Near-surface properties using seismograms from the GONAF-Tuzla vertical array, SE Istanbul

Monday, 15 December 2014: 9:30 AM
Christina Raub1, Peter E Malin2, Marco Bohnhoff1,3, Fatih Bulut1, Georg H Dresen1, Tugbay Kilic4, Recai F. Kartal4, Filiz Tuba Kadirioglu4 and Murat Nurlu4, (1)Helmholtz Centre Potsdam GFZ German Research Centre for Geosciences, Potsdam, Germany, (2)University of Auckland, Auckland, New Zealand, (3)Free University Berlin, Institute of Geological Sciences, Berlin, Germany, (4)Disaster and Emergency Management Presidency (AFAD), Earthquake Department, Ankara, Turkey
As part of the ICDP-GONAF project (Geophysical Observatory at the North Anatolian Fault) geophone arrays have been installed in 300 m deep boreholes around the eastern Sea of Marmara. The objectives of GONAF are to (1) monitor the North Anatolian Fault Zone (NAFZ) at the transition from the 1999 Izmit rupture to the Princes Islands segment offshore Istanbul, where a M ~ 7 earthquake can reasonably be expected to occur and (2) to determine ground-motion amplification and near-surface properties at the GONAF sites. Here we use recordings from the first GONAF borehole on the Tuzla peninsula in eastern Istanbul. The array consists of one 1 Hz 3C Mark Products L4 seismometer at the surface, three 1 Hz vertical Mark Products L4 seismometers at 75 m depth-spacing, and 2 Hz and 15 Hz 3C Geospace HS-1 and DM2400 seismometers at 288 m depth. During April - May 2013 this array recorded a microearthquake swarm located ~ 3.5 km epicentral distance south of the Tuzla site. By cross-correlating the continuous Tuzla data with the only swarm event detected by the regional network (20th of April 2013, Md 1.6) we retrieved an additional of 113 events. The swarm and an additional 15 events located throughout the eastern Marmara region were used to analyze the near-surface properties of the Tuzla site. We derive a velocity model from sonic-log measurements which were conducted upon completion of the Tuzla well. This model is confirmed by modeling travel-time curves with forward ray-tracing and by an analysis of spectral interference effects. In the uppermost ~80 m at this site we observe an unexpectedly high velocity lid (VP ~ 4000 m/s) over a much slower (VP ~ 3000 m/s) ~50 m thick zone. These velocity structure leads to interference effects between up and downgoing waves even at the deepest stations. As a result of this the standard spectral ratio technique for deriving the site response becomes poor approximation. However, waveform deconvolution techniques allow determination of the seismic propagator matrix, by which the downhole signals can be deconvolved using the surface signal, making it possible to derive the near surface properties by an inversion method. The results show that while signal amplitudes of the upcoming body waves are noticeably reduced, the unusual velocity structure results in unexpectedly large surface displacements.