Non Stationary Analysis of Extreme Events

Traditional approaches to the analysis of extreme hydrological series are based on the stationarity assumption for the underlying processes, namely that the probability distribution of the hydrological variable does not change with time. Over the last decade however, a growing interest has arisen both from a scientific as well as engineering point of view, toward the development of tools able to cope with the apparent non stationary features (either natural or anthropogenic) observed in many hydrological processes. Though most of the works deal with extreme precipitation and floods, less attention has been devoted to modeling droughts under non stationarity paradigm. In the paper, a brief review of the available tools for modeling non stationary series is presented. An extension of such methodologies to drought lenght modeling is developed, taking into account the non stationary nature of the underlying series and/or of the threshold level used for drought definition. An example of application of the developed methods to four precipitation series in Sicily, Italy, exhibiting different degrees of trends is also presented.