New refinements in attenuation measurements from free-oscillation and surface-wave observations

Abstract

This paper deals with improvements in the accuracy and precision of attenuation measurements in the Earth, obtained from observations of surface waves and free oscillations. We focus our study on the observation of the resonant frequencies of the Earth and propose an improved method for deriving attenuation data from the free-oscillation amplitude decays that reduces uncertainties. We test our method on different theoretical seismograms computed for two different Earth models, a spherically symmetric one (the PREM model) and a laterally heterogeneous one (the M84A model). We compute seismograms for: the spheroidal fundamental mode only, the fundamental mode and overtones, without noise and with actual noise. Some biases related to the method are systematically found and corrected. Our conclusion is that if we fulfill some conditions in attenuation measurements by taking into account different parameters, such as the observed resonance period, the relative amplitude of the peak for the particular great-circle path considered, the noise level of the component in the station considered, it is possible to reduce the uncertainty in amplitude measurements. Doing this, we are more confident in free-oscillation measurements than in those of surface waves. It is evident that the presence of noise may increase the mean Q inferred from long-time series if no care is taken relative to the signal-to-noise level. The method is then systematically applied to a dataset of about one thousand vertical seismograms provided by the GEOSCOPE network after various large earthquakes, which occurred from 1982 to 1995. We determine reliable estimates of the mean frequency and attenuation of the fundamental spheroidal mode for the angular order range l=21–51 (period range 175–336 s). Some results are discarded for reasons explained. We compare the consistency of this new dataset with previously published Q observations and commonly used models (PREM, Q M1, Q L6). We point out the difficulties of obtaining reliable Q factors and show the discrepancies between surface waves and normal modes measurements. It is noted that surface wave attenuation measurements may be strongly affected by the time window used in order to extract each particular surface wave train, but fit normal modes results, in some cases.