ABSTRACT

Changing hydroclimatological conditions lead to changes in hydrological risk. Recent hydrological extremes such as floodings along the Rhine, Mississipi, or Oder Rivers can to a large extent be explained by the occurrence of unusual hydroclimatological extremes. Whether these extremes are part of natural variability, indicate possible climatic fluctuations, or are signals of an anthropogenically induced climate change is the first question to be answered. For this purpose, time series of different hydrological variables (atmospheric circulation patterns, rainfall, and runoff) are investigated. As hydrological risk is related to extremes, the series are investigated from that viewpoint, and not only from that of their mean behavior. Different statistical methods including nonparametric methods and bootstrap are applied to selected series to test the hypothesis of stationarity. Whenever this hypothesis is rejected, assumptions about the future have to be made. This can either be a scenario based on present trends, an assumption of stationarity at the present level, or a scenario based on a general circulation model (GCM). In the GCM case, due to the coarse resolution, a downscaling method is also needed. The next step is to assess the probabilities of extremes under these changes. It is demonstrated that areal precipitation extremes should be evaluated using not only precipitation amount but also duration and persistence of events. An example of extreme areal precipitation demonstrates this part of the methodology. On the basis of scenarios based either on present conditions or assumed trends, one may obtain a direct assessment of the flood risk. In contrast, GCM-based scenarios do not yield runoff values, thus a hydrological model has to be used to transform downscaled hydroclimatological series into runoff.