The PEER Lifelines Program: Current Research in Design Ground Motions
Michael Riemer (PEER), Clifford Roblee (Caltrans), and Brian Chiou (Caltrans)

Design earthquake ground motions in California are typically governed by large magnitude events at close distances (approximately <20 km). Recorded motions from previous California earthquakes show variability approaching a factor of 10 for sites at the same distance and magnitude. This variability arises from the differences in source, path, and site effects.

Figure 1 illustrates the variability of near-field ground motions. Velocity time histories in the direction normal to the fault strike are presented for two sites that are both within 3 km of the fault trace during the 1979 Imperial Valley, California, earthquake. Clearly, the El Centro #6 (E06) recording shows a large velocity pulse that is not evident in the Bonds Corner (BCR) record. This difference in amplitude between the recordings at the two stations is primarily attributed to rupture directivity. A map of the region (fig. 2a) shows the location of the two stations relative to the trace; the star denotes the epicenter. The E06 station is in the forward directivity direction (i.e., the fault rupture propagates toward E06), while the BCR station is close to the epicenter. However, the response is complicated when further details of the rupture are considered. A vector plot (fig. 2b) of the distribution of co-seismic slip along the fault plane shows a large asperity about midway along the strike. The E06 station is located much closer to this zone where energy release is greatest, thus obscuring conclusions regarding the exact contribution of directivity to the observed velocity pulse. Regardless of the cause, the difference in motions between these sites can have serious ramifications for the response of built facilities, and point to the need for improved ground-motion models.


Fig. 1. Fault-normal velocity time histories from the 1979 Imperial Valley earthquake.
Both sites are within 3 km of the fault trace, but the El Centro array is further along in the direction of the rupture.


A primary theme of the PEER-Lifelines Program is the performance of a coordinated suite of research projects aimed at systematically developing a better understanding and accounting for variability in recorded ground motions. Individual projects are focused on acquisition of new field and laboratory data, calibration of numerical procedures of ground-motion simulation, and development of new design tools. Together, these projects are expected to yield improved design capabilities to estimate the seismic loads, and associated uncertainties, that are imposed on built facilities. These improved capabilities are providing the scientific underpinning for the demand-side of performance-based engineering design that is at the core of PEER’s mission. Several projects from the Lifelines Program that contribute to improved ground-motion models are introduced below. For more detailed information, refer to the project updates on the PEER website (http://peer.berkeley.edu/).

Ground Truth: Recorded Data

Direct observations of near-field (<20 km) ground motions from large magnitude (M >7) earthquakes provide the standard against which any predictive methods should be compared. Unfortunately, until recently there have been little data available, since large events are rare and strong motion instruments have been sparsely distributed. During 1999, however, the Kocaeli and Duzce earthquakes in Turkey, and the Chi-Chi earthquake in Taiwan provided a wealth of new information. The addition of recordings from these events expanded the database of near-field ground motions approximately tenfold.

The “raw” data recorded by accelerometers often requires substantial evaluation and processing to produce a reasonable estimation of actual ground motions. When considering many different recordings from multiple earthquakes around the world, this processing must be especially consistent while extracting the maximum engineering value from each. The PEER Strong Motion Database (http://peer.berkeley.edu/smcat/) has been developed with these goals in mind. The database provides a user-friendly, web-based search engine to access recordings that have been uniformly processed by Dr. Walter Silva (Pacific Engineering and Analysis). The PEER-Lifelines program has recently enhanced the database by supporting Silva’s work to incorporate main shock data from the Turkey and Taiwan earthquakes. Additionally, the Lifelines Program is supporting efforts under way by Dr. William Lee (formerly of U.S. Geological Survey) to appropriately screen and process the large body of data gathered from the Taiwan aftershock sequence recorded by the Taiwan Central Weather Bureau’s extensive instrumentation network.
To optimize the value of the recent recordings, the Lifelines Program is supporting the development of a new generation of attenuation relations. As a preliminary step, Dr. Brian Chiou (California Department of Transportation) recently completed an update of the Sadigh attenuation relationship and others, including data from the Turkey and Taiwan events. A long-term goal of the ground motion research is the participation of experts in the reformulation of their widely applied attenuation relationships, implementing the improved understanding of source, path, and site effects developed from the PEER-Lifelines Program as well as other relevant research.


Fig. 2a. Fault location and stations distribution

Site Characterization

To maximize the value of recorded data to improve ground-motion models, subsurface conditions beneath a recording site need to be characterized. This allows the recordings to be appropriately grouped for use in attenuation relationships and provides critical inputs to account for site-to-site variations when ground-motion numerical modeling is performed. The PEER-Lifelines Program is contributing to ongoing efforts to characterize key recording sites both in the U.S. and abroad. In the U.S., support is being provided to the ROSRINE effort to systematically collect and disseminate geologic, geotechnical, and geophysical data from instrumented sites, as described in a related article. Dr. Robert Nigbor (University of Southern California) is leading the field site investigations, and is working closely with Dr. J.-P. Bardet and Dr. Jennifer Swift (also at USC) in web posting the logs (http://geoinfo.usc.edu/rosrine/). Dr. Kenneth Stokoe and Dr. Ellen Rathje (University of Texas at Austin) and Dr. James Bay (Utah State University) have used the Spectral-Analysis-of-Surface-Waves (SASW) technique to measure the shear wave velocity profiles at selected strong motion sites in the U.S., Turkey and Taiwan (forthcoming). This noninvasive method has proven to be a valuable and cost-effective approach for characterizing site conditions. Dynamic geotechnical material properties for samples collected at instrumented sites are being determined by a team coordinated by Dr. Donald Anderson (CH2MHill) with state-of-the-art laboratory testing provided by Dr. Kenneth Stokoe (UT) and Dr. Mladen Vucetic (University of California at Los Angeles).

Modeling of Near-Field Ground Motions

Procedures to numerically simulate ground motions from an extended fault rupture have been developed over many years, with sophisticated procedures now capable of incorporating near-field effects. These effects include “directivity pulse” (wave field buildup in the direction of rupture propagation), “fling step” (motions very near the fault trace associated with permanent offset of the ground surface), and the polarization (radiation pattern) of energy release that causes different motions to occur in the strike-parallel and strike-normal directions. The simulation procedures are currently used in practice to guide the development of attenuation relationships and to generate synthetic motions for conditions where little or no recorded data are currently available. In the future, there will likely be an increased reliance on simulation approaches to account for the various physical processes that contribute to ground motion variability. Acceptance of results from numerical simulation procedures requires careful calibration and validation against available earthquake data and new experimental data sets.


Fig. 2b. Slip distribution of the composite source model

Toward this end, the PEER-Lifelines Program is supporting a comprehensive validation exercise involving three teams of researchers: Dr. Paul Somerville, Dr. Robert Graves and Dr. Arben Pitarka (URS Corporation); Dr. Walter Silva (Pacific Engineering and Analysis); and Dr. Yuehua Zeng and Dr. John Anderson (University of Nevada, Reno). In the first phase of the work, now complete, each team applied its own procedures to simulate five large earthquakes from which some near-field data were recorded (1979 Imperial Valley, 1989 Loma Prieta, 1992 Landers, 1994 Northridge, and 1995 Kobe). These calibrations focused on directivity issues, and the results showed that the procedures were comparable in terms of fitting to the recorded data, although significant differences between the simulations and the observed motions were observed at specific locations. The initial study also highlighted the need to treat site effects in a consistent manner during the development of the fault slip model and subsequent simulation of recorded motions.

Further validation of these methods is continuing with simulation of the Turkey and Taiwan earthquakes of 1999, and with simulation of physical model tests of fault rupture as described below. Once the validation studies are completed, it is anticipated that the strengths and limitations of each numerical procedure can be assessed with confidence, and that directivity effects on near-field ground motions can be incorporated into the attenuation relationships used in design practice.

Experimental Data for Directivity

The validation of numerical simulation techniques for directivity effects using field data is hampered by the lack of near-field recordings and difficulties in adequately characterizing complex subsurface conditions. To investigate an alternative approach, the PEER-Lifelines Program is sponsoring an investigation by Dr. James Brune and Dr. Rasool Anooshehpoor at the University of Nevada, Reno, designed to produce directivity effects in the laboratory under carefully controlled conditions. Fault rupture is being physically simulated by slippage between two large blocks of foam rubber while precise measurements of the acceleration and displacement time histories are made at critical locations, both along the “ground surface” of the fault and within the rupture plane. The low density and stiffness of the material allows high-resolution measurement of wave propagation over laboratory scales, while the uniform properties minimize complexities in the response caused by inhomogeneities. In addition to developing data for uniform slip, fault-plane roughness will be altered locally in a variety of patterns to examine the impact of asperities on the buildup and saturation of rupture directivity effects. In effect, these experiments allow for direct control of two separate physical phenomena that are difficult to distinguish in field recordings, such as the examples in figures 1 and 2.

Basin Studies

Ground motion characteristics, particularly the duration of strong shaking, can be affected by seismic energy being trapped in sedimentary basins. In recent years, a variety of numerical “basin modeling” procedures have been developed to model long-period (>1 sec) wave propagation in a 3-D medium. The PEER-Lifelines Program is sponsoring a national group of investigators in a joint calibration exercise headed by Dr. Steven Day at San Diego State University. Modeling is being performed by Dr. Jacobo Bielak of Carnegie Mellon University, Dr. Shawn Larsen and Dr. Douglas Dreger at Lawrence Livermore National Laboratory and UC Berkeley, respectively, Dr. Kim Olsen at UC Santa Barbara, and Dr. Robert Graves and Dr. Arben Pitarka. Each team is applying its own numerical codes to analyze a series of increasingly complex “standard” problems. An initial study, now complete, involved a series of only simple problems for which exact solutions were available. A surprising finding from this early work was the large discrepancies between some of the results for the simplest problem. The cooperative framework for the calibration enabled the teams to isolate and rectify errors in the analyses, and consistent results are now being obtained. The full set of problems has been made available via the World Wide Web to a wide community of international researchers, and is therefore providing a valuable standard beyond the direct limits of the funded research.

With basic validation complete, the models are now being applied to more realistic problems including simulation of the 1994 Northridge earthquake using the 3-D velocity model developed by Dr. Harold Magistrale (San Diego State University) and others for the Southern California Earthquake Center (SCEC). In this phase of calibration, the teams will evaluate the quality of simulation results by direct comparison to each other and to recorded data. In addition, the results will be contrasted to those simulated from site-specific 1-D velocity models to identify the benefit of using increasingly sophisticated models.

This stepwise approach to calibration is considered essential to acceptance of the results in practice. Once the individual models are fully calibrated, the codes will become useful tools both for predicting responses in particular regions due to specific events and for evaluating the parametric effects of basins on the amplitude and duration of motion. Ultimately the PEER-Lifelines Program, in conjunction with SCEC, plans to apply these codes to selected scenarios for both the Los Angeles basin and the greater San Francisco Bay Area.

Improved Emergency Response

Developments in ground-motion modeling are also providing benefits to the emergency response community. The PEER-Lifelines Program is supporting a research project headed by Dr. Douglas Dreger to integrate fundamental features of finite-fault and near-field modeling into the Shakemap software developed by the U.S. Geological Survey. The improved system will provide automated estimates of ground shaking immediately after an earthquake, even in areas that are currently only sparsely instrumented.

Implementation

For the research program to make a positive impact on the reliability of lifeline systems in future earthquakes, the improvements must be implemented into practice. This focus on implementation is a major feature of the PEER-Lifelines Program and drives the motivation to keep the various related tasks coordinated within an overall framework. Individual projects are already beginning to have impact on practice. However, the most significant payoffs are expected in the coming years when the calibrations and validations in light of the new high quality data are completed. Then, it is anticipated that we will have gained the confidence and experience to apply the improved models widely in practice both directly and in the formulation of simpler design guidelines.