Abstract
In an attempt to estimate the intensity of ground motion that the Marina District, San Francisco, California, experienced during the 1989 Loma Prieta earthquake, we investigate the 3D seismic response of a 2D model (referred to in the present article as ``2.5D model'') of a SW-NE trending cross section of the Marina Basin. As a first step in this endeavor, the simulated elastic response characteristics of the model are compared with recorded aftershock data. The comparison, in terms of peak amplitudes, duration, and frequency content of the time response, is favorable. In simulating the response of the Marina Basin to the Loma Prieta mainshock excitation, we account for the effect of soil nonlinearities by an iterative procedure, referred to as the ``equivalent linear approach'' according to which the values of soil damping and stiffness are selected to be consistent with the level of strain. Our results show that accelerations and velocities may have reached values as high as 0.23 g and 34 cm/sec, respectively; strains induced by wave propagation were of the order of 10-4, while spectral acceleration Sa for damping ratio 5% reached values as high as 0.8 g for periods in the range of 0.8 
T 1.2 sec, which contains the fundamental frequencies of the most heavily damaged structures in the Marina District. The simulations confirm the conjecture made by Hanks and Brady (1991) that the motion recorded at Treasure Island is the most likely strong-motion surrogate for the filled areas of the Marina District.
Based on the results of the simulations, it may be stated that for strong-motion (i.e., large strain) excitation, 3D focusing and lateral interferences, while still present, are not as prominent as in the weak-motion (i.e., small-strain) excitation case. The above conclusion suggests that, in general, the damping characteristics of soil deposits (particularly of poorly consolidated soft-soil deposits, i.e., mean shear-wave velocity of the upper 30 m of deposits, less than 200 m/sec), selected to be consistent with the level of strain induced by the seismic excitation, are a key factor in controlling the nature of the overall response of a sedimentary basin. Finally, in computing empirical amplification ratios based on recorded motions, selection of an appropriate ``free-field'' motion, representative of the incident excitation, is very crucial.