Applied

Applied Geochemistry 19 (2004) 1255–1293

Geochemistry

www.elsevier.com/locate/apgeochem

Natural organic matter in sedimentary basins and itsrelation to arsenic in anoxic ground water: the example

of West Bengal and its worldwide implications

J.M. McArthura,*, D.M. Banerjeeb, K.A. Hudson-Edwardsc, R. Mishrab,R. Purohitb, P. Ravenscroftd, A. Cronine, R.J. Howartha, A. Chatterjeef,

T. Talukderf, D. Lowryg, S. Houghtona, D.K. Chadhah

aDepartment of Earth Sciences, UCL, Gower St, London WC1E 6BT, UKbDepartment of Geology, University of Delhi, Chattra Marg, Delhi 110007, India

cSchool of Earth Science, Birkbeck, University of London, Malet St., London WC1E 7HX, UKdArcadis Geraghty & Miller International Inc., 2 Craven Court, Newmarket CB8 7FA, UK

eRobens Centre for Public and Environmental Health, University of Surrey, Guildford GU2 7XH, UKfCentral Ground Water Authority, Bidhan Nagar, Salt Lake City, Kolkata 700091, India

gDepartment of Geology, Royal Holloway University of London, Egham, Surrey TW20 0EX, UKhCentral Ground Water Authority, 18/11 Jamnagar House, Mansingh Road, Delhi 110011, India

Received 16 September 2002; accepted 16 February 2004

Editorial handling is by Fuge

Abstract

In order to investigate the mechanism of As release to anoxic ground water in alluvial aquifers, the authors sampled

ground waters from 3 piezometer nests, 79 shallow (<45 m) wells, and 6 deep (>80 m) wells, in an area 750 m by 450 m,

just north of Barasat, near Kolkata (Calcutta), in southern West Bengal. High concentrations of As (200–1180 lgL�1)

are accompanied by high concentrations of Fe (3–13.7 mgL�1) and PO4 (1–6.5 mgL�1). Ground water that is rich in

Mn (1–5.3 mgL�1) contains <50 lgL�1 of As. The composition of shallow ground water varies at the 100-m scale

laterally and the metre-scale vertically, with vertical gradients in As concentration reaching 200 lgL�1 m�1. The As is

supplied by reductive dissolution of FeOOH and release of the sorbed As to solution. The process is driven by natural

organic matter in peaty strata both within the aquifer sands and in the overlying confining unit. In well waters, thermo-

tolerant coliforms, a proxy for faecal contamination, are not present in high numbers (<10 cfu/100 ml in 85% of wells)

showing that faecally-derived organic matter does not enter the aquifer, does not drive reduction of FeOOH, and so

does not release As to ground water.

Arsenic concentrations are high (�50 lgL�1) where reduction of FeOOH is complete and its entire load of sorbed

As is released to solution, at which point the aquifer sediments become grey in colour as FeOOH vanishes. Where

reduction is incomplete, the sediments are brown in colour and resorption of As to residual FeOOH keeps As con-

centrations below 10 lgL�1 in the presence of dissolved Fe. Sorbed As released by reduction of Mn oxides does not

increase As in ground water because the As resorbs to FeOOH. High concentrations of As are common in alluvial

aquifers of the Bengal Basin arise because Himalayan erosion supplies immature sediments, with low surface-loadings

* Corresponding author. Fax: +44-20-7679-4166.

E-mail address: j.mcarthur@ucl.ac.uk (J.M. McArthur).

0883-2927/$ - see front matter � 2004 Published by Elsevier Ltd.

doi:10.1016/j.apgeochem.2004.02.001

1256 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

of FeOOH on mineral grains, to a depositional environment that is rich in organic mater so that complete reduction of

FeOOH is common.

� 2004 Published by Elsevier Ltd.

1. Introduction

In the Bengal Basin, some 25% of wells that tap

ground waters from alluvial aquifers contain high (>50

lgL�1) concentrations of As, the consumption of which

threatens the health of millions of people (e.g. Chatterjee

et al., 1995; Chakraborti et al., 2002): by, for example,

inducing cancers that may prove fatal (Smith et al.,

2000; Yu et al., 2003). One model for the release of As to

ground water, viz. reduction of sedimentary Fe oxides

(FeOOH without mineralogical connotations hereinaf-

ter) and release of the sorbed As, was outlined by Gulens

et al. (1979), adopted by Matisoff et al. (1982), and

systematised by Korte and Fernando (1991). The pro-

cess is driven by microbial reduction of organic matter

(Nealson, 1997; Lovley, 1997; Banfield et al., 1998;

Cummings et al., 1999; Chapelle, 2000; Lovely and

Anderson, 2000). This mechanism was used by Nickson

et al. (1998, 2000) to explain the occurrence of high

concentrations of As in aquifers in the Bengal Basin: it

was further developed by McArthur et al. (2001) and

Ravenscroft et al. (2001), who proposed that organic

carbon in buried peat was the driver of microbial re-

duction of FeOOH in the sediments, and that human

organic waste in latrines might also locally contribute to

As release by providing point-sources of organic matter

to enhance local reduction of FeOOH (thereby confus-

ing regional trends of As distribution by adding local

signals). These authors also discussed, and dismissed,

other mechanisms proposed to explain the release of As

to ground water, such as oxidation of sedimentary py-

rite, and competitive anion exchange of sorbed As with

inorganic PO4 from fertiliser; they thereby broke the

putative link between the development of irrigation with

ground water and the presence of As poison in ground

water.

Although the Fe-reduction model is now widely ac-

cepted as valid for the Bengal Basin (Bandyopadhyay,

2002; Dowling et al., 2002; Harvey et al., 2002; Anawar

et al., 2003; Tareq et al., 2003), the suggestion of

McArthur et al. (2001) and Ravenscroft et al. (2001) that

reduction is driven by buried peat has been disputed.

For example, Pal et al. (2002) found peat in an As-free

area, but did not find it in an As-affected area nearby.

Harvey et al. (2002) suggest that reduction of FeOOH is

driven by surface organic matter, from river beds,

ponds, and soils, that is drawn into the aquifer by irri-

gation drawdown: this model seeks to re-link irrigation

and high concentrations of As in ground water. Should

such a linking lead to renewed calls to curb or ban ir-

rigation with ground water, the livelihoods of the rural

poor will be affected adversely.

To better define the source of the organic matter

driving reduction of FeOOH in the Bengal Basin and

to further examine the behaviour of As and Fe in

ground water, the authors installed 3 piezometer nests

in a shallow (<45 m depth) aquifer near Barasat, in

West Bengal (Fig. 1); measured vertical profiles of

water composition in them; and took cores for study

from the piezometer sites. One piezometer was sited

where As poisoning of wells did not occur (site DP),

one was where it was thought to be minor (site AP),

and one where it was severe (site FP): the sites are less

than 350 m apart (Fig. 1). Wells tapping the shallow

(<45 m) and deep (>70 m) aquifers in the area were

also sampled in order to ascertain the lateral extent,

and degree, of changes in water quality, including As

pollution. The content of thermotolerant coliform

(TTC) bacteria in wells and surface ponds was mea-

sured in order to assess their content of faecally derived

micro-organisms, the presence of which indicates the

influence of human and animal organic waste. In order

to investigate whether the wells changed their compo-

sition with time, they were analysed (TTC apart) on 3

occasions in 2 years. The results of these studies are

presented here.

2. Location of area

The study area is located at 22� 44.430 N, 88� 29.450 E(Indian/Bangladesh datum) in the contiguous villages of

Joypur, Ardivok and Moyna (JAM hereinafter; Fig. 1)

in southern West Bengal. The area is just north of the

town of Barasat, Barasat-1 Block, 24 Parganas (North)

District, some 20 km NNE of Calcutta. The climate of

the area is tropical monsoonal with a dry season lasting

from December to May. The Hugli River flows from

north to south about 13 km to the west of the area, and

the Sunti River (dry in most months) runs north-south 2

km to the east, joining the east-west draining Sunti

Channel (also dry most months) at Mirathi, about 4 km

to the north of the area. The area looks flat but has a

regional slope to the north and east; i.e. towards the

rivers. Superimposed on this is a microtopography that

Fig. 1. Location of wells and piezometers in the study area on the northern outskirts of Barasat town, West Bengal. Waved areas are

ponds and rectangles represent some of the larger houses. Piezometer sites are double circles at AP, FP, DP. Arrowpoint in (b) shows

field area. Roads in C are double lines, prominent tracks are single lines. Unfilled circles are deep wells (>70 m depth), filled circles are

shallow wells (<45 m depth). N-34 is national highway 34.

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1257

influences local hydrology: relative relief of 2–3 m exists

where major roads are raised above ground level, and

where ponds, perched above the local water table by

several metres, have dry-season water levels up to 3 m

below ground level. These numerous ponds range in size

from <10 m diameter to around 100 m diameter, and are

excavated to a depth that preserves natural sealing by

the lower part of an aquitard that overlies the shallow

aquifer.

3. Methods

Piezometers consist of multiple, independent, wells

drilled to different depths and spaced no more than 2 m

apart. They were drilled using the reverse-circulation,

hand-percussion, method (Ali, 2003). Using this meth-

od, mud and clay samples are recovered with undis-

turbed cores. In sands, the samples are disturbed.

Separately at site AP (Fig. 1), percussive piston-sam-

pling into plastic core-tubes was alternated with rotary

wash-boring. The coring gave good recovery from pre-

cisely known depths in all strata. Cores were stored at

room temperature until sampled for analysis, as the local

groundwater temperature is 25 �C. A LECO C/S 125

Analyser was used to analyse the <0.5 mm fraction of

sediments for total S (TS), total C (TC), and total or-

ganic C (TOC), the last after leaching samples with 10%

v/v HCl. The difference between TC and TOC is re-

ported as a percentage of calcite. To obtain values of

d13C for calcite, samples were analysed with a Prism�mass spectrometer using a common acid bath and NBS

19, a calcium carbonate reference material supplied by

the National Bureau of Standards (now National Insti-

tute of Standards Technology).

Ground waters were sampled, after purging, in Feb-

ruary, 2002, November, 2002, and March 2003, from

hand-pumped domestic and public wells. The depth of

each well was recorded at the time of sampling by asking

the owner how many pipes were used to install it:

commonly, individuals buy 20-foot (6.1 m) lengths of

pipe and employ a driller to install them. The depths of

some wells are uncertain because of a change of owner,

or some other circumstance. Water samples drawn from

wells were not filtered: an accurate measurement was

sought of the amount of As in water consumed by well-

users, who do not filter their water. All samples were

crystal clear, i.e. without visible turbidity. Piezometers

fitted with 1.8 m screens were installed in March 2003 in

eastern JAM where As poisoning was absent (DP); in

south-eastern JAM where it was believed to be minor

(AP); and in northwestern JAM where it was severe (FP,

1258 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

Fig. 1). The level above ground of each well in each

piezometer nest, relative to an arbitrary level, was es-

tablished by theodolite survey. Water levels were mea-

sured in order to establish hydraulic heads and the

direction of ground-water flow. After extensive purging,

piezometers were sampled via hand-pumps between 5

and 10 days after installation. Piezometer samples were

filtered through 0.45 lm membrane filters. All samples

for cation analysis were acidified to pH 2 with HNO3.

Samples taken for anion analysis and for stable isotope

analysis were not acidified.

Ammonium and NO�3 were determined within 2

weeks of collection, and mostly much sooner. Mea-

surements of conductivity, and temperature were made

at the well head. Measurements of dissolved O2 are not

given as they are devalued by pump aeration. For

piezometers and some wells, alkalinity was measured in

the field by pH-titration with H2SO4 and is reported as

HCO3: for wells where this was not done, HCO3 has

been estimated by difference of charge balance. No well-

sample contained H2S that was detectable by smell after

acidification of the sample with H2SO4 and agitation.

Analysis of waters for cations, total-S (reported as SO4)

and total-P (reported as PO4) was by ICP-AES and

analysis for anions and NH4 was by ion chromatogra-

phy (including duplicate analysis for SO4). As a check

on ICP determinations of P, PO4 was also determined

spectophotometrically using the Mo-blue method.

Concentrations of As were measured on acidified sam-

ples using graphite-furnace AAS. In order to assess

particulate contributions to the dissolved load of well

water, the concentration of Al in waters was measured

and the turbidity of samples was recorded at collection.

All samples were crystal clear, i.e. without visible tur-

bidity, and concentrations of Al were <0.1 mgL�1, so it

was assumed that the analysis is unaffected by particu-

late contributions. Measurement of d13C of dissolved

inorganic C (DIC) was done with a Micromass Multi-

flow� preparation system on-line to an IsoPrime� mass

spectrometer. After He flushing of samples, and addition

of H3PO4, the isotopic composition of CO2 was mea-

sured after equilibration for 9 h at 30 �C. Drift correc-

tions were made using a reference-gas pulse that

bracketed each sample. Reference materials used were

an in-house carbonate and NBS-19. Every third sample

was accompanied by a standard, which was used to

make corrections for drift, which was assumed to be

linear; corrections were 6 0.4‰. Replicate analysis

showed that repeatability was 6 ±0.1‰ for solid ref-

erence materials and 6±0.2‰ for solutions made from

them. The average repeatability of the analysis was

�0.06‰ for d13C (n ¼ 134).

Microbiological analysis for thermotolerant coliform

bacteria was conducted in the field. The water was anal-

ysed, after purging, from 40 wells and 10 surface ponds in

JAM, and 14 wells in the As-free area of Mirathi, some

4 km to the north of JAM. The isolation and enumeration

of the TTC was carried out using membrane filtration

and growth on membranes with lauryl SO4 broth (Oxoid,

UK; Anon, 1994). The plates were incubated for 6 4 h at

ambient temperature to aid bacterial resuscitation, before

incubation at 44� 0.5 �C for between 16 and 18 h. Yellow

colonies were classified and counted as thermotolerant

coliforms. Sanitary risks were quantified for each well in

order to identify faecal sources, contaminant pathways

and other factors that might contribute to contamination

of wells (e.g. Lloyd and Bartram, 1991). Risk factors

comprise observable faults in well construction and the

proximity of sources of contamination, e.g. faecal mate-

rial, and the distance between wells and latrines.

4. Results

4.1. Sedimentology

A full description of the sediments and their com-

position will be presented elsewhere (Hudson-Edwards

et al., 2004), so only an outline is given here (Table 1).

Lithological logs of cored sediment were made on site

and are reproduced in Fig. 2 without change. An upper

aquitard of silts and silty clays (Fig. 2), of variable

thickness (9–24 m), overlies an unconsolidated sand

aquifer between 32 and 36 m in thickness (the shallow

aquifer): the aquifer is underlain by a clay aquitard

about 25 m thick, beneath which lies another sand

aquifer (the deep aquifer) that is >60 m thick.

The sediments were deposited between 22 and 1.2 ka

ago (14C and OSL ages from Hudson-Edwards et al.,

2004) by the Ganges/Brahamaputra Rivers and their

distributary channels. Dating of quartz in cores from the

AP site by the OSL method showed that the aquifer

sands are around 17 ka old below 32 m depth and <8 ka

old above it (Hudson-Edwards et al., 2004), which ex-

plains why the lower part of the aquifer is twice as hard

to core as the shallower levels, when judged by the

number of hammer blows required to penetrate 0.5 m of

sediment (around 90 below 32 m, around 50 above 32 m;

McArthur et al., unpublished data).

Correlations between the piezometer sites are shown

on Fig. 2 and are based on the abundance of calcite and

organic matter (TOC) in the sediments. The aquifer at

site DP, and the lower 13 m of the aquifer at sites AP

and FP, contains little or no calcite and almost no or-

ganic matter (<0.1%) and are therefore regarded as

being of equivalent age (around 17 ka years). In the

upper aquitard, the calcite-free level around 6–8 m depth

in both AP and DP is assumed to correlate with the

calcite-free level between 5 and 5.5 m in DP. Finally, in

AP and FP, the upper few decimetres of the lower

aquitard are brown, oxidised, and correlate (drilling at

DP did not reach the lower aquitard).

Table 1

Composition of sediment from piezometer sites in Joypur, Ardivok and Moyna (JAM), West Bengal, India

Site DP Site AP Site FP

m bgl IC % as Cal TS % TOC % m bgl IC % as Cal TS % TOC % m bgl IC % as Cal TS % TOC % d13C OM d13C IC

1.5 3.17 0.425 1.46 2.10 3.83 0.009 0.44 0.5 4.98 0.04 0.53 )22.3 )1.913.0 5.67 0.027 0.66 2.50 3.83 0.006 0.24 1.4 5.49 0.00 0.24 )21.2 )1.614.6 9.50 0.014 0.36 2.70 3.08 0.006 0.39 1.5 2.25 0.00 0.49

5.2 0.08 0.052 4.87 3.00 3.00 0.005 0.27 1.5 4.33 0.00 0.40

6.1 < 0.05 0.023 2.32 3.05 0.33 0.014 0.33 2.3 5.46 0.02 0.45

7.6 26.9 0.001 0.27 3.05 3.17 0.028 0.44 3.0 1.75 0.00 0.42

9.1 2.00 0.023 0.38 3.20 1.67 0.003 0.36 3.0 4.92 0.01 0.40

10.7 1.17 0.044 2.25 3.50 5.00 0.007 0.31 3.2 6.52 0.04 0.53

12.2 1.58 0.051 0.56 3.64 3.58 0.005 0.35 4.6 6.33 0.02 0.56

13.7 2.25 0.430 1.51 3.80 5.83 0.007 0.28 5.0 7.00 0.06 0.44

14.3 1.17 0.122 1.52 4.00 5.67 0.008 0.35 5.9 5.34 0.01 0.66

15.2 1.25 0.221 4.52 4.50 6.83 0.009 0.35 6.1 1.08 0.00 0.86

15.2 0.25 1.27 7.61 4.57 0.08 0.041 0.78 6.1 1.75 0.01 0.79

15.2 < 0.05 2.24 13.3 4.57 7.33 0.020 0.15 7.4 0.58 0.00 0.89

16.8 0.67 0.036 1.90 4.80 5.50 0.008 0.38 7.6 0.08 0.04 2.92

16.8 5.00 0.332 33.7 5.03 0.08 0.016 1.12 7.6 0.00 0.08 2.63

18.3 0.67 0.010 0.52 5.20 5.67 0.012 0.32 7.8 5.52 0.01 0.20 )20.9 )0.6919.8 6.50 0.006 0.29 5.50 4.25 0.021 0.39 9.1 5.67 0.06 0.62

19.8 7.75 0.000 0.20 5.57 6.58 0.007 0.39 9.1 3.83 0.01 0.58

21.3 0.08 0.000 0.16 5.70 6.08 0.011 0.39 9.1 6.17 0.03 0.42

21.3 0.33 0.001 0.13 5.86 0.25 0.010 0.78 9.5 4.99 0.00 0.72

22.9 0.08 0.007 0.28 5.92 3.33 0.009 0.67 10.7 5.33 0.05 0.58

24.4 < 0.05 0.007 0.10 6.00 0.25 0.029 0.61 11.3 5.57 0.00 0.35 )0.8825.9 0.17 0.007 0.07 6.00 0.42 0.119 0.72 12.2 3.67 0.01 0.60

27.4 0.08 0.012 0.10 6.03 0.25 0.015 0.60 12.2 3.83 0.01 0.78

29.0 0.08 0.007 0.09 6.10 0.33 0.018 0.70 13.3 4.75 0.00 0.17

30.5 0.17 0.006 0.08 6.10 0.42 0.039 1.58 14.2 5.73 0.01 0.17 )22.0 )2.0332.0 < 0.05 0.005 0.09 6.25 0.75 0.014 0.37 14.3 5.31 0.01 0.17

33.5 < 0.05 0.006 0.09 6.50 0.25 0.010 0.35 17.0 4.91 0.00 0.18

35.1 0.08 0.007 0.11 6.50 1.50 0.015 0.25 18.8 4.81 0.00 0.33 )21.4 )1.6236.6 0.08 0.005 0.07 6.75 0.67 0.020 0.53 20.8 3.18 0.00 0.14

38.1 < 0.05 0.008 0.10 7.00 0.58 0.040 0.44 22.6 3.78 0.00 0.13 )20.2 )1.5639.6 0.08 0.007 0.07 7.62 0.42 0.016 0.25 24.4 3.85 0.00 0.14

41.1 < 0.05 0.008 0.08 7.62 0.42 0.033 0.33 26.1 3.57 0.00 0.17

42.7 < 0.05 0.008 0.08 7.75 1.25 0.012 0.45 27.9 3.45 0.00 0.23

44.2 < 0.05 0.006 0.07 8.00 10.3 0.010 0.21 28.8 2.02 0.01 0.10 )23.6 )1.3745.7 < 0.05 0.005 0.05 8.50 1.58 0.028 0.35 29.5 2.71 0.62 7.20 )28.5 )9.7747.2 0.50 0.009 0.11 8.77 4.67 0.054 0.24 30.5 1.87 0.04 0.17 )25.0 )2.2748.8 0.83 0.007 0.06 9.14 0.17 0.023 0.33 31.6 0.87 0.00 0.18 )1.8050.3 < 0.05 0.006 0.11 9.14 0.75 0.065 0.78 33.6 0.65 0.00 0.12 )24.7 )3.9351.8 0.08 0.007 0.08 9.14 5.17 0.154 0.65 35.3 0.73 0.00 0.10 )2.5557.9 2.67 0.011 0.11 9.95 3.75 0.013 0.28 37.0 0.79 0.01 0.09 )24.3 )3.7059.4 0.08 0.006 0.10 11.35 4.42 0.014 0.22 39.0 0.79 0.02 0.08 )2.99

12.40 4.00 0.011 0.40 40.8 0.60 0.01 0.09 )5.2313.25 3.83 0.009 0.25 42.5 1.08 0.01 0.08 )4.57

J.M

.McA

rthuret

al./Applied

Geochem

istry19(2004)1255–1293

1259

Table 1 (continued)

Site DP Site AP Site FP

m bgl IC % as Cal TS % TOC % m bgl IC % as Cal TS % TOC % m bgl IC % as Cal TS % TOC % d13C OM d13C IC

13.50 4.25 0.005 0.16 44.4 2.09 0.01 0.06 )23.5 )3.7013.50 4.42 0.016 0.15 46.4 2.31 0.00 0.18 )2.6314.25 2.92 0.010 0.35 47.1 2.16 0.00 0.16 )3.0814.50 4.67 0.009 0.16 48.0 9.84 0.00 0.09 )4.1315.50 3.42 0.011 0.22 48.3 6.58 0.00 0.24

15.51 3.42 0.007 0.17 48.4 8.58 0.00 0.30

16.20 6.25 0.015 0.49 48.5 6.50 0.00 0.28

16.50 6.83 0.018 0.30 48.6 9.08 0.00 0.34

16.50 6.08 0.018 0.48 48.6 9.50 0.01 0.41

17.30 6.25 0.012 0.20 48.7 10.0 0.03 0.68

17.50 6.08 0.014 0.21 49.0 8.87 0.32 0.83 )24.9 )0.6118.50 5.00 0.008 0.24 50.0 9.79 0.28 0.72 )20.718.50 5.17 0.006 0.13 51.9 6.24 0.23 0.49

19.51 5.17 0.005 0.13 53.7 8.49 0.02 0.42 )22.9 )2.4720.30 4.17 0.009 0.32 55.5 9.02 1.21 1.57

20.50 4.83 0.006 0.16 57.2 8.16 1.33 1.48

21.15 5.25 0.013 0.42 59.0 8.84 1.09 1.30

21.45 4.75 0.009 0.52 60.8 9.53 0.50 0.69 )24.9 )0.8321.49 5.50 0.009 0.28 62.6 10.7 0.31 0.68

22.25 5.42 0.018 0.52 64.3 10.5 0.22 0.53

22.35 5.75 0.014 0.31 66.1 10.4 0.25 0.50

22.45 3.25 0.020 0.76 67.8 10.0 0.45 0.66 )0.4022.50 5.33 0.009 0.41 69.6 9.70 0.28 0.83

23.30 5.58 0.008 0.17

23.50 4.67 0.015 0.49

23.50 5.25 0.006 0.30

24.51 4.92 0.007 0.18

25.50 4.92 0.009 0.34

26.00 4.25 0.006 0.17

26.50 4.00 0.002 0.18

27.00 3.08 0.005 0.14

27.49 2.42 0.005 0.14

28.00 2.25 0.004 0.14

28.50 1.92 0.004 0.12

28.96 1.25 0.002 0.10

29.50 0.50 0.006 0.24

29.50 1.92 0.002 0.08

30.00 0.75 0.004 0.11

30.50 0.50 0.003 0.09

31.00 0.42 0.002 0.10

32.50 0.08 0.005 0.09

39.50 0.75 0.005 0.11

43.50 2.17 0.005 0.11

44.50 10.3 0.005 0.21

45.11 2.67 0.002 0.42

45.42 7.58 1.26 1.43

1260

J.M

.McA

rthuret

al./Applied

Geochem

istry19(2004)1255–1293

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1261

At site DP, in the east of the JAM area (Fig. 1), the

upper confining layer is 24 m thick, with the lower half

comprising impermeable clay (17–24 m). A calcite-free

interval occurs around 5 m depth. Above 17 m depth,

the confining layer contains abundant organic matter,

especially at depths of around 5, 15 and 17 m: the

deepest occurrence comprises peat sensu stricto with a

concentration of total organic C (TOC) of 34% (Table

1); the peat contained well-preserved tree-leaves. Sul-

phur concentrations in the OM-rich units reach 2.2% at

15.2 m depth, but is only 0.33% at 17 m depth. Calcrete

occurs at 19.8 m. Around 21 m depth, the clay is brown

and oxidised and shows in Fig. 2 as a minimum con-

centration of total sulfur: the oxidised unit passes gra-

dationally downward into the underlying unoxidised

grey clay. There is a sharp transition at 24 m into the

underlying aquifer sands, which are brown in colour

through the full thickness of the aquifer. The mean

concentration of TOC in the aquifer sands is 0.09%, with

no sample containing more than 0.11%, and the con-

centration of total S in the aquifer sand does not exceed

0.012%. At the greatest depth reached (59.4 m), cuttings

comprised brown sand mixed with grey silty clay, sug-

gesting that the lower confining clay was nearby. How-

ever, the poor sample quality at this depth made it

impossible to discover whether the underlying aquitard

had an oxidised upper surface as is the case at the other

piezometer sites.

At AP, in southern JAM, the upper confining layer

consists of silts and clayey silts. An interval free of cal-

cite occurs between 6 and 8 m depth. Around 9 m depth,

the silts grade downwards into fine sand in the upper

part of the aquifer (Fig. 2). The aquifer sands are grey in

colour above depths of 40 m, and brown in colour below

it. Organic-rich units exist in the upper confining silts

(1.6% TOC at 6 m depth; Table 1), and within the

aquifer sands at 17 m (0.49% TOC) and 23 m (0.76%

TOC). In aquifer sands, the mean concentration of TOC

is 0.27%. Concentrations of TOC appear to decrease

with depth below a depth of about 23 m, but an uncored

interval from 32.5 to 38.5 m depth prevents certainty in

this matter. Concentrations of total S reach 0.12% in the

upper confining silts but do not exceed 0.02% in the

aquifer sands. At a depth of 45 m, the aquifer sands give

way abruptly to the underlying clay of the basal aqui-

tard. The upper few decimetres of this aquitard are

brown, oxidised, and contain <0.1% TOC and <0.001%

of total S. The oxidised horizon grades downwards into

grey clay in which the concentrations of these constitu-

ents are higher. A similar oxidised upper aquitard occurs

at FP (see below): at both sites, the total S content of

these intervals (Fig. 2) is low over several metres of the

clay, suggesting that the oxidation is more pervasive that

it appears.

At FP, in central JAM, the sediments consist of silts

and clayey silts in the upper confining layer (0–12 m)

that grade into mostly medium sand in the main part of

the aquifer, which is 32 m thick. The upper aquitard is

free of calcite between depths of 6 and 8 m. The aquifer

sands are grey in colour, except at the base of the aquifer

(>40 m depth) where hints of brown occur in a general

grey colouration. Organic-rich units exist towards the

lower part of the upper confining silts (2.9% TOC at 8 m

depth; Table 1), and within the aquifer sands at 29.5 m

(7.2% TOC). A single sample at 22.9 m depth contains

0.71% TOC and may represent a lateral equivalent of the

organic-rich horizon at 23 m depth in AP. Elsewhere in

the sands, concentrations of TOC seldom exceed 0.3%

and are lower than this in the basal part of the aquifer.

Concentrations of total S reach 1.3% in the lower con-

fining clay and 0.6% in the peaty unit at 29.5 m, but do

not exceed 0.04% in the aquifer sands, nor do they ex-

ceed 0.06% in the upper aquitard. The underlying

aquitard of grey clay has been proven by drilling to be

25 m thick at this site. The upper part of this clay is

brown and oxidised, as it is at AP, and contains <0.1%

TOC and <0.001% of total S; it grades downwards into

grey clay in which concentrations of these constituents

are higher.

4.2. Piezometer waters

The composition of water from the 3 piezometers

(Figs. 3–5), which tap only the shallow aquifer, is pre-

sented in the order of increasing concentration of As in

the ground waters drawn from the aquifer (DP none; AP

some; FP much), which is in the order of increasing

concentration of organic matter in the aquifer sands (but

not in the overlying aquitard; Fig. 2). The piezometer

profiles reveal that the ground water is chemically

stratified (Figs. 3–5), as expected (Parker et al., 1982;

Karim et al., 1997; DPHE, 1999; McArthur et al., 2001).

In the piezometer at DP, concentrations of aqueous

As do not exceed 5 lgL�1 (Table 2; Fig. 3). Concen-

trations of Mn are high and peak in the middle of the

aquifer at 4.2 mgL�1, whilst concentrations of Fe show

an inverse trend to those of Mn, being lowest in mid-

aquifer (0.44 mgL�1 at 43.7 m) and highest (2.4 mgL�1)

in the lowest piezometer. Concentrations of PO4 are 0.35

mgL�1 at 25.8 m and decrease downwards. Concentra-

tions of SO4 are higher than at the other sites and in-

crease downward to be around 10 mgL�1 in the bottom

half of the aquifer. Values of d13C(DIC) are around

)2‰ in the upper part of the aquifer and )6‰ in the

lower part and show no obvious relation to other

parameters.

In ground water at AP from depths <40 m (Fig. 4),

concentrations of Fe (3.9–6.2 mgL�1), PO4 (0.8–3.2

mgL�1), and As (240–420 lgL�1) are high. The maxi-

mum concentrations occur at depths of around 38 m

(As), 26 m (Fe), and 21 m (P). The peak concentration

for Fe occurs just below the peak concentration of

Fig. 2. Lithological logs and profiles of total organic C (TOC, bold line), total S (TS, light line) and CaCO3 in sediment at the pie-

zometer sites DP, AP, FP, in the JAM area. Lithological logs are those made in the field and are reproduced without alteration. Lines

of correlation are shown dotted. Note the logarithmic scale for TS and TOC.

1262 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

organic matter (0.72% TOC) in the aquifer sands at 23 m

depth. At 26 m, the value of d13C(DIC) also decreases by

around 2‰, but overall values of between )2‰ and

)10‰ are less negative than might be expected for a

system so reducing. Towards the base of the aquifer

(from the piezometer at 38.3 m depth downwards),

concentrations of dissolved Fe, As, and PO4 decrease.

Concentrations of SO4 decrease downwards from the

uppermost piezometer, which is in the upper confining

layers, and are <10 mgL�1 below 32 m depth.

In the piezometer waters at FP, concentrations of Fe

are high (4–7 mgL�1, Fig. 5) over the entire aquifer

thickness, but are not related to those of dissolved As

nor to depth. Concentrations of As are around 430

lgL�1 in the upper part of the aquifer, but are much

higher in most of the lower part (>32 m depth), where

they are >900 lgL�1 in a considerable thickness of the

aquifer. Near the base of the aquifer (lowermost pie-

zometer at 43.4m depth), the As concentration (239 lgL�1) is lower, but is still higher than it is in the base of

the aquifer at AP (7 lgL�1). Values of d13C(DIC) range

from +0.6‰ to )10.8‰ but show no strong relation to

lithology, including the presence of units rich in organic

matter. Concentrations of SO4 decrease downwards

from the uppermost piezometer (in the upper confining

unit) and are <1 mgL�1 below 25 m depth. Concen-

trations of PO4 peak at 2.5 mgL�1 around 20 m depth

and decline downwards.

4.3. Ground water flow

In the aquifers of the Bengal Basin, the natural flow

regime combines 3 scales of flow-system, as described

by Ravenscroft (2003). The first, at the kilometre scale

in the shallow aquifer, is driven by local topography.

On this are superimposed local systems, around mic-

rotopographic features such as levees, channels and

ponds, that extend laterally for a few tens to perhaps a

hundred metres, and vertically for 10–20 m. The third

scale comprises natural flow to the south in the deeper

aquifer.

Water levels in the piezometers at the 3 sites in the

shallow aquifer, recorded in March 2003, are shown in

Fig. 6: each has a perched water table in the upper

aquitard. The head differences between the aquifer and

overlying aquitard are 0.32 m at site AP, 0.82 m at site

FP, and 3.70 m at site DP, and reflect the degree of

hydraulic connectivity between aquitard and aquifer.

Within the aquifer sands, piezometric levels vary by less

than 0.2 m and decrease downward, with the gradient

being greatest (0.0013) at site AP and lowest (0.0007) at

site DP. Ground water elevations in the lower part of the

shallow aquifer indicate a hydraulic gradient of 0.0006

to the NNE at the time of measurement – in the dry

season in early March, 2003. During the monsoon sea-

son, the direction reverses as local rivers to the north

and east fill with water (McArthur et al., unpublished

Fig. 3. Lithologic log and piezometer profiles of groundwater composition in March, 2003, at site DP, Barasat, West Bengal. Hor-

izontal dashed lines marks the boundary between the aquifer sands and the overlying confining unit: the base of the aquifer was not

definitely reached, so no lower boundary is drawn. Note the very much thicker upper confining unit, and the very much lower con-

centrations of As, compared to other sites (Figs. 4,5). Alkalinity is reported as HCO3.

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1263

piezometric data). Time-series measurements of hy-

draulic heads, and the hydrogeology of the area, will be

presented elsewhere (Hudson-Edwards et al., 2004).

4.4. Well waters

In Table 3 are given the results of the chemical and

microbiological analysis of the ground waters sampled

from domestic and public wells. The compositions are

typical of those found elsewhere in the Holocene sedi-

ments of the Bengal Basin (DPHE, 1999; McArthur

et al., 2001; Harvey et al., 2002; Dowling et al., 2003).

Shallow wells in the central and NW of JAM (Fig. 7) are

rich in As (200–1180 lgL�1), Fe (3–13.7 mg L�1), and

PO4 (1–6.5 mg L�1), and lack Mn and SO4. A group of

shallow wells (14–18, 40–44) in the immediate vicinity of

piezometer DP contain <10 lgL�1 of As and more than

2 mgL�1 of Mn. With few exceptions, high concentra-

tions of As are not spatially coincident with high con-

centrations of Mn (Fig. 7) nor with high concentrations

of SO4 (Fig. 7), and these parameters are not correlated

(Figs. 8(a) and (b)).

In cross-plots (Fig. 8) of species affected by redox

reactions (Fe, SO4) or Fe reduction (As, PO4, Fe), the

data for shallow wells (excluding wells near the pie-

zometer site DP, owing to their distinct sedimento-

logical setting; Fig. 2) appear to fall into 3 groupings,

on the basis of their As/Fe values, and their concen-

tration of As: those with <17 lg L�1 of As; those

with >32 lg L�1 of As and high As/Fe; and those

with >32 lgL�1 of As and low As/Fe. These 3 groups

of wells appear to bear a relation to depth, a relation

explained in a later section (Discussion). Wells with

high As/Fe also have high As/PO4 (Fig. 8(d)) and

distinctly less Fe and PO4 (Fig. 8(e)) than do wells

with low As/PO4 (Figs. 8(c)–(e)).

The ground waters are anoxic because they contain

dissolved Fe. Concentrations of nitrate are generally < 1

mg L�1 and the range in pH is small (6.76 to 7.32; Tables

2,3). Strong correlations exist between dissolved con-

centrations of Ca, Mg, and HCO3 (Figs. 9(a) and (b)).

Fig. 4. Lithologic log and piezometer profiles of groundwater composition in March, 2003, at Site AP, Barasat, West Bengal. Hor-

izontal dashed lines mark the boundaries of the aquifer sands with confining units. The uppermost piezometer is screened within the

silts of the upper confining unit. Alkalinity is reported as HCO3.

1264 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

Shallow wells have Na/Cl ratios that scatter widely

about the value for marine-salt (Fig. 9(c)) showing that

some wells have gained Na and some lost it, relative to

Cl. Many wells plot along a trend of increasing Ca and

Na (Fig. 9(d)), as sodic feldspar and detrital calcite are

weathered, but some plot to the right of this line, in a

direction compatible with ion-exchange of Ca in solu-

tion for Na on mineral ion-exchange sites (trend of line

B, Fig. 9(d)).

The deep ground waters are fresher (mean EC 753 lScm�1) than are shallow ground waters (mean EC 957

lS cm�1): they contain less Ca, Mg, Cl, and HCO3, but

more Na than most shallow wells (Fig. 9). Deep wells

(>70 m) have higher Na/Cl and Na/Ca values than do

shallow wells (Figs. 9(c) and (d)). Water from deep wells

contains <17 lgL�1 of As.

Saturation indices, calculated for waters from shal-

low wells with Geochemist’s Workbench (v. 3.0; Bethke,

1998), show that most of the Joypur waters are over-

saturated with calcite (SI +0.5 to +2.0), dolomite (SI

+0.1 to +2.0) and aragonite (SI +0.1 to +0.5). Water

samples with low concentrations of Fe are at equilibrium

or undersaturated with siderite (SI +0.2 to )0.9) and

vivianite (SI +0.3 to )3), and at equilibrium or over-

saturated with rhodochrosite (SI +0.02 to +0.7). By

contrast, samples high in Fe are saturated with vivianite

(SI +0.8 to +3.7) and siderite (SI +0.3 to +1.5) and

undersaturated with rhodochrosite (SI )0.3 to )1.3).Because Ba solubility is limited by the solubility of

barite, concentrations of Ba in a few well waters (Tables

2, 3) approach the guideline value of 0.7 mgL�1 (World

Health Organization, 1993) where concentrations of SO4

are low.

4.5. Ground water abstraction

Inhabitants draw water for domestic supply mostly

from private wells in the shallow aquifer, typically with

12-foot (3.6 m) screens that have their screen-centres at

depths around 25–45 m, but 6 wells in the study area tap

the deep aquifer (>70 m depth). The deep aquifer is also

exploited by 3 irrigation wells near the JAM area: one is

600 m to its east, a second is 1000 m to the north, and

the third is 200 m to its SW. The 3 deep wells irrigate 34

ha of agricultural land that borders the villages: the

shallow aquifer is not used much for irrigation. During

Fig. 5. Lithologic log and piezometer profiles of groundwater composition in March, 2003, at site FP, Barasat, West Bengal. Hori-

zontal dashed lines mark the boundaries of the aquifer sands with confining units. The uppermost piezometer is screened in the silts of

the upper confining unit. Alkalinity is reported as HCO3.

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1265

the development of both aquifers, abstractions must

have modified the natural ground water flow regime.

Nevertheless, natural flow at the kilometre scale appears

still to dominate the shallow aquifer, as irrigation water

is drawn from the deep aquifer only and shallow ab-

straction is but a small fraction of recharge. Deep ab-

straction for irrigation will induce regional down-flow

through the clay aquitard between 45 and 70 m, but its

thickness ensures that the effect is small. Human modi-

fications have applied, at most, for 40 a, whereas the

natural flow regime has existed in its present form for up

to 1.2 ka, since this is the age of OM in the upper

aquitards (Hudson-Edwards et al., 2004).

The deep irrigation wells to the north and east of the

JAM area are screened from 100 to 130 m and pump for

8 h a day in February and for 12 h a day in March and

April to irrigate about 27 ha of rice and 7 ha of vege-

tables. The third, sited 200 m to the SW of the area, is 95

m deep and has a comparatively small abstraction. No

abstraction volumes are available for these wells but,

assuming typical crop water requirements (Ravenscroft,

2003) and that deep percolation comprises 33% of ab-

straction, the authors estimate that the annual abstrac-

tion from these wells is of the order of 400,000 m3, with

around 130,000 m3 being returned to the shallow aquifer

as deep percolation losses. Because of the clay aquitard

at 45–70 m, the cone of depression created by these wells

must have spread widely in the deep aquifer as down-

ward leakage balances withdrawal of water from con-

fined storage. In JAM, there are about 100 hand

tubewells taking water from the shallow aquifer. By

assuming that each well is used by 10 people with an

average water use of 20 L/d, the annual abstraction from

the shallow aquifer is estimated to be of the order of

7300 m3, which is 2% of the abstraction from the irri-

gation wells.

5. Discussion

5.1. Piezometer sediments

The data and the correlations between the piezometer

sites (Fig. 2) suggest that the sediments below a depth of

32 m in FP and AP are equivalent in age, and of the

same age as the aquifer at DP; they also suggest that the

sediment above 32 m in FP and AP is distinctly younger

(<8 ka) than underlying sediment, which is at least 17 ka

Table 2

Composition and hydraulic heads of ground waters from piezometers in Joypur, Ardivok and Moyna (JAM), West Bengal, India

Depth

mbgl

Water

level

Temp.

�CEC lScm�1@

25 �C

pH As

lg l�1

Units are mg l�1 unless specified otherwise d13CDIC

‰

Bal

%Na K Ca Mg Sr Ba Fe Mn NH4 HCO3 Cl SO4 NO3 PO4 H4SiO4

Site FP: location 22 deg, 44.500 min N, 88 deg, 29.430 min E

6.9 94.45 26.7 1210 6.54 35 37.5 10.3 164 35.1 0.49 0.37 4.0 1.0 579 89.7 65.5 0.23 0.54 52 )8.4 )0.714.5 93.62 1018 7.21 94 29.7 10.3 135 34.2 0.50 0.33 4.7 0.8 562 58.0 33.2 0.53 0.72 47 )10.8 )1.019.4 93.58 26.7 764 6.84 443 21.1 2.2 107 25.4 0.43 0.23 7.4 0.1 3.50 487 20.9 0.7 0.33 2.50 70 0.6 1.7

22.4 93.57 26.7 795 6.96 411 18.0 2.8 111 30.7 0.50 0.26 4.8 0.2 2.30 526 26.5 1.5 0.40 1.57 66 )0.6 )0.927.9 26.7 975 6.96 447 20.0 3.6 138 37.4 0.63 0.22 6.0 0.2 2.30 601 52.4 0.4 0.18 1.84 61 )3.5 )0.231.2 93.58 26.8 1022 6.96 448 19.1 3.1 137 40.1 0.64 0.13 7.5 0.2 2.60 601 78.4 0.6 0.49 1.21 59 )4.3 )2.537.0 93.56 26.9 1028 7.10 960 25.2 3.4 130 45.3 0.74 0.11 5.2 0.2 5.00 584 72.4 0.4 0.56 1.05 51 )5.0 0.9

39.8 93.57 26.8 1024 7.08 958 26.3 3.1 131 46.6 0.84 0.17 4.7 0.2 1.25 598 63.3 0.4 0.17 1.08 57 )6.2 1.0

43.4 93.57 26.7 1139 7.04 239 39.6 3.1 138 47.7 0.84 0.28 4.8 0.1 1.30 611 94.2 0.1 0.47 0.88 68 )4.7 0.8

Site DP: location 22 deg, 44.511 min N, 88 deg, 29.561 min E

14.3 97.22

25.8 93.52 26.1 1142 6.88 4 76.8 0.9 146 27.5 0.44 0.19 1.8 1.6 0.08 744 51.2 0.4 0.58 0.35 58 )2.1 )2.231.9 93.54 26.3 1175 6.84 1 50.2 1.0 154 38.6 0.55 0.07 1.5 4.2 0.02 724 62.6 1.5 0.31 0.17 70 )1.9 )1.337.6 93.52 1048 7.08 1 53.6 0.6 139 37.3 0.53 0.08 0.6 2.7 0.13 670 58.8 5.4 1.32 0.09 77 )6.8 )1.243.7 93.52 26.3 1031 7.00 <1 55.6 1.3 129 33.8 0.51 0.09 0.4 2.6 0.12 640 40.2 11.5 0.41 0.02 75 )6.4 )0.250.1 93.53 26.3 999 6.96 1 47.6 1.6 118 38.8 0.48 0.09 0.5 2.1 0.45 606 47.6 6.3 0.19 0.04 65 )6.4 )0.555.6 93.51 26.4 1131 6.95 4 48.2 2.4 151 42.9 0.66 0.14 2.4 0.5 0.17 728 80.4 9.7 4.56 0.05 71 )5.5 )4.0

Site AP: location 22 deg, 44.347 min N, 88 deg, 29.505 min E

6.9 94.13 26.7 3152 6.70 135 250 3.2 364 87.2 0.97 0.31 2.0 4.1 701 558 490.3 0.12 0.24 54 )6.8 )1.214.3 93.81 26.5 1406 6.78 327 87.6 1.4 168 39.0 0.49 0.40 5.2 0.8 1.30 690 100 98.5 0.13 0.82 62 )7.3 )1.420.6 93.72 26.3 1307 7.04 258 58.8 2.1 186 51.4 0.83 0.39 5.3 0.3 1.30 670 74.1 80.1 0.18 3.16 64 )7.4 5.4

25.9 93.71 26.1 1075 7.04 226 42.1 2.3 143 39.6 0.67 0.37 6.2 0.2 1.95 562 66.8 53.0 0.36 3.12 67 )9.8 1.8

31.9 93.70 26.3 925 6.96 279 28.4 1.5 122 37.3 0.62 0.28 4.4 0.2 1.55 551 65.2 11.9 0.30 3.00 69 )6.6 )1.938.3 26.5 757 7.07 419 23.3 1.6 109 33.3 0.55 0.15 3.9 0.2 1.10 526 8.1 2.8 0.12 2.25 66 )2.5 2.8

41.0 26.5 859 6.91 187 33.4 1.4 112 35.3 0.56 0.09 2.0 0.3 1.30 579 13.4 2.4 0.35 1.15 65 1.3

41.8 26.4 919 6.88 110 34.4 1.5 116 36.3 0.57 0.09 1.5 0.5 1.50 593 30.0 8.4 0.16 0.49 68 )1.142.8 93.69 26.5 939 7.07 7 44.7 2.4 115 38.0 0.55 0.14 1.0 0.6 1.30 622 29.3 3.9 0.22 0.13 64 )3.6 )0.5

Pond A 1 20 8.8 11.7 18 6.1 0.07 0.04 0.3 0.24 10.1 6.0 0.35 1.5 30

Pond A 2 17 16.1 11.2 29 9.4 0.11 0.04 0.0 0.16 16.9 4.6 0.35 0.5 25

Rainwater 12 4.3 1.0 12 2.2 0.08 0.04 4.0 6.0 0.7 5

Notes: depth mbgl is depth in metres to mid-screen of each piezometer, Water levels are m above an arbitrary datum.

1266

J.M

.McA

rthuret

al./Applied

Geochem

istry19(2004)1255–1293

0

10

20

30

40

50

6093.4 93.6 93.8 94.0 94.2 94.4 94.6

Water level (m above datum)

Dep

th m

AP

DP

FP

97.22

Fig. 6. Water levels in March, 2003, piezometers at sites DP, AP, and FP, in metres relative to an arbitrary datum. At each site, the

uppermost piezometer was completed close to the base of the upper confining unit. The higher head in the upper confining unit at Site

DP arises from the impermeable clay unit that locally caps the aquifer at this site. At FP and AP, the aquifer sands are capped by more

permeable silts.

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1267

old. The lack of calcite in sediment at DP, and below 32

m depth at AP and FP, might be due to its removal

during oxidative weathering for a period of around 9 ka

prior to the deposition of overlying sediment, or to the

deposition of the calcite-poor units by the Brahamapu-

tra River: today, sediment of the Ganges River contains

detrital calcite but sediment of the Brahamaputra River

does not. This contrast is best seen by examining its

reflection in the Sr content of ground water in Bangla-

desh (P. Ravenscroft; Maps of Ground-Water Composi-

tion for Bangladesh, http://www/es.ucl.ac.uk/research/

lag/as): where detrital calcite occurs, concentrations of

Sr in ground water are high (>0.3 mgL�1); where calcite

is absent, concentrations of Sr are low (<0.3 mgL�1).

At DP, the aquifer sediments are brown over the

entire depth of the aquifer. At AP, the aquifer sands are

grey in colour (Fig. 10) above 40 m depth, and brown at

greater depth. At FP, the sediments are grey throughout

the full aquifer thickness (Fig. 10) but, at the very base

of the aquifer (>40m depth), the grey colour is accom-

panied by hints of brown at some levels (44.5 m). The

upper few decimetres of the underlying aquiclude at AP

and FP is also brown in colour and oxidised. As others

have done (van Geen et al., 2003a,b), the authors assume

that the brown colour arises from the presence of Fe-

OOH and that the grey colour occurs when these oxides

have been destroyed (or given a sequestering surface

coating) by reduction. Attempts to confirm this as-

sumption, and to quantify the cause of the colour dif-

ference, using sequential chemical leaches were

unsuccessful owing to uncertainty about the homoge-

neity, solubility, grain-size, and crystallinity, of likely

reduced phases and of FeOOH. Further work on this

issue will be reported elsewhere (Hudson-Edwards et al.,

2004). Nevertheless, the colour difference does suggest

an explanation for why concentrations of dissolved As

and Fe seldom correlate well in ground water from al-

luvial aquifers where FeOOH-reduction is the dominant

mechanism of As release: it is that most of the Fe(II)

that is reduced is sequestered in solid phases, such as;

siderite (Pal et al., 2002), Fe sulphide (Korte and Fer-

nando, 1991; Rittle et al., 1995; Chakraborti et al., 2001)

vivianite, magnetite (Harvey et al., 2002), and green rust

(Frederickson et al., 1998). That not all reduced Fe stays

in solution is clear from the fact that the highest con-

centration of Fe in the waters (13.7 mgL�1; Table 3)

requires the dissolution of only about 5 mg kg�1 of solid

FeOOH. Although the amount of FeOOH in the sedi-

ment is low (see later), it is not that low: the reduction of

FeOOH must, therefore, result largely in the production

of solid phases containing Fe(II), with only a small

fraction of the Fe remaining in solution. The notable

difficulty, then, in explaining the composition of ground

water in alluvial basins is to account for those instances

(as found in JAM; see later) where dissolved Fe and As

do correlate: the expectation is that they should not.

The brown colour occurs distally from the sources of

organic matter, which are in the upper aquifer and

overlying aquitard. At AP, only the base of the aquifer is

brown (i.e. contains FeOOH), a level well below that

where most of the organic matter is found. At site DP,

the entire aquifer is brown, yet has abundant organic

matter in the unit above it. An examination of that unit

explains why: the upper confining layer is 24 m thick,

but the lower half comprises impermeable clay (15–24

m) which, despite its high content of organic matter

between depths of 15.2 and 16.8 m (Table 1, Fig. 2),

prevents hydraulic connectivity being attained between

Table 3Chemical composition of, and content of thermotolerant coliform bacteria in, ground waters from water wells in Joypur, Ardivok and Moyna (JAM), West Bengal, India. Some TTC determinations werereplicated, mostly on different days

JAM, Shallow AquiferBa 1b 22 44.5097 88 29.4413 33.5 26.3 1087 7.18 361 30.3 3.3 133 45.1 0.72 0.04 8.8 0.15 594 557 83.0 0.0 0.3 1.7 48 -7.6 2.5Ba 2 22 44.5430 88 29.4145 8 0, 0, 1, 10 39.6 26.5 949 7.01 1180 25.3 3.3 120 36.1 0.66 0.12 6.3 0.06 2.9 533 513 64.7 0.0 0.2 1.5 59 -4.8 1.4Ba 3 22 44.5439 88 29.4295 11 0 18.3 26.4 1045 7.12 1090 26.4 3.5 128 40.1 0.72 0.16 7.0 0.05 2.8 547 565 83.5 0.0 0.3 1.3 58 -5.3 -1.3

New Ba 4 22 44.4748 88 29.4515 12 0 45.7 26.6 909 6.98 131 21.6 2.1 114 31.5 0.58 3.4 0.52 0.5 475 58.4 0.6 0.7 71Old Ba 4 22 44.4748 88 29.4515 30.5 26.2 1160 7.03 626 36.5 2.9 144 42.5 0.71 0.42 11.3 0.06 622 635 91.3 0.0 0.1 65 -6.9 -0.9

Ba 6 22 44.4748 88 29.3215 11 7 39.6 26.5 685 7.12 96 13.4 1.4 91 28.6 0.44 0.31 4.7 0.05 1.0 430 423 15.8 11.2 0.1 2.3 67 -10.1 0.6Ba 7 22 44.4830 88 29.2765 7 1 45.7 26.6 911 7.02 317 47.6 3.5 100 34.5 0.55 0.32 3.4 0.02 0.3 572 581 27.8 0.1 0.5 2.3 59 -7.9 -0.9Ba 8 22 44.5041 88 29.2965 10 0 39.6 26.9 937 7.04 310 47.5 4.5 105 33.8 0.54 0.35 4.6 0.14 0.2 582 583 30.2 0.3 0.4 2.1 60 -3.6 -0.2Ba 9 22 44.5067 88 29.3205 11 0 39.6 27.7 672 7.10 163 17.2 2.4 86 24.1 0.41 0.32 7.1 0.07 1.5 414 407 23.5 0.1 0.2 3.1 98 -9.0 0.6

Ba 10 22 44.4818 88 29.4979 3 0 42.7 26.2 1009 6.99 < 1 25.4 2.3 127 37.9 0.57 0.10 0.2 2.25 0.9 485 487 101 0.0 0.1 0.3 72 -6.6 -0.2Ba 11 22 44.5007 88 29.4745 8 1 26.0 1197 6.88 391 45.7 2.5 160 35.0 0.60 0.34 8.1 0.05 3.5 680 682 80.0 0.3 0.2 1.8 67 0.9 -0.2Ba 12 22 44.5300 88 29.4766 33.5 26.1 1107 7.00 459 46.8 1.7 132 37.5 0.59 0.10 13.7 0.14 625 658 71.4 0.1 0.4 53 -7.9 -2.2Ba 13 22 44.4673 88 29.5605 7 0 44.5 26.0 1090 6.91 32 38.2 1.3 143 37.5 0.69 0.05 0.1 1.76 0.1 627 612 60.6 0.3 0.3 0.9 77 -8.2 1.0Ba 14 22 44.4697 88 29.5505 11 5 45.7 26.1 1011 6.95 < 1 38.1 1.6 131 34.5 0.56 0.05 0.6 2.14 0.0 552 526 69.6 7.8 0.3 0.2 68 -5.4 1.9Ba 15 22 44.4777 88 29.5538 11 0 48.8 26.2 951 6.93 < 1 30.2 1.3 127 34.5 0.63 0.05 1.1 2.94 571 591 39.6 9.0 0.3 0.0 81 -7.4 -1.6Ba 16 22 44.4756 88 29.5661 8 61, 29 33.5 26.0 1265 6.84 < 1 49.5 1.1 181 35.1 0.35 0.06 0.6 5.29 0.1 768 819 60.2 0.1 0.3 0.1 67 -1.3 -2.8Ba 17 22 44.5284 88 29.5615 4 0 43.9 7.32 < 1 43.4 1.0 145 40.4 0.55 0.04 0.2 2.80 0.0 647 69.0 2.2 0.2 0.3 68 -6.4Ba 18 22 44.5200 88 29.5751 7 1 44.2 26.2 998 7.02 < 1 35.4 1.1 134 33.6 0.52 0.03 0.1 2.36 0.1 600 612 45.1 0.2 0.1 0.0 76 -5.6 -0.9Ba 19 22 44.5178 88 29.2745 30.5 26.8 1051 6.96 233 23.8 3.2 129 43.4 0.72 0.58 10.9 0.07 570 583 73.8 6.0 0.1 3.5 68 -6.6 -1.1Ba 20 22 44.5218 88 29.3115 6 0 33.5 26.9 1024 7.03 245 50.3 3.6 108 36.9 0.58 0.33 3.8 0.04 0.4 608 661 31.9 1.4 0.1 2.1 59 -1.2 -3.9Ba 21 22 44.5262 88 29.3085 7 0, 0 33.5 27.0 1003 7.04 330 43.7 3.6 120 37.2 0.60 0.41 7.2 0.04 0.5 638 646 30.8 0.0 0.3 1.1 67 -0.3 -0.6Ba 22 22 44.5352 88 29.3105 33.5 26.8 1133 7.08 415 43.9 3.8 136 43.2 0.68 0.58 8.3 0.06 1.4 703 736 43.4 0.1 0.1 3.3 67 0.3 -2.2Ba 24 22 44.5469 88 29.3095 7 0 39.6 27.0 1110 6.99 370 41.3 3.8 134 42.7 0.67 0.60 7.7 0.06 0.6 678 666 46.4 0.1 0.2 3.7 67 -6.2 0.6Ba 25 22 44.4540 88 29.4515 4 0 39.6 26.6 791 7.07 4 21.8 1.8 105 30.1 0.46 0.03 0.1 0.94 1.4 489 29.9 0.1 0.1 0.1 75 -7.7Ba 26 22 44.4334 88 29.4465 9 2 39.6 26.5 1038 7.18 < 1 25.6 2.7 128 43.5 0.66 0.14 0.2 0.56 1.0 527 87.6 6.3 0.3 0.1 68 -6.5Ba 27 22 44.4246 88 29.4347 3 0 41.1 26.6 957 7.01 42 22.3 2.5 124 39.2 0.60 0.05 0.2 0.74 1.0 521 70.9 1.1 0.1 0.1 71 -4.0Ba 28 22 44.4316 88 29.4245 10 6 26.6 971 7.15 98 33.1 2.3 117 36.4 0.58 0.08 2.2 0.64 525 63.6 2.8 0.5 0.3 71 -6.6Ba 29 22 44.4490 88 29.4331 33.5 7.07 56 35.5 2.0 109 32.5 0.54 0.06 1.6 0.76 514 46.2 3.7 0.1 0.0 67 -7.8Ba 30 22 44.4248 88 29.3953 12 7 39.6 26.4 848 7.16 < 1 21.3 1.9 111 34.0 0.54 0.08 0.1 1.17 0.7 498 43.3 0.1 0.6 0.2 66 -10.1Ba 31 22 44.4165 88 29.3725 12 1, 5 42.7 26.4 1212 6.99 167 39.8 3.1 143 48.1 0.78 0.14 3.1 0.41 1.8 574 116 22.1 0.4 0.2 68 -5.1Ba 32 22 44.4158 88 29.3875 8 0 29.3 26.2 991 7.04 135 22.3 1.9 118 41.2 0.60 0.52 7.9 0.07 1.8 512 78.5 3.8 0.4 3.0 66 -8.8Ba 33 22 44.4287 88 29.3705 11 6, 7, 6, 20 33.5 26.1 932 7.04 174 39.2 1.9 101 36.8 0.57 0.42 7.9 0.08 0.1 526 48.7 5.8 0.2 3.4 68 -9.0Ba 34 22 44.4321 88 29.3805 7 0 51.8 26.4 1217 6.99 380 30.6 2.6 141 45.9 0.70 0.18 5.1 0.14 502 144 7.4 0.3 1.1 66 -6.5Ba 35 22 44.4391 88 29.3675 11 1 29.0 26.2 743 7.08 270 19.2 1.7 91 31.8 0.49 0.25 5.7 0.05 0.7 478 16.1 0.1 0.7 2.9 63 -10.3Ba 36 22 44.4297 88 29.3460 8 1 21.3 26.3 939 7.09 590 25.2 2.2 113 38.3 0.63 0.09 5.7 0.07 1.8 531 56.0 0.0 0.3 1.1 59 -7.5Ba 37 22 44.4411 88 29.3425 10 63, 0 21.3 26.0 688 7.08 390 18.3 7.4 81 31.6 0.49 0.38 4.8 0.04 0.6 453 14.6 0.1 0.3 3.0 67 -7.8Ba 38 22 44.4604 88 29.3405 9 4 41.1 26.3 802 7.13 673 21.6 2.3 100 30.7 0.50 0.10 6.1 0.08 1.2 478 34.7 0.0 0.1 1.7 59 -6.2Ba 39 22 44.4459 88 29.5155 3 8 49.4 26.2 1090 7.22 6 31.8 1.9 141 40.4 0.59 0.05 0.1 1.40 1.5 539 88.0 29.2 0.3 0.1 72 -7.5Ba 40 22 44.5003 88 29.5776 6 1 45.7 26.0 1001 7.02 5 36.6 1.1 127 32.8 0.50 0.04 0.2 2.35 0.1 573 47.7 0.2 0.5 0.1 78 -7.1Ba 41 22 44.5333 88 29.5455 9 1 32.3 26.0 1226 7.02 < 1 58.7 0.9 163 32.5 0.50 0.24 0.1 2.97 700 72.4 0.3 0.3 0.2 63 -0.5Ba 42 22 44.5080 88 29.5538 5 0 < 1 41.9 1.1 136 36.1 0.53 0.03 0.4 2.36 0.0 0.1Ba 43 22 44.5051 88 29.5677 43.3 26.2 867 7.10 < 1 35.0 0.9 114 31.3 0.44 0.03 0.1 2.07 538 37.4 0.2 0.3 0.0 71 -2.6Ba 44 22 44.5100 88 29.5924 4 44.1 1.0 130 33.3 0.46 0.05 2.1 3.01 0.0 0.2Ba 45 22 44.4486 88 29.4996 45.7 909 6.94 < 1 22.8 1.7 125 33.3 0.48 0.10 0.2 2.1 0.0 615 0.2Ba 46 22 44.4918 88 29.4594 33.5 908 7.08 145 29.9 1.8 118 29.8 0.56 0.06 3.3 0.8 539 34.6 0.4 1.2 73Ba 47 22 44.5190 88 29.4437 33.5 988 6.78 152 26.1 2.9 130 36.4 0.64 0.14 7.2 0.2 565 60.8 0.2 1.5 63Ba 49 22 44.6118 88 29.2764 27.4 1181 6.88 138 45.4 4.0 147 41.2 0.64 0.56 10.8 0.1 750 33.0 0.5 5.8 72Ba 50 22 44.6058 88 29.3124 16.8 1195 7.01 310 56.7 3.6 134 42.9 0.63 0.30 7.4 0.1 703 53.5 0.6 4.9 63Ba 51 22 44.6238 88 29.3436 44.5 800 6.97 142 23.2 3.3 100 30.4 0.51 0.39 7.7 0.1 510 19.0 0.2 6.5 81

Well Lat.

(deg.)

Lat. Long.

(deg.)

Long.

(min) a

bSan.

Risk

TTC Depth Temp EC pH

As Na K Ca Mg Sr Ba Fe Mn NH4 HCO3

(diff)HCO3

Field

Cl SO 4 NO3 PO4 H4SiO 4

13 δ C DIC

Bal2(cfu/100ml) (m) (˚C) (µS cm )-1

µg L-1

Units are mg L–1 unless specified otherwise

(‰) c(min.)

1268

J.M

.McA

rthuret

al./Applied

Geochem

istry19(2004)1255–1293

Ba

5222

44.6

124

8829

.307

033

.510

977.

0323

730

.83.

813

742

.10.

620.

4210

.70.

165

251

.10.

96.

166

Ba

5322

44.6

208

8829

.229

632

.012

057.

0815

159

.14.

013

544

.30.

670.

465.

80.

169

964

.80.

32.

258

Ba

5622

44.6

118

8829

.226

025

.911

267.

0119

256

.14.

112

841

.10.

600.

446.

20.

176

63.

0B

a 62

2244

.555

488

29.2

296

33.5

872

6.98

405

41.5

3.5

102

31.5

0.49

0.34

4.8

0.0

595

2.9

Ba

6422

44.5

536

8829

.266

229

.010

857.

0044

447

.53.

912

438

.90.

620.

535.

80.

165

337

.60.

33.

362

Ba

8122

44.3

814

8829

.520

939

.611

206.

914

36.7

2.1

139

42.8

0.69

0.15

0.3

1.5

0.8

537

91.4

41.9

0.2

61B

a 82

2244

.366

488

29.4

807

44.5

1066

6.98

3433

.62.

113

640

.70.

660.

130.

30.

60.

957

361

.230

.60.

274

Ba

8322

44.3

766

8829

.486

444

.599

96.

931

34.9

2.1

128

38.8

0.63

0.12

0.2

0.8

0.2

685

0.1

Ba

8422

44.3

796

8829

.451

639

.610

697.

0649

31.9

2.3

140

40.1

0.66

0.09

0.8

0.7

0.0

568

70.9

24.2

0.4

69B

a 85

2244

.336

488

29.4

570

40.2

1154

7.03

134

.12.

614

744

.50.

690.

140.

30.

60.

254

795

.344

.60.

067

Ba

8622

44.3

250

8829

.421

044

.510

397.

03<

125

.22.

314

239

.00.

680.

070.

20.

80.

058

069

.80.

40.

269

Ba

8722

44.3

900

8829

.387

446

.010

676.

89<

132

.72.

513

239

.70.

640.

081.

90.

769

90.

1B

a 88

2244

.398

888

29.3

646

39.6

963

7.10

< 1

27.3

2.2

123

37.3

0.61

0.04

0.4

0.5

0.0

523

65.5

2.5

0.4

66B

a 89

2244

.353

888

29.3

268

28.0

615

7.05

135

13.0

4.3

8322

.10.

260.

195.

70.

139

413

.20.

42.

967

Ba

9022

44.4

174

8829

.298

638

.496

46.

9953

028

.52.

512

137

.70.

590.

146.

00.

157

046

.00.

81.

658

Ba

9122

44.4

552

8829

.278

838

.489

36.

8017

825

.12.

011

833

.10.

540.

388.

20.

161

52.

2B

a 92

2244

.440

888

29.2

458

26.2

600

7.16

8712

.84.

080

20.8

0.31

0.25

3.7

0.2

345

27.9

4.3

2.2

52B

a 22

222

44.2

965

8829

.487

038

.426

.587

97.

18<

135

.51.

611

235

.30.

580.

060.

30.

772.

460

661

511

.00.

90.

40.

269

-0.7

Ba

223

2244

.290

788

29.5

314

44.5

26.5

837

7.22

189

27.6

2.1

113

33.3

0.54

0.12

6.7

0.22

2.9

606

604

3.6

0.1

0.2

1.7

620.

1B

a 22

422

44.3

470

8829

.500

040

.826

.585

96.

9116

733

.91.

411

234

.90.

560.

092.

00.

352.

559

857

913

.41.

90.

41.

265

1.5

Ba

225

2244

.350

888

29.5

030

42.1

26.4

919

6.88

105

35.6

1.5

118

36.5

0.58

0.09

1.6

0.50

2.5

591

593

30.4

8.4

0.2

0.6

68-0

.2B

a 22

622

44.3

910

8829

.559

526

.433

27.

008

27.1

1.9

122

35.9

0.58

0.04

0.1

2.10

2.6

570

559

36.6

7.3

0.2

0.2

690.

8B

a 22

722

44.4

054

8829

.557

232

.026

.390

47.

0316

22.6

1.8

123

37.1

0.58

0.04

1.1

1.04

2.6

586

579

27.0

4.9

0.2

0.3

740.

5B

a 22

822

44.4

018

8829

.563

235

.126

.476

87.

03<

118

.21.

710

830

.50.

600.

040.

11.

842.

553

08.

30.

20.

10.

173

Ba

229

2244

.445

088

29.5

948

44.2

26.3

872

7.00

< 1

23.5

1.2

120

33.3

0.51

0.04

0.1

2.16

1.9

544

546

33.7

4.8

0.4

0.2

72-0

.1B

a 23

022

44.4

216

8829

.497

243

.626

.291

66.

99<

134

.32.

612

635

.70.

540.

080.

00.

882.

458

858

345

.23.

20.

20.

169

0.3

Ba

231

2244

.360

488

29.5

940

34.4

26.4

694

7.13

186

18.8

1.9

103

29.7

0.50

0.36

5.7

0.11

2.0

505

485

16.5

0.6

0.3

2.3

631.

8B

a 23

222

44.3

682

8829

.617

244

.525

.969

17.

0411

3016

.31.

510

529

.40.

450.

030.

60.

742.

351

352

66.

80.

10.

21.

085

-1.3

Ba

233

2244

.360

488

29.6

736

45.1

26.0

747

6.98

530

.61.

510

629

.40.

510.

050.

11.

461.

750

749

526

.09.

70.

50.

371

1.1

Ba

234

2244

.372

488

29.7

396

41.1

25.9

734

7.02

< 1

19.9

1.2

103

31.7

0.41

0.02

0.0

2.69

1.7

519

513

11.1

0.1

0.3

0.0

730.

6B

a 23

522

44.3

700

8829

.751

036

.626

.110

707.

01<

129

.11.

414

642

.90.

570.

052.

74.

332.

058

298

.39.

40.

40.

174

JAM

, Dee

p A

quife

rB

a 1a

2244

.505

388

29.4

405

90

137.

226

.776

66.

786

47.9

4.4

8624

.30.

510.

160.

80.

150.

149

249

816

.10.

70.

30.

081

-5.1

-0.5

Ba

522

44.4

735

8829

.464

57

011

5.8

26.8

780

7.15

1744

.33.

690

26.7

0.52

0.16

1.4

0.19

1.3

516

534

13.5

0.2

0.4

0.0

75-7

.0-1

.6B

a 23

2244

.560

088

29.3

148

76

125.

027

.774

97.

283

51.8

3.8

8222

.60.

510.

151.

00.

150.

047

747

414

.65.

80.

40.

076

-4.5

0.2

Ba

4822

44.4

324

8829

.308

288

.468

77.

207

44.9

3.7

7231

.40.

610.

180.

60.

10.

047

713

.91.

50.

064

Ba

6522

44.5

212

8829

.236

211

5.8

760

6.76

754

.03.

382

22.4

0.50

0.15

0.8

0.1

0.0

482

13.2

7.8

0.0

7117

6-T

2244

.419

8829

.827

121.

977

57.

303

56.5

4.6

8724

.20.

540.

160.

80.

131.

052

652

312

.22.

60.

10.

172

0.3

Not

es:

a San

. Ris

k is

San

itary

Ris

k; s

ee te

xt fo

r de

tails

.b T

herm

otol

eran

t col

iform

s; c

olon

y-fo

rmin

g un

its p

er 1

00 m

l.c

Bic

arbo

nate

cal

cula

ted

from

imba

lanc

e of

ioni

c ch

arge

.B

al =

ioni

c ba

lanc

e as

100

*(su

m c

atio

ns -

sum

ani

ons)

/(su

m c

atio

ns +

ani

ons)

as

meq

l-1

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1269

Fig. 7. Map of the distribution of As, Fe, Mn and SO4 in JAM wells. Symbols as on each figure. Wells near piezometer DP are rich in

Mn and lack As: wells in northwestern JAM are rich in As and Fe. For a discussion of why, see text.

1270 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

the peat and the underlying aquifer sands. The best

proof of the absence of downward transport is the

presence of the brown clay at 21 m depth, which resulted

from oxidative weathering; it has not been re-reduced

during burial, and the resumption of reducing condi-

tions, as would have happened had organic-rich water

from the overlying peaty layers been percolating

through it. The high water-level in the upper confining

unit at DP (>3 m higher than that at AP or FP; Fig. 6)

provides further proof of impermeability.

5.2. Piezometer waters

The development of anoxia must be driven by con-

sumption of O2 and NO3 in the units rich in organic

matter, and by the reductive effect of small organic

molecules (carboxylic acids, amines, sugars), derived by

fermentative degradation of the organic matter, which

are dispersed through the aquifer (Bergman et al., 1999).

Where the sediments contain FeOOH (DP, base of AP),

concentrations of dissolved As are low (<10 lgL�1),

despite the presence of dissolved Fe which attests to a

degree of reduction of FeOOH. The authors attribute

these low As concentrations to resorption, by residual

FeOOH, of As released by partial reduction of FeOOH.

For example, towards the base of the aquifer at AP

(>38.3 m), concentrations of dissolved Fe, As, and PO4

decrease markedly (Fig. 4). The Fe concentration (0.97

mgL�1) in the lowermost piezometer shows that reduc-

tion of FeOOH has occurred, but the brown colour of

the sediment shows it has not gone to completion. The

low concentration of As in the lowermost piezometer at

AP (7 lg L�1) is therefore attributed to resorption of As

to residual FeOOH. Given sufficient time, organic

matter migrating downwards will reduce all the Fe oxide

in the lower part of the aquifer at AP and release all

of its sorbed As to ground water, before continued

recharge flushes the system clear of As and other

constituents.

A contrast between brown and grey coloration of

sediment (particularly, between grey Holocene sediment

and orange/brown Pleistocene sediment) and an associ-

ation of brown coloration with low-As water has been

noted in the area of Araihazar, Bangladesh (van Geen

0

10

20

30

40

50

0 5 10 15

Fe mg L-1

SO

4 m

g L-1

0

1

2

3

4

5

6

7

0 5 10 15

0

250

500

750

1000

1250

0 2 4 6 8

PO

4 m

g L-1

0

250

500

750

1000

1250

0 5 10 15

0

250

500

750

1000

1250

0 10 20 30 40 50

SO4 mg L-1

PO4 mg L-1

0

250

500

750

1000

1250

0 1 2 3 4 5

As

µg

L-1

As

µg

L-1

As

µg

L-1

As

µg

L-1

Mn mg L-1

Fe mg L-1

Fe mg L-1

(a) (b)

(d)(c)

(e) (f)

Fig. 8. Relation in March, 2003, in JAM wells between (a) As and Mn (b) As and SO4, (c) As and Fe (d) As and PO4 (e) Fe and PO4,

(f) Fe and SO4. Wells are categorized using As/Fe and As concentration as proxies for depth. Open squares are wells with >32 lgL�1

of As and low As/Fe values; filled circles are wells with >32 lgL�1 of As and high As/Fe values; small open circles are wells containing

<16 lgL�1 of As. See text for an explanation of the relation of these groupings to well depth.

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1271

et al., 2003a), but in Araihazar, unlike in JAM, grey

sediment does not always host As-rich water. That a

grey Holocene aquifer overlies a Pleistocene aquifer of a

‘‘deep burnt orange’’ colour was also noted in Munshi-

ganj, Bangladesh (Harvey et al., 2002) but, again, low

concentrations of As prevailed over much of the thick-

ness of the grey Holocene aquifer. The differences be-

tween these sites and JAM awaits explanation.

The concentrations of As in most of the lower part

(37.0–39.8 m) of the aquifer at FP are higher (>900

lgL�1) than they are in the upper part of FP (94–448

lgL�1) and are higher in both than they are at AP

(maximum 420 lgL�1). The authors attribute the dif-

ferences to the fact that FP contains more organic

matter than does AP, and higher peak concentrations of

organic matter (e.g. 7.2% at 29.5 m in FP, 0.76% at 23 m

in AP). This additional organic matter must be driving

reduction of FeOOH in FP further than it has gone in

AP, and results in ground water with higher concen-

trations of both Fe and As. This organic distribution

0

50

100

150

200

0 20 40 60

Na mg L-1

Ca

mg

L-1

0

10

20

30

40

50

60

70

0 50 100 150 200

Cl mg L-1

Na

mg

L -1

A

B

300

400

500

600

700

800

50 100 150 200

Ca mg L-1

HC

O3

mg

L-1

50

100

150

200

15 25 35 45 55

Mg mg L-1

Ca

mg

L-1

(a) (b)

(d)(c)

Fig. 9. Relation in March, 2003, in JAM wells between (a) Ca and Mg (b) Ca and HCO3, (c) Cl and Na (d) Na and Ca. Line in (c)

shows Na/Cl in marine salt. Arrow A in (d) represents a weathering trend: arrow B represents stoichiometric ion-exchange of Ca in

solution with Na on mineral exchange-sites. Symbols as in Fig. 8, with the addition of large, open, circles for deep wells (>70 m), and

crosses for wells in the vicinity of piezometer site DP, which, owing to their proximity to piezometer site DP, are inferred to be tapping

an aquifer different sedimentologically from that tapped by other wells.

1272 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

does not, however, appear to explain the fact that the

maximum As concentrations in AP and FP occur well

below the maximum concentrations of PO4 and Fe: the

control on dissolved P and Fe appears to be partially

decoupled from the control on dissolved As. The dif-

ferent depths of the maxima in As, Fe, and PO4 is

probably due simply to differences of the As/Fe/PO4

ratios in the aquifer sands above and below 32 m, given

their different hardness and age (see Section 4): variable

lithology is another reason why As and Fe seldom cor-

relate well in ground water from alluvial basins.

In DP, the brown colour of the aquifers sands at all

depths (Fig. 10) attest to the presence of FeOOH

throughout the aquifer: that the composition of ground

water is quite different from those at sites AP and FP is

therefore not unexpected. Concentrations of As are low

(<5 lgL�1), whilst concentrations of dissolved Mn and

SO4 are high. The profiles and compositions at DP are

best explained as resulting from the preservation of

originally oxic connate water, which has subsequently

been subject to creeping reduction as organic matter

moves in laterally from surrounding sediments (e.g. site

FP, which is 250 m to the west) to cause (presently)

incomplete reduction of FeOOH and Mn(IV) oxides.

Given the U-shaped profiles of As, Mn and Fe with

depth (Fig. 3), organic matter must be diffusing upwards

from the underlying organic-rich clay at this site (so its

upper part cannot be oxidised?) as well as entering the

aquifer laterally at shallow depths, so that the reduction

potential decreases towards the mid-point of the aquifer:

there, a peak in dissolved Mn (2–4 mgL�1, Fig. 3) co-

incides with a minimum in dissolved As (<2 lgL�1) and

Fe. Reduction of FeOOH dominates over reduction of

Mn oxides towards the top and base of the aquifer.

Concentrations of As are kept low (<5 lgL�1) by re-

sorption of As to residual-FeOOH. Oxides of Mn sorb

As (Manning et al., 2002; Foster et al., 2003), so the As

released by reduction of MnO2 must be re-sorbed to

residual oxides of Mn and Fe, rather than remain in

solution (as noted by St€uben et al., 2003).

The values of d13C in dissolved inorganic C (DIC) are

high for a redox-dominated system. They are kept from

becoming more negative, and so from reflecting organic

inputs more strongly, by the low pH of the ground water

(6.76–7.32; Tables 2 and 3), which promotes rapid

equilibration between DIC and detrital calcite, with a

Fig. 10. Colour of wet sediment from West Bengal, showing the grey colour of sediments in which reduction of FeOOH has gone to

completion, and the brown colour of sediments that retain FeOOH. Groundwater from the former is enriched in As, that from the

latter is not. (For interpretation of the reference to colour in this figure legend, the reader is referred to the web version of this article.)

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1273

d13C around )1‰ (Table 1; Hudson-Edwards et al.,

2004). At AP, values of d13C(DIC) decrease by around

2‰ at 26 m, a change that may reflect an organic con-

tribution to DIC from the degradation of organic matter

at 23m depth. At FP, the abundance of calcite ap-

proaches zero below 30 m depth; below this level, the

more negative values of d13C(DIC) reflect this dimin-

ished buffering capacity (Table 1). Surprisingly, the or-

ganic matter at 29.5 m in FP appears not to influence

values of d13C(DIC). In the upper part of the aquifer at

DP, values of d13C(DIC) are around )2‰: they are

around )6‰ at greater depth. These high values must

reflect equilibration with traces of calcite in the aquifer,

although the analytical data suggests calcite is absent

from DP.

At sites AP and FP, concentrations of SO4 decrease

downwards from the uppermost piezometers, which are

in the upper confining layers (Figs. 4, 5), but it cannot be

said with certainty whether these profiles represent

downward passage (dilution and dispersion) of recent

anthropogenic SO4 from the surface, infiltration of

natural amounts of SO4 concentrated by evaporo-tran-

spiration, or remobilisation of S in Fe sulphide from the

organic-rich layers as oxic water passes through the

uppermost part of the upper confining layers. It seems

unlikely that they represent steady-state consumption of

natural SO4, at depth, by SO4 reduction, because the

sites are <350 m apart and would have similar near-

surface concentrations were the profiles natural. In fact,

concentrations at 7 m depth are 65 mgL�1 at FP and 490

mgL�1 at AP. The authors speculate that the high

concentrations of SO4 at 7 m at AP result from surface

loadings of gypsum as a result of house building, as it is

used in plaster and the locality has several substantial

houses. If this is the case, the profiles imply that SO4

from a surface source can penetrate to 30 m depth within

a few tens of years.

The chemical stratification revealed by the piezome-

ters explains why wells may differ in composition despite

being close to each other (<5 m) and of similar depth.

Arsenic concentrations in the bottom 4 piezometers at

the AP site decrease linearly with depth from 419 lgL�1

at 38.3 m to 7 lgL�1 at 42.8 m, a rate of change of 92

lgL�1 m�1. The rate is 200 lgL�1 m�1 in the basal part

1274 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

of the aquifer at FP. Such high gradients will cause wells

to differ substantially in As concentrations if they differ

by no more than a metre in depth, a distance within the

uncertainty with which a driller would record the depth

of a well.

Chemical stratification also compromises interpre-

tions drawn from well-surveys of As concentrations, in

the Ganges Plain and elsewhere: low concentrations of

As in wells (e.g. as found near AP; Fig. 7) do not indi-

cate that the aquifer they tap is free of As: using local

knowledge of aquifer colour, a driller may emplace wells

in the small thickness of good water in the brown sand

at the base of polluted aquifers, if it exists (as it does at

AP). Consequently, As in ground water would go un-

detected until breakthrough to wells, and its detection

by analysis or the development of arsenocosis in con-

sumers. Breakthrough and detection might be separated

substantially in time and have serious consequences for

health. Furthermore, the finding in an area of dangerous

concentrations of As in only a small percentage of ran-

domly distributed wells does not signal that the As is

localized and the aquifer generally safe for development:

it may show only that most wells tap water from above

or beneath a poisoned part of the aquifer.

5.3. Well waters

The chemical stratification of the ground water (Figs.

3–5), the differing depths of wells and the uncertainty

about some well-depths, and the differing amounts of

organic matter in the aquifer sands from place-to-place,

conflate to obscure areal patterns in water quality

(Fig. 7) and cross-plots of some dissolved constituents

(Fig. 8): nevertheless, some patterns emerge. Wells near

piezometer DP (14–18, 40–44) contain <10 lgL�1 of As

and more than 2 mgL�1 of Mn, so the aquifer they tap

must be sedimentologically and chemically similar to

that seen in the DP piezometer, and be in a setting that is

sedimentological distinct from those of other wells in the

JAM area that lack Mn. The wells near DP must tap a

part of the aquifer that is both free of organic matter

and protected from direct infiltration of organic matter

from above by a thick upper-aquiclude of clay. For wells

near DP, the low concentrations of As and Fe, and the

high concentration of Mn must be due to the fact that

reduction of MnO2 dominates over reduction of Fe-

OOH, and that where the latter occurs, it does not re-

duce all the FeOOH, so, in both cases, re-sorption keeps

concentrations of As low. That concentrations of As are

low when Mn is present is an observation not confined

to the DP area: as Fig. 8(a) shows, As and Mn appear to

be largely mutually exclusive in solution, a situation that

extends to most shallow wells within 2 km of the JAM

area (McArthur and Banerjee, unpublished data) and

more widely in Bangladesh (data in DPHE, 1999). High

concentrations of Mn might therefore indicate the

presence of As-free connate water in an aquifer where

reduction of FeOOH is incomplete. The exploitation of

water at such sites will constitute mining of a finite re-

source and may draw in As-rich water. Such transport

will be moderated by sorption to unreacted FeOOH.

Regular monitoring of water quality is essential in such

a complex situation.

The composition of wells away from site DP can be

interpreted if it is assumed that the vertical pattern of

chemical stratification in the aquifer across the rest of

the JAM area resembles that seen at AP and FP. The

assumptions are: that As concentrations are low (<50

lgL�1) where FeOOH is present to sorb As; that this

occurs only at the base of the aquifer; that As concen-

trations reach a maximum at 35–40 m and decrease

sharply below this depth; that ground water above this

depth contains less As, although concentrations are still

high; and that As concentrations reflect the amount of

reduction that has occurred, and so the amount of or-

ganic matter in the sediments. A well-by-well analysis of

As concentrations, however, does not reveal a clear re-

lation between concentrations of As and well depth,

probably because some well-depths are wrong. There is,

however, some relation between well-depth and both As/

Fe and As concentrations: if the lowest concentrations

of As (<20 lgL�1) are assumed to represent wells em-

placed in the base of the aquifer where FeOOH persists,

as at AP, and if the remaining wells are divided into two

groups on the basis of their As/Fe values, 3 divisions

emerge (Fig. 8), viz.

Group A; 19 wells with <1–16 lgL�1 of As.

Group B; 24 wells with >32 lgL�1 of As and As/Fe

between 13 and 56.

Group C; 24 wells with >32 lgL�1 of As and As/

Fe between 36 and 2025.

The mean well-depths for each group are: 135� 14

feet (41.1 ± 4.27 m) (A); 124� 24 feet (37.8 ± 7.32 m) (B);

and 108� 20 feet (32.9 ± 6.1 m) (C), where all uncer-

tainties are 1 s.d. Although these differences are not sta-

tistically significant, they are interesting because wells are

constructed from 20-foot (6.1 m) lengths of plastic pipe,

topped by 3–6 (0.91–1.82 m) feet of Fe pipe. The mean

depths of these 3 groups suggest that they reflect groups

of wells that are mostly made with 7, 6, or 5 pipes, with 7-

pipe wells (group A) being screened in brown sand at the

base of the aquifer, 6-pipe wells (group C) being screened

in the As maximum seen in the piezometer profiles to

occur in the lower part of the aquifer, and 5-pipe wells

(group B) being screened above the As maximum. The

high standard deviation probably arises because some

well depths are incorrect: nevertheless, enough are correct

to allow the overall depth-control to be seen. The use of

As/Fe, and As concentrations, as a proxy for depth seems

reasonable because the As/Fe values in AP and FP

(Fig. 11) change with depth, reflecting the fact that the

maximum dissolved concentrations of As occur at a

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1275

depth greater than do those of Fe or PO4, which

are themselves not much separated in terms of depth

(Figs. 4,5).

The different As/Fe/PO4 values present at different

depths (Fig. 11) probably arises because the younger

sediment (<32 m depth) releases more Fe and PO4 on

reduction, and less As, than the older sediment (>32 m

depth): such lithological differences, reflected in ground

water composition, must be one reason why Fe and As

often correlate poorly in the well waters of the Ganges

Plain. Dividing the wells onto groups A–C, however,

yields correlations in the As-rich groups between As and

Fe that are better than for the combined population,

probably because the effect of mixing populations from

different depths is reduced. The coefficient of correlation

between As and Fe in the 3 populations (Fig. 8(c)) are:

0.01 for A (pP 0:1); 0.58 for B (p ¼ 0:01); and 0.86 for

C (p ¼ 0:001, if one outlier, with 0.56 mgL�1 of Fe and

1131 lgL�1 of As, is excluded); the pooled population

has a coefficient of 0.58 (p ¼ 0:001 excluding the outlier).In a pooled population of 213 wells within 2 km of JAM,

the correlation coefficient between Fe and As is 0.39

(p ¼ 0:001; Fig. 12; Banerjee and McArthur, unpub-

lished data). If the correlations have meaning, it can be

only because the lower and intermediate concentrations

of As, and the accompanying Fe, reflect a dynamic

process that is proceeding extremely slowly, with the

dissolved As and Fe concentrations representing a

snapshot of slow, microbially mediated release, prior to

sequestration of Fe into solid diagenetic phases, and

resorption of As: the As and Fe are ‘kinetic species’ in

the sense used in Fig. 13 (see next section). Higher

concentrations of As may represent a state where all

reactive FeOOH has been consumed, leaving the As in

solution with no sorber to adhere to: this idea is also

developed in a later section. For groups A–C, the cor-

relation coefficients between PO4 and Fe (Fig. 8(e)) are

0.00 for A (p > 0:1); 0.50 for B (p ¼ 0:01); and 0.62 for

C (p ¼ 0:001): the coefficient for the pooled population

is 0.72 (p ¼ 0:001) and is enhanced by mixing popula-

tions B and C, one high in PO4 and Fe, the other lower

in both. In a pooled population including data for 213

other wells within 2 km of JAM (Fig. 12; Banerjee and

McArthur, unpublished data), the correlation coefficient

between Fe and PO4 is 0.76 (n ¼ 213, p ¼ 0:001).Given the discussion above, the high concentrations

of As and Fe in the northwestern part of JAM (Fig. 7) is

probably due to two factors viz. that most wells there are

emplaced at a depth near to the As maximum (are 6-pipe

wells) and that organic matter around 30 m depth (as

seen at FP) is more abundant in central and NW JAM

than it is elsewhere. In much of southern and eastern

parts of JAM, concentrations of As must be low (<50

lgL�1) because wells there are screened at the base of

the aquifer in the small thickness of brown sand revealed

at AP.

The considerations above show that real element

associations can be disguised or enhanced by pooling

data on too regional a scale, and that piezometer profiles

are essential to fully understand the source, transport,

and sinks, of As in the Bengal Basin. This discussion

also serves to emphasise that the results of well-surveys

must be evaluated in combination with a knowledge of

the local geology, particularly aquifer colour, and well

depth.

5.4. Diagenetic Iron sulphide

Diagenetic Fe sulphide is a sink for dissolved As

(Korte and Fernando, 1991; Rittle et al., 1995), and is

present in the sediments in West Bengal (Fig. 2; Chak-

raborti et al., 2001). As much of the ground water in the

JAM area lacks SO4, as do most ground waters in sur-

rounding areas (McArthur and Banerjee, unpublished

data) and in the Ganges Plain generally (DPHE, 1999,

2001; Nickson et al., 2000; McArthur et al., 2001;

Dowling et al., 2002), it must have been removed by SO4

reduction from most recharging water. As SO4 reduction

is driven by microbial metabolism of organic matter, it is

localised where organic matter is present. At all sites,

total S is almost absent from aquifer sands but is rela-

tively abundant in the organic-rich units (Fig. 2; Table

1), which are (or have been) the site of SO4 reduction

and formation of diagenetic Fe sulphides. The traces of

S in the sands probably represents Fe sulphide formed

very early in the history of anoxia from reduction of SO4

in connate water. Reduction of SO4 is probably not

occurring in the aquifer sands now, either because they

lack SO4 (FP) or because SO4-reducing bacteria are

outcompeted by Fe-reducing bacteria (Chapelle and

Lovley, 1992). The presence of SO4 in the upper aquifer

at AP and FP (Figs. 4 and 5), where it has penetrated to

levels below some organic-rich units, might show that

the organic matter has lost much of its original reducing

power through biodegradation. From these arguments,

and the fact that most of the organic-rich units are

stratigraphically above As-rich aquifers, the authors

infer that little As is being removed now from ground

water in the aquifer sands by sequestration in diagenetic

sulphide.

5.5. Resorption as a control on aqueous As

5.5.1. The shallow aquifer

Knapp et al. (2002) showed that aquifer sands from

Virginia, USA, sorbed SO4 least where FeOOH in the

sediments had been removed from grain surfaces by the

establishment of reducing conditions. In JAM, As con-

centrations are low where the sediments are coloured

brown by FeOOH, and are high where sediments are

grey because all FeOOH has been reductively consumed

(Figs. 10). Reduction is incomplete in those parts of the

0

10

20

30

40

50

600.0 0.2 0.4 0.6 0.8 1.0

As / PO4

Dep

th, m

etre

s bg

l

DP

FP AP

0

10

20

30

40

50

600.00 0.05 0.10 0.15 0.20

As / Fe

Dep

th, m

etre

s bg

l

Fig. 11. Relation of As/Fe to depth at piezometer sites AP and FP in March, 2003. Values of As/Fe, and As concentration, are low

where wells tap brown sediment at the base of the aquifer (more than about 40 m depth); are high where wells tap the aquifer between

35 and 40 m depth, and lower, but still high, where wells tap the aquifer at shallower depths.

1276 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

aquifer that are distal from sources of organic matter

(e.g. the base of the aquifer at AP), or where no hy-

draulic continuity exists between organic-rich sediments

and aquifer sands (at DP). Where reduction is incom-

plete, resorption to residual FeOOH will keep As con-

centrations low (as mooted by Ravenscroft et al., 2001),

until all sorption sites become saturated by re-sorbed As,

or all FeOOH has been consumed: this idea is implicit in

the model of Welch et al. (2000) that explains the gen-

eration of high concentrations of As by progressive re-

duction of FeOOH. In Fig. 13, this model is shown more

explicitly: As released to solution by reduction will exist

in solution temporarily (this is termed ‘kinetic As’) until

it is resorbed to residual FeOOH. Laboratory studies of

microbial FeOOH-reduction show that As resorbs over

a timescale of a few days to weeks (Figs. 3B and C in

Dowdle et al., 1996; Figs. 6 and 7 in Jones et al., 2000).

Resorption will decouple concentrations of As from

those of Fe and leads to poor correlations between these

constituents: this is one reason why these constituents

seldom correlate well in ground waters in alluvial aqui-

fers. Resorption of As will occur as ground water moves

away from the influence of peaty sediment, where re-

duction is complete, to areas where it is not complete

and FeOOH is available to sorb it (DPHE, 1999).

5.5.2. The deep aquifer

Resorption may be one reason why the deep aquifers

in JAM, and widely across the Bangladesh, are free of

As (a very limited exceedance of 50 lgL�1 of As in deep

wells may be explained as contamination or mis-identi-

fication of wells). Sands of the deep aquifer in Bangla-

desh are typically described as ‘‘burnt-orange’’ (Harvey

et al., 2002) or as ‘‘yellowish-brown to pale grey sands

0

250

500

750

1000

1250

0 2 4 6 8 10 12 14

Fe mg L-1

As

µg

L-1

0

1

2

3

4

5

6

7

0 2 4 6 8 10 12 14

Fe mg L-1

PO

4 m

g L-1

Fig. 12. Relation of As, PO4, and Fe, in ground waters sampled from 213 wells in JAM in November, 2002, within 2 km of the study

area (McArthur and Banerjee, unpublished data). By analogy with piezometer profiles, the poor correlation between As and Fe

(r ¼ 0:39, n ¼ 213) arises because As/Fe values in ground water reflect As/Fe values in aquifer sediments, and are lithology-dependent.

The better correlation between Fe and PO4 (r ¼ 0:76, n ¼ 213) reflects a more uniform distribution (less lithology-dependent con-

centration) of PO4 than of As in aquifer sands.

Fig. 13. Schematic of dissolved concentrations of As as a function of the percentage of FeOOH reduced. Kinetic-As is that released by

partial reduction of FeOOH but not immediately re-sorbed: its concentration represents a balance between the rates of FeOOH re-

duction and As sorption. Filled-arrow 1 shows position in the sequence of reduction of groundwater at site DP and the lowermost

piezometer (42.8 m) at site AP; filled-arrow 2 shows the likely state of sediments and ground waters at site FP and the middle and upper

aquifer at site AP.

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1277

1278 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

. . . cemented by secondary clay and Fe oxide’’ by Rav-

enscroft in DPHE (1999; Vol. S1-Main Report), because

they were subjected to a long period of oxidative

weathering during the lowering of sea-level that oc-

curred between 125 and 18 ka. This weathering must

have generated FeOOH by oxidation of biotite and

other Fe-bearing minerals in the sediment. As a conse-

quence, the higher content of FeOOH in Pleistocene

sands of the deep aquifer distinguishes them (Stanley

et al., 2000) from the Holocene sands of the shallow

aquifer. The authors suggest that the establishment of

anoxic conditions in the deep aquifer, as a consequence

of the deposition of overlying, organic-rich, Holocene

aquifers, has been sufficient to mobilise some Fe into

ground water (Table 3; DPHE, 1999), but has been in-

sufficient to reduce all its FeOOH, the remains of which

may now act as a sorptive buffer to prevent As from

accumulating in solution. An alternative explanation for

the lack of As in waters from the deep aquifer, previ-

ously suggested by Ravenscroft (DPHE, 1999), McAr-

thur et al. (2001), and Ravenscroft et al. (2001), and

repeated in Smedley and Kinniburgh (2002), is that

sorbed As was flushed from the deep aquifer during an

extended period of oxidative weathering and flushing

that occurred during the lowering of sea level between

125 and 18 ka. With available data, the authors cannot

be certain which explanation is correct.

6. The driver for FeOOH reduction

6.1. Natural organic matter in sediments

6.1.1. The JAM Area

The data for the JAM area suggest that reduction of

FeOOH is driven by natural, subsurface, organic matter,

in organic-rich units, formed at the periphery of local

peat basins, that are in hydraulic continuity with the

aquifer. The concentrations of organic matter in JAM

sediments (low% levels) is more than adequate to ac-

count for the amount of reduction present: at 25 �C (the

mean annual air temperature in JAM) infiltrating water

contains around 8.3 mgL�1 of dissolved O2, an amount

consumed by <0.001% of reactive organic matter in the

sediment. That peaty sediment is connected with the

presence of unwholesome water is confirmed by local

experience in JAM: Ashraf Ali (Ardivok driller, pers.

comm., 2003) has noted an association between bad

water and ‘khalu mati’ (black soil, which the authors

interpret to be peat, or peaty units) and between brown

sediment and good water. The association suggests that

connectivity (permeability) exists between aquifers and

peat deposits in some parts of the Ganges Plain, perhaps

where they are cut through by post-depositional chan-

nelling (H. Brammer, pers comm., 2001).

In JAM, however, (where CH4 is not recorded) the

DP profile shows that peat sensu stricto is encapsulated

in impermeable clay: the deposition of the peat must

have occurred in response to local waterlogging caused

by the clay. The low permeability of this clay now pre-

vents recharge from carrying soluble organic matter

downwards into the underlying aquifer sands. The sed-

iment profiles at FP and AP show that sites marginal to

the peat depocentre (DP) developed lesser amounts of

organic matter, in peaty layers that are lateral extensions

of the main peat (i.e. are of a similar age; Hudson-Ed-

wards et al., 2004). In these marginal environments, the

thinner, perhaps intermittent, clay will allow greater

hydraulic connectivity with underlying aquifer sands

and so allow ingress of organic matter. During sedi-

mentation, however, such sites would have been less

waterlogged and so would have accumulated less or-

ganic matter than basin centres, and so would host less

organic matter. Beyond the basin margin, permeability

will be higher, but the amount of organic matter will be

negligible. These ideas are illustrated in Fig. 14, in which

a schematic cross-section through the JAM area shows

the relation of permeability, organic matter, and sedi-

mentary architecture. This model suggests that chemical

reduction in the JAM aquifer (and by implication,

elsewhere in the Bengal Basin) is driven by organic

matter in peaty units that develop on the margins of peat

basins where there is enough organic matter to drive

redox, and sufficient permeability to allow organic

matter to invade the aquifer in soluble form.

6.1.2. Other areas in the Bengal Basin

The coincidence of high concentrations of As in

groundwater and wetland environments, where plant

productivity is high, was noted by Ravenscroft (DPHE,

1999) and has been reproduced in Smedley and Kinni-

burgh (2002), whilst both McArthur et al. (2001) and

Ravenscroft et al. (2001) argued that peat, which forms

in such wetlands, was the driver for reduction of Fe-

OOH in the Ganges Plain. One reason for an emphasis

on peat is that CH4 in combustible amounts is found at

depth in some ground waters (Ahmed et al., 1998;

Harvey et al., 2002) and the fact that CH4 in such

amounts can be generated only by anoxic degradation of

high concentrations of organic matter. If organic matter,

accumulating marginally to peat depocentres, drives

reduction of FeOOH (and so As release to ground wa-

ter) it is crucial to establish the extent of peat deposition

in the Bengal Basin.

Peat is common beneath the Old Meghna Estuarine

Floodplain in Comilla (Ahmed et al., 1998), Sylhet, and

the Gopalganj-Khulna Peat Basins (Reimann, 1993;

Brammer, 1996). Many wells in the area around Farid-

pur may be screened in waterlogged peat (Safiullah,

1998) and the shallow aquifer system in Lakshmipur

contains peat (DPHE, 1999). Peat is often found in

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1279

geotechnical borings (piston samples), although it is

rarely recorded during rotary drilling for water wells

because such drilling masks its presence, unless the peat

is very thick. Peat occurs extensively beneath the As-

affected areas of southern Samta village and Deulhi

village in southwestern Bangladesh at a depth of about

10 m (AAN, 2000; Ishiga et al., 2000). Peat has been

found in Holocene sediments around Mymensingh

(Ishiga et al., 2000) and is well documented at Panigati,

in SW Bangladesh (Islam and Tooley, 1999). Organic-

rich sediment has been reported to occur at 6 m and 25

m depth in Faridpur and at 28m depth in Sonargaon

(Fig. 3 of Tareq et al., 2003). Sediment from a depth of

2.1 m at Gopalganj (100 km SW of Dhaka) contained

6% TOC (Nickson et al., 1998); sediment from a depth

of 23m at Tepakhola (Faridpur municipality) contained

7.8% TOC (Safiullah, 1998). Peat is repeatedly men-

tioned by Umitsu (1987, 1993) and by Goodbred and

Kuehl (2000) as being present in Bangladesh sediments;

it has been reported to occur in southern West Bengal by

Banerjee and Sen (1987), and also by Pal et al. (2002)

who note that peat was present in the subsurface in an

As-free area, but not in a nearby As-affected area. These

authors do not state whether their peat was underlain by

impermeable strata, as at the DP site, but it seems likely.

It could be that the apparent lack of peat at their As-

affected site means that it was missed during coring.

Harvey et al. (2002) also reported that no peat was

found at an As-affected site in Bangladesh, but they

report the presence of CH4 (marsh gas) in combustible

amounts; this proves the presence of much organic

matter in the aquifer close to their site.

Thus, in the Bengal Basin, peat deposits are wide-

spread, and so the margins of such deposits must be

Fig. 14. Schematic of the relationship of peat basins to the occurren

drive FeOOH-reduction can invade the aquifer only at the margins o

allow its passage downwards in recharging ground water and the se

Towards the basin centre, permeability decreases; away from the ba

enrichment of As decreases.

common. Despite the importance of peaty sediment as

the redox driver in the Ganges Plain, there are two other

potential drivers of redox: these are organic waste in

latrines (McArthur et al., 2001; Ravenscroft et al., 2001),

and organic matter leached from surface sources, such

as soils, river beds, and ponds (Harvey et al., 2002).

Each is considered in turn.

6.2. Other sources of organic matter

6.2.1. Organic matter in latrines

Human organic waste in latrines does locally pro-

mote reduction of FeOOH in aquifers and contributes

As to groundwater to locally enhance As concentra-

tions, thereby confusing the regional signal of As de-

rived from reduction driven by natural organic matter.

The impact of point-sources can be seen in Singair, an

As hot-spot in Holocene sediments some 10 km west of

Dhaka, Bangladesh: of 15 shallow wells within 300 m of

Singair mosque (Safiullah and McArthur, unpublished

data), the 3 wells sited within 3 m of latrines contained

900 lgL�1 of As, whilst the remaining 12 wells, all >15

m from latrines, contained between 100 and 300 lgL�1

of As. Such an influence is not uncommon; new wells

drilled through new concrete pit-latrines have been ob-

served (McArthur, unpublished data). Point sources will

be particularly strong where the polluting organic mat-

ter is in hydraulic continuity with the aquifer e.g. where

the depth of a latrine is greater than the thickness of the

upper confining layer and/or where wells and latrines are

adjacent. But organic matter in latrines cannot be the

main driver of As release to ground water because it is

free of As in many areas of Bangladesh, despite the fact

that the country is densely, and fairly evenly, populated.

ce of As enrichment in groundwater. Soluble organic matter to

f peat basins, where the sediments are sufficiently permeable to

dimentary source of OM is sufficiently abundant to supply it.

sin, the abundance of organic matter decreases: in both cases,

1280 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

Where the depth of dug (pit) latrines is much less

than the thickness of the upper confining layer, where

wells are far from latrines, where sewage collection is

engineered, or where the unsaturated zone is very thick

(as it is on parts of the Madhupur and Barind Tracts),

the impact of human sources of organic matter will be

small, or negligible. Such is the case in JAM, where most

dwellings have concrete-lined holding tanks with tops

above ground level and where the depth of the upper

confining layer is between 9 m (at FP and DP) and 24 m

(at DP). Several considerations suggest that there is little

connectivity between latrines and aquifer water in JAM.

Firstly, thermotolerant (faecal) coliforms (presumptive

Escherichia coli) are bacterial indicators of faecal pol-

lution. Despite the very high risks of faecal pollution

from poor well-completion, open access of many wells to

animals, and the fact that there is generally a broad

relationship between log(TTC) and sanitary risk scores

(Cronin et al., 2003), the study wells mostly have a low

TTC count (Table 3; summary statistics in Table 4):

about 85% of the wells contained <10 cfu/100 ml of

water. Furthermore, there is no correlation between

TTC count and As concentrations in JAM wells. Given

the high number of observable risks to JAM wells, their

low TTC counts probably stem from the fact that most

are >30 m deep, are sealed around the annulus by the

clayey silt of the thick (P 9 m thick) upper confining

layer, and many have concrete completions of good

quality, although the completions are mostly small in

diameter. These factors reduce access of microbially

polluted water into the aquifer via the annulus of the

well. The implications of these results is that microbial

contamination of the aquifer in the JAM area is on a

small scale and so organic pollution of the aquifer by

human waste is limited in volume and distribution.

Hence, it is unlikely that faecal organics do much to

drive reduction of FeOOH in JAM aquifers.

6.2.2. Surface sources of organic matter

For reasons given above, it seems unlikely that re-

duction of FeOOH in the JAM area is driven by surface

Table 4

Summary table of content of thermotolerant coliform bacteria in wat

and Moyna (JAM), and Mirathi, West Bengal, India, (March 2003). C

count

Summary statistics JAM wells

Numbers are cfu/100 ml

Number, incl. repeats 49

Maximum TTC 63

Minimum TTC 0

Non-detects (TTC¼ 0) 23

Mean TTC counts in wells with positive detects 10

Notes. cfu is colony-forming units.

TTC is too numerous to count.

sources of organic matter (rivers, ponds, and soils),

drawn into the aquifer by irrigation drawdown, as has

been suggested to occur in Bangladesh (Harvey et al.,

2002). The authors note, however, that the occurrence of

CH4 in combustible amounts around 30 m depth in

Munshiganj, but in lessening amounts as depth dimin-

ishes (Harvey et al., 2002), and the presence of CH4 in

ground water elsewhere in the Bengal Basin (Ahmed

et al., 1998), can be due only to the presence of sub-

surface organic matter in high concentrations. Further-

more, if surface sources of organic matter in the Bengal

Basin were of much importance in driving the release to

ground water of large amounts of As, high As concen-

trations would not be confined to about 25% of wells

(DPHE, 1999), or to Holocene sediments (it is largely

absent from the Barind and Madhupur Tracts; DPHE,

1999; McArthur et al., 2001; Ravenscroft et al., 2001),

nor would they be highest around 20 to 40 m depth

across most of the Bengal Basin (Karim et al., 1997;

McArthur et al., 2001; Ravenscroft et al., 2001; Harvey

et al., 2002; van Geen et al., 2003a). Finally, although

the 14C ages of Harvey et al. (2002) for DIC show that

very young water is penetrating to depths of around

20 m, the piezometer profiles of those authors show that

concentrations of dissolved organic carbon (DOC) in-

creases with depth to around 30 m depth, below which

they scatter but remain high to 80 m depth, thereby

proving the DOC has a deep source. The As maximum

(and a maximum concentration of CH4) at 35–40 m

depth cannot be caused by organic matter migrating

downwards against a concentration gradient and then

reacting at depth. The discrepancy noted by Harvey

et al. (2002) between young 14C ages for DIC and older14C ages for DOC is explicable in terms of a reservoir

age for DOC and its derivation from older buried peat.

6.3. The capacity to reduce FeOOH

The results given here provide strong evidence that

reduction of FeOOH across the Ganges Plain is driven

overwhelmingly by buried, natural, organic matter.

ers from wells, ponds, and piezometers, from Joypur, Ardivok

olony forming units (cfu) per 100 ml. TNTC¼ too numerous to

Mirathi wells JAM ponds JAM piezos

15 11 21

65 TNTC TNTC

0 4400 0

1 0 1

14 – –

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1281

According to the equation for reduction of FeOOH,

where FeOOH is used to represent Fe oxides

8FeOOHþ CH3COOHþ 14H2CO3

$ 8Fe2þ þ 16HCO�3 þ 12H2O ð1Þ

Given: the stoichiometry of this reaction; a density of 1

for buried peaty sediment; a C content of 7% in the

sediment; a thickness of 50 cm for a peaty layer (as at

29.5 m depth in FP) with half the C bioavailable; and a

porosity of 40%, then there will be about 1.1 g of bio-

reactive C under each cm2 of aquifer. The organic

matter will consume 25 g of FeOOH which, given a

porosity of 40%, a density of FeOOH of 2.4, and an

aquifer containing 0.2% FeOOH, will be contained in an

8 m thickness of aquifer sand, which is half that of the

aquifer thickness at FP below the peaty layer at 29.5 m

depth, and is similar to the thickness in which concen-

trations of As exceed 900 lgL�1. Towards the base of

the aquifer at FP, concentrations of As decrease whilst

remaining high. Our calculation is very basic, but sug-

gests that few aquifer volumes of water have passed

vertically through the sands: had they done so, the or-

ganic matter, or the FeOOH, would be exhausted by

now and the aquifer would be clear of both dissolved Fe

and dissolved As.

Although approximate, the calculations above sug-

gest (and observation confirms) that there is enough

organic matter in the aquifers in JAM for the entire

aquifer thickness to be at risk: confusing the calculation

are 3 factors. Firstly, the organic matter in the sediments

at JAM is P 1.2 ka old (Hudson-Edwards et al., 2004),

so it has been degrading for a substantial period of time.

On deposition of the sediment, the amount of organic

matter would have been greater and it would have been

fresher, and more biodegradable, than it is now. Sec-

ondly, there is some uncertainty regarding the estimate

of 0.2% for the amount of FeOOH in the aquifer. This

figure represents the Fe, recalculated to FeOOH, that is

leachable by ammonium oxalate at pH 3 (Table 5) from

sands in mid-aquifer at site DP, which is the least re-

duced of the 3 piezometer sites (Figs. 3–5 and 10). The

figure of 0.2% is a maximum because the leaches con-

tained P, S, Al, Fe, Mg, (but not Ca, because Ca oxalate

is insoluble) and it is not possible to apportion the

leachable Fe between FeOOH and possible diagenetic

phases formed by its partial reduction viz. magnetite,

green rust, vivianite, siderite, Fe sulphide (Lovley et al.,

1987; Frederickson et al., 1998; Chakraborti et al., 2001;

Glasauer et al., 2003), or to possible detrital phases such

as Fe-rich chlorite. Thirdly, the formation of magnetite

(Fe2þFe3þ2 O4) by reduction of sedimentary FeOOH re-

duces only one-third of the available pool of Fe(III)

whilst sequestering the remaining two-thirds in the

mineral structure (Frederickson et al., 1998) where it

may be unavailable for reduction. Thus, the availability

of FeOOH to sorb As in Bangladesh sediments will be

some 70% less than supposed if magnetite is the main

product of FeOOH reduction across the Ganges Plain.

It can be inferred from the results of Harvey et al. (2002)

that magnetite comprised most of the Fe that was ex-

tractable with oxalic acid solution from aquifer sedi-

ments from, Bangladesh, and so magnetite constituted

up to 6% of total-Fe (although this inference is com-

promised by uncertainty about whether the mineral

formed by oxidation of green rust during storage, or by

reduction in the aquifer).

7. Other mechanisms of As release to ground water

This study confirms that reduction of FeOOH is a

good model for the release of high concentrations of As

to anoxic ground water in the Ganges Plain and, by

implication, worldwide, since reduction of FeOOH is

not a site-specific process. Other mechanisms, such as

reductive dissolution of Mn oxides (Smedley and

Kinniburgh, 2002), competitive exchange with HCO3

(Appelo et al., 2002), and pH-dependent desorption

from, or recrystallization of, FeOOH (Welch et al., 2000;

Smedley and Kinniburgh, 2002), are not needed. The

reasons why are discussed below.

7.1. Reduction of Mn oxides

In JAM wells, concentrations of dissolved As are low

(<33 lgL�1) where concentrations of Mn exceed 1

mgL�1 (Table 3; Fig. 8(a)). The reduction of Mn oxides

is thermodynamically more favourable than is the re-

duction of FeOOH (Edmunds, 1986; St€uben et al., 2003)

and occurs before it. It follows that As, originally sorbed

to Mn oxides but released by reduction, must be

resorbed to FeOOH (or residual Mn oxides) rather than

be released to ground water (St€uben et al., 2003). This

finding is contrary to that of Smedley and Kinniburgh

(2002), who state that ‘‘Manganese oxides also undergo

reductive desorption and dissolution and so could con-

tribute to the As load of ground waters in the same way

as Fe’’.

7.2. Exchange of sorbed As with HCO3 in ground water

Potentially toxic concentrations of As in ground

water are not common, and are particularly uncommon

in potable water from oxic aquifers, in which HCO3 is

the commonest and most abundant anion worldwide: it

follows that the suggestion that As is released to solution

by competitive exchange with HCO3 (Appelo et al.,

2002) is unlikely to be correct. Ground waters from a

variety of aquifer types in the UK contain HCO3 in

concentrations of hundreds of mL�1, yet As concen-

Table 5

Provenance, and content of oxalate-extractable Fe, of sediment from aquifers in alluvial and deltaic settings worldwide

Setting Source of sediment Nature of source % of oxalate-extractable Fe Data source

Min Max

Bangladesh Himalayas Ig, Met 0.1 0.5 DPHE (1999, 2001)

West Bengal Himalayas Ig, Met 0.05 0.16 This work, site DP,

oxidised

West Bengal Himalayas Ig, Met 0.12 0.12 This work, site AP,

oxidised

Campania, Italy Sarno & Latari

Mountains

Ig, Sed (Lmst) 0.83 1.6 Adam et al. (2003)

Coastal Virginia,

USA

Allegheny Mountains ? 0.4 41 Knapp et al. (2002)

Huhhot Basin,

Inner Mongolia

Da Qing & Man Han

Mountains

Met 0.67 3.4 Smedley et al. (2003)

North Island,

New Zealand

North Island uplands Ig, Sed. 3.7 8.1 Nguten and Sukias

(2002)

NW Mississippi Mississippi uplands Sed 1.2 1.9 Rhoton et al. (2002)

Lower Kissimmee

Basin, Florida

Florida uplands Sed Lmst? 0.07 0.25 Reddy et al. (1995)

1282 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

trations are typically <10 lgL�1 and natural concen-

trations above 50 lgL�1 are rare (BGS, 1989). More-

over, concentrations of HCO3 in JAM are highest (744

mgL�1 at site DP, Table 2; Fig. 3) where the ground

water contains less than 5 lgL�1 of As. Furthermore,

the large range of As concentrations at AP and FP bear

no relation to concentrations of HCO3 (Figs. 4, 5): this is

best seen in the lower part of the aquifer at AP (Fig. 15),

where concentrations of As decrease from 419 lg L�1

(38.3 m) to 7 lg L�1 (42.8 m) whilst pH varies by less

than 0.2 units, HCO3 increases from 526 to 622 mgL�1,

dissolved silica (as H4SiO4) varies by only 4 mgL�1,

dissolved PO4 decreases from 2.3 to 0.13 mgL�1, and dis-

solved Fe decreases from 3.9 to 0.97 mgL�1. These data

emphasise the minimal role played by pH, HCO3, and

silica, in affecting concentrations of As in ground water

in JAM and, by implication, in sediments elsewhere.

Competitive exchange with natural organic matter

(Redman et al., 2002), or other ligands (e.g. silica), may

become important when considering mechanisms of As

release that lead to concentrations of As of <50 lg L�1

but, even at low concentrations, partial reduction of

FeOOH and release of kinetic As (As released by re-

duction but soon resorbed to residual FeOOH, Fig. 13)

may be adequate to explain low concentrations of As in

anoxic environments.

7.3. Desorption of As induced by increasing pH

Sorption of As to solids, notably sedimentary Fe-

OOH, is generally held to be pH dependent, leading to

the suggestion that desorption of As from mineral sur-

faces may contribute As to ground water as pH rises

(Welch et al., 1988, 2000; Robertson, 1989; Smedley and

Kinniburgh, 2002). A general observation that waters of

higher pH tend to have higher concentrations of As

(Welch et al., 1988; Robertson, 1989) is not in itself

adequate to prove a causal connection between the two:

other factors, such as residence time, evaporation, and

weathering, may cause an increase in both. Whilst pH-

dependent desorption may release As to ground water in

plumes of landfill leachate, or hydrocarbons where pH

can exceed 10, there is little direct evidence that this

process releases As to potable ground water, in which

pH rarely exceeds 8.5. In waters from JAM, As con-

centrations range from <1 to 1180 lgL�1, despite the

narrow range of pH (6.8–7.3; Table 3), a fact best

illustrated by data from the base of the AP site (Fig. 15).

It follows that pH is not a control on any sorption or

desorption of As that may occur in JAM aquifers.

Arsenic concentrations were linked to pH by Welch

et al. (1988): the mechanism of pH-dependent desorp-

tion was used by Robertson (1989) to explain the pres-

ence of As in oxic potable ground water from 24 alluvial

basins in Arizona, and has since been widely invoked.

Robertson (1989) noted that concentrations of As in-

creased as pH increased along flowlines in aquifers in

several alluvial basins, and assumed the relation was

causal. Many indicators in his data suggest otherwise: in

Verde Valley ground waters, As concentration decreased

with increasing pH (as he noted); the overall correlation-

coefficient between pH and As concentration for his 24

sites was low (0.18, n ¼ 456; his Table 2); salinity in-

creased along flowlines by a factor of between 4 and 8

(judged from his Fig. 3) suggesting either extensive

evaporo-concentration, mixing with deeper more saline

waters, or weathering, occurred as the water passed

down the aquifer. It seems likely that As accumulated in

Fig. 15. Detailed profiles of water composition in the basal aquifer at Site AP, showing the invariance in pH and in the concentrations

of HCO3, and H4SiO4, in contrast with the change in concentration of dissolved As, PO4 and Fe.

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1283

solution along some flowlines in response to in-situ

weathering of volcanic rocks, and that the accompany-

ing increase in pH was a consequence of such weath-

ering, not a cause of As accumulation. Robertson (1989)

noted that ‘‘the basins containing the largest concen-

trations (of As) are bounded primarily by basalt and

andesite’’ and so the basin-fill must be weathered ma-

terials from those rocks, which may have moderated the

concentrations by resorption to weathered sediments.

Robertson’s (1989) observation that concentrations of

As correlated well with concentrations of Mo, F and V,

at sites examined in detail, parallels the observation of

Nicolli et al. (1989) that As, fluorine, and V, are asso-

ciated in As-enriched ground water in central Argentina;

both the association and the As enrichment were as-

cribed by Nicolli et al. (1989) to the weathering of dacitic

volcanic ashfall layers in the loess aquifers.

The relation of pH to As in solution can be further

illustrated by reference to ground waters in aquifers of

metamorphic bedrock in New England, USA (Ayotte

et al., 2003). Higher concentrations of As (>10 lgL�1)

occur in waters with higher pH (7.5 to 9.3), but many

samples, with pH ranging from 5 to 8.8, contain <2

lgL�1 of As (Fig. 4 of Ayotte et al., 2003): this would

not be the case if As concentrations were linked to pH-

related desorption. These authors note that the primary

control on As concentrations is bedrock type. Further-

more, concentrations of As in UK aquifers are not re-

lated to pH (BGS, 1989): concentrations of <4 lg�l

occur in aquifers with a pH range from 7.0 to 9.5 (e.g.

the Lincolnshire Limestone, Table 5.3 of BGS (1989))

and concentrations rarely exceeded 50 lg L�1, whatever

the pH. Thus, the observation made by Welch et al.

(1988) of a weak positive correlation between pH and

concentrations of As in a large dataset of compositions

of US ground waters may have explanations other than

pH-induced desorption.

Finally, Dixit and Hering (2003) show that sorption

of both As(III) and As(V) to Fe oxides becomes less

sensitive to pH as As concentrations, and Asaq/Fesediment

ratios, decrease. At their lowest As concentration (750

lgL�1), which approaches those found in the Bengal

Basin, these authors showed that sorption is almost in-

dependent of pH over the range of pH common in most

ground waters. Their data confirm that pH-dependent

sorption does not contribute significantly to the release

of As to anoxic potable ground waters of the composi-

tion, pH-range, and salinity typical of potable waters

from alluvial aquifers.

8. The Himalayan connection

8.1. The Bengal Basin

The authors postulate that the widespread occur-

rence of As concentrations above 100 lgL�1 occurs only

when sorption sites for As in an aquifer are saturated i.e.

when reduction of FeOOH has gone largely to comple-

tion. Because high concentrations of As are common in

Holocene sediments, they evidently have a particular

susceptibility to undergo complete reduction of FeOOH.

the authors explain this as follows. The Holocene sedi-

ment in the Bengal Basin derives from Himalayan rocks

1284 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

that are (mostly) igneous and metamorphic. Other than

adjacent to the Indo-Burman ranges, surface coatings of

FeOOH, with sorbed As, are not inherited from previ-

ous erosional cycles but must be created from crystalline

rocks by weathering. The aquifer sands largely derive

from erosion at a time of extended glaciation in the

Himalayas when physical weathering dominated and the

intensity of chemical weathering was limited by the low

(glacial) temperatures during erosion. The immature

sediment was transported rapidly to the lower reaches of

the basin and buried in a waterlogged, anoxic environ-

ment and so had little opportunity to acquire much

FeOOH through oxidative weathering. Further, much of

the fine fraction was winnowed during transport and

deposited offshore, thereby reducing the surface area of

sediment available to host coatings of FeOOH. The

abundance of fresh-looking muscovite and biotite in the

Holocene sediments attests to the limited degree of

chemical weathering they have undergone. Older sedi-

ments, such as the deep aquifer in Bangladesh, and

sediment at DP, and below 32 m depth in JAM, has

suffered more weathering.

In Table 5, the content of diagenetically available Fe

(as acid-oxalate leachable Fe) in oxidised brown sedi-

ments from JAM (from the base of the aquifer at AP,

and from DP) is compared to that in sediments from

elsewhere and shown to be low (<0.2%) in comparison.

As the brown sediment at depth in JAM is of pre-Ho-

locene age and has probably suffered a period of oxi-

dative weathering (that would have increased its content

of FeOOH), the original Holocene sediment may have

contained even less. Admittedly, such comparisons are

fraught with uncertainty: the leachable Fe may be in

vivianite, green rust, magnetite, Fe sulphides, chlorite,

and carbonate, as well as in oxides. Nevertheless, this

data lends some support to the idea that ground water in

the Bengal Basin is polluted by arsenic because its sed-

iments contain little FeOOH (to sorb As) but have a

superabundance of organic matter with which to reduce

it, so total reduction is common. In this context, the

comment of Harvey et al. (2002, p. 1604), that grey-

coloured sediment in an As-enriched profile in the

Ganges Plain ‘‘contains little, if any, purely ferric

[Fe(III)] oxyhydroxides’’, confirms that, where high

concentrations of As are common, reduction of FeOOH

has gone effectively to completion.

8.2. Implications for other areas

In aquifers that contain much FeOOH and/or little

organic matter, reduction of FeOOH will be incomplete

and the release of As by reduction will be moderated by

resorption to residual FeOOH. Such a process may be

moderating As concentrations in groundwater in, for

example, the alluvial aquifers of the Tualatin Basin,

Oregon (Hinkle, 1997), where ground waters contain

modest concentrations of As (<1–73 lgL�1), and where

those that exceed 50 lgL�1 also contain some Fe (0.16–

1.9 mgL�1) and PO4 (0.36–2.0 mgL�1). Hinckle and

Polette (1999) speculated that FeOOH reduction was

one plausible mechanism of As release to ground water

in reduced ground waters of the basin. The authors

agree, but add that resorption of As to residual FeOOH

has probably kept its concentration low. A more ex-

treme example of resorption control may occur in allu-

vial aquifers of eastern Arkansas, where ground water

contains up to 42 mgL�1 of dissolved Fe, but less than

52 lgL�1 of As (Kresse and Fazio, 2003). Although the

aquifer sediments contain lignite and peat, the amount

of organic matter available (i.e. biodegradable) appears

to have been insufficient to reduce all of the sedimentary

FeOOH, so concentrations of As are probably being

moderated by resorption to unreduced oxides. Con-

versely, the Song Hong (Red River) Basin, in the vicinity

of Hanoi, Vietnam, appears to be another basin where a

superabundance of peat (up to 5 m in thickness; BGS,

1996) occurs in the area most poisoned by As (Fig. 16)

and gives rise to ground water compositions similar

from those in the Bengal Basin (Fig. 17): ground water

in both areas is rich in As, Fe, NH4 and PO4, and

probably for the same reason – the organic matter in the

Song Hong basis is abundant enough to drive FeOOH

reduction to completion. Lesser amounts of peaty sedi-

ment occur elsewhere in the Song Hong delta (Tanabe

et al., 2003) making it likely that toxic concentrations of

As are widespread in the area’s ground water.

9. Temporal trends in As concentrations

9.1. Reasons why the composition of ground water may

change with time

All ground waters change their composition with

time: the important question is, at what rate? In the

Bengal Basin, some reasons for real change are clear

from the chemical stratification of the ground water:

water, and hence As, may move by bulk flow, both

vertically and horizontally, and so redistribute As.

With vertical gradients in As concentration of up to

200 lgL�1 m�1 (at FP in JAM), a small migration of

the As front will severely poison aquifer levels that are

presently safe to exploit. Additionally, seasonal fluctu-

ations of the water table might affect the As concen-

tration of wells screened near the water table, although

this has not been observed over a monsoonal cycle

elsewhere (van Geen et al., 2003a). In the longer term,

redox reactions may deplete the store of As, or Fe-

OOH, or redox driver, in the sediments and so cause

the water composition to change with time. One thing

that is unlikely to cause change is a change in the rate

of ground water flow. Microbial reduction of FeOOH

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1285

occurs on a timescale of hours to days (e.g. Dowdle

et al., 1996; Jones et al., 2000; Zobrist et al., 2000), as

do in-situ redox reactions in the Ganges Plain (Harvey

et al., 2002). Such rates are rapid in comparison to

rates of ground water movement in the Ganges Plain,

which are typically in the range of a few centimetres

per day or less (Ravenscroft et al., 2001; Harvey et al.,

2002), and so are likely to be de-coupled from rates of

ground-water flow.

Some reasons for apparent change with time in

ground water composition include:

(1) improving well-location as understanding of As pol-

lution improves;

(2) poor or inconsistent techniques for sampling, sample

storage, or analysis;

(3) inappropriate purging volume;

(4) misidentification of the well;

(5) transcription errors.

9.2. Long-term changes in the Bengal Basin

There are insufficient time-series data to demonstrate

or disprove a long-term change in water composition

with time in the Bengal Basin, but tentative conclusions

Fig. 16. Comparison of the distribution of As in ground waters arou

deposits (after BGS, 1996). The inner box shows the extent of the BG

There is some indication that As enrichment is greatest where the pea

where the peat is thickest, to the immediate south of Hanoi.

on long-term trends in composition can be inferred from

data on the ages and As concentrations of wells docu-

mented in regional surveys (e.g. DPHE, 1999, 2001; van

Geen et al., 2003a). For the shallow aquifer of Bangla-

desh, the proportion of wells falling above various

threshold concentrations of As, plotted as a function of

well age (Table 6; Fig. 18), suggests that, for the first ten

years of existence, the exceedance for most threshold

values of As increases as well-age increases, irrespective

of region or threshold. A similar effect is not seen in data

for Na (Fig. 18), an element that is largely conservative

in solution, unaffected by redox transformations, and so

is used as a comparator. For ages greater than 10 a, the

large uncertainties in the proportion of wells exceeding a

given threshold, which arise as a result of the paucity of

older wells, makes uncertain the validity of an apparent

increase in concentrations towards older ages. Similar

deductions were made for a data set of 6 000 wells from

the region of Araihazar (van Geen et al., 2003a), but

were based on linear regression to highly skewed data

(without log-transformation) and so may not be robust.

The trend to 10 a seen in the DPHE data probably does

not reflect the fact that an increasing awareness of the

As problem might have lead drillers to site wells more

nd Hanoi (after Berg et al., 2001) with the distribution of peat

S map, and the lightly stippled area is where peat is >5 m thick.

t is found, and further indication that the enrichment is greatest

0

10

20

30

40

0 200 400 600 800 0 200 400 600 800 0 200 400 600 800

HCO3 mg L-1 HCO3 mg L-1 HCO3 mg L-1

SO

4 m

g L-1

PO

4 m

g L-1

0 5 10 15 20

NH4 mg L-1 NH4 mg L-1NH4 mg L-10 5 10 15 20

0

2

4

6

8

10

12

0 5 10 15 20

JAM Area, West Bengal Hanoi Area, Viet Nam Faridpur Area, Bangladesh

0 5 10 15 20

Fe mg L-10 5 10 15 20

0

1

2

3

4

0 5 10 15 20

Mn

mg

L-1

Fe mg L-1 Fe mg L-1

Fig. 17. Comparison of the concentrations of Fe, Mn, PO4, NH4, SO4, and HCO3 in wells tapping the As-affected shallow aquifers of

Bangladesh (data from DPHE, 1999), West Bengal (this work) and the Song Hong (Red River) Basin around Hanoi (data from BGS,

1996, which did not report concentrations of As). All 3 areas contain waters rich in Fe, As, PO4, and NH4, which shows the influence of

much reduction of FeOOH and organic-matter degradation.

1286 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

carefully than before – perhaps on the basis of sediment

colour. Two considerations suggest that the trends may

be real. Firstly, sampling was competed by 1998, when

the As problem was not widely known about in rural

areas. Secondly, there is a distinct difference in trend

between redox-sensitive As and the more conservative

tracer Na (Cl was not analysed for the national survey of

Bangladesh, so could not be modelled).

In the deep aquifers of the Bengal Basin, the estab-

lishment of anoxic conditions in the deep aquifer, as a

consequence of the deposition of overlying, organic-rich,

Holocene aquifers, has been sufficient to mobilise some

Fe into groundwater in the deep aquifer (Table 3;

DPHE, 1999), but has been insufficient to reduce the

entire pool of FeOOH, which may now act as a sorptive

buffer to prevent P or As from accumulating in solution.

If FeOOH in the deep aquifer of the Bengal Basin does

host sorbed As (i.e. the deep aquifer has not been flushed

of it – see earlier), the potential exists for downward-

migrating organic matter to eventually reduce all its

FeOOH and so mobilize As. The data in Table 7,

compiled from DPHE (1999), suggest that this is not yet

happening. In the most As-polluted enriched areas of

Bangladesh, deep wells have remained free of As for up

to 16 a.

9.3. Short-term trends in JAM

In order to examine whether wells in the JAM area

changed over the short time encompassed by the present

study, the wells were sampled on 3 separate occasions

over 18 months; in February, 2002, November, 2002,

and March, 2003. In the most precise data viz. electrical

conductivity, Ca, Mg, and Cl, some differences are seen

in water composition between samplings. Changes are

small and best illustrated by the correlation in relative

Table 6

Non-linear models for the proportion of wells exceeding given

threshold concentration of As, as a function of age. The data

are fitted with the function P ¼ b� ðb� aÞct, where P is the

proportion exceeding a given threshold, a is the intercept, b is

the plateau value, c is a fitted constant and t is time in years.

Data from DPHE (2001)

Threshold

value

a b c t90

All Wells: As lg L�1

25 0.159 0.523 0.921 24

50 0.122 0.597 0.963 55

100 0.068 0.45 0.968 66

200 )0.005 0.184 0.899 22

400 )0.025 0.036 0.665 7

Depth <100 m: As

25 0.173 0.587 0.944 34

50 0.139 1.406 0.985 143

100 0.076 0.645 0.982 120

200 0.002 0.204 0.921 28

400 )0.019 0.041 0.736 9

Barisal: As

25 )0.135 0.51 0.797 11

50 )0.093 0.415 0.786 10

100 )0.044 0.425 0.858 16

200 )0.014 0.346 0.899 22

400 0.001 0.1 0.925 29

Khulna: As

25 )0.222 0.535 0.713 8

50 )0.333 0.457 0.682 7

100 )0.318 0.26 0.581 6

Chittagong: As

25 )0.199 0.661 0.895 23

100 )0.121 0.687 0.951 49

Rajshahi: As

25 )0.001 0.103 0.772 9

All wells: Na mg L�1

40 )0.048 0.416 0.568 4

60 )0.199 0.329 0.505 4

100 )0.281 0.233 0.391 3

200 )0.182 0.15 0.451 4

Khulna: As¼ 200, 400; no fit.

Chittagong: As¼ 50, 200 400; no fit.

Rajshahi: As¼ 50, 100, 200, 400; no fit.

t90¼ time in years to reach 90% of the plateau value.

0

0

.2

0.4

0.6

0.

8

Pro

port

ion

over

thre

shol

d

Wells < 100m deep

As > 50 µg L -1 a = 0.139b = 0.146c = 0.985

0

0

.2

0.4

0.6

0.

8

Pro

port

ion

over

thre

shol

d

All wells: As > 50 µg L-1

a = 0.122b = 0.597c = 0.963

0 10 20 30

0

0.2

0.4

0.6

0.

8

Pro

port

ion

over

thre

shol

d

Well age in years

All wells

Na > 60 mg L -1a = - 0.199 b = 0.329 c = 0.505

Fig. 18. Trend of As concentration in wells with well age,

represented by a model showing the proportion of wells above a

threshold concentration as a function of time. Fitted numeri-

cally by the function [proportion over threshold]¼ b�ðb� aÞct,where a is intercept, b is plateau value, c is a constant and t is

time in years.

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1287

change between these determinands (Fig. 19). The cor-

relation between Ca and Mg might be an analytical ar-

tefact – despite the authors’ belief in its reality–because

both were measured on the same solution, with the same

ICP-AES, at the same time. The measurements of Cl

and electrical conductivity were made independently of

Ca; Cl by ion chromatography in the laboratory, and

electrical conductivity by conductivity meter in the field.

The correlation between Ca and these determinands

must therefore be real. No changes were seen in con-

-4

-2

0

2

4

-15 -10 -5 0 5 10 15 20 25 30

Cl % difference

Ca

% d

iffer

ence

-4

-2

0

2

4

-5 -4 -3 -2 -1 0 1 2 3 4 5

Measured EC % difference

Ca

% d

iffer

ence

-4

-2

0

2

4

-5 -4 -3 -2 -1 0 1 2 3 4 5

Mg % difference

Ca

% d

iffer

ence

Fig. 19. Percentage difference in Ca concentration in wells,

sampled in February 2002 and March 2003, plotted against the

% change in Mg concentration (both measured by ICP-AES),

the % change in EC (field measurement) and the % change in Cl

concentration measured by ion chromatography.

1288 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

centrations of As: the analytical repeatability for As

analysis was around 10%, so the magnitude of any

changes must be less than this amount; the data do not

rule out changes in As concentration, but merely con-

strain the rate of change. Indeed, the piezometer profiles

suggest that the depths at which ground water is en-

riched will increase as organic matter moves downwards

through the aquifer, consuming FeOOH, although

subsequent migration of the As released may be slowed

by adsorption to such phases elsewhere in the aquifer.

9.4. Changes induced by irrigation pumping in the Bengal

basin

An important issue is whether pumping for irrigation

is spreading As through the shallow aquifers of the

Bengal Basin, or is flushing them of As by inducing

additional recharge. The authors believe the latter is the

case. This contention may be best illustrated by site FP

in the JAM area (Fig. 5). The piezometer profile shows

that As concentration increases steeply between 31.2 and

37.0 m depth (or, given that there are only two control

points, somewhere between). Given the large amount of

organic matter at 29.5 m depth, the increase should

occur across this unit and give higher concentrations of

As at 31.2 m. That it does not suggests that the entire

profile has been translocated downwards by between

about 2 and 9 m by abstraction of water from shallow

and deep aquifers. Such movement would also have

lowered As concentrations above 19.2 m depth (Fig. 5)

by drawing in new water to give the present profile.

Water above 19.2 m depth has concentrations of Ca and

Na quite different to those in deeper waters, adding

support to this suggestion.

Perhaps more convincingly, the depth profiles of

Harvey et al. (2002), for groundwater in, Bangladesh,

show that concentrations of dissolved constituents de-

rived from differing sources (e.g. As from reduction of

FeOOH; dissolved organic C, NH3 and CH4 from or-

ganic matter degradation; Ca from mineral weathering;

HCO3 from multiple sources) all increase with depth to a

peak around 35–40 m depth. Such a concordance of

profiles chemically different species can be induced only

by flushing, not by reactions in the aquifer. In this

context, it is interesting to compare the profiles in

Munshiganj and JAM.

Table 7

Summary of As concentrations, depth, and age, for deep wells in As-

Area No. of

deep wells

Age range

in 1998

Ra

As

Chandpur 4 2–3 0–

Lakshmipur 6 3–10 0–

Noakhali 6 5–15 2–

In JAM, As concentrations reach a maximum at

depth in FP (Fig. 5) because of the OM-rich layer at 29.5

m depth. At AP, however (Fig. 4), the maximum is not

affected areas of Bangladesh. Data from DPHE (2001)

nge of

lgL�1

Depth range (m) % of shallow wells

with >50 g L�1 As

6 221–269 98

8 183–318 64

10 246–292 79

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1289

developed strongly (if data from the upper aquitard are

excluded). The progression from a ‘flat’ profile, similar

to that seen at AP, to the ‘peaked’ profile seen in

Munshiganj can be achieved by depth-dependent flush-

ing, which would lower the concentration of all species

in the upper part of the aquifer (Fig. 20). Harvey et al.

(2002) propose or imply, instead, that the shape of many

element profiles results from redox reactions that are

driven by organic matter from surface sources; this is

unlikely, as dissolved species move down, not up, con-

centration gradients, and organic matter cannot move

downwards and increase in concentration as it is being

consumed by reaction. The suggestion of a surface

source of organic matter is also at odds with the distri-

bution of 14C-ages, as discussed by van Geen et al.

(2003b), who point out that Harvey et al. (2002) report

measurable bomb -14C to 19 m depth but not below.A maximum in dissolved As concentration around

20–40 m depth is common across the Bengal Basin

(Karim et al., 1997; DPHE, 1999; McArthur et al.,

2001), but not universal: for example, in Araihazar,

Bangladesh, it appears to be around 15 m depth (Fig. 6

of van Geen et al., 2003a), whilst in the region of Sylhet,

it is around 60 m and may even be bi-modal, although as

a result of differential subsidence, not pumping (Rav-

enscroft et al., 2001). The explanation that shallow

maxima in As concentrations are caused by flushing of

shallow levels in the aquifers seems reasonable because

natural flow, and abstraction, particularly for irrigation,

are confined to shallow depths – natural flow because of

the lack of topography and the lack of natural outlets at

deeper levels, the irrigation flow because shallow irri-

gation wells are at shallow depths in order to minimize

the cost of installation. Ground water dating by Dow-

ling et al. (2002) confirms that young water may be

found to depths of 30 m in the Bengal Basin. If correct,

the authors’ explanation for peaked profiles of As shows

that the extensive development of ground water for ir-

rigation is reducing the As problem in the ground water

across the Bengal Basin.

Fig. 20. Development of a peaked profile in As concentration

as a result of abstraction for irrigation and public supply.

Heavy pumping of the shallow aquifer for irrigation would

enhance the rate of renewal of water in the aquifer to the depth

of pumping. From an initial profile, A, similar to that seen in

JAM at site AP, would develop a peaked profile, B, similar to

that seen in Munshiganj by Harvey et al. (2002), by rapid re-

newal of water in the upper aquifer (shaded area between A and

B) to a degree that diminishes with depth. Pumping from the

deep aquifer (not shown) beneath the lower aquitard would

induce slow translocation downward of the entire profile B.

Localized OM-rich sediment shown as black stringers.

10. Wider considerations

The process of FeOOH-reduction is not site specific:

it can explain the As problem in many sedimentary ba-

sins, both coastal and continental, because Fe oxides

and sorbed As are ubiquitous components of clastic

sediments. Whether the reduction of FeOOH will push

As concentrations to hazardous levels depends on how

near the process goes to completion. That, in turn, is

governed by the balance between the amount of FeOOH

and the amount and reactivity of organic matter avail-

able to reduce it. Whilst concentrations of up to a few

tens of lgL�1 of what the authors have called kinetic As

may prove common where concentrations are moder-

ated by resorption to residual FeOOH, frequent high

concentrations of As (>100 lgL�1) require the accu-

mulation of organic matter in amounts sufficient to ex-

haust the entire store of FeOOH in a location that is in

hydraulic continuity with aquifer sands. This condition

is most easily met on the margins of buried peat basins,

where organic matter accumulates in low percentage

abundance in moderately permeable confining units or

in the aquifer sands themselves, or where significant

contributions from human and animal organic loading

exist locally.

As first proposed by P. Ravenscroft, as author of

DPHE (1999, V1–3, 5, 6), by Ahmed and Ravenscroft

(2000) and Ravenscroft et al. (2001), and since repeated

by Smedley and Kinniburgh (2002), the global rise in

sea-level between 18 and 6 ka, created settings world-

wide for As-enriched aquifers to form. The large volume

of accommodation space that resulted is now filled with

recent sediments. The rates of sea level rise and sediment

supply have been the primary controls on wetland for-

mation, where organic-rich muds and alluvial or deltaic

sands are closely juxtaposed. The problem of As in

ground water will therefore be most prevalent in depo-

sitional areas where climate was hot and wet during the

1290 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

climax of nearshore sediment deposition around 4–8 ka

and where there was limited chemical weathering in the

upper catchment. A set of globally applicable parame-

ters–the depth of pre-Holocene incision, the relative

rates of sea level rise and alluvial aggradation, the Ho-

locene climatic record, and the geology of the upper

catchment–therefore provide a general model for pre-

dicting the possible occurrence of As in coastal alluvial

basins. Inland basins lack the drainage control imposed

by global sea-level, so inland basins will be subject to

local base-level controls and application of the outline

model to such areas may be difficult. In a global context,

the physical and demographic characteristics of the

Bengal Basin represent an extreme, in which the extreme

topography of the hinterland, its weathering under ex-

tremes of climate, and its crystalline geology, have

conspired to exacerbate the pollution by minimising

chemical weathering and production of FeOOH. Other

areas that might be expected to suffer from the As

problem, although perhaps to a lesser degree, include the

deltas of tropical rivers such as the Mekong, Irrawaddy,

and Chao Phraya rivers, and the coastal plains of Java

and Sumatra.

11. Conclusions

In the Bengal Basin, As is released to solution by

reductive dissolution of FeOOH and release of its sorbed

As to ground water. In the study area near Barasat in

southern West Bengal, and by implication elsewhere in

the Bengal Basin and worldwide in alluvial aquifers, the

reduction is driven by natural organic matter buried in

sediments. In JAM, the organic matter exists in mod-

erate percentage abundance within the upper confining

units of the aquifers and/or in discrete organic-rich units

within aquifer sands. The organic matter accumulated in

peat basins; the portion that drives reduction in the

aquifer is that which accumulated on the margins of the

basin where the sediments are sufficiently permeable to

allow infiltration to the aquifer of the soluble organic

products of fermentative degradation of the organic

matter. Within the JAM area, which is typical of much

of the southern Bengal Basin, surface sources of organic

matter do not contribute to the reduction of FeOOH

and so do not result in ground water containing high

concentrations of As. In areas of the Bengal Basin where

dug latrines penetrate the upper confining layer, sewage

in hydraulic continuity with the aquifer may locally

promote additional reduction of FeOOH and add to the

As problem. Unless populations reach an urban density,

however, this process will not exacerbate the regional As

problem but will operate at the scale of individual wells.

Where the reduction of FeOOH goes to completion,

As concentrations often exceed 100 lgL�1. Where re-

duction of FeOOH is incomplete, concentrations of ki-

netic As in solution are moderated (to <100 lgL�1) by

resorption to residual FeOOH. The extent of As mo-

bilisation will be controlled by the balance between the

quantity and biodegradability of organic matter, and the

size of the FeOOH buffer available to sorb As. Reduc-

tion of FeOOH in the Bengal Basin commonly goes to

completion because the sediments contain abundant

natural organic matter to drive reduction but little Fe-

OOH, owing to the immaturity of the sediment, which is

produced and transported rapidly from a crystalline

Himalayan source to its place of accumulation in the

anoxic environment of the Bengal Basin.

Acknowledgements

We are indebted to the inhabitants of Joypur,

Ardivok, and Moyna for allowing us to sample their

wells. Particular thanks are due to Ali Ahmed Fakir,

for his assistance over several years with logistics, and

to him, Mrs. D. Chandra Ghatak, and Dual Ch. Shee,

for allowing us to install piezometers on their land. We

thank A. Osborn for assistance with analysis, much of

which was done in the Wolfson Laboratory for Envi-

ronmental Geochemistry (UCL-Birkbeck, University of

London). Additional analysis was done using the

NERC ICP-AES Facility at RHUL, with the permis-

sion of its Director, Dr. J.N. Walsh. Thanks are due

Sam Tonkin for providing Fig. 1. This work was

supported by the Royal Society, the British Council via

a LINK Scheme grant to Delhi University and a new

investigators grant to KHE (NERC Grant No. GR8/

04292). A.C. thanks Drs. S. Peddle and R. Taylor for

their support. We thank Laurent Charlet, Doris

St€uben, and an anonymous reviewer, for their helpful

comments about our work.

References

AAN, 2000. Arsenic contamination in groundwater and

hydrogeology, Samta village, western Bangladesh. Asian

Arsenic Network. Report July 2000.

Adam, P., Denaix, L., Terribile, F., Zampella, M., 2003.

Characterization of heavy metals in contaminated volcanic

soils of the Solofrana river valley (southern Italy). Geo-

derma (in press).

Ahmed, K.M., Hoque, M., Hasan, M.K., Ravenscroft, P.,

Chowdhury, L.R., 1998. Occurrence and origin of water

well CH4 gas in Bangladesh. J. Geol. Soc. India 51, 697–708.

Ahmed, K.M., Ravenscroft, P., 2000. Extensive arsenic con-

tamination of Bangladesh groundwater: curse of unman-

aged development? In: Inyang, H.I., Ogunro, V.O. (Eds.),

Proceedings of the 4th International Symposium on Envi-

ronmental Geotechnology and Global Sustainable Devel-

opment, vol. I. CEEST, University of Massachusetts, pp.

105–114.

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1291

Ali, M., 2003. Review of drilling and tubewell technology for

groundwater irrigation. In: Rahman, A.A., Ravenscroft, P.

(Eds.), Groundwater Resources and Development in Ban-

gladesh – Background to the Arsenic Crisis, Agricultural

Potential and the Environment. University Press Ltd,

Dhaka.

Anawar, H.M., Akai, J., Komaki, K., Terao, H., Yoshioka, T.,

Ishizuka, T., Safiullah, S., Kata, K., 2003. Geochemical

occurrence of arsenic in groundwater of Bangladesh: source

and mobilization processes. J. Geochem. Explor. 77, 109–

131.

Anon., 1994. The bacteriological examination of drinking water

supplies 1994. Reports on Public Health and Medical

Subjects No. 71: Methods for the examination of waters

and associated materials. HMSO, London.

Appelo, C.A.J., Van Der Weiden, M.J.J., Tournassat, C.,

Charlet, L., 2002. Carbonate ions and arsenic dissolution by

groundwater. Environ. Sci. Technol. 36, 3096–3103.

Ayotte, J.D., Montgomery, D.L., Flanagan, S.M., Robinson,

K.W., 2003. Arsenic in groundwater in Eastern New

England: occurrence, controls, and human health implica-

tions. Environ. Sci. Technol. 37, 2075–2083.

Bandyopadhyay, R.K., 2002. Hydrochemistry of arsenic in

Nadia District, West Bengal. J. Geol. Soc. India 59,

33–46.

Banerjee, M., Sen, P.K., 1987. Palaeobiology in understanding

the change of sea level and coastline in Bengal basin during

Holocene period. Indian J. Earth Sci. 14, 307–320.

Banfield, J.F., Nealson, K.H., Lovley, D.R., 1998. Geomicro-

biology: interactions between microbes and minerals. Sci-

ence 280, 54–55.

Berg, M., Tran, H.C., Nguyen, T.C., Pham, H.V., Schertenleib,

R., Giger, W., 2001. Arsenic contamination of ground and

drinking water in Vietnam: a human health threat. Environ.

Sci. Technol. 35, 2621–2626.

Bergman, I., Lundberg, P., Nilsson, M., 1999. Microbial carbon

mineralisation in an acid surface peat: effects of environ-

mental factors in laboratory incubations. Soil Biol. Bio-

chem. 31, 1867–1877.

Bethke, C.M., 1998. The Geochemist’s Workbench Release 3.0.

C.M. Bethke, Hydrogeology Program, University of Illi-

nois.

BGS, 1989. Trace element occurrence in British groundwaters.

Research Report SD/89/3.

BGS, 1996. NAFF Report WC/96/22. The effect of urbaniza-

tion on the groundwater quality beneath the city of Hanoi,

Vietnam. British Geological Survey.

Brammer, H., 1996. The Geography of the Soils of Bangladesh.

University Press Ltd, Dhaka.

Chakraborti, D., Basu, G.K., Biswas, B.K.,, Chowdhury, U.K.,

Rahman, M.M., Paul, K., Roy, Chowdhury, T., Chanda,

C.R., Lodh, D., 2001. Characterization of arsenic bearing

sediments in Gangetic Delta of West Bengal, India. In:

Chappell, W.R., Abernathy, C.O., Calderon, R.L. (Eds.),

Arsenic Exposure and Health Effects IV. Elsevier, Oxford,

pp. 53–77.

Chakraborti, D., Rahman, M.M., Paul, K., Chowdhury, U.K.,

Sengupta, M.K., Lodh, D., Chanda, C.R., Saha, K.C.,

Mukherjee, S.C., 2002. Arsenic calamity in the Indian

subcontinent – What lessons have been learned? Tantala 58,

3–22.

Chapelle, F.H., 2000. The significance of microbial processes in

hydrogeology and geochemistry. Hydrogeol. J. 8, 41–46.

Chapelle, F.H., Lovley, D.R., 1992. Competitive-exclusion of

sulphate reduction by Fe(III)-reducing bacteria – a mech-

anism for producing discrete zones of high-iron ground-

water. Ground Water 30, 29–36.

Chatterjee, A., Das, D., Mandal, B.K., Chowdhury, T.R.,

Samanta, G., Chakraborti, D., 1995. Arsenic in ground-

water in 6 districts of West Bengal, India – the biggest

arsenic calamity in the world 1. Arsenic species in drinking-

water and urine of the affected people. Analyst 120, 643–

650.

Cronin, A.A., Breslin, N., Taylor, R., Pedley, S., 2003.

Assessing the risks to groundwater quality from on-site

sanitation and poor sanitary well completion, ‘Ecosan –

closing the loop’. In: Proceedings of the 2nd International

Conference on Ecological Sanitation, Lubeck, Germany,

April.

Cummings, D.E., Caccavo Jr., F., Fendorf, S., Rosensweig,

R.F., 1999. Arsenic mobilization by the dissimilatory

Fe(III)-reducing bacterium Shewanella alga BrY. Environ.

Sci. Technol. 33, 723–729.

Dixit, S., Hering, J.G., 2003. Comparison of arsenic (V) and

arsenic (III) sorption to iron oxide minerals: implications

for arsenic mobility. Environ. Sci. Technol. 37, 4182–4189.

Dowdle, P.R., Laverman, A.M., Oremland, R.S., 1996. Bacte-

rial dissimilatory reduction of arsenic (V) to arsenic (III) in

anoxic sediments. Appl. Environ. Microbiol. 62, 1664–1669.

Dowling, C.B., Poreda, R.J., Basu, A.R., Peters, S.L., Aggar-

wal, P.K., 2002. Geochemical study of arsenic release

mechanisms in the Bengal Basin groundwater. Water

Resour. Res. 38, 1173–1190.

DPHE, 1999. Groundwater studies for arsenic contamination

in Bangladesh. Final Report, Rapid Investigation Phase.

Department of Public Health Engineering, Government of

Bangladesh. Mott MacDonald (P. Ravenscroft; V1, 2, 3, 5,

6) and British Geological Survey (V4).

DPHE, 2001. In: Kinniburgh, D.G., Smedley, P.L. (Eds).

Arsenic Contamination of Groundwater in Bangladesh.

BGS Technical Report, WC/00/19, British Geological Sur-

vey, Keyworth.

Edmunds, W.M., 1986. In: Brandon, T.W. (Ed.), Groundwater

Chemistry. Groundwater: Occurrence, Development and

Protection. Inst. Water Eng. Sci., Water Practice Manual 5

(Chapter 3).

Foster, A.L., Brown, G.E., Parks, G.A., 2003. X-ray absorp-

tion fine structure study of As(V) and Se(IV) sorption

complexes on hydrous Mn oxides. Geochim. Cosmochim.

Acta 67, 1937–1953.

Frederickson, J.A., Zachara, J.M., Kennedy, D.W., Dong, H.,

Onstott, T.C., Hinman, N.W., Li, S.-M., 1998. Biogenic

iron mineralization accompanying the dissimilatory reduc-

tion of hydrous ferric oxide by a groundwater bacterium.

Geochim. Cosmochim. Acta 62, 3239–3257.

Glasauer, S., Weidler, P.G., Langley, S., Beveridge, T.J., 2003.

Control on Fe reduction and mineral formation by a

subsurface bacterium. Geochim. Cosmochim. Acta 67,

1277–1288.

Goodbred Jr., S.L., Kuehl, S.A., 2000. The significance of large

sediment supply, active tectonism, and eustasy on margin

sequence development: late Quaternary stratigraphy and

1292 J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293

evolution of the Ganges–Brahmaputra delta. Sed. Geol.

133, 227–248.

Gulens, J., Champ, D.R., Jackson, R.E., 1979. Influence of

redox environments on the mobility of arsenic in ground-

water. In: Jenne, E.A. (Ed.), Chemical Modeling in Aque-

ous Systems. American Chemical Society Symposium

Series, vol. 93, pp. 81–95.

Harvey, C.F., Swartz, C.H., Badruzzaman, A.B.M., Keon-

Blute, N., Yu, W., Ali, M.A., Jay, J., Beckie, R., Niedan, V.,

Brabander, D., Oates, P.M., Ashfaque, K.N., Islam, S.,

Hemond, H.F., Ahmed, M.F., 2002. Arsenic mobility and

groundwater extraction in Bangladesh. Science 298, 1602–

1606.

Hinkle, S.R., 1997. Quality of shallow ground water in alluvial

aquifers of the Willamette Basin, Oregon, 1993–95. US

Geol. Survey Water-Res. Investigations Report 97-4082–B.

Hinckle, S.R., Polette, D.J., 1999. Arsenic in groundwater of

the Willamette Basin, Oregon. USGS Water Resour. Invest.

Rep. 98-4205.

Hudson-Edwards, K., Banerjee, D., Ravenscroft, P., McAr-

thur, J.M., Mishra, R., Purohit, R., Lowry, D., 2004.

Sediments of the West Bengal and their relation to arsenic

pollution (in preparation).

Ishiga, H., Dozen, K., Yamazaki, C., Ahmed, F., Islam, M.B.,

Rahman, M.H., Sattar, M.A., Yamamoto, H., Itoh, K.,

2000. Geological constrains on arsenic contamination of

groundwater in Bangladesh. Abstract of 5th Forum on

Asian arsenic contamination, Yokohama, Japan, RGAG,

Saitama, pp. 53–62.

Islam, M.S., Tooley, M.J., 1999. Coastal and sea level changes

during the Holocene in Bangladesh. Quatern. Int. 55,

61–75.

Jones, C.A., Langner, H.W., Anderson, K., McDermott, T.R.,

Inskeep, W.P., 2000. Rates of microbially mediated arse-

nate reduction and solubilization. Soil Sci. Soc. Am. 64,

600–608.

Karim, M., Komori, Y., Alam, M., 1997. Subsurface As

occurrence and depth of contamination in Bangladesh. J.

Environ. Chem. 7, 783–792.

Knapp, A.P., Herman, J.S., Mills, A.L., Hornberg, G.M., 2002.

Changes in the sorption capacity of Coastal Plain sediments

due to redox alteration of mineral surfaces. Appl. Geochem.

17, 387–398.

Kresse, T., Fazio, J., 2003. Occurrence of arsenic in ground

waters of Arkansas and implications for source and release

mechanisms. Water Quality Report, Arkansas Ambient

Ground-Water Monitoring Program, Arkansas Department

of Environmental Quality.

Korte, N.E., Fernando, Q., 1991. A review of arsenic(III) in

groundwater. Crit. Rev. Environ. Contr. 21, 1–39.

Lovley, D.R., 1997. Microbial Fe(III) reduction in subsurface

environments. FEMS Microbiol. Rev. 30, 305–313.

Lovely, D.R., Anderson, R.T., 2000. Influence of dissimilatory

metal reduction on the fate of organic and metal contam-

inants in the subsurface. Hydrogeol. J. 8, 77–88.

Lovley, D.R., Stolz, J.F., Nord, G.L., Phillips, E.J.P., 1987.

Anaerobic production of magnetite by a dissimilatory iron-

reducing micro-organism. Nature 330, 252–254.

Lloyd, B., Bartram, J., 1991. Surveillance solutions to micro-

biological problems in water quality control in developing

countries. Water Sci. Technol. 24 (2), 61–75.

Manning, B.A., Fendorf, S.E., Bostick, B., Suarez, D.L., 2002.

arsenic (III) oxidation and Arsenic (V) adsorption reac-

tions on synthetic birnessite. Environ. Sci. Technol. 36,

976–981.

Matisoff, G., Khourey, C.J., Hall, J.F., Varnes, A.W., Strain,

W., 1982. The nature and source of arsenic in Northeastern

Ohio ground water. Ground Water 20, 446–455.

McArthur, J.M., Ravenscroft, P., Safiullah, S., Thirlwall, M.F.,

2001. Arsenic in groundwater: testing pollution mechanisms

for sedimentary aquifers in Bangladesh. Water Resour. Res.

37, 109–117.

Nealson, K.H., 1997. Sediment bacteria: Who’s there, what are

they doing, and what’s new? Annu. Rev. Earth Planet. Sci.

25, 403–434.

Nguten, L., Sukias, J., 2002. Phosphorus fractions and reten-

tion in drainage ditch sediments receiving surface runoff and

subsurface drainage from agricultural catchments in the

North Island, New Zealand. Agric., Ecosyst. Environ. 92,

49–69.

Nickson, R., McArthur, J., Burgess, W., Ahmed, K.M.,

Ravenscroft, P., Rahman, M., 1998. Arsenic poisoning of

Bangladesh groundwater. Nature 395, 338.

Nickson, R.T., McArthur, J.M., Ravenscroft, P., Burgess,

W.S., Ahmed, K.M., 2000. Mechanism of arsenic poisoning

of groundwater in Bangladesh and West Bengal. Appl.

Geochem. 15, 403–413.

Nicolli, H.B., Suriano, J.M., Gomez Peral, M.A., Ferpozzi,

L.H., Baleani, O.A., 1989. Groundwater contamination

with arsenic and other trace elements in an area of the

Pampa, Province of C�ordoba, Argentina. Environ. Geol.

Water Sci. 14, 3–16.

Pal, T., Mukherjee, P.K., Sengupta, S., 2002. Nature of arsenic

pollutants in groundwater of Bengal basin – a case study

from Baruipur area, West Bengal, India. Curr. Sci. 82, 554–

561.

Parker, J.W., Perkins, M.A., Foster, S.S.D., 1982. Groundwa-

ter quality stratification – its relevance to sampling strategy.

‘Commisie voor Hydrologisch Onderzoek’. TNO, Nether-

lands, Publication No. 31, pp. 43–54.

Ravenscroft, P., 2003. Overview of the hydrogeology of

Bangladesh. In: Rahman, A.A., Ravenscroft, P. (Eds.),

Groundwater Resources and Development in Bangladesh

– Background to the Arsenic Crisis, Agricultural

Potential and the Environment. University Press Ltd,

Dhaka.

Ravenscroft, P., McArthur, J.M., Hoque, B.A., 2001. Geo-

chemical and palaeohydrological controls on pollution of

groundwater by arsenic. In: Chappell, W.R., Abernathy,

C.O., Calderon, R.L. (Eds.), Arsenic Exposure and Health

Effects IV. Elsevier, Oxford, pp. 53–77.

Redman, A.D., Macalady, D.L., Ahmann, D., 2002. Natural

organic matter affects arsenic speciation and sorption onto

hematite. Environ. Sci. Technol. 36, 2889–2896.

Reddy, K.R., Diaz, O.A., Scinto, L.J., Agami, M., 1995.

Phosphorus dynamics in selected wetlands and streams of

the lake Okeechobee Basin. Ecol. Eng. 5, 183–207.

Reimann, K.-U., 1993. The Geology of Bangladesh. Gebruder

Borntraeger, Berlin.

Rhoton, F.E., Bigham, J.M., Lindbo, D.L., 2002. Properties of

iron oxides in streams draining the Loess Uplands of

Mississippi. Appl. Geochem. 17, 409–419.

J.M. McArthur et al. / Applied Geochemistry 19 (2004) 1255–1293 1293

Rittle, K.A., Drever, J.I., Colberg, P.J.S., 1995. Precipitation of

arsenic during bacterial sulphate reduction. Geomicrobiol.

J. 13, 1–11.

Robertson, F.N., 1989. Arsenic in ground-water under oxidis-

ing conditions, south-western United States. Environ. Geo-

chem. Health 11, 171–185.

Safiullah, S., 1998. CIDA Arsenic Project Report: Monitoring

and Mitigation of Arsenic in the Groundwater of Faridpur

Municipality. Jahangirnagar University, Dhaka, Bangla-

desh.

Smedley, P.L., Kinniburgh, D.G., 2002. A review of the source,

behaviour and distribution of arsenic in natural waters.

Appl. Geochem. 17, 517–568.

Smedley, P.L., Zhang, M., Zhang, G., Lou, Z., 2003. Mobil-

isation of arsenic and other trace elements in fluviolactus-

trine aquifers in the Huhhot Basin, Inner Mongolia. Appl.

Geochem. 18, 1453–1477.

Smith, A., Lingas, E., Rahman, M., 2000. Contamination of

drinking-water by As in Bangladesh. Bull. WHO 78, 1093–

1103.

Stanley, D.J., Hait, A.K., Jorstad, T.F., 2000. Iron-stained

quartz to distinguish Holocene deltaic from Pleistocene

alluvial deposits in small core samples. J. Coastal Res. 16,

357–367.

St€uben, D., Berner, Z., Chandrasekharam, D., Karmakar, J.,

2003. Arsenic enrichment in groundwater of West Bengal,

India: geochemical evidence for mobilization of As under

reducing conditions. Appl. Geochem. 18, 1417–1434.

Tanabe, S., Hori, K., Saito, Y., Haruyama, S., Vu, V.P.,

Kitamura, A., 2003. Song Hong (Red River) delta evolution

related to millennium-scale Holocene sea-level changes.

Quatern. Sci. Rev. 22, 2345–2361.

Tareq, S.M., Safiullah, S., Anawar, H.M., Rahman, M.M.,

Ishazuka, T., 2003. Arsenic pollution in groundwater: a self-

organizing complex geochemical process in the deltaic

sedimentary environment, Bangladesh. Sci. Total Environ.

313, 213–226.

Umitsu, M., 1987. Late quaternary environment and landform

evolution in Bengal. Geog. Rev. Japan B 60, 164–178.

Umitsu, M., 1993. Late quaternary environment and landforms

in the Ganges Delta. Sed. Geol. 83, 177–186.

van Geen, A., Zheng, Y., Versteeg, R., Stute, M., Horneman,

A., Dhar, R., Steckler, M., Gelman, A., Small, C., Ahsan,

H., Graziano, J.H., Hussain, I., Ahmed, K.M., 2003a.

Spatial variability of arsenic in 6000 tube wells in a 25 km2

area of Bangladesh. Water Resour. Res. 39, art. no. 1140.

van Geen, A., Zheng, Y., Stute, M., Ahmed, K.M., 2003b.

Comment on ‘‘Arsenic mobility and groundwater extraction

in Bangladesh’’. Science 300, 5619, 584C.

Welch, A.H., Lico, S.M., Hughes, J.L., 1988. Arsenic in

groundwater of the Western United States. Ground Water

26, 333–346.

Welch, A.H., Westjohn, D.B., Helsel, D.R., Wanty, R.B., 2000.

Arsenic in groundwater of the United States: occurrence

and geochemistry. Ground Water, 38–589-604.

WHO, 1993. Guidelines for Drinking Water Quality, vol. 1.

World Health Organization, Geneva.

Yu, W.H., Harvey, C.M., Harvey, C.F., 2003. Arsenic in

groundwater in Bangladesh: a geostatistical and epidemio-

logical framework for evaluating health effects, and

potential remedies. Water Resour. Res. 39, 1146–1163.

Zobrist, J., Dowdle, P.R., Davis, J.A., Oremland, R.S., 2000.

Mobilization of arsenite by dissimilatory reduction of

adsorbed arsenate. Environ. Sci. Technol. 34, 4747–4753.