Evaluation of the return period of the September 1992 flood in Genova, Italy

The September 1992 flood in Genova caused great damage and loss of life. The flood peak, at the outlet of a 12.5 km² basin, was estimated at 200 m³/s. Detailed rainfall measurements in a nearby raingage station during the flood were used in a rainfall- runoff model (RRM), which was adapted to reproduce the estimated maximum discharge. It was necessary to consider various sets of parameters, with the common ability to reproduce the observed maximum discharge. Thirteen sets were selected and assumed to have the same likelihood of being the true parameter set. In order to find the return period of the event under consideration, an equivalent population of rainfall was examined, namely, the family of constant-intensity rainfalls which, for different durations, had the same distribution for the annual maxima as the known distribution of the historical rains. First, equivalent rains for different durations were evaluated such as to reproduce the observed peak discharge, for each of the 13 parameter sets. The return period was then estimated by using the Monte Carlo method, i.e., generating sequences of annual maxima for the equivalent rainfalls for different durations and enumerating the cases in which at least one of them was, each year, greater or equal to those previously calculated. The comparison between the statistics of the generated sequences of rainfall maxima and those of the historical records necessitated modification of the generation procedure. This was done by introducing a distortion in the conditional distributions, which yielded a closer fit to the historical statistics.