Content uploaded by Pham Hung Viet
Author content
All content in this area was uploaded by Pham Hung Viet on Jan 20, 2014
Content may be subject to copyright.
LETTER
doi:10.1038/nature12444
Retardation of arsenic transport through a
Pleistocene aquifer
Alexander van Geen
1
, Benjamı
´
n C. Bostick
1
, Pham Thi Kim Trang
2
, Vi Mai Lan
2
, Nguyen-Ngoc Mai
2
, Phu Dao Manh
2
,
Pham Hung Viet
2
, Kathleen Radloff
1
{, Zahid Aziz
1
{, Jacob L. Mey
1,3
, Mason O. Stahl
4
, Charles F. Harvey
4
, Peter Oates
5
,
Beth Weinman
6
{, Caroline Stengel
7
, Felix Frei
7
, Rolf Kipfer
7,8
& Michael Berg
7
Groundwater drawn daily from shallow alluvial sands by millions of
wells over large areas of south and southeast Asia exposes an esti-
mated population of over a hundred million people to toxic levels of
arsenic
1
. Holocene aquifers are the source of widespread arsenic
poisoning across the region
2,3
. In contrast, Pleistocene sands depo-
sited in this region more than 12,000 years ago mostly do not host
groundwater with high levels of arsenic. Pleistocene aquifers are
increasingly used as a safe source of drinking water
4
and it is there-
fore important to understand under what conditions low levels of
arsenic can be maintained. Here we reconstruct the initial phase of
contamination of a Pleistocene aquifer near Hanoi, Vietnam. We
demonstrate that changes in groundwater flow conditions and the
redox state of the aquifer sands induced by groundwater pumping
caused the lateral intrusion of arsenic contamination more than 120
metres from a Holocene aquifer into a previously uncontaminated
Pleistocene aquifer. We also find that arsenic adsorbs onto the aqui-
fer sands and that there is a 16–20-fold retardation in the extent of
the contamination relative to the reconstructed lateral movement of
groundwater over the sameperiod. Our findings suggest thatarsenic
contamination of Pleistocene aquifers in south and southeast Asia as
a consequence of increasing levels of groundwater pumping may
have been delayed by the retardation of arsenic transport.
This study reconstructs the initial phase of contamination of an
aquifer containing low levels of arsenic (low-As) in the village of Van
Phuc, located 10 km southeast of Hanoi on the banks of theRed River.A
key feature of the site is the juxtaposition of a high-As aquifer upstream
of a low-As aquifer in an area where pumping for the city of Hanoi has
dominated lateral groundwater flow for the past several decades
(Fig. 1a). Many residents of the village of Van Phuc still draw water
from their 30–50-m-deep private wells. In the western portion of the
village, the wells typically contain less than 10 mg of As per litre of water
and therefore meet the World Health Organization guideline for As in
drinking water, whereas As in the groundwater from most wells in
eastern Van Phuc exceeds this guideline by a factor of 10–50 (ref. 5).
Drilling and sediment dating in the area has shown that low-As
groundwater is drawn from orange-coloured sands deposited over
12,000 years ago, whereas high-As groundwater is typically in contact
with grey sands deposited less than 5,000 years ago
6,7
. We examined to
what extent the boundary between the low-As and high-As aquifers of
Van Phuc has shifted in response to groundwater withdrawals in
Hanoi. This large-scale perturbation spanning several decades has
implications for low-As aquifers throughout Asia that are vulnerable
to contamination owing to accelerated groundwater flow.
The collection of sediment cores and the installation of monitoring
wells was concentrated along a transect trending southeast to north-
west that extends over a distance of 2.2 km from the bank of the Red
River (Fig. 1b). Groundwater heads, and therefore the groundwater
velocity field, withinVan Phuc respond rapidlyto the daily and seasonal
cycles in the water level of the river (Supplementary Information).
Before large-scale groundwater withdrawals, rainfall was sufficient to
maintain groundwater discharge to the river, as is still observed else-
where along the Red River
8
. In Van Phuc, however, the groundwater
level was on average 40 cm below that of the water level of the Red River
in 2010–11 and the hydraulic gradient nearly always indicated flow
from the river into the aquifer. The reversal of the natural head gradient
is caused by the large depression in groundwater level centred 10 km to
the northwest that induces groundwater flow along the Van Phuc
transect from the river towards Hanoi (Fig. 1a). This perturbation of
groundwater flow is caused by massive pumping for the municipal
water supply of Hanoi
9–11
, which nearly doubled from 0.55 million to
0.90 million cubic metres per day between 2000 and 2010 owing to the
rapid expansion of the city (Supplementary Fig. 1).
A change in the colour of a clay layer capping sandy sediment along
the transect defines a geological boundary between the two portions of
the Van Phuc aquifer. Up to a distance of 1.7 km from the river bank,
the clay capping the aquifer is uniformly grey with the exception of a
thin brown interval at the very surface (Fig. 2b). In contrast, a readily
identifiable sequence of highly oxidized bright yellow, red and white
clays was encountered between 12 m and 17 m depth at all drill sites
along the transect beyond a distance of 1.7 km from the river bank.
This oxidized clay layer is probably a palaeosol dating to the last sea-
level low-stand about 20,000 years ago
7,12
.
The colour of aquifer sands below the upper clay layer also changes
markedly along the Van Phuc transect. Sand colour in fluvio-deltaic
deposits is controlled primarily by the extent to which Fe(
III)hasbeen
reduced to Fe(
II) by the decomposition of organic carbon
13
.Uptoa
distance of 1.6 km from the river bank, sandy drill cuttings within the
20–40 m depth range are uniformly grey. The predominance of orange
sands beyond 1.6 km indicates oxidation during the previous sea-level
low-stand. After the sea level rose back to its current level, the nature of
the remaining organic carbon precluded a new cycle of Fe(
III) reduction
14
.
Independently of sediment colour, the calcium (Ca) content of sand
cuttings collected while drilling along the Van Phuc transect confirms
that a geological boundary extends to the underlying aquifer sands.
Within the southeastern portion of the aquifer that is not capped by the
presumed palaeosol, X-ray fluorescence measurements indicate Ca
concentrations of over 2,000 mg Ca per kg of sand in cuttings to a
depth of 30 m (Fig. 2a). The groundwater in this portion of the aquifer
is supersaturated with respect to calcite and dolomite
6
, suggesting that
authigenic precipitation is the source of Ca in the grey drill cuttings, as
previously proposed elsewhere
12
(Supplementary Fig. 2). At a distance
of 1.7 km from the river and further to the northwest, instead, the Ca
1
Lamont-Doherty Earth Observatory (LDEO), Columbia University, Palisades, New York 10964, USA.
2
Research Centre for Environmental Technology and Sustainable Development (CETASD), Hanoi
University of Science, Vietnam National University, Hanoi, Vietnam.
3
Department of Physical Sciences, Kingsborough Community College, Brooklyn, New York 11235, USA.
4
Massachusetts Institute of
Technology, Cambridge, Massachusetts 02139, USA.
5
Anchor QEA, Montvale, New Jersey 07645, USA.
6
Earth and Environmental Sciences, Vanderbilt University, Nashville, Tennessee 37235, USA.
7
Eawag,
Swiss Federal Institute of Aquatic Science and Technology, 8600 Du
¨
bendorf, Switzerland.
8
Institute of Geochemistry and Petrology, Swiss Federal Institute of Technology, Zurich ETHZ 8092, Switzerland.
{Present addresses: Gradient, Cambridge, Massachusetts 02138, USA (K.R.); Sadat Associates, Trenton, New Jersey 08610, USA (Z.A.); Earth and Environmental Sciences, California State University,
Fresno, California 93740, USA (B.W.).
204 | NATURE | VOL 501 | 12 SEPTEMBER 2013
Macmillan Publishers Limited. All rights reserved
©2013
content of orange sand cuttings systematically remains lessthan 100 mg
Ca per kg and the groundwater is undersaturated with respect to calcite
and dolomite. Unlike surficial shallow grey clays, the Ca content of the
presumed palaeosol is also very low (,100 mg Ca per kg) and consist-
ent with extensive weathering.
The redox state of the aquifer has a major impact on the composi-
tion of groundwater in Van Phuc, as reported elsewhere in Vietnam
15
and across south and east Asia more generally
3
. High but harmless
Fe(
II) concentrations in groundwater (10–20 mg per litre) associated
with grey reducing sediments are apparent to residents of eastern Van
Phuc as an orange Fe(
III) precipitate that forms in their water upon
exposure to air (Supplementary Fig. 3). In contrast, the high and toxic
concentrations of As in groundwater at 20–30 m depth within the
same portion of the transect, ranging from 200 mg per litre near the
river to levels as high as 600 mg per litre at 1.2–1.6 km from the river
bank, are invisible (Fig. 2c). The groundwater in contact with Pleisto-
cene sands in northwestern Van Phuc is also anaerobic but contains
less than 0.5 mg Fe(
II) per litre and less than 10 mg As per litre and
shows little indication of organic carbon mineralization compared to
the Holocene aquifer (Supplementary Fig. 4).
The Pleistocene portion of the Van Phuc aquifer adjacent to the
Holocene sediment is not uniformly orange or low in As. Of particular
interest is a layer of grey sand at 25–30 m depth extending to the
northwest at a distance of 1.7–1.8 km from the river bank (Fig. 2b).
The intercalation of grey sand between orange sands above and below,
combined with the low Ca content of sand cuttings within this layer,
indicate that it was deposited during the Pleistocene and therefore
until recently oxidized and orange in colour. Within the portion of
the Pleistocene aquifer that became grey and is closest to the geological
boundary, groundwater As concentrations are therefore presumed to
have been originally very low (,5 mg per litre). Actual As concentra-
tions of 100–500 mg per litre, as high as in the adjacent Holocene
aquifer, indicate contamination extending over a distance of about
120 m into the Pleistocene aquifer (Fig. 3a).
A subset of the transect wells was sampled in 2006 and analysed for
tritium (
3
H) as well as noble gases in order to measure groundwater
ages and determine the rate of As intrusion into the Pleistocene aqui-
fer. Atmospheric nuclear weapons testing in the 1950s and 1960s is the
main source of
3
H that entered the hydrological cycle
16
. The distri-
bution of
3
H indicates that only groundwater in the southeastern high-
As portion of the aquifer contains a plume of recharge dating from the
1950s and later. Concentrations of
3
He, the stable decay product of
3
H,
were used to calculate groundwater ages for eight wells in the 24–42-m
depth range with detectable levels of
3
H. In 2006, the oldest water dated
by the
3
H–
3
He method (Supplementary Fig. 5) was sampled at a dis-
tance of 1.6 km from the river, which is the most northwestern location
along the transect where the aquifer is uniformly grey (Fig. 2b, d).
Younger ages of 15 years and 17 years were measured closer to the
river at 1.3 km and 1.5 km, respectively. Concentrations of
3
H, ground-
water
3
H–
3
He ages, and hydraulic head gradients consistently indicate
that the Holocene aquifer has been recharged by the river from the
southeast within the past few decades.
Drilling and geophysical data indicate that the main groundwater
recharge area extends from the centre of the Red River to the inland
area where a surficial clay layer thickens markedly, that is, from 100 m
southeast to 300 m northwest of the river bank (Supplementary Fig. 6).
The relationship between groundwater ages and travel distance from
the recharge area implies accelerating flow drawn by increased Hanoi
pumping (Supplementary Fig. 7). A simple transient flow model
for the Van Phuc aquifer yields average advection rates of 38 m yr
21
and 48 m yr
21
towards Hanoi since 1951 and 1971, respectively (Sup-
plementary Discussion). According to these two pumping scenarios,
–10 m
–12 m
–14 m
60º
240º
30º
210º
0
180º
330º
150º
300º
120º
270º
90º
Van Phuc
Red River
Hanoi
2 km
N
a
b
c
10,000
10,000
30,000
30,000
20,000
20,000
10,000
30,000
20,000
*
*
*
Figure 1
|
Map of the Hanoi area extending south to the study site.
a, Location of the village of Van Phuc in relation to the cone of depression
formed by groundwater pumping for the municipal water supply of Hanoi
(white contours, adapted from ref. 10). Urbanized areas are shown in grey;
largely open fields are shown in green. b, Enlarged view of Van Phuc (box shows
location in a) from Google Earth showing the location of the transect along
which groundwater and sediment were collected, with tickmark labels
indicating distance from the Red River bank in kilometres. Symbol colour
distinguishes the uniformly grey Holocene aquifer (red), the Pleistocene aquifer
contaminated with As (yellow), the Pleistocene aquifer where the groundwater
conductivity and dissolved inorganic carbon concentrations are high but As
concentrations are not (green), and the Pleistocene aquifer without indication
of contamination (blue), all within the 25–30-m depth interval. Three white
asterisks identify the wells that were used to determine flow direction. Image
copyright 2012 Digital Globe Google Earth. c, Rose diagram frequency plot of
the head gradient direction based on data collected at 5-min intervals (numbers
indicate the number of observations) from these three wells from September
2010 to June 2011.
LETTER RESEARCH
12 SEPTEMBER 2013 | VOL 501 | NATURE | 205
Macmillan Publishers Limited. All rights reserved
©2013
groundwater originating from the Holocene portion of the aquifer was
transported 2,000–2,300 m intothe Pleistocene sands by 2011, when the
transect was sampled for analysis of As and other groundwater constituents.
The sharp decline in As concentrations between 1.60 km and
1.75 km from the river bank indicates that migration of the As front
across the geological boundary was retarded by a factor of 16 to 20
relative to the movement of the groundwater (Fig. 3a). Without
retardation, attributable to As adsorption onto aquifer sands, the entire
Pleistocene aquifer of Van Phuc would already be contaminated. The
retardation is derived from several decades of perturbation and is at the
low end of previous estimates by other methods, typically measured
within days to weeks
17–22
, and therefore predicts greater As mobility
than most previous studies. The retardation measured in Van Phuc
integrates the effect of competing ions typically present at higher con-
centrations in the Holocene aquifer (Supplementary Fig. 4) as well as
the impact of Fe oxyhydroxide reduction. However, the extent to which
contamination was caused by either As transport from the adjacent
Holocene aquifer or reductive dissolution of Fe(
III) oxyhydroxides and
in situ As release to groundwater cannot be determined from the
available data (Supplementary Fig. 8).
The sharp drop in dissolved organic carbon concentrations across
the geological boundary from 9 mg per litre to about 1 mg per litre
indicates rapid organic carbon mineralization coupled to the reduction
of Fe(
III) oxyhydroxides and explains the formation of a plume of grey
sands within the Pleistocene aquifer (Fig. 3b). On the basis of a stoi-
chiometric Fe/C ratio of 4 (ref. 15), the dissolved organic carbon sup-
plied by flushing the aquifer 30 times with groundwater from the
Holocene aquifer would be required to turn Pleistocene sands from
orange to grey by reducing half of their 0.1% reactive Fe(
III) oxyhydr-
oxide content
23
, assuming a porosity of 0.25. Given that groundwater
was advected over a distance of 2,000–2,300 m across the geological
boundaryover the past 40–60 years, we would predict that the plume of
grey sands extends 65–75 m into the Pleistocene aquifer. This is some-
what less than is observed (Fig. 3), possibly due to additional reduction
by H
2
advected from the Holocene portion of the aquifer
14
. The Van
Phuc observations indicate that dissolved organic carbon advected
from a Holocene aquifer can be at least as important for the release
of As to groundwater as autochthonous organic carbon
12,24–27
.
Contamination of Pleistocene aquifers has previously been invoked
in the Red River and the Bengal basins
11,12,28
, but without the benefit of
0
100
200
300
400
500
As concentration (μg per litre)
R = 1
0
2
4
6
8
10
1.51.71.92.1
DOC concentration (mg per litre)
Distance from river (km)
R = 5
R = 16
R = 20
R = 40
a
b
Figure 3
|
Distribution of arsenic and dissolved organic carbon in
groundwater within the 25–30-m depth interval along the Van Phuc
transect. Symbols are coloured according to the classification in Fig. 1. Grey and
orange shading indicates the extent of the grey Holocene aquifer and the portion
of the Pleistocene aquifer that is still orange, respectively. The intermediate area
without shading indicates the portion of the Pleistocene aquifer that became grey.
Shown as dotted lines are predicted As concentrations bracketing the
observations with retardation factors R of 16 and 20 and an average advection
velocity of 43 m yr
21
over the 50 years preceding the 2011 sampling
(Supplementary Discussion). a, Also shown are predicted concentrations for As
assumingretardationfactors of 1, 5 and 40 and thesame averagerate of advection.
b, For visual reference, predicted dissolved organic carbon concentrations are
shown as dotted lines according to the same advection velocity and retardation
factors of 16, 20 and 40, assuming there was no detectable dissolved organic
carbon in the Pleistocene aquifer before the perturbation.
0
20
40
60
0
20
40
60
0
20
40
60
0
20
40
60
Depth (m)Depth (m)Depth (m)Depth (m)
2.0 1.8 1.6 1.4
Distance from river bank (km)
Holocene clay
Holocene aquifer
Pleistocene aquifer
10,000
8,000
6,000
4,000
2,000
0
0.8
0.6
0.4
0.2
0
600
500
400
300
200
100
0
50
40
30
20
10
0
Ground water age (yr)Groundwater As (μg per litre)Reectance 530–520 nmCuttings Ca (mg kg
–1
)
a
b
c
d
Palaeosol
Figure 2
|
Contoured sections of sediment and water properties based on
data collected between 1.3 km and 2.0 km from the Red River bank. The
location and number of samples indicated as black dots varies by type of
measurement. a, Concentration of Ca in sand cuttings measured by X-ray
fluorescence. Also shown are the boundaries separating the two main aquifers
and the palaeosol overlying the Pleistocene aquifer. ‘2000’ labels the contour for
2,000 mg Ca per kg. b, Difference in diffuse spectral reflectance between 530 nm
and 520 nm, indicative of the colour of freshly collected drill cuttings
13
. The
contour labels correspond to the percentage difference in reflectance shown by
the colour scale. c, Concentrations of As in groundwater collected in 2006 with
the needle sampler and in 2011 by monitoring wells along the transect. ‘10’
labels the contour for the WHO guideline, 10 mg As per litre. d, Groundwater
ages relative to recharge determined by
3
H–
3
He dating of groundwater samples
collected from a subset of the monitoring wells in 2006. The portion of the
Pleistocene aquifer that became reduced and where As concentrations
presumably increased over time is located within the large white arrow pointing
in the direction of flow. The plot was drawn with Ocean Data View
(http://odv.awi.de/).
RESEARCH LETTER
206 | NATURE | VOL 501 | 12 SEPTEMBER 2013
Macmillan Publishers Limited. All rights reserved
©2013
a well-defined hydrogeological context. The Pleistocene aquifer of Van
Phuc was contaminated under the conducive circumstances of accel-
erated lateral flow. Although downward groundwater flow and there-
fore penetration of As will typically be slower, the Van Phuc findings
confirm that the vulnerability of Pleistocene aquifers will depend on
the local spatial density of incised palaeo-channels that were subse-
quently filled with Holocene sediments
12
. Owing to retardation, con-
centrations of As in a Pleistocene aquifer will not increase suddenly but
over timescales of decades even in the close vicinity of a Holocene
aquifer. This is consistent with the gradual increase in groundwater
As concentrations documented by the few extended time series avail-
able from such a vulnerable setting
29
. However, concentrations of As
could rise more rapidly where flow accelerates beyond the rate docu-
mented in Van Phuc, closer to Hanoi for instance.
METHODS SUMMARY
A total of 41 wells were installed in Van Phuc in 2006–11. The water levels of the
river and in the wells were recorded from September 2010 to June 2011 using
pressure transducers and adjusted to the same elevation datum after barometric
corrections. The magnitude and direction of the head gradient within the 25–30-m
depth interval was calculated from water level measurements in three wells
(Fig. 1b). In 2006, a subset of the wells was sampled for noble gas and tritium
(
3
H) analysis at a high flow rate using a submersible pump to avoid degassing. The
samples were analysed by mass spectrometry in the Noble Gas Laboratory at ETH
Zurich.
3
H concentrations were determined by the
3
He ingrowth method
30
.
Groundwater As, Fe and Mn concentrations measured by high-resolution induc-
tively coupled plasma mass spectrometry at LDEO represent the average for
acidified samples collected in April and May 2012. Further details are provided
in the Supplementary Information.
Full Methods and any associated references are available in the online version of
the paper.
Received 17 December 2012; accepted 11 July 2013.
1. Ravenscroft, P., Brammer, H. & Richards, K. Arsenic Pollution: A Global Synthesis
(RGS-IBG Book Series, Wiley-Blackwell, 2009).
2. BGS/DPHE. (British Geological Survey, Department of Public Health Engineering)
Arsenic Contamination of Groundwater in Bangladesh Final Report,
http://www.bgs.ac.uk/arsenic/bangladesh/ (British Geological Survey, 2001).
3. Fendorf, S., Michael, H. A. & van Geen, A. Spatial and temporal variations of
groundwater arsenic in south and southeast Asia. Science 328, 1123–1127
(2010).
4. Ahmed, M. F. et al. Epidemiology: ensuring safe drinking water in Bangladesh.
Science 314, 1687–1688 (2006).
5. Berg, M. et al. Magnitude of arsenic pollution in the Mekong and Red River deltas—
Cambodia and Vietnam. Sci. Total Environ. 372, 413–425 (2007).
6. Eiche, E. et al. Geochemical processes underlying a sharp contrast in groundwater
arsenic concentrations in a village on the Red River delta, Vietnam. Appl. Geochem.
23, 3143–3154 (2008).
7. Funabiki, A., Haruyama, S., Quy, N. V., Hai, P. V. & Thai, D. H. Holocene delta plain
development in the Song Hong (Red River) delta, Vietnam. J. Asian Earth Sci. 30,
518–529 (2007).
8. Larsen, F. et al. Controlling geological and hydrogeological processes in an arsenic
contaminated aquifer on the Red River flood plain, Vietnam. Appl. Geochem. 23,
3099–3115 (2008).
9. Thu, T. M. & Fredlund, D. G. Modelling subsidence in the Hanoi City area, Vietnam.
Can. Geotech. J. 37, 621–637 (2000).
10. Berg, M. et al. Hydrological and sedimentary controls leading to arsenic
contamination of groundwater in the Hanoi area, Vietnam: the impact of iron-
arsenic ratios, peat, river bank deposits, and excessive groundwater abstraction.
Chem. Geol. 249, 91–112 (2008).
11. Winkel, L. H. E. et al. Arsenic pollution of groundwater in Vietnam exacerbated by
deep aquifer exploitation for more than a century. Proc. Natl Acad. Sci. USA 108,
1246–1251 (2011).
12. McArthur, J. M. et al. How paleosols influence groundwater flow and arsenic
pollution: a model from the Bengal Basin and its worldwide implication. Wat.
Resour. Res. 44, W11411 (2008).
13. Horneman, A. et al. Decoupling of As and Fe release to Bangladesh groundwater
under reducing conditions. Part I: evidence from sediment profiles. Geochim.
Cosmochim. Acta 68, 3459–3473 (2004).
14. Postma, D. et al. Groundwater arsenic concentrations in Vietnam controlled by
sediment age. Nature Geosci. 5, 656–661 (2012).
15. Postma, D. et al. Arsenic in groundwater of the Red River floodplain,Vietnam:
controlling geochemical processes and reactive transport modeling. Geochim.
Cosmochim. Acta 71, 5054–5071 (2007).
16. Kipfer, R., Aeschbach-Hertig, W., Peeters, F. & Stute, M. Noble gases in lakes and
ground waters. Rev. Mineral. Geochem. 47, 615–700 (2002).
17. Stollenwerk, K. G. et al. Arsenic attenuation by oxidized aquifer sediments in
Bangladesh. Sci. Total Environ. 379, 133–150 (2007).
18. van Geen, A. et al. Flushing history as a hydrogeological control on the regional
distribution of arsenic in shallow groundwater of the Bengal Basin. Environ. Sci.
Technol. 42, 2283–2288 (2008).
19. Nath, B. et al. Mobility of arsenic in the sub-surface environment: an integrated
hydrogeochemical study and sorption model of the sandy aquifer materials.
J. Hydrol. 364, 236–248 (2009).
20. Itai, T. et al. Variations in the redox state of As and Fe measured by X-ray absorption
spectroscopy in aquifers of Bangladesh and their effect on As adsorption. Appl.
Geochem. 25, 34–47 (2010).
21. Radloff, K. A. et al. Arsenic migration to deep groundwater in Bangladesh
influenced by adsorption and water demand. Nature Geosci. 4, 793–798 (2011).
22. Jessen, S. et al. Surface complexation modeling of groundwater arsenic mobility:
results of a forced gradient experiment in a Red River flood plain aquifer, Vietnam.
Geochim. Cosmochim. Acta 98, 186–201 (2012).
23. Dhar, R. K. et al. Microbes enhance mobility of arsenic in Pleistocene aquifer sand
from Bangladesh. Environ. Sci. Technol. 45, 2648–2654 (2011).
24. Islam, F. S. et al. Role of metal-reducing bacteria in arsenic release from Bengal
delta sediments. Nature 430, 68–71 (2004).
25. Polizzotto, M. L., Kocar, B. D., Benner, S. B., Sampson, M. & Fendorf, S. Near-surface
wetland sediments as a source of arsenic release to ground water in Asia. Nature
454, 505–508 (2008).
26. Neumann, R. B. et al. Anthropogenic influences on groundwater arsenic
concentrations in Bangladesh. Nature Geosci. 3, 46–52 (2010).
27. Mailloux, B.J. et al. Advection of surface-derived organic carbon fuels microbial reduc-
tion in Bangladesh groundwater. Proc. Natl Acad. Sci. USA 110, 5331–5335 (2013).
28. Mukherjee, A. et al. Elevated arsenic in deeper groundwater of the western Bengal
basin, India: extent and controls from regional to local scale. Appl. Geochem. 26,
600–613 (2011).
29. McArthur, J. M. et al. Migration of As, and
3
H-
3
He ages, in groundwater from West
Bengal: implications for monitoring. Water Res. 44, 4171–4185 (2010).
30. Beyerle, U. et al. A mass spectrometric system for the analysis of noble gases and
tritium from water samples. Environ. Sci. Technol. 34, 2042–2050 (2000).
Supplementary Information is available in the online version of the paper.
Acknowledgements This study was supported by NSF grant EAR 09-11557, the Swiss
Agency for Development and Cooperation, grant NAFOSTED 105-09-59-09 to
CETASD, and NIEHS grants P42 ES010349 and P42 ES016454. This is
Lamont-Doherty Earth Observatory contribution number 7698.
Author Contributions A.v.G., M.B., P.T.K.T., P.O. and B.C.B. conceived the study. V.M.L.,
N.-N.M, P.D.M., P.T.K.T. and P.H.V. were responsible for organizing the field work and
carrying out the monitoring throughout the study. K.R., Z.A. and B.W. participated in the
field work in 2006. M.O.S. processed the hydrological data and carried out the flow
modelling under the supervision of C.F.H. and P.O. J.L.M. was responsible for
groundwater analyses at LDEO, C.S. for those at Eawag, and F.F. for noble gas
measurements in R.K.’s laboratory. A.v.G. drafted the paper, which was then edited by
all co-authors.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare no competing financial interests.
Readers are welcome to comment on the online version of the paper. Correspondence
and requests for materials should be addressed to A.v.G.
(avangeen@ldeo.columbia.edu).
LETTER RESEARCH
12 SEPTEMBER 2013 | VOL 501 | NATURE | 207
Macmillan Publishers Limited. All rights reserved
©2013
METHODS
Drilling. A first set of 25 wells, including two nests of nine and ten wells tapping
the depth range of the Holocene and Pleistocene aquifers, respectively, were
installed in Van Phuc in 2006 (ref. 6). Another 16 monitoring wells were installed
between December 2009 and November 2011. Three additional holes were drilled
to collect cuttings without installing a well. All holes were drilled by flushing the
hole with water through a rotating drill bit.
Needle sampling. In 2006, drilling was briefly interrupted at seven sites to increase
the vertical resolution of both sediment and groundwater data using the needle
sampler
31
. Groundwater was pressure-filtered under nitrogen directly from the
sample tubes. As a measure of the pool of mobilizable As, sediment collected with
the needle sampler was subjected to a single 24-hour extraction in a 1 M PO
4
solution at pH 5 (ref. 32).
Water level measurements. A theodolite elevation survey of the well and river
measurement points were carried out in June 2010 by a surveying team from
Hanoi University of Science. Water level data in both the wells and river were
recorded using Solinst Levelogger pressure transducers. A barometric pressure
logger was also deployed at the field site. Water level and barometric data were
recorded at 5-min intervals and all water level data was barometrically corrected.
The barometrically corrected water level data from each logger was then adjusted
to the surveyed elevation of their respective measurement point so that all of the
data was referenced to the same elevation datum.
Groundwater flow. The magnitude and direction of the head gradient within the
25–30-m depth of the aquifer at Van Phuc was calculated using the barometrically
adjusted and survey-referenced water level data collected at 5-min intervals from
September 2010 to June 2011 in three wells located near the centre of the transect
(Fig. 1b). A least-squares fit of a plane was calculated for each set of simultaneous
water levels at these three wells, and from this set of planes the magnitude and
direction of the head gradient at 5-min intervals was directly computed.
Groundwater analysis. In 2006, a subset of the monitoring wells was sampled
along a vertical transect for noble gas and tritium (
3
H) analysis. After purging the
wells, the samples were taken using a submersible pump. To avoid degassing of the
groundwater owing to bubble formation during sampling the water was pumped
at high rates to maintain high pressure. The samples for noble gas and
3
H analysis
were put into copper tubes and sealed gastight using pinch-off clamps. All samples
were analysed for noble gas concentrations and the isotope ratios
3
He/
4
He,
20
Ne/
22
Ne and
36
Ar/
40
Ar using noble gas mass spectrometry in the Noble Gas
Laboratory at ETH Zurich
30,33
.
3
H concentrations were determined by the
3
He
ingrowth method using a high-sensitivity compressor-source noble gas mass
spectrometer.
3
H–
3
He ages were calculated according to the equations listed in
ref. 34, taking into account an excess air correction. When comparing the recon-
structed original
3
H content of each sample as a function of
3
H–
3
He age with the
3
H input function for south and southeast Asia (Supplementary Fig. 5), most
samples follow the trend expected from simple plug flow
34,35
.
Several days before analysis by high-resolution inductively coupled plasma
mass spectrometry at LDEO, groundwater was acidified to 1% Optima HCl in
the laboratory
36
. This has been shown to re-dissolve entirely any precipitates that
could have formed
37
. In most cases, the difference between duplicates was within
the analytical uncertainty of ,5%. With the exception of needle-sample data and
the nest of ten wells in the Holocene portion of the aquifer, which had to yield to
construction, groundwater As, Fe and Mn concentrations reported here represent
the average for samples collected without filtration in Apriland May 2012. Ground-
water data from 2006 were previously reported in refs 6 and 31.
Dissolved organic carbon samples were collected in 25-ml glass vials combusted
overnight at 450 uC and acidified to 1% HCl at the time of collection. Dissolved
inorganic carbon samples were also collected in 25-ml glass vials with a Teflon
septum but were not acidified. Both dissolved organic carbon (‘‘NPOC’’) and
dissolved inorganic carbon (by difference of ‘‘TC-NPOC’’) were analysed on a
Shimadzu TOC-V carbon analyser calibrated with K phthalate standards.
Ammonium samples were collected in polypropylene bottles after passing
through 0.45 mm cellulose acetate membrane filters and preserved by acidifying
to pH , 2 with HNO
3
.NH
4
1
concentrations were analysed on a spectrophot-
meter (UV-3101, Shimadzu) at a wavelength of 690 nm after forming a complex
with nitroferricyanide
38
.
Methane (CH
4
) samples were filled up to about half of the pre-vacuumed glass
vials and immediately frozen in dry ice. The analyses were performed no longer
than ten days after sampling. Headspace CH
4
in the vials was measured on a
Shimadzu 2014 gas chromatograph with a Porapak T packed column
14
.
Sediment analysis. As a measure of the redox state of Fe in acid-leachable oxy-
hydroxides, the diffuse spectral reflectance spectrum of cuttings from all sites was
measured on samples wrapped in Saran wrap and kept out of the sun within 12
hours of collection using a Minolta 1600D instrument
13
. Starting in 2009, the
coarse fractions of the drill cuttings were analysed by X-ray fluorescence for a
suite of elements including Ca using an InnovX Delta instrument. The drill cut-
tings were resuspended in water several times to eliminate the overprint of Ca-
enriched clays contained in the recycled water used for drilling. The washed
samples were run as is, without drying or grinding to powder. Analyses of NIST
reference material SRM2711 (28,800 6 800 mg Ca per kg) analysed by X-ray
fluorescence at the beginning and end of each run averaged 30,200 6 400 mg Ca
per kg (n 5 16).
31. van Geen, A. et al. Comparison of arsenic concentrations in simultaneously-
collected groundwater and aquifer particles from Bangladesh, India, Vietnam,
and Nepal. Appl. Geochem. 23, 3244–3251 (2008).
32. Zheng, Y. et al. Geochemical and hydrogeological contrasts between shallow and
deeper aquifers in two villages of Araihazar, Bangladesh: implications for deeper
aquifers as drinking water sources. Geochim. Cosmochim. Acta 69, 5203–5218
(2005).
33. Frei, F. Groundwater dynamics and arsenic mobilization near Hanoi (Vietnam)
assessed using noble gases and tritium. Diploma thesis, ETH Zurich (2007).
34. Klump, S. et al. Groundwater dynamics and arsenic mobilization in Bangladesh
assessed using noble gases and tritium. Environ. Sci. Technol. 40, 243–250
(2006).
35. Stute, M. et al. Hydrological control of As concentrations in Bangladesh
groundwater. Wat. Resour. Res. 43, W09417 (2007).
36. Cheng, Z., Zheng, Y., Mortlock, R. & van Geen, A. Rapid multi-element analysis of
groundwater by high-resolution inductively coupled plasma mass spectrometry.
Anal. Bioanal. Chem. 379, 512–518 (2004).
37. van Geen,A.et al. Monitoring 51 deep community wellsin Araihazar,Bangladesh,
for up to 5 years: implications for arsenic mitigation. J. Environ. Sci. Health A 42,
1729–1740 (2007).
38. Koroleff, F. In Methods of Seawater Analysis (ed. Grasshoft, K.) 126–133 (Chemie,
1974).
RESEARCH LETTER
Macmillan Publishers Limited. All rights reserved
©2013
WWW.NATURE.COM/NATURE | 1
SUPPLEMENTARY INFORMATION
doi:10.1038/nature12444
2
Figure S1 (A) Government estimates of total municipal pumping in 2000 and 2005, with a
projection for 2010 provided in 2006 by the Hanoi Water Works. (B) Water levels in the
Phapvan (red) and Maidich (blue) well fields of Hanoi used to model subsidence (9).
3
Series8
AMS12
NS7
NestHiA
s
NS3
AMSNS5
AMS3
AMS1
AMS11
VPNS4/A
M
AMS2
NestLoA
VPNS6
0.2
1.1
1.3
1.5
1.6
1.68
1.69
1.7
1.8
1.9
2.0
2.2
Distancefrom river
(km)
10
20
30
40
50
60
‐2.5 ‐2.0 ‐1.5 ‐1.0 ‐0.5 0.0 0.5 1.0
Depth(m)
Saturationindexwithrespecttocalcite
Figure S2 Saturation index with respect to calcite calculated using Visual MINTEQ 3.0
(http://www2.lwr.kth.se/English/OurSoftware/vminteq/) on the basis of concentrations of Ca,
alkalinity and pH measured in groundwater from the Pleistocene and Holocene aquifers of Van
Phuc. Symbols shapes and colors as in Fig. 3. Similar calculations for dolomite are presented in
(6).
SUPPLEMENTARY INFORMATION
2 | WWW.NATURE.COM/NATURE
RESEARCH
4
Figure S3 Contoured sections of groundwater (A) Fe and (B) Mn concentrations between
distances of 1.3 and 2.0 km from the Red River bank. Also shown are the boundary separating
the two main aquifers and the paleosol overlying the Pleistocene aquifer. The portion of the
Pleistocene aquifer that became reduced and where As concentrations presumably increased over
time is located within the large arrow pointing in the direction of flow.
WWW.NATURE.COM/NATURE | 3
SUPPLEMENTARY INFORMATION RESEARCH
5
(A)
(B)
(C)
(D)
Figure S4 Distribution of (A) conductivity, (B) dissolved inorganic carbon (DIC), (C)
ammonium (NH
4
+
), and (D) methane (CH
4
) within the 25-30 m depth interval along the Van
Phuc transect. Symbols are colored according to the classification in Fig. 1. Grey and yellow
shading indicates the extent of the grey Holocene aquifer and the portion of the Pleistocene
aquifer that is still orange, respectively. The intermediate area without shading indicates the
portion of the Pleistocene aquifer that turned grey. For visual reference, predicted conductivity
and concentrations in groundwater are shown as dotted lines according to an average advection
velocity 43 m/yr over 50 years using a 5-40 range of retardation as well as the 16-20 range
calculated for As from this study.
6
1
10
100
1000
J‐60 J‐65 J‐70 J‐75 J‐80 J‐85 J‐90 J‐95 J‐00 J‐05
3
H+
3
He(TU)
1960197019801990 2000
Figure S5 Comparison of sum of
3
H and
3
He levels and this “stable
3
H” with predicted levels
based on the history of bomb input and dispersion. For 3 out of the 8 samples along the transect
that were dated, stable
3
H levels significantly deviate from the model, presumably because of
mixing and/or deviation from simple plug flow. The single point with a deficit in stable
3
H could
be due to mixing with old water recharged before the 1950s (35) whereas the two points elevated
in stable
3
H may reflect mixing with a component of groundwater recharged in the 1960-1970
when the levels of
3
H in precipitation were particularly high. For one additional sample,
problems were already recorded during the analysis and the measurement was not taken into
account (33).
SUPPLEMENTARY INFORMATION
4 | WWW.NATURE.COM/NATURE
RESEARCH
7
Figure S6 Variations in EM conductivity across the Red River meander of Van Phuc. Areas of
low EM conductivity indicate proximity to the surface of sand; areas of high EM conductivity
correspond to a thick surficial clay layer (39).
WWW.NATURE.COM/NATURE | 5
SUPPLEMENTARY INFORMATION RESEARCH
8
Figure S7 Prediction of the distribution of groundwater age in 2006 as a function of distance
from the southern river bank of Van Phuc for different groundwater flow scenarios (see
Supplementary Discussion). Horizontal error bars span the 400 m width of the likely recharge
area. Vertical error bars indicate the measurement error in groundwater ages. Symbols are
colored according to the classification in Fig. 1. Samples without detectable
3
H are shown
assuming the minimum age of 55 years. Solid lines indicate predicted age-distance relationships
for various combinations of start dates and annual increments for enhanced groundwater flow
attributable to Hanoi pumping.
9
Figure S8 Phosphate-extractable As levels in Holocene and Pleistocene sediment collected with
the needle-sampler in 2006, from (31) and additional data for orange Pleistocene aquifer (green
and blue symbols). Despite the contrast in P-extractable As between the two formations, the
extent to which the contamination was caused by As transport from the adjacent Holocene
aquifer or reductive dissolution of Fe(III) oxyhydroxides and in situ As release to groundwater
cannot be determined. Release of ~0.2 mg/kg As from the Pleistocene sands following the
reduction of Fe oxyhydroxides could result in an increase in groundwater As concentrations by
as much as 500 g/L (2).
SUPPLEMENTARY INFORMATION
6 | WWW.NATURE.COM/NATURE
RESEARCH
10
Supplementary Discussion
Groundwater Age Modeling: The mean age and
3
H-
3
He age for the different pumping scenarios
outlined in the paper were solved numerically to test how dispersion could influence the results.
The average difference over the modeled domain between the analytical solution for
groundwater ages and the numerical solution for mean groundwater ages was less than one-
percent for all of the modeled scenarios. The average difference over the modeled domain
between the numerical solution for
3
H-
3
He ages and the analytical solution for the groundwater
ages was less than 10-percent for all the modeled scenarios. Comparison of these numerical
solutions that include dispersion with the analytic solution for purely advective transport indicate
that dispersion had a small effect on the calculated ages.
The one-dimensional advective-dispersive equation with a reaction term for the aging process
was evaluated using finite-volumes with an explicit time-stepping method. In each scenario the
initial velocity was set equal to zero and the velocity in all subsequent time-steps was assigned
according to the groundwater acceleration rate for the scenario being modeled. A dispersivity
value of 5m, representative of a sandy aquifer, was used in all of the modeled scenarios. For the
mean age calculations the inflowing water was assigned an age of zero. In case of the
3
H-
3
He age
calculations, atmospheric tritium data reported monthly by the IAEA from 1961 to 2009 for
Hong Kong were used to assign tritium values to the inflowing water (IAEA/WMO (2006).
Global Network of Isotopes in Precipitation. The GNIP Database. Accessible at:
http://www.iaea.org/water). Mean age calculations for each pumping scenario were run for two
different sets of initial conditions. In the first case the initial ages were set equal to zero
throughout the modeled domain. In the second case the initial ages throughout the domain were
set equal to 200 years, a reasonable age estimate at 30 m-depth in the aquifer under conditions of
only vertical groundwater recharge that likely dominated prior to the influence of pumping.
EQ1 was used solved numerically to model the mean groundwater ages and EQ2 and EQ3 were
solved numerically to model the He-Tritium ages of the groundwater:
Where A is age [T]; V is velocity [L/T]; α is dispersivity [L]; R is the aging constant
[dimensionless]; Tr is the tritium activity [TU]; K is the tritium decay constant [1/T]; He is the
tritiogenic
3
He concentration, i.e. the
3
He that is produced by
3
H decay [TU].
Reconstruction of past flow induced by pumping:
The main source of recharge must be known in order to relate
3
H-
3
He ages to groundwater flow
velocities. Recharge of the Van Phuc aquifer could enter directly from the sandy river bottom or
enter the aquifer above the river bank during flooding of low-lying areas, which occurred as
recently as 2002 even though the level of the Red River is largely controlled upstream by a series
of dams. A geophysical survey of the area combined with drilling logs shows that the sandy
WWW.NATURE.COM/NATURE | 7
SUPPLEMENTARY INFORMATION RESEARCH
11
aquifer extends almost to the surface in the eastern uninhabited portions of the meander and,
unlike most of the remaining area, is not capped by a thick clay layer (Fig. S6). On the basis of
these observations, the likely recharge area for the groundwater sampled along the transect
extends from the center of the Red River to the inland area where the surficial clay layer thickens
markedly, i.e. from 100 m southeast to 300 m northwest of the river bank (Fig. 1B).
The increase in
3
H-
3
He groundwater age with travel distance from the recharge area is used to
estimate how groundwater velocity has shifted in recent decades. The distribution of
groundwater ages along the transect indicates that the rate of recharge of the aquifer at the river
has not been constant (Fig. S7). Time series of groundwater level indicate a roughly linear
increase in the depth of the cone of depression below Hanoi since at least 1988 and imply a
constant acceleration of groundwater flow over time (Fig. S1). Steady flow divergence could
also explain the apparent decrease in velocity along the transect, but this interpretation is
inconsistent with pressure transducer records at the geological boundary. The simplest
explanation for the observed distribution of ages as a function of distance from the recharge area
is therefore accelerating flow drawn by increased Hanoi pumping.
A
T
V0
flowreversal
Hanoipumping
time
On the basis of this assumption and neglecting dispersion, the distance of the front from recharge
at the upstream end of the transect is X
F
= KT
2
/2, where K is the slope of the acceleration and T
is the time since the gradient reversal, calculated by integrating the expression Vdt = Kt dt from
time zero to time T. In order to calculate the groundwater age A of the water behind the front
(i.e. X<X
F
), the same expression is integrated from time T-A to time T and solved for A = T –
(T
2
– 2X/K)
1/2
. A numerical model including Fickian dispersion shows that the age-distance
relation is not particularly sensitive to dispersion. Two model scenarios illustrated in Fig. S7
reflect government pumping rate estimates which back-extrapolate linearly to zero groundwater
extraction in 1983 (Fig. S1). Annual increases in groundwater velocities by 3.5 and 5.5 m/yr
starting in 1983, respectively, encompass the envelope of distances and ages defined by the 3
samples within the faster moving core (24-28 m depth). Although economic liberalization in
Vietnam and, therefore possibly an increase in municipal pumping, started in the early 1980s, the
two scenarios do not seem plausible because groundwater pumping reportedly started in Hanoi
as early as 1909 (9). In addition, a 1983 start date for the perturbation cannot generate
groundwater that has been isolated from the atmosphere for more than 23 years (the 2006
sampling date minus the presumed 1983 onset of intense groundwater pumping) and therefore do
SUPPLEMENTARY INFORMATION
8 | WWW.NATURE.COM/NATURE
RESEARCH
12
not extend to the sample of 35-yr-old water at 32 m depth and a distance of 1380 m from the
recharge area (Fig. S7).
A scenario that fits the range of all groundwater ages measured by the
3
H-
3
He method requires a
reversal of the head gradient taking place in Van Phuc no later than 1971 and, therefore, a slower
annual increase in advection of 2.4 m/yr (Fig. S7). The 1971 reversal date corresponds roughly to
the end of what is referred to in Vietnam as the American War. In order to extend the model to
the minimum age of 55 yr established for groundwater in the Pleistocene aquifer, the reversal of
flow directions has to be anticipated by another 20 years to 1951 and the annual increase reduced
to 1.2 m/yr. Advection velocities of 75 and 95 m/yr in 2010 predicted for flow reversals starting
in 1951 or 1971, respectively, combined with the average head gradient measured in Van Phuc in
2010-11 of 5.2 x 10
-4
(Fig. 1C), yield hydraulic conductivities of 100-125 m/day (assuming a
porosity = 0.25) that are within the range expected for sandy aquifers. The two transient flow
models yield average advection rates of 38 and 48 m/yr, respectively, over the time spans of 40
and 60 years separating the onset of the flow perturbation and groundwater sampling for
chemical analysis in 2011.
Changes in groundwater chemistry across the Holocene-Pleistocene boundary:
The single most abundant constituent of Van Phuc groundwater is dissolved inorganic carbon
(DIC), primarily in the form of bicarbonate ion (Fig. S4). Given the circumneutral pH of
groundwater, the distribution of DIC and conductivity are therefore closely coupled. Much lower
concentrations of DIC, methane (CH
4
), and ammonium (NH
4
+
) in the Pleistocene portion of the
Van Phuc transect compared to the Holocene are consistent with the association between
sediment age and organic matter reactivity recently documented elsewhere in the Red River
basin (14). DIC is present at an order of magnitude higher concentration than DOC and therefore
unlikely to be significantly enhanced by DOC remineralization along the transect. Some
contribution to the sharp drop in DOC as groundwater crosses the geological boundary due to
adsorption onto the sediment cannot be ruled out, although field observations indicate retardation
of DOC is not likely to be as pronounced as for As (40). Other groundwater constituents elevated
in the Holocene portion of the aquifer such as DIC and NH
4
+
migrated at least twice as far as As
into the Pleistocene aquifer (Fig. S4).
Although surface-complexation models have been developed to predict the retardation of As due
to adsorption in both Holocene and Pleistocene sands (e.g. 22), the available data do not justify
assumptions beyond that of a linear adsorption isotherm described by a distribution coefficient
K
d
that is the ratio of the As concentration in the particulate phase per mass of sediment divided
by the As concentration in the dissolved phase (18). For a simple one-dimensional advection-
diffusion model that assumes no As is initially present along the flow path and that recharge
water enters the aquifer at one end of the flow path containing an As concentration C
i
, the
solution of the governing equation is:
WWW.NATURE.COM/NATURE | 9
SUPPLEMENTARY INFORMATION RESEARCH
13
where C is the concentration of As in the dissolved phase, x the distance along the flow path, v
the advection velocity, D is a hydraulic dispersion coefficient, and R the retardation factor (41).
Assuming As initial concentrations of 486 g/L in the Holocene aquifer at a distance of 1.6 km
from the river bank and 1 g/L of Pleistocene aquifer, two retardation factors R but the same
dispersion coefficient D were calculated for the two advection scenarios by iteratively
minimizing the sum of the squared residuals between the observations and the model predictions.
For an optimized hydraulic dispersion of 1.8 10
-6
m
2
/s, retardation factors of 16 and 20 are
calculated for advection at 48 m/yr for 40 years and at 38 m/yr for 60 years, respectively. The
trends in predicted As concentrations based on these two scenarios are essentially
indistinguishable. The same retardation factors were applied to an average advection scenario of
43 m/yr for 50 years to illustrate that they span the likely range of uncertainty in As retardation
(Fig. 3). The distribution coefficient K
d
is related to the retardation factor R by the equation K
d
=
θ/(R, where θ is a typical aquifer porosity of 0.25 and is the corresponding aquifer bulk
density assuming a dry sediment density of 2.65 kg/cm
3
. With these assumptions, retardation
factors of 16-20 are equivalent to distribution coefficients of 1.7-2.1 L/kg.
The pattern of Ca concentrations in sand cuttings provides an independent check on groundwater
flow across the Holocene-Pleistocene boundary in Van Phuc. Detrital carbonate is rarely
observed in fluvial sediment, even in karstic terrain such as the Red River watershed. Unlike the
Pleistocene aquifer, groundwater within the Holocene aquifer of Van Phuc is highly
supersaturated with respect to calcite due to a combination of elevated Ca (2 mmol/L) and DIC
concentrations (10 mmol/L) at pH 7 driven by the metabolism of organic carbon paired with
reductive dissolution of Fe oxyhydroxides (Fig. S2). Elevated Ca concentrations in Holocene
sand cuttings of Van Phuc therefore most likely indicates authigenic precipitation of calcite or
dolomite. The deficit in conductivity and DIC between 1.7 and 1.9 km from the river bank
relative to the model prediction may indicate that, in addition to adsorption of DIC, some
precipitation of calcite occurs as groundwater from the Holocene aquifer enters the Pleistocene
aquifer (Fig. S4). This process provides an independent check on the onset of strong
northwesterly groundwater flow in Van Phuc. On the basis of the Ca concentration in the
northernmost well tapping the Holocene aquifer, precipitation from almost 1000 pore volumes of
advected groundwater would be required to build up a concentration of 2000 mg/kg Ca in the
sediment. Given the 2 km advection of the groundwater front, elevated Ca concentrations in
cuttings would not be expected beyond a few meters downstream of the Holocene-Pleistocene
transition even if calcite precipitation were very rapid. The lack of detectable Ca in cuttings from
the first set of wells within the Pleistocene aquifer that are contaminated with As is therefore
consistent with a recent acceleration of northwesterly flow.
Supplementary References
39. Aziz, Z. et al. Impact of local recharge on arsenic concentrations in shallow aquifers inferred
from the electromagnetic conductivity of soils in Araihazar, Bangladesh, Water Resources
Research 44, doi:10.1029/2007WR006000 (2008).
40. McCarthy, J. F. et al. Field tracer tests on the mobility of natural organic matter in a sandy
aquifer. Water Resources Research 32, 1223-1238 (1996).
41. van Genuchten, M. Th. & Alves, W. J. Analytical solutions of the one-dimensional
convective-dispersive solute transport equation; USDAARS Technical Bulletin 1661; U.S.
Salinity Laboratory: Riverside, CA (1982).
SUPPLEMENTARY INFORMATION
10 | WWW.NATURE.COM/NATURE
RESEARCH
TableS1.Noblegasand
3
Hdata
Well Latitude Longitude Depth(m) He +/‐
Ne Ar
3
He/
4
He
+/‐
20
Ne/
22
Ne
+/‐
40
Ar/
36
Ar
+/‐
3
He
tri
+/‐
3
H
+/‐
3
H‐
3
Heage
+/‐
(10
‐8
cm
3
STP
/g)
(%) (%)
(10
‐6
) (10
2
)
(TU) (TU) (yr)
MLA1 20.92326 105.89198 24 1.70 0.02 ‐71 ‐67 1.29 0.02 9.81 0.07 2.98 0.02 1.87 0.22 0.0 0.6 >55
MLA2 20.92326 105.89198 27 1.93 0.02 ‐68 ‐68 1.33 0.02 9.85 0.04 2.93 0.02 1.82 0.18 ‐ ‐ (>55)
MLA3 20.92326 105.89198 30 1.93 0.02 ‐68 ‐68 1.33 0.02 9.83 0.03 2.95 <0.01 1.81 0.19 0.0 1.4 >55
MLA4 20.92326 105.89198 33 2.24 0.02 ‐49 ‐27 1.39 0.02 9.88 0.04 2.92 0.01 1.16 0.21 0.0 0.5 >55
MLA5 20.92326 105.89198 36 2.66 0.03 ‐35 ‐16 ‐ ‐ 9.81 0.02 2.96 0.01 ‐
‐
0.4 1.0 >55
MLA6 20.92326 105.89198 39 2.44 0.02 ‐38 ‐20 1.35 0.01 9.77 0.03 2.96 0.01 0 0.22
‐
‐
(>55)
MLA7 20.92326 105.89198 41 2.55 0.03 ‐35 ‐17 1.35 0.02 9.79 0.01 2.96 0.01 0 0.26 0.2 0.4 >55
MLA8 20.92326 105.89198 45 2.12 0.02 ‐58 ‐57 1.32 0.02 9.79 0.03 2.95 0.01 0.28
‐
0.0 0.6 >55
MLAC 20.92326 105.89198 54 2.11 0.02 ‐60 ‐48 1.25 0.02 9.82 0.03 2.97 0.01 1.18
‐
0.0 2.2 >55
MLB2 20.91982 105.89751 21 3.66 0.04 ‐14 ‐4 1.92 0.02 9.75 0.04 2.94 0.01 7.73 2.04 0.0 1.1 >55
MLB3 20.91982 105.89751 24 3.79 0.04 ‐13 ‐3 1.80 0.01 9.79 0.02 2.96 <0.01 6.41 2.94 4.8 0.9 15.2 5.1
MLB4 20.91982 105.89751 27 4.95 0.05 11 10 1.64 0.02 9.80 0.01 2.96 <0.01 5.25 0.58 2.3 1.1 21.0 5.9
MLB5 20.91982 105.89751 34 5.16 0.05 16 13 1.64 0.01 9.78 0.02 2.96 <0.01 5.56 0.42 2.3 0.6 22.2 3.4
MLB6 20.91982 105.89751 36 5.02 0.05 17 15 1.88 0.01 9.78 0.01 2.96 <0.01 9.17 0.5 26.7 1.2 5.3 0.3
MLB7 20.91982 105.89751 41 3.32 0.03 ‐21 ‐11 1.63 0.01 9.77 0.02 2.96 0.01 3.02 0.29 1.2 0.4 22.9 4.4
MLB8 20.91982 105.89751 45 3.53 0.04 ‐17 ‐7 1.64 0.01 9.79 0.02 2.95 <0.01 3.63 0.34 0.0 1.2 >55
MLBC 20.91982 105.89751 57 3.10 0.03 ‐30 ‐6 1.31 0.01 9.82 0.01 2.95 <0.01 0 ‐ 0.2 1.0 >55
NS3 20.92081 105.89613 26 4.59 0.05 4 7 2.41 0.02 9.77 0.01 2.96 0.01 18.25 0.74 11.2 0.6 17.3 0.8
NS9 20.92489 105.89791 26 6.23 0.06 40 17 1.62 0.01 9.78 0.01 2.97 <0.01 5.59 0.73 3.5 0.6 17.1 2.3
NS6 20.92429 105.89043 28 3.55 0.04 ‐19 ‐11 1.43 0.01 9.80 0.01 2.95 0.01 1.37 (>1.37) 4.6 0.9 4.6 (>4.1)
NS5 20.92152 105.89495 31 4.92 0.05 14 14 2.53 0.02 9.77 0.02 2.95 0.01 21.7 0.54 3.5 0.5 35.3 2.1
NS1 20.92326 105.89198 31 5.34 0.05 14 ‐1 1.31 0.01 9.83 0.02 2.95 0.01 4.18 ‐ 0.0 0.6 >55
NS4 20.92205 105.89338 38 4.80 0.05 9 9 1.36 0.01 9.76 0.01 2.96 <0.01 0 0.5 0.1 0.6 >55
TableS1.Noblegasand
3
Hdata
Well Latitude Longitude Depth(m) He +/‐
Ne
Ar
3
He/
4
He
+/‐
20
Ne/
22
Ne
+/‐
40
Ar/
36
Ar
+/‐
3
He
tri
+/‐
3
H
+/‐
3
H‐
3
Heage
+/‐
(10
‐8
cm
3
STP
/g)
(%) (%)
(10
‐6
) (10
2
)
(TU) (TU) (yr)
MLA1 20.92326 105.89198 24 1.70 0.02 ‐71 ‐67 1.29 0.02 9.81 0.07 2.98 0.02 1.87 0.22 0.0 0.6 >55
MLA2 20.92326 105.89198 27 1.93 0.02 ‐68 ‐68 1.33 0.02 9.85 0.04 2.93 0.02 1.82 0.18
‐
‐
(>55)
MLA3 20.92326 105.89198 30 1.93 0.02 ‐68 ‐68 1.33 0.02 9.83 0.03 2.95 <0.01 1.81 0.19 0.0 1.4 >55
MLA4 20.92326 105.89198 33 2.24 0.02 ‐49 ‐27 1.39 0.02 9.88 0.04 2.92 0.01 1.16 0.21 0.0 0.5 >55
MLA5 20.92326 105.89198 36 2.66 0.03 ‐35 ‐16
‐
‐
9.81 0.02 2.96 0.01
‐
‐
0.4 1.0 >55
MLA6 20.92326 105.89198 39 2.44 0.02 ‐38 ‐20 1.35 0.01 9.77 0.03 2.96 0.01 0 0.22
‐
‐
(>55)
MLA7 20.92326 105.89198 41 2.55 0.03 ‐35 ‐17 1.35 0.02 9.79 0.01 2.96 0.01 0 0.26 0.2 0.4 >55
MLA8 20.92326 105.89198 45 2.12 0.02 ‐58 ‐57 1.32 0.02 9.79 0.03 2.95 0.01 0.28
‐
0.0 0.6 >55
MLAC 20.92326 105.89198 54 2.11 0.02 ‐60 ‐48 1.25 0.02 9.82 0.03 2.97 0.01 1.18
‐
0.0 2.2 >55
MLB2 20.91982 105.89751 21 3.66 0.04 ‐14 ‐4 1.92 0.02 9.75 0.04 2.94 0.01 7.73 2.04 0.0 1.1 >55
MLB3 20.91982 105.89751 24 3.79 0.04 ‐13 ‐3 1.80 0.01 9.79 0.02 2.96 <0.01 6.41 2.94 4.8 0.9 15.2 5.1
MLB4 20.91982 105.89751 27 4.95 0.05 11 10 1.64 0.02 9.80 0.01 2.96 <0.01 5.25 0.58 2.3 1.1 21.0 5.9
MLB5 20.91982 105.89751 34 5.16 0.05 16 13 1.64 0.01 9.78 0.02 2.96 <0.01 5.56 0.42 2.3 0.6 22.2 3.4
MLB6 20.91982 105.89751 36 5.02 0.05 17 15 1.88 0.01 9.78 0.01 2.96 <0.01 9.17 0.5 26.7 1.2 5.3 0.3
MLB7 20.91982 105.89751 41 3.32 0.03 ‐21 ‐11 1.63 0.01 9.77 0.02 2.96 0.01 3.02 0.29 1.2 0.4 22.9 4.4
MLB8 20.91982 105.89751 45 3.53 0.04 ‐17 ‐7 1.64 0.01 9.79 0.02 2.95 <0.01 3.63 0.34 0.0 1.2 >55
MLBC 20.91982 105.89751 57 3.10 0.03 ‐30 ‐6 1.31 0.01 9.82 0.01 2.95 <0.01 0 ‐ 0.2 1.0 >55
NS3 20.92081 105.89613 26 4.59 0.05 4 7 2.41 0.02 9.77 0.01 2.96 0.01 18.25 0.74 11.2 0.6 17.3 0.8
NS9 20.92489 105.89791 26 6.23 0.06 40 17 1.62 0.01 9.78 0.01 2.97 <0.01 5.59 0.73 3.5 0.6 17.1 2.3
NS6 20.92429 105.89043 28 3.55 0.04 ‐19 ‐11 1.43 0.01 9.80 0.01 2.95 0.01 1.37 (>1.37) 4.6 0.9 4.6 (>4.1)
NS5 20.92152 105.89495 31 4.92 0.05 14 14 2.53 0.02 9.77 0.02 2.95 0.01 21.7 0.54 3.5 0.5 35.3 2.1
NS1 20.92326 105.89198 31 5.34 0.05 14 ‐1 1.31 0.01 9.83 0.02 2.95 0.01 4.18 ‐ 0.0 0.6 >55
NS4 20.92205 105.89338 38 4.80 0.05 9 9 1.36 0.01 9.76 0.01 2.96 <0.01 0 0.5 0.1 0.6 >55
