Application of a multivariate statistical index on series of weather measurements at local scale

In this study an innovative statistical procedure based on multivariate techniques is applied. It is able to identify and to interpret the correlation structure among different meteo-climatic descriptors measured in different sampling sites. The test case of Agri valley (Southern Italy), is presented, analyzing daily data of temperature, relative humidity and precipitation measured in seven monitoring sites, from 2000 to 2013. The Principal Component Analysis technique is recursively applied for each year and for each station, for calculating the synthetic Normalized Principal Component Index. This index allows to characterize quantitatively the information content of descriptors and/or of sites and to compare the behavior of the different variables in the correlation structure. The results show that temperature and relative humidity have high weights in the correlation structure, while the precipitation represents a singular variable. Moreover sampling sites may be classified according to their degree of stability in the correlation structure. A low degree of stability may be correlated with the occurrence of climate changes at the microscale.