GNGTS 2015 - Atti del 34° Convegno Nazionale

26 GNGTS 2015 S essione 2.1 CGPS data. Concerning the continuous GPS data analysis, the spatially correlated Common Mode Error (CME) is one of the major noise sources in GPS position time series (Wdowinsky et al. , 1997; Dong et al. , 2006) and it has been recently recognized and accounted by Serpelloni et al. (2013) to improve the accuracies of GPS measurements of vertical ground motion rates. In order to obtain an improvement in the daily repeatability of GPS positions, and an improvement in the signal-to-noise ratio, we adopted a PCA (Principal Component Analysis) method to estimate and remove the spatially correlated Common Mode Noise error in GPS time-series, following the procedure described by Dong et al. (2006) and by Serpelloni et al. (2013). This step required to enlarge the number of GPS stations included in the analysis, because the CME in the Euro-Mediterranean area has been described as non-spatially uniform (Serpelloni et al. , 2013). Thus, we used state-of-the-art GPS data processing methods to derive a uniform and homogeneous GPS solution over the Euro-Mediterranean area, analyzing data from >2500 continuous GPS stations for the 1994/01/01 – 2015/05/26 time span. Additionally, in order to avoid the cancelation of interesting tectonic signals from the study area, the CME was estimated using a wider network, including tectonically stable regions in central Europe. We performed this analysis on residual time-series of high-quality GPS stations, after removing the linear trend, jumps and seasonal signals, excluding from the analysis the investigated area and all sites with clear non-linear or noisy data. The rawGPS data, in the form of daily, 30-seconds sampling, RINEX (Receiver INdependent EXchange) files, were analyzed with the GAMIT/GLOBK software. Stations velocities were obtained following a three-step procedure, consisting in: 1) the phase data reduction, 2) the combinations of solutions and realization of the reference frame, and 3) the time-series analysis. In the first step, we used daily GPS phase observations to estimate site position, adjustments to satellite orbital parameters (EOPs) and time-variable piecewise linear zenith and horizontal gradient tropospheric delay parameters by means of the GAMIT (V10.5) software, applying loose constraints to the geodetic parameters, using IGS final orbits. GPS phase data are weighted according to an elevation-angle dependent error model, using an iterative analysis procedure whereby the elevation dependence is determined from the observed scatter of phase residuals. We applied the Ocean-loading and a pole-tide correction model FES2004. We used the Vienna Mapping Function (VMF1) for modeling the tropospheric delay and the Global Pressure and Temperature 2 (GPT2) model. We also accounted for non-tidal atmospheric loading in the phase data processing step, and for 2nd/3rd order ionosphere effects. We used the IGS absolute antenna phase center model for both satellite and ground-based antennas. In the second step we used the ST_FILTER program of the QOCA software (http://qoca.jpl.nasa.gov) to combine the regional daily loosely constrained solutions obtained in Step 1 and simultaneously realize a global reference frame by applying generalized constraints. Specifically, we defined the reference frame by minimizing the velocities of the IGb08 global core stations (https://igscb. jpl.nasa.gov/network/refframe_core.html), while estimating a seven-parameter transformation with respect to the IGS realization of the ITRF2008 NNR frame, named IGb08 (https://igscb.jpl . nasa.gov). In the third step, we analysed the position time-series, defined in the IGb08 reference frame. Changes in stations positions were modelled using standard functional model. Position outliers were cleaned from the time-series adopting a post-fit RMS criterion (values larger than 3 times the post-fit Weighted Root Mean Square, WRMS, were discharged). A few stations recorded the coseismic offsets and postseismic transients due to the May 2012 Emilia earthquake sequence. For those sites, we modeled two coseismic offsets and an exponential decay function, with a 30-day decay constant. Non-tectonic jumps, mainly due to changes in the stations equipment, were defined from the analysis of station log-files, when available, and from visual inspections of the time series. Outliers were cleaned adopting a post-fit RMS criterion, in particular, we discarded values larger than 3 times the post-fit Weighted Root Mean Square (WRMS).