A Zeta Model for the Frequency-Dependent Decay of the Fourier Amplitude Spectrum of Acceleration at High Frequencies

Session: Near-Surface Effects: Advances in Site Response Estimation and Its Applications
Type: Oral
Date: 4/28/2020
Time: 05:00 PM
Room: 110 + 140
Description: 

Fourier amplitudes of acceleration rapidly decrease at frequencies beyond the source corner frequency, mainly due to attenuation along the propagation path and below the study site. This decay is commonly modeled with a single parameter kappa introduced by Anderson and Hough (1984). The "standard model" to explain many observed features of kappa is with attenuation, characterized by the quality factor Q, that is frequency independent, combined with an omega-square source and amplification that is on average flat at high frequencies. This standard model does not explain many recent studies in which the observed high-frequency slope is not exactly linear in log-linear space but curved, resulting in estimated values of kappa that strongly depend on the chosen frequency band of analysis. At the same time, many seismological studies of high-frequency attenuation have found that Q is frequency dependent.

We explore the possibility of substituting a frequency-dependent Q model into the "standard model" to explain the dependence of kappa on the frequency band. The result is an alternative to the original kappa model of Anderson and Hough (1984). The spectral slope in log-linear space is then described by two parameters instead of the single parameter kappa. We refer to this new approach as the "zeta" model for the high frequency spectral shape. Equivalent to the kappa model, one of the model parameters can be split into a path-dependent and a path-independent part to account for different effects between source and site.

We apply the zeta model to vertical array data of the Euroseistest site in Greece. A comparison with the "standard model" shows consistent results if the same frequency band of analysis is considered. Yet, the new model is better able to capture the non-linear spectral decay of the data in terms of misfit reduction and a better visual match between the data and the model.

Presenting Author: Annabel Haendel

Authors