Evaluating the performance of CMIP5, CMIP6 and EURO-CORDEX ensembles in capturing recent climate trends in Slovenia

Climate projections for regions with complex topography, such as Slovenia, typically rely on high-resolution Regional Climate Models (RCMs) such as the EURO-CORDEX ensemble. However, the added value of dynamical downscaling in representing recent historical trends remains a subject of ongoing evaluation. We compare the performance of various climate model ensembles – CMIP5 and CMIP6 Global Climate Models (GCMs), CMIP5 EURO-CORDEX RCMs and bias-corrected subselection of CMIP5 EURO-CORDEX RCMs – against gridded observations over Slovenia.

The analysis focuses on ten-year moving average of temperature and precipitation anomalies since 1950 (relative to the 1981–2020 baseline) and linear trend estimates for the period 1981–2025. Our results indicate a significant discrepancy: the EURO-CORDEX RCMs underestimate the accelerated warming observed in recent decades while GCMs align more closely with observed trends. Almost all RCM ensemble members fall outside the two standard error range of the observed trend at 0.47–0.71 °C per decade. In contrast, GCM ensembles reflect the magnitude of recent observed temperature increase over Slovenia more accurately, especially the latest CMIP6 generation with approximately half of the ensemble members falling within two standard error range of the observed trend, while for CMIP5 this ratio is approximately one fifth (depending on the underlying emission scenario). Due to high interannual precipitation variability, resulting in less pronounced trends in precipitation, differences between the RCM and GCM ensembles in capturing observed precipitation trends are less significant. Some differences appear on a seasonal scale for spring and summer, where GCMs slightly outperform RCMs.

Our findings suggest that while RCMs provide necessary spatial detail, they may fail to capture certain large-scale drivers of recent rapid warming. Many RCMs in the EURO-CORDEX ensemble use constant aerosol distributions or simplified versions that do not decrease over time, despite the observed decrease in aerosol pollution in recent decades. In addition, many CMIP6 models are set at higher equilibrium climate sensitivity, resulting in more warming than CMIP5 generation and associated EURO-CORDEX models driven by respective boundary conditions. The results highlight the importance of model selection, consideration of model physics and potential limitations of relying solely on RCM ensembles for assessing climate change impacts in the sub-Alpine and sub-Mediterranean transition zones.