GNGTS 2015 - Atti del 34° Convegno Nazionale

90 GNGTS 2015 S essione 1.2 (Rebischung, 2012), on which the orbits, satellite clocks and antenna models are based. Of particularly relevance for our work is the ETRF2000 (Boucher and Altamimi, 2011), which is related at any epoch to the IGb08 by a 14 parameter Helmert transformation. The ETRF2000 frame is defined, in compliance with the EU Directive INSPIRE on Conventional Reference Systems (INSPIRE, 2009), in such a way that the velocities of GNSS sites in a stable part of Europe are minimized. Because the reference GNSS sites used for datum definition may suffer from accidental discontinuities (e.g. antenna changes, coseismic offsets and similar) it is very important for consistency to adopt in the time series analysis conventional solution numbers, as published regularly by (Kenyeres, 2014) the EPN, to avoid that sudden jumps in the time series of reference sites affect those of other GNSS sites. Likewise a careful analysis of discontinuities in the time series of all the GNSS sites is necessary to ensure that the velocity which is computed for each site represents a long term coordinate change, in the given reference frame.The visualization of the time series of the GNSS sites we process is available at http:// regegnssveneto.cisas.unipd.it/scidata and is directly linked to the .STA file used by the Bernese Software to track the history of each processed site, including coordinate discontinuities and setup of new solution numbers, edited periods and similar. The noise affecting the time series can be described by white noise and flicker phase noise (Caporali, 2003; Williams, 2003). This affects the estimate of the uncertainty in the velocities and is properly taken into account to correctly weigh the contributions of GNSS sites e.g. with different time histories. Estimating a realistic uncertainty in the velocity of each site is crucial to properly weight its contribution in the least squares collocation. From velocities to strain rates. Next we address the spatial analysis of the velocities, and in particular their horizontal gradient, or 2D strain rate. To compute a strain rate it is necessary to identify an area containing a sufficient number of GNSS sites of known velocity. Although the minimum number of sites is three, the statistical reliability of the estimates and an appropriate monitoring of the uncertainties imposes that a larger number of sites is used. In general one can expect that comparing the sites which are nearest to each other, on the length scale of the deformation, leads to neglect the contribution of other sites which have a coherent signal. On the other hand, including sites spread on a large area relative to the scale length of the deformation, will tend to attenuate the deformation signal due to the loss of coherence. Consequently, for each seismic province containing a number of GNSS sites there must exist a typical distance defining an area of maximum coherence. Because the distribution of the GNSS sites is not optimized on the fault geometries, it is necessary to examine case by case the area on which the computation of horizontal deformation can be meaningfully carried out. Following previous work (Caporali et al. , 2011), in this paper we adopt least squares collocation as an optimal algorithm to map velocity into strain rate. This algorithm requires a covariance function with zero derivative at the origin and going to zero ad infinite distance. The scale distance of the fall off needs to be determined by the data themselves, as it is a statistical indication of the coherence of their signal content. Because we are interested in deformation, we propose the use of the maximum shear strain rate (Savage and Simpson, 1996) as a coherence indicator. For example, for any given GNSS site of known velocity we compute by least squares collocation the shear strain rate using a covariance function with scale distance varying from 10 to 100 km at steps of say 10 km. We expect that the curve of the shear strain rate as a function of the scale distance is bell shaped, with the maximum at the scale distance which defines the maximum coherence. Failure to do so can be interpreted as an area subject to negligible deformation, or an area which is undergoing deformation but is occupied by GNSS sites unable to pick up the deformation. As a consequence, we discard sites for which the shear strain rate as a function of the correlation distance is not bell shaped, or such that within the optimal correlation distance there are less than three additional sites of known velocity. The analysis of the velocities is made with a Matlab program. As input a file containing