Future changes in runoff over western and central Europe: disentangling the hydrological behavior of CMIP6 models

Abstract. A large ensemble of climate projections from the Coupled Model Intercomparison Project Phase 6 is analyzed to characterize changes in runoff over western and central Europe in the late 21st century under a high-end emissions scenario. Our second objective is to gain a better understanding of the mechanisms responsible for the inter-model uncertainties. For this purpose, the models are grouped according to their hydrological response using a hierarchical classification algorithm. Additional sensitivity experiments from two Model Intercomparison Projects are examined to better assess the role of the soil moisture–precipitation feedback and of the physiological impact of CO2 in this context. Half of the clusters show no significant change or a slight increase in annual runoff, while the others show a substantial decrease. Even when models agree on the annual changes in runoff, the changes in precipitation and evapotranspiration that drive them can be very different, even in terms of sign. Seasonal changes further differentiate the hydrological behavior of the different clusters. It is difficult to reject any cluster of models based on their accuracy in representing climatological averages and recent trends. The link between present-day averages or trends and future changes is generally weak, and there are in general no major inconsistencies with reference datasets, partly because of large observational uncertainties. Finally, we show that large-scale circulation in winter and the representation of the physiological impact of CO2 in summer are important for the unusually large hydrological changes projected by some models. The soil moisture–precipitation feedback is important in summer for the multi-model ensemble mean but not for the inter-model spread.