Stochastic models of floods

There are essentially two different approaches to the problem of flood analysis. One corresponds to the streamflow annual flood series (AFS); and the other to the streamflow partial duration series (PDS). Here we discuss characteristics of these approaches. Three stochastic models of floods presented in this paper are based on the streamflow PDS. Each model depends on certain assumptions concerning properties of exceedances of a base level x0. The second and third models represent improvement vis-a-vis the first one in the sense that they are based on less restrictive assumptions. The time when the largest exceedance occurs is also studied, and the distribution of this time is determined. Each exceedance is characterized, roughly speaking, by its duration and its volume. Here a method is proposed for determining the distribution of this volume. The distribution function of the largest volume in an interval of time [0,t] is also given. Good agreement between theoretical and observed distributions shows that the assumptions concerning the exceedances are not unduly restrictive.