Vertical-profile sampling through a ~33 m-thick glacial overburden aquifer in central Massachusetts reveals distinct hydrogeochemical units. The upper ~12 m is characterized by low oxidation-reduction potential (ORP) and high dissolved iron and arsenic. Between ~14 and 26 m, ORP is positive and iron and arsenic levels drop markedly. Below ~26 m and into bedrock, ORP is again negative, arsenic is elevated, but iron remains low. Extensive physical characterization suggests that the hydrology strongly influences subsurface redox conditions, and, therefore, the distribution of dissolved arsenic.

Hydraulic conductivity, K, has been characterized in historic pumping tests and associated calibration of a basin-scale numerical model, as well as by slug tests, grain-size analyses, and specific capacity measurements collected in the present study. The upper ~12 m comprises sands and silts with K of the order of a few m/d. The effective vertical conductivity in this layer is likely one to two orders of magnitude lower due to silty interbeds. The underlying layer (from ~12 to ~21 m) exhibits conductivity of the order of 100 m/d, and K decreases with depth below this interval.

In the upper ~12 m, relatively static groundwater in the presence of organics in glacio-lacustrine facies has resulted in development of reducing conditions. Reductive dissolution of iron oxides within this zone releases sorbed arsenic. Groundwater in the underlying, high-K zone has a relatively short residence time, resulting in oxidizing conditions, and low dissolved iron and arsenic. Pumping of supply wells screened in the high-K unit enhances the contrast with the overlying reducing zone. With increasing depth, lower K and longer pathways from recharge areas again lead to reducing conditions and elevated dissolved arsenic.

We conclude that extraction of groundwater from the high-K, low-As layer induces leakage from the overlying low-ORP, high-As zone. Mixing of these compositions occurs as water approaches the supply wells. Numerical simulations based on an idealized hydrostratigraphy, along with backward particle tracking from the pumping wells, support this conceptual model.