Spatial and temporal distributions of shear wave anisotropy and analysis of repeating earthquakes in the Karadere-Duzce branch of the north Anatolian fault

We analyze crustal shear wave anisotropy from seismograms recorded by a PASSCAL seismic network deployed along and around the Karadere-Duzce branch of the north Anatolian fault for about 6 month, starting a week after the August 17, 1999, Mw7.4 Izmit earthquake. On November 12, 1999, the Mw7.2 Duzce earthquake started and propagated eastward from the Karadere-Duzce fault. Our temporary seismic network straddles the eastern end of the Izmit earthquake and the western end of the Duzce earthquake and recorded about 26000 earthquakes. We use the technique of Silver and Chan (1991) to estimate the fast polarization direction and delay time of the splitted shear waves. Stations at several km off the fault generally have fast polarization direction sub-parallel to the regional tectonic stress direction (roughly NW-SE). Stations within several hundred meters of the surface rupture show polarization directions that are sub-parallel to and change with the local fault strike. This suggests stress-induced cracks aligned by nearby faulting during a major earthquake as a source for the observed anisotropy for these stations. The time delay shows no systematic relationship with either focal depth or hypocentral distance. This indicates that seismic anisotropy in our study area is confined primarily to the top 3-4 km of the shallow crust. In an effort to detect temporal changes associated with the occurrence of the Izmit and Duzce earthquakes, we perform analysis of shear wave splitting of repeating earthquakes in our study area. Repeating earthquakes are identified using an equivalency class algorithm (e.g., Aster and Scott, 1993). The similarity measure is based on the mean cross-correlation values of all waveforms between event pairs. The waveform cross-correlation is performed over a time window of 1 sec before and 5 sec after the P arrivals. Our data set can be divided to fault zone events that are in the vicinity of the Karadere-Duzce branch of the Izmit rupture zone and the reminder regional events (located around Stations CH and BV). Depending on the similarity criteria, approximately 15-40% of events in the fault zone set belong to similar event clusters. The percentage is about 8-20% for the regional events off the fault. The analysis done so far does not show a clear precursory change before the Duzce earthquake. However, splitting measurements from several repeating earthquake clusters indicate slight changes in both fast polarization direction and delay time before and after the Duzce earthquake. The observed temporal changes might be caused by the increasing of the crack density due to the static stress change or dynamic shaking effect associated with the Duzce earthquake. Updated results will be presented in the meeting.