NATURE GEOSCIENCE | VOL 4 | OCTOBER 2011 | www.nature.com/naturegeoscience 655

correspondence

To the Editor — Neumann et al.1 claim that dissolved organic carbon (DOC) in isotopically light recharge leaking from ponds in Bangladesh causes arsenic pollution of groundwater. They thereby link human activity to arsenic pollution of groundwater that adversely affects millions of people2. We suggest that the data of Neumann et al. show that isotopically light recharge is flushing arsenic from the aquifer at their two field sites in Munshiganj, Bangladesh.

Relying on Fig. 1 in their paper, Neumann et al. state that the minimum δ18O value in groundwater coincides with the depth of maximum arsenic at 30 m below ground level (mbgl), and so conclude that DOC in isotopically light ‘pond recharge’ is driving arsenic pollution. Their Fig. 1 actually shows peak concentrations of

arsenic at 36.6 mbgl, where δ18O is −3.5‰. The minimum δ18O value of −6.5‰ occurs 24 mbgl, where the arsenic concentration is 50% lower (Fig. 1a, where we laterally juxtapose Figs 1a and 1g of Neumann et al.). Their model outputs, arbitrarily doubled to make them approach reality1, show that 70% of the arsenic pollution resides in isotopically heavy water that was recharged before irrigation pumping began (Fig. 1a inset). Vertical profiles through the aquifer at Bejgaon (Fig. 1c), 800 m northwest of Bashailbhog, also show that appreciable arsenic resides in deeper, isotopically heavy groundwater. These findings suggest that isotopically light recharge is not driving arsenic pollution.

Our view — that isotopically light water is flushing arsenic from the aquifer in Munshiganj — is supported by the fact

that, through a flushing zone 0–30 mbgl at Bejgaon (data for Bashailbhog are not available), concentrations of dissolved species plot along linear mixing lines between predevelopment water and fresher recharge (Fig. 1b). Finally, if DOC in isotopically light recharge was driving arsenic release, as proposed by Neumann et al., the concentration of DOC would decrease along the flow path as it reacts to reduce sedimentary iron oxyhydroxides and so release arsenic; in reality, concentrations increase along the flow path and with depth to 30 mbgl (ref. 3).

Our proposal on flushing agrees with previous indications that ponds do not cause arsenic pollution in the Bengal basin4, and that the upper part of the shallow aquifer at Munshiganj is being flushed of arsenic5,6. ❐

Aquifer arsenic source

0

a b cδ18O ( ) δ18O ( )

10

–6

Unmineralizedrecharge

Isotopicallynormalrecharge

Isotopicallylight recharge

–5 –4 –3 600

400

200

00 5 10

DOC (mg l–1)

As

(mg

l–1)

15

0

10

–4 –3 –2 –1 0

20

30

40

50

60

70

80

600

400

200

00 5 10

Ca (mg l–1)

As

(mg

l–1)

15

600

400

200

00 0 250 500 7505 10

Cl (mg l–1)

As

(mg

l–1)

15 20 25

20

30

40

Dep

th (m

)

Dep

th (m

)

50

60

70

80

00 50

% water types

Post-developmentrecharge

Predevelopmentrecharge

100

1020304050

Dep

th (m

bgl)

607080

0 500 1,000As (mg l–1)

0 250 500 750As (μg l–1) As (μg l–1)

1,000

Isotopicallylight recharge

Figure 1 | a, Aquifer profiles from Neumann et al. of arsenic and δ18O (Figs 1a and 1g in their paper). Yellow areas denote arsenic-rich water in the aquifer at Bashailbhog. Dark blue area is water identified by Neumann et al. as having δ18O < −4.75‰ and so is, in their interpretation, derived from ponds. In our interpretation, areas of both shades of blue indicate water that is flushing the aquifer of arsenic. Inset shows model outputs of Neumann et al. predicting the proportion of predevelopment water with depth (their Fig. 1e), overlain by the distribution of arsenic in yellow. b, Cross-plots of DOC, dissolved calcium, chloride and arsenic in the Munshiganj aquifer at Bejgaon, 800 m northwest of Bashailbhog (data from ref. 7; similar data are not available for Bashailbhog) for depths to 30 mbgl — the depth of penetration of water that is isotopically light — and so, according to Neumann et al., derives from ponds. We do not plot deeper waters from beneath the zone of flushing as they are of different derivation and age. The linear trends show mixing between a mineralized groundwater around 30 mbgl and less-mineralized shallower recharge. c, Aquifer profiles of arsenic and δ18O in the aquifer at Bejgaon; data for arsenic from ref. 7; data for δ18O from ref. 8, includes samples 30(2), 30(3) and 61(2).

© 2011 Macmillan Publishers Limited. All rights reserved

656 NATURE GEOSCIENCE | VOL 4 | OCTOBER 2011 | www.nature.com/naturegeoscience

correspondence

References1. Neumann, R. B. et al. Nature Geosci. 3, 46–52 (2010).2. Argos, M. et al. Lancet 376, 252–258 (2010).3. Harvey, C. F. et al. Science 298, 1602–606 (2002).4. Sengupta, S., McArthur, J. M., Sarkar, A. & Leng, M. J.

Environ. Sci. Technol. 42, 5156–5164 (2008).5. McArthur J. M. et al. Appl. Geochem. 19, 1255–1293 (2004).

6. Klump, S. et al. Environ. Sci. Technol. 40, 243–250 (2006).7. Swartz, C. H. et al. Geochim. Cosmochim. Acta 68, 4539–4557 (2004).8. Harvey, C. F. et al. Chem. Geol. 228, 112–136 (2006).

J. M. McArthur1*, P. Ravenscroft2 and O. Sracek3

1Earth Sciences, University College London,

Gower Street, London WC1E 6BT, UK, 2AMEC Entec, 17 Angel Gate, City Road, London EC1V 2SH, 3Department of Geology, Faculty of Science, Palacký University, 17. listopadu 12, 771 46 Olomouc, Czech Republic. *e-mail: j.mcarthur@ucl.ac.uk

Authors’ reply — McArthur et al. argue that a single process, termed aquifer flushing, explains the pattern of dissolved arsenic concentrations we observe in groundwater in Bangladesh. We concur that rice-field recharge has the potential to flush out arsenic-contaminated groundwater at shallow depths; we have shown that rice-field recharge carries little arsenic or biologically available organic carbon that could mobilize arsenic from the aquifer1–3. However, we contend that their interpretation explains neither the origin of the high arsenic concentrations in groundwater located at intermediate depths, nor why concentrations decline at greater depths.

We argue that McArthur et al.’s interpretation is inconsistent with the physics that force groundwater layering. Pond recharge must flow horizontally beneath low-arsenic recharge from rice fields, to reach the irrigation wells and river channels where it discharges. Thus pond recharge should predominately occupy arsenic-contaminated intermediate depths2,4.

In their Fig. 1a, McArthur et al. compare water isotope data from one location with arsenic concentration data from multiple locations, where flow patterns differ. In fact, the δ18O minimum does align with the arsenic peak, within the resolution of the data, when data from the same wells are compared, contradicting their assertion of a

mismatch (see Supplementary Information and their Fig. 1c). Our interpretation, however, does not rely on this alignment.

Plumes of different solutes that originate from the same source are often not collocated. Solutes follow different patterns because surface sorption retards transport to varying degrees and many solutes, including arsenic, are mobilized from the aquifer. Reactive transport of most solutes in groundwater is not explained by ‘endmember’ mixing, as suggested by McArthur et al. Indeed, according to their proposed model of linear mixing, the consistent decline in arsenic concentrations with depth, below its peak, implies more flushing towards the bottom of the aquifer, an implication that we feel is physically implausible.

In their Fig. 1b, McArthur et al. do not present our measurements5 from 30 m and below, arguing that they are of different derivation and age. However, the origin of the contaminated groundwater is the question being pursued; all samples have different ages; and ponds existed before the advent of irrigation pumping. When the complete data set is plotted, the trends apparent in their Fig. 1b are no longer evident (see Supplementary Information).

Our interpretation — that water within the intermediate contaminated zone originates from pond recharge — is also supported by chemical analysis showing

that ponds provide dissolved organic carbon that is biologically available; carbon-dating analysis suggesting that organic carbon concentrations are maintained by old organic carbon released from the aquifer; and the observation that only pond water from early in the dry season can provide the isotopically light recharge matching contaminated groundwater. ❐

References1. Harvey, C. F. et al. Science 300, 584D–U3 (2003).2. Neumann, R. B. et al. Nature Geosci. 3, 46–52 (2010).3. Neumann, R. B. et al. Environ. Sci. Technol. 45, 2072–2078 (2011).4. Harvey, C. F. Nature 454, 415–416 (2008).5. Swartz, C. H. et al. Geochim. Cosmochim. Acta

68, 4539–4557 (2004).

Additional informationSupplementary information accompanies this paper on www.nature.com/naturegeoscience.

Rebecca B. Neumann1,2, Khandaker N. Ashfaque1, A. B. M. Badruzzaman3, M. Ashraf Ali3, Julie K. Shoemaker4 and Charles F. Harvey1*1Massachussetts Institute of Technology, Cambridge, Massachusetts 02139, USA, 2University of Washington, Seattle, Washington 98195, USA, 3Bangladesh University of Engineering and Technology, Dhaka 1000, Bangladesh, 4Harvard University, Cambridge, Massachusetts 02138, USA. *e-mail: charvey@mit.edu

© 2011 Macmillan Publishers Limited. All rights reserved

Comment on “Anthropogenic influences on groundwater arsenic concentrations in Bangladesh”

by Rebecca Neumann et al., Nature Geoscience, 2009: DOI 10.1038/NGO685

J.M. McArthur1, P. Ravensroft

2, and O. Srececk

3

1. Earth Sciences, UCL, Gower Street, London WC1E 6BT, UK (j.mcarthur@ucl.ac.uk);

2. AMEC, Trinity House, Cambridge Business Park, Cowley Road, Cambridge, CB4 0WZ (peter.ravenscroft@entecuk.co.uk);

3. Dep. of Geology, Palacký University, 17. listopadu 12, 771 46 Olomouc, Czech Republic (srondra@yahoo.com).

Sirs,

Neumann et al. (1a) claim that biodegradable organic carbon, mobilised into water infiltrating from man-

made ponds and concentrated by irrigation-pumping, is responsible for generating As-pollution in a

shallow aquifer in Bangladesh and, by extension, much of the Bengal Basin. The claim links human

action and As-pollution in groundwater. The reality of such a link is of profound significance for tens of

millions of domestic consumers of groundwater in the region (1b), so such claims should be well founded.

We show below, both in our main text and our accompanying reference notes, that their claim is not well-

founded, and that Neumann et al. also overlook the simplest, and best, interpretation of their data: that

isotopically-light recharge is flushing As from the aquifer in both (or all) field-sites they have studied in

the region of Munshiganj.

The claim of Neumann et al., that human activity causes As pollution, centres on a claimed association

between high concentrations of As and groundwater that is isotopically-light in composition. The

association, they believe, can be accounted for only by recharge by pond water that mobilizes As. In

making that claim, they do not characterise the isotopic signals for other sources of recharge, either in total

or for critical parts of the year, omissions that make their claim unproven at best. They do not make clear

the fact that some of their data relates to a site, or sites, other than Bailshibhog (2). They open their paper

by citing as fact a number of matteres that are, at best, contestable assertions (3a), they do not accurately

represent previous work (3b), and they force 14C data to fit a desired outcome whilst overlooking the

limitations that preclude such a use (3c). Some of their opening assertions reiterate past papers that were

strongly questioned, even if cited widely (4). In contravention of the disclosure policy of their publishing

journal, Neumann et al. have not made available the data on which their claims are based (5), despite two

requests by e-mail to provide it.

Regarding their mathematical models, we note simply that the output of their flow model depend on the

inputs and that many of the input parameters are poorly constrained, if constrained at all, by ground-truth

as opposed to bulk, pump-test, averaging e.g. the hydraulic conductivity of the upper aquitard in non-

irrigated fields, and of pond bottoms and sides. Nor has their model been calibrated rigorously against

heads and flows (6), apparently relying for calibration on a 7 month period (7) in 2006-07 that ignores

their longer-term data.

The critical ‘pond recharge water’ of Neumann et al., collected 2.7 m beneath an (unidentified) pond, has

not been shown to derive from the pond, nor has any DOC in it been shown to derive from the organic-rich

bottom of a pond (8) rather than the aquitard underlying it. No analysis of pond water is provided for

comparison to the ‘pond water recharge’. Using their figures, with a pond depth of 3 m and an aquitard 4

m thick, the water was collected 1.7 m beneath the base of the upper aquitard, in an upper aquifer in which

Kh/Kv is between 733 and 1958 (their Table 1.1). No information is given on the content or distribution of

organic carbon in the upper aquitard beneath the pond, or the pond bottom, so the source of the DOC in

their ‘pond recharge water’ might be either, or neither given the strong anisotropy of flow.

When comparing their model predictions to measured values of conservative tracers δ18O and δ2H in

water, and dissolved Cl (9), Neumann et al. do not use their model output (their Fig. 1c), which cannot

reproduce measured profiles of these tracers in the aquifer. Instead they use the ‘refined’ model output

shown in their Fig. 1e (see their Methods Section, last paragraph). The refinement includes doubling the

modelled proportion of ‘pond recharge water’ at 24 mbgl from 50 to 100 %. Their model outputs in

Figs. 1b and 1c also predict that a substantial component of ‘pond recharge water’ (8) should occur at

36.6 mbgl, but their Fig. 1.14c shows groundwater at that level around IR8, their pump-test site, to be at

least 400 years old. If this water is 400 years old, it was not drawn to that level by a diesel-powered

irrigation well (10), so the claim that pumping causes As-pollution cannot be correct.

In our Fig. 1, we compare their refined proportions of subsurface water, given in their Fig. 1e, to their

profiles of dissolved As in the aquifer at Bashailbhog given in their Fig 1a: for clarity we have combined

their two types of post-development recharge (putative ‘pond’ and ‘rice-field’ water). This combined

figure shows that around 70% of arsenic in the aquifer at Bashailbhog resides in pre-development water

i.e. was recharged prior to the construction of ponds or the introduction of pumping bores for irrigation:

concentrations of As decrease upward as the proportion of post-development recharge increases.

As δ18O is Neumann et als’ tracer of pond water, in our Fig. 2 we combine Neumann et al.’s profiles of As

(their Fig. 1a) and δ18O (their Fig. 1g). The dark blue area depicts the water that has a value of δ18O more

negative than – 4.75 ‰ (their Fig. 4.1) and so is, according to Neumann et al., partly derived from ponds

(entirely so at 24 mbgl according to their Fig. 1e). Neumann et al. state (page 57 of the Supplementary

Information) that this isotopically light water co-occurs with “the depth of maximum arsenic (~ 30m)”. In

fact, peak concentrations of As are close to 1000 µg/L and occur at 36.6 mbgl, where the δ18O has a value

of – 3.5 ‰. The minimum in δ18O of – 6.5 ‰ is at 24 mbgl, where the concentration of As is half as

much. The simplest interpretation of our Figs. 1 and 2 (and so their data) is that the proportion of

isotopically-light water is peaking at 24 mbgl and, together with recharge that is isotopically less depleted,

is flushing older, As-rich, water from the aquifer.

The Flushing Model

This flushing model has been proposed previously (11) for Bejgaon, some 800m NW of Bashailbhog in

Munshiganj (12, 13) and more widely (14a). With respect to Bejgaon, the authors of Ref 14a noted that

“concentrations of dissolved constituents derived from differing sources…all increase with depth to a peak

around 35 – 40 m. Such a concordance of profiles of chemically different species can be induced only by

flushing, not by reactions in the aquifer.” That view was summarised in a diagram, reproduced here as

Fig. 3 beside the profile of 18O and As in the aquifer at Bejgaon (from Refs 13, 15). As at Bashailbhog,

most of the As-pollution at Bejgaon is found below 30 mbgl in what is inferred, on the basis of isotopic

composition, to be pre-development water (11). The flushing model can be refined by adding to the

flushing recharge all water above 40 mbgl with δ18O more negative than – 3.25 ‰, the value around 36 to

45mbgl; i.e. by adding the light-blue area in Fig. 2.

Testing the Flushing Model

If, as we contend, the aquifer at Bejgaon and Bashailbhog is being flushed of As as post-development

water invades the aquifer, then the composition of water between the surface and pre-development water

should plot along mixing lines that define the end members. This supposition cannot be tested for

Bashialbhog beyond the result seen in Figs. 1 and 2 because Neumann et al. provide no data with which to

do so. It can be tested further at Bejgaon using data from Ref. 15 presented in Fig 4 as covariation

diagrams for dissolved constituents. There is a good linear correlation between most constituents, and a

reasonable linear correlation between DOC and As. Taken together, these cross-plots indicate that mixing

is occurring between a pre-development water high in Ca, K, NH4, DIC, DOC and As, and fresher

recharge lower in all these constituents (as noted in 11 on other criteria) that may be from more than one

source, each of which is poorly mineralized. The alternative explanations of groundwater composition

presented by Neumann et al. contradict previous interpretation (14b), or adduce an all-inclusive mixture of

processes, evidence for the operation of which either does not exist (14c) or has not been shown

convincingly to operate in an aquifer (14d), or are quoted uncritically (14e). Finally, if surface sources of

DOC (e.g. pond infiltration) were important in driving FeOOH reduction (16), the concentration of DOC

would decrease along the flowpath (with increasing age and depth) through its consumption in driving

FeOOH reduction. However, DOC increases with depth as concentrations of As increase (Fig. 1 of Ref.

12; see also Ref 15), which is why they correlate positively (our Fig. 4).

Fig. 2

The Isotopically-Light End-Member of Recharge

Is the end-member of the isotopically light recharge (both hues of blue in Fig. 2) really pond water mixed

with recharge from fields and river? Unfortunately, because Neumann et al. did not adequately

characterize the 18O value of all potential sources of recharge, the question cannot be resolved.

Regarding field recharge, the isotopic composition of pore water beneath irrigated rice fields is presented

(diagrammatically) only for March 2004 and January 2008, with no information concerning infiltration of

monsoonal rain or flood water on the land. Further, no information is presented on water that infiltrates

through the 27% of the area that is non-irrigated agricultural land. Irrigated land has a ploughpan that

reduces deep percolation (17). Non-irrigated land planted with deep-water rice would have a poorly

developed ploughpan; that planted with dryland crops that would have none (18). Being less affected by

evaporation, water infiltrating through non-irrigated fields would be isotopically lighter than that

infiltrating through irrigated fields. The failure of Neumann et al. to characterise these water sources (19),

which could account for the isotopically light water at around 24 mbgl, in itself invalidates their claim.

Regarding pond water, our Fig. 6 compares the pond-water data for δ18O of Neumann et al. with published

data for West Bengal and the Bengal Basin Meteoric Water Line (20, 21). The reduced major-axis-

regression lines for each data set are indistinguishable in slope and intercept. Through the dry season

(October to May), when ponds are capable of recharging the aquifer, values of δ18O for pond water in

Munshiganj range from – 8 to + 1 ‰ as a result of evaporation of 2.0 m per year (Fig. 3D of 13). The

mean value is about – 3.1 ‰, significantly heavier than the – 6.8 ‰ assumed by Neumann et al. to

represent ‘pond recharge water’ (their Fig. 1 and 4.1, our Fig. 2; 22). To generate aquifer water with a

18O value of – 6.8‰, Neumann et al. offer explanations that are internally inconsistent, conflict with their

own data, and discard data that does not fit with their hypothesis that ponds drive As-pollution. Because

the mean isotopic composition of ponds through the year, at around – 3.1 ‰ for 18O, does not match the

value of – 6.5 measured at 24 mbgl, Neumann et al. allow ponds to recharge groundwater only at the end

of the monsoon season when the aquifer is full and the pond is isotopically light, and not later when the

aquifer’s water table is low and water in the pond is isotopically heavy from evaporation. The justification

for this selectivity is that only “water from the middle of the pond” is lost by seepage through the edges

and that ponds are stratified. Ponds may well lose water laterally, and it is a good way for pond leakage to

bypass organic-rich pond-bottoms, so organic matter on pond bottoms cannot drive As-pollution (8, 16).

Despite this conflict, we note that their claim that their ponds are stratified is contradicted by the

concordance with the pond evaporation line for West Bengal (Fig 6; 21) of that part of their stable isotopic

data (22) not obviously affected by analytical error (23): samples from ponds in West Bengal were

collected using a depth-integrating sampler, a point they overlook. The suggestion that ponds are

significantly stratified is further contradicted by pond-water profiles containing >5mg/L of DO2 (their Fig.

4.7), which proves ponds overturn, and a pond-profile for 18O that is heavier at the bottom than at the top,

where evaporation occurs, which also does so (but see 23)

With regard to river water, Neumann et al. minimise the contribution to recharge of water from the

Ichamati River and its tributaries, which flank the study area on three sides and join the Padma River 7 km

south of Bashailbhog. Reference 12 notes that in the field area at Bejgaon, “water levels rose in May 2002

from the levels in April by ~ 1.25 m before the monsoon” and that this rise “confirms that surface water

bodies recharge the aquifer.”. The pre-monsoonal rise of the Ichamati River (Fig. 4.2 of Neumann et al.)

may be driven by the Brahmaputra, which rises strongly in April, May and June, several months before the

Ganges. Its rising would direct water into the Ichamati (Fig. 7, from 24), delivering water that is

isotopically much lighter than that in the Ganges (Table 1). Given that the upper aquitard at Munshiganj is

4 m thick, and the Ichamati River is 6 m deep (p14 of SI for 1a), lateral infiltration through the 2 m of

aquifer exposed in the river bank seems a real possibility, as its Kh is 436 times that assigned to the river

bottom (their Table 1.1). Because Neumann et al. provide no isotopic data for the Ichamati River in April,

May, or June, this possibility cannot be tested.

How Typical of Bangladesh is Bashailbhog?

Finally, we consider as unjustified the assertion of Neumann et al that their site is ‘typical’ either of

Bangladesh or any major part of the Bengal Basin. Given that the aquifer at Munshiganj is being flushed

of As, this atypicallity is to be regretted because such flushing would be beneficial to groundwater

resources. We explain our view below:

looking locally:

Munshiganj is annually flooded (in parts) by rainwater to a depth of several metres. Such flooding is

typical only of the 18% of the country that is similarly flooded annually (25).

The field area at Munshiganj is surrounded on three sides by the Ichamati River and its distributaries,

from which no part of the field area is more than 2000 m distant. Such proximity to rivers is not

typical of large parts of the Bengal Basin.

The positioning of irrigation wells within 20 m of a pond is not typical of Bangladesh. Irrigation wells

are usually sited away from ponds, excepting fish ponds, because ponds commonly substitute for

wells as a source of irrigation water.

The models of Neumann et al. presented in their Figs. 1e ( used to match aquifer profiles) and 1c,

omit recharge from the 27% of their area that is not rice-field. Such fields, lacking a well-developed

plough-pan, will be more permeable than rice fields. This omission is a tacit acceptance that neither

the site at which their large-scale pump-test was conducted (at IR8), nor its immediate surroundings,

are typical even of 27% of Munshiganj.

More widely, the agro-ecosystems, and to a lesser extent the aquifers, of the Bengal Basin are extremely

diverse in sedimentology (e.g. 21 – 23), and on that basis alone the study area of Neumann et al. can no

more represent the whole of it than can the site in West Bengal where Sengupta et al. (21) reached the

exact opposite conclusion regarding the role of DOC from ponds in causing As-pollution.

References 1a. Neumann R.B., Ashfaque K.N., Badruzzaman A.B.M., Ashraf Ali A., Shoemaker J.K. and Harvey C.F. 2009.

Anthropogenic influences on groundwater arsenic concentrations in Bangladesh. Nature Geoscience, 3, 46, 2010.

1b. Argos M., Kalra T., Rathouz P.J., Chen Y., Pierce B., Parvez F., Islam T., Ahmed A., Rakibuz-Zaman M., Hasan R.,

Sarwar G., Slavkovich V., van Geen A., Graziano J., Ahsan H. 2010. Arsenic exposure from drinking water, and all-

cause and chronic-disease mortalities in Bangladesh (HEALS): a prospective cohort study. The Lancet,

DOI:10.1016/S0140-673(10)60481-3. This paper implies that around 1 in 5 deaths in As-polluted areas of

Bangladesh may be related to As-poisoning.

2. Neumann et al. present a study of a field site at Bashailbhog, yet much of their information (e.g. the Cl profile in

Fig. 1, the 14C data, the 3H/3He dating) are for a site at Bejgaon, 800 m to the north-west of Bashailbhog.

Furthermore, in their Table 7.1, the composition for the ‘Aquifer (30 m) is for Bejgaon, 800m away from

Bashailbhog.

3a. Neumann et al. state that “The patterns of dissolved arsenic observed at a variety of sites have not been explained by

local differences in the composition of the solid aquifer material…” The reader can test the validity of this assertion

by reading the papers they cite: the pattern of dissolved As in groundwater most certainly has been explained at many

sites; for example Moyna, in West Bengal (Ref. 14 below); Matlab, Bangladesh (von Brömssen et al. 2008.

Geochemical characterisation of shallow aquifer sediments of Matlab Upazila, Southeastern Bangladesh -

implications for targeting low-As aquifers. Jour. Contam. Hydrology, 99, 137–149) and central Bangladesh

(Stollenwerk et al. 2007. Arsenic attenuation by oxidized aquifer sediments in Bangladesh. Sci. Tot. Environ., 379,

133 – 150).

3b. Citing Klump et al. (2006), Neumann et al state on page 1 of their paper that “Tritium-helium-3 analysis has shown

that the water at the depth of the arsenic peak is roughly 50 years old”. They present no such data for their field site

at Bashailbhog. The dating by Klump et al. was done for a site at Bejgaon, some 800 m NW of Bashailbhog, and at

which the a minimum in 18

O, and a peak in the concentration of As, occur at 30 m depth (Our Fig 3), rather

shallower than at Bashailbhog. Of Bejgaon, Klump et al. state that “The As concentrations are greatest at a depth of

30 m where… groundwater age significantly exceeds 30 years, indicating that recharge of most of the contaminated

water occurred before groundwater irrigation became established in Bangladesh.” and on page 247 of their article

“the fact that below a depth of 35 m the groundwater is virtually free of 3H indicates that this groundwater was

recharged prior to the atmospheric bomb peak and, hence, is at least 55 years old.”.

3.c The raw data on 14

C activity, and 13

C for fractionation correction, are not provided but presented as interpretations

only. Without raw data, the interpretations cannot be checked for accuracy nor alternative interpretations made.

Neumann et al. state that “analysis of carbon-14 dates of methane and dissolved inorganic carbon (DIC) has shown

that the carbon promoting microbial respiration was recently transported from the surface”. No such thing has been

shown, nor could be shown, with dates on methane or DIC. No 14C dating are presented for Bailshibhog. Carbon-

14 dates for DOC (from Bejgaon, Ref 12) show that the DOC is several thousand years old (uncorrected age) and so

cannot be “from the surface” in recent times, or from ponds.

4a. by Klump S. et al. 2006. Groundwater dynamics and arsenic mobilization in Bangladesh assessed using noble gases

and tritium. Environ. Sci. Technol., 40, 243-250.

4b. by McArthur et al. 2004. Natural organic matter in sedimentary basins and its relation to arsenic in anoxic

groundwater: the example of West Bengal and its worldwide implications. Applied Geochemistry, 19, 1255–1293

4c. by Aggarwal et al. 2002. Comment on "Arsenic mobility and groundwater extraction in Bangladesh" (I), Science,

300, 5619, 584B-U1.

4d. by van Geen A. et al. 2002. Comment on "Arsenic mobility and groundwater extraction in Bangladesh" (II) Science,

300, 5619, 584C-584C.

5. Almost none of the data used to plot figures are given in 89 pages of Supplementary Information of Neumann et al;

they are presented only on those figures. Neither the figures, nor the absence of raw data, allow data to be assigned

to piezometers, ponds, depths, locations or times of sampling. There is no indication which ponds were sampled, or

where in relation to pond perimeters their putative ‘pond recharge water’ was collected. They do not give the

screened interval for irrigation well IR8, on which they undertook their ‘large-scale pumping test’ etc.

6. A comparison of heads is shown in SI Figs1.10, which has an obvious error in the time axis, and Fig 1.11.

7. As shown in their Fig. 4.2, the authors have longer term water levels in ponds and groundwater, but do not appear to

have used them in model calibration.

8. Neumann et als (2009) repeated use of the term ‘pond recharge water’ is unfortunate as it is not shown to derive

from a pond. At the collection point of the ‘pond recharge water’, neither the thickness of the aquitard under the

pond, nor the concentration of organic carbon in the aquitard, are given by Neumann et al., so possible contributions

to DOC across the field area from organic matter in, or distributed through, the upper aquitard cannot be assessed.

Neither is analysis of TOC presented for bottom sediment in the pond.

9. It is not clear whether Fig. 1c refers to Bejgaon or Bashailbhog. Measured values of conservative tracers δ18

O and

δ2H in water plotted in Fig. 1f and g were collected in 2005 from an unspecified location that might be either or

neither. The dissolved Cl profile (Fig. 1h) is from Bejgaon, 800 m to the NE, according to Fig. 3 of Ref 13. In their

Fig. 1 Neumann et al use the Cl profile from Bejgaon, but do not use the 18

O profile from that site, given in Ref 13,

and provide no explanation of why they have used one but not the other.

10. Irrigation well IR8 was used for the large scale pumping test that discharged around 1300 m3 of groundwater over 24

pumping hours after a seasons use for irrigation pumping. On page 17 of the SM, Neumann et al. state that “… flow

is strongly focused at the depth of the irrigation pumping…”. They do not say what that depth is for IR8, although

this information is available because an “average depth” is given in Fig. 1.

11. Klump S., Kipfer R., Cirpka O., Harvey C.F., Brennwald M., Ashfaque K.N., Badruzzaman A., Hug S.J., and

Imboden D., 2006. Groundwater dynamics and arsenic mobilization in Bangladesh assessed using noble gases and

tritium. Environ. Sci. Technol., 40, 243-250.

12. 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. and Ahmed M.F. 2002. Arsenic mobility and

groundwater extraction in Bangladesh. Science, 298, 1602 –1606.

13. Harvey et al. 2006. Groundwater dynamics and arsenic contamination in Bangladesh. Chemical Geology 228 (2006)

112–136. Note that lack of data precludes doing the same for Bashailbhog.

14a. McArthur J. M., Banerjee D. M., Hudson-Edwards K. A., Mishra R., Purohit R., Ravenscroft P., Cronin A., Howarth

R. J., Chatterjee A., Talukder T., Lowry D., Houghton S., and Chadha D. K., 2004. Natural organic matter in

sedimentary basins and its relation to arsenic in anoxic groundwater: the example of West Bengal and its worldwide

implications. Applied Geochemistry, 19, 1255–1293 and McArthur et al. 2008.

14b. Neumann et al. state that in the aquifer at Munshiganj methane seems to behave conservatively. The evidence that it

is conservative in solution is not presented. In Ref 12, methane is claimed to derive in part from reduction of DIC in

the aquifer. It cannot both be generated in the aquifer by reduction and be conservative in solution.

14c. Neumann et al. state that “Apatite dissolution can also explain the increased calcium concentrations and the tight

correlation between arsenic and calcium at our field site.”. The statement is geochemically untenable, as well as

leaving open the reason for the tight correlation between the concentrations of K and As (our Fig. 4).

14d. ‘Desorption’ of As has not been shown to contribute to the generation of As-pollution in aquifers. The idea, and its

employment, overlooks the fact that As cannot ‘desorb’ from aquifer minerals to produce high concentrations of As

in groundwater (hundreds of µg/L) unless those minerals have first adsorbed the element from solutions of As many

times more concentrated. The process of ‘desorption’ sensu stricto will, however, lengthen the time required to flush

the aquifer of As by adding some As to the flushing tail.

14e. If weathering of mica generates As-pollution, the pollution would be universal in the aquifers of Bangladesh. It is

not present everywhere, so mica weathering cannot be important in contributing significantly to As-pollution. The

degree to which mica weathering contributes to As-pollution is also constrained to be insignificant by K/As ratios in

groundwater that around 1000 (i.e. 103) times lower than in mica.

15. Swartz C.H., Blute N.K., Badruzzman B., Ali A., Brabander D., Jay J., Besancon J., Islam S., Hemond H.F. and

Harvey C.F. 2004. Mobility of arsenic in a Bangladesh aquifer: inferences from geochemical profiles, leaching, data,

and mineralogical characterization. Geochim. Cosmochim Acta, 68, 4539 – 4557

16. Their experiments on consumption of BDOC were made under oxic conditions and have limited relevance, or none at

all, to microbial metabolism in an anoxic aquifer. See also 5 above.

17 The recent trend of replacing bullock ploughs with power tillers may break up the ploughpan and increase

infiltration.

18. Hugh Brammer, pers. comm., 2010.

19. When discussing the isotopic composition of rainfall, Neumann et al. use isotopic composition (estimated), rather

than volume-weighted isotopic composition. Their values of δ18O for rain water are not ‘data’, but rather are

interpolations made between widely separated sites, including Shillong at an elevation of 1500 m, using a computer

programme available on the website of the University of Purdue (Bowen, 2010 www.watersiotopes.org.). The

suitability of such data to the task in hand has not been established.

20. Sengupta S. and Sarkar A. 2006. Stable isotope evidence of dual (Arabian Sea and Bay of Bengal) vapour sources in

monsoonal precipitation over north India. Earth and Planetary Science Letters 250, 511 – 521

21. Sengupta S., McArthur J.M., Sarkar A., Leng M.J., Ravenscroft P., Howarth R.J., & Banerjee D.M. 2008. Do ponds

cause arsenic-pollution of groundwater in the Bengal Basin? An answer from West Bengal. Environmental Science &

Technology, 42, 5156 – 5164.

22. As expected, the mean stable isotopic composition of ponds in Munshiganj (– 3.1‰) is lighter than the mean value

for ponds in West Bengal (– 0.2 ‰) because ponds in Munshiganj, being overtopped by floodwater, are more

influenced by late monsoon rain, which is isotopically lighter than early-monsoon rain.

23. The pond-profile data (their Fig. 4.7) of Neumann et al. are wrong for either

18O, or

2H, or both, because many data

plot above the MWL for the Bengal Basin (Fig. S1, Supplementary Information). The error is probably in D; for

this reason we use 18

O in our discussion. An error in 18

O, however, might explain the upward decrease in 18

O in

the pond water profile, which is the reverse of that expected as ponds evaporate from their surfaces. The arrow on

Fig S1 points to the ‘January 2008 sample of pond water’ of Neumann et als’ Fig. 4.6. , and represents their mean

data of two profiles (Fig. 4.7 of their Supplementary Material), but data for only one pond profile seem to be figured.

24. Jian J., Webster P.J. and Hoyos C.D. 2009. Large-scale controls on Ganges and Brahmaputra river discharge on

intraseasonal and seasonal time-scales. Q. J. R. Meteorol. Soc. 135: 353–370. DOI: 10.1002/qj.384

25. Chowdhury M.R. (2003). The El Niño-Southern Oscillation (ENSO) and seasonal flooding – Bangladesh. 2003.

Theor. Appl. Climatol. 76, 105–124. DOI 10.1007/s00704-003-0001-z. Note that Fig. 5 of Chowdhury has an

incorrect scale on its right-hand side; the scale on the left-hand side is correct.

26. Stute M., Zheng Y., Schlosser P., Horneman A., Dhar R.K., Datta S., Hoque M.A., Seddique A.A., Shamsudduha M.,

Ahmed K.M. and van Geen A. 2007. Hydrological control of As concentrations in Bangladesh groundwater, Water

Resour. Res., 43, W09417, doi: 10.1029/2005WR004499.

Table 1. Isotopic composition of river water. From Dray M. 1981 (In: Symposium on isotope techniques in river basin

development, Dhaka, Bangladesh. IAEA.) and also Lambs L., Balakrishna K., Brunet F. and Probst J.L. 2005. Oxygen

and hydrogen isotopic composition of major Indian rivers: a first global assessment. Hydrol. Process., 19, 3345–3355.

DOI: 10.1002/hyp.5974.

δD δ18

O

Ganges average – 6.6 – 47.6

Brahmaputra average – 12.0 – 85.4

Fig. 1. Superposition of the modelled proportion of recharge at Neumann et als’ field site (Fig. 1e)

and their profiles of dissolved As through the aquifer (their Fig. 1a). For clarity, the latter

has been modified by combining the two types of recharge water (putative ‘pond recharge

water’, and ‘rice-field water’) shown in their Fig. 1e.

Fig. 2. Comparison of aquifer profiles of As and 18

O in the aquifer at Bashailbhog, from Neumann

et al. 2009. Dark blue area is that supposed by Neumann et al to be water falling in the field

of As-rich water. Light and dark blue are waters considered here to be participating in

aquifer flushing.

Fig. 3. Comparison of aquifer profiles of As and 18

O in the aquifer at Bejgaon, 800 m NW of

Bashailbhog. Data for As from Ref 15; data for 18

O from Ref. 13.

Fig. 4. Cross-plots of dissolved species in the Munshiganj aquifer at Bejgaon, 800m NW of

Bashailbhog (date from Ref 15) for depths to 30mbgl. We mix only to 30 m because that is

the depth of penetration of isotopically light water. The straight-line plots indicate mixing

between an end member rich in As, K, Ca, NH4, DIC, and DOC, and one or more

unmineralized end-members. The plot excludes the one piezometer at 30 mbgl that contains

2.7 mg/L SO4, and so surface-derived contamination not present in the others at 30 m.

Fig. 5. Field site of Neumann et al. 2009, cleared of model grid to allow a clear view of the relation

of ponds and piezometers, and or both to irrigation well 8 (IR8). Dark areas are ponds; a

white open circle is drawn around IR8 at a radius of 20 m.

Fig. 6. Stable isotopic composition of ponds of Neumann et al. for Munshiganj plotted together with

the Meteoric Water Line for southern West Bengal (20, modified in 21); the filled large red

circle is the ‘pond recharge water’ of Neumann et al. and represents their estimate of the

18

O of isotopic water of aquifer recharge from ponds. Also plotted are the pond data of 21

for large ponds, as their composition is better buffered than is that of small ponds. Inclusion

of all ponds from Ref 21 yields essentially the same regression, but with more noise. Ponds

at Araihazar, Bangladesh, (26) also plot along the pond regression (21).

Fig. 7. Annual average monthly hydrograph of flow in the rivers Ganges and Brahmaputra (from

Ref 24). Note that peak flows are out of phase.

Fig. 1

Fig. 2

Fig. 3.

Fig. 4.

Fig. 5.

Fig. 6.

Fig. 7.

Supplementary Material

Fig. S1. Enlargement of part of our Fig. 6 with addition of δ18

O and δ2H (crosses) values forming the

‘pond profile’ of Neumann et al. (their Fig. 4.7.). Blue line is the meteoric water line for the Bengal

Basin (data of 20 modified by 21). The pond profile values of Neumann et al. (+) scatter randomly

and many are above the BBMWL, showing that they are affected by analytical error.