CMStatistics 2020: Start Registration
View Submission - CMStatistics
Title: Use of compound Poisson-gamma distribution as a model for precipitation: A case study from the south of Sweden Authors:  Anastassia Baxevani - University of Cyprus (Cyprus) [presenting]
Christos Andreou - University of Cyprus (Cyprus)
Abstract: Many statistical models exist for modelling precipitation. The main challenge when modelling precipitation is that one needs to model both the presence of exact zeros that correspond to dry days and the amounts of precipitation on the wet days. We suggest using the compound Poisson gamma distribution, which belongs to the family of distributions known as the Tweedie family, and has the ability to produce exact zeros and marginals that follow the gamma distribution. We discuss different methods for fitting the \emph{compound Poisson gamma} distribution to daily precipitation data from southern Sweden. Then, we generate realistic artificial precipitation sequences to estimate different precipitation characteristics.