A climate version of the nonhydrostatic fifth-generation Penn State/NCAR Mesoscale Model has been used to downscale a global climate scenario to cloud-resolving scales over complex terrain (the Alps). After first describing the model and methodology, we then present comparison results from the model-predicted and ensemble-averaged regional-scale winter and summer season precipitation distribution with results from the global simulation and analyzed observed precipitation climatologies. Finally, results from the cloud-resolving simulations are compared to the regional simulations. It is shown that the degree to which the terrain is resolved in the various runs significantly alters the simulation of the precipitation climatologies. This is caused not only by the complex interaction of the flow with the topography but also by the different treatment of the convective processes (resolved versus nonresolved) in the model. Even in winter, the model-simulated seasonally averaged precipitation patterns change drastically with every increase in horizontal resolution. Furthermore, with a horizontal grid resolution of 1 km, when seen on a local scale and over complex terrain, the model-simulated precipitation patterns are not guaranteed to converge to one solution. This behavior is still more complicated in summer. Here it is shown that parameterized convection in the regional model simulation tends to be locked to the mountains, while in the cloud-resolving simulations the convection moves with the upper level flow, producing precipitation maxima away from the mountain tops. ¿ 2000 American Geophysical Union