The long-standing dilemma of European summer temperatures at the Mid-Holocene and other considerations on learning from the past for the future using a regional climate model

Abstract. The past as an analogue for the future is one of the main motivations to use climate models for paleoclimate applications. Assessing possible model limitations in simulating past climate changes can lead to an improved understanding and representation of the response of the climate system to changes in the forcing, setting the basis for more reliable information for the future. In this study, a Regional Climate Model (RCM) is used for the investigation of the Mid-Holocene (MH, 6000 years ago) European climate, aiming to contribute to end the long-standing debate on the reconstruction of MH summer temperatures for the region, and gaining more insights on the development of appropriate methods for the production of future climate projections. Two Physically Perturbed Ensembles (PPEs) are first built by perturbing model physics and parameter values, consistently over two periods characterized by different forcing (i.e. the MH and Pre-Industrial (PI)). The goal is to uncover possible processes associated with the considered changes, that could deliver a response in MH summer temperatures closer to evidence from continental-scale proxy reconstructions. None of the investigated changes in model configuration produces remarkable differences with respect to the mean model behaviour. This indicates a limited sensitivity of the model to changes in the climate forcing, in terms of its structural uncertainty. Additional sensitivity tests are further conducted for the MH, by perturbing the model initial soil moisture conditions at the beginning of spring. A strong spatial dependency of summer near surface temperatures on the soil moisture available in spring is evinced from these experiments, with particularly remarkable differences evident over the Balkans and the areas north of the Black Sea. This emphasizes the role of soil-atmosphere interactions as one of the possible drivers of the differences in proxy-based summer temperature evident between Northern and Southern Europe. A deficiency of the considered land scheme of COSMO-CLM in properly retaining spring soil moisture, evinced from the performed tests and further confirmed by the evidence of present-day studies, suggests that the consideration of more sophisticated schemes may help bridging the gap between models and proxy-reconstructions. Finally, the distribution of the PPEs with changes in model configuration is analyzed for different variables (T2, PREC, TCLC). In almost all of the considered cases the results show that what is optimal for one period, in terms of a model configuration, is not the best for another characterized by different radiative forcing. These results raise a concern about the usefulness of automatic and objective calibration methods for RCMs, suggesting that a preferable approach is the production of small PPEs that target a set of model configurations, properly representing climate phenomena characteristic of the target region and that will be likely to contain the best model answer under different forcing.