Multimodel ensembles, whereby different global climate models (GCMs) and regional climate models (RCMs) are combined, have been widely used to explore uncertainties in regional climate projections. In this study, the extent to which information can be enhanced from sparsely filled GCM RCM ensemble matrices and the way in which simulations should be prioritized to sample uncertainties most effectively are examined. A simple scaling technique, whereby the local climate response in an RCM is predicted from the large-scale change in the GCM, is found to often show skill in estimating local changes for missing GCM RCM combinations. In particular, scaling shows skill for precipitation indices (including mean, variance, and extremes) across Europe in winter and mean and extreme temperature in summer and winter, except for hot extremes over central/northern Europe in summer. However, internal variability significantly impacts the ability to determine scaling skill for precipitation indices, with a three-member ensemble found to be insufficient for identifying robust local scaling relationships in many cases. This study suggests that, given limited computer resources, ensembles should be designed to prioritize the sampling of GCM uncertainty, using a reduced set of RCMs. Exceptions are found over the Alps and northeastern Europe in winter and central Europe in summer, where sampling multiple RCMs may be equally or more important for capturing uncertainty in local temperature or precipitation change. This reflects the significant role of local processes in these regions. Also, to determine the ensemble strategy in some cases, notably precipitation extremes in summer, better sampling of internal variability is needed.