Understanding the Magnitude Dependence of PGA and PGV in NGA-West 2 Data

Abstract

The Next Generation Attenuation‐West 2 (NGA‐West 2) 2014 ground‐motion prediction equations (GMPEs) model ground motions as a function of magnitude and distance, using empirically derived coefficients (e.g., Bozorgnia et al. , 2014); as such, these GMPEs do not clearly employ earthquake source parameters beyond moment magnitude ( M ) and focal mechanism. To better understand the magnitude‐dependent trends in the GMPEs, we build a comprehensive earthquake source‐based model to explain the magnitude dependence of peak ground acceleration and peak ground velocity in the NGA‐West 2 ground‐motion databases and GMPEs. Our model employs existing models (Hanks and McGuire, 1981; Boore, 1983, 1986; Anderson and Hough, 1984) that incorporate a point‐source Brune model, including a constant stress drop and the high‐frequency attenuation parameter κ , random vibration theory, and a finite‐fault assumption at the large magnitudes to describe the data from magnitudes 3 to 8. We partition this range into four different magnitude regions, each of which has different functional dependences on M . Use of the four magnitude partitions separately allows greater understanding of what happens in any one subrange, as well as the limiting conditions between the subranges. This model provides a remarkably good fit to the NGA data for magnitudes from 3< M <8 at close rupture distances ( R rup≤20 km). We explore the trade‐offs between Δ σ and κ in ground‐motion models and data, which play an important role in understanding small‐magnitude data, for which the corner frequency is masked by the attenuation of high frequencies. That this simple, source‐based model matches the NGA‐West 2 GMPEs and data so well suggests that considerable simplicity underlies the parametrically complex NGA GMPEs. Online Material: Figures providing detail on the V S 30 distribution in the subset of the Next Generation Attenuation‐West 2 (NGA‐West 2) data used, the quarter‐wavelength amplifications used in the model, the statistical test for the large magnitude portion of the model, and the magnitude dependence.