Probabilistic Seismic Source Inversion of the 1886 Charleston, South Carolina, Earthquake from Macroseismic Evidence: A Major Updating

Abstract

The source of the 1886 Charleston, South Carolina earthquake influences the computed seismic hazard of the Southeastern U.S. and thus impacts public policy and engineering practice. However, because the 1886 rupture predated seismic instruments, its source is highly uncertain. This study presents probabilistic seismic-source inversions of the Charleston earthquake from liquefaction evidence and historical intensity reports. Using the latest predictive models and a novel inversion approach, we seek to constrain the magnitude, location, and orientation of the 1886 rupture. Probability distributions of rupture magnitude are conditioned on both the “Woodstock Fault” – a commonly inferred source of the 1886 event – and on an unknown source, wherein the uncertainties of fault location and orientation are considered. These distributions are compared to the M_w6.7-M_w7.5 distribution adopted by the U.S. National Seismic Hazard Model Project (NSHMP). Collectively, the results do not provide strong support for the hypothesized Woodstock Fault. This is not to say the Woodstock Fault does not exist, but rather, that the position of the 1886 source model cannot be constrained by the data and models studied herein, given the large uncertainties inherent to each. While this is at odds with the underlying assumption of many prior studies, the results nonetheless generally uphold the magnitude distribution assumed by the NSHMP. The largest uncertainties inherent to this distribution are identified and could be diminished in the future. Finally, we note that the inversion methodology used here is not specific to any region, or to certain types of evidence, but can be applied to any seismic zone and to any co-seismic response. This methodology allows for uncertainty to be accounted for in a more complete and transparent manner when inverting seismic source parameters from macroseismic data. Of course, any limitations, biases, or unmodeled uncertainties inherent to these data must be understood, and their implications acknowledged, as further discussed herein.