Low flow in Bavaria: derivation of drought characteristics and their future development in a hydrological single-model large ensemble

In the past two decades, Europe has been hit by major summer heat waves and droughts, with heavy impacts on ecology, economy and civil society.In addition to increased risk of crop failure, forest fires and danger to human health, extensive dry conditions may lead to riverine low flows and general water scarcity. Low flow conditions can restrict river navigation, hydropower production, and limit water use for power plant cooling and irrigation agriculture. Furthermore, the ecological state of the river is impaired.To address these challenges, setting up a hydrological model based on a large ensemble climate simulation provides the required data to evaluate the water availability under future heat and drought conditions. Therefore, we create a hydrological large ensemble with 50 realizations for the periods 1990 – 2099 featuring the Water balance Simulation Model (WaSiM). The single-model initial condition large ensemble (SMILE) CRCM5-LE (CRCM5-Large Ensemble) used consists of 50 transient simulations (50 members) of a regional climate model of 150 years each (1950-2099, 7500 model years, hourly time step, 0. 11° spatial resolution) and provides the meteorological forcing data, after bias correction and statistical downscaling to the hydrologic model application scale, for 98 gauges simulated with the WaSiM-ETH water balance model in hydrological Bavaria. Due to the high number of model years, this model chain on the one hand provides a novel way to transfer and assess the non-linear relationships of the natural variability of the climate system within the hydrological system, and on the other hand results in a sufficiently large number of extreme events to conduct a robust statistical analysis.Based on the modeling results, the dynamics of the low flow situation in Bavaria is mapped for the reference period (1981-2010), spatial patterns of drought are highlighted, and regional correlations are identified. To allow for seasonal comparisons of the negative anomalies of the runoff event, the variable-threshold approach is used. Here, the threshold is defined as the 15th percentile for the 30-day moving average of the discharge value for each day of the year, averaged over the reference period. An undershoot of this threshold for at least 20 days is considered a drought event. The use of climate simulation data allows for an analysis of how these characteristics (intensity, duration, spatial occurrence of the drought event) will change in the future due to climate change. Emphasis is placed on the potential change in the seasonal regime and the associated impacts on river system usage. By accounting for the natural variability of the climate system through the ensemble approach, the results become more robust, particularly with respect to extremes, and strengthen confidence in the change signals that are observed.Results of these analyses are presented using a representative sample of watersheds for the entire study area, highlighting common features as well as unique characteristics. The evaluations provide important evidence for the basic definition of low-flow events and a robust estimate of how their intensity, frequency, and seasonality changes in the future as a result of climate change impacts.