Integrating statistical information concerning historical floods: ranking and interval return period estimation

The importance of the historical information in flood analysis has previously been underlined. In this context, we present an integral methodology aimed at the establishment of return periods of different flood units on the unique basis of historical data. Specifically, the reconstruction of the flood chronology extended back to 1900, complemented with a new (historical data-based) event intensity index, and the return period estimation will be addressed. Since some of the historical data are collected from interviews and other sources with different degrees of precision and reliability, two kinds of uncertainty will be considered; namely the statistical variability and the imprecision. We propose an innovative methodology involving intervals (ranges), fuzzy sets, and weights to formalize and average the criteria that determine the importance of the different events. On the other hand, to take into account the statistical variability, we propose to estimate the return period in a flexible and efficient way by considering bootstrap confidence intervals. The methodology is particularly useful at ungauged, or partially gauged, flood inundation areas, where the existing flow gauge stations do not give the flood series at the point of interest. A case study developed in Spain is discussed. The results are supported by two recent events, which have been mapped at 1:5000 scale.