Clay-plug sediment as the locus of arsenic pollution in Holocene alluvial-plain aquifers

arsenic in an anoxic environment, and subsequent migration of the desorbed arsenic to, and stratigraphic entrapment in, adjacent sandy point-bar aquifers. To assess the magnitude of the arsenic source in clay-plug, bulk sediment volume calculations were made of twenty clay plugs on the Middle Ganges Plain of Bihar (India), by combining clay-plug surface area analysis of Sentinel-2 satellite data, side-scan sonar depth profiling of oxbow lakes and the Ganges River, and sedimentological data from five cored shallow wells. ICP-MS based elemental analysis of 36 core sub-samples, complemented with published concentration data in a similar geomorphological setting in West Bengal, India, yielded an average arsenic content of 28.75 mg/kg sediment in the 12-m-thick clay plugs, which amounts to a total arsenic volume of 0.07 – 3.13 . 10 6 kg per clay plug. A scenario is presented for the release of arsenic from the clay-plug sediment by microbial metabolism, followed by migration of the des- orbed arsenic to the bordering point-bar sands.

The source of arsenic enrichment in shallow Holocene aquifers is linked to infiltration of dissolved arsenic from river water to aquifer in the depositional basin (Postma et al., 2016;Trung et al., 2020), and to the in-situ microbially-steered reductive dissolution of arsenic (Ghosh and Bhadury, 2018;Ghosh et al., 2021) from, e.g., biotite (Seddique et al., 2008); iron-oxyhydroxides (Nickson et al., 1998), arsenic-bearing pyrites and arsenopyrites (Acharyya et al., 2000(Acharyya et al., , 2007Shah, 2017). Mukherjee et al. (2019) presented a comprehensive overview in which the worldwide occurrence of arsenic pollution in the shallow-aquifer domain of Holocene foreland basins is linked to primary arsenic sources in mountain belts of convergent continental margins in Alpine and older orogenic systems. Documented primary natural arsenic sources are, among others, volcanic rocks, salts and thermal waters (Tapia et al., 2019), arsenic-enriched magmatic rocks (Mukherjee et al., 2014), brecciated marbles (Göd and Zemann, 2000), hydrothermal arsenic enrichment of calcite veins in altered fault rocks (Horton et al., 2001;Campbell et al., 2004). Upon weathering of the orogenic rocks, the arsenic that is chemically-bound to rock fragments is river-transported to the sedimentary sink, where it is deposited as part of the siliciclastic alluvial-basin sediment.
The geomorphological juxtaposition of (a) compacted, low-porosity and low-permeability, organic matter-rich clay silt and clay deposits in abandoned and sediment-filled meandering-river bends (or: clay-plugs), and (b) high-porosity, high-permeability sandy point-bar deposits (see: Graphical Abstract) forms a dynamic combination in which microbiallysteered reductive desorption of arsenic takes place in the anoxic clayplug environment . Biogeochemical factors such as redox conditions, pH, presence of different ions and metabolic activity of indigenous microbial communities play a crucial role in the mobilization of arsenic (Islam et al., 2004;Harvey et al., 2006;Bhattacharya et al., 2007;Ghosh et al., 2015aGhosh et al., , 2021. The desorbed arsenic subsequently migrates along the porosity-permeability gradient into the adjacent sandy point-bar aquifer (Donselaar et al., 2017;Ghosh et al., 2021), where it is stratigraphically entrapped by the permeability contrast with the surrounding fine-grained alluvial-plain and clay-plug sediment (Donselaar and Overeem, 2008;Weinman et al, 2008;Sahu and Saha, 2015;Donselaar et al., 2017;Jakobsen et al., 2018).
To get to grips with the magnitude of the pollution problem, the present paper aims to quantify the total arsenic volume contained in clay-plug sediment, and that can potentially be desorbed in the anoxic clay-plug environment, and migrate to the point-bar aquifer. To reach the aim, the arsenic content of sediment samples was analysed from five shallow (15 to 50 m depth), fully-cored wells, drilled in Holocene clayplug and point-bar sediment along the Ganges River in Bihar, India. The quantitative data set was complemented with published data (Ghosh et al., 2015b) on arsenic concentrations in clay-plug sediment in the Nadia District in West Bengal (India). Next, the concentration values were assigned to the clay-plug bulk sediment volumes as calculated from Sentinel-2 satellite surface area data in combination with side-scan sonar depth profiling of oxbow lakes and the Ganges River, and sedimentological data from the cored wells. The results of the study provide a quantitative data set on potential arsenic availability in the shallow aquifer domain, which is vital to envisage the magnitude of the pollution problem, and to develop mitigation strategy models accordingly.

Geomorphological setting
The study area is in the alluvial plain of the Ganges River in the Middle Ganges Plain (MGP) in Uttar Pradesh and Bihar (India) (Fig. 1A-B), part of the larger Indo-Gangetic foreland basin between Peninsular India in the south and the Himalayas in the north. The alluvial landscape in the MGP is shaped by Holocene floodplain and meandering-river deposits of the Ganges River and its tributaries. The process of chutechannel shortcutting (Bridge, 2003;Nichols, 2009) has detached meander loops from the active river and converted these crescentshaped oxbow lakes ( Fig. 1C-D), where hydrophytes and animal life thrive in the standing bodies of lake water, and contribute to the high rates of organic matter deposition into the lake sediment (Lawson et al., 2013;Ghosh and Biswas, 2015;Ghosh et al., 2021). Oxbow lakes eventually fill-in with sediment by aggradation of a mixture of organic debris and clay and silt that settles out of suspension from the lake water column and, upon completion of the infill process, convert to so-called clay plugs (Fig. 1D). The clay and silt easily compact, and are characterized by very-low porosity and permeability. In the active river period, prior to meander-loop cut-off, sand is accumulated by bed-load sediment transport in point bars on the inner river bends (Bridge, 2003). The sandy point-bar deposits are characterized by high porosity and permeability values (Hartkamp-Bakker and Donselaar, 1993), that strongly contrast with the surrounding clay-plug and floodplain fines, and result in differential compaction where the point-bars stand out as a positive relief in the flat and clay-dominated alluvial-plain landscape. Human settlements are concentrated on the point bars because the elevated position is a protection against seasonal monsoonal river flooding. In the study area, piped water-supply grids are absent, and groundwater for consumption and irrigation is directly extracted from the point-bar aquifer by hand-pump wells of approximately 30 m depth (Donselaar et al., 2017).

Data and methods
Remote sensing and ground-truth geomorphological and sedimentological data sets were collected along the alluvial plain of the Ganges River ( Fig. 1; Table 1). Sedimentological descriptions from five shallow (15 to 50 m depth), fully-cored wells provided details of the point-bar and clay-plug lithofacies succession (Fig. 1C,D). In total 36 sediment samples were recovered from the cores for further analyses. Samples were stored in individual zip-lock polyethylene bags for analyses of total organic carbon (TOC), elemental extraction and quantification with inductively coupled mass spectrometer (ICP-MS). The methodology of Bojko and Kabala (2014) was followed for TOC analysis, and the Swedish Standard Method (SIS, 1993) for the elemental extraction. For the latter, sample digestion was carried out by treating about 0.5 g dried sediment sample with 10 ml 7 M HNO 3 at 100 kPa and 121 • C for 30 min. The solution was cooled down and centrifuged at 14,000 RCF for 15 min. The CRM-601 standard was used as a certified reference sediment to calculate the errors. The extracts were diluted before analysing the elemental levels in an inductively coupled mass spectrometer (ICP-MS-Thermo Scientific X-Series II).
As a first step for the estimation of potential arsenic volume in clay plug sediments, the bulk volume of clay plugs was calculated ( Fig. 2 and Graphical Abstract) by multiplying (1) the surface areas of twenty clay plugs, extracted from Sentinel-2 satellite images ( Fig. 1A) with (2) crosssectional shape models from four side-scan sonar surveys across oxbow lakes and the Ganges River (Fig. 1B, 3). Next, sediment proportion ratios of the three sediment classes (clay, silt and sand) from own core data and from three complementary cored well data sets (Ghosh et al., 2015b were used to assign (3) bulk sediment volumes to the clay plugs. The ICP-MS analysis results gave the solid-state arsenic content in each of the three sediment classes, and this allowed to (4) calculate the potential arsenic volume in each of the clay plugs.

Clay-plug surface area
Surface areas of twenty clay plugs were extracted from Sentinel-2 satellite raster images with ArcGIS. Sentinel-2 multispectral band data ranging from 0.49 μm to 1.00 μm (visible light to Near Infrared) with 10 m spatial resolution were used for self-iterated unsupervised classification in ArcGIS. The output was rectified with a 3x3 edge-detection filter. The unsupervised raster images were converted to vector data in ArcGIS for automated calculation of the surface areas. Data consistency was cross-verified with Google Earth-Pro images.

Clay-plug cross-sectional area calculations
Four side-scan sonar surveys across the oxbow lakes and Ganges River served to generate the cross-sectional depth profile model for clay plugs. The asymmetric profiles (Fig. 3A-D) with a steep outer bank (or: cut bank) near the channel thalweg, and a gently sloping inner bank (or: point bar), were converted in MATLAB to a parabolic synthetic cross section (Fig. 3E). To assign the clay-plug thickness to the cross-section, the total height difference between channel thalweg and top of the point bar (or: bank-full depth) was measured. The largest measured present-  day channel depth measured with side-scan sonar is 9.5 m in profile SS-4 (Fig. 2B, 3D). A Jacob's staff survey yielded a height difference of 2.5 m from the waterline to the top of the point bar at the location of SS-4 (Fig. 2B), which gives the total height of 12 m from channel thalweg to point-bar top. The height was corroborated by the clay-plug thickness in wells BKP-3 and BKP-4 (Fig. 1D, 4). Similar thicknesses were described by Ghosh et al. (2015b) and Ghosh et al., (2017) in wells through clay plugs on the alluvial plain of the Ganges River in West Bengal (India). For the cross-sectional area calculations, the width of the synthetic cross-sectional profile was scaled to the measured width in ArcGIS of each individual clay-plug; the depth was fixed at 12 m, assuming that the scouring depth of the Ganges River is constant in the along-river equilibrium profile in the low-gradient part of the MGP. The synthetic cross profiles, in combination with the clay-plug surface radius were used to calculate the contact area between the clay plug and the point bar.

Bulk sediment weight
As a next step, the calculated clay plug bulk volume was converted to bulk sediment weight. The sedimentological logs of the two wells penetrating the clay plugs (BKP-3 and BKP-4; Fig. 1D, 4) provided the ratio of the different sediment types in the clay plugs. The sediments were deposited in the standing body of oxbow lake water, and therefore it is assumed that the sediments are fully water-saturated and the pore spaces are filled with water (Bear, 1972). Standard specific weight values for sand, silt and clay (Hillel, 1980;Linsley et al., 1982;Yu et al., 1993), and average effective porosity values for unconsolidated sediment (McWhorter and Sunada, 1977) were assigned for the calculations (Table 2). Porosity values for coarse, medium and fine sand were grouped, because of their close similarity in porosity and permeability (Freeze and Cherry, 1979). The initial porosity, i.e., before compaction, of clay-rich sediments ranges from 40% to 95% (Holbrook, 2001). For clay-rich sediments in oxbow lakes, initial porosities range from 70% to 90% (Singer and Müller, 1983). The initial porosity of unconsolidated sand varies from 35% to 50% (Stone and Siever, 1996). As the sediments in the cores showed compaction, the average values of representative porosity for compacted sand, clay and silt as described by Yu et al. (1993) were used in the present calculations.
The bulk density ρ bulk is calculated as: in which φ is the porosity, and clay%, silt% and sand% the specificweight values.
The total weight of a clay plug (W) is calculated as:

Arsenic weight per clay plug
Elemental analysis (ICP-MS based) of four core samples from the clay-plug intervals in wells BKP-4 and SHY-1 (Fig. 1C-D) yielded the arsenic weight fraction in the sediments (Table 3; Fig. 5). The average arsenic content from the samples in the upper clay plug interval (12 m in wells BKP-4, and 8 m in SHY-1; Fig. 4) was calculated, and this value was used to calculate the total potential arsenic weight per clay plug: Arsenic weight = clay − plug weight * Average As concentration (3)

Results
The twenty clay plugs in the study area have a wide variety in surface area, ranging from 0.42 ⋅10 6 m 2 to 18.78 ⋅ 10 6 m 2 (Table 4) and, consequently, the total sediment volume of the individual clay plugs varies between 2.44 ⋅ 10 6 m 3 and 113.01 ⋅ 10 6 m 3 . The total solid-state arsenic weight contained in the clay plug sediment is calculated from the average content of wells BKP-4 and SHY-1 clay-plug samples of 28.75 mg/kg, and amounts to 0.07 ⋅ 10 6 kg to 3.13 ⋅ 10 6 kg per clay plug ( Table 4). The arsenic and iron content in the wells show a decrease with depth (Fig. 5), with a correlation coefficient between both elements of more than 0.9 in wells BKP-2 and SHY-1. This is consistent with earlier studies (Ghosh et al., 2015b), which suggest that the geogenic arsenic in this region is mostly present in the form of arsenic-bearing iron minerals such as pyrite, Fe-(hydro)oxides (Mukherjee et al., 2011). The lower correlation coefficient of 0.75 in well BKP-4 is cause by a spike in arsenic and iron content in highly porous and permeable unconsolidated conglomerate at 30.7-30.8 m, just below the sequence boundary with overlying low-permeable, fine-grained sand (Donselaar et al., 2017, Kumar et al., 2019. A similar trend of decrease with depth is observed in the TOC content in wells BKP-2, BKP-4 and SHY-1, with again a spike below the sequence boundary in BKP-4 (Fig. 6).
The measured arsenic concentrations in the clay-plug samples ranges from 19.8 to 38.9 mg/kg and is in the range of published data of arsenic concentrations in sediment cores. Bundschuh et al. (2004) Berg et al. (2007) documented arsenic concentration ranges from 2 to 33 mg/kg in Holocene sediments of the Red River Delta, Cambodia and Vietnam; in two of the three published wells the arsenic concentrations sharply decrease below approximately 15-20 m depth. Ghosh et al. (2015b) documented arsenic concentrations in 12-m-thick Holocene clay-plug sediment of Nadia District in West Bengal (India) that range from 2 to 21 mg/kg. Seddique et al. (2008) reported arsenic concentrations of 0 to 12 mg/kg (with an outlier of 53.4 mg/kg) in Holocene Meghna River sediment.

Discussion
Spot measurements in Holocene aquifers highlight the heterogeneity of arsenic pollution concentrations, both on the surface, as with depth (e.g., Bundschuh et al., 2004;Zheng et al., 2005;Acharyya and Shah, 2007;Berg et al., 2007;Seddique et al., 2008;Shah, 2008Shah, , 2010Shah, , 2017Ghosh et al., 2015b;Sahu and Saha, 2015;Donselaar et al., 2017, Jakobsen et al., 2018. In this paper the concentrations have been extrapolated to total arsenic load contained in a single clay-plug morphological element. The total load is considered here as static, and no recharge by introduction of additional arsenic into the clay plug is taken into consideration. The remaining questions are: (a) whether the bulk of the arsenic that is adsorbed to sediment grains in the clay plug, has potential of release and migration to the aquifer waters, and (b) the nature of migration of the released arsenic from low-porosity, low permeability clay-plug sediment to the aquifer domain in the adjacent porous and permeable point-bar sand.

Arsenic release
Arsenic release is driven by microbial reduction and mobilization of arsenic associated with iron-oxy(hydroxide) minerals (Ghosh et al., 2015a. Microbial metabolism is controlled by the availability of organic carbon (OC). The sandy alluvial sediments are generally lean in OC. By contrast, clay-plug sediment in the study area has high TOC concentrations of 0.7-0.8% in wells BKP-4 and SHY-1 (Fig. 6). Ghosh et al. (2015a) reported similar TOC concentrations (of 0.7%) in clayplug deposits at depths to12 m in Haringhata (Nadia district, West Bengal, India). The presence of high macrophytes growth coverage in the oxbow lake (Ghosh and Biswas, 2015;Ghosh et al., 2021) blocks the penetration of light, and prevents water circulation in the lower part of the water column in the oxbow lake, and contributes to the development of an anaerobic environment in the hypolimnion favourable for the desorption of As from FeOOH (e.g., Bhattacharya et al., 2002). Decayed oxbow-lake macrophytes are the source of TOC, and are considered vital for the metabolism process . The OC associated with the sediment provides nutrition to the microorganisms (bacteria) in the system (Ghosh et al., 2015a). During the heterotrophic metabolism in anaerobic or facultative aerobic environment, the organic molecules act as electron donors and this subsequently reduces the arsenate [As(V)] which enters into the microbial cell via Pit transporters and converted to arsenate [As(III)] by arsenate reductase enzymes within the cell, and then released into the environment (Fig. 7). The gradual decrease of OC coupled with decreased arsenic and iron concentrations in the sediment with depth below the clay plugs (Fig. 6) testifies to the effectiveness of the arsenate reduction process.

Diffusion process in the clay plug
Under the assumption that microbially-steered reductive dissolution of arsenic is effective in the anoxic clay-plug environment, diffusion is the principal process of arsenic expulsion in the gravitational, compaction-driven environment, and advection transport driven by concentration differences takes over in the adjacent point-bar sand past the permeability boundary at the clay-plug to point-bar interface. Because of the permeability anisotropy between the very lowpermeable, compacted clay sediment and the porous and permeable point-bar sands, discharge of dissolved arsenic will be directed towards  the clay-plugpoint-bar interface by the process of diffusion. The hydraulic conductivity is linearly related to the diffusion flux. The arsenic released in the clay plug by metabolic arsenate reduction leads to accumulation with respect to the adjacent point-bar sand. At the interface of the compacted, low-permeable clay and the highly permeable point-bar sand, arsenic transport occurs through molecular diffusion which levels the concentrations of arsenic by redistributing elements with higher concentration to elements with lower concentration (Bear, 1972;Kumar et al., 2019). Once the reduced arsenic molecules enter the permeable point-bar sand, advection is the process through which transport of arsenic occurs within the aquifer system along the pressure gradient. The latter is amplified by water extraction through the handpump wells.
To get to grips with the potential of dissolved arsenic diffusion volume along the permeability gradient and across the permeability interface, the contact area between the clay plug and bordering pointbar sand was calculated in MATLAB. It is assumed that the arsenic in the clay plug is not replenished, and that the amount and discharge will decrease exponentially over time. For the estimate of the diffusion flux, Fick's first law of diffusion is used (Eq. (4)) with two assumptions, firstly that the system is in a steady-state, and secondly that the medium is homogeneous.
D is the diffusion coefficient, a material-specific property, and C is the concentration. With some modifications, Fick's first law for transport through a membrane is given by Eq. (5): where J is diffusion flux [mol m − 2 s − 1 ], K the hydraulic conductivity of the medium [m s − 1 ] and ΔC the difference in concentration [mol m − 3 ]. In the hydraulic conductivity, crucial to the modified version of Fick's first law, the following range was used: a minimal hydraulic conductivity of 10 − 8 ms − 1 and maximal value of 10 − 7 ms − 1 (Bear, 1972). The results of the calculations are thatwith contact area between the clay plug and the adjacent point bar varying between 0.41 ⋅ 10 6 m 2 and 21.80 ⋅ 10 6 m 2 -the potential dissolved arsenic load that could pass through the contact area amounts to 0.136 -0.175 kg/m 2 . The initial diffusion flux through the contact area, based on the application of Fick's First Law, ranges between 10 1 and 10 2 g/(m 2 y − 1 ).

Point-bar advective flow along the pressure gradient
The lithofacies distribution of point-bar deposits is characterized by the presence of (a) high-permeable clean sand lag deposits at the base (which in hydrocarbon enhanced recovery scenarios are referred to as so-called 'thief zones', (b) fining-upward grain-size sequences towards the top of the point-bar sediment body, that form as response to the energy anisotropy of the helicoidal flow in the meandering-river bend, and (c) inclined clayey lateral accretion surfaces in the upper two-thirds of the point-bar morphology that form in peak runoff periods when the river water expands over the adjacent alluvial plain (see the Graphical Abstract; Hartkamp-Bakker and Donselaar, 1993;Donselaar and Overeem, 2008). This very specific spatial arrangement of sedimentological and related poro-perm characteristics determines the direction of advective fluid flow and transport of dissolved arsenic in the aquifer  system beyond the clay-plugpoint-bar interface. Flow direction is along the thief zone at the base of the point-bar sediment body, and up the point-bar slope along the pressure gradient formed by water extraction through hand-pump wells by the villagers living on the elevated point-bars (see the dashed lines with arrows in the Graphical Abstract) and, consequently, draws the arsenic-polluted water to the well-head, thereby aggravating the pollution problem for the inhabitants.

Conclusions
With the aim to understand the magnitude of the arsenic-pollution problem in the shallow-aquifer domain of Holocene meandering-river geomorphology, the potential total arsenic load contained in clay-plug sediment was calculated by the integrated analysis of bulk sediment volume of twenty clay plugs (i.e., abandoned and sediment-filled meandering-river bends) in the Holocene alluvial plain of the Ganges River in Bihar, India with (a) remote-sensing imagery analysis of the clay-plug surface area, (b) four side-scan sonar surveys in present-day oxbow lakes and in the active Ganges River for the calculation of clayplug cross-sectional area, (c) sedimentological data of five fully-cored wells in clay plugs, point-bar sand, and a partly-filled oxbow lake, all in Bihar, complemented with sediment-type and arsenic content data from three wells in Haringhata, West Bengal, India, and (d) Elemental (ICP-MS based) data of arsenic distribution in the sediment columns. The total arsenic load stored in sediment of individual clay plugs was calculated to range from 0.07 ⋅ 10 6 kg to a staggering 3.13 ⋅ 10 6 kg, depending on the size of the clay plug. The juxtaposition of clay-plug and point-bar geomorphological elements with their lithofacies differences and inherent porosity and permeability anisotropy, is proposed as a recipe for arsenic release, transport and storage. Based on the arsenic, iron and TOC concentration trends in the boreholes, a scenario is presented for the desorption of arsenic by the process of respiratory arsenate reduction by microbial metabolism, which is favoured by the high TOC concentrations in clay-plug sediment. The released arsenic is transported to the interface between low-permeable clay-plug sediment and high-permeable point-bar sand by the process of molecule diffusion. The initial diffusion flux across the interface ranges between 10 1 and 10 2 g/(m 2 y − 1 ). Direction of advective fluid flow and arsenic transport in the adjacent point-bar sediment body is conditioned by the lithofacies distribution and intricate porosity and permeability anisotropy. Water extraction by hand-pump wells creates a pressure gradient in the pointbar aquifer system, and draws the arsenic-polluted water to the wellhead.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.