Statistical analysis and modeling of precipitation data Original Research Article

The objective of the present paper is to perform statistical analysis and modeling of actual precipitation data collected in 20 stations in Greece. Although our analysis and statistical modeling procedure utilizes precipitation data from Greece, the same methodology can be applied to similar data or other environmental variables. We identify the probability density function (PDF) that best characterizes the behavior of the data, that being the Frechet PDF, a member of the family of the Generalized Extreme Value (GEV) family of distributions. In addition to obtaining useful information, such as the expected rainfall along with its variance and confidence limits, we are able to cluster the stations into 7 groups that have similar response. Furthermore, we introduce time dependence into the Frechet probability distribution by characterizing the location and scale parameters to be functions of time and other well known, but more complex covariates such as the North Atlantic Oscillation index. This approach enables a more flexible and realistic model to quantify the uncertainty inherent into the system, and allows for more complex dependence patterns.