What Controls the Duration of Aftershocks, and Why It Matters for Probabilistic Seismic Hazard Assessment

Abstract:

A fundamental problem confronting hazard modelers in slowly deforming regions such as the central and eastern United States, Australia, and inner Honshu, is whether the current seismicity represents the steady state earthquake potential, or is instead a decaying potential associated with past mainshocks. If the current seismicity were composed of long-lived aftershock sequences, it might then be anti-correlated with the next large earthquakes. While aftershock productivity is known to be a property of the mainshock magnitude, aftershock duration (the time until the aftershock rate decays to the pre-mainshock rate) should, according to rate/state friction theory of Dieterich[1994], be inversely proportional to the fault stressing rate. If so, slowly deforming regions would be expected to sustain long aftershock sequences. Most tests have supported the Dieterich hypothesis, but use ambiguous proxies for the fault stressing rate, such as the mainshock recurrence interval. Here we test the hypothesis by examining off-fault aftershocks of the 2011 M=9 Tohoku-oki rupture up to 250 km from the source, as well as near-fault aftershocks of six large Japanese mainshocks, sampling a range of receiver faults, from thrusts slipping 80 mm/yr, to normal faults slipping 0.1 mm/yr. We find that aftershock sequences lasted a month on the fastest-slipping faults, have durations of 10-100 years on faults slipping 1-10 mm/yr, and are projected to persist for at least 200 years on the slowest faults. Although the Omori decay exponent for short and long sequences is similar, the very different background rates account for the duration differences. If the stressing rate is generally proportional to fault slip rate, then aftershock durations indeed support the Dieterich hypothesis. The test means that the hazard associated with aftershocks depends on local tectonic conditions rather than on the mainshock magnitude alone. Because declustering approaches do not remove such long-lived aftershocks, these shocks can be misinterpreted as independent events. Aftershock sequences may thus masquerade as background seismicity and misguide hazard assessments in slowly deforming regions. Depending on how the hazard is calculated, this can lead to over- or underestimates of hazard in aftershock zones.