SUPPLEMENTARY INFORMATION doi: 10.1038/ngeo739 Supplementary Information to variability and distributed deformation in the Marmara Sea fault system Tobias Hergert 1 and Oliver Heidbach 1,* 1 Geophysical Institute, Universität Karlsruhe (TH), Hertzstr. 16, 76187 Karlsruhe, Germany. * now at GFZ German Research Centre for Geosciences, Telegrafenberg, 14773 Potsdam, Germany. Fault geometry We constructed the fault geometry using fault maps 7,15 and results from seismic imaging 6,14,15,16. We assumed a through-going MMF that is locally non-vertical along the edges of the basins in its upper part but vertical below 15 km bsl. As a major plate boundary fault it extends to the bottom of the model at 38 km depth. The Middle branch of the NAF extends to 20 km depth and the other faults to 15, 10 and 7.5 km, with the longer faults extending deeper than the smaller ones. Faults dip steeply at angles between 70 and 90, with faults on the northern shelf dipping to the south and those south of the MMF to the north. Rock properties The basement-topography and Moho were constructed using the constraints from seismic velocities and from reflections/refractions of seismic waves 6,14. Representative values for elastic parameters and density are assigned to sediments, basement and mantle, respectively (Suppl. Table 1). Sediments Basement Upper Mantle Density [g/cm³] 2.2 2.65 3.3 Poisson s ratio 0.35 0.25 0.25 Young s modulus [GPa] 10 70 150 Modelling interseismic velocities and fault slip rates A coefficient of effective friction of µ =0.05 is assigned to fault portions deeper than 15 km bsl. Above this depth we consider two cases. The first is the locked-fault case with infinite fault friction in the seismogenic layer to simulate interseismic deformation for comparison with GPS observations (Suppl. Fig. 1a). The second is the unlocked-fault case with uniform friction of µ =0.05 throughout the whole depth range of the fault in order to obtain the long- nature geoscience www.nature.com/naturegeoscience 1
supplementary information doi: 10.1038/ngeO739 term fault slip rates (Suppl. Fig. 1b). The value of µ =0.05 was found appropriate for the NAF 23,24. In contrast to the common approach 25 we do not impose slip on the faults below the seismogenic depth. Instead, boundary conditions at the sides of the model govern slip on faults according to the Coulomb friction law τ = µ σ n where τ is shear stress and σ n fault normal stress. ocked fault case with infinite effective coefficient of friction in the seismogenic layer in the uppermost 15 km and. 5 below. t the surface interseismic deformation evolves while there is no fault slip in the seismogenic layer., nlocked fault case with uniform of. 5. emporally continuous fault slip evolves. Note, that in both cases slip is not prescribed on the faults as commonly done. Instead slip evolves in response to remote kinematic boundary conditions applied at the sides of the model. Kinematic boundary conditions from a 3D regional model In order to obtain appropriate boundary conditions for our Marmara model we apply the socalled sub-modelling technique, i.e. we drive the model by the velocity field derived from a larger-scale 3D model of Northwest Anatolia that we term in the following regional model (Suppl. Fig. 2). We keep the northern side of this regional model fixed and allow lateral velocities at the eastern and western model boundary north of the NAF only in NS direction, not perpendicular to it (Suppl. Fig. 2). South of the northern branch of the NAF we apply a modified rotation around the Euler Pole of the Anatolian Plate 5, with the velocity between 29 and 33 E at the southern model boundary kept constant. Additionally, south of the NAF and east of 28 E we applied the west-component of the Euler rotation at the bottom of the model. The bottom of the model is otherwise laterally unconstrained while vertical velocities are constrained to zero. The velocity field resulting from this regional model largely reproduces the observed velocity field (Suppl. Fig. 2). Comparison of the kinematic results of the regional model with GPS velocities revealed that a friction coefficient of µ =0.05 gave the best-fit results. This modelled best-fit velocity field is used to drive the Marmara model at its lateral boundaries. 2 nature geoscience www.nature.com/naturegeoscience
doi: 10.1038/ngeo739 supplementary information Technically, the modelled velocity field from this regional model is mapped onto the nodes of the finite elements that form the lateral boundaries of the Marmara model. Black lines denote faults implemented in the regional model as contact surfaces to allow for frictional sliding according to Coulomb friction. Grey symbols and arrows denote applied displacement boundary conditions for the regional model in order to fit the GPS velocities 5. Triangles denote that the regional model is fixed on its northern side, circles indicate that displacement is only allowed parallel to the sides. Red arrows show modelled velocities at GPS sites in comparison with GPS derived velocities 5 (blue arrows). elocities from this best fit regional model with friction coefficient of. 5 are applied to the lateral boundaries of the Marmara model (black box). Model loads Besides the kinematic boundary conditions the model is subjected to the body force of gravity. Furthermore, the load of the seawater on the sea floor is applied. MMF slip rate at depth While Fig. 3b provides fault slip rates at the surface, Suppl. Fig. 3 shows the MMF slip rate at 15 km below sea level. on the MMF at depth is less than ~1 greater than at the surface. This is due to the fact that the smaller faults end at shallower depths so that the relative motion accommodated by these faults in the uppermost kilometres is taken up by the MMF at greater depths. Furthermore, the MMF in the model becomes somewhat straighter with depth, which facilitates fault slip. nature geoscience www.nature.com/naturegeoscience 3
supplementary information doi: 10.1038/ngeO739 Black lines mark fault traces at the surface of the model. Deviations between MMF slip rate of this study and previous estimates Supplementary Table 2 provides a quantitative comparison of the MMF slip rate from this study with previous estimates. Depending on location and study our slip rates are 10-45 % smaller than previous estimates. When taking our upper bound slip rate of ~2 more (see below) the s become correspondingly smaller and our rates would be similar to those of reference 2. Study Ganos Segment Central Segment Prince s Islands Seg. Izmit Segment this study 18 15-17 13-15 15-18 Ref. 2 20-10 % 19-16 % 17-21 % 18-10 % Ref. 3 23-22 % 23-30 % 23-41 % 23-30 % Ref. 4 24.8-27 % 24.5-35 % 24.5-45 % 24.4-34 % Ref. 5 26.5-32 % 26.9-27.9-42 % 24.6-45 % 27.1-40 % * Basis for the comparison are the fault slip rates displayed in Figure 3b. ** Locations of the fault segments are given in Supplementary Figure 3. Uncertainty of the modelled MMF slip rate We explored end-member cases with regard to the uncertainties of the model parameters to constrain an upper bound slip rate for the MMF. Greater slip rates on the MMF are conceivable if (1) Internal deformation between the faults is reduced. This means lower coefficient of friction on the faults and/or stiffer rock, i.e. increased Young s modulus E. (2) The MMF slip rate would increase if slip partitioning is reduced, i.e. if the other faults besides 4 nature geoscience www.nature.com/naturegeoscience
doi: 10.1038/ngeo739 supplementary information the MMF contribute less to relative plate motion. (3) Higher velocities at the boundaries of the model would result in greater slip rate on the MMF. To quantify the effect of the uncertainties of the model parameters on the MMF slip rate we calculate a misfit m of the modelled interseismic velocities (locked-fault model) and the GPS data using the definition where m = i v v GPS mod i v + v GPS i mod v and v denote GPS velocities and the modelled interseismic horizontal velocities, GPS mod respectively. In the following we discuss the results of the aforementioned three possibilities to increase the MMF slip rate. (1) Uncertainties of the Young s modulus and of the coefficient of friction Both, increased Young s modulus E and decreased coefficient of friction µ lead to less internal deformation. For µ =0 the MMF slip rate increases by 0-2 since slip is facilitated (Suppl. Fig. 4a). We tested doubled Young s modulus compared to the values listed in Supplementary Table 1, which leads to less internal deformation due to greater rock stiffness. The MMF slip rate increases by 1 (Suppl. Fig. 4b). The corresponding velocity fields are almost unchanged compared to the reference model in Fig. 3a. The velocity misfit (m=0.12 for the model presented in Fig. 3a) becomes little worse for µ =0 (m=0.13) and is unchanged for doubled Young s modulus (m=0.12). Notably, µ =0 and doubled Young s modulus represent end-member cases since this would imply completely frictionless faults and much greater seismic velocities than observed. (2) Reduced fault complexity Were we to neglect some of the second order faults, the maximum possible increase of the MMF slip rate would be to add the slip that is taken up by these smaller faults. However, the faults besides the MMF are seismically active 22,26,27,28,29. Thus, they contribute to relative plate motion in expense of slip on the MMF since there is no evidence for left-lateral slip in the region. Moreover, inactive secondary faults would neither produce the sharp rims of the bathymetric depressions in the Marmara Trough nor the observed vertical offsets in sedimentary strata in seismic profiles across these faults 6,15. nature geoscience www.nature.com/naturegeoscience 5
supplementary information doi: 10.1038/ngeO739 (3) Uncertainties of the boundary conditions Even though the applied boundary conditions are taken from the best-fit regional model, they have uncertainties in the order of 10% resulting from the GPS data that were fitted. We investigated the impact of 10% increased kinematic boundary conditions of the regional model on the MMF slip rate. In response to the increased boundary conditions the slip rate on the MMF increases by 2 (Suppl. Fig. 4c). The misfit decreases slightly to m=0.10. Figures on the left show modelled interseismic velocities (red arrows) in comparison with GPS data 5 (blue arrows, with 2σ uncertainties). Figures on the right show the corresponding right lateral fault slip rates at the surface (Suppl. Fig. 1). 0 Doubled Young s modulus compared to Suppl. Table 1. Increased velocity boundary conditions in the regional model (Suppl. Fig. 2) by 10%. Conclusion of the uncertainty study The analysis of the uncertainties of the model parameters shows that their impact on the MMF slip rate is small in general. We conclude that an upper bound of the MMF slip rate is ~2 greater than the slip rates shown in Fig. 3b, hence ~15-20 depending on 6 nature geoscience www.nature.com/naturegeoscience 6
doi: 10.1038/ngeo739 supplementary information location along strike of the fault. However, the main conclusions of this study remain unaffected: (1) The upper bound of the MMF slip rate resulting from the uncertainty analysis is still considerably smaller than previously estimated 3,4,5 and would be similar to the slip rate given by reference 2. (2) The variability of the MMF slip rate along strike of the fault is characteristic for all the considered alternative models. (3) Slip on secondary faults and internal deformation in the volume between the faults contribute to relative plate motion. Further details and constraints on the kinematics beneath the Sea of Marmara may be deducible from deformation measurements at the sea floor. Additional references 23. Provost, A.-S., Chéry, J. & Hassani, R. 3D mechanical modeling of the GPS velocity field along the North Anatolian fault, Earth Planet. Sci. Lett. 209, 361-377 doi:310.1016/s0012-1821x(1003)00099-00092 (2003). 24. Jiménez-Munt, I. & Sabadini, R. The block-like behavior of Anatolia envisaged in the modeled and geodetic strain rates, Geophys. Res. Lett. 29(20), 1978, doi:1910.1029/2002gl015995 (2002). 25. Savage, J. & Burford, R. Geodetic determination of relative plate motion in Central California. J. Geophys. Res. 78, 832-845 (1973). 26. Bulut, F., Bohnhoff, M., Ellsworth, W.L., Aktar, M. & Dresen, G., Microseismicity at the North Anatolian Fault into the Sea of Marmara offshore Istanbul, NW Turkey. J. Geophys. Res. 114, B09302, doi:10.1029/2008jb006244 (2009). 27. Ambraseys, N.N. & Jackson, J.A. Seismicity of the Sea of Marmara (Turkey) since 1500. Geophys. J. Int. 141, F1-F8 (2000). 28. Altınok, Y. & Alpar, B. Marmara Island earthquakes, of 1265 and 1935; Turkey. Nat. Hazards Earth Syst. Sci. 6, 999 1006 (2006). 29. Kürçer, A. et al. The Yenice Gönen active fault (NW Turkey): Active tectonics and palaeoseismology. Tectonophysics 453, 263 275 (2008). nature geoscience www.nature.com/naturegeoscience 7