A distance profile-based window-rejection algorithm for microtremor HVSR measurements

Microtremor horizontal-to-vertical spectral ratio (HVSR) analysis is widely used to identify the predominant frequency of a site. However, HVSR curves calculated from microtremor records can be affected by non-stationary ambient vibrations, transient disturbances, and varying measurement conditions. Therefore, time-window selection is a critical step for improving the stability and reliability of HVSR analysis. This paper proposes a distance profile window-rejection algorithm (DPWA). The method quantifies the deviation of each window from the current mean HVSR curve by jointly considering the peak-frequency shift distance and spectral-shape distance. Windows with large combined distance profile values are then identified as anomalous and rejected iteratively. A weight-sensitivity analysis indicates that ω 1 = 0.5 and ω 2 = 0.5 provide a balanced trade-off between peak-frequency stability and spectral-shape consistency. For the representative CentrePort dataset, compared with the unprocessed algorithm (UA), DPWA reduces the log-standard deviation of peak frequencies and the standard deviation of peak amplitudes by 33.6% and 38.2%, respectively. Compared with the frequency-domain window-rejection algorithm (FDWA), DPWA further reduces the peak-amplitude standard deviation by 31.7%, indicating a more balanced control of peak-frequency stability and spectral-shape consistency. Finally, 81 archived microtremor datasets from CentrePort in Wellington, New Zealand, are used to further assess the robustness of DPWA. Across these datasets, both peak-frequency and peak-amplitude dispersion decrease after DPWA screening, further supporting its robustness for HVSR time-window selection within the evaluated dataset.