abstractThe 222Rn concentration simulated by the regional atmospheric model REMO over Europe and western Siberia is compared to in-situ records in Europe, and discussed in the context of site effects for stations that are also part of a CO2 observing network. The REMO model has a limited spatial domain, forced at its lateral boundaries with meteorological fields of the European Centre for Medium-Range Weather Forecasts and with tracer concentrations issued from the TM3 global transport model. The modelled 222Rn field is compared to measurements at six stations: two coastal ones (Atlantic Ocean and Baltic Sea), two low-elevation sites in plains, one mountain station and one high-altitude station. We show that the synoptic and diurnal 222Rn variability as simulated by REMO (55 km by 55 km) is realistic. In some cases REMO performs better than TM3, which is of coarser resolution, but this is not always true. At Mace Head, a station located near the western edge of the REMO domain, we show that the 222Rn “baseline” concentration is strongly influenced by boundary conditions, reflecting 222Rn transport from North America across the Atlantic Ocean. At Schauinsland, a mountain station in south-western Germany, even though the spatial resolution of REMO is not fine enough to reproduce transport processes induced by local topography, a fairly good agreement between model and measurements can be obtained, provided that one can determine from comparison of observed and modelled diurnal temperature changes which layer of the model is suitable for comparison with the data. Finally, the implications of modelling 222Rn are discussed here in the broader context of interpreting site effects that may also affect CO2 continental observations in Europe.