TRAVEL-TIME TOMOGRAPHY OF THE EUROPEAN MEDITERRANEAN MANTLE DOWN TO 1400 KM

The 3-D P-wave velocity structure of the mantle below Europe, the Medi terranean region and a part of Asia Minor is investigated. This study is a considerable extension of an earlier tomographic experiment that was limited to imaging upper-mantle structure only. Here, the Earth's volume under study encompasses the mantle to a depth of 1400 km, and w e increase the number of International Seismological Centre (ISC) data for inversion by a factor of four by taking more years of observation , and by including data from teleseismic events. The most important de parture from the earlier study is that we do not use the Jeffreys-Bull en model as a reference model, but an improved radially symmetric velo city model, the PM2 model, which is appropriate for the European-Medit erranean mantle. Our inversion procedure consists of two steps. First, the radial model PM2 is determined from the ISC delay times by a nonl inear trial-and-error inversion of the data. As opposed to the Jeffrey s-Bullen model, the new reference model has a high-velocity lithospher e, a low-velocity zone, and seismic discontinuities at depths of 400 a nd 670 km. Next, the ISC data arc corrected for effects related to the change in reference model and inverted for 3-D heterogeneity relative to the PM2 model. We follow this two-step approach to attain a better linearizable tomographic problem in which ray paths computed in the P M2 model provide a better approximation of the actual ray paths than t hose computed from the Jeffreys-Bullen model. Hence, the two-step sche me leads to a more credible application of Fermat's Principle in linea rizing the tomographic equations. Inversion results for the 3-D hetero geneity are computed for both the uncorrected ISC data and for the PM2 data. The data fit obtained in the two-step approach is slightly bett er than in the inversion of ISC data (using the Jeffreys-Bullen refere nce model). A comparison of the tomographic results demonstrates that the PM2 data inversion is to be preferred. To assess the spatial resol ution an analysis is given of hit count patterns (sampling of the mant le by ray paths) and results of sensitivity tests with 3-D synthetic v elocity models. The spatial resolution obtained varies with position i n the mantle and is studied both in map view and in cross-section. In the well-sampled regions of the mantle the spatial resolution for larg er-scale structure can (qualitatively) be denoted as reasonable to goo d, and at least sufficient to allow interpretation of larger-scale ano malies. A comparison is made of the results of this study with indepen dent models of S-velocity heterogeneity obtained in a number of invest igations, and with a prediction of the seismic velocity structure of t he mantle computed from tectonic reconstructions of the Mediterranean region. In the context of this comparison, interpretations of large-sc ale positive anomalies found in the Mediterranean upper mantle in term s of subducted lithosphere are given. Specifically addressed are subdu ction below southern Spain, below the Western Mediterranean and Italy, and below the Aegean. In the last region a slab anomaly is mapped dow n to depths of 800 km.