The Distribution of Seismic Velocities and Attenuation in the Earth

Estimates of the radial distribution of seismic velocities and density and of seismic attenuation within the earth are obtained through inversion of body wave, surface wave, and normal mode data. The effect of attenuation related dispersion on gross earth structure, and on the reliability of eigenperiod identifications is discussed. The travel time baseline discrepancies between body waves and free oscillation models are examined and largely resolved. As preliminary steps in this study, a technique is developed for determining S wave arrival times and applied to records from several large nuclear explosions. The resulting low-scatter travel times are combined with other high resolution body wave results to help define a gross earth model, designated C2, which fits 86% of the normal mode data to within their 95% confidence limits. The second stage considers the effect of attenuation on seismic dispersion and shows the perturbation of phase velocity to be approximately an order of magnitude greater than the observational error. Inclusion of an attenuation correction in the normal mode data and subsequent inversion results in an elimination of the baseline dis­crepancies. The final portion of this research covers the inversion of all available seismic Q data to obtain a better estimate of the radial distribution of seismic absorption in the earth. Prominent features of the resulting Q models, designated SL1 and SL2, are low Q zones in both the upper mantle and in the 150 kilometers of the mantle just above the core-mantle boundary and finite compressional dissipation in the inner core. Model SL1 is used to compute the attenuation corrections for the normal mode data for a final inversion for seismic veloc­ities and density. The resulting attenuation-corrected earth model, QM3, fits the corrected observations to the same precision as model C2 fits the raw data. Moreover, QM3 represents a better match to the travel time data than previous earth models. The reliability of existing eigenperiod identifications in light of excitation criteria and computed attenuation is examined. For completeness, an appendix is included in which the relative excitations of a large set of spheroidal modes (T ≥ 45 sec, ℓ ≤ 150, n ≤ 30) and toroidal modes (T ≥ 45 sec, ℓ ≤ 150, n ≤ 7) is presented for both the Alaska earthquake (1964) and the Columbia earthquake (1970) sources.