Site response at Yerba Buena Island, San Francisco Bay, California analyzed with weak motion recordings

We utilize weak motion recordings to evaluate the site response of Yerba Buena Island (YBI). YBI is a rock outcrop in San Francisco Bay, California, and is often used as a reference site in research and engineering studies. We identified 18 earthquakes in the area for which weak motion was recorded at the bottom and top of a 61-m borehole. Examining this data, we found that: (1) near-surface rock above the downhole site does not have a significant site response of its own; (2) there is a regional effect on seismic energy that severely attenuates frequencies above about 2 Hz. We suspect that this is due to propagation through the highly heterogeneous basement Franciscan formation. (3) Plots of the spectra for most of the recorded earthquakes show evidence for a spectral hole in the downhole recordings between 8 and 11 Hz as compared to the uphole recordings. We attribute this to interference between the up and downgoing waves. The downhole recordings would therefore cause a bias if used as a reference site without consideration of this interference. From these three observations, we conclude that at YBI the surface site provides the best recordings for reference rock sites or input into soils or engineering models. However, the effect of a regional site effect causing near-constant corner frequencies for recordings of M<4.0 earthquakes means that there is no true reference site, flat in spectral amplifications, available in the region. In this paper, we also compare site response and transfer function calculations using an auto-regressive moving average (ARMA) transfer function, and the horizontal-average, cross-spectra, complex signal, and horizontal-to-vertical spectral ratio methods. We found that the complex signal spectral ratio and ARMA transfer function calculations were more stable as both site response calculations and as transfer functions. The horizontal-to-vertical method best revealed the lack of site response at YBI, and the calculation with just background noise gave as good results as that with seismic signal.