Statistics of Long Paleoearthquake Records: Comparison of San Andreas Fault and Dead Sea Fault Behavior

Long paleoseismic records provide great opportunities to explore recurrence patterns of large, ground-rupturing earthquakes because the length of the record tends to compensate for uncertainty in paleoearthquake identification and dating. In this study, we apply two nonparametric tests that distinguish periodic and clustered patterns from Poisson (random) sequences to records from the San Andreas and Dead Sea faults. Although lithologically more complex, the San Andreas fault has a ~6X faster slip rate, >3X greater total slip and a simpler fault geometry than the Dead Sea fault, and thus offers a useful contrast for examining the effect of rate-dependent processes and fault evolution on recurrence behavior. The longest record for the San Andreas fault is from the Wrightwood site. This record is complete for two periods, 5,000 to 3,500 yr BP and 1,500 yr BP to present, which together include up to 29 events >M6.5 (Scharer et al., 2007). The tests show this record has a quasi-periodic recurrence that is too regular to be the likely result of a Poisson distribution and no evidence of tempo changes suggestive of clustering. To explore the sensitivity of these results with respect to the geologic record, we also tested alternate paleoseismic chronologies in which up to five paleoearthquakes with weaker evidence were removed and up to five synthetic earthquakes were added to stratigraphic intervals during which preservation of ground rupture could be poor. All of the alternate chronologies also exhibited quasi-periodic recurrence, indicating that the record is sufficiently long that a robust pattern is resolved. For comparison, we also tested two long paleoearthquake records from the Dead Sea fault. In the southern Arava Valley the Elat fault has a ~45,000 year record of 20 ~Mw 5.9-7.1 paleoearthquakes (Hamiel et al., 2009). Our test results support a clustered recurrence largely due to a period of quiescence between 30,000 and 18,000 yr BP and show no evidence of periodic behavior. In the Dead Sea area, a combination of paleoseismites, cave damage, and historic earthquakes creates a 16 event record of >M6.5 earthquakes since 53,500 yr BP (Hamiel et al., 2009; Begin et al., 2005; Kagan et al., 2005). Our test results confirm previous observations of strong clustering due to a change from fast to slow recurrence around 46,500 yr BP. Considering their differences, a quasi-periodic San Andreas fault but clustered Dead Sea fault record support laboratory and modeling results in which fault maturity and/or high loading rate contribute to a greater predictability in earthquake recurrence.