Experimental validation of a geographical information systems-based procedure for predicting pesticide exposure in surface water.

A GIS-based procedure for predicting pesticide exposure in surface waters has been applied on a pilot river basin characterized by intensive agricultural activity. The predictive approach has been validated through experimental monitoring, performed by collecting manual and automatic water samples during the productive season. Five active ingredients (terbuthylazine, metolachlor, alachlor, linuron, fenitrothion) were selected for analysis to validate the predictive approach. Comparison between predicted and experimental values showed good agreement for terbuthylazine and metolachlor (used in large volumes within the basin), demonstrating the reliability of the approach. However, some anomalous results were obtained for some of the other chemicals, which serve to highlight the difficulties in getting reliable input data, in particular on application patterns (rate and time). Furthermore, the value of mapping pesticide exposure on the medium-large scale is described, and the limitations of the reported predictive approach are discussed.