Using a multi-model ensemble of climate-change impacts over the Iberian Peninsula, we identified regions and cropping options for which the uncertainty levels of projected impacts on crop productivity were either high or low. The ensemble consisted of 10 members per combination of scenario, climate model, location and cropping option, and was used to quantify the uncertainty of impacts on crop yield. CERES wheat and maize simulation models were linked to the control run and 1 scenario provided by 10 regional climate models (RCMs): control and A2 future (2070-2100) climate. The contribution of RCMs, locations and cropping options to uncertainty on yield projections was analysed. Differences between the sign of the response and 30 yr time series of projections generated by each member of the ensemble were compared. The largest response to A2 scenarios also resulted in the smallest uncertainty, and vice versa. Low uncertainty was found for the sign of the yield response, which was mainly positive for spring and winter cropping options and negative for the summer option. Uncertainty was lower for A2 than for control projections. Uncertainty was largest in northern, coastal and mountain regions, and smallest for inland southern regions, and depended on seasonal cropping options. Minimum and maximum uncertainty were found for maize and irrigated spring wheat, respectively. Water availability was the determinant for interannual variability and its uncertainty. Choice of RCM contributed less to uncertainty than choice of location, and choice of cropping option contributed more to uncertainty than both of these factors. Interannual variability showed larger uncertainty than mean impact magnitude, and this uncertainty was larger than that of the sign of the yield response. Regions with high uncertainty could benefit from higher-resolution simulations.