Testing and Reconciling Stress Drop and Attenuation Models for Southern California

Peter M. Shearer, Rachel E. Abercrombie, & Daniel T. Trugman

Submitted August 14, 2017, SCEC Contribution #7566, 2017 SCEC Annual Meeting Poster #054

Earthquake stress drop is a fundamental source parameter, implicit in many SCEC science goals. It is relatively easy to estimate from seismic data, but hard to measure reliably. The large uncertainties and scatter in results affect strong ground motion prediction, and also limit our understanding of the physics of earthquake rupture. We study two complementary approaches to investigate sources of consistency and discrepancies in stress drop estimates, and quantify uncertainties: (1) the spectral decomposition method of Shearer et al. [2006] and Trugman and Shearer [2017], a large-scale, automated approach involving stacking and averaging spectra to obtain parameters for large numbers of events; and (2) the more traditional empirical Green's function (EGF) method of Abercrombie [2014] and Abercrombie et al. [2017], a focused approach that attempts to obtain optimal results for a small number of the best-recorded earthquakes. Our longer-term aim is to develop improved methods to estimate more accurate and reliable stress drops, together with regional attenuation and site effects.

We apply the two approaches independently to over a thousand earthquakes in two test regions, one near the 1992 Landers earthquake epicenter and one around the Cajon Pass borehole. We find strong correlation between the results, but also significant differences of up to a factor of ~1.5 in corner frequency. The results from the focused approach have a larger range of corner frequency and stress drop compared to the large-scale stacking approach. Disagreement between the methods appears largely related to station selection and signal-to-noise criteria (particularly for less well recorded events), and misfit of the source model, rather than any fundamental flaw or bias in either method. We repeat the analyses using the same calculated input spectra to ensure this is not an issue. We implement bootstrap resampling approaches, and vary the available frequency bandwidth, to compute more realistic uncertainties for our corner-frequency estimates. Varying the tradeoff between self-similarity and spectral fall-off in the spectral decomposition method results in a frequency-dependent effect on the mean stress drops, and on the shape of the source spectra. We also study the possibility of using the best-recorded and modeled events from the individual EGF approach to help constrain the tradeoffs between these parameters in the spectral decomposition analysis.

Key Words
stress drop, attenuation, corner frequency

Shearer, P. M., Abercrombie, R. E., & Trugman, D. T. (2017, 08). Testing and Reconciling Stress Drop and Attenuation Models for Southern California. Poster Presentation at 2017 SCEC Annual Meeting.

Related Projects & Working Groups