In order to determine the stability of the weak-motion site response and test the ability of geotechnical measurements to predict site response, empirical transfer functions and theoretical site amplification functions were compared at several sites across Turkey Flat Site-Effects Test Area near Parkfield, California. Observed site response relative to a bedrock reference site was determined for a given site by using the Fourier amplitude spectral ratio method. Two separate studies were conducted: a profile study and a site-specific study. Surface bedrock site response between two cross-valley rock stations separated by 2 km agreed to within 30%. Sensor separations of 10 and 20 m were used in the profile study at several sites spaced 100 m apart out onto the valley of the test area. Surface stiff-soil site response over a sensor separation of 10 m is reproducible to within 30%. For sensor separations of 20 m there are occasional significant variations in stiff-soil site response by a factor of 2 to 3 in the 10 to 20 Hz band. For sensor separations of 100 m there are systematic variations in stiff-soil site response by a factor of 3 to 10 in the 5- to 20-Hz band related to variations in soil thickness. Site-specific weak-motion mean spectral ratios and their standard deviations (Sd's) serve as empirical predictions of strong-motion transfer functions for the stiff-soil conditions found at Turkey Flat.
Observed transfer functions were modeled using relative amplification functions from the computer program SHAKE. A better fit to the amplitude of resonance peaks is obtained using damping values close to those measured in situ as opposed to damping values determined by laboratory tests. Inaccuracies in the standard geotechnical model for Turkey Flat (Real, 1988) lead to inaccuracies in calculated weak-motion transfer functions when compared with observations. Modeling of observed transfer functions indicates that the model fit is sensitive to 5% changes in S-wave velocity. Uncertainties of 10 to 20% in field measurements of S-wave velocity at Turkey Flat suggest that accurate and consistent prediction of weak motion may not be possible when based on field geotechnical measurements alone.