Quantitative Uncertainty Estimation for the Coseismic Fault Model and its Slip Heterogeneity Using Real-time GNSS Data

Earthquakes beneath the sea-floor possibly generate tsunamis. After the 2011 Tohoku-Oki earthquake, several researches have stressed on the advantages of real-time GNSS data for rapid earthquake size estimation and its utilization for tsunami early warning. In contrast, when we only use the onshore real-time GNSS data, resolution of the estimated coseismic fault model in the offshore will be degraded. Quantitative the uncertainties estimation will directly relate to uncertainties of the resulting tsunami height. Based on these backgrounds, we try to estimate the quantitative uncertainty for the coseismal fault model by real-time GNSS data. We estimate single rectangular fault model and coseismic slip distribution along the plate interface using Bayesian inversion approach. We constructed a discrete representation of the posterior probability density function (PDF) by sampling with a Markov Chain Monte Carlo (MCMC) method. In addition, considering utilization in real time, we make sampling more efficient using Parallel-tempering approach: the samples can be obtained more effectively compared with single chain MCMC because of a parallel search by the PDF family being annealed by different temperatures.

We applied it to the 2011 Tohoku-Oki earthquake (Mw9.0). In the single rectangular fault model estimation, we successively obtained coseismic fault parameters as PDF. We also found the clear trade-off between the fault dimension and slip amount even though the estimated moment magnitude was stable. This results strongly suggested that the difficulty of the strict separation between the fault dimension and its slip amount using only onshore GNSS data. In the slip distribution, we divided into 482 sub-faults along the subducting plate interface. We found that it is difficult to obtain the reasonable result within realistic computation time even though the using of the parallel tempering approach without appropriate a priori information in each sub-fault. To estimate the realistic coseismic slip uncertainties along the plate interface, appropriate a priori information will be important. We will discuss how to assume and utilize the a priori information for the uncertainties estimation in the presentation.