Project 1 - Hydrology of extraordinary floods – event analysis

The subproject 1 “Hydrology of extraordinary floods – event analysis” examines how extreme floods are generated, which factors affect their peaks, volumes and shapes and how these factors interact with each other. In this context a detailed analysis of extreme flood events in different regions is conducted. For this purpose, the largest floods in selected river basins in Germany and Austria are analysed in regards to flood generating processes and the interaction between these processes. Particular consideration is given to precipitation, soil moisture and the interaction with runoff generation and the routing of the flood waves. The subproject consists of two phases. In the first phase, regional flood events of the Mulde and Inn Rivers are selected and analysed. In the second phase the results of the first phase are tested, validated and, if necessary, adapted. For this purpose, the basins of the Aller-Leine and Neckar are used. At the beginning of the first phase, available hydrological and meteorological data of the Mulde and Inn basins are processed. Peak-Volume-Relationships are used to characterise the similarities and differences between flood types. In this context, the antecedent moisture at selected catchments is determined and an analysis of the precipitation runoff conditions and the coincidence of floods in these areas is performed. This analysis is done according to Schumann et al. (2016). The analysis is divided into three sectors.

The first step is the determination of the antecedent precipitation in the catchments. For this, the 5, 6-10, and 30 days antecedent precipitation values of the areal precipitation are identified and a weighting of the antecedent precipitation is performed. This evaluation not only regards the temporal distances to the peak, but also the different conditions of the catchments in different seasons. The determined weighted antecedent precipitation is put in relation to their maximal values to identify which flood events have similar antecedent precipitation conditions. Afterwards, the discharge previous to the flood event is examined. The initial discharge of the flood events in relation to the mean discharge of the season in which the event took place (winter or summer) is calculated. Through this, the base flow at the beginning of the flood events becomes apparent. Thereafter, the spatial distribution of the flood triggering precipitation events are assessed. In this regard the precipitation events of varying durations at the selected catchments are compared by the 72h, 48h, 24h, 12h, and 6h maximal amount of the areal precipitation.

Figure 1 shows exemplary the maximal amount of areal precipitation belonging to different sub basins for the flood events in 1954, 1958, 1995, 2002 and 2013. Differences between the extreme events of 2002 and 2013 are distinct. The convective part of the precipitation and the recipitation duration in the catchments are calculated based on the heterogeneity of the rainfall distribution. For this, the maximal amount of the areal precipitation (6h, 12h, 24h, and 48h) for each flood event is compared with the next higher duration period. This leads to the heterogeneity quotient.

Figure 1: 24h precipitation amount for 15 catchments and five floods at the Mulde basin (Schumann et al. 2016).

Following the antecedent moisture considerations, the precipitation discharge conditions of the flood events are outlined. Concerning this matter the discharge coefficient, this means the relation between direct flow and the precipitation amount, are calculated by using a homogeneous regression model. After that, a classification of the flood hydrographs in intensity and amount dependent event developments is done to show the influence of the precipitation conditions on the flood event. To concretise the connection between discharge generation and concentration, the peak discharge per unit area and the discharge coefficient are compared.

This is done by the consideration of amount and intensity dependent peak discharges per unit area with the means of the heterogeneity quotient of the precipitation. Subsequently, the flood coincidence is examined. Thereby, the effect of coincidental flood waves on the peak discharge and therefore the annuality of the flood event is evaluated. The comparison between Figure 2 and 3 shows that during an assumed critical overlay of the flood peaks in 2002 a clearly higher peak discharge at the Golzern gauge occurs. Potential superimpositions and their probabilities can be assessed by multivariate statistic models.

Figure 2: Observed superimposition of the temporal shifted flood hydrograph of the Zwickauer Mulde, Zschopau and Freiberger Mulde (gauge: Nossen) 2002 (Schumann et al. 2016).

Figure 3: An assumed superimposition of the peaks of the temporal shifted flood hydrograph of the Zwickauer Mulde, Zschopau and Freiberger Mulde (gauge: Nossen) during the Flood event of 2002 would have created a 13% higher peak (Schumann et al. 2016).

Following from this, the extreme floods of the Mulde and Inn basins are chosen for analysis. Thereafter, a data based flood analysis takes place. For this purpose, the analytical framework of Woods and Sivapalan (1997) and Viglione et al. (2010) is used and adapted by a GIS based modelling approach. Such modelling approach is needed to estimate the flood characteristics that cannot be deduced directly from observed data. For this, large-scale precipitation fields are analysed using machine learning algorithms (e.g. support vector machines (SVM)). This is done to determine special precipitation distribution patterns and to classify the heterogeneity of hydrological loads. On this basis a spatial differentiation of the catchment response is performed. That means, the catchment response is determent depending on the precipitation loads and a categorisation of these reaction patterns is done on the basis of the discharge hydrograph. Figure 4 shows an example for different precipitation loads and consequent distinctions in discharge reactions. Despite the similarity in the flood peak discharges, the temporal distributions of precipitation are very different (e.g. high intensities and low overall duration in 2002, low intensities and long overall duration in 2013).

Figure 4: Flood triggering precipitation of the 2002 (above left) and 2013 (above right) flood at the Mulde basin and their flood hydrographs (below) at the Zwickau/ Zwickauer Mulde gauge.

These response patterns are used to identify events with similar runoff formation mechanisms in the range of past floods in the respective catchments. For this purpose, a model based analysis of the process controls of extreme floods in the Mulde and Inn basins is done. Therefore, the factors that are responsible for the extremisation of the flood such as soil moisture, intersection of flood waves and extreme precipitation intensity or volume are estimated. On this basis a derivation of extreme flood scenarios is done for observed smaller floods. Figure 5 shows the different hydrographs of the floods in 2002 and 2013 for selected gauges in the Mulde basin. Differences in the hydrograph forms between these events alongside the Mulde River are probably referable to the above-mentioned extremisation factors.

Figure 5: Flood hydrographs of the 2002 (left) and 2013 (right) Mulde flood along the Mulde River and its tributaries at selected gauges.

This phase will lead to a method to classify and to determine the causes of extreme flood events in a region. The factors of extremisation are characterised causally and statistically and potential influences in regards to flood protection planning is shown. In the second phase the comparative analysis includes the Aller-Leine and Neckar basins. The deductions from the first phase are transferred to other regions to understand flood responses in a comparable way given the different catchment characteristics and hydro-meteorological drivers. The aim of this study is to create a generalised understanding of extreme floods and the specific process controls that depend on the catchment characteristics and hydro-meteorological conditions.


Schumann, A.; Fischer, B.; Büttner, U.; Bohn, E.; Walther, P.; Wolf, E. (2016): Vergleich der größten Hochwasser im Muldegebiet (Schriftenreihe des Landesamtes für Umwelt, Landwirtschaft und Geologie Freistaat Sachsen, 18). Online available, last updated 2016, last proven 29.09.2016.
Viglione, A.; Chirico, G. B.; Komma, J.; Woods, R.; Borga, M.; Blöschl, G. (2010): Quantifying space-time dynamics of flood event types. Journal of Hydrology 394 (1-2), S. 213–229. DOI: 10.1016/j.jhydrol.2010.05.041.
Woods, R.A; Sivapalan, M. (1997): A connection between topographically driven runoff generation and channel network structure. Water Resources Research (33 (12)), S. 2939–2950.