Mechanisms behind large-scale inconsistencies between regional and global climate model-based projections over Europe

Important discrepancies in the large-scale summer climate change projections were recently detected between the global and regional climate models (RCM/GCM) in the EURO-CORDEX ensemble for several variables including surface temperature, total precipitation, and surface solar radiation. In this study, we use a new experimental framework inspired by the Big-Brother-Little-Brother protocol to explore the mechanisms responsible for generating large-scale discrepancies in future projections between GCM/RCM pairs over Europe in summer. Starting from past and future simulations with a perfect GCM/RCM pair (same resolution, same physics, same forcings), we then disentangle the role of potential sources of GCM/RCM inconsistency by carrying out targeted sensitivity studies. We show that by following such a perfect approach, it is possible to obtain a GCM/RCM pair without statistically significant inconsistencies in projected climate change. Such discrepancies are mainly generated by differences in aerosols representation and atmospheric physics. The role of plant physiology is limited and unlikely to be the dominant factor in the detected discrepancies. Finally, it is unlikely that the discrepancies in the EURO-CORDEX ensemble projections are a result of the upscaled added value, as we show that the effect of increased resolution is not strong enough and mostly limited to areas with complex topography. These findings raise important questions about the current practices in regional climate modelling. In the short term, implementing RCM external forcings consistent with the driving GCM can significantly improve the situation at low cost. In the long term, adopting a seamless strategy in developing the GCM/RCM models should be questioned.