2.1. Geological Setting Different regions of Bangladesh are not affected to the same extent by elevated As concentrations in groundwater. Broad patterns in the proportion of low- and high-As wells across distances of 10 to 100 km are related largely to the depth of older Pleistocene sediments. Groundwater As concentrations associated with these orange-brown sandy deposits, frequently referred to as the Dupi Tila formation, are typically below 10 g/L (BGS and DPHE, 2001). In areas where this formation is located no more than 30 m (100 ft) below the surface, the vast majority of existing tube wells are low in As. In regions where these deposits are overlain with a thick layer of more recent Holocene sediment, often gray in color, the groundwater pumped from existing tube wells less than 30 m (100 ft) deep is frequently elevated in As (BGS and DPHE, 2001). The samples collected for the incubations of the present study were collected from a transitional area where the depth of the upper portion of the Dupi Tila formation rapidly changes over a few kilometers (van Geen et al., 2003a). Figure 1 is a map of a portion of Araihazar upazila, where a long-term health, social, and earth science study of the origin and impact of elevated groundwater arsenic, coupled to a mitigation program, is underway (http:// superfund.ciesin.columbia.edu). Also shown is a depth section of As concentrations based on the same set of wells that illustrates the two-layered structure of the subsurface in the area. The boundary between a mixed population of shallow wells that are either high or low in As and a deeper aquifer that is typically low in As dips downwards from 30 m to 90 m (100 to 300 ft) in eastward direction. The combination of the map and depth section also shows that over distances of 0.1 to 1 km, i.e., from one village to the other, and even within a village, there are considerable variations in groundwater As concentrations within shallow Holocene aquifers.
2.2. Collection of Sediment and Groundwater Four samples of aquifer material used in the incubations were collected during drilling of community wells CW1 and CW6 in two villages of Araihazar Upazila in June 2001 and January 2002, respectively. A rhythmic drilling technique used throughout Bangladesh and called the “hand-flapper” or “sludger” method generates a stream of sediment-laden slurry originating from a depth known from the length of PVC pipe that extends into the drill hole. The sand and silt particles that rapidly settled from this slurry constituted the starting material for 3 of the 4 series of incubations. These samples may be biased towards the coarse fraction of the sediment, although the aquifers that groundwater can be extracted from are typically well-sorted and contain 90% sand (BGS and DPHE, 2001; Zheng et al., in review). The grain size of one additional sediment sample is representative of the original sandy aquifer material since it was collected with an improvised PVC coring device equipped with a check valve at the top to prevent the loss of sediment. For all incubations, groundwater used to dilute the sediment was hand-pumped from a well in the region (ID 4135) that taps into the Dupi Tila aquifer at 40 m (135 ft) and contains little dissolved As and Fe compared to the range of groundwater concentrations in the study area (Table 1). Additional characteristics of groundwater from well 4135 determined by methods described in Zheng et al. (2004) include a barely detectable dissolved oxygen content of 0.02 mg/L (CHEMet detection limit 0.01 mg/L); Eh 25 mV; pH 6.29; and Alk 2.74 0.03 meq/L. The first sediment sample collected from a slurry at 4.5 m (15 ft) depth at CW6, in the eastern portion of the study area, was a mixture of gray and orange-coated sands containing Fe(II) and Fe(III) oxyhydroxides, as indicated by ferrozine analysis of the acid leachate of similar material (Horneman et al., this volume). This relatively oxic aquifer material is referred to hereon as CW6-G&O and is considered representative of the early stage of sediment transformation towards a more reduced state. Another slurry sample from a slightly greater depth was uniformly gray (CW6-G), indicating a predominance of Fe(II) in the solid phase, and was evidently further along the path towards reducing conditions (Horneman et al., this volume). The third sample of shallow aquifer material, from the western portion of the study area, was collected at 4.5 m (15 ft) with the coring device and was also uniformly gray (CW1-G). The last slurry sample collected from CW1 at a depth of 34 m (115 ft) was uniformly orange/brown (CW1- O), indicating a particularly high proportion of Fe(III) oxyhydroxides in the sand coatings (Horneman et al., this volume). Radiocarbon dating of sediment in the region indicates that the three shallow samples of aquifer material sediment used in the incubations were deposited within the past several thousand years (Zheng et al., in review). In contrast, the deeper sands of CW1-O were deposited at least 40000 yr ago and probably were exposed to extensive weathering under oxic conditions during a previous Sealevel low-stand (Zheng et al., in review).
2.3. Preparation of the Incubations Slurry samples CW6-G&O and CW6-G were diluted on the day of collection and placed inside a nitrogen-filled glove bag with anoxic groundwater in a series of six 1-L Nalgene high-density polyethylene bottles. Sodium acetate (1 g/L) and an antibiotic (40 mL of Guillard reagent/L), respectively, were also immediately added to suspensions of the two aquifer samples from CW6. The Guillard reagent had been prepared by adding 16250 U of penicillin G, 2 mg of chloramphenicol, and 5 mg of streptomycin to 100 mL of MQ water. During the first week of the incubation, the suspensions were kept in a nitrogen-filled glove bag in Bangladesh. They were then shipped by air to France without special precautions. After each sampling, the headspace of each bottle was flushed with nitrogen before closing. The bottles themselves stood in ambient air throughout this first set of experiments, however. Samples CW1-G and CW1-O, a core sample, and a slurry sample, respectively, were collected in plastic bottles that were then filled with groundwater from well 4135 until no headspace was left. The samples, in addition with more water from well 4135, were then sent by air to France. The second set of incubations sediment was started by diluting aquifer material sampled from the center of a 2 kg mass of sediment with groundwater in 1-L Nalgene bottles under a 98% N2/2% H2 atmosphere 14 d after collection in Bangladesh. The suspensions were again amended in the same manner with sodium acetate and Guillard reagent, but the latter was unfortunately prepared with expired reagents. These suspensions remained in the glove chamber for the continuation of the experiment. The glove chamber was flushed several times with 98% N2/2% H2 after each sampling. The laboratory temperature varied between 20 to 25o C during the incubations, slightly below the range of groundwater temperatures in Araihazar (25–27o C).
2.4. Sampling and Analyses Although the suspensions were not agitated continuously during the incubations, they were vigorously shaken and then allowed to settle for 30 min each time before sampling. Over a period of two months, aliquots of the supernatant were taken from the incubations with a syringe at intervals that increased in duration from hours in Bangladesh to weeks in France. In ambient air, the aliquots were filtered immediately out of the same syringe though 0.45 um Nuclepore syringe filters. Typically, 3 filters had to be used to filter an entire 5 mL sample. For the second batch of experiments, an aliquot of filtered sample was also immediately passed through a small disposable ion-exchange column that retains As(V) but not As(III) marketed by Metalsoft Center, Highland Park, New Jersey (Meng et al., 1998). After filtration or passage through the ion exchange column, all samples were acidified to 1% by volume with Optima HCl. In ambient air in the case of the first set of incubations and inside the glove chamber for the second set, pH, Eh, and oxygen saturation were measured in the suspensions using a Bio block Scientific glass electrode and Consort P107 pH meter, a Schott platinum electrode, and a YSI Model 55 oxygen sensor. These electrodes were calibrated with appropriate solutions before each use. The filtered and acidified samples were analyzed for dissolved As, Fe, Mn, P, and S at LDEO by High-Resolution Inductively-Coupled Plasma Mass Spectrometry (HR ICP-MS) on a VG Axiom instrument. The samples were diluted 1:5 in a 2% Optima HNO3 solution containing 10 g/L Ge added to monitor any drift in the sensitivity of the instrument. The instrument was calibrated with mixed As, Fe, Mn, P, and S standards added to a 2% HNO3/Ge solution. Standard additions performed on a limited number of samples showed no detectable dependence of instrument response on the nature of the matrix. The method predates but is very similar to the multielement method described by Cheng et al. (in press). Detection limits were calculated from the variability of blank counts for repeated analysis of HNO3/Ge solutions: 0.1 g/L for As and below 0.5 mol/L for Fe, Mn, P, and S. Replicate analyses and the inclusion of a consistency standard with each run indicate an overall reproducibility of the method on the order of 5% for all analytes. The quantity of As, Fe, Mn, P, and S that is moderately bound to the surface of the sediment used in the incubations was estimated with a hot HCl leach. The extractions were performed by adding 50 mg of freeze-dried sediment in 1.5 mL centrifuge tubes, adding 1 mL of 1.2 N HCl, and heating in hot water bath at 80o C for 30 min (Horneman et al., this volume). After leaching, the suspensions were centrifuged at 10000 rpm and the supernatant was transferred to a new set of acid-leached centrifuge tubes. The acid leachates were diluted in a 2% HNO3/Ge solution and also analyzed by HR ICP-MS. This leaching procedure is likely to have released As, Fe, and P adsorbed on the sediment and the fraction of the same elements associated with poorly crystallized Fe oxyhydroxides (Frederickson et al., 1998; Roden and Urrutia, 2002; Zachara et al., 2002). A remarkable correspondence between the reflectance spectrum of Bangladesh sediment and the Fe(II)/(III) ratio of the hot-leachable fraction documented in the companion paper by Horneman et al. (this volume) suggests that, probably because of the heating step, the leaching procedure used in this study dissolved some crystalline phases such as goethite or hematite.