| description abstract | The hydrological cycle on a regional scale is poorly represented with a present-day coarse resolution general circulation model (GCM). With a dynamical downscaling technique, in which a regional higher-resolution climate model (RCM) is nested into the GCM, this starts to become feasible. Here the authors go one step further with a double nesting approach, applying an RCM at 19-km horizontal resolution nested into an RCM at 57-km resolution over an area covering the Scandinavian Peninsula. A 9-yr-long time-slice simulation is performed with the driving boundary conditions taken from a fully coupled ocean?atmosphere GCM experiment, the recently completed ECHAM4/OPYC3 control simulation performed by the Max Planck Institute for Meteorology in Hamburg. With increasing resolution, local effects playing a significant role in the hydrological budget become better and better resolved and are more realistically simulated. It is found in particular that in mountainous regions the high-resolution simulation shows improvements in the simulation of hydrologically relevant fields such as runoff and snow cover. Also, the distribution of precipitation on different intensity classes is most realistically simulated in the high-resolution simulation. It does, however, inherit certain large-scale systematic errors from the driving GCM. In many cases these errors increase with increasing resolution. Model verification of near-surface temperature and precipitation is made using a new gridded climatology based on a high-density station network for the Scandinavian countries compiled for the present study. The simulated runoff is compared with observed data from Sweden extracted from a Swedish climatological atlas. These runoff data indicate that the precipitation analyses are underestimating the true precipitation by as much as 96% on an annual basis in the most mountainous region of Sweden. This fact as well as estimates of the underestimation in other areas of Scandinavia make the high-resolution RCM simulations appear more realistic. | |
