Improving deformation models by discounting transient signals in geodetic data: 2. Geodetic data, stress directions, and long-term strain rates in Italy

The lithosphere of Italy is exposed to a number of different short-term strain transients, including but not limited to landslides, postseismic relaxation, and volcanic inflation/deflation. These transients affect GPS velocities and complicate the assessment of the long-term tectonic component of the surface deformation. In a companion paper we present a method for anticipating the principal patterns of nontectonic, short-term strains and building this information into the covariance matrix of the geodetic velocities. In this work we apply this method to Italian GPS velocities to build an augmented covariance matrix that characterizes all expected discrepancies between short- and long-term velocities. We find that formal uncertainties usually reported for GPS measurements are smaller than the variability of the same benchmarks across a geologic time span. Furthermore, we include in our modeling the azimuths of most compressive horizontal principal stresses (SHmax) because GPS data cannot resolve the active kinematics of coastal and offshore areas. We find that the final tectonic model can be made relatively insensitive to short-term interfering processes if the augmented covariance matrix and SHmax data records are used in the objective function. This results in a preferred neotectonic model that is also in closer agreement with independent geologic and seismological constraints and has the advantage of reducing short-term biases in forecasts of long-term seismicity.