Redox trapping of arsenic during groundwater discharge in sediments from the Meghna riverbank in Bangladesh
- aLamont-Doherty Earth Observatory of Columbia University, 61 Route 9W, Palisades, New York, NY 10964;
- bKansas State University, Department of Geology, Manhattan, KS 66506;
- cBarnard College, Department of Environmental Sciences, New York, NY 10027;
- dQueens College, School of Earth and Environmental Sciences, City University of New York, Flushing, New York, NY 11367; and
- eUniversity of Dhaka, Department of Geology, Dhaka, 1000 Bangladesh
See allHide authors and affiliations
-
Communicated by Charles H. Langmuir, Harvard University, Cambridge, MA, July 30, 2009 (received for review September 5, 2007)

Abstract
Groundwater arsenic (As) is elevated in the shallow Holocene aquifers of Bangladesh. In the dry season, the shallow groundwater discharges to major rivers. This process may influence the chemistry of the river and the hyporheic zone sediment. To assess the fate of As during discharge, surface (0–5 cm) and subsurface (1–3 m) sediment samples were collected at 9 sites from the bank of the Meghna River along a transect from its northern source (25° N) to the Bay of Bengal (22.5° N). Bulk As concentrations of surface sediment averaged 16 ± 7 mg/kg (n = 9). Subsurface sediment contained higher mean concentrations of As of 4,000 mg/kg (n = 14), ranging from 1 to 23,000 mg/kg As, with >100 mg/kg As measured at 8 sites. X-ray absorption near-edge structure spectroscopy indicated that As was mainly arsenate and arsenite, not As-bearing sulfides. We hypothesize that the elevated sediment As concentrations form as As-rich groundwater discharges to the river, and enters a more oxidizing environment. A significant portion of dissolved As sorbs to iron-bearing minerals, which form a natural reactive barrier. Recycling of this sediment-bound As to the Ganges-Brahmaputra-Meghna Delta aquifer provides a potential source of As to further contaminate groundwater. Furthermore, chemical fluxes from groundwater discharge from the Ganges-Brahmaputra-Meghna Delta may be less than previous estimates because this barrier can immobilize many elements.
In the Ganges-Brahmaputra-Meghna Delta (GBMD) of India and Bangladesh, more than 30 million people have been drinking groundwater with elevated levels of arsenic (As) for at least the last decade (1). Many studies have focused on the biogeochemical processes that release As in the GBMD aquifers, but little work has been done on the fate of As during groundwater discharge. In Bangladesh, rivers are potential areas of groundwater discharge (2). Because of the change in redox conditions at the interface, As may precipitate or sorb onto iron (Fe)-bearing sediments in the riverbank or riverbed.
It has been recently demonstrated that when anoxic groundwater discharges to the ocean or to a lake, Fe in groundwater is oxidatively precipitated onto the sands (3), immobilizing many elements, including As and phosphorus (4). Growing evidence supports the important role that processes in the hyporheic zone have in regulating the composition of groundwater discharging to rivers (5–7). In this study, we examine accumulations of As bound to riverbank sediment and implications these enrichments have on As cycling in Bangladesh. We determined the As concentrations of shallow sediments along the entire length of the Meghna River. This system was chosen because most of the shallow groundwater As east of the Meghna River between 22 and 24° N (Fig. 1) contains >50 μg/L As (1). Our results have important implications for cycling of As in deltaic environments and groundwater discharge as a source of elements to the ocean.
Locations of sediment sampling sites along the Meghna River are indicated by circles with dots in the center. White, light gray, and dark gray circles represent HCl-leachable As concentrations in subsurface sediments at a depth between 1 and 3 m, with values of <10 mg/kg, between 100 and 1,000 mg/kg, and between 1,000 and 23,000 mg/kg, respectively. There are no samples with concentrations between 10 and 100 mg/kg. At 14 other sites reported in a previous study (15), maximum HCl-leachable As concentrations in shallow sediments (1–3 m) are indicated by open squares for those with values between 6 and 30 mg/kg and by light gray squares for those with values between 100 and 758 mg/kg. Groundwater As concentrations for wells with depth <25 m (n = 743) are plotted as blue, red, and dark red circles to represent <50, 50 to <100, and 100 to 1,090 μg/L (1).
Results
Aquifers in the GBMD.
The Ganges and Brahmaputra rivers coalesce northwest of Dhaka and then join the Meghna River south of Dhaka before flowing into the Bay of Bengal (see Fig. 1). Bangladesh is less than 10 m above sea level, except for the uplifted Pleistocene terraces, the Chittagong Hills, and the hilly area in the northeast. The sandy, unconsolidated Pleistocene to Holocene fluvial and deltaic sediments that underline much of Bangladesh form prolific aquifers in this highly energetic depositional environment. The shallow Holocene aquifers are typically elevated in As, whereas the older Pleistocene aquifers appear to contain little As. The shallow aquifer sediments have been deposited since ≈10,000 to 11,000 years B.P. (1). During the early Holocene (11,000–7,000 years B.P.), the river sediment flux was sufficient to infill 50-m incisions formed during glacial periods of low sea level. The deeper Pleistocene aquifers are typically separated from the shallow Holocene aquifers by multiple layers of silt and clay. Aquifers are recharged from precipitation directly or through flooding during the wet season (1). Shallow aquifers are discharged via flow into the rivers and other low-lying surface water bodies, as well as by evapotranspiration.
Sediment samples were collected on the Meghna riverbank at 9 locations in January 2003 (Table 1) and at 2 locations in January 2006 (Fig. S1). In 2006, continuous sediment cores from surface to up to 6-m depth were obtained. In 2003, surface (0–5 cm) and subsurface (variable depth between 1 and 3 m) sediment samples were obtained at each site. Four sites are located on the tributaries to the Meghna River at ≈25° N at an elevation of 3 to 10 m. Five sites are located on the main channel of the Meghna River between 22.5° N and 24° N at 0 to 3 m elevation. Eight sites are located on sandy deposits along the riverbank. RS-1 is located on a sandbar in the river. At RS-4, an additional sediment core was obtained from the riverbed (see Table 1). Samples from the top 5 cm and bottom 5 cm were subjected to leaching and acid digestion (see SI Analytical and Spectroscopic Methods) before determination of As (see Table 1).
Composition of the Meghna riverbank sediment
As in Surface Sediment.
Bulk As concentrations of surface sediment samples subjected to total acid dissolution ranged from 7 to 27 mg/kg, averaging 16 ± 7 mg/kg (n = 9) (see Table 1). These values were similar to As concentrations in soils from around the world (8) and in Bangladesh (9).
As in Subsurface Sediment.
High concentrations of As (≈100 to ≈20,000 mg/kg) were extracted from the sediment using 1.2 N hydrochloric acid (HCl) at 8 of the 9 sample sites (see Table 1 and Fig. 2). The concentrations of As released by 1-M phosphate leaches were elevated (≈200 to ≈1,000 mg/kg) at 7 sites (see Table 1). Six samples with 3 to 1,000 mg/kg of phosphate-extractable As displayed comparable HCl-extractable As concentrations (R2 = 0.99) (see Table 1). Solid residues from the phosphate extractions of these samples had low concentrations of As ranging from 3 to 30 mg/kg (see Table 1). Collectively, these results indicate that the majority of the As in these 6 subsurface sediment samples are easily extracted.
Comparison of concentrations of As in surface sediment obtained by acid digestion and in subsurface sediment obtained by HCl-leach.
For 5 samples with very high HCl-extractable As (≈4,000 to ≈20,000 mg/kg), As concentrations were also elevated in the residues of the phosphate extracts, ranging from 500 to ≈3,000 mg/kg (see Table 1). However, the sum of the phosphate-extractable As and the residual As after phosphate-extraction, should be higher than, or at least equal to the HCl-extractable As. This was not the case for these 5 samples. This discrepancy is most likely caused by leaving ≈1 ml of phosphate solution to react with the sediments for about a year before the residual sediment was separated for total acid dissolution. We suspect that As continued to be leached by this discarded aliquot of phosphate solution. Nevertheless, the results are consistent with greater concentrations of As bound in the phases that are leached either by phosphate or HCl in subsurface sediments than is bound in residual more refractory phases.
Depth Profiles of As and Fe(II)/Fe.
Enrichment of sediment As was found at shallow depths, corresponding to a redox transition zone, as indicated by the increasing sediment Fe(II)/Fe ratio with depth (Fig. 3). Samples were analyzed for HCl-leachable Fe(II)/Fe within 12 h to characterize sediment redox condition. At RS-30, the depth interval from 0.75 to 0.90 m displayed increasing bulk sediment As concentration from ≈30 mg/kg to ≈200 mg/kg. The sediment from this depth interval was gray and characterized by a high Fe(II)/Fe ratio of ≈0.85, although the sediment immediately above at 0.3 m had a low Fe(II)/Fe ratio of 0.25. At RS-33, depth intervals of 0.24 m to 0.30 m, and 1.45 to 1.5 m displayed sediment As enrichment up to ≈100 mg/kg. The shallower interval had sediment Fe(II)/Fe of 0.48, whereas the deeper interval had sediment Fe(II)/Fe of 0.24. Sediment Fe(II)/Fe reached 0.9 at 3 m, suggesting that this location has a thick redox transition zone. One sample displayed higher HCl-leachable As concentration than bulk sediment As content (see Fig. 3). The difference likely reflects the focused analysis of the extraction on a 0.5-g sample compared to the concentration averaged over a 5-cm depth interval.
Depth profiles of bulk (solid square) and HCl-leachable (open square) sediment As concentrations, and sediment HCl-leachable Fe(II)/Fe for RS-30 (0–2 m) and RS-33 (0–6.5 m), located close to RS-11 (see Fig. S1).
Acid Leached Fe and Manganese in Sediment.
In contrast to the 5 orders of magnitude variations of HCl-extractable As concentrations (see Fig. 2), HCl-extractable Fe varied little, averaging 2.9 ± 1.0% (see Table 1). HCl-extractable manganese concentrations were 570 ± 370 mg/kg (see Table 1).
Mineralogy of As and Fe in Subsurface Sediment.
The X-ray absorption near-edge structure (XANES) spectra of 6 subsurface sediment samples from 5 sites indicated that As is present as arsenate and arsenite. These 2 species accounted for 79 to 98% of As in the sample (Table 2). Only in RS-9, which is located in the downstream area of the Meghna River where tidal influences are stronger and which can provide sulfate in seawater (see Fig. 1), was where 56% of As found to be associated with sulfides.
Arsenic species in subsurface sediment by XANES
That As in the subsurface sediment is of mixed arsenate and arsenite form is consistent with the differential pulse cathodic stripping voltammetry (DPCSV) analysis of phosphate-extract solutions (see Table 1). Compared to the total As determined by high-resolution inductively coupled plasma (HR ICP)-MS on the phosphate extract, the DPSCV analyses indicate that most samples have only a small fraction of phosphate-extractable As as arsenate (see Table 1). Results of the DPSCV and XANES analyses indicate similar arsenic speciation. In RS-4, 85% of phosphate-extractable As is arsenate, whereas XANES identified 79%. In RS-11, 12% of phosphate-extractable As is arsenate, whereas XANES identified 0%. RS-9 displayed much higher phosphate-extractable As(III) than total As quantified by HR ICP-MS beyond the 1 sigma error (≈25%) of this measurement, possibly because of a very large dilution factor (250 times) used in the analyses.
The extended X-ray absorption fine structures (EXAFS), near edge (XANES), and X-ray diffraction spectra of 11 subsurface sediment samples from 8 sites show several Fe-bearing minerals, including biotite, hornblende, goethite, magnetite, vivianite, but not siderite (see SI Analytical and Spectroscopic Methods). Fe-bearing clays were present at all sites and represented the sum of chlorite, illite, and hydrated smectite. Goethite was identified at sites along the tributaries of the Meghna River but not along the main channels. Goethite was used to represent all of the Fe (III) hydroxides, as the spectra are similar for these minerals. Further work is needed to confirm that reduced Fe minerals, such as green rust or vivianite, are important components of reactive barrier along the main channel of the Meghna River, which is consistent with sediment Fe(II)/Fe data (see Fig. 3).
Discussion
A Natural Reactive Barrier for Groundwater As.
The enrichment of As, several orders of magnitude higher than the crustal abundance of As, observed in subsurface sediment along nearly the entire length of the Meghna River, cannot be explained by processes of sediment deposition. Suspended particulate matter sampled from dozens of locations from the Ganges, Brahmaputra, and Meghna Rivers in Bangladesh displayed crustal values of As from 4 mg/kg to 5.5 mg/kg (10, 11). Surface sediment samples from all 9 sites also show bulk sediment concentrations <20 mg/kg, suggesting the source of enrichment is in the subsurface.
Therefore, enrichment of As in subsurface sediment implies that groundwater is the source of As. The enrichment occurs in a redox transition zone that is saturated during the wet season but unsaturated during the dry season. Hydrological and geochemical comparisons between Waquiot Bay, Massachusetts (4) and the Meghna River in this study (Table 3) support this interpretation. Sediment samples were obtained at sandy locations along the Meghna River, similar to the setting at Waquiot Bay, where the shallow aquifer consists of permeable sand. Because of the high permeability, a short residence time of <10 years was observed for the upper 10 m of the aquifer at Waquiot Bay and in Bangladesh. Charette and Sholkovitz (3) reported that oxidative precipitation of groundwater Fe at a few m depths below the surface at Waquiot Bay (4). This “iron curtain” acted as a barrier, preventing a large plume of groundwater-derived phosphorus and several other elements from entering the surface water of the bay. We postulate that a similar natural reactive barrier consists of secondary Fe minerals also formed along the Meghna River. The redox transition zone may extend to different depths, depending on variation of the seasonality, with a particularly dry year corresponding to a very deep penetrating unsaturated zone, and thus may explain the multiple layers of enrichment observed in RS-33. Secondary Fe minerals formed in situ, goethite, and vivianite were present in our samples (Table S1). Magnetite and Fe-bearing clays were also present, but they could be secondary or primary. Green rust may also be part of a consortium of secondary Fe minerals in this barrier.
Comparison between Waquoit Bay and the Meghna River
Can the natural reactive barrier in the Meghna riverbank sediment accommodate the high As/Fe ratios observed? In the upstream area, where 4 of our sites are located and the groundwater As data are few (see Fig. 1), a survey by UNICEF found that 20 to 80% of the wells contain >50 μg/L As (12). An average sediment HCl-extractable As/Fe ratio of 89 ± 114 mmole/mole was determined for Meghna River sediments, compared to 0.7 ± 1.1 mmole/mole for the Waquiot Bay (see Table 3). In other words, ≈100 times more As was trapped by the reactive barrier along the Meghna riverbank on a per mole basis of Fe. This high As/Fe ratio, however, is still within the sorption capacity for arsenate and arsenite on a number of iron oxides, such as ferrihydrite and goethite (13, 14). Further study of the Meghna riverbank sediment is needed to determine the sorption capacity for arsenate and arsenite in these natural sediments.
Recycling of As.
Enrichment of As up to 758 mg/kg associated with ferric oxides in shallow subsurface sediments (1–2 m) were also reported for 8 other sites located within ≈80 km of each other (see Fig. 1), mostly east of the Meghna River (15). The As-enriched ferric oxides form by oxidation of Fe(II) and As(III) near the top of the capillary fringe, where exposure to oxidants enables bacterial oxidation and coprecipitation (15). We realize that further study is needed to determine the extent of As enrichment in near-surface sediment, but our results and previous studies suggest a systematic feature driven by seasonally active hydrological and biogeochemical interactions between discharging anoxic groundwater and more oxic river water or air. Recently, a unique yet unproven explanation of As release was proposed, invoking mobilization during recharge in near-surface sediment and followed by transport to depth (16). If near-surface sediment As enrichment is found to be widespread, then it represents part of an As cycle that is also responsible for elevated As in groundwater. In this scenario, reversal of flow from river to aquifer at the onset of the wet season between April and June (17), or reversal of flow from riverbank to aquifer because of irrigation pumping at the peak of the dry season between January and March (18), has the potential to remobilize As from the riverbank sediment.
Alternatively, sediments enriched with labile As may eventually act as an As source to the groundwater system. In one scenario, in the modern prograding GBMD, the enriched layers are likely to be buried and result in a sedimentary As hot spot in the aquifer. By placing the enriched zone into a strongly reducing environment deeper in the subsurface, As is supplied to groundwater. In a slightly different scenario, sediment As may be resuspended and redeposited, and will no longer be highly enriched after the homogenization but would remain in an aquifer as dispersed grains of potentially labile As. The time scale for sediment redistribution is likely to be determined by the tectonic and sedimentary processes that result in major river avulsions (1–3 hundred thousand years) (19) or local river migrations (tens to hundreds of years) (20). In these scenarios, arsenic is not discharged directly from the delta following initial release to groundwater.
Aquifer flushing over time was thought to have lowered sedimentary As concentrations of the Pleistocene deeper aquifer that today contains low groundwater As (21, 22). Flushing should also lower the sedimentary As inventory in the shallow Holocene aquifer. However, our results suggest that not all of the As is leaving the aquifer system with the water. In other words, the flux of As discharge cannot be calculated using discharge fluxes of water multiplied by groundwater As concentrations. A simple back-of-the-envelope calculation illustrates this point. For the last 1,000 years, assuming groundwater is discharging at a rate of 1 m per year with an average As concentration of 114 μg/L, the riverbank sediment As concentration can increase to 1,800 mg/kg if all of the As is removed in a reactive zone of 0.1-m thickness with 25% porosity and a grain density of 2.5 g/cm3. Previously it was thought that this As left the delta. Our results imply that this As remains in the sediment and can be a potential source for As in groundwater.
Is the amount of As trapped on the riverbank significant compared to the sediment As inventory in the shallow aquifer sediment? The shallow aquifer in Bangladesh is ≈30-m thick, with an area of ≈40,000 km2 and a sediment As concentration of ≈3 mg/kg (1, 22, 23). Therefore, the sediment As inventory is estimated to be 7 × 109 kg, assuming a 25% porosity and 2.5 g/cm3 grain density. The Meghna River discharges 500 m3/s in the dry season (1), most of which probably represents groundwater discharge, and therefore an annual discharge rate is estimated to be 250 m3/s, equivalent to 0.2 m per year. (This is lower than the recharge rate because groundwater also discharges via evapotranspiration.) Assuming all of the groundwater As (114 μg/L) (see Table 3) is immobilized, then 9 × 105 kg of As per year is trapped in the riverbank sediment. Over 10,000 years, this is equivalent to all of the shallow aquifer sediment As inventory.
Implications for Submarine Groundwater Discharge.
Submarine groundwater discharge has been shown to impact coastal (24) and perhaps global elemental budgets. A large flux of submarine groundwater discharge, mostly as recirculated saline water, was implied based on high fluxes of radium and barium during low river-discharge periods in the GBMD (25). Fluxes of strontium by groundwater discharge was determined to be substantial based on high groundwater strontium concentrations and high groundwater recharge rates (26), although the recharge rates used may contain substantial error because of irrigation-induced recharge. Nevertheless, the behavior of elements across the iron curtain during groundwater discharge must be carefully evaluated before their fluxes via submarine groundwater discharge to the ocean can be calculated. This is because even for relatively conservative element like strontium, nonconservative behavior was observed (27).
Materials and Methods
Sediment Collection and Preservation.
The location of each sediment sampling site was recorded with a handheld GPS. Surface sediment (0–5 cm) was usually dry and was collected using a spatula into polyethylene Ziploc bag. An auger was advanced into the subsurface until water-saturated gray sediment was intersected, which was usually at ≈1 m below surface but could be as deep as 3 m. One subsurface sediment core was then collected by hammering (AMS compact slide hammer) a soil probe with a seven-eighth inch outside diameter with an ≈30 cm long plastic core liner (AMS) into the hole. The exact depth for each sample varied among sites, but all were within 1 to 3 m. Usually this method retrieved a 25 to 30 cm wet core that was placed into a Mylar bag with O2 absorbers. All samples were stored at 4 °C upon returning to the laboratory. Riverbed sediment samples were collected similarly but from a boat.
On January 25, 2006, a similar coring method was used near RS-11 (see Fig. S1) to recover 4 continuous 30-cm cores from the ground surface to 1.2 m at RS-30, and 6 sections of 30-cm cores with bottom depths of 0.3 m, 0.6 m, 1.5 m, 3.0 m, 4.6 m, and 6.1 m at RS-33.
Methods for leaching and acid digestion of sediment samples for As analysis are described in the SI Analytical and Spectroscopic Methods, as well as the spectroscopic methods for sediment As speciation analysis and Fe mineralogy.
Acknowledgments
We thank George Breit of the United States Geological Survey and two anonymous reviewers, and W. Rahman and M. Rahman for their assistance with fieldwork. This work was supported by Grant P42ES10349 of the United States National Institute of Environmental Health Sciences/Superfund Basic Research Program (to Y.Z.). S.D. was a Columbia Earth Institute and a Mellon Foundation Post-Doctoral Fellow; B.M. was a Columbia Earth Institute Post-Doctoral Fellow; H.-B.J. was a City University of New York Graduate Center Science Fellow. This is Lamont-Doherty Earth Observatory contribution 7298.
Footnotes
- 1To whom correspondence may be addressed. E-mail: yzheng{at}ldeo.columbia.edu or yan.zheng{at}qc.cuny.edu
-
Author contributions: Y.Z. designed research; S.D., H.-B.J., M.A.H., M.S., K.M.A., and Y.Z. performed research; S.D., B.M., and H.-B.J. contributed new reagents/analytic tools; S.D., B.M., H.-B.J., and Y.Z. analyzed data; and S.D., B.M., and Y.Z. wrote the paper.
-
The authors declare no conflict of interest.
-
This article contains supporting information online at www.pnas.org/cgi/content/full/0908168106/DCSupplemental.
References
- ↵
- Kinniburgh DG,
- Smedley PL
- British Geological Survey, Bangladesh Department Public Health Engineering
- ↵
- ↵
- ↵
- ↵
- Feris KP,
- et al.
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- Ahmed MF,
- Ali MA,
- Adeel Z
- Chowdhury MAI,
- Ahmed MF,
- Ali MA
- ↵
- ↵
- Dixit S,
- Hering JG
- ↵
- Meng X,
- Bang S,
- Korfiatis GP
- ↵
- Wanty RB,
- Seal RRI
- Breit GN,
- et al.
- ↵
- Polizzotto ML,
- Harvey CF,
- Sutton SR,
- Fendorf S
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- Basu AR,
- Jacobsen SB,
- Poreda RJ,
- Dowling CB,
- Aggarwal PK
- ↵
Citation Manager Formats
Article Classifications
- Physical Sciences
- Geology