Ingredients of German flood events

Floods can be caused by a large variety of different processes, such as short, but intense rainfall bursts, long rainfall events, which are wetting up substantial parts of the catchment, or rain on snow cover or frozen soils. Although there is a plethora on studies analysing or modelling rainfall-runoff processes, it is still not well understood, what rainfall and runoff generation conditions are needed to generate flood runoff and how these characteristics vary between catchments. In this databased approach we decipher the ingredients of flood events occurred in 161 catchments across Germany. For each catchment rainfall-runoff events are separated from observed time series for the period 1950-2013, resulting in about 170,000 single events. A peak-over-threshold approach is used to select flood events out of these runoff events. For each event, spatially and temporally distributed rainfall and runoff generation characteristics, such as snow cover and soil moisture, as well as their interaction are derived. Then we decipher those event characteristics controlling flood event occurrence by using machine learning techniques.

On average, the most important event characteristic controlling flood occurrence in Germany is, as expected, event rainfall volume, followed by the overlap of rainfall and soil moisture and the extent of wet areas in the catchment (area with high soil moisture content). Rainfall intensity is another important characteristic. However, a large variability in its importance is noticeable between dryer catchments where short rainfall floods occur regularly and wetter catchments, where rainfall intensity might be less important for flood generation. To analyse the regional variability of flood ingredients, we cluster the catchments according to similarity in their flood controlling event characteristics and test how good the flood occurrence can be predicted from regionalised event characteristics. Finally, we analyse the regional variability of the flood ingredients in the light of climate and landscape catchment characteristics.