On the shot-noise streamflow model and its applications

In this paper we consider the shot-noise model of streamflow series. We show how design discharge can be obtained by the stochastic intensity of thinned Poisson processes describing the peaks over a threshold. The main result concerns the stationary distribution of peaks. We derive an explicit expression for this limit distribution in terms of its Laplace transform. Approximation formulas are developed making use of the saddle point method for the asymptotic evaluation of contour integrals and the Post-Widder formula for inversion of Laplace transforms. We illustrate this methods on the case of Gamma-distributed shots. The stationary peak distribution is used to approximate the maximum value distribution for larger time intervals.