The robustness of the Numerical Atmospheric-dispersion Modelling Environment (NAME) for forecasting the dispersion of volcanic ash clouds is investigated by comparing the output from different Volcanic Ash Advisory Centre (VAAC) models initialised using the parameters for the 2004 Grimsvötn, Iceland, volcanic eruption. London, Darwin, Washington, Montreal and Toulouse VAAC dispersion models are all run operationally as if responding to the eruption. Comparison of the model set-ups reveals differing approaches between the VAACs for model averaging times, ash release rates, and thresholds for defining the ash cloud, amongst others. The importance of these factors is considered in detail. Despite using different weather conditions and having different structures, the models all demonstrate strong similarities for forecasting regional ash cloud transport. The dispersal of volcanic ash is simulated over Scandinavia and as far as Eastern Europe in all cases. Greater variations are seen between the forecast ash concentrations for different aircraft flight levels. The model forecasts are highly dependent on the amount of eruption information available at the time. Copyright © 2007 Royal Meteorological Society