Dynamics of exceptional peak discharges of Bavarian rivers in a changing climate

Over the last 30 years heavy precipitation events, which were either locally bound or spacious in extent have caused several extreme floods in Bavaria, such as the Pentecost flood in 1999 affecting the entire Danube region, followed by floods in 2002, 2005, and 2013. All these floods have been described as events which statistically should only occur once in a hundred years (100-year flood) or longer. The time span between these high flow events, separated by only a few years, indicates that these severe events have become more frequent. In populated areas these events cause severe damage and often involve human casualties leading to an increased attention from the general public and science. According to their definition, extreme events such as the 100-year flood occur rarely; thus, they are only sparsely covered in discharge observations. The 100-year flood is frequently used in Bavaria and elsewhere as a design criterion for the development and construction of flood protection measures or hydro-power facilities. Hence, a reliable estimation of its current value and future dynamics due to a changing climate is important. These critical thresholds are generally derived from the existing discharge time series by applying various methods of extreme value statistics. However, since most available discharge time series are too short for an empirical estimation, the statistical methods need to extrapolate beyond the observed record to estimate the 100-year flood magnitude. For a more reliable quantification of the 100-year flood magnitude and in order to account for any changes in response to climate change, new approaches employ hydrological models to create long time series of discharge data based on meteorological inputs provided by weather generators or large climate model ensembles.
\nThis dissertation investigates the impact of climate change on extreme flood events for all major Bavarian catchments. A hydro-meteorological model chain for the assessment of climate change impacts on the hydrology (scenario - global climate model (GCM) - regional climate model (RCM) - hydrological model) is employed. The model chain uses a Single-Model Initial Condition Large Ensemble (SMILE) of the Canadian Regional Climate Model, version 5 (CRCM5-LE) forced by the Representative Concentration Pathway 8.5 (RCP8.5) emission scenario, to drive the hydrological model WaSiM (Water Balance Simulation Model, formerly WaSiM-ETH). Three scientific publications address different aspects of this model chain regarding its application to simulate high flow events and their change in dynamics (i.e., frequency and intensity) in response to a changing climate.
\nThe first publication addresses the development of the hydrological model itself. The process-based and fully-distributed model WaSiM was set up for 98 catchments of the Bavarian Danube and Main, as well as their tributaries (such as the Inn). Since some of these catchments extend beyond the political borders of Bavaria, the entirety of them all is further referred to as the Hydrological Bavaria. To account for the spatial and temporal dynamics of hydrological extreme events, the model was set up in a high spatiotemporal resolution. Furthermore, regionalized model parameters were determined using a semi-global and semi-automatized approach, focusing on the representation of high flows. To determine the model’s performance to simulate these events a confidence value (Level of Trust, LOT) was introduced which shows the deviation of discharge values of selected return periods (1 in 5-, 10-, 20-years) between model data and observations at the respective gauge. The discharge values were estimated using an extreme value distribution (Generalized Pareto Distribution with Peak over Threshold sampling and L-Moments for parameter estimation). The results show that the model performs sufficiently well with values of the Nash Sutcliffe Efficiency and Kling-Gupta Efficiency above 0.6 for most of the gauges. The results regarding the LOT, which represents the capability of the model to reproduce high return period events, depict moderate (between 20% and 30% deviation) to very high (less than 10% deviation) confidence for the majority gauges. However, the number of gauges yielding trustworthy results (above moderate LOT) reduces with an increasing return period.
\nRCM data often exhibit systematic deviations from long term mean observations (bias) which should be removed for climate change impact studies. Hence, the second publication investigated which of the selected methods for bias correction (BC; linear scaling, local intensity scaling, quantile-mapping, qm; yearly and monthly correction factors) is best suited for the adaptation of raw RCM data (by means of their impact on different hydrological indicators) and how these methods affect the climate change signal of different hydrological indicators for a selection of catchments within the Hydrological Bavaria. Although a BC is inevitable in many cases due to a strong bias in precipitation (amounts, seasonal course) and/or temperature, its application is often critically discussed as most approaches result in incoherence between variables and may alter the original climate change signal. As shown in the second publication, the qm approach with monthly correction factors is recommended for the adjustment of RCM outputs for the catchments of the Hydrological Bavaria as it yields either the best adjustment to simulations using observations or performs similarly well than other methods. Further, the presented results of this study illustrate that the employed BC methods affect the change signal of the presented hydrological indicators. Change signals for extreme event indicators are more affected by different BC methods, with more 100% difference in absolute change values in extreme cases, than those of long term mean flow indicators, with differences in relative CCS between 0 and 15 percent points.
\nThe third publication focuses on the main scope of this dissertation: the impact of climate change on the dynamics (i.e., frequency and intensity) of extreme flood events for the 98 catchments of the Hydrological Bavaria focusing on the 100-year flood. For this purpose, the model introduced in the first publication is driven by the CRCM5-LE climate simulations which have been corrected using the qm approach as recommended in the second publication but adapted for daily correction factors. The resulting hydrological SMILE (hydro-SMILE) provides a large database of 1,500 model years per 30-year period (50 members x 30 years) for the analysis of extreme events. A comparison between values for the 100-year flood obtained by a Generalized Extreme Value distribution (GEV) and the empirical probability of exceedance is made to illustrate the benefit of the hydro-SMILE for a robust estimation of extreme flood events. The robust estimation of 100-year flood events further allows for the assessment of possible changes in the frequency and intensity by direct comparison between present and future values. The presented results show the benefit of the hydro-SMILE for the robust estimation of extreme high flow events using empirical probabilities compared to statistical estimates using an extreme value distribution. Furthermore, the results show that for catchments exhibiting a nival (snow) component in their runoff regime a considerable to severe increase in frequency (between 7 and 12 times as frequent) and intensity (between 36% and 104%) of 100-year flood events is expected until the end of the century. In catchments exhibiting a more pluvial influence in their flow regime (especially north of the Alps) these dynamics are less pronounced (at least 10% to 25% increase in intensity for more than 50% of the gauges, up to a maximum between 20% and 44%; at least 1.5 times as frequent for more than 50% of the gauges, up to 3 times as frequent at the maximum) or in individual cases even show a decline in frequency and intensity. Other studies also show this behavior in the dynamics of extreme floods for the upper Danube. However, the methods employed in this dissertation allow for a better quantification of a dynamically changing hydrological system under a transient changing climate.
\nThis dissertation illustrates the results of a state-of-the-art modelling chain employing a single RCM large ensemble driving a single hydrological model under a strong emission scenario to study the changes in dynamics of high flow events in the Hydrological Bavaria.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.