SMIP 2006 Seminar

The September 28, 2004 Parkfield earthquake, arguably the best recorded earthquake ever, allows for detailed investigation of finite- source models, and their resolution. We have developed models using GPS and InSAR geodetic data, and CSMIP and USGS strong motion seismic waveforms independently and jointly. In order to focus on better resolved long wavelength features the seismic data was lowpass filtered at 0.5 Hz. We have investigated the sensitivity of the finite-source models due to station coverage, weighting and smoothing parameters, and seismic velocity structure. The resulting model is consistent with preliminary results using regional broadband and strong motion waveform data (e.g. Langbein et al., 2005; Dreger et al., 2005) as well as with higher frequency seismic waveform only inversions (e.g. Custódio et al., 2005; Liu et al., 2006). The scalar seismic moment was found to be 1.28e25 dyne cm (Mw6.0), the overall rupture length 25 km, the peak slip 58 cm, and the average rupture velocity 2.6 km/s. The rupture was predominantly unilateral to the NW with a small component to the SE. The obtained slip model was used to simulate 3D wave propagation taking into account the velocity contrast across the San Andreas Fault, and a narrow low-velocity fault zone. The results of this analysis suggest observed waveform complexity and amplitudes at stations located in the fault zone could be due to fault zone guided waves.

The 2004 Mw6 Parkfield earthquake is the last in a series of several strike-slip earthquakes that have occurred on the same fault located in a zone that marks the transition between a creeping section and a locked section of the San Andreas Fault in central California. Ground motion data recorded at a dense network of near-fault stations installed by California Geological Survey (CGS) and United States Geological Survey (USGS) are unprecedented in terms of quality and characteristics for this type of earthquake in California. Although of moderate size, the earthquake produced near-fault ground motion acceleration that exceeds predictions from empirical ground motion models. At three sites the recorded acceleration was more than 1.0 g (Shakal et al., 2005). Very large peak ground velocities of up to 83 cm/s were also recorded at both ends of the fault. On the other hand, most of the stations located very near to the fault recorded ground motion with very low acceleration and velocity. In this study we investigate the implication of the rupture kinematics and dynamics, and local site effects in the amplitude and spatial variation of the near-fault ground motion for this earthquake.

Gravity imposes a lower bound on the strength needed for stable response. Collapse spectra are plots of this strength vs. period for constant values of a parameter that characterizes gravity. The paper contains formulas for collapse spectral ordinates for near fault conditions and shows that the collapse mechanism in buildings is not statistically dependent on whether the excitation is near fault or far field. Safety against instability can be predicted using the provided expressions and results from a pushover analysis. The near fault condition is not found to be a critical consideration from an instability perspective.

An improved measure of ground motion intensity that is well correlated with structural and many kinds of nonstructural damage is presented. The proposed intensity measure is based on the peak interstory drift demand computed using a simplified continuous model that consists of a combination of a flexural beam and a shear beam. This new intensity measure accounts for the influence of higher modes and for concentrations of lateral deformation demands along the height of buildings. It is then proposed to compute this new intensity measure at all stations that recorded a seismic event in order to generate improved ShakeMaps for loss estimation and emergency response. The 2004 Parkfield event is used to illustrate both concepts.

In order to develop seismic codes that can effectively mitigate damage to wood-frame construction under seismic activity, the dynamic characteristics of wood-frame buildings must be well understood. Toward this end, this data interpretation project focuses on the dynamic behavior of low-rise wooden shearwall buildings under large seismic motions. The procedure includes determining the modal parameters and extracting hysteretic characteristics from the available records.

A Caltrans designed concrete slab bridge is located southwest of the 2004 Mw 6.0 Parkfield rupture zone. Peak horizontal accelerations of 1.0 g and absolute displacements of 4 inches were measured with six CSMIP accelerometers on the bridge, and one free-field accelerometer east of the bridge. Ground motions resulted in longitudinal soil displacements in front of the abutments and around the bent piles due to the structure swaying back and forth. Based on our displacement analyses using the measured ground motions, we found that the current Caltrans seismic design approach results in a close match with the measured bridge displacements.

Recognizing the wide variability of methods and often conflicting results of seismic response analyses used for design and construction in the early 1980’s, the California Geological Survey (CGS) established the Turkey Flat Site Effects Test Area in 1987, and within two years a blind test was conducted to predict the test area’s low-strain seismic response. Test results focused on the need to reduce uncertainties in the geotechnical parameters that drive site response codes. Fifteen years later, the array recorded the September 28, 2004 M6.0 Parkfield Earthquake at a fault-rupture distance of only 5 km. A blind test has been conducted to evaluate the ability of current practice to determine the test area’s moderate-strain seismic response. This paper provides an overview of the Turkey Flat test site, and describes the rationale for what has become an evolving blind test experiment.

A blind prediction experiment was conducted for the strong-motion data recorded at the Turkey Flat test area during the September 28, 2004 M>6.0 earthquake. The motion was predicted at several sites by 15 prediction teams, first based on the observed motion at the edge of the valley, and secondly, based on the observed motion in the rock underlying the valley. Predictions were received from geotechnical firms and researchers, both in the US and internationally. A workshop was held to preliminarily review and compare the predictions to each other and the recorded data. In general, the predictions based on the valley-edge motion exceed the observed data. Predicted peak ground acceleration at the center of the valley exceeded the observed by about 50% and predicted response spectra exceeded the observed by as much as 3–5 times at periods near 0.5 sec. In the second phase, involving predictions based on the recorded motion beneath the valley sediments, much closer results were obtained. In both phases, the predictions by different investigators were quite similar to each other. The use of nonlinear vs. equivalent-linear models did not significantly improve the predictions for this stiff-soil, relatively low strain motion.