GNGTS 2015 - Atti del 34° Convegno Nazionale

28 GNGTS 2015 S essione 2.1 It is worth noting that, in order to reduce the percentage of missing data when forming the PCA analysis, we have restricted the time-span processed with PCA to the 2006-2012 time interval, before the beginning of the May 2012 Emilia earthquake sequence and when most of the CGPS sites were operative. Moreover, we have excluded all GPS stations with a number of epochs <50% of the total expected. Results. An interesting analysis method of space-time distributed data is the Principal Component Analysis (PCA) method. In brief, PCA compresses data by finding an intrinsic orthogonal coordinate system for the observation data that preserves the greatest possible amount of variance ( ‘spread’ of the distribution). PCAprojects the data from a high-dimensional space (data or sensor space) onto a low-dimensional space, and allows detecting signals in the geodetic time series that show a spatial coherence higher than a desired threshold. In the considered case study, the first PC clearly represents a long-term linear motion, mainly associated with the subsidence of areas located in proximity of cities like Bologna (PC1 in Fig. 3). Such a signal is dominating the time series. For some frames, the second PC shows a random temporal evolution, with a spatial pattern mainly focused around the edges of the frame. This result may represent the influence of the ground control points for the whole data set. Finally, the third PC (for some frame sorted as second, and switched with the previous one) shows a temporal evolution described by two linear functions, intersecting around the epoch 2003. Again, the spatial region mainly affected by this PC is close to subsiding areas. This peculiar temporal evolution is likely reflecting the corrections performed in order to unify the ERS and ASAR data sets in a single time series database. It is worth noting that the second PC pattern could depend on the choice of the Ground Control Points in the InSAR filtering step. In fact, these points are generally selected over the frame border (not including areas affected by known subsidence or uplift phenomena) to better estimate orbital ramps and tropospheric artefacts. Conclusions. We processed a large number of SAR images both for the inter (pre)- seismic phase that for the post-seismic phase. Similarly the available CGPS time series has been elaborated, subtracting the CME and seasonal trends. The link between SAR retrieved displacement time series (1992-2010) with respect to the GPSs, provides results concerning the main deformation patterns. These are characterized by strong subsidence, mainly related to the underground water pumping. In addition, some feeble tectonic signals can be observed, principally localized in correspondence of the Mirandola anticline. We then performed the Principal Component Analysis (PCA) to both GPS and InSAR displacement time-series, attempting to discriminate the different contributions to the observed ground movements. The PC Analysis of the GPS time series shows how the main contribution to the displacement can be explained by a seasonal signal with an annual frequency, and by a multi-annual seasonal signal. The PC Analysis of the InSAR time series shows that the principal contributions to the displacement can be explained by a linear trend (major part of the variance) and a seasonal signal (more evident for the ASAR case). In this test area, although there are some difficulties to calibrate InSAR and GPS finely, both techniques provide interesting data concerning ground deformation patterns in time and space. However, the tectonic signals are very low with respect to the subsiding and noise ones and their interpretation could be very arduous. For the same reasons, we think that the PCA analysis is not able to highlight trends or patterns related to tectonics phenomena in this area as the involved values are very small and probably masked by larger signals. Acknowledgments. Results carried out using COSMO-SkyMed® PRODUCTS, © ASI (Italian Space Agency) - provided under license of ASI in the framework of the S3 Project “Short term earthquake prediction and preparation” (DPC-INGV, 2013-14). The Envisat images are provided by ESA (European Space Agency) under the CAT.1P 5605. This study has benefited from funding provided by the Italian Presidenza del Consiglio dei Ministri – Dipartimento della Protezione Civile (DPC). This paper does not necessarily represent DPC official opinion and policies. The InSAR processing are carried out using the Sarscape© software (Sarmap, CH). We thank Dr. Riccardi P. and Dr. Cantone A. (Sarmap, CH) for the fruitful collaboration and useful suggestions during the SAR data processing.