Arsenic rich AMDs (Acid Mine Drainages) represent a major source of pollution for aquatic ecosystems. Microbially driven iron (Fe) and arsenic (As) oxidation and precipitation represent a promising strategy to treat this pollution. A better understanding of the biogeochemical mechanisms involved is required prior any further exploitation of this microbial potential. A field-scale pilot was implemented at the Carnoul`es mine (France) for the treatment of AMD. It is an ergonomic and passive aerobic system: five treatment units of 1.5 m2 are combined vertically in series and fed with the AMD water by gravitation flow. Biogenic precipitates (corresponding to Fe- and As-rich biofilms) covered the bottom of the units. Inlet water and biogenic precipitates were collected over a 7 months period. We determined the bacterial community structure in the precipitates by fingerprint (ARISA), metabarcoding (16S rRNA gene) and qPCR targeting arsenite oxidase gene aioA. Chemical and mineralogical analyses were conducted on the precipitates and on the feed water. The bacterial communities in the precipitates developed from the indigenous communities of the AMD used to feed the pilot. Our results showed an evolution of these communities over time associated with an increase of the potential genetic for As oxidation. The proportion of As(V) in the precipitates and arsenic removal efficiency fluctuated, with maximum levels of 99% and 97 % respectively. This work provided information about microbial dynamics and pollution removal efficiency in a treatment pilot under field conditions. It will serve for future design of a bioremediation system to treat As-rich AMD.