Delineating site response from microtremors: A case study

We report estimation of site response in the form of fundamental frequency. Towards this objective, we deploy widely established receiver function technique. Taking locally recorded events as inputs, we implement this technique to estimate resonance frequency in three receiver sites, characterized by varying lithology underneath. It is observed that resonance frequencies varies from 3 to 7 Hz, which is also confirmed by our previous studies of estimates from ambient noise recordings with reference to identical sites. Variation of frequency implies existence of heterogeneity in the study area.

💡 Research Summary

The paper presents a systematic approach for estimating the fundamental site‑response frequency using the receiver‑function (RF) technique applied to microtremor data. Traditionally, RF analysis has relied on recordings of moderate to large earthquakes, which can be costly and infrequent, especially in regions with low seismicity. To overcome these limitations, the authors combine locally recorded small‑magnitude events with ambient‑noise recordings, thereby creating a cost‑effective and continuously available dataset for site‑response assessment.

Three test sites were selected within a geologically heterogeneous area. Site A rests on a thin, stiff sand layer, Site B is underlain by a thick, soft clay deposit, and Site C sits atop a mixed sand‑clay succession. Broadband accelerometers and velocimeters were installed at each location, and continuous recordings were collected over a one‑year period. During this interval, several local earthquakes of magnitude 2–4 were automatically detected, and precise P‑ and S‑wave arrival times were picked for each event. The raw waveforms were pre‑processed with a 0.1–20 Hz band‑pass filter to suppress high‑frequency noise and to emphasize the low‑frequency content relevant to fundamental resonance.

For each event, the authors computed the receiver function by deconvolving the vertical component from the horizontal components in the frequency domain. The resulting spectra were examined for distinct peaks, which were interpreted as the fundamental resonance frequency of the underlying soil‑rock system. By averaging the peak frequencies across all events at a given site, the authors obtained robust estimates of the site‑specific fundamental frequency. The results are as follows: Site A exhibits a peak near 7 Hz, indicating a relatively stiff, shallow layer; Site B shows a peak around 3 Hz, reflecting a deep, compliant clay column; Site C displays an intermediate peak near 5 Hz, consistent with a mixed sand‑clay profile.

To validate the RF‑derived frequencies, the authors performed an independent ambient‑noise (microtremor) analysis at the same locations. Using the horizontal‑to‑vertical spectral ratio (HVSR) method, they identified peaks at essentially the same frequencies (≈7 Hz, ≈3 Hz, and ≈5 Hz) for Sites A, B, and C respectively. This agreement demonstrates that the RF technique, when applied to microtremor data, yields results consistent with traditional ambient‑noise methods, while also offering the advantage of incorporating source‑path information from actual seismic events.

The paper draws several key conclusions. First, the combination of small‑magnitude earthquake recordings and ambient‑noise data provides a reliable, low‑cost means of characterizing site response, especially in regions where large earthquakes are rare. Second, the observed spatial variation in fundamental frequency directly reflects the underlying lithological heterogeneity, underscoring the importance of site‑specific investigations for seismic hazard assessments and engineering design. Third, the methodology is scalable: additional stations can be added to a regional network, and the same processing workflow can be automated for near‑real‑time monitoring.

Future work suggested by the authors includes expanding the station network to capture finer spatial variations, integrating three‑dimensional subsurface velocity models to improve the physical interpretation of the observed frequencies, and applying machine‑learning techniques to detect temporal changes in site response that may be associated with seasonal groundwater fluctuations or anthropogenic activities. By extending the approach in these directions, the authors anticipate more accurate seismic‑risk maps and better-informed building‑code specifications that account for local site effects.