ANALYSIS OF RAINFALL MONITORING NETWORK BY EVOLUTIONARY POLYNOMIAL REGRESSION

In Italy there is a rainfall monitoring network that has been built incrementally over the last 150 years following different logics and largely tied to the mere coverage of the territory, within the limits of the budget availability that has progressively become available over the years. In light of the changes in rainfall patterns resulting from the so-called "climate changes" underway, the need has emerged to understand whether this network is suitable for following the new dynamics in terms of land coverage and to allow detailed rainfall monitoring. This study considers a part of this national network, consisting of 25 rainfall stations located along the coast of the Ionian Sea in Puglia, managed by the Decentralized Functional Center of the Department of Civil Protection of the Puglia Region, Italy. The study analyzes precipitation data recorded at these stations over the last decades through a machine learning approach based on Evolutionary Polynomial Regression (EPR), aimed at revealing potential correlations between observations at different stations. This analysis will be useful to analyze the level of land coverage by the stations and the possibility of predicting missing observations (gaps in the data) through observations in neighboring/related stations. The research has produced models with highly satisfactory levels of accuracy. Furthermore, the analysis of the correlations between the observations at the various stations has allowed to highlight situations in which the network needs integration.