Attenuation coefficients of Rayleigh and Lg waves

Abstract

Analysis of the frequency dependence of the attenuation coefficient leads to significant changes in interpretation of seismic attenuation data. Here, several published surface-wave attenuation studies are revisited from a uniform viewpoint of the temporal attenuation coefficient, denoted by χ. Theoretically, χ( f) is expected to be linear in frequency, with a generally non-zero intercept γ = χ(0) related to the variations of geometrical spreading, and slope dχ/df = π/Qe caused by the effective attenuation of the medium. This phenomenological model allows a simple classification of χ( f) dependences as combinations of linear segments within several frequency bands. Such linear patterns are indeed observed for Rayleigh waves at 500–100-s and 100–10-s periods, and also for Lg from ~2 s to ~1.5 Hz. The Lgχ( f) branch overlaps with similar linear branches of body, Pn, and coda waves, which were described earlier and extend to ~100 Hz. For surface waves shorter than ~100 s, γ values recorded in areas of stable and active tectonics are separated by the levels of \(\gamma _{D} \approx 0.2 \times 10^{-3}\) s − 1 (for Rayleigh waves) and 8 ×10 − 3 s − 1 (for Lg). The recently recognized discrepancy between the values of Q measured from long-period surface waves and normal-mode oscillations could also be explained by a slight positive bias in the geometrical spreading of surface waves. Similarly to the apparent χ, the corresponding linear variation with frequency is inferred for the intrinsic attenuation coefficient, χi, which combines the effects of geometrical spreading and dissipation within the medium. Frequency-dependent rheological or scattering Q is not required for explaining any of the attenuation observations considered in this study. The often-interpreted increase of Q with frequency may be apparent and caused by using the Q-based model of attenuation and following preferred Q( f) dependences while ignoring the true χ( f) trends within the individual frequency bands.