Long Paleoseismic Records at Plate Boundaries: Clustering, Segmentation, Supercycles and More (Invited)

Long paleoseismic records from plate boundary earthquake series afford uncommon opportunities to examine recurrence models, clustering, segmentation, interaction with other faults, and long term strain history. For example, the 10,000 year marine record in Cascadia suggests at least four seismic segments defined by correlation of turbidites along strike and compatible temporal links to onshore evidence. The linkage makes use of subsurface log correlation techniques, radiocarbon ages, and the similarity of records in isolated settings along the Cascadia margin. The temporal and spatial record reveals a pattern of decreasing recurrent time southward, from ~ 500-530 years in northern Cascadia, to ~ 240 years or less in southern Cascadia. The decrease in recurrence times southward is consistent with a southward tapering sediment supply, which in turn allows lower plate roughness to define segment boundaries. The temporal record also reveals clusters of 4-5 events, with gaps of 700-1200 years between clusters. These temporal clusters are supported by regional high-resolution seismic reflection records, and appear to extend to Late Pleistocene time, deeper than reachable with gravity and piston coring. Long temporal records also afford opportunities to test linkages between adjacent fault systems. Long turbidite records for Cascadia and the Northern San Andreas Fault suggest that these two faults have virtually the same average recurrence interval through the Holocene. The physical connection at the Mendocino Triple Junction may be augmented by a stress connection between the earthquake cycles of these two great faults. This connection is suggested by a close event by event temporal correlation. Finally, energy in plate boundary fault zones may have long term cycling involving many seismic cycles. Several lines of evidence, both direct and indirect suggest that the connection between interevent time and earthquake magnitude, and models predicated on this relationship, may be very weak. If a measure of earthquake magnitude can be inferred from a paleoseismic record, then recurrence models and long term-cycling can be explored. In Cascadia, coseismic energy may be modeled as proportional to the mass of turbidites triggered in seismic shaking. We infer that turbidite mass is a suitable proxy for energy release because of its consistency along strike at multiple sites. If turbidite mass (energy release) is plotted to balance plate convergence (energy gain) the result is a 10ka energy time series for Cascadia. The pattern reveals that the earthquake clusters apparent in the time series have variable behavior. Some events appear to release less, while others release more energy than available from plate convergence (slip deficit). Those that are larger may have borrowed stored energy from previous cycles. Cycle variations may explain mismatches between deformation models based on interevent times in the last 4600 years and coastal paleoseismic data.