Drivers of Extreme European Summer Temperature Variability

We use the MPI-ESM Grand Ensemble to evaluate European summer temperatures and their main driving mechanisms under global warming. This 100-member ensemble from a fully coupled Earth system model provides the large samples of extreme European summer temperature events needed to robustly evaluate variability changes and quantify the contributions of both dynamical and thermodynamical drivers to this variability. Under a moderate warming scenario (RCP4.5), we find that by the end of the century European summer temperature variability increases up to 40% over central Europe, leading to a maximum increase in extreme temperature events with respect to the mean decadal climate state of almost 150% compared to 20th century conditions.

To represent the main drivers of summer temperature variability, we use geopotential height at 500 hPa, sea level pressure, soil moisture, evapotranspiration, and global mean temperature as predictors in a multiple regression model. Including both these dynamical and thermodynamical drivers we are able to account for up to 90% of the simulated variability. We find that the main contributors to extreme European summer temperature variability are the two dynamical drivers, geopotential height at 500 hPa and sea level pressure; by contrast, the local thermodynamical drivers, soil moisture and evapotranspiration, contribute significantly on a smaller scale to this variability only in northern and southern Europe, respectively. We also find that the strength of the correlation of extreme temperatures to both sets of drivers does not change substantially with warming; while the correlation to global mean temperatures increases slightly.

Our findings highlight the importance of simultaneously considering both dynamical large-scale drivers as well as local thermodynamical drivers as source of extreme temperature variability. Considering a multiple regression model with only soil moisture and evapotranspiration as drivers leads to a twofold overestimation of their correlation to extreme temperatures, as well as of the area of significance of this correlation, while only explaining about 25% of the variability. On the other hand, considering only dynamical drivers leads to slightly higher correlation to extreme temperatures than in the full regression model, and accounts for only around 60% of the variability.