Abstract
The basis of this study is the acceleration, velocity, and displacement wave-forms of the Loma Prieta earthquake (18 October 1989; M = 7.0) at two rock sites in San Francisco, a rock site on Yerba Buena Island, an artificial-fill site on Treasure Island, and three sites in Oakland underlain by thick sections of poorly consolidated Pleistocene sediments. The waveforms at the three rock sites display a strong coherence, as do the three sedimentary sites in Oakland. The duration of strong motion at the rock sites is very brief, suggestive of an unusually short source duration for an earthquake of this size, while the records in Oakland show strong amplification effects due to site geology. The S-wave group at Treasure Island is phase coherent with the Oakland records, but at somewhat diminished amplitudes, until the steps in acceleration at approximately 15 sec, apparently signaling the onset of liquefaction. All seven records clearly show shear-wave first motion opposite to that expected for the mainshock radiation pattern and peak amplitudes greater than expected for sites at these distances (95 ± 3 km) from an earthquake of this magnitude.
While the association between these ground motion records and related damage patterns in nearby areas has been easily and eagerly accepted by seismological and engineering observers of them, we have had some difficulty in making such relationships quantitative or even just clear. The three Oakland records, from sites that form a nearly equilateral triangle about the Cypress Street viaduct collapse, are dominated by a long-period resonance (≃ 1 1/2-sec period) far removed from the natural frequency of the structure to transverse motion (2.5 Hz) or from high-frequency amplification bands observed in aftershock studies. A spectral ratio arbiter of this discrepancy confuses it further. The failure of the East Bay crossing of the San Francisco-Oakland Bay Bridge cannot be attributed to relative displacements of the abutments in Oakland and Yerba Buena Island, but the motions of the Bay Bridge causing failure remain unknown. The steps in acceleration at Treasure Island present unusual strong-motion accelerogram processing problems, and modeling suggests that the velocity and displacement waveforms are contaminated by a spurious response of the filtering operations to the acceleration steps. A variety of coincidences suggests that the Treasure island accelerogram is the most likely strong-motion surrogate for the filled areas of the Marina District, for which no mainshock records are available, but the relative contributions of bad ground, poor construction and truly strong ground motion to damage in the Marina District will never by known in any quantitative way. The principal lesson of all of this is that until a concerted effort is mounted to instrument ground and structures that are likely to fail during earthquakes, our understanding of the very complex relationships between strong ground motion and earthquake damage will, in general, remain rudimentary, imprecise, and vague.
Site correction of a high-frequency strong-ground-motion simulation based on an empirical transfer function