Global phase velocity maps of Love and Rayleigh waves between 40 and 150 seconds

SUMMARY Although much is known of the 3-D structure of the Earth, existing models do not make use of much that is known about the large structural perturbations near the surface. It has long been known, for example, that continental and oceanic crustal structures are quite different, and that these differences are evident in the dispersion of Love and Rayleigh waves sampling continental and oceanic paths. Such differences are largest at periods of less than about 100 s. Existing global models do not adequately account for such data, and make allowances for crustal structure in a very approximate way, owing to the incompleteness of information on the global distribution of crustal parameters. As a result, variations in, for example, crustal thickness translate themselves into model artefacts extending to great depth. This can be seen as one aspect of the imperfect resolution of the existing global models. In order to construct higher resolution models of the Earth's outer shell (0-200 km depth), it is necessary to gain more precise knowledge of near-surface structure by incorporating data that have sensitivity to the details of the depth distribution of heterogeneity near the surface. As a first step we analyse a large data set of fundamental-mode Rayleigh and Love waveforms to obtain global phase-velocity maps in the period range 40–150 s. Minor and major arc phase velocities have been determined from about 24 000 digital GDSN and GEOSCOPE seismograms recorded between 1980 and 1990. In order to make such measurements in an automatic way, we have developed a method, using non-linear waveform inversion, in which velocity and amplitude, as a function of frequency, are expanded in B-splines. The waveform data are inverted for the B-spline coefficients, with the application of an explicit smoothness constraint that protects against unwanted effects, such as those due to notches in the amplitude spectra, and avoids some of the problems associated with the phase ambiguity. The cost function (which is minimized in a least-squares sense) presents many local minima, and a good initial model is needed; this is derived by integration of group velocities. The measurements made using this new technique are then used in a global inversion for phase-velocity distributions of Love and Rayleigh waves, expressed in terms of a spherical harmonic expansion. We show resulting phase-velocity maps up to degree and order 40. These maps are corrected for possible artefacts due to the truncation of the spherical harmonic expansion. We present a detailed resolution analysis which shows that global lateral resolution for surface-wave tomography is of the order of 2000 km. Love-wave phase velocities show a high correlation with known upper mantle structure at long periods and with crustal structure at shorter periods. Similarly, Rayleigh-wave phase velocities correlate well with known tectonic features, but show no clear crustal signature owing to their different sampling of the structure with depth.