Abstract Ground response analyses (GRA) model the vertical propagations of SH waves through flat-layered media (1DSH) and are widely carried out to evaluate local site effects in practice. Horizontal-to-vertical spectral ratio (HVSR) technique is a cost-effective approach to extract certain site-specific information, e.g., site fundamental frequency (f0), but HVSR values cannot be directly used to approximate the levels of S-wave amplifications. Motivated by the work of Kawase et al. (2019), we propose a procedure to correct earthquake HVSR amplitudes for direct amplification estimations. The empirical correction compensates HVSR by generic vertical amplification spectra categorized by the vertical fundamental frequency (f0v) via k-means clustering. In this investigation, we evaluate the effectiveness of the corrected HVSR in approximating observed linear amplifications in comparison with 1DSH modellings. We select a total of 90 KiK-net (Kiban Kyoshin network) surface-downhole sites which are found to have no velocity contrasts below their boreholes and thus of which surface-to-borehole spectral ratios (SBSRs) can be taken as their empirical transfer functions (ETFs). 1DSH-based theoretical transfer functions (TTFs) are computed in the linear domain considering uncertainties in VS profiles through randomizations. Five goodness-of-fit metrics are adopted to gauge the closeness between observed (ETF) and predicted (i.e., TTF and corrected HVSR) amplifications in both amplitude and spectral shape over frequencies from f0 to 25 Hz. We find that the empirical correction to HVSR is highly effective and achieves a “good match” in both spectral shape and amplitude at the majority of the 90 KiK-net sites, as opposed to less than one-third for the 1DSH modelling. In addition, the empirical correction does not require a velocity model, which GRAs require, and thus has great potentials in seismic hazard assessments.

中文翻译：

---

评估地震 HVSR 在特定地点放大估计中的新应用

摘要 地面响应分析 (GRA) 模拟 SH 波通过平坦层状介质 (1DSH) 的垂直传播，并在实践中广泛用于评估局部场地效应。水平垂直频谱比 (HVSR) 技术是一种提取特定站点信息（例如站点基频 (f0)）的经济高效的方法，但 HVSR 值不能直接用于近似 S 波放大的水平. 受到 Kawase 等人的工作的启发。(2019)，我们提出了一种程序来校正地震 HVSR 振幅以进行直接放大估计。经验校正通过通过 k 均值聚类按垂直基频 (f0v) 分类的通用垂直放大谱来补偿 HVSR。在这次调查中，与 1DSH 建模相比，我们评估了校正后的 HVSR 在近似观察到的线性放大方面的有效性。我们总共选择了 90 个 KiK-net（Kiban Kyoshin 网络）地表井下站点，这些站点被发现在其钻孔下方没有速度对比，因此可以将其中的地表与钻孔光谱比 (SBSR) 作为其经验传递功能（ETF）。基于 1DSH 的理论传递函数 (TTF) 是在线性域中计算的，考虑到 VS 剖面中的不确定性，通过随机化。采用五个拟合优度指标来衡量观察 (ETF) 和预测 (即 TTF 和校正的 HVSR) 放大在 f0 到 25 Hz 频率范围内的幅度和频谱形状之间的接近程度。我们发现对 HVSR 的经验校正非常有效，并且在 90 个 KiK-net 站点中的大多数站点上实现了光谱形状和幅度的“良好匹配”，而 1DSH 建模则不到三分之一。此外，经验校正不需要GRAs需要的速度模型，因此在地震危险性评估中具有很大的潜力。