Five regional scale models with a horizontal domain covering the European continent and its surrounding seas, one hemispheric and one global scale model participated in an atmospheric mercury modelling intercomparison study. Model-predicted concentrations in ambient air were compared against mercury species observed at four monitoring stations in Central and Northern Europe and a station on the Irish west coast. The modelled concentrations of total particulate mercury (TPM) were generally consistent with the measurements at all sites. The models exhibited significant ability to simulate concentrations of gaseous elemental mercury (GEM), but some of the short-duration peaks at the Central European stations could not be consistently reproduced. Possible reasons for these discrepancies include (1) errors in the anthropogenic emissions inventory utilized; (2) coarse spatial resolution of the models; and (3) uncertainty of natural and re-emitted mercury sources. The largest discrepancies between measurements and modelled concentrations were found for reactive gaseous mercury (RGM). For these models, the uncertainty in predicting short-term (two-week episode) variations of mercury species in air can be characterized by the following overall statistics: 90% of the results for TGM are within a factor of 1.35 of the measurements; for TPM, 90% are within a factor of 2.5; and for RGM, 90% are within a factor of 10.