This report presents the latest version of the Rossby Centre regional atmospheric model, RCA3, with focus on model improvements since the earlier version, RCA2. The main changes in RCA3 relate to the treatment of land surface processes. Apart from the changes in land surface parameterizations several changes in the calculation of radiation, clouds, condensate and precipitation have been made. The new parameterizations hold a more realistic description of the climate system. Simulated present day climate is evaluated compared to observations. The new model version show equally good, or better, correspondence to observational climatologies as RCA2, when forced by perfect boundary conditions. Seasonal mean temperature errors are generally within ±1oC except during winter in north-western Russia where a larger positive bias is identified. Both the diurnal temperature range and the annual temperature range are found to be underestimated in the model. Precipitation biases are generally smaller than in the corresponding reanalysis data used as boundary conditions, showing the benefit of a higher horizontal resolution. The model is used for the regionalization of two transient global climate change projections for the time period 1961- 2100. The radiative forcing of the climate system is based on observed concentrations of greenhouse gases until 1990 and on the IPCC SRES B2 and A2 emissions scenarios for the remaining time period. Long-term averages as well as measures of the variability around these averages are presented for a number of variables including precipitation and near-surface temperature. It is shown that the changes in variability sometimes differ from the changes in averages. For instance, in north-eastern Europe, the mean increase in wintertime temperatures is followed by an even stronger reduction in the number of very cold days in winter. This kind of performance of the climate system implies that methods of inferring data from climate change projections to other periods than those actually simulated have to be used with care, at least when it comes to variables that are expected to change in a non-linear way. Further, these new regional climate change projections address the whole 21st century.