Revisiting the early postseismic deformation of the 2003 Tokachi-oki earthquake

A logarithmic function is the popular model of temporal evolution of afterslip, derived from the rate-and-state friction law (RSF) under the steady-state assumption (Marone+1991JGR). Relaxing this assumption, self-accelerating aseismic slip is predicted prior to subsequent decay even with velocity velocity-strengthening setting (i.e., a–b> 0; PerfettiniAmpuero2008JGR). The only natural observation example of such an accelerating stage of afterslip following large earthquakes is the case of the 2003 Tokachi-oki earthquake (M 8.0) in Japan, presented by Fukuda2009JGR (F09) with the data analysis performed by LarsonMiyazaki2008EPS (LM08). They reported that the early postseismic deformation emerged ~1 hour after the mainshock. We revisit this earthquake’s early postseismic deformation with a modern kinematic GNSS processing workflow by Gipsy-X v2.3 because many default and/or recommended settings and products have evolved from the time when these previous works were carried out. This revisit will align the Tokachi-oki case with other earthquake cases analyzed by GNSS processing strategies closer to ours than LM08’s.

Among all the parameters/settings of GNSS processing we tested, the most impactful parameter was the position random walk (RW) parameter. We tested a wide range of values from 1 to 1e-5 m/sqrt(s) for this parameter with switching to the white noise during the mainshock and the M 7.1 largest aftershock (1.3-h later). Comparing our test results with F09’s dataset, the largest mismatch was found between the mainshock and the 7.1 largest aftershock when we attempted to reproduce F09’s cumulative displacements. During this interevent window, F09’s dataset shows tiny deformation, while our solutions show significant deformation. On the other hand, our test solutions exhibit the acceleration at similar timings as F09’s, with the RW parameter same as F09’s (1e-5 m/sqrt(s)), but our cumulative displacements are much smaller than F09’s after the largest aftershock coseismic step was removed. This is because of a trade-off between early postseismic deformation and the largest aftershock step, caused by the very tight RW not allowing sites to move other than at the coseismic timing. Therefore, we recommend careful testing position RW parameter to accurately resolve early postseismic deformation, rather than taking a value introduced in other studies. With our test results, we concluded that no parameters could satisfactorily reproduce the early postseismic deformation presented in F09; in other words, the acceleration of early afterslip reported in F09 was absent in our solutions. Our results imply that the transition between the interseismic and postseismic stage of velocity strengthening faults would happen within several minutes at the longest, implying that the very beginning of afterslip is concurrent with the dynamic ruptures of the mainshock.