Climate change effects on river droughts in Bavaria using a hydrological large ensemble

Abstract. Europe and Germany have been hit by severe meteorological droughts in recent years, which have resulted in extreme low flow conditions in streams. Climate projections expect an intensification and increase in the frequency of the associated meteorological drivers. In this study, we investigate the impact of climate change on rare and extreme river droughts in a catchment with a pluvial regime, a northern tributary to the Danube (Wörnitz river) and in a river catchment with nivo-pluvial regime in the Pre-Alps (Ammer river). We employ a unique physically-based modelling chain, where the hydrological model WaSiM (Water balance Simulation Model) is driven by 50 members of the single model initial-condition large ensemble (SMILE) of the Canadian Regional Climate Model version 5 under the high-emission scenario RCP8.5. This results in a hydrological SMILE yielding 1500 simulated years for each of the investigated 30-year periods of a reference (1980 – 2009; REF), a current (2010 – 2039; CUR), a near future (2040 – 2069; NF), and a far future climate (2070 – 2099; FF). We investigate the seasonality, univariate and bivariate return periods of peak low flow and duration, and explore the climatic drivers causing the alterations.

The Wörnitz catchment shows a summer low flow regime, with climate change affecting the seasonality so that the river droughts are projected to extend further into the autumn. The typical bivariate 100-year event (REF: 7-day peak low flow = 1.96 m³/s; event duration = 171 d) shifts to a 30-year (CUR), 17-year (NF), and 6-year (FF) event. In the reference climate of the Ammer catchment, the intensity of winter low flows dominates over summer low flows. However, this is projected to switch during the current climate with more intense summer river droughts in the near and far future. While there is a tendency for cold-dry winters to cause low flow conditions in February during the reference climate, future winter low flows shift towards November/December and are triggered by the hot and dry pre-conditions of an antecedent summerly river drought. The most probable bivariate 100-year summer low flow event (REF: 7-day peak low flow = 4.9 m³/s, event duration = 60 d) is drastically altered to occur every 34 years (CUR), 8 years (NF), and 2.5 years (FF).

In both catchments, there is an increase in the autocorrelation of peak low flows from one summer to the next, which emphasises the causality and increasing importance of lagged effects and preconditions in the course of climate change. We identify hotter and drier summer seasons as the main driver, with the positive interdependency between heat and drought in climate change intensifying, further exacerbating extremes.

Thereby, the study highlights the opportunities of a hydrological SMILE for the investigation of river droughts. Due to the large sample size of the hydrological SMILE, we can robustly assess very rare events and generate bivariate design values narrowing down uncertainties of extreme value statistics in the light of a well characterized internal climate variability. In turn, all results are subject to scenario and model uncertainties, as the simulations are carried out with one hydrological model driven by one climate model under one emission scenario.