Dynamical downscaling was performed with the use of ALADIN model, driven by the ERA-40 reanalysis, at 8 km horizontal grid spacing during a 10-yearly period (1992-2001). Thereupon, a simplified and cost-effective model version, so-called dynamical adaptation, was carried out with a 1-hourly frequency at 2 km horizontal grid spacing. Complimentary moment-based and spectral verification, performed on a number of surface stations in different climate regions of Croatia, suggested that downscaling was successful. The greatest average wind speeds are associated with areas where gap flows and gravity-wave breaking take place during bora flows. Systematic errors of 10-m wind speed are close to 1% in flat terrain and reach up to 10 % for coastal stations in the vicinity of Dinaric Alps, the latter due to underestimation of the strongest wind speeds. The shape of kinetic energy spectrum follows theoretical considerations regardless of the season and generally relaxes towards the orography spectrum as approaching the ground. Near the surface, divergent flows show more energetic than rotational at wavelengths smaller than 200 km. The main improvement of both mesoscale model versions is found for diurnal circulations, whereas the energy variance of subdiurnal flows is largely underestimated. Dynamical adaptation shows beneficial primarily for cross-mountain winds determined by the pressure gradients over the mountain range. 2 on 4-27-2012-->