Extracting Site Effects From P‐wave spectra: Application of Cluster‐Event Method 2 (CEM2) in the Tohoku Region, Japan

Abstract Theories related to the Earth's attenuation effect and earthquake source properties predict a similar spectral falloff of body waves, creating severe trade‐off problems when determining these two processes from waveform spectra. Previous efforts attempted to suppress this trade‐off by either applying an empirical Green's function or conducting linear inversion with a specially designed event–station configuration. However, both approaches encounter difficulties when dealing with large datasets or site effect contamination. In this study, we modify a previously used cluster‐event method (CEM), combining both the spectral amplitude and the spectral ratio of the P‐waves for the clustering of events (CEM2); we then apply CEM2 to the High‐Sensitivity Seismograph Network data in the Tohoku region of Japan. An initial inversion with CEM2 yields the first estimates of the source and path parameters, that is, the corner frequency ( f c ) and the quality factor ( Q ). The residuals of this inversion demonstrate distinct spectral characteristics that are well correlated with vertical‐to‐horizontal (V/H) spectral ratios of the P‐coda waves at each station, suggesting that they represent a form of site response unaccounted for by theoretical predictions of Q and f c . After removing this particular site response, the second inversion yields an improved determination of Q . We hypothesize that this site effect is dominated by P to SV conversion upon P‐wave arrival and develops and amplifies toward the coda waves. Furthermore, the site effect demonstrates systematic regional characteristics, some of which correlate well with those derived from the V/H spectral ratios of the P‐coda waves.