Measurements of group velocities by means of band-pass filtering lead to systematic errors when group in velocity changes rapidly with frequency. A new method hereafter referred to as “residual dispersion measurement” avoids this difficulty by transforming the signal prior to the filtering. An observed seismogram is cross-correlated with a theoretical seismogram in which the dispersion approximates the observed dispersion within a few per cent. Dispersion of the resulting pulse can be measured with a much higher precision, since du/dω has been reduced by at least an order of magnitude. In addition, the method can be employed iteratively to obtain a greater precision of measurement.
The dispersion of mantle Rayleigh waves is measured from a sum of 48 auto-correlograms of the recordings of the Alaskan earthquake of March 28, 1964. Group and phase velocities are obtained for order numbers between 0S9 and 0S47. It is estimated that the absolute error of the group-velocity measurement does not exceed 0.015 km/sec.
The analysis of a sum of 13 auto-correlograms of horizontal component seismograms with predominantly transverse motion reveals that the fundamental-mode Love-wave data may be contaminated by higher torsional modes. Comparison of several sets of average Love-wave phase velocities shows a discrepancy of the order of 0.4 per cent in the period range from 350 to 170 sec. The corresponding figure for Rayleigh waves is less by at least an order of magnitude.