This study was undertaken as part of a larger project to evaluate the impact of soil erosion on soil productivity in Costa Rica, Nicaragua and Mexico. An important part of the overall project consists of the use of the event-based EUROSEM model (European Soil Erosion Model) to predict soil erosion rates. This paper evaluates the use of the model both for single event and yearly soil loss estimations using erosion plot data from Nicaragua and data obtained through rainfall simulator experiments in Costa Rica and Mexico. EUROSEM was calibrated based on the hydrographs followed by the sedigraphs in Costa Rica and Mexico and this was followed by a model validation. In Nicaragua, model calibration was done using total soil loss values for 1993 and the model was consecutively validated using plot data for 1994 and 1995. The study stresses the importance of calibrating the model for individual catchments, and that the total area of plant stems and soil cohesion are crucial calibration parameters when modelling grassland with cover percentages above 60%. EUROSEM generally did not perform well on single event simulations in terms of simulating hydrographs and sedigraphs. Whereas the difference between observed and simulated total soil loss was between 0.0% and 100.0%, differences in total discharge, peak run-off rate and peak soil loss ranged between 2.0% and 326.5%. The difference was attributed to the model's inability to model crusting. The application of the model for yearly soil loss predictions looks promising with simulated and observed total soil loss values in Nicaragua differing by between 2.5% and 5.0%.