Conçus initialement pour le traitement des eaux usées domestiques, les filtres plantés de roseaux ont été adaptés pour le traitement des effluents de lavage de salles de traite pour fournir une alternative au stockage et épandage. Pour simplifier les système en vue de réduire les coûts d'investissement, plusieurs sites expérimentaux ont été construits, modifiés si nécessaire et suivis. Placé à la suite d'une fosse toutes eaux de 3 à 9 m3, la configuration la plus aboutie consiste en 2 étages des filtres plantés à flux verticaux chacun étant composé de 2 filtres alimentés en alternance chaque semaine. La surface totale active correspond à 0.25m2 par vache laitière pour les eaux blanches seules (machine à traire et tank à lait) et 0.4m2 par vache lorsque s'y ajoutent les eaux vertes de lavage des quais et murs de la salle de traite. Les filtres du 1er étage occupent 65% de la surface totale. Observées dans 3 exploitations agricoles laitières, les concentrations obtenues atteignent une fourchette de 50 à 180 mg/L en DCO, 5 à 75 mg/L en DBO5 et 10 à 60 mg/L en MES. La nitrification qui est considérée comme un bon indicateur du potentiel d'oxydo-réduction dans le réacteur n'est certes pas complète mais beaucoup plus élevée que dans le cas de filtres à sable étudiés pour le même objectif alors que plusieurs d'entre eux ont colmaté. / Initially designed for the treatment of domestic wastewater, Vertical Flow Reed Bed Filters [VFRBF] have been adapted for washing parlour effluents to study the feasibility of the treatment of such effluents instead of their storage and land spreading. It was necessary to adapt the design of VFRBF to minimise the investment costs. Several experimental sites have been built, monitored and progressively adapted. Following a septic tank of 3 to 9 m3, the best design was considered to be: 2 stages of VFRBF in series each composed of 2 filters which are alternately fed on a weekly rhythm. The total active area is approximately equivalent to 0.25 m2 per milking cow for the washing effluent of the milking machine and milk-storage tank and 0.4 m2 per milking cow when the washing effluents from the floor and walls of the milking parlour are added to the previous ones. The 1st stage filters takes up 65 % of the total area. Based on a survey done in 3 farms, the concentrations achieved a range of 50 to 180 mgL-1 in COD, 5 to 75 mgL-1 in BOD5, 10 to 60 mgL-1 in SS. The nitrification, which can be considered as a good indicator of the redox potential within the filter media, was not complete but much more effective than in sand filters studied for the same purpose but some of them have clogged.

Phosphorus removal from wastewater has been of growing interest for some decades to avoid eutrophication in surface water. In subsurface constructed wetlands precipitation and adsorption are the main mechanisms responsible of P uptake. The use of calcareous media is attractive because of p-ca interactions. Two media (calcite and recycled crushed concrete (rcc)) were examined in batch and continuous systems. They show attractive sorption capacities using batch experiments, however experiments carried out in open reactors pointed out some limitation in retention capacities and effluent quality. rcc was sensitive to a strong dissolution leading to a quick phosphorus precipitation but induced high conductivity and ph values in the treated water. Calcite efficiency depended on the material dissolution rate, directly linked to the carbonate equilibrium of the solution. Microscopic observations of the calcite surface show crystal making of phosphorus precipitate. Crystal growth seems to be the main P uptake once materials surfaces are covered. No experimental condition allows to reach an effluent concentration below 2 mgp.l-1.