

[2] O. Sracek, P. Bhattacharya, G. Jacks, J.P. Gustafsson, M. Von Brömssen, Behav-ior of arsenic and geochemical modeling of arsenic enrichment in aqueousenvironments, Appl. Geochem. 19 (2004) 180–269.

[3] K.G. Stollenwerk, G.N. Breit, A.H. Welch, J.C. Yount, J.W. Whitney, A.L. Foster,M.N. Uddin, R.K. Majumder, N. Ahmed, Arsenic attenuation by oxidized aquifersediments in Bangladesh, Sci. Total Environ. 379 (2007) 133–150.

[4] M.U. Sharif, R.K. Davis, K.F. Steele, B. Kim, T.M. Kresse, J.A. Fazio, Inverse geo-chemical modeling of groundwater evolution with emphasis on arsenic in theMississippi River Valley alluvial aquifer, Arkansas (USA), J. Hydrol. 350 (2008)41–55.

[5] G. Cornelis, S. Poppe, T. Van Germen, E. Van den Broeck, M. Ceulemans, C. Van-decasteele, Geochemical modelling of arsenic and selenium leaching in alkalinewater treatment sludge from the production of non-ferrous metals, J. Hazard.Mater. 159 (2008) 271–279.

[6] P.G. Whitehead, D. Butterfield, A.J. Wade, Simulating metals and mine dis-charges in river basins using a new INtegrated CAtchment model for metals:pollution impacts and restoration strategies in the aries-mures river system inTransylvania, Romania, Hydrol. Res. 40 (2009) 323–346.

[7] R.L. Runkel, B.A. Kimball, Reactive solute transport in streams: a surface com-plexation approach for trace metal sorption, Water Resour. Res. 35 (1999)3829–3840.

[8] C. Le Guern, P. Baranger, C. Crouzet, F. Bodénan, P. Conil, Arsenic trapping by ironoxyhydroxides and carbonates at hydrothermal spring outlets, Appl. Geochem.18 (2003) 1313–1323.

[9] J. Molson, M. Aubertin, B. Bussière, M. Benzaazoua, Geochemical transport mod-elling of drainage from experimental mine tailings cells covered by capillarybarriers, Appl. Geochem. 23 (2008) 1–24.

10] J. Sánchez-Espana, E.L. Pamo, E. Santofimia, O. Aduvire, J. Reyes, D. Barettino,Acid mine drainage in the Iberian Pyrite Belt (Odiel River watershed, Huelva,SW Spain): geochemistry, mineralogy and environmental implications, Appl.Geochem. 20 (2005) 1320–1356.

11] J.M. Nieto, A.M. Sarmiento, M. Olías, C.R. Canovas, I. Riba, J. Kalman, T.A. Del-valls, Acid mine drainage pollution in the Tinto and Odiel rivers (Iberian PyriticBelt, SW Spain) and bioavailability of the transported metals to Huelva Estuary,Environ. Int. 33 (2007) 445–455.

12] A.M. Sarmiento, J.M. Nieto, C. Casiot, F. Elbaz-Poulichet, M. Egal, Inorganicarsenic speciation at river basin scales: The Tinto and Odiel Rivers in the IberianPyrite Belt, SW Spain, Environ. Pollut. 157 (2009) 1202–1209.

13] K. Fukushi, M. Sasaki, T. Sato, N. Yanase, H. Amano, H. Ikeda, A natural atten-uation of arsenic in drainage from an abandoned arsenic mine dump, Appl.Geochem. 18 (2003) 1267–1278.

14] C. Casiot, M. Leblanc, O. Bruneel, J.C. Personne, K. Koffi, F.O. Elbaz-Poulichet,Geochemical processes controlling the formation of As-rich waters withina tailings impoundment (Carnoulès, France), Aquat. Geochem. 9 (2003)273–290.

15] A.G. Gault, D.R. Cooke, A.T. Townsend, J.M. Charnock, D.A. Polya, Mechanismsof arsenic attenuation in acid mine drainage from Mount Bischoff, westernTasmania, Sci. Total Environ. 345 (2005) 219–228.

16] V. Oliveira, A.M. Sarmiento, J.L. Gómez-Ariza, J.M. Nieto, D. Sánchez-Rodas, Newpreservation method for inorganic arsenic speciation in acid mine drainagesamples, Talanta 69 (2006) 1182–1189.

17] T. To, K.D. Nordstrom, K. Cunningham, J. Ball, B. McCleskey, New method forthe direct determination of dissolved Fe(III) concentration in acid mine waters,Environ. Sci. Technol. 33 (1999) 807–813.

dous M

[

[

[

[

[

[

[

[

[

[[

[

[

[

[

[

[

[

[

[37] A.M. Sarmiento, V. Oliveira, J.L. Gómez-Ariza, J.M. Nieto, D. Sánchez-Rodas, Diel

M.P. Asta et al. / Journal of Hazar

18] G.A. Waychunas, B.A. Rea, C.C. Fuller, J.A. Davis, Surface-chemistry of ferrihy-drite 1. EXAFS studies of the geometry of coprecipitated and adsorbed arsenate,Geochim. Cosmochim. Acta 57 (1993) 2251–2269.

19] S. Dixit, J.G. Hering, Comparison of arsenic (V) and arsenic (III) sorption ontoiron oxide minerals: Implications for arsenic mobility, Environ. Sci. Technol. 37(2003) 4182–4189.

20] M.P. Asta, C. Ayora, G. Román-Ross, J. Cama, P. Acero, A.G. Gault, J.M. Charnock,F. Bardelli, Natural attenuation of arsenic in the Tinto Santa Rosa acid stream(Iberian yritic Belt, SW Spain): the role of iron precipitates, Chem. Geol. 271(2010) 1–12.

21] P.C. Singer, W. Stumm, Acidic mine drainage: the rate-limiting step, Science167 (1970) 1121–1123.

22] R.B. McCleskey, D.K. Nordstrom, A.S. Maest, Preservation of water samples forarsenic(III/V) determinations: an evaluation of the literature and new analyticalresults, Appl. Geochem. 19 (2004) 995–1009.

23] R.B. Johnston, P.C. Singer, Redox reactions in the Fe–As–O2 system, Chemo-sphere 69 (2007) 517–525.

24] M.T. Emett, G.H. Khoe, Photochemical oxidation of arsenic by oxygen and ironin acidic solutions, Water Res. 35 (2001) 649–656.

25] P. Acero, C. Ayora, C. Torrentó, J.M. Nieto, The role of trace elements duringschwertmannite precipitation and subsequent transformation into goethiteand jarosite, Geochim. Cosmochim. Acta 70 (2006) 4130–4139.

26] J. Sánchez-Espana, E. López Pamo, E. SantofimiaPastor, The oxidation of ferrous

iron in acidic mine effluents from the Iberian Pyrite Belt (Odiel Basin, Huelva,Spain): Field and laboratory rates, J. Geochem. Explor. 92 (2007) 120–132.

27] U. Schwertmann, R.M. Cornell, Iron Oxides in the Laboratory, New York, 1991.28] J.M. Bigham, U. Schwertmann, S.J. Traina, R.L. Windland, M. Wolf, Schwerman-

nite and the chemical modelling or iron in acid sulphate waters, Geochim.Cosmochim. Acta 12 (1996) 2111–2121.

[

aterials 177 (2010) 1102–1111 1111

29] J.Y. Yu, B. Heo, I.K. Choi, J.P. Cho, H.W. Chang, Apparent solubilities of schwert-mannite and ferrihydrite in natural stream waters polluted by mine drainage,Geochim. Cosmochim. Acta 63 (1999) 3407–3416.

30] K. Fukushi, T. Sato, N. Yanase, Solid-solution reactions in As(V) sorption byschwertmannite, Environ. Sci. Technol. 37 (2003) 3581–3586.

31] K. Fukushi, T. Sato, N. Yanase, J. Minato, H. Yamada, Arsenate sorption on schw-ertmannite, Am. Mineral. 89 (2004) 1728–1734.

32] D.L. Parkhurst, C.A.J. Appelo, User’s Guide to PHREEQC (Version 2), a computerprogram for speciation, batch reaction, one-dimensional transport, and inversegeochemical calculations, Water Resources Research Investigations Report 99-4259, 1999, p. 312.

33] J. Ball, D. Nordstrom, User’s manual for WATEQ4F with revised thermodynamicdatabase and test cases for calculating speciation of major, trace and redox ele-ments in natural waters, U.S. Geological Survey Water-Resources InvestigationReport 91-183, 1991.

34] P.C. Singer, W. Stumm, Kinetics of the oxidation of ferrous iron, in: 2nd Symp.Coal Mine Drainage Res., Pittsburgh, Pennsylvania, 1968, pp. 12–34.

35] D.M. McKnight, B.A. Kimball, K.E. Bencala, Iron photoreduction and oxidationin an acidic mountain stream, Science 240 (4852) (1988) 637–640.

36] D.M. McKnight, K.E. Bencala, Reactive iron transport in an acidic mountainstream in Summit County, Colorado: a hydrologic perspective, Geochim. Cos-mochim. Acta 53 (1989) 2225–2234.

cycles of arsenic speciation due to photooxidation in acid mine drainage fromthe Iberian Pyritic Belt (SW Spain), Chemosphere 66 (2007) 677–683.

38] M.P. Asta, J. Cama, M. Martínez, J. Giménez, Arsenic removal by goethite andjarosite in acidic conditions and its environmental implications, J. Hazard.Mater. 171 (2009) 965–972.