I outline a referenced empirical approach to the development of ground-motion prediction equations (GMPEs). The technique is illustrated by using it to develop GMPEs for eastern North America (ENA). The approach combines the ENA ground-motion database with the empirical prediction equations of Boore and Atkinson (2008) for the reference region of western North America (WNA). The referenced empirical approach provides GMPEs for ENA that are in agreement with regional ground-motion observations, while being constrained to follow the overall scaling behavior of ground motion that is observed in better-instrumented active tectonic regions. They are presented as an alternative to the commonly used stochastic ground-motion relations for ENA. The motivation of the article is not to supplant stochastic GMPEs but is rather to consider other approaches that might shed light on their epistemic uncertainty.
Differences between the referenced empirical GMPEs of this study and the stochastic GMPEs of Atkinson and Boore (2006), along with inconsistencies between both of these studies and inferences based on intensity observations, suggest that uncertainty in median ENAGMPEs is about a factor of 1.5–2 for M≥5 at distances from 10 to 70 km. Uncertainty is greater than a factor of 2 for large events (M≥7) at distances within 10 km of the source. It may be that saturation effects not modeled in the stochastic predictions, but inferred from observations in other regions, cause overestimation of near-source amplitudes from large events in Atkinson and Boore (2006). On the other hand, these saturation effects cannot be directly verified in ENA data. Differences in predictions according to the approach taken are also significant at distances from 40 to 150 km, due to uncertainty in the shape of the attenuation function that will be realized in future earthquakes.