"Flood analysis for regions, like Southern Italy, where the annual flood series exhibits outliers (and, then, high skewness), associated with disastrous storms, requires building suitable stochastic models. In such cases, the usual simple model (Model A), which assumes the largest annual flood to be the maximum of a Poissonian number of indepen dent random variables with common exponential distribution function, proves to be inadequate. Better models can be built by replacing the hy_ potheses on which Model A is based with others, phenomenologically closer to reality, namely, that the number of exceedances in a year is still a non-homogeneous Poisson process, but the exceedance values are not identically distributed random variables. Of the two models considered, i.e., a time-dependent distribution for the exceedancemagnitude (Model B) and a mixed exponential distribution (Model C), the latter is found to give a better statistical fit. There is also better phenomen-ological support for Model C in that disastrous storms occur more rarely but with much larger intensities than others, and they are accordingly better modelled as belonging to different populations."
Analysis of Flood Series by Stochastic Models
ROSSI, Fabio
1982-01-01
Abstract
"Flood analysis for regions, like Southern Italy, where the annual flood series exhibits outliers (and, then, high skewness), associated with disastrous storms, requires building suitable stochastic models. In such cases, the usual simple model (Model A), which assumes the largest annual flood to be the maximum of a Poissonian number of indepen dent random variables with common exponential distribution function, proves to be inadequate. Better models can be built by replacing the hy_ potheses on which Model A is based with others, phenomenologically closer to reality, namely, that the number of exceedances in a year is still a non-homogeneous Poisson process, but the exceedance values are not identically distributed random variables. Of the two models considered, i.e., a time-dependent distribution for the exceedancemagnitude (Model B) and a mixed exponential distribution (Model C), the latter is found to give a better statistical fit. There is also better phenomen-ological support for Model C in that disastrous storms occur more rarely but with much larger intensities than others, and they are accordingly better modelled as belonging to different populations."I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
