A study of horizontal-to-vertical component spectral ratio as a proxy for site classification in central Asia

Abstract

SUMMARY Tien Shan of central Asia is known as one of the world's largest, youngest and most active intracontinental orogens. In this study, we implemented the horizontal-to-vertical spectral ratio (HVSR) technique as a widely used first-order approximation of the site effect parameters (i.e. fundamental frequency and site amplification). A set of data including 2119 strong-motion recordings from 468 earthquakes with hypocentral distances up to 500 km and small to moderate moment magnitudes ($ {M_{\rm{w}}}\sim $3.0–5.5) recorded by 24 broad-band stations from five different networks, located in Afghanistan, Tajikistan and Kyrgyzstan was deployed to investigate site-specific characteristics. We fitted a Gaussian-shape pulse function to evaluate fundamental frequencies and site amplifications. The HVSRs analysis revealed that although the majority of the stations (16 out of 24) show flat amplification functions, there are few stations with single sharp amplification functions. Then, we classified the stations based on the predominant frequency. Furthermore, we approximated the time-averaged shear wave velocity in the uppermost 30 m (${V_{{\rm{S}}30}}$) using the fundamental frequency and its corresponding amplitude. Moreover, we compared the HVSRs obtained from P waves, S waves, coda and pre-event noise. All peak frequencies including the fundamental frequency estimated from different seismic phases are in good agreement; whereas generally, the amplitude of the P-wave window is the lowest, the amplitudes of the S wave and noise windows are similar to the whole record and the amplitudes of early and late coda windows are the highest. We also observed that the HVSRs of noise using a 5 s window may have anomalous high amplitudes and peaks. These anomalous high amplitudes and peaks in the noise HVSRs indicate the existence of some unnatural sources or artefacts such as traffic and wind with specific resonance frequencies, suggesting 5 s ambient noise window is insufficient to capture site characteristics. Finally, to assess the reliability of the determined geotechnical results, we implemented a blind theoretical HVSR inversion to obtain representative shear wave velocity profiles as well as ${V_{{\rm{S}}30}}$ along with associated uncertainties for stations characterized by a single-peak HVSR curve using a Bayesian statistical framework.