[P03] 地殻変動・GNSS・重力

[P03-02] 2003年十勝沖地震の初期余効変動再訪

*伊東 優治¹、Twardzik Cédric²、Vergnolle Mathilde²、Maubant Louise¹ (1. 東京大学地震研究所、2. Géoazur)

Large earthquakes induce postseismic deformation, which is a relaxation process of coseismic elastic stress change. High-precision daily GNSS coordinates have been typically employed to quantify and model postseismic deformation. However, daily coordinates cannot access the earliest stage of postseismic deformation during 12 to 36 hours following the mainshock (e.g., Twardzik et al., 2019). Afterslip at the earliest stage within several days following the mainshock might offer observational constraints to the existence of transient acceleration of afterslip predicted by the full rate and state friction law solution without the steady-state approximation (Perfettini & Ampuero, 2008, Helmstetter & Shaw, 2009). Despite vigorous investigations on the early postseismic deformation following many large earthquakes, very limited examples on such acceleration are observationally reported. A famous example is a transient acceleration of postseismic deformation delayed by about 1.2 hours following the 2003 M_w 8.0 Tokachi-oki earthquake in northeastern Japan, reported by Larson and Miyazaki (2008) and modeled by Fukuda et al. (2009). In this study, we revisit their postseismic deformation using modern GNSS analysis techniques to validate their observational findings.We employed Gipsy-X version 2.3 to perform Precise Point Positioning analyses of coordinates before and after the mainshock. The setting we primarily investigated is the filtering algorithm of temporal evolution of position, in particular the random walk scale of Kalman Filter – smoother algorithm. Our preliminary analysis results did not quantitatively reproduce Larson and Miyazaki (2008)’s early postseismic deformation: (1) although the acceleration of postseismic deformation following the largest aftershock appeared with the same random walk parameter as theirs (1e-5 m/sqrt(s)), the cumulative deformation over the 5 hours was significantly underestimated than theirs. (2) With a larger random walk parameter (e.g., 1 m/sqrt(s)), we were able to see the similar amount of deformation over the 5 hours, but our postseismic deformation started right after the mainshock, lacking the temporal delay of its acceleration. Our results indicate that a choice of random walk scale may significantly bias the temporal evolution of early postseismic deformation.