S U M M A R YWe present a new 3-D transversely isotropic shear wave velocity model of the European and Mediterranean upper mantle obtained by analysis of surface waves. Data used are fundamentalmode Rayleigh and Love group velocity measurements in the period range 35-170 s, taken on seismograms recorded by European stations for regional earthquakes. The tomographic inversion to map the 3-D earth structure is split into two steps. First, we regionalize the group velocity dispersion measurements, obtaining distinct geographical group velocity maps at different periods; then, each local dispersion curve is inverted separately to find the shear wave velocity structure at depth. The inversion benefits from using a priori information from a 3-D global mantle model (S20RTS) and a new detailed European crustal model (EPcrust) to constrain the shallower layers. The inversion scheme follows a non-linear iterative algorithm by which Rayleigh and Love group slowness are inverted simultaneously for the best-fitting isotropic Voigt shear wave speed ( 2 3 v SV + 1 3 v S H ) and radial anisotropy parameter (v SH − v SV ). Final merging of the v S profiles results in a new higher resolution 3-D model of European upper mantle. We find that Western Europe and Mediterranean Sea are mainly characterized by relatively low velocities, strongly contrasting with the fast roots of the Eastern European Craton. Many regional scale structures are also evident in the model, thus providing insights into the complex geodynamic framework of the European continent. Most prominent are the low-velocity West Mediterranean spreading basins and European Cenozoic rift system, and seismically fast features connected to subduction of Adria microplate, Hellenic Arc and Calabrian Arc. Radial anisotropy does not vary very significantly with respect to the PREM profile, as available data only resolve lateral variations to a limited degree due to trade-off with velocity. EPmantle has the potential to provide a reliable seismological reference for the upper-mantle structure in the broad European region.

S U M M A R YWe present a new surface-wave tomographic study of the broad European and Mediterranean region. Our goal is to enhance the resolution of previously published group velocity models using new data from European permanent seismic networks and a dense broad-band array in Northern Apennines (RETREAT). We measure fundamental mode Rayleigh and Love wave group velocities from long-period seismograms recorded at regional distance (between 600 and 7000 km). Our measurement technique is based on iterative application of multiple filters and phase-matched filters; we accurately estimate dispersion curves for more than 1500 Rayleigh wave and about 850 Love wave paths in the period range 35-170 s. Consistency of measurements is evaluated by comparing ray clusters from sample earthquakes to closely spaced RETREAT stations. In the whole data set, measurement errors in group velocity decrease with increasing distance and show to be caused by inaccuracy in the estimate of group arrival time. We calculate maps of Love and Rayleigh group velocity at selected periods by linear tomographic inversion, accounting for group arrival time errors and evaluating a posteriori group slowness errors. Data coverage in this region is not uniform, and it is highly influenced by the uneven distribution of earthquakes and seismic stations. We therefore build a laterally heterogeneous reference model by inverting a global data set of group velocity derived from the phase velocity library of Ekström et al. (1997). Use of this reference as an a priori model during inversion improves preliminary data coverage at the borders of our study region and warrants consistency with global models. The implications of different regularization constraints (mathematically equivalent to norm damping or smoothing with different criteria) are analysed and compared. Group velocity maps confirm the large-scale geological lineaments known for the region: short-periods maps differentiate well among thinner oceanic and thicker continental crust; the most dominant feature in long-period maps is the difference between the fast Precambrian East European Platform and the low velocity signature of central Europe and western Mediterranean, separated by a sharp gradient in correspondence of the TornquistTesseyre Zone. The seismically active Tethyan Belt is clearly marked by a continuous slow anomaly. Smaller scale, possibly thermally related, low velocity anomalies are found under Iceland and Mid-Atlantic Ridge, Rhine Graben and Tyrrhenian back-arc basin, whereas the Hellenic Arc is characterized by fast velocity.

[1] Earthquake occurrence stems from a complex interaction of processes that are still partially unknown. This lack of knowledge is revealed by the different statistical distributions that have been so far proposed and by the different beliefs about the role of some key components as the tectonic setting, fault recurrence, seismic clusters, and fault interaction. Here, we explore these issues through a numerical model based on a realistic interacting fault system. We use an active fault system in central Italy responsible for moderate to large earthquakes, where geometric and kinematic parameters of each structure can be confidently assessed. Then, we generate synthetic catalogs by modeling different seismogenic processes and allowing coseismic and postseismic fault interaction. The comparison of synthetic and real seismic catalogs highlights many interesting features: (1) synthetic seismic catalogs reproduce the short-term clustering and the long-term modulation observed in the historical catalog of the last centuries; (2) a recurrent model of earthquake occurrence on faults is more effective than a Poisson model to explain such short-term and long-term time features; (3) a realistic fault pattern is a key component to generate stochasticity in the seismic catalog, preventing a systematic time ''synchronization'' of strongly coupled faults; (4) such a stochasticity may put strong limits to the forecasting capability of models based on fault interaction, even though the latter is a key component of the process. Finally, the model allows explicit predictions on future paleoseismological observations to be made.

We here exploit fundamental mode Rayleigh and Love seismic wave information and the high resolution satellite global gravity model GGM02C to obtain a 1° × 1° 3‐D image of: (a) upper‐mantle isotropic shear‐wave speeds; (b) densities; and (c) density‐vScoupling below the European plate (20°N–90°N) (40°W–70°E). The 3‐D image of the density‐vScoupling provides unprecedented detail of information on the compositional and thermal contributions to density structures. The accurate and high‐resolution crustal model allows us to compute a reliable residual topography to understand the dynamic implications of our models. The correlation between residual topography and mantle residual gravity anomalies defines three large‐scale regions where upper mantle dynamics produce surface expression: the East European Craton; the eastern side of the Arabian Plate; and the Mediterranean Basin. The effects of mantle convection are also clearly visible at: (1) the Eastern Sirt Embayment; (2) the West African Craton northern margins; (3) the volcanically active region of the Canarian Archipelago; (4) the northern edge of the Central European Volcanic Province; and (5) the Northeastern part of the Atlantic Ocean, between Greenland and Iceland. Strong connections are observed among areas of weak radial anisotropy and areas where the mantle dynamics show surface expression. Although both thermal and additional dependencies have been incorporated into the density model, convective down‐welling in the mantle below the East European Craton is required to explain the strong correlation between the estimated negative mantle residual anomalies and the negative residual topography.