Earthquake Relocation and Error Estimate for Aftershocks of 2003 M 7.0 Northeastern Japan Earthquake

We investigate the aftershock sequence of a M 7.0 earthquake that occurred on 26 May, 2003 in northeastern Japan, where the characteristic double seismic zone (DSZ) was first identified by Hasegawa in 1978. The preliminary locations(from Tohoku University) show that these aftershocks are distributed between 65km-80km depth, not along the upper plane of the DSZ, but cut cross the slab at a steep angle. Catalog data from 227 aftershocks and waveform data from 69 large aftershocks (M>3.8) are used to refine earthquakes location. After double-difference earthquake location (hypoDD), the aftershocks are found to distribute along plane with dip angle about 53 (degree), but with significant scatter (RMS deviation is about 1.524km).These aftershocks are right along the eastern coast of Japan, in an area with significant velocity heterogeneity. We test the ability of hypoDD to recover the correct locations using synthetic data without and with noise added. The synthetic data are created based on the real data with the same set of event-station observation but with varying geometry. First, all the aftershocks are projected onto a plane (strike angle = -10 (degree), dip angle =53.79 (degree)) to generate the synthetic events, and synthetic travel time are calculated with 3-D velocity model. Several variations are then created to estimate the relocation error, and examine the effects of initial source locations, Gaussian noise and correlated noise on the relocation results.The results show that hypoDD does well in recovering the correct locations (RMS∼=224m) and the initial source locations have little effect on relocation if arrival time errors are small. The absolute relocation error mainly comes from model error (true model is 3-D, hypoDD model is 1-D), and Gaussian noise represents the noise well.