MCMC法による地震後変動モデルの推定手法の開発

抄録

The latest Geodetic Reference Frame （GRF） ITRF2014 has been available since Jan. 2016. In ITRF2014, Post Seismic Deformation （PSD） models were introduced in coordinate time series models to account for coseismic and postseismic deformations caused by major earthquake. PSD models are indispensable in modeling coordinate time series of Japanese GNSS observation network （GEONET） stations since they are subject to active crustal deformations. In this paper, we developed Bayesian method to estimate PSD models that was composed of non-linear functions based on ensemble Marcov Chain Monte Carlo （MCMC） method. We compared the estimated PSD model and PSD model of ITRF2014 products at TSKB station, which is one of IGS GNSS stations. We found that standard deviation of differences between the estimated PSD model and observation was as small as （E, N, U） ＝ （7.9, 13.2, 1.6） mm, indicating that the estimated PSD models sufficiently explain observation data. We also evaluated prediction capability of the PSD model in Miyako station, which had been severely impacted by the 2011 off the Pacific coast of Tohoku Earthquake. We estimated the PSD model of Miyako station with observations spanning two years from the earthquake （fitting period） and compare estimated PSD model with observations for post-fitting period. As a result, we found that the differences were smaller than 1 cm for horizontal component and 2 cm for vertical component from end of fitting period to 2 years later if Log＋Exp＋Log model was employed.