Arsenic in groundwaters of the Lower Mekong

Luyen4& Tuyen Tran Thanh5 1 Australia-based water resources consultant, currently with the UNDP, Box 551, Sanaa´, Yemen 2 Sub-Institute for Water Resources Planning, 253A An Duong Vuong,

Arsenic in groundwaters of the Lower Mekong

Gordon Stanger1,6, To VanTruong2, K.S Le Thi My Ngoc3, T.V Luyen4& Tuyen Tran Thanh5

1

Australia-based water resources consultant, currently with the UNDP, Box 551, Sanaa´, Yemen

2

Sub-Institute for Water Resources Planning, 253A An Duong Vuong, Quan 5, TP Ho Chi Minh, Vietnam 3

Sub Institute of Hydrometeorology of South Vietnam, Phu Trach Tram, Tram Thuc Nghiem, KTTVNN, Dong Bang, Song Cuu Long, Vietnam

4

Centre for Nuclear Techniques, 217 Nguyen Trai St., Quan 1, TP Ho Chi Minh, Vietnam

5

Department of Environmental and Natural Resources Management, Can Tho University, 3/2 Street, TP Can Tho, Vietnam

6

Author for correspondence (e-mail: hydrodocgeo@yahoo.co.uk)

Received 14 October 2004; Accepted 17 March 2005

Key words: arsenic, arsenicosis, Cambodia, Cuu Long Delta, groundwater, Mekong, Vietnam

Abstract

Increasing incidence and awareness of arsenic in many alluvial aquifers of South-east Asia has raised concern over possible arsenic in the Lower Mekong Basin Here, we have undertaken new research and reviewed many previous small-scale studies to provide a comprehensive overview of the status of arsenic in aquifers of Cambodia and the Cuu Long Delta of Vietnam In general natural arsenic originates from the Upper Mekong basin, rather than from the local geology, and is widespread in soils at typical concentrations of between 8 and

16 ppm (dry weight) Industrial and agricultural arsenic is localised and relatively unimportant compared to the natural alluvial arsenic Aquifers most typically contain groundwaters of no more than 10 lg L)1, although scattered anomalous areas of 10 to 30 lg L)1are also quite common The most serious, but possibly ephemeral arsenic anomalies, of up to 600 lg L)1, are associated with iron and organic-rich flood-plain sediments subject to very large flood-related fluctuations in water level, resulting in transient arsenopyrite dissolution under oxidizing conditions In general, however, high-arsenic groundwaters result from the competing interaction between sorption and dissolution processes, in which arsenic is only released under reducing and slightly alkaline conditions High arsenic groundwaters are found both in shallow water-tables, and in deeper aquifers of between 100 and 120 m depth There is no evidence of widespread arsenicosis, but there are serious localised health-hazards, and some risk of low-level arsenic ingestion through indirect pathways, such as through contaminated rice and aquaculture An almost ubiquitous presence of arsenic in soils, together with the likelihood of greatly increased groundwater extraction in the future, will require continuing caution in water resources development throughout the region

Introduction

In recent years, high-profile discussion of the

arsenic problems in Bangladesh, West Bengal and

Nepal has raised fears of impending arsenic

problems in other regions of comparable

hydrog-eology High amongst these concerns is the Cuu

Long (Mekong) Delta in Southern Vietnam, and

adjacent areas of Cambodia where the Lower

Mekong and its tributaries connect with significant alluvial aquifers deposited by the ‘palaeo-Mekong
and other rivers The arsenic threat has been recognised since the late 1990s, and many studies have been undertaken by regional and central government departments and concerned NGOs, partly using off-the-shelf test kits, and partly util-ising laboratory facilities in Phnom Penh, Can Tho and Ho Chi Minh Cities This paper is an attempt DOI: 10.1007/s10653-005-3991-x

to draw together the existing data from numerous

sources, supplemented by our own field and

lab-oratory work, to ‘fill in the gaps
The overall

finding is that of a less dire arsenic distribution

than in Bangladesh, Uttar Pradesh, Nepal and the

Red River Basin of Northern Vietnam

Neverthe-less, there are localised arsenic ‘hot-spots
, a more

general risk of low-level arsenic ingestion

(argu-ably sub-clinical in effect), additional potential

pathways of arsenic exposure through the food

chain, and a potential for increasing arsenic

con-centrations in groundwater over time In short,

there are a few sites requiring urgent attention, and

a generally low-level problem elsewhere The latter

requires continuing vigilance, but is of relatively

minor importance compared to microbiological,

PAH1and organophosphate pesticide

contamina-tion in Cambodia and Vietnam

Geology, hydrogeology and drainage of the Lower

Mekong

The Lower Mekong in Cambodia and Vietnam

naturally divides into three geologically and

physiographically distinct areas; the main channel

area, the Tonle Sap, and the Cuu Long Delta (i.e

the Mekong Delta)

Between the Lao-Cambodian border and

Phnom Penh the flood plain is narrow to

com-pletely absent, with heavily forested hills of up to

600 m in elevation constricting the Mekong River

on both banks These hills are of Triassic age,

comprising partly basalts and partly sediments of

the Lower- to Mid-Indosinias group In this area,

the Indosinias consists of andesitic and dacitic

lavas with predominantly continental sandstones,

silty shales, red shales, marls and conglomerates

with subordinate breccias Gleyic and ferralic

cambisols, or rhodic and humic ferralsols have

developed on these sediments and volcanics, with

several localised low-yielding aquifers being

tapped by shallow wells Six wells from this

envi-ronment were tested but none exhibited significant

arsenic concentrations, (i.e >10 lg L)1) Only

further downstream, where the flood plain begins

to broaden, does significant arsenic begin to

appear in groundwaters This arsenic is restricted

to areas where sedimentary deposition clearly originates from the Mekong

To the north-west of Phnom Penh the ‘reversible tributary
of the Mekong, the Tonle Sap River, drains into, and out of, the Tonle Sap Lake (the

‘Great Lake
) Thirteen other river basins, com-prising some 60% of Cambodia, also drain into this lake from the highlands which form the perimeter of the north-western provinces (Figure 1) To the south-west of the Great Lake the four main river basins consist of uninhabited and densely forested mountains up to 1500 m high The highland headwaters of these catch-ments are of heterogeneous facies comprising Cretaceous to Palaeocene sandstones and con-glomerates of the Upper Indosinias, Triassic sandstones, conglomerates, tuffs and shales of the Mid-Indosinias, and Devonian quartzites, shales, schists and gneiss None of these yield sediments containing arsenic-rich groundwaters However, the eastern part of the massif is dominated by the

‘Aoral granite
and Triassic marine shales The latter are presumed to be marine in origin, and possibly arsenic-rich, although suspected mine-fields prevented close inspection and sampling Northward drainage from this area has deposited more than 150 m of low-lying intercalated iron-rich clays and coarse quartz sands associated with typical groundwater arsenic concentrations of about 20 lg L)1 This anomaly is marked ‘A
in Figure 2 It is a moot point whether the arsenic originates from a palaeochannel of the Mekong or from arsenious shales in the upper catchment Perhaps there is intercalation of sediments from both sources, but drilling records suggest that a relatively local provenance is more likely

If this is so, then the ‘Pursat–Kampong Chhn-ang
arsenic anomaly, ‘A
, is the only known case

of natural non-Mekong arsenic throughout Cam-bodia and Southern Vietnam Of the 40 wells tes-ted in this area the highest arsenic concentration was 30 lg L)1

North-east of the Tonle Sap the hills are rela-tively low, occasionally reaching about 500 m elevation, and merge eastwards with the forested hill country of the Mekong proper Likewise the geology is of the more easterly Indosinias facies, comprising quartzite, sandstone and basalt Groundwater sampling in this area was sparse, and arsenic concentrations were all less than

10 lg L)1, with the majority being below the

1 Polycyclic aromatic hydrocarbons, in this case mainly

con-sisting of dioxin residues.

detection limit Red clay and lateritic sediments

are widespread in this area, providing a potential

substrate for arsenic adsorption, but no evidence

was found for high arsenic within or downstream

of these northern hills

Between the south-western mountains and

north-eastern hills lies the sedimentary depression

of the Great Lake and surrounding low-lying

alluvium This basin is partially fault-bounded on

its south-western edge, but is of unknown

thick-ness A roughly concentric synclinal pattern of

sedimentation surrounds the lake, with older

coarser ferruginous silts, sands and grits around

the perimeter conformably overlain by younger

red-clayey and silty sediments towards the lake During the monsoon season, from about June to October, flood waters from the Mekong back up into the lake, which then resumes normal drainage back into the Mekong between November and March This results in a seasonal lake-level fluc-tuation of between 6.7 and 8.5 m in amplitude Rainfall into and evaporation from the lake are closely balanced so the dry-season outflow of the Tonle Sap closely matches the combined inflow of the Tonle Sap and the 13 other circum-lacustrine river basins The sediment flux appears to be less well balanced, although available data is some-what imprecise Carbonnel and Guiscafre
s Fig 1 Geography, and distribution of actual and potential acid sulphate soils.

estimates (1963) were an inflow of 4.6 million

tonnes of sediment, of which the Tonle Sap River

contributed 2.7 million tonnes (58%) A more

recent estimate by the Mekong River Commission

cites an average annual inflow to the ‘Great Lake

of 5.7 million tonnes, of which the Tonle Sap

River remobilises a seasonal outflow of 4.7 million

tonnes (82%), thereby yielding between 0.10 and

0.23 mm of deposition per year, depending upon

the degree of compaction and effective area of

deposition This compares with Carbonnel and

Guiscafre
s estimates of 0.3 mm of deposition per year, averaged over the previous 5000 years Both estimates indicate a net annual gain of Mekong-derived argillaceous sediment, and hence the possibility of ‘modern
arsenic importation of far-upstream provenance However, there is an alternative possibility Tectonic evidence from the ‘Golden Triangle
area (of Thailand– Myanmar–Laos) suggests that the Mekong River configuration has remained stable for about

5 million years, prior to which the palaeo-Mekong Fig 2 Distribution of arsenic concentrations in groundwater.

entered the Gulf of Thailand somewhat to the east

of Bangkok During the transition from the old to

the new channel configuration it is very likely that

it flowed through the north-west Cambodian/Thai

border, and thence along the current axis of the

‘Great Lake
and into the Cuu Long Delta Whilst

details of this palaeo-channel are sparse, it

would explain the low relief along that part of the

Thai–Cambodian border, the slight arsenic

groundwater anomaly (marked ‘B
on Figure 2),

and the distribution of low concentrations of

arsenic throughout sediments of the northwest

Cambodian provinces

Further downstream there is no clearly defined

apex to the Cuu Long Delta Rather, the broad

floodplain downstream of Phnom Penh broadens

still further to become the Delta at about the

Cambodian–Vietnam border All of the very high

arsenic concentrations encountered in this survey,

i.e >100 lg L)1, occur within this

floodplain-delta complex

At 144 to 160 million tonnes per year the

Mekong has the seventh largest sediment load of

any river (Ta et al 2002a/b) This heavy

sedimentary load has been deposited as the Cuu

Long Delta through numerous cycles of marine

regression and transgression, the latter

corre-sponding to Pleistocene glacial and interglacial

sea-level stands, respectively Hence, this alluvial

flat, the world
s third largest flood-plain delta, is

geologically very young Little is known of the

sub-alluvial topography Within the floodplain at

the Cambodian–Vietnam border the sediment is as

little as 40 m deep, whereas it is at least 650 m

thick in the Ben Tre area (the outer delta)

Attempts to rationalise the available

hydrogeo-logical data suggest that there are four accessible

silty and/or sandy aquifers sandwiched between

thicker clay-rich aquicludes These are:

• QRRecent thin shallow to surface sediments

• QIV Holocene, incised river silts and coastal

sand dunes up to 20 m thick

• QII–III Upper to mid Pleistocene, freshwater

to saline, 10–30 m thick, widely used

• QI Lower Pleistocene, freshwater to saline,

underlies about 90% of the delta

Some arsenic is associated with all four of these

aquifers Deeper Neogene aquifers exist but are

seldom exploited and nothing is known of their

arsenic risk Neither the extent of lateral aquifer

continuity, nor the effectiveness of vertical hydraulic separation between aquifers are adequately known Comparison with better-known deltaic environments suggests that the sedimentology is almost certainly more complex than indicated on the available litho-stratigraphic sections

The current delta, from the surface to a depth of between )10 and )20 m has all been deposited during the past 5000 years (Ta et al 2001; Tanabe

et al 2003) Hence during the last glacial maxi-mum high stand the coast was some 250 km further inland, near the Cambodian border Prior

to this, several cycles of erosion and deposition have buried subaerial laterites and floodplain silts

in inland areas, together with beach sands and mangrove swamps and incised channels along a prograding shoreline This has resulted in many localised areas of oxidised palaeosols and more widespread strongly reducing organic clayey sediments, all of which potentially provide substantial adsorption substrates for arsenic An association of finely disseminated pyrite within the organic cambisols and gleysols has facilitated the development of widespread ferruginous acid-sul-phate soils, which have becoming a problem over many parts of the delta (Figure 1)

Arsenic in soils

Industrial sources of arsenic in the Lower Mekong are trivial in comparison to natural sources Huy et al (2002) noted slightly elevated arsenic concentrations in sediments of the Luong Canal, which drains the industrial area of northwest Ho Chi Minh City, but their maximum observed enrichment factor relative to the back-ground concentration was only 8.2 at a single sampling point A potential arsenic source is from fertilisers, since arsenic is readily adsorbed onto phosphate, and much of the phosphate fertiliser used in the Cuu Long Delta is from the Red River basin of Northern Vietnam, which is known to be arsenic-rich To investigate the possibility of such artificial arsenic enhancement a set of surface soils from an area of variable fer-tiliser use in rice paddies between O
Mon and Can Tho was analysed by neutron activation The results are plotted in Figures 3 and 4 Apart from

a single unexplained phosphorus anomaly, the

lack of any correlation between arsenic and

phosphorus suggests that arsenic from fertiliser is

not a significant factor

A better, albeit still weak relationship, exists

between total iron and arsenic, Figure 4 This

could be related to the natural presence of either

finely disseminated arsenious pyrite or to arsenic

adsorbtion onto an iron oxy-hydroxide substrate

It is evident from the distribution of arsenic in

groundwater that the source of arsenic is more or

less ubiquitous throughout the soils and sediments

of the region No primary source of arsenic

min-eralization has yet been identified, but we have

analysed major and trace element concentrations,

by neutron activation, from 129 soils and near-surface sediments, including two modern sus-pended-sediment samples from the Bassac river, the second largest of the eight distributaries of the delta, (see Figure 1) The geographic distribution

of soil-arsenic concentrations within the delta is fairly uniform with no discernible clusters of high concentrations The histogram of concentrations is shown in Figure 5

Arsenic is present in surface or near-surface soils

at concentrations varying from 3 to 47 ppm, but the great majority of measurements lie within the range of 8–16 ppm Hence, aside from a few high values of >25 ppm, the natural variation only

0.01 0.10 1.00 10.00

Arsenic (ppm)

Fig 3 Phosphate versus As in surface soils from the O
Mon–Can Tho farming area.

0.0 1.0 2.0 3.0 4.0 5.0

Arsenic (ppm)

Fig 4 Total iron versus arsenic in surface soils from the O
Mon–Can Tho farming area.

exceeds the measurement error by a factor of

about four The average inter-sampling distance in

this survey was about 30 km, which yielded no

discernible pattern of concentration across the

delta About 35 soil samples were also taken from

north-east of the Mekong Delta, in and around Ho

Chi Minh City, where smaller rivers clearly derive

their sediments more locally (from Binh Phuoc and

Tay Ninh provinces) These showed a marked

contrast in soil-arsenic concentrations (Table 1)

Between the Mekong Delta sensuo stricto and

the north-eastern river basins lies the ‘Plain of

Reeds
, an area strongly affected by the Mekong
s

annual flood, but which may partially have derived

sediment from other sources in the past One of the

key features of the Plain of Reeds is the high

prevalence of acid sulphate soils, with water pHs

as low as 2.9 The obvious presence of widespread

sulphides, oxidising at or near the surface as part

of a jurbanite-acidic-sulphate controlled

equilib-rium system2 raises the suspicion that aqueous

arsenic might be derived from the dissolution of

arsenious pyrite, and conditionally precipitated

with jarosite Acid sulphate soils affect some 41%

of the Cuu Long Delta, and contain abundant

jarosite, (K,Na)Fe3Æ(SO4)2Æ(OH)6, in which

arse-nate anions are known to substitute for sulphate

(Savage et al 2000) However, there is insufficient

evidence to conclude that arsenic concentrations

from acidic soils are significantly different from

those from non-acidic areas of the delta Rather, it appears that all of the Mekong sediments, with or without pyrite and jarosite, (i.e all parts of the Delta), are naturally arsenic-rich

With such large inter-sampling distances, the scale of natural variation, and hence of the typi-cality of samples, was considered To test this variation, samples were taken from 18 farms in the O
Mon area near Can Tho City, with an inter-sample distance of about 5 km Unlike point-sampling from the main survey, composite soil samples were taken at each farm, from predomi-nantly rice paddies These samples were homog-enised prior to INAA analysis, to give an average value over tens of hectares

The results, plotted in Figure 6, are only slightly more uniform than the main regional sampling This degree of local variation is probably influenced by differential ‘dilution
with organic matter in such intensively cultivated top-soils

0 5 10 15 20 25 30 35

Range (ppm)

Fig 5 Frequency histogram of surface soil arsenic concentrations in the Cuu Long Delta and adjacent alluvial areas, (mg Kg)1 dry weight).

Table 1 Contrast in soil-arsenic concentrations.

Mean sediment concentrations, total

As ppm

58 samples from the Mekong Delta

13.8

35 samples from river basins to the north-east of the delta

4.8

5 samples from the Plain of Reeds, and downstream

15.0

2 samples of suspended sediment from the Bassac river

9.5

2 Jurbanite, Al(SO4)ÆOHÆ5H2O is the phase controlling the

equilibrium water chemistry.

The arsenic distribution map, Figure 2, is based

upon 932 water quality analyses collected between

1998 and 2003 as part of 12 different sources and

studies3 Some of these were purely field studies

using ‘Hach
field test kits, but in most instances

semi-quantitative field indications of greater than

10 lg L)1 (‘ppb
) were followed-up by more

accurate laboratory analysis upon HNO3)spiked,

freshly pumped groundwater samples The classes

of groundwater arsenic concentrations depicted in

Figure 2 are based upon the WHO revised

‘acceptable limit
of 10 lg L)1 This limit arises

more from the constraints of realistic measurement

than from any proven limit of ‘no-adverse effect
,

although in practice concentrations below this limit are probably safe

The sampling distribution was uneven, being pragmatically based upon the availability of accessible dug wells and boreholes As found in numerous other studies, in Bangladesh, Assam and Nepal, analyses from clusters of wells in villages and towns do not yield consistent or spatially uniform results Rather, it was commonly found that one or two groundwater samples would yield greater than 10 lg L)1 total arsenic within a local scatter of wells with less than detectable concentrations4 That is within about

1 km2 There were areas of very low groundwater sampling density within the northern and western Cuu Long Delta (in Vietnam) This is partly due to difficulty of access, but mainly because the abun-dance of surface water, and the expense of sinking wells, does not yet justify the development of groundwater Consequently there are no existing wells to sample However, as surface pollution and dry-season water-demand increases, the future

Fig 6 Variation in soil–arsenic concentrations from the farmed area of O
Mon-Can Tho.

3

These comprise (1) The AusAID-funded Vietnam–Australia

Water Resources Management Assistance Project, Component

3, 2003, (2) NWIS Project (ADB) TA-3758 Cambodia, 2002, (3)

The World Bank-funded SIWRP reconnaissance data for

arsenic in Groundwaters of the CLD, (4) Haskoning and

ARCADIS Euroconsult, Groundwater Study of the Mekong

Delta, (1999) (5) JICA, The Study of Groundwater

Develop-ment in Central Cambodia, 2002, (6–9) Unpublished data from

the Pursat et al Meanchey Departments of Rural Water Supply,

Cambodia, (10) Unpublished data from the Cambodian

Min-istry of Rural Development, (11) the AusAID-funded Cuu

Long Delta Regional WSS Project, and (12) new data collected

and analysed by the authors of this paper.

4 For laboratory analyses the limit of detection was estimated at 1.4 lg L)1 For field kits the limit of detection is subjective, but approximately 5 lg L)1total arsenic.

demand for groundwater is likely to increase very

substantially, thus requiring a continuing vigilance

to deal with rising arsenic at an early stage

In Cambodia there are few deep boreholes apart

from a few town water supplies although there is

an abundance of shallow dug wells in most

prov-inces This has facilitated a reasonably

represen-tative spatial sampling throughout the lowland

agricultural areas, and to a lesser extent, along the

Mekong riparian zone as far north as Kratie The

only area with unsatisfactory coverage is in Siem

Reap, north of the ‘Great Lake
Overall, the

average sampling densities were 1 per 101 km2in

the Cuu Long Delta, and 1 per 111 km2 in the

inhabited parts of Cambodia The total studied

area was 105km2

For the purposes of this reconnaissance study, localities in which at least one sample was greater than 10 lg L)1 total arsenic are mapped at this concentration, even though the same area may have contained a majority of samples at less than

10 lg L)1 Three factors, the widespread occur-rence of aqueous arsenic concentrations at close to the limit of detection, the local soil and ground-water variation in arsenic concentrations, and the consistency of the fluvial source area, all suggest that the concentrations of arsenic are governed more by processes of arsenic mobilisation than by its geochemical availability

Literature from better known and longer stud-ied areas (BGS, 1999); McArthur et al 2004) seem

to have reached a consensus that there are two

Fig 7 Distribution of the annual flooding depth of the Mekong–Tonle Sap system.

main processes of arsenic mobilization in shallow

non-thermal groundwater environments, namely

(a) oxidation of arsenious pyrite from microscopic

pyrite/arsenopyrite framboids, and (b) release of

adsorbed arsenic through the process of reductive

dissolution of FeOÆOH In Bangladesh, Bengal,

Nepal and the Red River basin of Northern

Viet-nam the second of these processes is now regarded

as by far the more important (Berg et al 2001;

Smedley 2003; McArthur et al 2004) In the Lower

Mekong, however, there is circumstantial evidence

to suggest that oxidation has at least a

contribut-ing influence upon the highest arsenic anomalies in

groundwater, measured in hundreds of

micro-grams per litre, and may also be a secondary

process contributing to the episodic release and

continuing mobility of arsenic over wide areas at

concentrations of tens of micrograms per litre

Comparison of Figures 2 and 7 show that there

is some correspondence between the depth of

annual flooding and the occurrence of arsenic

groundwater anomalies In particular, the annual

monsoonal flood of the Mekong results in a stage

amplitude variation of between 8 and 11 m at

Phnom Penh, decreasing to about 3 m, some

125 km further downstream at Chau Doc, i.e at

the delta
s apex, just south of the

Cambodia-Vietnam border Typical hydrographs are shown

in Figure 8

In response to this flooding, there is an annual cycle in groundwater levels, which varies from at least 6 m in some riparian environments, to only centimetres in more distant parts of the floodplain The sediments are mostly organic-rich, originating from both former reeds and mangroves, and modern cultivation, with anoxic preservation at greater depth Therefore, there is likely to be an annual cycle of near-surface oxidising and reduc-ing conditions, more or less correspondreduc-ing to the dry and wet seasons The oxidising conditions would, of course, be consistent with pyrite oxida-tion and partial dissoluoxida-tion Against this lies two strands of evidence regarding redox, and potential substrates

Firstly, Figure 9 illustrates the relationship between redox potential and groundwater arsenic concentrations At low concentrations of up to

10 lg L)1, the arsenic appears to be stable across virtually the entire redox spectrum – a range of some 500 mV On the other hand, high arsenic concentrations are obviously dominated by reducing, or only slightly oxidizing conditions Given the availability of arsenic in virtually all soils, Figure 9 suggests that some other major co-variable is required to mobilise high concen-trations of arsenic Curiously the iron–redox relationship was much less clear-cut, with 1 to

10 mg L)1 total iron across the same redox

0.0 1.0 2.0 3.0 4.0 5.0 6.0

1

-1

r a M -1

r p -1

y M

-un 1 l u J -l u J -1 3

g A

p e S -0

t c O

-o v

-1 D c

M onth

Fig 8 Typical annual stage hydrographs of the Mekong river at Chau Doc.