Providing reliable information on climate change at local scale remains a challenge of first importance for impact studies and policymakers. Here, we propose a novel hybrid downscaling method combining the strengths of both empirical statistical downscaling methods and Regional Climate Models (RCMs). In the longer term, the final aim of this tool is to enlarge the high-resolution RCM simulation ensembles at low cost to explore better the various sources of projection uncertainty at local scale. Using a neural network, we build a statistical RCM-emulator by estimating the downscaling function included in the RCM. This framework allows us to learn the relationship between large-scale predictors and a local surface variable of interest over the RCM domain in present and future climate. The RCM-emulator developed in this study is trained to produce daily maps of the near-surface temperature at the RCM resolution (12 km). The emulator demonstrates an excellent ability to reproduce the complex spatial structure and daily variability simulated by the RCM, particularly how the RCM refines the low-resolution climate patterns. Training in future climate appears to be a key feature of our emulator. Moreover, there is a substantial computational benefit of running the emulator rather than the RCM, since training the emulator takes about 2 h on GPU, and the prediction takes less than a minute. However, further work is needed to improve the reproduction of some temperature extremes, the climate change intensity and extend the proposed methodology to different regions, GCMs, RCMs, and variables of interest.

Since a decade, convection-permitting regional climate models (CPRCM) have emerged showing promising results, especially in improving the simulation of precipitation extremes. In this article, the CPRCM CNRM-AROME developed at the Centre National de Recherches Météorologiques (CNRM) since a few years is described and evaluated using a 2.5-km 19-year long hindcast simulation over a large northwestern European domain using different observations through an added-value analysis in which a comparison with its driving 12-km RCM CNRM-ALADIN is performed. The evaluation is challenging due to the lack of high-quality observations at both high temporal and spatial resolutions. Thus, a high spatio-temporal observed gridded precipitation dataset was built from the collection of seven national datasets that helped the identification of added value in CNRM-AROME. The evaluation is based on a series of standard climatic features that include long-term means and mean annual cycles of precipitation and near-surface temperature where CNRM-AROME shows little improvements compared to CNRM-ALADIN. Additional indicators such as the summer diurnal cycle and indices of extreme precipitation show, on the contrary, a more realistic behaviour of the CNRM-AROME model. Moreover, the analysis of snow cover shows a clear added-value in the CNRM-AROME simulation, principally due to the improved description of the orography with the CPRCM high resolution. Additional analyses include the evaluation of incoming shortwave radiation, and cloud cover using satellite estimates. Overall, despite some systematic biases, the evaluation indicates that CNRM-AROME is a suitable CPRCM that is superior in many aspects to the RCM CNRM-ALADIN.