pore 2 (Fig. 1) (18). As expected from spe-cific guanidinium-phosphate interactions(19), KD

min � 2 �M determined for ATPdemonstrated increased sensitivity (by a fac-tor of �1000) of apyrase sensor 2 comparedto 1 � Mg2�

n (17, 18).The most important adaptability of pore 2 to

sense various enzymes was exemplified withreal-time detection of selected classes of en-zymes for which noninvasive assays compati-ble with high-throughput substrate screeningfor the development of enzymatic carbohydrateand oligosaccharide synthesis (10–12) are need-ed. As with apyrase, rabbit muscle aldolase(RAMA) (10) and calf intestine alkaline phos-phatase (CIAP) (11) converted strong channelblockers into poor ones (Table 1, entries 5 to 7).Enzyme-catalyzed substrate consumption wasthus reported by the decreasing ability of thereaction mixture to hinder CF efflux through 2with increasing reaction time, i.e., “enzyme-gated” pore formation (Fig. 3, A to C). Bovinemilk galactosyltransferase (11, 12), in contrast,produced uridine diphosphate (UDP) with KD-min � 1.2 mM from two poor blockers (Table1, entry 8). This enzyme was therefore detect-able by decreasing activity of pore 2 with in-creasing reaction time (Fig. 3D). Use of CIAPto eliminate UDP blockage (Fig. 3D, arrow)corroborated the proposed mechanism anddemonstrated the possibility to detect two en-zymes simultaneously or, from a different pointof view, to turn synthetic multifunctional pores“off ” and “on” using enzymes only.

Taken together, these results demonstratepractical usefulness of synthetic supramo-lecular pores for fluorometric detection ofenzyme activity, a topic of current scientificinterest (13). The most important character-istic of our assay is adaptability of the samesupramolecular sensor to a broad variety ofsubstrates and enzymes. There is no need toprepare new, substrate- or product-specificantibodies [as for catalytic assays using en-zyme-linked immunosorbent assay (cat-ELISA) and related assays] or synthetic re-ceptors for individual enzymes; no need tolabel substrates as in fluorogenic and radio-

assays; and no need to calibrate coupled mul-tienzyme assays (13, 14). Whereas intrinsiclimitations concerning, e.g., hydrophobic orlytic substrates, certain more complex ana-lytes (18), continuous assay, or in vivo detec-tion will be challenging to overcome, at leastin a general manner, there is much room toimplement various in vitro formats compati-ble with high-throughput screening and min-iaturization (1).

References and Notes1. H. Bayley, P. S. Cremer, Nature 413, 226 (2001).2. S. Cheley, L. Gu, H. Bayley, Chem. Biol. 9, 829 (2002).3. P. Scrimin, P. Tecilla, Curr. Opin. Chem. Biol. 3, 730

(1999).4. G. W. Gokel, A. Mukhopadhyay, Chem. Soc. Rev. 30,

274 (2001).

5. G. J. Kirkovits, C. D. Hall, Adv. Supramol. Chem. 7, 1(2000).

6. S. Matile, Chem. Soc. Rev. 30, 158 (2001).7. G. Das, S. Matile, Proc. Natl. Acad. Sci. U.S.A. 99,

5183 (2002).8. G. Das, H. Onouchi, E. Yashima, N. Sakai, S. Matile,

Chembiochem 3, 1089 (2002).9. Without high-resolution crystallographic evidence, it

is not incorrect to perceive the depicted suprastruc-tures as, at worst, productive working hypothesesthat are consistent with all (6–8) functions discov-ered so far.

10. T. D. Machajewski, C.-H. Wong, Angew. Chem. Int. Ed.39, 1352 (2000).

11. M. M. Palcic, Methods Enzymol. 230, 300 (1994).12. N. Wymer, E. J. Toone, Curr. Opin. Chem. Biol. 4, 110

(2000).13. D. Wahler, J.-L. Reymond, Curr. Opin. Chem. Biol. 5,

152 (2001).14. A. W. Czarnik, Acc. Chem. Res. 27, 302 (1994).15. M. Komoszynski, A. Wojtczak, Biochim. Biophys. Acta

1310, 233 (1996).16. G. M. Whitesides et al., Acc. Chem. Res. 28, 37

(1995).17. All reported KD values are conditional to the method

(Hill analysis of concentration dependence) specifiedin (8).

18. See supporting data on Science Online.19. T. S. Snowden, E. V. Anslyn, Curr. Opin. Chem. Biol. 3,

740 (1999).20. G. Das, P. Talukdar, S. Matile, data not shown.21. We thank J.-L. Reymond and N. Sakai for advice, J.

Gruenberg for access to still video camera and fluo-rescence counter, and the Swiss NSF (2000-064818.01 and National Research Program “Su-pramolecular Functional Materials” 4047-057496)for financial support.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/298/5598/1600/DC1Materials and MethodsReferences and Notes

14 August 2002; accepted 14 October 2002

Arsenic Mobility andGroundwater Extraction in

BangladeshCharles F. Harvey,1 Christopher H. Swartz,1*

A. B. M. Badruzzaman,2 Nicole Keon-Blute,1 Winston Yu,1

M. Ashraf Ali,2 Jenny Jay,1 Roger Beckie,3 Volker Niedan,1

Daniel Brabander,1† Peter M. Oates,1 Khandaker N. Ashfaque,1

Shafiqul Islam,4 Harold F. Hemond,1 M. Feroze Ahmed2

High levels of arsenic in well water are causing widespread poisoning in Ban-gladesh. In a typical aquifer in southern Bangladesh, chemical data imply thatarsenic mobilization is associated with recent inflow of carbon. High concen-trations of radiocarbon-young methane indicate that young carbon has drivenrecent biogeochemical processes, and irrigation pumping is sufficient to havedrawn water to the depth where dissolved arsenic is at a maximum. The resultsof field injection of molasses, nitrate, and low-arsenic water show that organiccarbon or its degradation products may quickly mobilize arsenic, oxidants maylower arsenic concentrations, and sorption of arsenic is limited by saturationof aquifer materials.

In an effort to prevent waterborne disease,Bangladesh shifted its drinking water supplyfrom surface to groundwater. Many of the

recently installed 6 to 10 million drinking-water wells contain high concentrations ofarsenic (1), causing widespread arsenicosis

Fig. 3. Real-time enzyme screening with(A) apyrase, (B) aldolase, (C) alkalinephosphatase, and (D) galactosyltrans-ferase without and (arrow) with CIAP,using 2. Fractional pore activity Y (18)after addition of 20 �l of reaction mix-ture and 2 to LUVs � CF is shown as afunction of reaction time. Reaction mix-tures: (A) 0 (�) and 4 (F) U/ml apyrase,40 mM ATP; (B) 0 (�), 2 (F), and 16 (E)U/ml RAMA, 750 U/ml triosephosphateisomerase (TIM), 40 mM D-fructose 1,6-diphosphate (FDP); (C) 0 (�), 66 (F), and670 (E) U/ml CIAP, 800 mM UDP; (D) 0(�), 8 (F, arrow: addition of 66 U/mlCIAP), and 32 (E) U/ml galactosyltrans-ferase, 250 mM N-acetylglucosamine (Gl-cNAc), 375 mM UDPGal [or 375 mM NAcUDPGal (20)] (18).

R E P O R T S

22 NOVEMBER 2002 VOL 298 SCIENCE www.sciencemag.org1602

and threatening to increase cancer rates (2).Sediments typically contain only modest lev-els of arsenic (1), and the cause and timing ofarsenic mobilization from these sedimentsremain unclear (3). Some have argued thatarsenic is mobilized by slow reduction of ironoxyhydroxides or sorbed arsenate by detritalorganic carbon (1, 4, 5). Others have suggest-ed that arsenic may have recently been re-leased through sulfide oxidation reactions in-duced by the massive increase in dry-seasonirrigation pumping (6, 7). Here, we charac-terize aqueous and solid phases at one site toshow that young carbon brought to depth byrecent irrigation pumping plays a role inmobilization.

Our study site was located in the Munshi-ganj district, 30 km south of Dhaka and 7 kmnorth of the Ganges River. We installed 17

sampling wells within 5 m of each other todepths from 5 to 165 m, sampled 54 moni-toring wells in a 16-km2 region, and collectedthree sediment cores [e.g., (8)]. The site istypical of southern Bangladesh (9) and in-cludes 3 m of surficial clay, a 100-m aquiferof gray sand (Holocene aquifer), a 40-m aqui-tard of marine clay, and a deep burnt-orangesandy aquifer (Pleistocene aquifer). The aqui-fers are �85% quartz and feldspar by weight,with the remainder composed of biotite,hornblende, magnetite, and a variety of otherminerals. We found no peat, which has beenhypothesized to be associated with dissolvedAs (10). The pH of the groundwater was 6.6(4.6 m) to 7.1 (107 m) in the upper aquifer,and 6.8 in the lower aquifer (165 m). In bothaquifers, redox measurements by platinumelectrode are between 20 and 90 mV (correct-ed to the standard hydrogen electrode), andonly one depth (14 m) has a dissolved O2

concentration (0.33 mg/liter) greater than 0.2mg/liter (our approximate limit of detection).

Dissolved As concentrations [�90%As(III)] peaked at a depth of 30 to 40 m (Fig.1A). Solid-phase As exceeded 27 nmol/g (2ppm) in only one sample from the sandyaquifers and nevr exceeded 100 nmol/g (8ppm) in the clays. Solid-phase As (Fig. 1B) isadsorbed, coprecipitated in solids leachableby mild acids and reductants, incorporated insilicates, and incorporated in solids that are

dissolved only by cold and hot nitric acid(most likely crystalline sulfides) [e.g., (8,11)]. The distribution of arsenic among thesepools is similar throughout the deep and shal-low sediments, with no apparent correspon-dence to the peak in dissolved As.

The inverse relation of dissolved sulfatewith arsenic (Fig. 1C) in the natural ground-water, and the presence of acid volatilesulfide (AVS) in the sediments near the dis-solved As peak, suggest that oxidative disso-lution of pyrite has not liberated arsenic. In-stead, low dissolved sulfur levels appear tolimit the precipitation of arsenic sulfides nearthe arsenic peak. To test whether oxidantsmay mobilize arsenic, we injected nitrate intothe aquifer. Dissolved As levels declined(Fig. 2A), probably as a result of adsorptionon iron oxyhydroxides precipitated by a mi-crobial process such as described in (12).Arsenic concentrations remained low evenafter the volume extracted exceeded the vol-ume injected, perhaps indicating arsenic re-moval in a geochemically altered zone ofincreased sorption capacity around the injec-tion well.

Solid-phase As and Fe appear to be asso-ciated. Arsenic (Fig. 1B) and iron releasedfrom sediments in the first five extractants ina sequential extraction procedure are corre-lated (r2 � 0.64) among samples from differ-ent depths, although dissolved Fe(II) (ranging

1Ralph M. Parsons Laboratory, Department of Civil &Environmental Engineering, Massachusetts Instituteof Technology, Cambridge, MA 01239, USA. 2Ban-gladesh University of Engineering and Technology,Dhaka 1000, Bangladesh. 3Department of Earth andOcean Sciences, University of British Columbia, BCV6T 1Z1, Canada. 4Department of Civil and Environ-mental Engineering, University of Cincinnati, Cincin-nati, OH 45221, USA.

*Present address: Silent Spring Institute, Newton, MA02458, USA.†Present address: Environmental, Coastal and OceanSciences Department, University of Massachusetts,Boston, MA 02125, USA.

Fig. 1. Vertical profiles of subsurface geochemical characteristics in Munshiganj. (A) Dissolved As levels peak at 8.5 �M or 640 ppb, well above theBangladeshi standard of 50 ppb. (B) Arsenic dissolved by the sequence of extractions that target different solid phases [e.g., (8, 11)]. (C) Dissolved Asand sulfate are inversely related. (D) Dissolved As, ammonium, and calcium are closely correlated. (E) DIC, DOC, and methane increase with depth to31 m.

R E P O R T S

www.sciencemag.org SCIENCE VOL 298 22 NOVEMBER 2002 1603

from 27 to 180 �M) and solid-phase Fe (rang-ing from 1 to 3 wt % in the upper aquifer and 1wt % in the lower aquifer) have no apparentsystematic depth dependence. The sedimentcontains little, if any, purely ferric [Fe(III)]oxyhydroxides, which have been hypothesizedto sequester and release arsenic (1, 4). Extrac-tion results show that 60% of total solid-phaseFe was bound in silicates and pyritic phases,relatively immobile pools in anoxic water. The10% of total iron extractable by 1 M HCl,which targets amorphous iron oxides, carbon-ates, and phosphates, was ferrous [Fe(II)] iron,as determined with ferrozine. Oxalic acid ex-traction, which targets amorphous iron oxyhy-droxides and magnetite, and Ti(III)-citrate-EDTA extraction, which reductively dissolvesiron oxides (and perhaps forms complexes withamorphous sulfides), each released 6% of totaliron. Most of the iron in these pools can beexplained as magnetite (a mixed-valencephase), which was measured by magnetic sep-aration. Thus, if ferric oxyhydroxides controlarsenic mobility, they do so in small quantities.

Surface sorption sites appeared to be sat-urated with arsenic where dissolved concen-trations were high. Measured dissolved Asconcentrations from a field test, where waterwas withdrawn after injecting low-Asgroundwater into a high-As depth (Fig. 2B),were closely fit by a model that incorporatesa Langmuir isotherm describing saturation ofsurface sites, but not by a linear isotherm [e.g.(8)]. The fitted Langmuir isotherm agreesremarkably well with dissolved and sorbedAs concentrations measured at differentdepths (Fig. 2C). These results suggest thatsorption sites saturate when dissolved Asconcentrations exceed 100 ppb; the implica-tion is that when aqueous concentrations arehigh, little additional arsenic can be sorbed.Further arsenic inputs would likely be mani-fest as increased dissolved concentrationsthat are readily transported by flowinggroundwater. High concentrations of phos-

phate and silicate (�50 �M and �700 �M,respectively, and nearly uniform with depth)as well as carbonate may contribute to thelimited capacity to sorb arsenic oxyanions(13, 14). Iron phosphates and amorphous sil-ica were both calculated to be supersaturatedfrom measured aqueous concentrations, andwe observed these phases by scanning elec-tron microscopy as rims on host grains.

Ambient geochemical concentrations andthe results of a field redox manipulationexperiment both suggest that respiration of or-ganic carbon plays a role in arsenic mobiliza-tion. Mobilization may be driven by reductionwith organic carbon despite the paucity of ferricoxyhydroxides: Equilibrium chemical model-ing indicates the dissolved As could have beensorbed by ferric oxyhydroxides in quantities ofless than 5% of the iron liberated by oxalic acid.

Mobilization may also occur by displacementof arsenic by carbonate, a process recently pro-posed for ferric oxyhydroxides (13). Either pro-cess, or a combination, is consistent with theobserved geochemical profiles. DissolvedNH4

� and Ca2� profiles follow the dissolvedAs profile remarkably closely throughout theshallow aquifer (Fig. 1D), and dissolved or-ganic and inorganic carbon (DOC and DIC)increase with increasing arsenic to the peakdepth (Fig. 1E). The correlation betweenNH4

� and Ca2� may be explained by respi-ration of DOC producing ammonia and CO2,with the latter inducing calcite dissolution.The molar ratio of carbonate to ammoniumwas �70 (50 after subtracting carbonate lib-erated from calcite, assuming a 1:1 molarratio of calcite carbonate to dissolved calcium),within the range for terrestrial organic material.

Fig. 2. Arsenic levelsresponded within hoursafter chemical perturba-tions in three injection-withdrawal experiments:(i) nitrate, 500 liters fromthe 38-m well [7.2 �MAs, with nitrate (200 mg/liter) added] into a 31-mwell (concentration atthe well before injectionC0 � 6.5 �M); (ii) molas-ses, 8000 liters from the160-m well [�0.01 �MAs, with molasses (300mg/liter) added] into a 105-m well (C0 � 0.8 �M As) filling �40 m3 ofaquifer material; and (iii) low-As water, 6052 liters of clean water from the160-mwell (�0.01 �MAs) into a 31-m well (C0 � 8.5 �MAs). All injectionsincluded bromide (100 mg/liter) as tracer. (A) Arsenic concentrations (nor-malized by C0) in withdrawn water after nitrate and molasses injections. (B)The arsenic rebound after injection of low-As water is well fit by a Langmuirisotherm, but not by a linear isotherm that does not represent limited

surface sites [e.g. (8)]. (C) The shape of the Langmuir isotherm is confirmedby the load of sorbed arsenic from core samples (shown in Fig. 1B) normal-ized by the easily extracted (HCl and oxalate) Fe and plotted against thedissolved As (Fig. 1A) for each depth. Error bars represent standard devia-tions of at least three samples from the 3-m well screen interval for eachdepth. The sorbed As predicted by the Langmuir model is normalized by 0.6wt % Fe, the approximate value from core samples.

-30 -25 -20 -15 -10 -5 0

Inorganic

Carbon

Organic

Carbon

30m

19m3mMethanogenesis

CaCO3

(Atmospheric ∆14C valuesover last 50 years)

60m

-1000

-500

0

500

1000

-70

Methane

Surface Water Carbon

0700

4000

2000

Rad

iocarb

on

Ag

e

13C (‰)

∆14 C

(‰

)

δ

Fig. 3. Carbon isotope valuesfrom groundwater samples.The �14C ratios have beencorrected for fractionationwith the 13C ratios from thesame samples. The circle onthe orange line indicates theweighted average of meth-ane (38 m depth) and DIC (31m) isotope values, represent-ing the isotopic values thatDIC would have had beforefractionation during carbon-ate reduction. The green linerepresents mixing betweenyoung terrestrial organic car-bon (changing values due tobomb testing are shown bythe rising and falling arrow)and geologic carbonate,which is found in the sedi-ments. Because the inflowingDIC likely mixes with someold DIC, the radiocarbon dateshown on the right axis mayoverestimate the age of theinflowing carbon. Symbolscorrespond to aquifer depth, as shown.

R E P O R T S

22 NOVEMBER 2002 VOL 298 SCIENCE www.sciencemag.org1604

Field injection of molasses-amendedgroundwater produced an initial rise in ar-senic concentration (Fig. 2A) concurrent withdrops in platinum electrode readings (100 to–230 mV ) and pH (6.7 to 5.6), followed by adrop in arsenic concentrations. Dissolved sul-fate, which was introduced from the molas-ses, also fell (from �300 �M to �5 �M).These chemical changes are consistent with ascenario where the introduction of labile or-ganic carbon caused microbially mediateddissolution of iron solids and liberation ofarsenic, followed by precipitation of sulfidesthat sequestered arsenic.

As a further test, we analyzed carbon iso-topes of groundwater. The radiocarbon agesfor DOC are 3000 to 5000 years, consistentwith wood fragments found at similar depthsin nearby sediment (15). However, the radio-carbon ages for DIC in the same water sam-ples are much younger, about 700 years at31 m depth (Fig. 3), and 14C is at bomb levelsat 19 m, indicating carbon younger than 50years. Because the young DIC cannot be theoxidative product of the older DOC, the pres-ence of young DIC and old DOC in the samewater samples must be explained either bymixing of invading younger water with olderresident water, or by mobilization of olderorganic carbon from aquifer sediment intoinvading younger water. The similar profilesof DIC and DOC (Fig. 1E) are inconsistentwith pore water mixing, which would create anegative correlation—recently rechargedwater would contain predominantly youngcarbon, whereas old resident water would

contain predominantly old carbon. The cor-relation instead is consistent with mobiliza-tion of old carbon through geochemical per-turbations caused by the invasion of youngwater transporting young carbon, perhaps byrelease from iron oxyhydroxides, which areknown to sequester organic carbon (16).

Methane concentrations reached 1.3 mM at38 m (sufficient to effervesce flammable gas)and have approximately the same radiocarbonage as the DIC (Fig. 3). Because methane is anend product of carbon metabolism under anoxicconditions, recent biogeochemical reactions ap-pear to have been driven by relatively youngcarbon carried by recent inflow of groundwater.The deuterium ratio of the methane (D �197‰) is consistent with fractionation duringmethanogenesis by carbonate reduction (17).The similar radiocarbon ages of methane andDIC also suggest that the methane was derivedfrom the DIC pool. Ongoing carbonate reduc-tion is consistent with high levels of dissolvedhydrogen (10 nM at 38 m). Finally, the 13Cvalues of both methane and DIC can be ex-plained by fractionation during carbonate re-duction. The weighted mean 13C value frommethane at 38 m and carbonate at 31 m is–25‰, a reasonable value for terrestrial carbon.Dissolution of radiocarbon-dead solid carbon-ate, as we hypothesize, may have also movedthe isotopic ratios along the mixing line inFig. 3.

Groundwater pumping may have recentlydrawn young water into the upper aquifer atthe Munshiganj site, where irrigation pump-ing largely controls groundwater flow during

the dry season. Bangladesh is self-sufficientin food supply largely because irrigated dry-season crops are grown in addition to thetraditional crops after the monsoon. Duringthe irrigation season, January through March2002, the average water table elevationdropped 2.3 cm/day (interpolated from the 45monitoring wells in the 16-km2 area aroundthe sampling-well cluster; Fig. 4A). Thedrawdown calculated from the pumping ratesof the 38 irrigation wells was similar: 4.9cm/day if applied over only the 20% of landirrigated, and 1.2 cm/day when distributedover the entire area (18). These pumping-driven downward velocities imply plug-flowtravel times to 30 m of 6.8 and 28 years,respectively. Thus, young carbon could bequickly transported to depth. Before the ad-vent of irrigation pumping (�25 years ago),flow to depths of 30 m was extremely slow,limited by the negligible natural (i.e., withoutpumping) hydraulic gradients in the flat to-pography. Very little groundwater flow oc-curs during flooding ( June to November)because inundation erases hydraulic gradientsand downward flow by compression of wateris negligible.

Recharging water is composed of a mix-ture of carbon-rich surface water (which en-ters the aquifer during the dry season) andrainwater (which enters the aquifer fromMarch until flooding begins in June). Com-parison of rainfall, evaporation, and changesin groundwater storage at a nearby long-termmonitoring site (Fig. 4B) indicate that surfacewater contributes �0.3 m (�40% of re-

2

1.5

2.5

3

3.5

2

2.5

3

21.5

3

3.5

2

2.51.5

2.5

2.5

21.5

Watertable Drawdown Jan. 2002 [cm/day]

500 meters N

Irrigation WellDrinking WellMonit. BridgeField Site

Monitored Sites

A

Monit. Pond

Jan '85 Jul Jan Jul Jan Jul Jan Jul Jan Jul '90

0

20

40

-20

cm/m

on

th Net

Dis

char

ge

Net

Rec

harg

e Change in Groundwater Storage

Precipitation minus Evaporation

Flo

odin

g B

Recharge from Surface WaterRecharge from Surface Water

Fig. 4. (A) Water table drawdown contoured on an IKONOS satelliteimage with well positions. (B) Groundwater storage changes (draw-down times porosity) and net atmospheric recharge (rainfall minusevapotranspiration) determined from meteorological data. The dif-ference between the change in storage (blue line) and the netatmospheric recharge (red line) is attributed to exchange with surfacewater. Surface water entered the aquifer when the change in ground-water storage was greater than net atmospheric recharge (shown inyellow). Inflow is evident during periods of water table decline (greenstrips) that typically occur from November to April, as well as duringperiods of water table rise (white strips), typically from April to June.During flooding (blue strips), the recorded hydraulic heads representflood stage rather than changes in groundwater storage.

R E P O R T S

www.sciencemag.org SCIENCE VOL 298 22 NOVEMBER 2002 1605

charge) yearly, whereas precipitation (lessevaporation) contributes �0.5 m (�60%).The recent observation that water levels rosein May 2002 from the levels in April by�1.25 m before the monsoon confirms thatsurface water bodies recharge the aquifer.Infiltrating surface water receives heavyloadings from the untreated waste of the�17,600 inhabitants of the 16-km2 area.Hence this water contains high concentra-tions of dissolved carbon, as would water thatinfiltrates from rice paddies and through or-ganic-rich pond and river sediments.

The observed arsenic mobility appears re-lated to recent inflow of carbon through ei-ther organic carbon–driven reduction or dis-placement by carbonate. The distinctly olderradiocarbon ages of DOC relative to DIC andmethane imply that mobilization is not drivenby detrital organic carbon. Water budgetsindicate that the advent of massive irrigationpumping has drawn relatively young waterinto the aquifer over the last several decades,as may be typical of Bangladesh (19). Thus,irrigation pumping may affect arsenic con-centrations, but not by the oxidation of sul-fides as has been proposed. The low arsenicconcentrations from the deeper aquifer indi-cate that deep wells may provide a source ofclean water, an option that is already beingimplemented on an ad hoc basis (20). How-ever, the apparent relation of arsenic mobilityto inflow of organic carbon raises concernsabout the appropriate depth of new drinking-water wells and their position relative to irri-gation wells.

References and Notes1. D. G. Kinniburgh, P. L. Smedley, Eds., Arsenic Contam-

ination of Groundwater in Bangladesh (British Geo-logic Survey Report WC/00/19, 2001) (www.bgs.ac.uk/arsenic/Bangladesh).

2. A. H. Smith, E. O. Lingas, M. Rahman, Bull. WHO 78,1093 (2000).

3. C. F. Harvey, Environ. Sci. 8, 491 (2001).4. R. Nickson et al., Nature 395, 338 (1998).5. P. L. Smedley, D. G. Kinniburgh, Appl. Geochem. 17,

517 (2002).6. T. R. Chowdhury et al., Nature 401, 545 (1999).7. D. Das et al., Bore-hole soil-sediment analysis of

some As affected areas, in Proc. Int. Conf. on Arsenicin Groundwater: Cause, Effect and Remedy, Calcutta,December 1995.

8. See supporting data on Science Online.9. F. H. Khan, Geology of Bangladesh (University Press,

Dhaka, Bangladesh, 1991).10. J. M. McArthur et al., Water Resour. Res. 37, 109

(2001).11. N. E. Keon et al., Environ. Sci. Technol. 35, 2778

(2001).12. D. B. Senn, H. F. Hemond, Science 296, 2373 (2002).13. C. A. J. Appelo et al., Environ. Sci. Technol. 36, 3096

(2002).14. D. K. Nordstrom, in Minor Elements 2000: Processing

and Environmental Aspects of As, Sb, Se, Te, and Bi, C.Young, Ed. (Society for Mining, Metallurgy, and Ex-ploration, Littleton, CO, 2000), pp. 21–30.

15. S. L. Goodbred Jr., S. A. Kuehl, Geology 27, 559(1999).

16. C. J. Yapp, H. Poths, Geochim. Cosmochim. Acta 50,1213 (1986).

17. M. J. Whiticar, Chem. Geol. 161, 291 (1999).18. No rivers flow from the area during the dry season, so

pumped groundwater that is not evaporated ortransferred into rice fields or ponds must re-infil-trate back into the aquifer. Thus, the drawdown(the net change in aquifer storage) during thisperiod is controlled by the rate of evapotranspira-tion and transfer of groundwater with rice fieldsand ponds. The drawdown due to evapotranspira-tion, estimated by applying the Penman-Monteithequation to local meteorological data collected bythe Bangladesh Water Development Board, is con-sistent with this water budget, increasing from 1.1cm/day in January to 2.2 cm/day in March. Al-though no regional gradient is evident from themeasured hydraulic heads, strong lateral flowscontrolled by the position of irrigation wells andrecharge areas may create complex pathways forinvading water. Applying Darcy’s law to interpolat-ed hydraulic heads and hydraulic conductivity es-

timated from pump tests gives an average lateralcomponent to groundwater velocity of 1.7 cm/day,similar to the vertical component.

19. C. F. Harvey, Science 296, 1563 (2002).20. W. H. Yu, C. M. Harvey, C. F. Harvey, Water Resour.

Res., in press.21. Supported by NSF grant EAR-0001098, NSF graduate

fellowships, a grant from the Alliance for GlobalSustainability, and National Institute of Environmen-tal Health Sciences grant P30ES02109.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/298/5598/1602/DC1Materials and MethodsReferences

5 August 2002; accepted 21 October 2002

Grasping Primate OriginsJonathan I. Bloch* and Doug M. Boyer

The evolutionary history that led to Eocene-and-later primates of modern aspect(Euprimates) has been uncertain. We describe a skeleton of Paleocene plesiadapi-form Carpolestes simpsoni that includes most of the skull and many postcranialbones. Phylogenetic analyses indicate that Carpolestidae are closely related toEuprimates. C. simpsoni had long fingers and an opposable hallux with a nail. Itlacked orbital convergence and an ankle specialized for leaping. We infer that theancestor of Euprimates was primitively an arboreal grasper adapted for terminalbranch feeding rather than a specialized leaper or visually directed predator.

Extant primates are distinct from other euther-ian mammals in having large brains, enhancedvision brought about in part by optical conver-

gence, the ability to leap, nails on at least thefirst toes, and grasping hands and feet (1, 2).Several adaptive scenarios have been proposedto explain these specializations: (i) “grasp-leap-ing” locomotion (3), which predicts simulta-neous evolution of grasping and leaping; (ii)visually directed predation (4), which predictssimultaneous evolution of forward-facing orbits

Museum of Paleontology, University of Michigan,1109 Geddes Road, Ann Arbor, MI 48109–1079, USA.*To whom correspondence should be addressed. E-mail: carpo@umich.edu

Fig. 1. Hypothesis of phy-logenetic relationshipsamong select archontans,illustrating phylogeneticposition of Carpolestidaebased on cladistic analysisof 65 postcranial charac-ters (25). Many fossil ar-chontans were excludedfrom the analysis becausepostcranial skeletons forthese groups have not yetbeen described. Cladisticanalysis yielded a singlemost-parsimonious cla-dogram generated by anexhaustive search algo-rithm and rooted withAsioryctes: tree length �117, consistency index �0.75, retention index �0.67. All characters wereunordered. As for previ-ous cladistic analyses(10), the topology supports a plesiadapiform-euprimate link, whereas the cladogram based onpostcranial data presented here specifically allies Carpolestidae with Euprimates (Omomyidae plusAdapidae). Unambiguous synapomorphies supporting a carpolestid-euprimate link include mor-phology of distal humerus (character 10), a nail on the hallux (character 21), a sellar joint betweenmetatarsal I and the entocuneiform (character 50), and lateral torsion of the distal metatarsal I(character 65).

R E P O R T S

22 NOVEMBER 2002 VOL 298 SCIENCE www.sciencemag.org1606

Comment on “Arsenic Mobilityand Groundwater Extraction in

Bangladesh” (I)Harvey et al. (1) suggested that the elevatedarsenic levels found in Bangladesh aquifersare caused by desorption of arsenic solidsaccompanying the influx of fresh, labile car-bon-laden recharge water. They further con-cluded that rapid aquifer recharge is only arecent phenomenon, resulting from extensiveirrigation pumping over the last 25 years.

This latter conclusion was based on theirobservation of “young carbon” to depths ofabout 30 m, indicating relatively quick verti-cal travel times of �7 to 28 years (comparedwith flow times preceding the advent of irri-gation pumping). Harvey et al. then usedshort-term water level records to estimateexcess recharge resulting from irrigationpumping. However, they did not cite anyspecific evidence for their assumption of adrastic increase in dry-season vertical flow.Here, we show that irrigation has had at besta negligible impact on vertical groundwaterflow and, therefore, arsenic concentrations inBangladesh aquifers.

Our studies (2, 3) agree with the con-clusion in (1) that desorption due to con-tinuing recharge causes elevated arsenicconcentrations in Bangladesh aquifers.

Strontium isotope data and groundwaterage estimations (3–5) indicate that arsenicconcentrations generally increase with in-creasing groundwater residence time in theaquifer. As we have shown previously (2,5), the exponential increase in irrigationpumping since the mid-1970s did not sig-nificantly change stable oxygen and hydro-

gen isotope compositions of shallowgroundwaters between 1979 and 1999. Thisindicates that the source and mechanism ofrecharge remained essentially the same dur-ing this period (domestic and irrigationwells from across Bangladesh were used inthese surveys, and some wells were com-mon to both surveys). The depth of tritiumpenetration in 1979 was about 45 m, andbomb tritium was clearly present at depthsof �22 m (Fig. 1). These data suggestvertical travel times of �20 to 27 years, todepths of about 45 m under pre- and post-irrigation pumping conditions in 1979 and1999 [assuming piston flow and pre-bombatmospheric tritium levels of 5 tritium units(TU)], similar to those estimated by Harveyet al. (1). Alkalinity and stable carbon iso-tope values were also similar in the 1979

and 1999 investigations (6), which indi-cates that carbon loadings and bio-geochemical processes related to organiccarbon oxidation and carbonate dissolutionremained the same before and after signif-icant irrigation pumping. Based on theseisotopic data, we conclude that irrigationpumping in Bangladesh is not necessarilyresponsible for the transport of fresh re-charge or labile carbon loadings to depthsof 30 m or more, contrary to the assumptionmade by Harvey et al. (1).

Long-term records of water level fluctua-tion provide further evidence for the lack ofsignificant irrigation pumping impacts ondry-season vertical hydraulic gradients inBangladesh. Fig. 2 shows weekly measure-ments of water level between 1967 and 1997in a 31-m deep observation well in the Far-idpur area, southwest of the capital city ofDhaka (3). This hydrograph is typical of shal-low aquifer observation wells across Bang-ladesh (3). It shows that seasonal fluctuations ofabout 5 m in the water level of shallow aquifers,where elevated arsenic concentrations are ob-served today, have occurred consistently overthe 30-year period—even during the late1960s, before significant groundwater with-drawals by irrigation pumping. For the locationshown in Fig. 2, the mean depth to groundwaterremained at about 3.7 m, but the standard de-viation decreased from 2 m between 1967 and1983 to 1.6 m between 1984 and 1997. Lowerseasonal fluctuations over the last 25 years,when extensive irrigation pumping took place,clearly indicate that vertical hydraulic gradientsand dry-season recharge did not increase duringthe dry season, refuting the basic premise forthe conclusion reached in (1). By using a 5-yearrecord (1985 to 1990) from a similar observa-tion well to argue that irrigation pumping hasresulted in substantial increase in recharge, Har-vey et al. may have presented a flawed analy-sis—unless it could be shown that seasonalfluctuations were much less in the pre-irrigationpumping period. However, this appears unlike-ly, based on the isotopic data (3, 5) and ourreview (3) of water level records of a number ofobservation wells across Bangladesh similar tothat shown in Fig. 2.

Pradeep K. AggarwalIsotope Hydrology Section

International Atomic Energy AgencyPost Office Box 100

A1400 Vienna, AustriaE-mail: P.Aggarwal@IAEA.ORG

Asish R. BasuDepartment of Earth and

Environmental SciencesUniversity of Rochester

Rochester, NY 14627, USA

Fig 1. Tritium concentrations in Bangladesh groundwater in 1979 and 1999. Domestic or irrigationwells from northwestern and southern Bangladesh were sampled in both events. Measurabletritium concentrations (�1 TU) in 1979 clearly indicate the presence of modern recharge to depthsof �45 m, while bomb tritium is observed at depths of 22 m. Tritium profiles indicate similarvertical travel times in pre- and post-irrigation pumping periods.

TECHNICAL COMMENTS

www.sciencemag.org SCIENCE VOL 300 25 APRIL 2003 584b

Kshitij M. KulkarniIsotope Hydrology Section

International Atomic Energy Agency

References1. C. F. Harvey et al., Science 298, 1602 (2002).2. P. K. Aggarwal et al., EOS 81 (fall suppl.), F523 (2000).3. P. K. Aggarwal et al., “Isotope hydrology of ground-

water in Bangladesh: Implications for Characteriza-tion and Mitigation of Arsenic in Groundwater”(IAEA-TC Project Report: BGD/8/016, InternationalAtomic Energy Agency, Vienna, 2000; www.iaea.org/programmes/ripc/ih/publications/bgd_report.pdf ).

4. A. R. Basu, S. B. Jacobsen, R. J. Poreda, C. B. Dowling,P. K. Aggarwal, Science 293, 1470 (2001).

5. A. R. Basu, S. B. Jacobsen, R. J. Poreda, C. B. Dowling,P. K. Aggarwal, Science 296, 1563a (2002).

6. IAEA, unpublished data.

20 December 2002; accepted 3 March 2003

Fig. 2. Water-level fluctuations in observation well FA-01 (depth � 31 m) located in ThanaFaridpur, southwest of Dhaka (3). Seasonal fluctuations of nearly 5 m are observed between 1967and 1997, before and after the advent of irrigation pumping. The standard deviation of seasonalfluctuation is slightly lower for the period between 1984 and 1997, indicating that verticalhydraulic gradients may have decreased due to irrigation pumping and return recharge—in contrastto the increase proposed by Harvey et al.

T E C H N I C A L C O M M E N T S

25 APRIL 2003 VOL 300 SCIENCE www.sciencemag.org584b

Comment on “Arsenic Mobilityand Groundwater Extraction in

Bangladesh” (II)

Harvey et al. (1) recently presented detailedvertical profiles for groundwater arsenic and asuite of other water and sediment properties froma study site in southern Bangladesh. This infor-mation, supplemented with carbon isotopic data(13C and 14C) for dissolved organic carbon(DOC) and dissolved inorganic carbon (DIC),was used to support the notion that the DOCintroduced into subsurface aquifers with ground-water recharge could lead to reduction or disso-lution of iron oxyhydroxides and the subsequentrelease of associated arsenic into groundwater.This is an important issue for tens of millions ofpeople in Bangladesh and other South Asiancountries, who are at risk of contracting variouscancers from drinking water with elevated ar-senic levels (2). The process of arsenic mobili-zation could be much more dynamic than previ-ously thought if, as Harvey et al. have proposed(1), the leading agent for arsenic release fromiron oxyhydroxides is indeed DOC penetratingsandy aquifers on time scales of tens of yearsrather than particulate organic carbon depositedover several thousand years throughout theGanges-Brahmaputra-Meghna delta (3, 4). Onthe basis of particulate and dissolved sulfur data,Harvey et al. also have effectively laid to rest theargument that the oxidation and dissolution ofparticulate sulfide is the dominant cause of ele-vated arsenic in groundwater, at least for theirstudy site.

We are troubled, however, by the argumentthat groundwater pumping for irrigation in-creased the penetration of DOC into the subsur-face and therefore caused the release of arsenicinto the groundwater. We believe that the data in(1) and related information from other regions ofBangladesh could be equally well explainedwithout invoking a response of groundwaterarsenic to irrigation pumping. This is an impor-tant point, because policy-makers in Bangladeshand elsewhere should not be confronted with thefalse dilemma of elevated arsenic in groundwa-ter caused by irrigation or insufficient agricul-tural production.

The Harvey et al. data appear to be inconsis-

tent with recent mobilization of arsenic by irri-gation pumping, if their model of vertical flow istaken at face value. As Harvey et al. point out,the concentration of radiocarbon in the atmo-sphere increased to unprecedented levels startingin the 1950s because of atmospheric testing ofnuclear bombs (5). The penetration of radiocar-bon above pre-testing levels in subsurface aqui-fers is an indicator of groundwater recharge overthe past 50 years. The onset of this particularanthropogenic perturbation therefore precededthe onset of massive irrigation in Bangladesh byabout 25 years. According to their vertical flowmodel, this would mean bomb-produced 14Cshould have penetrated at least to the depth ofmaximum arsenic mobilization if this featurewas indeed caused by an enhanced supply ofDOC linked to irrigation. Yet, according to fig-ures 1A and 3 in (1), the maximum in arsenicmobilization is observed at about 40 m depth,while only the two shallowest radiocarbon sam-ples, at 3 and 19 m, indicate addition of bombradiocarbon. Thus, the Harvey et al. data do notnecessarily indicate a direct connection betweenarsenic mobilization and increased irrigationpumping in their study area.

Even if the portion of the Harvey et al.argument that we dispute were correct in theirstudy area, there is convincing evidence that itdoes not apply to other parts of Bangladesh.Tritium (3H) is another radionuclide whose con-centration in the atmosphere increased dramati-cally in response to atmospheric bomb testing (6,7). The distribution of 3H in water has thereforealso been used extensively to study oceanic cir-culation as well as groundwater recharge. A re-cent report from the International Atomic EnergyAgency (8) lists eight samples with significantlyelevated arsenic levels that do not contain anydetectable 3H. We have collected and analyzedan additional 49 groundwater samples from otherwells in Bangladesh that include a set of 5 pairedarsenic and 3H analyses indicating high arseniclevels without detectable 3H (9, 10). The impli-cation of the combined data set is that over adozen carefully analyzed samples indicate that

Bangladesh groundwater was elevated in arsenicwell before the onset of massive irrigation. Wetherefore believe that increased irrigation overthe past 25 years is unlikely to have causedwidespread arsenic mobilization in Bangladeshgroundwater through the sequence of steps pro-posed by Harvey et al. in their otherwise veryvaluable contribution.

Alexander van GeenLamont-Doherty Earth Observatory

Columbia UniversityRoute 9W

Palisades, NY 10964, USAE-mail: avangeen@ldeo.columbia.edu

Yan ZhengLamont-Doherty Earth Observatory

and Queens College,City University of New York

Flushing, NY 11367, USA

Martin StuteLamont-Doherty Earth Observatory

and Barnard College, Columbia University3009 Broadway New York, NY 10027, USA

Kazi Matin AhmedDepartment of Geology

University of DhakaDhaka 1000, Bangladesh

References and Notes1. C. F. Harvey et al., Science 298, 1602 (2002).2. D. G. Kinniburgh, P. L. Smedley, Eds., Arsenic Contam-ination of Groundwater in Bangladesh (British Geo-logical Survey Report WC/00/19, 2001; www.bgs.ac.uk/arsenic/Bangladesh).

3. R. Nickson et al., Nature 395, 338 (1998).4. J. M. McArthur et al., Water Resour. Res. 37, 109 (2001).5. I. Levin et al., Radiocarbon 27, 1 (1985).6. W. Weiss, J. Bullacher, W. Roether, “Evidence of

pulsed discharges of tritium from nuclear energyinstallations in central European precipitation: Behav-iour of Tritium in the Environment” (InternationalAtomic Energy Agency, Vienna, 1979), pp. 17–30.

7. M. Stute, “Tritium (3H) in precipitation in Bang-ladesh” in Groundwater Contamination in the BengalDelta Plain of Bangladesh [Royal Technical Institute(KTH) Special Publication TRITA-AMI Report 3084,2001], pp. 109–117.

8. P. K. Aggarwal et al., “Isotope hydrology of ground-water in Bangladesh: Implications for Characteriza-tion and Mitigation of Arsenic in Groundwater”(IAEA-TC Project Report, BGD/8/016, InternationalAtomic Energy Agency, Vienna, 2000; www.iaea.org/programmes/ripc/ih/publications/bgd_report.pdf ).

9. Y. Zheng et al., Appl. Geochem., in press.10. M. Stute, P. Schlosser, unpublished data.11. Our work in Bangladesh has been supported by grant

P42ES10349 of the U.S. NIEHS/Superfund Basic Re-search Program and NSF grant EAR 0119933.

2 December 2002; accepted 3 March 2003

TECHNICAL COMMENTS

www.sciencemag.org SCIENCE VOL 300 25 APRIL 2003 584c

Response to Comments on“Arsenic Mobility and Groundwater

Extraction in Bangladesh”Aggarwal et al. (1) and van Geen et al. (2)point to selected data in (3) to craft contrast-ing regional generalizations about the causeof high arsenic concentrations in groundwaterin Bangladesh. Aggarwal et al. argue thatarsenic mobilization is related to continuinggroundwater recharge, but that groundwaterflow is not influenced by groundwater extrac-tion, whereas van Geen et al. argue thatarsenic mobilization occurs in stagnantgroundwater. Both claims are contradicted bymultiple sources of hydraulic and geochemi-cal data from our site (4), contain seriousinconsistencies, and ignore basic hydrologicprocesses.

Aggarwal et al. (1) argue that extensiveextraction of groundwater for irrigation doesnot affect groundwater flow at our study site.They incorrectly assert that we inferred (4)the effects of pumping from water isotopesand water table movement. Three points areimportant to consider. First, our value ofpumping-induced flow was calculated fromknowledge of the location and pumping ratesof irrigation wells [figure 4 in (4)]. There isno need to infer pumping-induced flowthrough indirect means. Second, Aggarwal etal. provide no physical principle supportingtheir contention that irrigation withdrawaldoes not cause groundwater flow. Pumpingfrom a well does draw water to the wellscreen and must drive flow through the aqui-fer. Third, site data clearly show the effects ofpumping. Hydraulic head declines sharply atirrigation onset (Fig. 1A), and flow converges

at the depth of the arsenic peak later in theirrigation season (Fig. 1B), two effects forwhich we see no plausible explanation otherthan pumping.

Despite the physical implausibility of theircontention that pumping does not affect flow,Aggarwal et al. interpret water isotopes (tritiumand stable isotopes) from other areas of Bang-ladesh to support their argument. Contrary totheir interpretation, the tritium data are consis-tent with increased groundwater flow to depthafter the onset of groundwater irrigation. Noneof the wells below 25 m had tritium valuesabove 1 tritium unit (TU) in 1979. However,four wells showed measurable tritium in 1999,consistent with migration of young water toextraction depths after the marked increase inirrigation with pumped groundwater from1979 to 1999. As for the stable isotope data,irrigation could create heavier groundwater ifincreased evaporation fractionates irrigationwater before it reinfiltrates the aquifer. Fromtheir plot of paired deuterium and oxygenisotope ratios in 1979 and 1999 [figure 1 in(5)], Aggarwal et al. conclude that “pumpingsince the mid-1970s did not significantlychange stable oxygen and hydrogen isotopecompositions...” However, simple statisticaltests confirm that the 1999 waters do differfrom the 1979 waters, in direct conflict withtheir inference. A t test on the mean residualsaround the meteoric line (6) rejects the hy-pothesis that the two data sets derive from thesame population (P �0.01). Most of the 1979data plot above the meteoric line, and most of

the 1999 plot below. This shift in stableisotopes provides compelling evidence forthe effects of irrigation pumping becausethree processes work to mute the change.First, crop transpiration, which accounts formuch of the evaporation during irrigation,will not fractionate return water (7). Second,irrigation wells were sampled in 1979, clearlyindicating that these data do not representpreirrigation conditions. Third, recharging ir-rigation water is greatly diluted by surfaceand monsoon waters.

Aggarwal et al. also argue that pumpingdoes not affect flow by relying on a singlegroundwater hydrograph collected 40 kmfrom our site on the other side of the GangesRiver. However, 183 of the 252 groundwaterhydrographs collected over the 10-year peri-od (1988 to 1997) show trends of falling min-imum water levels (8). According to the Ban-gladesh Ministry of Water Resources, “Theimpact of increasing groundwater develop-ment can often be clearly observed from hy-drograph” (9). Furthermore, at Sreenagar, 4km from our site, the water table minimumduring the irrigation season dropped 1 m overthe last 17 years (P �0.001). The water tableminimum is a far better indicator of pumpingwithdrawals than the mean and variance usedin (1), which largely reflect rainfall and floodlevel during the wet season. However, eventhe minimum cannot be directly equated withflow below the water table. Evaporation inthe flat topography of Bangladesh can causewater table decline by simply removing waterfrom the near surface without driving ground-water flow. In contrast, pumping must driveflow, but the resulting water table declinemay be substantially diminished by infiltra-tion of irrigation return water. These hydro-logic dynamics, compounded by exchangewith ponds and rivers and rapid lateral flowinduced by pumping, argue strongly for rely-ing on direct measurement of pumping rates.

Van Geen et al. (2) overlook our hydraulicdata and most of our chemical data to disputethe importance of pumping. First, they avoidthe physical fact that pumping must affectgroundwater flow. Hydraulic data in Fig. 1provide an example at the arsenic peak. Sec-ond, they discount bomb levels of 14C indissolved inorganic carbon (DIC) at 19 m,despite the fact that at this depth arseniclevels sharply increase towards the peak. Tri-tium measurements support these radiocar-bon measurements and show that a portion ofthe groundwater throughout the arsenic peak,to 61 m, must be young (10). Third, they basetheir argument on the false premise that theonset of pumping must precede the radiocar-bon age of the bulk dissolved carbon forpumping to play a role in mobilization.

Fig. 1. (A) Hydraulic head at 32 m, the arsenic peak, 2001-2002. (B) Vertical head profile in March2002 showing convergence at �40 m.

TECHNICAL COMMENTS

www.sciencemag.org SCIENCE VOL 300 25 APRIL 2003 584d

Pumping may transport carbon, perhaps fromnearby sediment, that can drive mobilizationregardless of its bulk radiocarbon age. Pump-ing also transports young carbon from thesurface and mixes it with old. Indeed, vanGeen et al. do not object to our inferencesthat (i) pumping has altered groundwaterflow; (ii) clearly, DIC has been transportedby groundwater flow to the depth of thearsenic peak because the DIC radiocarbonage is much younger than the local sedimentage, and (iii) this carbon was involved inarsenic mobilization. By accepting these in-ferences, van Geen et al. apparently contra-dict their claim that arsenic was mobilized instagnant water unaffected by pumping.

In addition to neglecting the data andprocesses noted above, van Geen et al. dis-regard the accepted fact that dissolved carboncontains a mixture of radiocarbon ages as anecessary consequence of solute mixing andexchange with solid carbon. Five types ofgeochemical and hydrologic evidence indi-cate that beneath 19 m, a portion of the DICand methane derives from recently producedorganic carbon: (i) Tritium data show thatrecent recharge constitutes at least a portionof the groundwater to 61 m (10). Surfacecarbon transported by this recent recharge islikely also modern. (ii) Pumping rates aresufficient to have drawn water from the sur-face into the arsenic peak. (iii) The calciumpeak, which very closely follows the arsenicpeak, is likely the product of carbonate dis-solution driven by production or inflow ofDIC. Calcium concentrations indicate that30% of the DIC is derived from carbonate.Assuming that this carbonate has a radiocar-bon age 3000 years, the remaining 70% ofDIC must be primarily modern, as demon-strated by the mixing line in figure 3 in (4).(iv) The DOC is much older than the DIC andmethane in the same groundwater, so if DOCof this age has been converted, substantialyoung carbon must also have also been in-corporated to obtain the observed bulk radio-carbon ages. (v) The convergence of ground-water from above and below the arsenic peak(Fig. 1B) accelerates the lateral flow within astreamtube passing through this depth, andlikely mixes young carbon from above andold carbon from below. In essence, invadingyoung carbon must mix with resident oldcarbon, yet van Geen et al. ignore this factand present no hypothesis to reconcile theirclaim of stagnant groundwater in light of themeasured DIC age being a fraction of the ageof DOC and sediment.

Van Geen et al. argue against the impor-tance of groundwater flow regionally by se-lectively focusing on only eight out of morethan 50 samples from IAEA data, and five outof 49 unpublished values, where arsenic con-centration is high but tritium is undetected.For these samples, they boldly claim to know

arsenic concentrations decades before waterwas actually sampled, neglecting the possi-bility that solutes that interact with arsenic, orsimply arsenic itself, have been transported inthe old water. They then extrapolate theirinference from these selected recent samplesto generalize that arsenic concentrations havenot changed throughout the whole countryduring the last 30 years of pumping.

We disagree with the contention that ar-senic concentrations are immune to ground-water dynamics throughout Bangladesh. Datafrom our site point to the importance of trans-port and mixing due to flow and show nosimple uniform relation between arsenic con-centrations and water age. Arsenic levels arelow at shallow depths where water is young-est and most affected by pumping, but arealso low where water is oldest, deep in theaquifer. The highest concentrations occurwhere young and old water appear to mix,and tritium is still detected. Furthermore,mean arsenic levels in irrigation wells (120ppb, n � 23), which receive the most rapidflow from the surface, is significantly (P�0.05) lower than in domestic wells (200ppb, n � 32). This is consistent with thehypothesis that, although flow to irrigationwells may help mobilize arsenic, pumpinghas begun to flush arsenic from streamtubesleading to these irrigation wells. In someareas, irrigation pumping may be flushingarsenic from the system or introducing oxi-dants that immobilize arsenic. The ground-water flow systems in Bangladesh are com-plex, and conditions vary. Natural topograph-ic flow in some areas may control arseniclevels, as per Aggarwal et al., whereas inother stagnant areas (i.e., below clay layers),arsenic mobilized by detrital carbon (11) mayremain constant, as per van Geen et al. How-ever, data support neither of those two situa-tions at our site.

Our interpretation that the band of higharsenic has been caused by an inflow ofcarbon should not be misinterpreted to meanthat wherever wells are pumping, they arecurrently increasing arsenic concentrations.Such a misreading both oversimplifies andoverextends our results. Likewise, our find-ings that arsenic levels are low at shallowdepths most affected by irrigation, and lowerin irrigation than household wells, should notbe misconstrued to argue that irrigationpumping will remedy the problem acrossBangladesh. We do, however, hope that ourpaper will help to dispel assumptions thatgroundwater arsenic concentrations are im-mune to changes in groundwater flow, re-charge, and consequent chemical changes.Billions of dollars have been spent trackingthe movement of groundwater contaminantsunder the premise that groundwater flow af-fects solute concentrations. We see no reasonwhy Bangladesh would be different—ample

evidence documents strong spatial gradientsin dissolved arsenic and groundwater extrac-tion guarantees groundwater flow.

C. F. HarveyRalph M. Parsons Laboratory

Department of Civil &Environmental Engineering

Massachusetts Institute of TechnologyCambridge, MA 01239, USA

E-mail: charvey@mit.edu

C. Swartz*Ralph M. Parsons Laboratory

Massachusetts Institute of Technology

A. B. M. BadruzzamanBangladesh University of Engineering and

TechnologyDhaka 1000, Bangladesh

N. Keon-BluteW. Yu

Ralph M. Parsons LaboratoryMassachusetts Institute of Technology

M. A. AliBangladesh University of Engineering and

Technology

J. JayRalph M. Parsons Laboratory

Massachusetts Institute of Technology

R. BeckieDepartment of Earth and Ocean Sciences

University of British ColumbiaBC V6T 1Z1, Canada

V. NiedanD. Brabander†

P. OatesK. Ashfaque

Ralph M. Parsons LaboratoryMassachusetts Institute of Technology

S. IslamDepartment of Civil and

Environmental EngineeringUniversity of Cincinnati

Cincinnati, OH 45221, USA

H. HemondRalph M. Parsons Laboratory

Massachusetts Institute of Technology

M. F. AhmedBangladesh University of

Engineering and Technology

*Present address: Silent Spring Institute, Newton,MA 02458, USA. †Present address: Environmental,Coastal and Ocean Sciences Department, Universi-ty of Massachusetts, Boston, MA 02125, USA

T E C H N I C A L C O M M E N T S

25 APRIL 2003 VOL 300 SCIENCE www.sciencemag.org584d

References and Notes1. P. K. Aggarwal, A. R. Basu, K. M. Kulkarni, Science 300,

xxx (2003); www.sciencemag.org/cgi/content/full/300/5619/584b.

2. A. van Geen, Y. Zheng, M. Stute, K. M. Ahmed, Science300, xxx (2003); www.sciencemag.org/cgi/content/full/300/5619/584c.

3. P. K. Aggarwal et al., “Isotope hydrology of groundwaterin Bangladesh: Implications for Characteriza tion andMitigation of Arsenic in Groundwater” (IAEA-TC ProjectReport: BGD/8/016, International Atomic Energy Agen-cy, Vienna, 2002; www.iaea.org/programmes/ripc/ih/publications/bgd_report.pdf ).

4. C. F. Harvey et al., Science 298, 1602 (2002).5. A. R. Basu, S. B. Jacobsen, R. J. Poreda, C. B. Dowling,

P. K. Aggarwal, Science 296, 1563a (2002).6. Data extracted from figure 1 in (5).7. J. R. Gat, Annu. Rev. Earth Planet. Sci. 24, 225 (1996).8. Water Resources Planning Organization, Ministry

Water Resources, People’s Republic of Bangladesh,National Water Management Plan (Annex C, Appen-dix 9, 2000), vols. 1 to 11.

9. Water Resources Planning Organization, MinistryWater Resources, People’s Republic of Bangladesh,National Water Management Plan (Annex C, Appen-dix 6, p. 52, 2000), vols. 1 to 11.

10. Tritium values were 6.5 TU at 19 m; 4.10 at 24 m;

0.42, 0.74, and 0.93 in three wells at 32 m; 0.20 at38 m; 0.81 at 61 m; and 0.02 at 107 m, withmeasurement errors of approximately 0.1 IU. Thesevalues indicate that most of the groundwater at 19-and 24-m depths infiltrated in the last 50 years, andthat from 24 m to 61 m groundwater contains aportion of recent recharge. Analysis performed by W.Aeschbach-Hertig and R. Kipfer, EAWAG and ETH,Zurich.

11. P. Ravenscroft et al., in Arsenic Exposure and HealthEffects IV, W. Chappell, C. Abernathey, R. Calderon,Eds. (Elsevier, Oxford, 2002).

23 January 2003; accepted 11 March 2003

T E C H N I C A L C O M M E N T S

www.sciencemag.org SCIENCE VOL 300 25 APRIL 2003 584d