A stochastic model for the analysis of the temporal change of dry spells

In the present paper a stochastic approachwhich considers the arrival of rainfall events as a Poissonprocess is proposed to analyse the sequences of no rainydays. Particularly, among the different Poisson models, anon-homogeneous Poisson model was selected and thenapplied to the daily rainfall series registered at the Cosenzarain gauge (Calabria, southern Italy), as test series. The aimwas to evaluate the different behaviour of the dry spellsobserved in two different 30-year periods, i.e. 1951–1980and 1981–2010. The analyses performed through MonteCarlo simulations assessed the statistical significance of thevariation of the mean expected values of dry spellsobserved at annual scale in the second period with respectto those observed in the first. The model has then beenverified by comparing the results of the test series with theones obtained from other three rain gauges of the sameregion. Moreover, greater occurrence probabilities for longdry spells in 1981–2010 than in 1951–1980 were detectedfor the test series. Analogously, the return periods evaluatedfor fixed long dry spells through the synthetic data ofthe period 1981–2010 resulted less than half of the correspondingones evaluated with the data generated for theprevious 30-year period.