Analysis of Regional Ground Motion Variations for Engineering Application

Abstract

An important question for many ground motion hazard analyses is the degree to which ground motion prediction equations (GMPEs) developed for one region may have bias for a different region. A closely related problem is the applicability of multi-regional GMPEs to a particular region, even if that region contributed some fraction of the database. It is well known that ground motions show distinct characteristics for stable continental regions, subduction zones, and active tectonic regions with shallow crustal earthquakes. Here I consider variations among active regions with shallow crustal earthquakes. For such regions having sufficient data that meaningful comparisons are possible, I review four approaches for evaluating regional variations: (1) direct comparisons of medians from GMPEs; (2) analysis of variance; (3) overall goodness of fit metrics; and (4) verification of specific GMPE attributes relative to regional data. For engineering application, the objective of the comparison should be to evaluate whether median predictions show statistically similar trends with respect to magnitude-scaling, distance-scaling, and site effects across the range of magnitudes and distances controlling the seismic hazard, as well as consistent standard deviation terms.