Application of the data clusterization approach to the satellite measures of altimetric variation in the Emilia-Romagna Region (Italy)

Clusterization and decomposition techniques for time series have been around for mor than 50 years, but is now that the computational power at disposition makes them an appealing strategy to deal with massive data sets comprehending million of entries. Subsidence, as natural and anthropogenic originated phenomenon, constitutes the perfect sample to put them in practice since it requires thousand and million of points in order to properly characterize an entire region in terms of ground vertical movement, and is, for this reason, we decided to address its analysis with a fresh and novel approach that, until the moment of publication of this thesis, hasn’t been adopted before. The Emilia-Romagna region at the northeastern part of Italy, due to historical, geological, and technical criteria, constitutes a perfect study case to work with for which more than a century of subsidence studies exists and up to date, local authorities have developed successful strategies to characterize and establish its criticality all over the area, not only with the intention of quantifying it but to set policies in order to mitigate and prevent its negative effects on population and properties. Through a combined approach that include decomposition analysis of normalized time series data for subsidence satellite InSAR measures, kmean and hclust clusterization (4 and 8 classes each), and GIS mapping the phenomenon will be studied in order to validate an algorithm that allows, at the moment, to identify the seasonal behavior of measure points located over underground storage facilities in order to capture the seasonal pattern of the ground movement. The analysis is carried out at three scales: macro (regional), meso (local), micro (areal) starting with a dummy validating test and presenting a more complex analysis for the areal case. Comparison and conclusions are drafted in terms of correlation between the clustered points, their seasonal subsidence behavior and their geographic correspondence. Results are promising and pave the way to further steps on which a greater zone comprehending near two million points and 13.000 km2 will be analyzed not only looking for ugs activities but for a more precise characterization in terms of extraction of other fluids (oil, gas and water) aiming to identify a characteristic regional trend, independent from anthropic activities, that allows performing more accurate studies and design projects taking into account the particularities of the area unaccounted for at the moment.