Pesticides and nitrates represent the main sources of aquifer contamination in agricultural zones. In many regions, nitrate concentration levels reach and exceed the water quality criteria (50 mg NO3/L). The increasing use of mineral fertilizers (which has doubled during the 20 last years) and the intensive exploitation of the aquifers for crop irrigation (1,1 million ha in France) have led to groundwater contamination by nitrates. The dynamics (long-term persistence) and extensiveness (regional contamination) of this contamination make it a sensitive environmental issue. Comprehensive environmental management is needed in order to limit the increase of the concentration levels and to reduce the extent of the contaminated areas. During the last few years, research has been done in the field of watershed management, from laboratory experiments to field investigations. At the same time, numerous simulation models have been developed at different investigation scales. Banton et al. (1993) developed a model specifically devoted to environmental management. Their model, AgriFlux, is based on a mechanistic approach to the processes, using a stochastic method that takes into account the spatial variability of the parameters. AgriFlux calculates the nitrate concentrations as well as the water fluxes in the unsaturated zone. The concentrations in the unsaturated zone (obtained by modeling or measurement) are generally dissimilar to those observed in the saturated zone (i.e. in the aquifer) because the infiltration water is diluted in the aquifer water. This difference indicates that the concentrations in the unsaturated zone cannot be used to accurately evaluate the actual risk of groundwater contamination. Hydrogeological conditions such as the recharge limits, the flow direction and the flow rate should be incorporated into the evaluation. In this paper, the modeling results obtained previously (Dupuy et al., 1997) with AgriFlux for the La Jannerie watershed are used to determine the concentrations in the aquifer and to compare them with the concentrations measured in the observation wells. This watershed (160 ha) is used exclusively for agriculture. The fractured carbonate strata (Superior Oxfordian) constitute a phreatic aquifer with a vertical extension of about 20 m. First, the temporal evolution of the annual mean concentrations in the aquifer is compared with the evolution of the annual precipitation. The results show that the mean concentrations tend to follow precipitation levels. However, the differences observed at different locations in the watershed cannot be explained by these results. The spatial evolution of the concentrations from the upstream to the downstream part of the aquifer was studied in order to explain the concentration distribution in the watershed. On the main flow line, the concentrations observed from 1985 to 1989 show a decrease from the P7 well (upstream) to the P26 well (downstream). This phenomenon can be attributed to two factors. First, denitrification may occur in the aquifer during flow. However, it is recognized in literature that the denitrification rate is usually low and a long period of time is required to obtain a significant decrease in the nitrate level. The observed attenuation cannot be imputed to this factor alone. The second possible cause is related to the dilution of the nitrates in the water contained in the aquifer.
In order to test this hypothesis, a dilution model was elaborated using the watershed division as indicated in Dupuy et al. (1997). In each area, the resulting concentration is obtained by diluting the fluxes of water and nitrate leaching in the unsaturated zone in the fluxes of water and nitrates flowing from the upstream area. The concentrations in the aquifer are calculated from upstream areas to downstream areas for the period between 1985 and 1989. The pattern of the concentration curves obtained in this manner agrees with the trend measured in the different wells. The results clearly show a decrease of the concentration in the aquifer water leached from the unsaturated zone. For the downstream area (well P26), the calculated concentrations are higher than the observed ones. This difference could be due to the fact that the lateral fluxes (flow convergence into the median part) are not taken into account and the concentrations may thus be overestimated. However, the mean resulting error (12%) remains low considering the lack of knowledge of the aquifer characteristics. It is therefore possible to accurately estimate the nitrate concentrations in the saturated zone from the concentrations simulated in the unsaturated zone using a simple dilution model. However, this method is only valid for simple hydrogeological conditions.
Non point source pollution, watershed, unsaturated zone, saturated zone, dilution model.
Alain Dupuy, Laboratoire d'Hydrogéologie, JE 274 MESR, Université de Poitiers, 40 avenue du Recteur Pineau, 86022 Poitiers, FRANCE
Email : email@example.com