Understanding the processes causing herbicide transport to surface waters is crucial to determine mitigation options to reduce these losses. To this end, we investigated the atrazine (2-chloro-4-ethylamino-6-isopropylamino-1,3,5-triazine) transport in three agricultural catchments (1.1-2.1 km2) in the watershed of Lake "Greifensee" (Switzerland). In 1999, atrazine application data were recorded for all three catchments. Time proportional samples were taken at a high temporal resolution at the catchment outlets. Extremely wet conditions caused large relative losses from the catchments, ranging between 0.6 and 3.5% of the amount applied. Most of the atrazine load was due to event-driven diffuse losses from the fields. Farmyard runoff contributed less but caused the highest concentrations (up to 31 microg L(-1)) in the brooks. The maximum concentrations due to diffuse losses varied between 1.2 and 8.2 microg L(-1) among the catchments. Despite different absolute concentration levels, the concentration time-series were very similar. It seems that the travel-times within the catchments were mainly controlled by the rainfall pattern with little influence of the catchment properties. These properties, however, caused the relative losses to vary by a factor of 6 between the catchments. This variability could be partly explained by differences in the connectivity of the fields to the brooks and by their hydrological soil properties. A comparison of the losses from the three catchments with those from the entire watershed of Lake Greifensee demonstrated that they were representative for the larger area. Hence, the study results provide a good data set to evaluate distributed models predicting herbicide losses.