Occurrence analysis of daily rainfalls through non-homogeneous Poissonian processes

A stochastic model based on a non-homogeneous Poisson process, characterised by a time-dependent intensityof rainfall occurrence, is employed to explain seasonaleffects of daily rainfalls exceeding prefixed threshold values.The data modelling has been performed with a partitionof observed daily rainfall data into a calibration periodfor parameter estimation and a validation period for checking on occurrence process changes. The model has been applied to a set of rain gauges located in different geographical areas of Southern Italy. The results show a good fit fortime-varying intensity of rainfall occurrence process by 2-harmonic Fourier law and no statistically significant evidence of changes in the validation period for different threshold values.