Issues related to ground motion prediction equation development

Introduction

Detailed descriptions of the project and the GMPEs are readily available in a special issue of Earthquake Spectra (volume 24, No. 1), as well as other papers, such as Campbell et al. (2009), and for that reason this article only gives a capsule description of the essential details of the NGA project. New to this article will be a number of figures not contained in the Earthquake Spectra special issue. Because of length limitations, this article will primarily be composed of extended

Figure captions.

The NGA-Empirical Project

Personnel

The new GMPEs were derived by five developer teams: Abrahamson and Silva (AS), Boore and Atkinson (BA), Campbell and Bozorgnia (CB), Chiou and Youngs (CY), and Idriss (I). Because the Idriss GMPEs are of limited use (they are only for rock sites), they will not be discussed in this article. A number of working groups performed studies in support of the derivation of the GMPEs (see Power et al., 2008, for details).

Scope

The developer teams were given the task of developing GMPEs for a median measure of ground motion ("GMRotI50", as defined in Boore et al., 2006; see Beyer and Bommer, 2006; Campbell and Bozorgnia, 2007; Watson-Lamprey and Boore, 2007; and Huang et al., 2008, provided equations to convert GMRotI50 to maximum spectral amplitude), including peak acceleration, velocity, and displacement, as well as 5%-damped pseudo spectral acceleration (PSA) for periods from 0.01 s to 10 s. The equations were to be valid for magnitudes ranging from 5.0-8.5 (for strike slip faulting) and 5.0-8.0 for reverse slip faulting) and distances from 0 to 200 km. Models for the lavatory variability were to be included.

Database

A major effort was put into developing the database to be used by the developers (see Chiou et al., 2008, for details). Generally, the data are from shallow earthquakes located in tectonically active, shallow lithosphere, with reliable earthquake metadata being available.

Model Development

Dataset Selection

The developers used subsets of the full database, with justifications for the data not used. For example, BA excluded data from aftershocks, records for which metadata were missing or for which only one horizontal component was available, none "free-field" installations, etc. Figure 1 shows the magnitude-distance distribution of data for PGA and for PSA at T=10.0 s period. Note that there are many fewer data at longer periods than at shorter periods, a natural consequence of the low-cut filtering used in processing the data. Also note that there are no normal fault data for T=10.0 s. For these reasons, the GMPEs at longer periods (and for normal faults) will be less certain than at shorter periods.

Functional Forms

Each developer team chose different functional forms for the GMPEs, in order to capture effects that they thought should be modeled. The functional forms are tradeoffs between simplicity of use and being able to represent the complexity in ground motions, due to many physical effects. Some of the effects captured in the functional forms are given below, as divided into source, path, and site contributions:

Source

• Fault mechanism (all developers)

• Aftershock ...