Generally, spatially distributed erosion models are 'validated' using only data on runoff and sediment load collected at the catchment outlet. This implies that one of the most important aspects of such models, i.e. the prediction of the spatial variation of erosion and deposition within the catchment is not at all tested. After an extreme event. in a small agricultural catchment in the Belgium loam belt, erosion and deposition patterns were mapped by measuring rill and gully volumes and the thickness of sediment deposits. Volumes of interrill erosion were estimated. From these data, the erosion and deposition budget was calculated. The physically based soil erosion model LISEM was then used to simulate the event. Model parameters which could not be derived from the field data were estimated using guidelines for LISEM proposed for South Limburg (Netherlands), an area very near and similar to the Belgian catchment. The model was calibrated on total erosion within the catchment by varying the infiltration parameters. The results of the simulations show that the model predicts reasonably well the overall sediment delivery ratio of the event (ca. 60%). However, the observed variation in erosion rates for different crop types is not well-predicted. In general, LISEM strongly overpredicts erosion rates on fields with an important vegetation cover. Furthermore, the correlation between predicted and measured erosion rates per crop type at locations of the measured rill transects is weak. The predicted pattern of deposition is generally similar to the mapped deposition pattern. Discrepancies are mainly due to the fact that, in some cases, important (unpredicted) deposition was observed upstream of vegetation barriers or alongside roads. The study clearly shows that it is not possible to evaluate the performance of spatially distributed erosion models by using catchment outlet data alone. Such a 'validation' can mask important spatial variations within the catchment which are not accurately predicted. It is therefore possible to obtain a good agreement between observed and simulated outlet data, while the spatial pattern erosion and deposition within the catchment is not well-described. (C) 1999 Elsevier Science B.V. All rights reserved.