Over the past few years, contributors to the present study have taken a more empirical approach by trying to determine if there is a systematic relationship between As concentrations in the groundwater and a plausible reservoir of exchangeable As in aquifer particles (Swartz et al., 2004; Zheng et al, 2005). A key tool developed with this goal in mind has been the needle-sampler, a simple device that collects groundwater as well as sediment as a slurry from precisely the same depth interval (van Geen et al., 2004, 2006). Deploying the needle-sampler overcomes the problem when collecting groundwater from a well and aquifer sediment by coring nearby that the two intervals may be geochemically distinct because of the highly variable nature of fluvio-deltaic deposits. Using the needle-sampler in a broad range of settings in Bangladesh, van Geen et al. (2008) recently documented that a roughly linear relationship between dissolved As in groundwater and P-extractable As in the particulate phase held across 3 orders of magnitude in concentrations and interpreted this finding as an indication of adsorptive equilibrium. The importance of reductive dissolution as a pre-condition for As release to groundwater was acknowledged, as this relationship was shown to hold only for intervals that were sufficiently reducing.

The new evidence for adsorptive equilibrium may guide future studies of the specific mechanisms of As mobilization but, perhaps more significantly, provided the first quantitative estimate of the likely time scale over which an aquifer can be flushed of its initial As content by groundwater flow (BGS/DPHE, 2001). The distribution coefficient inferred from the data in Bangladesh (K_d = 4 mL/g) shows that, on average, the As content of groundwater accounts for ~3% of the combined pools in the dissolved and mobilizable particulate phase. Converted to a retardation factor, the estimated distribution coefficient indicates that As is transported through an aquifer ~40 times more slowly than groundwater because of adsorption. Using this information and a simple advection-diffusion model, van Geen et al. (2008) showed that shallow aquifers deposited during the late Holocene could be flushed of their initial As content in a few hundred to a few thousand years, depending on the velocity of groundwater flow (BGS/DPHE, 2001; McArthur et al., 2004; Ravenscroft et al., 2005; Klump et al., 2006). An argument was presented, based on a representative range of groundwater flow velocities, that the regional and local distribution of As in shallow aquifers of the Bengal Basin could therefore reflect different flushing histories rather than a difference in the initial As content of aquifer particles (Meharg et al., 2006).

It is worth emphasizing that a subset of intervals sampled in Bangladesh characterized by leachable Fe(II)/Fe ratios <0.5 in the particulate phase or dissolved Fe <0.2 mg/L in groundwater containing <10 μg/L dissolved As did not fit the relationship corresponding to a K_d = 4 mL/g. The outliers, characterized by a significant deficit in dissolved As compared to the prediction from the relationship and the level of P-extracted As, were tentatively attributed to the persistence of binding sites with a higher affinity for As under insufficiently reducing conditions, possibly a very small proportion of Fe(III) oxyhydroxides. Much less reducing conditions corresponding to aquifers containing orange-colored sands of likely Pleistocene age and particularly low leachable Fe(II)/Fe ratios were not considered in this analysis (BGS/DPHE, 2001; Horneman et al, 2004; Zheng et al., 2005).

A transect of 7 new needle-sampler profiles was collected in Vietnam in April 2006 from Van Phuc village, located in the northern portion of the Red River delta and 13 km south of Hanoi (Fig. 1, Berg et al., 2001, 2008). The Red River originates from the Yunnan Plateau in southern China, but sediment reaching the Hanoi area is derived from the entire catchment. There is a sharp contrast in As concentrations in shallow (<30 m) groundwater of the study area, with levels measured in the laboratory <10 μg/L in the northwestern portion of the village and concentrations up to 500 μg/L in the remaining portion to the south (see Fig. 6 in Berg et al., 2007; also Eiche et al., 2008). The discussion here is limited to grey sands of likely Holocene age and data are reported for 4 needle-sampler profiles extending to 30 m depth and spaced equally along the 0.5 km southern portion of the transect (Table 1).

As in the past, the small headspace remaining above the groundwater and sediment slurry in each needle-sampler tube was flushed with N₂ immediately after collection in Vietnam and Nepal. After allowing the sediment to settle for a few minutes, 5–10 mL of groundwater was gently pressure-filtered under N₂ through 0.4 μm syringe-filters and into acid-cleaned scintillation vials with PolySeal caps.

Several days before analysis for As, Fe, Mn, S, Na, Mg, Ca and K by High Resolution Inductively Coupled Plasma-Mass Spectrometry (HR ICP-MS) at LDEO, filtered groundwater was acidified to 1% Optima HCl in the laboratory (Cheng et al., 2004). The delay in acidification, which reduces the chance of contamination and eliminates the need for shipping strong acid, has been shown to entirely re-dissolve any precipitates that could have formed (van Geen et al., 2007).

Concentrations of As in filtered groundwater from grey aquifers of Van Phuc village in Vietnam range widely (200±130 μg/L; n=25), with several samples at one end of the spectrum containing ~10 μg/L and several samples approaching 500 μg/L at the other extreme. Concentrations of P-extractable As (2.3±1.5 mg/kg; n=24) and HCl-extractable As (2.5±2.3 mg/kg; n=23) are comparable and vary in concert at this location, but are not related in any systematic way to dissolved As (Fig. 2a, b). Concentrations of dissolved Fe in Van Phuc span more than two orders of magnitude (<0.1 to 40 mg/L), with most groundwater samples containing >100 μg/L As associated with >3 mg/L dissolved Fe (Fig. 2c). An advanced state of reduction in the particulate phase as well as groundwater anoxia are indicated by leachable Fe(II)/Fe ratios >0.7 and dissolved Mn>0.2 mg/L, respectively, in all sampled intervals (Figs 2c, e). Comparing dissolved S concentrations in groundwater with typical SO₄²⁻ levels in the Red River (~5 mg/L, Berg et al., 2008), a likely source of recharge for these shallow aquifers, suggests considerable SO₄²⁻ reduction has taken place in at least half the groundwater samples from Van Phuc (Fig. 2e).

It is not clear why concentrations of As in groundwater are roughly proportional to the P-extractable As content of aquifer particles across Bangladesh but not at the more recently sampled locations. One notable feature of the combined set of observations is that the relationship established in Bangladesh appears to place an upper bound on the concentration of dissolved As expected for a given level of P-extractable As (Fig. 2b). One possible explanation is that the 1 M phosphate extraction, which constitutes merely an operational definition of the pool of exchangeable As, may be too aggressive outside Bangladesh. Laboratory experiments have shown that even As adsorbed to Fe(III) oxyhydroxides is remobilized in 1 M phosphate because of the strong affinity of P(V) for the same sites (Jung and Zheng, 2006). Thoral et al. (2005) also showed that the structure of Fe(II) oxyhydroxides is limited to very small clusters in the presence of As(III) during Fe oxidation experiments. This suggests that extraction in 1 M phosphate could also release As bound deeply within the natural Fe oxyhydroxides coatings of aquifer particles. It can only be speculated that a milder extraction procedure might be more appropriate for quantifying readily desorbable As and could perhaps have produced a more consistent relationship across a wider range of settings.

An alternative explanation is that dissolved As levels are not controlled by adsorptive exchange with an identifiable pool in the particulate phase, even when considering exclusively Fe and SO₄ reducing environments. Differences in mineralogy and/or affinity of surface sites for As are another possibility, but it is not obvious why the mineralogy of sediment reaching different regions of Bangladesh should be any less heterogeneous than sediment deposited in India, Vietnam and Nepal.

Some of the differences in the relative efficiency of HCl and P-extractions documented here are consistent with previous observations and may therefore reflect regional trends worth exploring further. BGS/DPHE (2001) reported that oxalate extractions of aquifer sands, which can be considered to target a fraction equivalent to that mobilized by the 1 M phosphate extractions, release only 10–50% of As in the sediment measured by total digestion. Swartz et al. (2004) and Zheng et al. (2005) also noted that 1 M phosphate extractions release <30% of total As present in shallow grey sediment at several locations in Bangladesh. Within the suite of samples collected with the needle-sampler in India and Bangladesh, P-extracted As concentrations in sufficiently reducing intervals are on average 0.4–0.5 times lower than HCl-extractable As levels, respectively (Table 2). It therefore appears that 1 M phosphate extractions systematically release a modest fraction of HCl-extractable As present in aquifer solids from Bangladesh – and this phase appears to be in adsorptive equilibrium with As present in groundwater. Berg et al. (2008), in contrast, report that a similar P-based extraction releases >50% of As in the sediment at several locations in Vietnam. Phosphate extractions also leaves little residual As in the solid phase extractable in HCl for aquifer solids collected with the needle-sampler in Vietnam, as suggested by the comparable P- and HCl-extractable As levels (Table 2). The average P- to HCl- extractable As ratio for aquifer solids from Nepal is similar to the one measured in Vietnam, but more variable and less well constrained. For reasons that are presently not understood, P-extractable As concentrations are significantly lower than HCl-extractable As levels in two regions that are most affected by elevated As groundwater levels, Bangladesh and West-Bengal, India, whereas the two extractions release comparable amounts of As from the solid phase in two regions where the problem appears to be less widespread. It is speculated that the underlying cause may be related to the extent of transformation of As-containing phases in aquifer particles. Observations from Vietnam and Nepal may indicate a relatively mature system, whereas the significant fraction of As contained in the solid phase in the Bangladesh and West Bengal that is not mobilized in 1 M phosphate may indicate incomplete transformation and, consequently, further delay in flushing. One practical implication is that a milder form of extraction, possibly the less aggressive extraction carried out in 0.05 M phosphate by Postma et al. (2007), might be more appropriate for identifying the exchangeable fraction of As in aquifer material across the entire affected area.

Despite the scatter, the new data can be used to place some constraints on the rate at which shallow aquifers are likely to be flushed of their mobilizable As content, assuming it is relatively uniform initially. The simple one-dimensional advection-diffusion model recently presented by van Geen et al. (2008) is used with one modification: the distribution coefficient K_d is varied between 4 and 40 mL/g to accommodate the new observations (Fig. 2a). Assuming that dissolved and particulate-phase As concentrations are initially equilibrated along the entire flow path at a concentration C_i and that recharge water entering the aquifer at one end of the flow path contains no As, the solution of the governing equation is: where C is the concentration of As in the dissolved phase, x the distance along the flow path, v the advection velocity, D is a hydraulic dispersion coefficient, and R the retardation factor, which is related to the distribution coefficient K_d by the expression R = 1 + ρ K_d/θ where ρ is the aquifer bulk density, K_d is the ratio of the As concentration in the particulate phase on per mass of sediment divided by the As concentration in the dissolved phase, and θ is the aquifer porosity (van Genuchten and Alves, 1982).

The focus is on the sensitivity of the model between distances of 50–500 m, i.e. within the range of length of likely flow paths that lead from an area of recharge to the depth range of shallow aquifers and eventually to a discharge area, typically a local stream (Stute et al., 2007; Aziz et al., 2008). Only the central scenario based on a groundwater flow velocity 5 cm/d (5.8 × 10⁻⁷ m/s) and a dispersion coefficient D = 5.8 × 10⁻⁶ m²/s are considered here. Justification for these parameters and model sensitivity to these choices were discussed previously by van Geen et al. (2008). For K_d of 4 and 40 mL/g, retardation factors R = 36.8 and 359, respectively, are calculated from an aquifer bulk density of 2.24 g/cm³, derived in turn from a porosity of 0.25 and a particle density of 2.65 g/cm³.

The model assumes a uniform initial P-extractable As concentration of 4 mg/kg, which corresponds to dissolved As concentrations of 1000 and 100 μg/L for K_d = 4 and 40 mL/g, respectively. Under conditions of greatest As mobility, representative of shallow aquifers in Bangladesh, the model shows that a considerable portion of the 50–500 m range of distances along the flow path is already flushed by 0.5 ka and entirely so by 5 ka (Fig. 3). Whereas As concentrations start at a lower level under the alternative scenario associated with a 10-fold higher K_d, flushing of the aquifer is also considerably slower. The model predicts elevated As concentrations across the 50–500 distance range, and significant but incomplete flushing of the same shallow aquifers after 5 ka. The scatter in the needle-sampler data suggests that the reality under natural conditions is probably somewhere in between these two extreme scenarios, assuming all sampled deposits are of Holocene age (this has been confirmed in Vietnam and Nepal on the basis of optically stimulated luminescence dating, B. Weinman and A. Singhvi, pers. comm.). The only upside to the fact that As appears to be particularly mobile in groundwater of Bangladesh is that shallow aquifers are likely to be flushed more rapidly in Bangladesh compared to India, Vietnam and Nepal. This is, however, not relevant in terms of human exposure since, hopefully, much of the population in affected regions will no longer consume untreated groundwater within the next few decades. Flushing would be delayed further if, in addition, the fraction of As in the solid phase released in hot HCl is gradually converted to exchangeable As.