Kinematic Inversion of Large Relative Motion Datasets Derived from Combined Geodetic Velocity Fields: an Application to the Mediterranean Region
MetadataShow full item record
Previous applications of the Spakman-Nyst method to inverse problems of slow regional crustal deformation have been technically restricted to tens or hundreds of geodetic velocities within a single observation network. Here, we test the relative motion inversion technique on thousands of GPS observations compiled from tens of different studies. Our case study is the Alpine-Mediterranean plate boundary zone, for which 8969 station vectors from 45 papers are merged into one lateral velocity field. To avoid pairing all GPS data through ~40 million integration paths, we utilize a new dartboard site connection algorithm that allows for decreasing the number of relative motion observations to a few million without impairing model quality. The model parametrization coupling the corresponding integral equations comprises a possible maximum of 4*691 and 2*98 unknown velocity gradient tensor and surface fault slip components, respectively, at 584 model nodes spanning 1111 spherical triangles. Large triangles characterize data-deficient areas where we constrain the inverse problem with synthetic data and strain rate damping. Subsequently, 43 GPS datasets are corrected for network Euler rotations (i.e. 3*43 unknowns) relative to a chosen reference network. This enables us to average co-located observations into single vectors, which reduces the computation time of the forward problem, and to estimate rigid Euler velocities from four datasets enclosed by plate boundary faults that outline the Adriatic, Aegean, Anatolian, and Calabrian sub-regions in the model. Finally, we present a new approach to investigating intraplate deformation by comparing inversion of the geodetic observations with inverting the Euler motions at all data sites from the four sub-regions and major African, Arabian, and Eurasian plates. We find that inversion for the velocity gradient field and inactive or active fault creep results in the largest kinematic differences between the GPS and Euler motion inversions, reflecting internal deformation. The two inversion types show a significant fault-local data misfit especially in the Eastern Mediterranean Sea, being related to trade-offs between almost perfectly resolved velocity gradient and fault slip parameters where observations near faults are lacking. Our main conclusion is that the Spakman-Nyst method and its implementation can be efficiently applied to thousands of geodetic data points. We suggest to explore the trade-off complication before making further model interpretations.