Search

Throughout the last century, the North Anatolian Fault (NAF) produced mostly west migrating sequence of M>7 earthquakes, and in this period of time, the NAF did not break beneath the Marmara Sea. The Main Marmara Fault (MMF), which is the north branch of the North Anatolian Fault in the Marmara Region, is the most active branch in this area. The interseismic behavior of the MMF beneath the Marmara Sea is crucial with regards to assessments for the earthquake risk of the Marmara Region which has a population over 24 million. As a result, the prediction of the locking depth and the fault slip rate have great importance as they provide a comprehension concerning the seismic moment deficit, that can be released at the time of a future event.In this study, we modeled the interseismic locking of the MMF by making use of the interseismic GPS velocities and implementing a 3D finite-element analysis. Our kinematic model has a realistic 3D fault geometry, in which each and every fault section of MMF is confined to a fault slip rate down below a given locking depth which ranges from 2.5 to 15 km. We improve the fits to GPS velocity data just by changing the interseismic locking depth of each segment and placing more complexity to the locking structure in cases where the data needs. Our preliminary models reveal that a change in locking depth is required in between the Ganos Segment at the western end of the Marmara Sea and Central Segment. A model with a locking depth of 10 km for the Ganos Segment and 2.5 km for the Central Segment fits the GPS velocities well. We outrule the likelihood of similar locking patterns for the two segments. This result is consistent with the earlier work utilizing 1D profiles demonstrating that there is no significant strain accumulation beneath the Central Segment and also seismicity studies show that the background seismicity is very different in these two segments. Under the Princes' Islands Segment, the data is more compatible when we add some shallow creep. For the Izmit Segment, we predict that the locking depth is around 8 km, yet further analysis is needed. Our results show that in most cases, the variation in the seismic behavior of each segment is in accordance with its interseismic behavior in relation to the fault locking. [ABSTRACT FROM AUTHOR]

It has been two decades since the publication of the 5-degree global crustal model CRUST5.1in 1998. More recently, the 1-degree model CRUST1.0 was released for wide use in 2013.The more recent model was based on a comprehensive, updated database of crustal structureinformation from active source experiments as well as receiver function studies, Moho depthmaps from regional seismic tomography and, in some instances, results from gravity studies.A change in strategy to compile CRUST1.0 also was an iterative model validation andsubsequent update to fit global surface wave dispersion data. Yet, in an attempt to updateCRUST1.0 to the final version, some discrepancies with surface wave constraints remained.This project zooms in on the Mediterranean Sea and compares model predictions withobservations. CRUST1.0 subsequently was used as a starting model to compile a global model of theentire lithosphere, LITHO1.0 that was released in 2014. This model is based on surfacewave data alone, with an updated and larger database that includes proprietarydata in the Eastern Mediterranean/Asia Minor area. While intermediate-period(40 – 29 s) Rayleigh wave phase velocity is well-explained by CRUST1.0, groupvelocities are underpredicted in the western Mediterranean and overpredicted inthe eastern Mediterranean, by up to nearly 10%. Both CRUST1.0 and LITHO1.0fail to model Love wave group velocities adequately, even at long periods (T> 50s). Recently, new surface wave constraints collected around the Mediterranean area allowtomographic imaging to unprecedented detail. The extension to shorter periods than what isavailable on global scale allows assessment of structure not only on the crust in general butalso in areas with thick sediment cover, such as the Nile Delta in the eastern Mediterranean.At this level, we validate the 1997 Laske and Masters global sediment model that wasadopted into CRUST1.0. [ABSTRACT FROM AUTHOR]

Turkey poses a complex crustal structure as a result of the convergence of the Arabian and African plates. The North Anatolian Fault Zone (NAFZ) represents 1600 km long active right-lateral strike-slip fault system of the country that has caused several destructive earthquakes. Location of historical earthquakes indicates that the activity on the fault tends to migrate westward. In the western part of NAFZ, there is a seismic gap inside the Sea of Marmara. Reliable knowledge of physical properties of the crustal structure of the unruptured segment within the Sea of Marmara is crucial for proper seismic hazard assessments considering that the megacity Istanbul with its millions of habitants, is closely located to this seismic gap. In this study, ambient seismic noise tomography, for the first time, has been applied to obtain a high-resolution image of the crustal structure of northwestern Turkey. For forward estimation of traveltimes, we employed the Fast Marching Method. A Bayesian framework was utilized to invert traveltime residuals extracted from Green's functions for the Rayleigh wave group velocity models at various frequencies. Besides the Transdimensional Bayesian tomography imaging, we also applied a rather simple and computationally fast 2-D nonlinear tomography scheme on the same data-set to gain a rough idea on group velocity variation in the regions. Our high-resolution tomographic images included group velocities of Rayleigh waves resolved between 1 and 20 s. Lateral variation of group velocities successfully revealed the basin structures such as Çınarcık, Tekirdağ, and Adapazarı Basins between periods 1 and 10 s that are characterized by typically low group velocities with low uncertainties. The low-velocity zones appear to mark well weakness zones along the NAFZ in the crust and the complex structure due to the active tectonics of the Anatolia. Finally, different tectonic units, in particular, to the east were identified from each other as our maps highlight Istanbul-Zonguldak Zone from Armutlu-Almacik Block with a sharp high to low-velocity change in N-S direction. [ABSTRACT FROM AUTHOR]

In this study we present the results of an ongoing 3D full-waveform inversion for theArabian-Eurasian collision zone in eastern Turkey and Iran. This is intended to extract asmuch information as possible from available seismograms in order to constrain seismicstructure in both the crust and the upper mantle. In our method we simulate the 3D visco-elastic wavefield using the newly developedspectral-element solver Salvus. Our numerical mesh honors topography of the surface andinternal discontinuities. We compare observed and synthetic waveforms using time frequencyphase misfits. Using adjoint techniques, we then compute sensitivity kernels with respect tothe model parameters, which are VSV, VSH, VPV, and VPH. Finally, the kernels enable theiterative solution of the nonlinear inverse problem with the help of the L-BFGSalgorithm. For this study we obtained seismic waveform data of 143 earthquakes within themagnitude range of Mw 4.5 to 6.3 that occurred in the region between 2012 and 2016. Theseevents were recorded by 271 broadband seismic stations belonging to the two nationalIranian networks and freely available seismic stations of the Turkish Network, madeavailable by IRIS. For data management and workflow organisation we employ theLarge-Scale Seismic Inversion Framework (LASIF), modified to operate with theAdaptable Seismic Data Format (ASDF), which facilitates simulations on large HPCclusters. Starting from the first generation of the Collaborative Seismic Earth Model, we firstconstrain longer-wavelength structure. To this end, we consider 3-component recordingsfrom a subset of 41 events in the period range from 50 to 80 s. This period range issuccessively broadened to include shorter periods; for instance, the lower end of theperiod range was meanwhile extended to 35 s. For each period band, the number andthe length of measurements are increased to ultimately comprise almost completeseismograms, and also the number of events is increased to use complete of dataset. [ABSTRACT FROM AUTHOR]