GNGTS 2015 - Atti del 34° Convegno Nazionale

GNGTS 2015 S essione 2.1 27 Here the time-series analysis is not performed with the goal of estimating station velocities, but with the goal of generating several types of “residual” time-series, where with the term “residuals“ we intend the time-series obtained after removing a certain model from the raw data-set. In particular, we first generated the full residual time-series. These residual time-series ideally would contain only noise, and among the most important sources of time-correlated noise, the so-called Common Mode Error is the most important one. Thus, as previously mentioned, we estimated the CME using a Principal Component Analysis (PCA) technique on the residual time-series, and using the PCA outputs to apply a space-time filtering on the time-series, improving daily repeatability, accuracies and precisions of the velocity estimates. We generated two different sets of filtered (CME removed) time-series, for two different sets of stations. Two areas were identified, for which we generated two different sets of filtered (CME-removed) time- series: Area1) encompassing the Alpine and northern Apennines, Po Plain area, for which we generated de-trended time-series, where only the non-tectonic signals (i.e., instrumental offsets) have been removed, so the time-series contains the seasonal terms, and Area2) corresponding to the area where InSAR displacement time-series are available, for which we generated de- trended residual time-series, where also the seasonal terms are removed. Fig. 3 – The first three (and most representative) principal components of the PCA method (from left to right). U represents the spatial response with respect to the temporal trend (V), as V is the response of the function in time. Concerning the first and second components, one-year-lasting signals are extracted and a seasonal component prevails, while for the third component a cyclical multi-annual basis signal is present.