The ESCENA (2008 to 2012) project is a Spanish initiative, which applies the dynamical downscaling technique to generate climate change scenarios based on an ensemble of Regional Climate Models (RCMs) consisting of PROMES, WRF, MM5 or REMO over Peninsular Spain and the Balearic and Canary Islands using a high resolution of 25 km. We describe the mean fields and interannual variability for temperature and precipitation in an ensemble of simulations forced by the high resolution ERA-Interim reanalysis (1990 to 2007) and compare them to the Spain02 observed data set. Maximum surface air temperature shows seasonal cold biases up to -2.5K in all models and it is clearly underestimated during the coldest seasons, but less so during summertime (JJA). Generally, there is a better agreement between observed and simulated minimum surface air temperature, which is slightly overestimated (up to + 2K) especially during wintertime (DJF). Regarding precipitation, all models except PROMES tend to show low dry biases during all seasons, especially for autumn on the Mediterranean coast of the Iberian Peninsula. With respect to the interannual variability, the PROMES simulations overestimate the standard deviation of maximum surface air temperature, while the remaining models tend to slightly underestimate it, and most models tend to underestimate the standard deviation of precipitation. The results highlight the ability of these RCMs to reproduce the mean fields and the interannual variability in a very complex terrain such as the Iberian Peninsula, showing a great diversity of climatic behavior. The evaluation of the ensemble results indicates a great improvement in the temporal correlation and the representation of the spatial patterns of temperature and precipitation for all seasons with respect to the individual models.