This PhD. research was set up to investigate the use of rainfall-runoff models for simulation of high water events in hillslope areas. First, dominant parameters for runoff production during high water events have been identified. Subsequently, the influence of antecedent conditions on runoff percentages during the event has been assessed. Finally, the required amount of spatial information in rainfall-runoff models was analysed. A procedure is developed to analyse the relation between model performance and the scale of the model elements. Based on the results, guidelines are presented for the development of rainfall runoff models as well as for measurement network design. Pilot studies are the 114 km Zwalm catchment (Belgium) and the 28000 km Mosel basin (France, Germany, Luxembourg). Results show that even for a small area like the Zwalm catchment, the use of a single raingauge is not enough to guarantee reliable estimates of the area-average rainfall. However, if a reliable estimate of catchment average parameter values and rainfall intensities are available, a lumped model performs as good as distributed model. For the Mosel basin on the other hand, simulation errors are introduced if spatial heterogeneity is not taken into account by a rainfall-runoff model. For a number of high water events it is shown that about four model elements are required to capture heterogeneity effects on the basin runoff almost completely. For some moderate events with large spatial differences, more spatial detail was needed. Finally, it is found that reliable estimates of the areal precipitation in the Mosel basin are guaranteed if a network of 10 well-distributed raingauges is available.