Frequency-Dependent Attenuation of P and S Waves in Southern California

Abstract

Accurate models of crustal attenuation structure are important for simulating seismic wavefields at high frequencies (f > 1 Hz). In this study, we collected P and S waveforms from 160 regional earthquakes (3.3 ≤ M ≤ 5.7) recorded at 218 broadband stations of Southern California Seismic Network and measured spectral amplitudes of P and S waves in the 1- to 10-Hz band using integrals over wavelet transforms. We accounted for source structure and geometrical spreading by referencing the spectral amplitudes to values computed from 1-D synthetic seismograms. We inverted the spectral amplitude ratios for a 1-D, frequency-dependent crustal attenuation model. When plotted against wavenumber, the depth averages of both QP and QS are described by the same function, Q(k) = (240 ± 20) k0.40 ± 0.05 (1 km−1 ≤ k ≤ 17 km−1). Our results are consistent with attenuation dominated by first-order Born scattering from an isotropic, self-affine distribution of elastic heterogeneities with an outer scale of ~10 km, a root-mean-square relative amplitude of ~8% and a fractal dimension of ~3.8. The inversions accounted for frequency-dependent variations in the source spectra and site response. The source spectra roll off at an average rate of about 2 for both wave types. The station residuals, which reflect unresolved lateral structure, show stronger attenuation in the Los Angeles region and weaker attenuation in the batholithic terrains to the north and south, as well as in the Proterozoic terrains on the eastern side of the study region.