Dispersion and attenuation of mantle waves through waveform inversion

Abstract

We propose a new approach to the determination of elastic and anelastic parameters of the Earth's structure from seismic data. Instead of measuring such functionals of the medium response as the phase delay or spectral ratios, we perturb parameters of the model to satisfy the observed waveforms. The advantage of our waveform inversion technique (WIT) is that it uses the properties of the Earth as a smoothing filter. Also, because the Earth's structure is the common denominator, the method allows simultaneous interpretation of different functionals of this structure; for example, the waveforms of Rayleigh and Love waves for the same source-receiver pair.Following extensive testing of the method on synthetic and actual data, we subject results of our analysis of 37 recordings for various sources and stations to 'pure path' decomposition. The period range of analysis extends from 160 to 630s. We distinguish four types of regions: stable continents, areas tectonically active within the last 400Myr, ocean floors younger than 38 Myr, and old ocean floors older than 38 Myr.The results indicate significantly different responses of the oceanic and continental areas at long periods. Stable continental and tectonic regions have nearly the same dispersion for periods greater than 300-35Os, and then diverge rapidly, with stable continents showing higher velocities. The young and old oceans, on the other hand, become distinct at periods as long as 500 s; the old oceans are faster. In terms of shear velocity models, the data are consistent with a difference of about 4 per cent between the young and old oceans in a depth range from 400 to 670 km, while the continental regions are similar and close to the global average. At shallower depths our structures are conceptually similar to those inferred from previous 'pure path' analyses and short-period surface wave studies.While we cannot, yet, establish statistically significant differences in Q for the four regions, such differences are obtained if one distinguishes only between continents and oceans: the Q for the latter is 10-20 per cent lower in the period range of the analysis.

Absorption band Q model for the Earth

“…Period, sec means for mantle Rayleigh waves are probably accurate to better than 5% [Dziewonski and Steim, 1982;Chael and Anderson, 1982]. Most of the other data has an estimated subjective uncertainty, in the context of global means, of about 20%.…”

“…The main differences of the absorption band model, compared with the constant Q models, are the higher theoretical values for 0 S 2 , 0 S 3 , and 0 S 4 and the more rapid decrease of Q from 0 S 5 to 0 S 10 • Unfortunately, there is considerable uncertainty in the Q values for the low-order spheroidal modes. The data are compiled from Anderson and Hart [1978a, b], Chael and Anderson [1982], Nakanishi [1979a], and Dziewonski and Steim [1982]. In general, the fit of ABM is as good as or better than the constant Q models.…”

“…There have been numerous studies of the attenuation of mantle Rayleigh waves over various great circle paths, and good global average values are now available for 0 S 9 through oS76, which covers the period range from 634 to 125 s. The global means (Table 5) are probably good to better than 5% [Dziewonski and Steim, 1982;Chael and Anderson, 1982]. The absorption band model fits these data with an average error of 2.3%.…”

Measurements of mantle wave velocities and inversion for lateral heterogeneities and anisotropy: 3. Inversion

“…Nakanishi (1981) found no significant difference between young and old oceans below 300 km. Dziewonski and Steim (1982), using waveform inversion, inferred the existence of substantial shear velocity differences (0.2 km/s) between young and old oceans for depths from 400 km to 670 km. In our models, old oceans are faster than young oceans by about 0.1 km/s at the same depths, but we have seen that the a posteriori standard deviation was about as large.…”

Global maps of Rayleigh wave attenuation for periods between 40 and 150 seconds

“…ocean, shield) showed a correlation between phase velocity variations and quality factor variations and the existence of large differences between quality factors beneath continents and oceans [Ben-Menahem, 1965;Nakanishi, 1979;Mills and Hales, 1978;Dziewonski and Steim, 1982]. [Durek et al, 1988[Durek et al, , 1989[Durek et al, , 1993Romanowicz, 1990Romanowicz, , 1994Romanowicz, , 1995 led to frequency dependent quality factor maps QRayl½ign(w) and Q•:ove(w).…”