An Overview of the NGA Project - Hong Kong University of ...ihome.ust.hk/~gwang/CIVL575/Spectra_NGAPapers/Overview.pdf · An Overview of the NGA Project Maurice Power,a) M.EERI, Brian

An Overview of the NGA Project

Maurice Power,a)M.EERI, Brian Chiou,b) Norman Abrahamson,c)

M.EERI,

Yousef Bozorgnia,d)M.EERI, Thomas Shantz,b)

and Clifford Roblee,b)M.EERI

The “Next Generation of Ground-Motion Attenuation Models” (NGA)project is a multidisciplinary research program coordinated by the LifelinesProgram of the Pacific Earthquake Engineering Research Center (PEER), inpartnership with the U.S. Geological Survey and the Southern CaliforniaEarthquake Center. The objective of the project is to develop new ground-motion prediction relations through a comprehensive and highly interactiveresearch program. Five sets of ground-motion models were developed byteams working independently but interacting with one another throughout thedevelopment process. The development of ground-motion models wassupported by other project components, which included (1) developing anupdated and expanded PEER database of recorded ground motions, includingsupporting information on the strong-motion record processing, earthquakesources, travel path, and recording station site conditions; (2) conductingsupporting research projects to provide guidance on the selected functionalforms of the ground-motion models; and (3) conducting a program ofinteractions throughout the development process to provide input and reviewsfrom both the scientific research and engineering user communities. Anoverview of the NGA project components, process, and products is presentedin this paper. �DOI: 10.1193/1.2894833�

INTRODUCTION

The “Next Generation of Ground-Motion Attenuation Models” (NGA) project wasundertaken to develop new attenuation models (ground-motion prediction relations) forshallow crustal earthquakes in the western United States and similar active tectonic re-gions. The project has been accomplished through a comprehensive and highly interac-tive research program that has included the following components:

• Developing separate ground-motion models by five teams (each team referred toas a “developer”). Each team independently developed their models but inter-

a) Geomatrix Consultants, 2101 Webster St., 12th Floor, Oakland, CA 94612b) California Department of Transportation, Office of Infrastructure Research, 5900 Folsom Blvd, MS-5,Sacramento, CA 95819

c) Pacific Gas & Electric Company, Geosciences Department, 245 Market St., Rm. 403, PO Box 770000, MCN4C, San Francisco, CA 94177

d) Pacific Earthquake Engineering Research Center, University of California, 325 Davis Hall, MC1792,
Berkeley, CA 94720
3Earthquake Spectra, Volume 24, No. 1, pages 3–21, February 2008; © 2008, Earthquake Engineering Research Institute

4 M. POWER ET AL.

acted extensively with one another and with other NGA project componentsthroughout the development process.

• Developing an updated and expanded Pacific Earthquake Engineering ResearchCenter (PEER) ground-motion database that provided the recorded ground-motion data and the supporting information (metadata) about the recordings thatwere used by the five developers in developing the ground-motion models.

• Conducting a number of supporting research projects to provide an improved sci-entific basis for evaluating the functional forms of and constraints on the ground-motion models.

• Conducting a series of project interaction meetings that provided a frameworkfor discussion and review of project results by the ground-motion model devel-opers, the scientific research community and the engineering user community,and an external peer review team. The meetings included workshops, workinggroup meetings, developer meetings, and external review meetings.

Organizational leadership for the NGA project was provided by representatives ofPacific Gas & Electric Company (PG&E), California Department of Transportation(Caltrans), and Pacific Earthquake Engineering Research Center (PEER). The LifelinesProgram of PEER served as the lead organization for project coordination, and partnerorganizations were the U.S. Geological Survey (USGS) and the Southern CaliforniaEarthquake Center. The collaboration of these and other organizations and the projectinteractive framework provided a unique opportunity for the community of strong-motion seismologists and geotechnical and structural engineers to work together andmake a significant step forward in the prediction of strong ground motions for shallowcrustal earthquakes in active tectonic regions.

The following sections of this paper briefly summarize the project components listedabove. Other papers in this special issue of Earthquake Spectra present the NGA projectresults in detail, including the developed ground-motion prediction models, database de-velopment, and supporting research. The development, formulations, and ground-motionpredictions of the individual NGA models are described in papers in this issue by eachof the five developers (listed below). In addition, the five NGA models are compared insome detail with respect to the data sets utilized, model parameterizations, and ground-motion predictions in the paper jointly authored by the developers (Abrahamson et al.2008, this issue).

GROUND-MOTION PREDICTION MODEL DEVELOPMENT

GROUND-MOTION MODEL DEVELOPERS

Developers of five pre-existing and widely used ground-motion models have devel-oped NGA models. The developers are listed below, with references to their pre-existingground-motion models shown in parentheses:

• Norman Abrahamson and Walter Silva (Abrahamson and Silva 1997)

• David Boore and Gail Atkinson (Boore et al. 1997)

AN OVERVIEW OF THE NGA PROJECT 5

• Kenneth Campbell and Yousef Bozorgnia (Campbell 1997, Campbell andBozorgnia 2003)

• Brian Chiou and Robert Youngs, representing the model of Sadigh et al. (1993,1997)

• I. M. Idriss (Idriss 1991)

REQUIREMENTS FOR APPLICABILITY OF NGA MODELS

To meet the needs of the earthquake engineering community, all NGA models wereinitially required to be applicable to the following:

• Ground-motion parameters of peak ground acceleration (PGA), peak ground ve-locity (PGV), and 5% damped elastic pseudo-response spectral accelerations inthe period range of 0 to 10 seconds;

• Average horizontal component of ground motion, as well as ground motion in thefault-strike-normal (FN) and fault-strike-parallel (FP) directions;

• Shallow crustal earthquakes (strike-slip, reverse, and normal earthquakes) in thewestern United States;

• Moment magnitude range of 5 to 8.5 (strike-slip earthquakes) and 5 to 8 (reverseand normal earthquakes);

• Distance range of 0 to 200 km; and

• Commonly used site classification schemes, including the NEHRP classificationscheme.

The above requirements were later modified to postpone the quantification of FN andFP components of ground motion in order to focus on completion of models for the av-erage horizontal component. Implementation of the above requirements resulted in NGAground-motion models that are applicable to a wider range of conditions than pre-existing models. For the older relations, response spectral values were developed tolongest periods varying from 2 to 5 seconds, PGV was not addressed for most relations,largest magnitudes of stated applicability varied from 7.5 to 8, and distances of appli-cability varied from 60 to 100 km.

MAIN TECHNICAL ISSUES ADDRESSED IN NGA MODEL DEVELOPMENT

The main technical issues addressed in the NGA model development and in the sup-porting research are (1) consideration of various effects on ground motions, and (2) con-siderations in the statistical analysis of data.

Consideration of Effects on Ground Motions

• Moderate-to-large magnitude scaling at close distances;

• Distance scaling at both close and far distances;

• Rupture directivity;

• Footwall vs. hanging wall for dipping faults;

• Style of faulting (strike-slip, reverse, normal);

6 M. POWER ET AL.

• Depth to faulting (e.g., buried vs. surface rupture);

• Static stress drop (or rupture area);

• Site amplification relative to a reference “rock” condition; and

• Three-dimensional (3-D) sedimentary basin amplification/depth to basementrock.

Based on their analyses and judgment, the developers decided which of these effectswould be incorporated in their NGA relations and how the effects would be modeled. Inaddition to magnitude scaling, distance scaling, style of faulting, and site amplificationeffects (incorporated by all developers in their models), most developers explicitly in-corporated effects of footwall vs. hanging wall for dipping faults, depth of faulting, andbasin amplification/depth to basement rock. Decisions on whether and how to modelthese effects required considerable analysis and judgment because of uncertainties re-garding the relative influence of multiple effects on the observed ground motions. Ef-fects of directivity are not incorporated in the current NGA models. However, directivityeffects on the average horizontal component of ground motion have been analyzed bySpudich and Chiou (2008, this issue) working with the developers, and these results maybe used to extend the basic developer ground-motion predictions to include these effects.

A decision made by all developers was to use the average shear-wave velocity in theupper 30 meters of sediments, VS30, as the parameter for characterizing effects of sedi-ment stiffness on ground motions. Use of this parameter is considered to be more diag-nostic in determining site amplification than the broad and ambiguous soil and rock cat-egories used in the pre-existing relations (with the exception of the relation of Booreet al. (1997), which used VS30). In models of four of the developers, site amplificationsof ground motions relative to a reference rock condition are continuous functions ofVS30. These models also incorporate the effect of soil nonlinearity, i.e., dependence ofamplification on the level of ground shaking. The model of Idriss estimates ground mo-tions for two higher-velocity categories of VS30 (450 m/s�VS30�900 m/s, and VS30

�900 m/s), and extension of the model to lower-velocity soils is in progress.

As summarized later, supporting research carried out for the NGA project providedevaluations of effects that were considered in adopting functional forms and constrainingpredictions of ground-motion models. It is important to also note that developers usedbasic seismological theory, simple seismological models, and supplemental analyses inmaking decisions about their models. These additional efforts were important in charac-terizing magnitude and distance scaling of ground motions and extension of responsespectra models to ten-second period. (The number of records with reliable long-periodinformation was progressively smaller for periods beyond about one second.) Some de-velopers analyzed supplemental data sets of broadband ground-motion recordings thatare not available in strong-motion databases to evaluate attenuation rates at farther dis-tances (greater than about 50 km).


Considerations in the Statistical Analysis of Data

• Uncertainties in predictor variables (e.g., uncertainty in earthquake magnitudeand site VS30); and

• Dependencies of standard errors on magnitude, distance, soil type �VS30�, andshaking level.

As a result of these evaluations, developers variously recommended (1) lower stan-dard deviations in ground-motion models for sites where VS30 is measured rather thanestimated using correlations, (2) a formulation for reduced standard deviations for soilsites for higher reference rock motions, due to nonlinear soil response, and (3) standarddeviations to either decrease with increasing earthquake magnitude in the magnituderange of 5 to 7–7.5 or be independent of magnitude.

AVAILABILITY OF GROUND-MOTION MODELS AND REPORTS

Reports documenting NGA models are available electronically from the PEER website (http://peer.berkeley.edu/products/rep_nga_models.html) and in hard copy fromPEER. Computer-coded NGA models in various forms, convenient for use, are availablewith each report on the web site.

NGA DATABASE DEVELOPMENT

To provide a common database of recorded ground motions and supporting informa-tion for the ground-motion model developers, the NGA project conducted an extensiveupdate and expansion of the PEER database (Chiou et al. 2008, this issue). Figure 1shows the magnitude-distance distribution of the new earthquake data that have been

Figure 1. Magnitude and distance distribution of ground-motion records in the NGA database.

8 M. POWER ET AL.

added to the PEER database, superimposed on the pre-existing data. The new earth-quakes include the 1999 Hector Mine, California, earthquake; 1999 Kocaeli and Duzce,Turkey, earthquakes; 1999 Chi-Chi, Taiwan, earthquake and five major aftershocks; sev-eral well-recorded moderate earthquakes in California; 2003 Nenana Mountain and De-nali, Alaska, earthquakes, and several earthquakes from extensional tectonic regimes.The expanded data set includes 173 earthquakes, 1,456 recording stations, and 3,551multicomponent recordings.

Each developer team was required to use a subset of the database developed duringthe NGA project and supporting information in the database (e.g., source parameters,source-to-site distance, and local site condition of the recording station). From this com-mon database, individual records could be selected or excluded at the discretion of eachteam. Although the database consists of ground recordings intended to be from shallowcrustal earthquake environments similar to that of the western United States, neverthe-less there were decisions required by the developers on records to be used in developingNGA models. Factors considered by the developers in finalizing their data sets includedthe following:

• Records possibly from a different tectonic environment (e.g., subduction zone,deep earthquake, or stable continental region);

• Earthquakes not sufficiently well defined in terms of magnitude, distance, etc.(especially for some older earthquakes);

• Records with identified problems (e.g., S-wave trigger or second trigger) or ques-tions regarding data quality;

• Records judged not sufficiently representative of free-field conditions (e.g.,records obtained in basements, on the ground floor of heavy or tall buildings, onmassive foundations, or on dam abutments); and

• Records from aftershocks.

A key decision for each developer was whether to include records from the ex-tremely well recorded Chi-Chi M7.6 main shock (421 record sets) and aftershocks(1,392 record sets). This topic received considerable discussion in developer meetings.Ultimately, all five developer teams chose to use main-shock records from Chi-Chi, andthree of the five teams chose to use Chi-Chi aftershock data. The data selection criteriaand resulting data sets used by each developer are summarized in the developers’ papersin this issue and described in detail in the PEER reports of NGA models. The data setsare compared by Abrahamson et al. (2008, this issue).

As described by Chiou et al. (2008, this issue), a process was undertaken to evaluatethe techniques used in processing the ground-motion records with respect to the accu-racy of the response spectral values of the processed data. The lowest usable frequency(longest usable period) for each record was evaluated, and recommended values aregiven in the database. Most records in the database have been rotated to FN and FP di-rections. A method was developed and implemented for records in the database thatuniquely defines the average horizontal component independent of the orientation of thetwo horizontal components as recorded (Boore et al. 2006), and this method was
adopted and used for developing NGA models. Other measures of horizontal motion in-


cluding measures of the maximum horizontal component in comparison to the averagehorizontal component were also evaluated by NGA researchers and others, includingBeyer and Bommer (2006), Campbell and Bozorgnia (2007; 2008, this issue), Watson-Lamprey and Boore (2007), and Huang et al. (2008, this issue).

An extensive effort was made to compile, evaluate, and extend the metadata onearthquake sources, travel paths, and local site recording conditions. Some of the sig-nificant accomplishments include the following:

• Characterizations of site conditions of recording stations using different param-eters and classification systems, such as VS30 (average shear-wave velocity to 30-meter depth), NEHRP classification, surface geology classification, and Geoma-trix classification;

• Estimation of VS30 values using correlations at stations not having measuredVS30 values (about 30% of the recording stations have measured VS30);

• Systematic and consistent evaluation of earthquake magnitudes, type of faulting,fault rupture geometry, classification of stations for hanging wall and footwallconditions, rupture directivity parameters using both Somerville et al. (1997) andSpudich and Chiou (2008, this issue) parameterizations, and source-to-sitedistances using different distance measures;

• Estimation of depth to earthquake rupture;

• Compilation of depths to bedrock characterized by Vs=1.0, 1.5, and 2.5 km/secat recording station sites; and

• Compilation of basin information for stations in basins.

One especially significant effort in developing the NGA database was the estimationof VS30 at recording station sites based on correlation of measured VS30 values withmapped geology and other characteristics. Studies used in developing these estimatesincluded estimation of VS30 values for 1994 Northridge, California, earthquake record-ing sites by the USGS (Borcherdt and Fumal 2002; Borcherdt 2003, personal commu-nication) and VS30 values estimated for station sites throughout California by the Cali-fornia Geological Survey (CGS) (Wills and Clahan 2004, personal communication;Wills and Clahan 2006). Similar correlation studies were made to estimate VS30 at otherrecording sites, including sites that recorded the 1999 Chi-Chi, Taiwan, earthquake andaftershocks. In addition, as part of the NGA project, measurements of VS30 were madeby the USGS at selected stations (Kayen et al. 2005) and some of the resulting VS30values were incorporated in the NGA database.

The strong-motion data and supporting metadata in the NGA database are availablefor public use at PEER’s web site. For access to the entire data set used for the NGAproject, a flat file is available via http://peer.berkeley.edu/products/nga_project.html.Definition of each data column is available via this web site and also included in Ap-pendix A (online) of Chiou et al. (2008, this issue).

10 M. POWER ET AL.

SUPPORTING RESEARCH ON EARTHQUAKE EFFECTS

Supporting research has been carried out by the project partners, PEER-LL, USGS,and SCEC, to provide an improved scientific basis for evaluating the functional formsand constraints for NGA models. This research includes:

• Simulations of rock motions,

• Simulations of site response amplification of ground motions,

• 3-D simulations of basin response amplification of ground motions, and

• Development of a formulation for directivity effects on ground motions usingisochrone theory.

Brief descriptions of these studies are given below.

SIMULATIONS OF ROCK MOTIONS

Simulations of rock motions were carried out by three groups to provide theoreticalguidance to the NGA developers on the scaling of ground-motion response spectra withmagnitude, distance, and rupture area, and on effects of rupture directivity, hanging wallvs. foot wall, and buried vs. shallow faulting. The procedures and results of simulationsby URS Corporation (URS), Pacific Engineering and Analysis (PEA), and University ofNevada, Reno (UNR), are described by Graves and Pitarka (2004), Silva et al. (2002),and Zeng et al. (1994) and Zeng (2002), respectively. The three sets of results are com-pared and discussed by Somerville et al. (2006) as revised by Collins et al. (2006). Thesimulations provided some constraints on the scaling of ground motions with varioussource parameters, but differences among the simulations were significant in many cases(see Collins et al. 2006). An example of results obtained by the three groups—effects ofmagnitude on response spectral values at several periods normalized to values for mag-nitude 7—is shown in Figure 2. The magnitude scaling shown in Figure 2 is for constantstress drop, and results for another source model are also given in Collins et al. (2006).

SIMULATIONS OF SITE RESPONSE

As summarized by Walling et al. (2008, this issue), site response simulations wereconducted by Silva (2008) for shear-wave velocity profiles representative of differentVS30 values, NEHRP soil categories, and geologies. The equivalent linear analyses in-cluded four different soil models (i.e., different relations for variation of shear modulusand damping with shear strain) and variations in profile depth. Figure 3 is an example ofamplification of response spectral acceleration at 0.2-sec period relative to referencerock for soil and rock profiles of different shear-wave velocities as a function of refer-ence rock PGA �VS30 for reference rock=1,100 m/sec� for the “Peninsular Range” soilmodel. The VS30 values of 1,100 m/sec, 560 m/sec, and 270 m/sec correspond ap-proximately to mid-range values for NEHRP Site Classes B, C, and D, respectively, andthe VS30 value of 750 m/sec is approximately the NEHRP B/C boundary. Results simi-lar in form to those in Figure 3 were developed for a second soil model (“EPRI” model),which gave greater reduction in amplification with increase in rock PGA than the “Pen-insular Range” model (i.e., greater soil nonlinearity).


Using the site response simulation results, Walling et al. (2008, this issue) developedtwo parametric models (one for each soil model mentioned above) of nonlinear soil am-plification of response spectral accelerations for different periods of vibration as a func-tion of VS30 and reference rock PGA. Two NGA developers used the results of Wallinget al. (2008, this issue) to constrain the nonlinear soil response. The simulation resultswere also used in evaluating effects of depth to rock on ground-motion amplification.

3-D BASIN RESPONSE SIMULATIONS

Day et al. (2005; 2006; and 2008, this issue) and Day (2005) carried out a compre-hensive series of 3-D basin response simulations for deep sedimentary basins in south-ern California. Fault geometries were based on the SCEC Community Fault Model andthe 3-D velocity structure based on the SCEC Community Velocity Model (Magistrale etal. 2000). Response spectral accelerations from the simulations were normalized tothose from simulations for reference hard-rock models. Figure 4 illustrates means andstandard deviations of source-averaged basin amplifications calculated from the simula-

Figure 2. Magnitude scaling of spectral acceleration for strike-slip earthquakes from 1-D rockmotion simulations using constant stress drop model, normalized at M7 (Collins et al. 2006).

12 M. POWER ET AL.

tions for three different periods as a function of depth to the 1.5 km/sec shear-wave ve-locity interface (Z1.5). Amplifications were similarly calculated for 26 periods from 2 to10 seconds and depths to the 1.0 km/sec, 1.5 km/sec, and 2.5 km/sec shear-wave ve-locity interfaces (Z1.0, Z1.5, and Z2.5). Day et al. (2005; 2006; and 2008, this issue) andDay (2005) developed a parametric model to characterize these amplifications, which isillustrated by the dashed curves in Figure 4. For use in developing and incorporating abasin depth effect term in ground-motion relations, it is necessary to assess how the av-erage rock site departs from the idealized reference model, and how basin depth maycorrelate with other predictor variables affecting amplification, such as VS30 of the basinsoils. Some developers used these results in developing models of effects of basin depth/depth to rock on ground-motion amplification.

DIRECTIVITY FORMULATION USING ISOCHRONE THEORY

Based on the isochrone formulation of earthquake ground motion (Spudich andFrazer, 1984), Spudich and Chiou (2008, this issue) introduced the physically based iso-chrone directivity predictor (IDP) as an alternative to the one developed by Somerville etal. (1997). The advantage of IDP is that it can be used for any location near faults havingany dip or rake angles (faulting mechanism). Generalization to multi-segment faults isgiven in Appendix A of Spudich and Chiou (2008, this issue).

Spudich and Chiou also investigated the preferred formulation of the directivity ef-fect in the NGA data set. Using data with magnitudes greater than 5.6 and distances lessthan 70 km, they concluded that the directivity effect can be modeled as a linear func-tion of IDP. Using this formulation and the total residuals from four NGA models, em-pirical models for 10 spectral periods in the range of 0.5 to 10 seconds were developedfor each of the four NGA relations. The resulting isochrone directivity models predictroughly half of the strike-slip effects predicted by Somerville et al. (1997). The spatialpatterns of predicted effects also differ in several important ways. These empirical mod-

Figure 3. Example of amplifications of response spectral acceleration at 0.2-sec period fromsite response simulations using Peninsular Range soil model (Walling et al. 2008, this issue).


els may be applied as correction factors to the corresponding NGA relations as a meansto incorporate directivity effects in the median predictions for the average horizontalcomponent of the NGA relations. The isochrone directivity formulation may be ex-tended to characterize ground-motion polarization, which can be used as a method toconvert the predicted average horizontal motion to motion in a specific orientation suchas the direction perpendicular to fault strike (i.e., FN component).

PROJECT INTERACTIONS AND REVIEW

The NGA models development process has occurred over a period of five years, be-ginning with a kickoff meeting in October 2002. A key component of this process hasbeen input from the scientific and engineering communities as well as collaborative in-teraction among developer teams. Several activities were organized to foster discussionof technical issues. These are briefly summarized below.

WORKSHOPS

Eight one- to two-day workshops, each attended by 40 to 80 scientists and engineers,were conducted during the time period 2003–2005. The workshops provided for presen-tations by NGA developers, researchers, and working groups; and review of preliminaryresults by the scientific and engineering communities. Early workshops focused mainlyon total project scope and results of database development, supporting research, andworking group evaluations. Later workshops focused on the NGA models being devel-oped.

Figure 4. Illustration of basin amplification as a function of depth to 1.5 km/s isosurface from3-D basin response simulations. Solid symbols represent mean amplification factors and barsgive the standard deviations for three periods. Dashed curves are the corresponding basin fac-tors calculated from the parametric model (Day et al. 2008, this issue).

14 M. POWER ET AL.

WORKING GROUPS

Six working groups, each comprising 4 to 13 scientists and engineers having exper-tise in the respective working group areas of activity, were formed and met frequently toreview specific results being developed by the project and discuss key issues. Table 1shows the interrelationships between the working groups and the project technical tasks.The working groups made significant contributions to the project, and results of theirdeliberations were reported at the workshops. As examples, Working Group #4, Sourceand Path Effects, reviewed finite fault models (e.g., geometry and other characteristics ofthe fault rupture) for faults causing moderate-to-large magnitude earthquakes (see Chiouet al. 2008, this issue). Working Group #4 also recommended several evaluations in sup-port of NGA development that were subsequently carried out by researchers and NGAdevelopers, including supplemental evaluations of attenuation rates at far distances usingbroadband data; directivity characterization using the isochrone method; evaluation ofdata from multiple Chi-Chi earthquake aftershocks for use by the NGA developers; andreview of precarious rock studies. Working Group #5, Site Effects, examined differentpotential site classes and parameters that could be used as predictor variables for siteamplification in developing NGA models, and their work influenced the decisions of allof the developer teams to adopt VS30 as the preferred site characterization parameter.They recommended estimation of VS30 at sites not having measured values, and theyreviewed results of estimations. Other Working Group #5 activities included review of

Table 1. Technical tasks and related working groups for NGA project

Tasks Working Groups

1 • Evaluation of record processing WG#1 Record Processing• Data set development WG#2 Ground Motion Data Set

WG#4 Source/Path EffectsWG#5 Site Effects

2 • 1-D rock simulations• Validation of simulation procedures

WG#3 Validation of 1-D RockSimulation Procedures

WG#4 Source/Path Effects

3 • 3-D simulations of basin effects WG#4 Source/Path EffectsWG#5 Site Effects

4 • Evaluation of alternative source/path predictorvariables

WG#4 Source/Path Effects

5 • Evaluation of site classification schemesand site effects

WG#5 Site Effects

6 • Evaluation of site response analysis procedures• Development of site amplification factors

WG#5 Site Effects

7 • Development of statistical methods and toolsfor NGA applications

WG#6 Statistical Modeling of Data


analytical methodology used for site response simulations (Wang 2004), and compila-tion of results of empirical studies of site amplification (Power et al. 2004).

DEVELOPERS INTERACTION MEETINGS

Numerous meetings were held by the five NGA developer teams to exchange viewson technical issues and compare in-progress development of the NGA relations. Thisprocess was invaluable in sharing with all the developers the viewpoints held by eachteam and the modeling approaches being used. Each developer team made independentdecisions regarding their ground-motion model, but because of the interactions, deci-sions resulted in the models being similar in several respects.

EXTERNAL REVIEW

The USGS, led by the Golden, Colorado, office and working cooperatively with theCGS, conducted a review of the NGA models. The review was focused on the potentialuse of the NGA models in the current round of national seismic hazard mapping. Al-though earlier review meetings were held in 2005 and 2006, the principal and final re-view meeting occurred in September 2006 and involved, in addition to USGS and CGSreviewers, an external review panel formed by the USGS. Prior to that meeting, NGAdevelopers responded to review questions from USGS and CGS reviewers and providedthe responses to all reviewers. The questions and responses are posted on the PEER website. Based on the review process including responses to questions and input receivedfrom the external review panel, the USGS decided to adopt fully documented NGAmodels for the current revision of the U.S. national seismic hazard maps.

Many past assessments of seismic hazard have addressed modeling (or epistemic)uncertainty in ground-motion models by using a weighted combination of published re-lationships. The implicit (or sometimes explicit) assumption in this approach is thatmodels developed by different researchers represent alternative approaches to modelingsource, path, and site effects developed from somewhat “independent” viewpoints andthis captures much of the epistemic uncertainty in ground-motion modeling. As men-tioned in the preceding section, the NGA project involved a great deal of interactionamong the development teams and there are many common approaches for model com-ponents that appear in several of the final models. Thus the USGS was advised that justusing the alternative median models should not be assumed to adequately captureepistemic uncertainty and that additional epistemic uncertainty should be considered inapplication of the NGA models (or a subset) in seismic hazard assessments. Accord-ingly, the USGS developed and implemented a procedure for incorporating epistemicuncertainty in NGA models in national ground-motion mapping (Petersen et al. 2008).

SUMMARY

The principal features and accomplishments of the NGA project are summarizedbelow.

• Five sets of ground-motion models were developed for shallow crustal earth-quakes in the western United States and similar active tectonic regions. The mod-els were developed for wider ranges of magnitudes, distances, site conditions,

16 M. POWER ET AL.

and response spectral periods of vibration than had been used in pre-existingground-motion relations. As with the previous relations, the NGA models werein terms of the average horizontal component of ground motion. However, unlikethe previous models, a method of defining the average horizontal component foreach ground-motion record was developed that is independent of the as-recordedorientation of the two components. Each NGA developer team developed itsmodel independently but with frequent interaction with the other developers.

• For development of ground-motion models, developers systematically evaluateda list of predictor parameters to consider for predicting earthquake effects. Thepredictive parameters variously incorporated in the developers’ models includedearthquake magnitude, style of faulting, depth to top of fault rupture, source-to-site distance, site location on hanging wall or foot wall of dipping faults, near-surface soil stiffness, and sedimentary basin depth/depth to rock. One of the mostsignificant decisions made by all developers was to use the average shear-wavevelocity in the upper 30 meters of sediments, VS30, as the parameter for charac-terizing soil-stiffness effects on ground motions, whereas only one of the devel-opers had used VS30 in a pre-existing relation. Four of the developers incorpo-rated soil nonlinear effects in their NGA models. The use of VS30 withnonlinearity is considered to result in a much-improved characterization ofsite amplification effects as compared to characterizations in the pre-existingrelations.

• A series of supporting research projects were conducted to provide an improvedscientific basis for defining and constraining the functional forms used in theground-motion models to predict effects on ground motions. The researchprojects included simulations of rock motions (source-to-site ground-motionmodeling); simulations of site response amplifications of ground motions; simu-lations of deep basin amplification effects on ground motions; and developmentof a formulation and model for predicting directivity effects on ground motionsusing isochrone theory. In addition to using these research results, developersmade greater use of seismological theory in developing their models. They alsomade use of ground-motion data sets from broadband recordings to supplementthe strong-motion database developed for the NGA project to better defineattenuation rates at relatively far distances.

• One of the major accomplishments of the NGA project was the expansion andupdating of the PEER database of ground-motion recordings, which provided thedatabase used by the developers. Many significant recordings, including thosefrom international earthquakes, were added. Record processing was evaluatedand longest usable periods were defined for every record. Especially importantwas the systematic effort made to compile and extend the supporting information(metadata) about the causative earthquakes, travel paths, and site conditions atrecording stations, including the estimation of VS30 values using correlations atevery station not having measured values. The development of the NGA modelswould not have been possible without the PEER-NGA database, which is now a


valuable resource for the selection of ground-motion records for many applica-tions in practice and research.

• The NGA project was conducted within a framework of project interactions thatgreatly improved the transfer and review of information and project resultsamong project participants for the different project components, including data-base development, research results, and ground-motion model development. Theproject interactions included eight one- to two-day workshops, working groupmeetings for six working groups, developer interaction meetings to facilitatecommunication among the developer teams, and external project review meet-ings (the latter summarized in the paragraph below).

• A review of the developed NGA models was conducted by the USGS, working incooperation with the CGS. The review took place over an approximately one-year period and involved several meetings, of which the principal and final meet-ing also involved the participation of an external review panel formed by theUSGS. Based on the review, the USGS adopted fully documented NGA modelswith additional epistemic uncertainty for use in the current round of nationalground-motion mapping.

ACKNOWLEDGMENTS

The authors acknowledge sponsorship by the Pacific Earthquake Engineering Re-search Center’s (PEER’s) Program of Applied Earthquake Engineering Research of Life-lines Systems supported by the California Department of Transportation, the CaliforniaEnergy Commission, and the Pacific Gas and Electric Company. This work made use ofthe Earthquake Engineering Research Centers Shared Facilities supported by the Na-tional Science Foundation, under Award Number EEC-9701568 through PEER. Anyopinions, findings, and conclusions or recommendations expressed in this materialare those of the authors and do not necessarily reflect those of the National ScienceFoundation.

In addition to acknowledging the collaborative efforts of the project partners—Pacific Earthquake Engineering Research Center-Lifelines Program (PEER-LL), U.S.Geological Survey (USGS), and Southern California Earthquake Center (SCEC)—theNGA project would like to thank the numerous organizations and individuals who con-tributed to the project. Special appreciation is extended to the following organizations:the California Geological Survey (CGS) for contributions in many areas, including da-tabase development, project research, external project review with USGS, and partici-pation on working groups and in workshops; the Central Weather Bureau (CWB) in Tai-wan for providing the strong-motion records from the 1999 Chi-Chi main shock and fiveaftershocks; and the National Center for Research on Earthquake Engineering in Taiwanfor sharing with us borehole data at more than 200 CWB strong-motion sites. Manyother organizations who contributed to the development of the NGA database are ac-knowledged in the paper by Chiou et al. (2008, this issue).

The participation of numerous individuals has been vital to the energy and success ofthe NGA project. We thank the following persons for their contributions:

18 M. POWER ET AL.

Coordinators among the partnering and sponsoring organizations

Robert Anderson, David Chambers, Lloyd Cluff, William Ellsworth, Thomas Jordan,Jack Moehle, Stuart Nishenko, Clifford Roblee, William (Woody) Savage, ThomasShantz, and Paul Somerville.

NGA project coordination group

Norman Abrahamson, Yousef Bozorgnia, Brian Chiou, Maurice Power (coordinator),Michael Riemer, Clifford Roblee, and Thomas Shantz.

Ground-motion model developers

Norman Abrahamson and Walter Silva; David Boore and Gail Atkinson; KennethCampbell and Yousef Bozorgnia; Brian Chiou and Robert Youngs; and I. M. Idriss.

Project researchers

John Anderson, Jacobo Bielak, John Boatwright, David Brillinger, Chih-ChengChin, Brian Chiou, Kevin Clahan, Nancy Collins, Robert Darragh, Steven Day, DouglasDreger, Robert Graves, Nick Gregor, Shawn Larsen, Kim Olsen, Arben Pitarka,Leonardo Ramirez-Guzman, Walter Silva, Paul Somerville, Paul Spudich, Joseph Sun,Melanie Walling, Zhi-Liang Wang, Jennie Watson-Lamprey, Chris Wills, and YuehuaZeng.

Database developers and contributors

John Boatwright, David Boore, Roger Borcherdt, Jonathan Bray, Kenneth Campbell,Brian Chiou, Kevin Clahan, Robert Darragh, Douglas Dreger, William Ellsworth,Vladimir Graizer, Robert Graves, Nick Gregor, Moh-Jiann Huang, Robert Kayen, Will-iam H. K. Lee, Martin Mai, Robert Nigbor, Mark Petersen, Maurice Power, Ellen Rathje,Clifford Roblee, Kyle Rollins, Linda Seekins, Anthony Shakal, Jamison Steidl, WalterSilva, Paul Somerville, Paul Spudich, Christopher Stephens, Jonathan Stewart, KennethStokoe, David Wald, Jennie Watson-Lamprey, Donald Wells, Kuo-Liang Wen, ChrisWills, and Yuehua Zeng.

Working Group chairs and members

Working Group #1: Robert Graves and Maurice Power (chairs), Norman Abraham-son, Ralph Archuleta, David Boore, Brian Chiou, Wilfred Iwan, Robert Nigbor, and An-thony Shakal.

Working Group #2: Brian Chiou (chair), Norman Abrahamson, David Boore, YousefBozorgnia, Kenneth Campbell, I. M. Idriss, Walter Silva, Christopher Stephens, andRobert Youngs.

Working Group #3: Steven Day (chair), Norman Abrahamson, John Boatwright,Brian Chiou, and Ruth Harris.

Working Group #4: John Anderson (chair), Norman Abrahamson, John Boatwright,Brian Chiou, Steven Day, William Foxall, Ruth Harris, Mark Peterson, Walter Silva, PaulSomerville, and Paul Spudich.

Working Group #5: Roger Borcherdt (chair), Norman Abrahamson, Jonathan Bray,


Kenneth Campbell, Brian Chiou, Edward Field, I. M. Idriss, Maurice Power, CliffordRoblee, Walter Silva, Jonathan Stewart, Joseph Sun, and Chris Wills.

Working Group #6: Robert Youngs (chair), Norman Abrahamson, David Brillinger,and Brian Chiou.

USGS, CGS, and expert panel participants in USGS external review

John Anderson, Ralph Archuleta, John Boatwright, Roger Borcherdt, Tianqing Cao,Martin Chapman, Chris Cramer, C. B. Crouse, William Ellsworth, Arthur Frankel,Vladimir Graizer, Robert Graves, Steven Harmsen, Thomas Heaton, William Holmes,Nicolas Luco, Arthur McGarr, Charles Mueller, Mark Petersen, Michael Reichle, BadieRowshandel, Paul Spudich, Jonathan Stewart, David Wald, Robert Wesson, and YuehuaZeng.

Last, but not least, all those who participated in the eight NGA workshops and thekickoff meeting.

REFERENCES

Abrahamson, N. A., and Silva, W. J., 1997. Empirical response spectral attenuation relations forshallow crustal earthquakes, Seismol. Res. Lett. 68, 94–127.

Abrahamson, N., Atkinson, G., Boore, D., Bozognia, Y., Campbell, K., Chiou, B., Idriss, I. M.,Silva, W., and Youngs, R., 2008. Comparisons of the NGA ground-motion relations, Earth-quake Spectra 24, 45–66.

Beyer, K., and Bommer, J. J., 2006. Relationships between median values and aleatory vari-abilities for different definitions of the horizontal component of motion, Bull. Seismol. Soc.Am. 96, 1512–1522.

Boore, D. M., Joyner, W. B., and Fumal, T. E., 1997. Equations for estimating horizontal re-sponse spectra and peak acceleration from western North American earthquakes: A sum-mary of recent work, Seismol. Res. Lett. 68, 128–153.

Boore, D. M., Watson-Lamprey, J., and Abrahamson, N. A., 2006. Orientation-independentmeasures of ground motion, Bull. Seismol. Soc. Am. 96, 1502–1511.

Borcherdt, R. D., and Fumal, T. E., 2002. Shear-Wave Velocity Compilation for NorthridgeStrong-Motion Recording Sites, U.S. Geol. Surv. Open-File Rep. 2002-107.

Campbell, K. W., 1997. Empirical near-source attenuation relationships for horizontal and ver-tical components of peak ground acceleration, peak ground velocity, and pseudo-absoluteacceleration response spectra, Seismol. Res. Lett. 68, 154–179.

Campbell, K. W., and Bozorgnia, Y., 2003. Updated near-source ground-motion (attenuation)relations for the horizontal and vertical components of peak ground acceleration and accel-eration response spectra, Bull. Seismol. Soc. Am. 93(1), 314–331.

––—, 2007. Campbell-Bozorgnia NGA Ground Motion Relations for the Geometric MeanHorizontal Component of Peak and Spectral Ground Motion Parameters, Report PEER2007/02, Pacific Earthquake Engineering Research Center.

––—, 2008. NGA ground motion model for the geometric mean horizontal component of PGA,PGV, PGD, and 5% damped linear elastic response spectra for periods ranging from 0.01 to10 s, Earthquake Spectra 24, 139–171.

20 M. POWER ET AL.

Chiou, B., Darragh, R., Gregor, N., and Silva, W., 2008. NGA project strong-motion database,Earthquake Spectra 24, 23–44.

Collins, N., Graves, R., and Somerville, P., 2006. Revised Analysis of 1D Rock Simulations forthe NGA-E Project, Report to the PEER-Lifelines Program, Project 1C02d by URS Corpo-ration.

Day, S. M., 2005. 3D Ground motion simulation in basins, Supplemental models for the depthsto the 1.0 km/s and 2.5 km/s shear-wave velocity horizons, written communications.

Day, S. M., Bielak, J., Dreger, D., Graves, R., Larsen, S., Olsen, K. B., and Pitarka, A., 2005.3-D ground motion simulations in basins, Final report prepared for the Pacific EarthquakeEngineering Research Center, Project 1A03.

Day, S. M., Graves, R., Bielak, J., Dreger, D., Larsen, S., Olsen, K. B., Pitarka, A., andRamirez-Guzman, L., 2008. Model for basin effects on long-period response spectra insouthern California, Earthquake Spectra 24, 257–277.

Day, S. M., Bielak, J., Dreger, D., Graves, R., Larsen, S., Olsen, K. B., Pitarka, A., andRamirez, L., 2006. Numerical simulation of basin effects on long-period ground motion,Proceedings of the 8th National Conference on Earthquake Engineering, San Francisco, Ca-lif.

Graves, R. W., and Pitarka, A., 2004. Broadband time history simulation using a hybrid ap-proach, Proceedings of the 13th World Conference on Earthquake Engineering, Vancouver,Canada, Paper No. 1098.

Huang, Y. N., Whittaker, A. S., and Luco, N., 2008. Maximum spectral demands in the near-fault region, Earthquake Spectra 24, 319–341.

Idriss, I. M., 1991. Selection of Earthquake Ground Motions at Rock Sites, Report prepared forthe Structures Division, Building and Fire Research Laboratory, National Institute of Stan-dards and Technology, Department of Civil Engineering, University of California, Davis.

Kayen, R., Thompson, E., Minasian, D., and Carkin, B., 2005. Shear-Wave Velocity of theGround near Sixty California Strong Motion Recording Sites by the Spectral Analysis of Sur-face Waves (SASW) Method and Harmonic-Wave Sources, U.S. Geol. Surv. Open-File Rep.2005-1366.

Magistrale, H., Day, S. M., Clayton, R., and Graves, R. W., 2000. The SCEC Southern Cali-fornia three-dimensional seismic velocity model version 2, Bull. Seismol. Soc. Am. 90, S65–S76.

Petersen, M. D., Frankel, A. D., Harmsen, S. C., Mueller, C. S., Haller, K. M., Wheeler, R. L.,Wesson, R. L., Zeng, Y., Boyd, O. S., Perkins, D. M., Luco, N., Field, E. H., Wills, C. J., andRukstales, K. S., 2008. Documentation for the 2008 Update of the United States NationalSeismic Hazard Maps, U.S. Geological Survey Open-File Report 2008-1128.

Power, M., Borcherdt, R., and Stewart, J., 2004. Site Amplification from Empirical Studies,NGA project report, Working Group #5.

Sadigh, K., Chang, C.-Y., Abrahamson, N. A., Chiou, S. J., and Power, M. S., 1993. Specifi-cation of long-period ground motions: updated attenuation relationships for rock site condi-tions and adjustment factors for near-fault effects, Proceedings of the ATC-17-1 Seminar onSeismic Isolation, Passive Energy Dissipation, and Active Control, San Francisco, Calif.,59–70.


Sadigh, K., Chang, C.-Y., Egan, J. A., Makdisi, F., and Youngs, R. R., 1997. Attenuation rela-tionships for shallow crustal earthquakes based on California strong motion data, Seismol.Res. Lett. 68, 180–189.

Silva, W. J., 2008. Site response simulations for the NGA project, report prepared for the PacificEarthquake Engineering Research Center.

Silva, W., Gregor, N., and Darragh, R., 2002. Validation of 1-d Numerical Simulation Proce-dures, final report to PEER-Lifelines Project.

Somerville, P. G., Collins, N., Graves, R., Pitarka, A., Silva, W., and Zeng, Y., 2006. Simulationof ground motion scaling characteristics for the NGA-E Project, Proceedings of the 8th Na-tional Conference on Earthquake Engineering, San Francisco, Calif.

Somerville, P. G., Smith, N. F., Graves, R. W., and Abrahamson, N. A., 1997. Modification ofempirical strong ground motion attenuation relations to include the amplitude and durationeffects of rupture directivity, Seismol. Res. Lett. 68(1), 199–222.

Spudich, P., and Chiou, B. S. J., 2008. Directivity in NGA earthquake ground motions: analysisusing isochrone theory, Earthquake Spectra 24, 279–298.

Spudich, P., and Frazer, L. N., 1984. Use of ray theory to calculate high frequency radiationfrom earthquake sources having spatially variable rupture velocity and stress drop, Bull.Seismol. Soc. Am. 74, 2061–2082.

Walling, M., Silva, W., and Abrahamson, N. A., 2008. Nonlinear site amplification factors forconstraining the NGA models, Earthquake Spectra 24, 243–255.

Wang, Z., 2004. Examining site response issues for NGA, PEER Project Summary for Year 7,Project 2M01.

Watson-Lamprey, J. A., and Boore, D. M., 2007. Beyond SaGMRotI: Conversion to SaArb, SaSN,and SaMaxRot, Bull. Seismol. Soc. Am. 97, 1511–1524.

Wills, C. J., and Clahan, K. B., 2006. Developing a map of geologically defined site-conditioncategories for California, Bull. Seismol. Soc. Am. 96, 1483–1501.

Zeng, Y., 2002. Final Technical Report on Validation of 1-D Numerical Simulation Procedures,PEER Project 1C02, Task 1: Earthquake Ground Motion, Seismological Lab, University ofNevada-Reno.

Zeng, Y., Anderson, J. G., and Yu, G., 1994. A composite source model for computing realisticsynthetic strong ground motions, Geophys. Res. Lett. 21, 725–728.

(Received 2 July 2007; accepted 25 January 2008�

Documents

An Overview of the NGA Project - Hong Kong University of ...ihome.ust.hk/~gwang/CIVL575/Spectra_NGAPapers/Overview.pdf · An Overview of the NGA Project Maurice Power,a) M.EERI, Brian