UCLA

Frequency-dependent horizontal-to-vertical spectral ratios (HVSRs) of three-component recordings can provide useful information on site resonant frequencies, which may be useful for predicting site response. Despite this potential, HVSR-conditioned site response models have not been adopted in the ground motion models most often employed for ground motion prediction in the United States and most global regions. We suggest that this stems from HVSR data resources being unavailable to support model development until recently, ongoing questions related to how HVSR should be measured and interpreted, and lack of consensus on how HVSR data should be utilized for site response prediction. Despite those challenges, we argue that the stage is set for future implementation of HVSR-conditioned models because a public dataset is now available, it is increasingly recognized that models should be conditioned on microtremor-based HVSR (mHVSR), and several modeling approaches have appeared in recent literature. This brief article focuses on site response modeling approaches, which include machine learning methods, models conditioned on mHVSR ordinates, and models conditioned on attributes of mHVSR peaks (i.e., peak frequency, amplitude, and width) or the lack of peaks.