Local to Regional Time-Frequency Relations, Attenuation, and Kappas for S Waves in Southeastern Iran

Abstract:

Most empirical models of distance- (travel-time) and frequency dependencies of ground-motion amplitudes do not accurately fit the data at local and regional distances simultaneously. The reasons for this inaccuracy are in the predominance of recordings at regional distances and model parameterizations based on the frequency-dependent Q. To overcome this problem, we propose an empirical model based on a more detailed form of geometrical spreading, frequency-independent Q, and joint inversion for all parameters including source and receiver coupling, and high-frequency spectral parameters kappa. In addition to data fitting, additional constraints are used to control model artifacts such as spurious correlations of data residuals with source-receiver distances and frequencies. The model is applied to transverse-component and orientation-independent horizontal amplitudes of body S waves in southeastern Iran. The resulting data residuals are somewhat smaller than obtained by conventional methods, and the model fits both the local and regional ranges. For transverse-component spectral amplitudes, the geometrical spreading with distance d is much steeper than usually assumed and proportional to about d^–1.8 for d < 90 km and d^–2.5 for d > 115 km. Between these distances, S-wave amplitudes increase by a factor of r ~ 3, which is likely due to Moho reflections and scattering within the crustal waveguide. Kappa values of 30 to 50 ms and site and spectral seismic responses are also inverted for. With the more accurate geometric-spreading and kappa models, the inverted Q-factor for the crust exceeds 2000. Estimation of model uncertainties by bootstrapping and principal-component analysis shows that the principal trade-off consists in positive correlations between kappas and the exponents of geometrical spreading at local and regional distances, and also their anti-correlation with path attenuation (1/Q). The second principal component shows a positive trade-off between the regional-range geometrical-spreading exponent and parameter r mentioned above. The general conclusions (steep geometrical spreading, amplification r, high Q, and trade-offs) were also tested on a large synthetic dataset and appear to be applicable to other areas around the world.