Poster #090, Earthquake Forecasting and Predictability (EFP)

Statistical Analysis of Paleoseismically Determined Earthquake Recurrence

Poster Image: 

Poster Presentation

2020 SCEC Annual Meeting, Poster #090, SCEC Contribution #10496
We present a statistical analysis of 11 paleoseismic data sets from continental strike-slip faults in California (San Andreas [5], San Jacinto [2], Garlock [2]) and China (Altyn Tagh [1], Haiyuan [1]) to reveal general characteristics of earthquake recurrence. Recurrence intervals are sampled from the distribution of event ages, as constrained from bounding layer ages and stratigraphic ordering. We find that recurrence intervals show no central tendency, contradicting the characteristic earthquake recurrence model. Instead, the recurrence interval distributions are long-tailed, with the average greater than the median, and the longest recurrence interval >5x the shortest interval. We mod...el each distribution using Normal, Exponential, Brownian passage-time (Matthew et al., 2002), and Weibull distributions. We find that the normal distribution poorly fits earthquake recurrence, as expected from the lack of a central tendency. A minority of data sets are well fit by the exponential distribution, consistent with a Poisson process with uniform hazard over time and no memory of prior events. Most of the paleoseismic records, however, indicate an increase in hazard over time. We find that the Brownian passage-time model, formulated as perturbations added to steady tectonic loading, systematically underpredicts the young tail of earthquake recurrence at all but two sites. We find better fits to the majority of paleoseismic data sets using a Weibull distribution. The Weibull model may be envisioned as arising from an ensemble of many possible failure points (earthquake nucleation sites) along a given length of the fault, at each of which the hazard is growing over time. Once initiated, a rupture grows and propagates through the point of observation at a paleoseismic trench. Earthquake distributions vary widely along the length of the 1857 Fort Tejon earthquake rupture and suggest that interactions with the Garlock Fault and the San Jacinto Fault affect the time to failure on the San Andreas fault.