CHARACTERIZING THE LONG-TERM VULNERABILITY OF INTERMEDIATE AND DEEP AQUIFERS TO ARSENIC CONTAMINATION IN BANGLADESH UNDER THE INFLUENCE OF URBAN PUMPING

In Bangladesh, millions of people continue to consume groundwater, typically from 20-50 m depth, with arsenic (As) concentrations that exceed the 10 µg/L safe drinking water limit set by the World Health Organization. Cities pump from the deep (>150 m), low-As aquifers, whereas surrounding rural populations rely on the shallow (<50 m), intermediate (50-150 m) and deep aquifers for their household needs. The latter two aquifers, which typically contain much lower As concentrations, are the sole source of safe drinking water sources for the rural populations and are under threat. Decades of pumping in Dhaka has depressurized the intermediate and deep aquifers in surrounding rural areas. This threatens to increase As concentrations in these aquifers by the in-situ release of As through reductive dissolution of iron oxides by dissolved organic carbon (DOC) that advects downward from the shallow aquifer between breaks in clay layers or is expulsed by the compaction of clay layers at depth. Our study region is Araihazar sub-district, 15-35 km east of Dhaka, where blanket testing of As concentrations in 49,000 wells was performed using field kits in 2012-13. Forty-five wells spanning all depths down to 220 m were equipped with pressure transducers recording water levels every 20 minutes from 2012 to 2019. Over this time period, hydraulic heads in aquifers deeper than 50 m declined as the Dhaka drawdown cone expanded. This accentuated vertical hydraulic gradients between the shallow and intermediate aquifers which could drive the downward advection of shallow groundwater and the compaction of deeper (>50 m) clay layers. A groundwater model is developed in MODFLOW to evaluate processes that have led to the observed spatial distribution of As concentrations by explicitly incorporating clay compaction, DOC expulsion and advection, and reductive dissolution of iron oxides. The model is constrained by a 3-D hydrostratigraphic model of 94 lithological logs using the transition probability algorithm, hydraulic heads, and measured hydraulic and geochemical properties. By integrating these processes into one model the results should enhance understanding of the vulnerability of the intermediate and deep aquifers under the impact of Dhaka pumping.