Comparison of inversion methods in seismic tomography: application to tectonic structures in Northwestern Italy

In this paper we apply various inversion methods to a set of teleseismic data collected by a network operating along the Ligurian Belt in the transition region between Alps and Apennines. In particular, we consider the regularization method, the truncated singular value decomposition, the Landweber method (with the Related Simultaneous Iterative Reconstruction Technique) and the conjugate gradient method. All the methods provide rather similar velocity models which are well approximated by that provided by back-projection (used with an appropriate normalization constant). A drawback of these models seems to be the large discrepancy (of the order of 40%) between the observed time residuals and those computed from the model itself. However, for each station of the network, the azimuth dependence of the computed time residuals reproduces rather well the observed one so that it is believable that the most significant information contained in the data has been expIoited. The computed velocity models indicate strong heterogeneities in the first 200 km below the Apennines.