Foreshock Cascades to Failure

Abstract:

Our current understanding of how earthquakes nucleate leaves open critical questions about the physical processes that occur before dynamic rupture including whether or not foreshocks have other than weak statistical value as a precursor. The preslip and cascade models are two end-member hypotheses that take opposing views on the role of aseismic deformation in the nucleation process. Here we examine foreshocks to several well-observed earthquakes including the M_w 7.6 1999 Izmit, Turkey (Ellsworth and Bulut, 2018), M_w 7.1 1999 Hector Mine, California (Yoon et al., 2019), M_w 6.5 2016 Norcia, Italy, M_w 6.4 2019 Ridgecrest, California and M_w 5.8 2020 Owens Lake, California earthquakes. We used template matching to identify the events, cross correlation and first-motion timing with hypoDD to precisely determine hypocentroids and the point of initial mainshock rupture, and spectral ratio measurements to determine foreshock source dimensions.

All of these earthquakes began abruptly without direct evidence of a nucleation process in the initial P-wave arrival for over 6 orders of magnitude below the mainshock. In each case, the mainshock nucleated on the edge of the foreshock rupture area, in agreement with the cascade model. The observations also agree with the predictions of rate-and-state theory if the laboratory measured meter-scale dimension of the nucleation zone applies to faults in nature.

Constraining aseismic slip is notoriously difficult. For the M 6.4 July 4, 2019 Ridgecrest earthquake we did not detected slow seismic or aseismic slip before rupture initiation on either seismometers or a borehole tensor strainmeter located 18 km from the epicenter. We can, however, place a limit on undetectable aseismic slip during the half hour between the M 4.0 foreshock and mainshock. A strain with an amplitude greater than 0.2 nanostrain would have been observed, corresponding to an upper limit for aseismic slip equivalent to M 3.5. We searched for evidence of repeating foreshocks but found none. If any occurred, they would be M > 0.6 with an average slip of a few mm. Consequently, if aseismic slip occurred it played at most a minor role in this foreshock sequence and we can rule out slip acceleration as the time to failure approached. Collectively, the evidence from these foreshock sequences strongly favors the cascade model.