Measurements of Mantle Wave Velocities and Inversion for Lateral Heterogeneity and Anisotropy 1 . Analysis of Great Circle Phase Velocities

Long-period (100-330 s) fundamental-mode Love and Rayleigh waves have been processed to measure the great circle phase velocities for about 200 and 250 paths, respectively. The observations are inverted for regionalized phase velocities and for an even-order harmonic expansion of the lateral velocity heterogeneity. The regionalized inversions achieve a maximum variance reduction of about 65% and 85% for the Love and the Rayleigh wave data, respectively. The Im., = 2 inversions give a maximum variance reduction of about 60% and 90% for Love and Rayleigh waves, respectively. The /max = 8 inversion does not 1?ake a large improvement in the fit. The Love wave phase velocities have more power in l = 4 and 6, relative to l = 2, than the Rayleigh waves. For both Love and Rayleigh wave data the sectoral component dominates the I = 2 harmonics, and this component is stable if we increase lmax from 2 to 6. Heat flow also has strong sectoral components (Im = 22), which are approximately in phase with those of the phase velocities. The l = 2 harmonics of the nonhydrostatic geoid are dominated by large zonal (Im = 20) and m.oderate sectoral components. The sectoral components are in phase with those of the phase velocttles. The sectoral pattern of heat flow and phase velocity is controlled by high heat flow-low velocity of the East Pacific Rise and western North America, which is reinforced by low velocities in the antipodal region (Red Sea-Gulf of Aden-East African Rift). By contrast the geoid l = 2 pattern is dominated by geoid highs over the western Pacific subduction zones. A spherical harmonic expansion of regionalized phase velocities shows that they have l = 2 variations similar to those of the l = 2 nonregionalized inversions. This means that the regionalization approach is appropriate as a first ~~p for studying lateral heterogeneity of the earth. However, the great circle phase velocities are not sufficient by themselves to umquely locate the lateral heterogeneity. The same is true for free oscillation data. Regions of convergence have the interesting property of being slow for short-period waves and fast, faster than shields, for long-period waves.