Long period amplification of seismic waves inIstanbul metropolitan area

Istanbul is a megacity located very close to the North Anatolian Fault and thus is highly vulnerable to seismic hazards. During the 2019 Mw5.7 Silivri earthquake the recorded displacements show site amplification at long-periods and prolonged duration corresponding to the frequency range of 0.1-1.0 Hz. The excitations in these frequency range cannot be explained with variations in very shallow structure only and requires an analysis that considers deeper sedimentary basins. Amplifications in this frequency range cause resonance effects and serious damage tall and high-rise buildings and infrastructure. The Mw 5.9 Silivri earthquake of September 26, 2019, provided a good dataset for investigating site-specific amplification effects. This thesis analyzes acceleration records from nine AFAD strong-motion stations distributed over different geological units to evaluate the characteristics of long-period amplification. As an initial test, we compare synthetic waveforms using two velocity: (i) a crustal model from Karabulut et al. (2020), and (ii) an AFAD-based station specific shallow velocity model which includes near-surface velocity layers from AFAD station reports. We show that neither of the models work well for amplified waveforms especially on the motion recorded in stations on the European part of Istanbul. Instead we generate synthetic seismograms for each station. We try two approaches. First, we generate synthetic waveforms at the surface of the crustal model which is assumed to represent a hard-rock reference waveform. Then we randomly generate 2 layer structures on top of this structure and optimize the fit to the data in the frequency domain. As a second approach we calculate the synthetic waveforms at a depth of 2 km. We then again optimize for each station the layered structure that represent first two km beneath the station. For each trial we calculate the amplified waveforms using the transfer function of these two layers for SH waves. For each station we generate 5000 velocity models and optimize the fit between the acceleration spectrum of data and synthetics. The best-fitting velocity structure was obtained by spectral fitting of synthetic and recorded waveforms. Results show that the site amplification largely varies with local geological conditions: stations located on stiff geological units (e.g., 3405, 3413, 3417) have minimal amplification, while soft sediment sites (e.g., 3411, 3412, 3416) undergo strong amplification and long shaking. Station 3412 displays anomalous amplification and longer oscillations probably due to very thick unconsolidated sediments or local site complexities. While the AFAD-based velocity model underestimated the observed amplification, the crustal model fits better for hard rock sites but failed to capture the amplification at soft sediment stations. Based on our modeling results, we estimate that stations 3407, 3411, 3413, and 3416, which are situated on soft sediment, have sediment thicknesses ranging between 120 and 200 meters with underlying layers exhibiting shear wave velocities of 248 m/s, 87 m/s, 420 m/s, 188 m/s for first layer, respectively. For the second layer values for each station 254 m/s, 294 m/s, 789 m/s, 220 m/s for station 3407, 3411, 3413, 3416 respectively and the layer thickness for second layer ranging between 200 m to 1000 m. The better fit at a depth of 2 km for these stations suggests that deeper sedimentary layers significantly influence wave amplification. In comparison, AFAD's velocity models provide shear wave velocities of 597 m/s for 3407, 323 m/s for 3411, 452 m/s for 3413, and 420 m/s for 3416 for first layer and 2000m/s for 3407, 523 m/s for 3411, 772 m/s for 3413 and 849 m/s for 3416 for the second layer, indicating some discrepancies between our estimated values and the reference models. These differences highlight the need for further refinement, considering potential lateral variations and deeper structural influences. When compared to AFAD's reference models, our synthetic seismograms show good agreement for stations 3407, 3411, 3413, 3416 where both amplitude and spectral content are well captured. However, discrepancies at other stations suggest that additional factors, such as lateral heterogeneities or path effects, may be influencing the results. In general, for significantly amplified stations on the Anatolian side, it is to refine these models further, it is necessary to incorporate additional earthquake records from different directions to assess whether the observed variations persist across different source locations and azimuths.