| The inference of Earth structure by seismic tomography has become a focus of intensive research efforts over the past decade. However, some fundamental questions regarding both the tomographic process and the nature of the 3-D velocity structure of the mantle are still largely unanswered. The feasibility of a tomographic inversion is controlled by ray illumination, corrections of traveltime data, the way to formulate the inversion system, and inversion methods. These issues are tackled in this dissertation on a data base of nearly 3 million teleseismic P-wave rays from the International Seismological Centre catalogs.; We first evaluate station static corrections for 1343 stations worldwide. Patterns of the zero-degree statics correlate well with surface tectonic features and with previous results. At most stations over subduction zones the slowest direction of the second-order statics is perpendicular to the trench.; The reliability of cell inversion is then assessed through simulations using realistic raypaths but synthetic traveltimes. The simulations show that cell inversion is reliable at well-covered cells and the low-degree components are better resolved by cell inversion than by spherical harmonic inversion. We propose to replace the mathematical rays used in conventional traveltime tomography by the Fresnel volume rays.; Finally, whole-mantle inversions are conducted using both the conventional rays and Fresnel-volume rays for over 1.6 million selected P-wave arrivals. Detailed analysis reveals that the Fresnel-volume ray inversion is more robust. Slowness anomalies in the uppermost mantle correlate well with surface features and the results of previous studies. Our model indicates that lateral heterogeneities in the mantle may not be dominated by very long-wavelength anomalies as suggested previously. The fast-velocity anomalies beneath most shield regions persist to the upper/lower mantle boundary, and slow anomalies appear below many ridges. Most high-velocity slabs become stagnant in the transition zone, while lower-mantle slab penetration is inferred below the Java trench and Central to South America. The overall pattern of the model suggests that mantle geodynamics may be in a mixture of layered convection and whole mantle convection. |
