The runoff coefficient for a T-year design flood, using data from Austrian catchments

The runoff coefficient is the percentage of rainfall that becomes runoff (and river discharge) during a storm, and it has a relevant role in engineering practice. It is, in fact, a key parameter in the so called "design-storm" method, in which hypothetical storm events of a given return period are transformed into flood discharges that are assumed to have the same return period. In order to investigate what values of runoff coefficient should be used in the design-storm method, in this study we analyse the correlation between event runoff coefficients and catchment characteristics, accounting for the relationship between the return periods of the storms and of the flood peaks. We estimate the runoff coefficient for flood events from hourly rainfall and runoff timeseries for several catchments in eastern Austria. Each event is identified with an automatic procedure based on three main steps: (1) baseflow separation from direct runoff in the discharge timeseries; (2) identification of every single event starting from the largest peak flow to the smallest ones exceeding a pre-defined threshold; (3) estimation of the event runoff coefficient using a rainfall-runoff model that minimizes the root mean square between the observed and the modelled runoff. For each event, we then estimate the return period associated with the respective peak flow (TQ) and maximum precipitation for different durations (TP). Results indicate that TQ can be much higher than the corresponding TP and that their relationship is related mainly to the wetness of the system, represented by the event runoff coefficient. It is also recognisable a dependency of the runoff coefficient with the return periods of rainfall and runoff, since the runoff coefficient increases moving towards extreme peak discharge values. In engineering practice, the median value of the distribution of runoff coefficients is usually adopted, in order to transform the design-storm into a flood peak with the same return period, but this may not be the right choice. Our results show that, because of the dependency of the runoff coefficient on the return periods and the high variability found in real events, it is difficult to choose the proper value in a simple way. We find that the value of the runoff coefficients for which TQ is close to TP in real events varies very much for small events but, for large events with return periods of the order of 10 years or more, tends to stabilise around the value that is exceeded 80% of the times, considering the distribution of runoff coefficients for all events. In the region of interest, i.e., eastern Austria, these runoff coefficients have values close to 0.4.