Great Circle Rayleigh and Love Wave Dispersion from 100 to 900 Seconds

Summary When a seismic record containing multiple sets of world‐circling surface waves generated by a major earthquake is auto‐correlated, the resulting time function is composed of groups of correlated waves whose phase delay functions are proportional to the difference between the corresponding epicentral distances. Phase information sufficient to calculate the dispersion of waves which travelled exactly one Earth circumference is available from that portion of the auto‐correlogram related to the differential distance of interest. This new method also features an enhanced signal to noise ratio because of the superposition of these correlated waves and the simultaneous cancellation of uncorrelated random noise. The auto‐correlation method has been applied to the measurement of phase and group velocities of Rayleigh and Love waves for two great circle paths characterized by similar oceanic and continental portions (two‐thirds and one‐third, respectively) and by negligible tectonic segments. For Path I, three‐component recordings were analysed for two Australian stations of the World Wide Standard Seismograph Network which were collinear with respect to the Kurile Islands earthquake of 1963 October 13. The Matsushiro Observatory vertical component recording of the 1960 May 22 Chilean shock was used for Path II. Phase velocities derived from these auto‐correlograms were compared with those derived from the spectral peaks of the corresponding periodograms. The data for the auto‐correlograms are more reliable at shorter periods, while spectral analysis leads to better results at the longer periods. The estimated error of measurement for the trigonometrically smoothed phase velocities of Rayleigh and Love waves is 0–003 km s −1 in the period range from 180 s to 450 s. The total period interval for these new measurements extends from 100 s to 900 s for both types of waves. An important test of the correctness of the entire set of data processing operations is performed by the comparison of group velocities derived from the trigonometrically smoothed free period and phase velocity data with velocities directly measured by the multiple filter technique. The discrepancies between the group velocities determined by these two methods are usually of the order of 0–02 km s −1 or less. A number of free periods of the first and second spheroidal overtones were found, using a statistical approach which involved the summation of information from the spectral analyses of five seismograms. The observed higher mode phase velocities, with the exception of the gravest orders of the first overtone, are similar to dispersion curves for a spherical non‐rotating Gutenberg‐Bullen A′ model. A number of features are common to the deviations of trigonometrically smoothed Rayleigh wave phase velocities derived from (a) the filtered auto‐correlograms, (b) the corresponding periodograms, and (c) the free periods compiled by Pekeris in 1966, when these data are subtracted from values calculated for the non‐rotating spherical Gutenberg‐Bullen A′ Earth model. The deviations are negative for T > 500 s, slightly positive for 370 < T < 500 s, and again negative for T < 370 s. A negative minimum of ∼ 0.025 km s −1 occurs for T = 250 s, and the present measurements indicate a local decrease in the negative deviations near 180 s. Rayleigh wave group velocity deviations are positive for T > 450 s, and are generally negative for shorter periods with a sharp minimum of ≃ 008 km s −1 near 325 s and a local maximum near 225 s. Deviations of the trigonometrically smoothed Love wave phase velocities are negative for T > 350 s and positive for T < 350 s, rising to approximately 0–06 km s −1 near 150 s. The corresponding group velocity deviations change from negative to positive as T becomes less than 500–600 s; they then increase continually, reaching ≃ 0.1 km s −1 near 250 s and remaining near this value for periods as short as 100 s. The deviations found for T > 225 s may well be world‐wide phenomena which indicate that revisions will be required at corresponding depths within the Earth, taking proper account of the required corrections for ellipticity and rotation discussed by Dahlen in 1968. Shear velocities lower than those of the Gutenberg distribution will be needed at depths in the vicinity of 200–500 km, but alterations of both signs may be required at other depths when all of the data are considered. Short period dispersion, group velocities, overtone measurements and body wave observations should prove useful for this task.