Although groundwater arsenic constitutes a major hazard to the health of the people of Southeast Asia, the exact mineralogical origin of the arsenic in these fluvial aquifers is still under debate. Fe(III) oxides are the dominant hosts of mobilizable arsenic in the sediments, with the role of secondary Fe(II)-bearing phases like mackinawite, siderite, vivianite, magnetite, and carbonate green rust (fougerite) still unclear. Based on published field data from Chakdaha (India), the importance of the phases for arsenic mobility is evaluated quantitatively using models of growing complexity. Arsenic heterogeneity can be explained by the presence of two contrasted redox zones in the aquifers, with Fe(III) oxides being the dominant sorbent for arsenic in the less reduced zones and Fe(II) sulfides and/or Fe(II) carbonates being the solid-phase hosts for arsenic under more reduced conditions below impermeable soils or close to rivers where sulfate is reduced. A 1D reactive transport model which simulates the transition between the two environments has been developed and compared to field data. The results show that microbial sulfate reduction followed by abiotic and/or biotic reduction of As(III)-bearing iron oxides accounts for the spatial heterogeneity of arsenic in such reduced aquifers.