Shallow groundwater, the primary water source in the Bengal Basin, contains up to 100 times the World Health Organization (WHO) drinking-water guideline of 10 μg l−1 arsenic (As), threatening the health of 70 million people. Groundwater from a depth greater than 150 m, which almost uniformly meets the WHO guideline, has become the preferred alternative source. The vulnerability of deep wells to contamination by As is governed by the geometry of induced groundwater flow paths and the geochemical conditions encountered between the shallow and deep regions of the aquifer. Stratification of flow separates deep groundwater from shallow sources of As in some areas. Oxidized sediments also protect deep groundwater through the ability of ferric oxyhydroxides to adsorb As. Basin-scale groundwater flow modelling suggests that, over large regions, deep hand-pumped wells for domestic supply may be secure against As invasion for hundreds of years. By contrast, widespread deep irrigation pumping might effectively eliminate deep groundwater as an As-free resource within decades. Finer-scale models, incorporating spatial heterogeneity, are needed to investigate the security of deep municipal abstraction at specific urban locations.
The Bengal Basin hosts the largest case of mass poisoning in the world1. Excessive concentration of As occurs in shallow groundwater2 used for domestic supply by 70 million people, 30% of the combined population of Bangladesh and West Bengal, India. Half the shallow hand-pumped wells have As concentrations of 10–1,000 μg l−1 (ref. 2) and most inhabitants have no alternative water source. The use of groundwater was initiated in the 1960s; as a result much of the adult population has been exposed to toxic levels of As for three decades. The health impacts are potentially catastrophic3. Arsenic-affected groundwater has also been identified in fluvio-deltaic settings elsewhere in southeast Asia1,4,5.
The enormous scale of As contamination of shallow wells became apparent during the 1980s in West Bengal6 and the 1990s in Bangladesh7. No solutions have since been implemented that provide As-free water to most of the affected population. (By As-free, we mean water containing less than 10 μg l−1 As, the WHO drinking-water guideline, rather than the regulatory limit in Bangladesh and West Bengal, which is 50 μg l−1 As). Of the mitigation options, installation of wells to As-free depths in the aquifer8, usually taken to be greater than 150 m (ref. 9), offers the most popular, practical and economic solution1,10. In Bangladesh, more than 75,000 deep hand-pumped wells had been installed1 by 2007. Since 2000, deep wells yielding 2,500 m3 d−1 have been installed by local initiatives in over 100 rural supply schemes11. Previously, the Bangladesh Department of Public Health Engineering (DPHE) had fitted deep wells with pumps, each capable of yielding 4,500 m3 d−1, at more than 20 towns (DPHE, personal communication). Deep groundwater continues to be targeted, however there is concern12 that it may be vulnerable to invasion of As from shallow depths as a consequence of pumping.
Deep wells offer a solution to another problem: saline groundwater occurring at intermediate depth across most of the coastal region13,14,15,16. Here, pumping 'deep groundwater' may induce invasion by saline groundwater, which would be expected to precede arrival of As at deep wells, owing to the different distributions and geochemical behaviour of As and salinity. Our purpose is to review the deep groundwater environments of the Bengal Basin (Box 1). With reference to recent modelling results17,18, we focus on the vulnerability of deep groundwater to invasion by As. Any development of deep groundwater should be accompanied by chemical monitoring and consideration of the possible requirement for treatment to mitigate constituents such as iron, manganese and boron (ref. 19).
Distribution of arsenic
Arsenic-rich groundwater occurs in reducing9, grey-coloured, Holocene sediments20 at depths less than 150 m (Box 1). The As originates in association with a ferric oxyhydroxide coating of Himalayan-derived sediment21. Reducing conditions, sustained by organic carbon, favour As release to groundwater by microbially mediated22 reductive dissolution21 of the ferric oxyhydroxide. Spatial variability of As at the 10–20 km2 scale has been related to organic-carbon availability23, local sedimentology24,25,26 and groundwater flow27,28,29,30. Present uncertainties include the sources of carbon23,31, the As sorption capacity of aquifer sediments32 and future trends in As concentrations33.
Where reductive As release is absent, and yellowish-brown oxidized sediments with a capacity for As sorption exist, groundwater is As-free, notably in Pleistocene and older deposits deep beneath reduced Holocene sediments34,35 and at shallow depth in the vicinity of Pleistocene inliers2,36 (Box 1). Arsenic-free conditions also occur in grey, reduced Pleistocene sediments at depth37. In a national survey of Bangladesh2, of the wells at a depth greater than 150 m (which were principally in the coastal region), fewer than 1% exceeded 50 μg l−1 and 95% had less than 10 μg l−1. Arsenic concentration was generally negligible at depths greater than 200 m, attributed to the geochemical context31,37, the refractory nature of sedimentary organic matter27 and/or history of groundwater flushing2,38. A compilation of surveys2,12,19,39 giving a broader coverage of the basin (Fig. 1) indicates that As exceeds 10 μg l−1 in 18% of deep hand-pumped wells sampled, but whether this is a result of breached well casings or hydrological response to pumping remains uncertain.
Deep groundwater in the Bengal Aquifer System
The shallow floodplain aquifer is locally separated from deeper groundwater by a silt-clay aquitard in places, for example in West Bengal25,39, western Bangladesh33 and Khulna in the southwestern coastal region14. An aquitard has been reported as regionally persistent across central40 and southeast Bangladesh16. At Khulna, groundwater has been pumped from 200 to 350 m depth for municipal supply for over 30 years without inducing vertical flux of As or chloride-rich water from shallower levels14; similar experiences13 at other coastal towns have encouraged the view that a 'deep aquifer' might be developed more widely1,13.
Data from more than 2,000 deep boreholes12 have recently allowed a better-constrained interpretation of prevalent but laterally discontinuous aquitards (see Supplementary Fig. S1), with multiple layering leading to an effective large-scale vertical anisotropy41 in hydraulic conductivity. The resulting isolation of deep groundwater flow from shallow flow, and maintenance of vertical differences in hydraulic head, is similar to the effects of extensive aquitards. The distinction is that discontinuous aquitards could locally focus vertical flow25, providing pathways for invasion of deeper sediments by shallow groundwater where deep pumping imposes a downward hydraulic gradient. At issue is the conceptualization of lithological heterogeneity and its representation in models. The effective anisotropy provided by multiple discontinuous layers of silty-clay has been applied to describe a single, anisotropic aquifer at the scale of the entire Bengal Basin18. We summarize the As-free, deep groundwater environments of the Bengal Aquifer System (BAS) in the context of the basin's geological evolution (Box 1).
Defences against invasion of deep groundwater by As
Groundwater flow paths to deep pumping wells provide an element of protection if As concentrations in recharge areas are low, or travel times to deep wells are long. This is called the 'flow-pattern defence'. The potential for sediments along induced flow paths to adsorb or otherwise trap As provides a 'geochemical defence'. Wells should optimally be screened in locations protected by a combination of flow pattern and geochemistry, but where deep groundwater is ultimately vulnerable, the elapsed time before As arrival at pumping wells is critical. Geological interpretation can define the contexts of deep groundwater vulnerability, and the timescale of As arrival may be estimated using groundwater models and geochemical analyses, but acceptable timescales for security of supply are an economic and political consideration1.
Flow-pattern defence of deep groundwater. Groundwater flow paths to wells are controlled by spatial patterns of aquifer properties, hydrological surface conditions, aquifer geometry and the distribution of pumping. Flow patterns in the BAS have recently been evaluated by groundwater model analysis17,18 using a single, basin-scale, vertically anisotropic, homogeneous aquifer representation41, calibrated against groundwater heads and ages17. Conditions at the coast were represented by prescribing the equivalent freshwater head of water with a density of 1.025 kg l−1 and depth determined by bathymetry over an extensive offshore region, with a no-flow boundary at the southern limit of the model. Model robustness was demonstrated across a variety of boundary conditions and a range of parameter values17. The recharge provenance of 'deep' groundwater relative to shallow As sources, and the travel time from recharge to deep wells, were evaluated for basin-wide groundwater development and a range of development scenarios17,18 (Fig. 2). Deep pumping for domestic supply, with and without shallow irrigation pumping, was found to minimally perturb the subhorizontal flow paths from distant recharge zones. The flow-pattern defence protected groundwater at depths greater than 150 m across more than 90% of the As-affected area (Fig. 2c) indefinitely (modelled as more than 1,000 years18) if deep groundwater abstraction was limited to domestic supply and distributed among hand-pumped wells18. A south-central subregion stands out as more vulnerable to vertical flow on account of basin geometry, consistent with the south-central As anomaly (Fig. 1).
Suggestions that high-As irrigation water leads to accumulation of As in rice grains42, human exposure, and threats to sustainable agriculture43, might prompt widespread use of deep As-free groundwater for irrigation. However, the rate of irrigation pumping is about ten times that of domestic pumping44. Simulations17,18 show that deep irrigation pumping would amplify downward flow, considerably shortening travel times to deep wells, to as little as 30 years, and would create large drawdowns in water level (for example, of 20 m to 40 m at pumping depth), disabling deep hand-pumps and rendering some powered pumping uneconomic. Deep irrigation pumping thereby risks eliminating deep As-free groundwater as a source of domestic supply (Fig. 2b). In contrast, shallow irrigation pumping does not compromise the flow-pattern defence of deep groundwater (Fig. 2c), but provides extra protection by creating a hydraulic barrier against downward As migration.
Basin-scale analysis captures large-scale flow processes, but geological heterogeneity might locally allow more rapid penetration of As to deep groundwater. A zone of excessive As in deep groundwater of west-central Bangladesh, where more than 10% of wells at depths greater than 200 m have concentrations of As higher than 50 μg l−1, is attributed to the presence of thick Pleistocene palaeochannel sands of the proto-Ganges allowing migration of As to depth1. Future modelling analysis at subbasinal scale should be applied within the larger-scale framework and incorporate spatial heterogeneity, particularly where layers of low hydraulic conductivity are rare and earlier As breakthrough might occur. While spatial heterogeneity continues to generate uncertainty, factors of safety are desirable, such as the 1,000-year timeframe of the basin-scale models.
Geochemical defence of deep groundwater. The geochemical defence is restricted neither by depth nor stratigraphy but by reactivity within the sediment. The hydraulic characteristics of grey and yellowish-brown BAS sediments are similar, but their distinct chemical characteristics, evident in contrasting groundwater compositions25,35, reflect the geochemical processes that may retard As in the groundwater flow field.
Ferric oxyhydroxides have a large capacity for adsorbing dissolved As (refs 45,46) (Fig. 3a) as demonstrated for oxidized, Pleistocene sediments west of Dhaka35 and in central Bangladesh32. Adsorption capacity depends on sediment composition and history of exposure to As-rich, reducing water. Oxidized sediments proximal to the Pleistocene inliers and boundary hills are likely to have high adsorption capacity because of sustained oxic recharge; laboratory experiments35 confirm this, and indicate that adsorption is enhanced by the oxidizing potential of in situ manganese oxides. In contrast, isolated lenses of oxidized sand are likely to have been exposed to variable amounts of reduced groundwater, and their oxidizing and adsorption capacities correspondingly depleted.
Grey Pleistocene sediments37 also have geochemical attributes contributing to security against As invasion47,48. Deep, grey sediment in southern Bangladesh contains less than 1 to 210 ppm As, where groundwater consistently contains less than 10 μg l−1 As (ref. 47). The highest As contents were detected in grey micaceous silts, in which As is bound principally in authigenic pyrite37 (Fig. 3b). Arsenian pyrite has also been identified, at lower abundance, in deep sands, consistent with a strong correlation of As and sulphur (ref. 37). Accumulation of As in pyrite is a progressive diagenetic change constituting a refractory As sink in reducing environments. The effectiveness of As sequestration by actively forming pyrite within the deep sands is uncertain however, as is the adsorption capacity of framework grains within grey sediment. The contribution of these processes to the geochemical defence of deep groundwater requires further evaluation.
Conclusions and uncertainties
Studies of the hydraulics13,14,44,49 and geochemistry14,32,37,47 of BAS at depths greater than 150 m are few. Further lithological descriptions, measurement of hydraulic properties and groundwater head profiles, sediment mineralogical and sorption properties, water ages, and groundwater modelling analysis are necessary to improve evaluations of the vulnerability of deep groundwater to invasion by As.
Present groundwater models addressing the flow-pattern defence are limited by the paucity of hydraulic head data available for depths greater than 150 m, but they strongly suggest that without invoking the geochemical defence, widespread deep irrigation pumping might effectively eliminate deep groundwater as an As-free resource for domestic supply, possibly in less than 100 years. Consensus in evaluating the long-term security of deep hand-pumped wells remains to be achieved, but modelling indications are favourable. Deep municipal abstraction may be deemed economically and socially acceptable if secure for a more limited, but still substantial period before invasion by As or salinity. Modelling approaches need to be refined to elaborate the timescales of the security both of deep municipal and hand-pumped abstraction. To maximize the security of deep As-free groundwater, domestic wells should be screened as deep as possible within oxidized sediments. Domestic abstraction of shallow As-free groundwater in oxidized Pleistocene sediments relies solely on the geochemical defence19. In relation to both situations, the processes and limitations of As sorption need further investigation.
We thank J. Davies for discussion of concepts informing the description of the Bengal Aquifer System. M.A.H. is in receipt of a scholarship (BDCS 2006-37) from the Commonwealth Scholarship Commission.
About this article
Conceptualizing leakage and storage contributions from long open interval wells in regional deep basin flow models
Hydrological Processes (2019)