Seismic stratigraphy of the north-western Sea of Marmara shelf along the North Anatolian Fault system

The Ganos Fault, a part of the Northern strand of the North Anatolian Fault system, is an active-strike slip fault and divides the narrow NW shelf of the Sea of Marmara into two parts near the town of Gaziköy. This paper presents recently collected shallow high-resolution seismic data to discriminate the sedimentary successions, each characterized by distinctive stratigraphic patterns on both sides of the Ganos Fault. Two main units, namely U1 and U2, and three para-sequences (U1a, U1b and U1c) were identified, depending on their internal reflection patterns, accommodation depths as well as the presence of conformity and the unconformity surfaces. The thickness of Unit U1 reaches its maximum at the northern side of the Ganos Fault; it is much thinner to the south. The para-sequences of U1b and U1c have “progradation” and “aggradation to progradation” depositional characters, respectively. This probably implies fluvial deposition controlled by sea-level fluctuations. Unit U1b can only be observed at the northern side of the Ganos Fault, while Unit U1c at the southern side. Units U1a and U1b were deposited during the transgressive system tract, while Unit U1c was deposited during a sea-level fall and/or a lowstand phase marked by an erosional surface. The marine terraces in the study area are shallower than those along the northern shelf of the Sea of Marmara, possibly due to successive tectonic displacements along the Ganos Fault, which also controls the distribution and thickness of the parasequences identified in this study.


Introduction
Fluctuations in global sea level and associated changes in sediment supplies are the primary effects that control the depositional conditions between shelves to continental margins through transgressive-regressive cycles. Highresolution seismic exploration is one of the most effective methods associated with sequence-stratigraphic concepts; it provides the best clues about the spatial and temporal record of stratal development and the surrounding environmental processes that occurred during their formation [Van Wagoner et al., 1986]. Therefore, shelf-edge depositions, especially related with late Pleistocene and Holocene units, could be clearly identified and analyzed accordingly to eustatic sea-level fluctuations using high-resolution seismic data. When sediment cores and associated sequence stratigraphic analysis correlate with high-resolution seismic data [Vail 1987;Posamentier and Vail 1988;Posamentier et al., 1992], interpretation to wider areas is more accurate.
Water transfer from the Sea of Marmara to the adjacent seas (Black Sea and Aegean Sea), i.e. its paleooceanographic conditions in the late Quaternary, are controlled mainly by the morphology of the connecting straits (İstanbul and Çanakkale) in response to global sea level changes (Figure 1a). Water exchange between the Sea of Marmara and the Aegean, for example, has been cut off by a sill [Aksu et al., 1999;Çağatay et al., 2009]. Thus, during the Last Glacial Maximum (LGM), for example, the Sea of Marmara turned into a lacustrine, and its shelves were sub-aerially exposed. It was a fresh/brackish water lake between early MIS4 to MIS1, and an important regression occurred during MIS 2 [Çağatay et al., 2015]. Connection was re-established with a transition to a warming period [Çağatay et al., 2009] between 14.7 cal kyr BP [Vidal et al., 2010] 12.55 ± 0.35 cal kyr BP [Çağatay et al., 2015].
The sedimentary succession deposited along the northern shelf of the Sea of Marmara consists of a young and relatively thin layer, overlying an acoustic basement not penetrated by high-frequency seismic sources [Aksu et al., 1999;Çağatay et al., 2009;Tur et al., 2014;Vardar et al., 2018;Nasıf et al., 2019]. The coarser basal sediments, however, derive from the erosion of tectonic highs, terrestrial inputs, and new hydrodynamic conditions after the LGM. The stratigraphic settings of the seismic sequences at the southern outlet of the Istanbul Strait [e.g. Aksu et al., 1999;Algan et al., 2001;Hiscott et al., 2002;Gökaşan et al., 2005;Eriş et al., 2007;Köprülü et al., 2016] are thicker and more complex, due to variable hydrodynamic conditions and high sediment input, compared to the other parts of the Sea of Marmara [Alavi et al., 1989]. In addition, several recent seismic data sets have also proven complex stratal developments in the lagoons of Büyükçekmece and Küçükcekmece

Materials and Methods
A total of 450km of high-resolution single-channel, CHIRP (compressed high-intensity radar pulse) seismic reflection profiles were collected in April 2019 ( Figure 1b) between the settlements of Tekirdağ and Şarkoy,

Results
The bathymetric map obtained from the seismic data ( Figure 1b) indicates that the shelf is relatively wide between Tekirdağ-Kumbağ (~10 km) and Gaziköy-Şarkoy, compared to the Kumbağ-Gaziköy sector (< 500 m). The GF is a crustal weakness zone representing the main tectonic element affecting the study area. It is well-defined on land (see Seeber et al., 2004), separating the Eocene clastics/carbonates from the Miocene clastics ( Figure 1c).
We identified two main seismic units (U1 and U2) separated by a sequence boundary (the reflector "SB"; see Figures 2a,b;5a,b,c,d). The SB is characterized as an erosional unconformity at the depths shallow than -85 m; below this critical depth it is conformable with overlying sediments (Figure 2a, 2b). This implies that the erosional part of SB (i.e. above 85 m bsl) underwent subaerial conditions during the LGM lowstand. Moreover, as its paleo-morphology was shaped under the control of the GF, its topography looks rather similar to the recent bathymetry ( Figure 3a).
The upper unit U1 over the SB consists of the sediments deposited since the LGM to the present. As shown by the thickness map (Figure 3b), the unit is rather thin above the wave-cut terraces (Figures 4, 5a) and thickens towards the coast, depending on the increment of terrestrial and fluvial inputs accompanied by local tectonic movements. In fact, unit U1 reaches the greatest thickness at the northern part of the GF (Figures 3a, 5b), while it gets thinner to the south of the GF. The maximum thickness was observed in front of the town of Gaziköy.
According to its internal reflections, truncations, clinoform settings and depositional levels, Unit U1 consists of three parasequences; U1a, U1b and U1c (Figure 4). U1a is characterized by weak, continuous and parallel internal reflections, while U1b consists of roughly parallel to progradational sigmoidal reflections downlapping over SB (Figures 4; 5a, b). On the other hand, U1c has a stratified aggradational to progradation progressive internal reflection character (Figures 4; 5c), and the interface surface with U1a is highly eroded (Figure 5c).

Discussion
The total thickness of unit U1 reaches 30 ms TWT (~26 m) near the shore, implying high rates of sediment transportation. These units become thinner towards the shelf break. Unit U1 (Figure 1b) is rather thin above the terraces too, which could be explained by subaerial exposure. It may have been eroded during relative sea-level still stands or transported alongshore by currents, a well-known hydrodynamic factor in the region [Chiggiato et al., 2012]. Unit U1 reaches its maxima (~40 ms TWT, ~ 34 m) between the towns of Gaziköy and Kumbağ, at the northern side of the GF. The differences in sediment thickness depend on the rate of sediment input, which unknown due to lack of core data, and the variations of the accommodation space under controlled by the tectonic displacement along the GF.

5
Seismic stratigraphy of the north-western Sea of Marmara shelf    [Williams and Ferrigno, 2012]. The dates given on the x-axis of "b", a and the y-axis of "d" are kyrs before present.

Seismic stratigraphy of the north-western Sea of Marmara shelf
The thickness of sub-unit U1a, much greater to the north of the GF, should be directly related to cumulative displacements along the fault and its northward-dipping geometry. Tekirdağ and Kumbağ, and also between Gaziköy and Şarköy. We note that these two units have a common fluvial origin, as witnessed by their seismostratigraphic characters, including thickness, distribution, accumulation depths, stratification geometries, and that the clinothem patterns and upper surface factors indicate a difference in depositional medium. In this context, the succession geometry of the clinothems of unit U1b is associated with the global sea-level rise; the absence of any erosion on its top also indicates that it is not affected by erosional conditions. Therefore, sub-unit U1b must have been accumulated during a transgressive phase.
Conversely, unconformity above sub-unit U1c displays an erosional character, indicating that this unit was deposited during the "falling" and/or "lowstand" phase. Unit U1b can be correlated with such seismo-stratigraphic deltas such as our Unit U1c are not very common in the Sea of Marmara, those reported by Smith [1995] and Hiscott and Aksu [2002] are an exception. The internal character of Unit U1a also suggests that this sub-unit is of marine origin and was deposited during a transgressive phase. We did not observe stratigraphic unconformity below -85 m, which is consistent with other studies and suggets the presence of a paleo-shoreline and paleodeltas around the Sea of Marmara [Hiscott et al., 2002;Çağatay et al., 2003;Polonia et al., 2004;Gökaşan et al., 2005;Eriş et al., 2007;Gasperini et al., 2011Gasperini et al., , 2018Köprülü et al., 2016, Vardar et al., 2018, Nasıf et al., 2019.
Four levels that can be correlated with the relative still-stand identified at 31, 49, 60, 72 m bsl. According to the global sea-level curves (Figure 7), these levels can be dated at 8.850, 9.500, 10.000, 11.200 BP respectively. A number of wave-cut terraces were observed at 40, 50, 85, 93, and 105 m bsl along the northern shelf of the Sea of Marmara [Eris et al., 2007, McHugh et al., 2008, Gökaşan et al., 2008, Çağatay et al., 2009. Moreover, Alp et al.
[2018] argued that the shallowest terrace is located at 65 m bsl while others are at the depths of 87, 94 and 108 m bsl. Similar marine terraces can be seen in the study area at the 49 m and 60 m bsl. The differences in depth between the terraces compared to the other levels in the northern shelf of the Sea of Marmara can be explained by the relative amount of uplift [e.g. Bulkan et al., 2020, Vardar et al., 2021 under the influence of the GF. The sigmoidal-shaped inner reflections of Unit U1b between Tekirdağ and Kumbağ is progradational, the ascending trajectories of rollover points can be correlated with sea level rise during clinothem development, and delta front is at 43 ms (TWT) bsl. Therefore, unit U1b could probably be classified as a Transgressive System Tract (TST), like U1a. Unit U1a seems to have been evolving since 11,200 BP according to correlations with the sea-level curve  Ergin et al. [2007] with CHIRP data.
Seismic stratigraphy of the north-western Sea of Marmara shelf confirm fluvial and terrestrial siliciclastic sources. The texture of the sea bottom sediments gets thinner, from coarse-grained sediments on the coast to fine-grained ones at the shelf-break. The general lithological characteristic of the cores given by Ergin et al. [2007] indicate higher-energy conditions over the eroded substrate of near shore, while lower-energy conditions predominate on the slope.
Although our seismic profiles cannot be associated directly with cores 9-10 of Ergin et al. [2007], the CHIRP data ( Figure 8) is well-matched with their multichannel seismic profile. Depending on this similarity and using the core dates given by Ergin et al. [2007], the bottom of unit U1a can be dated between 11.200 and 11.585 BP.
Similarly, using the dates calculated for Core 9, the SB can be dated 24.815 BP (Figure8). There are time and sealevel differences in the effects of global sea-level changes on the deposition and erosional stages in the Sea of Marmara. During the LGM period, the sea level in the North Aegean Sea was about 120 m bsl [Simaiakis et al., 2017] whilst it was about 100 m bsl on the Romanian Black Sea coast [Lericolieas et al., 2007]. Yanchlina et al. [2017] have shown an ancient shoreline varying between 80-100m bsl for the entire Black Sea. In the Sea of Marmara, the sea level associated with the last glacial period is around 85m bsl; meaning that the Sea of Marmara has a unique depositional and erosional system that differs from its neighbouring seas. In this context, the sea levels obtained in this study and dating corrections made by the previous studies are rather important for defining this marine basin, as well as other similar inland seas, which have their own distinctive marine characteristics.

Conclusions
Distribution and associated seismic facies of seismic units along the NW shelf of the Sea of Marmara, above an erosional surface (Sequence Boundary, SB) indicate inputs of fluvial, terrestrial and marine origin, starting from the LGM. A stratigraphic analysis of the high-resolution reflection profiles led to recognition of two main seismic units (U1 and U2) and three para-sequences (U1a, U1b and U1c), mainly controlled by water-level changes and regional tectonics. In fact, the thickness of the units, as well as their distribution along the shelf, has been controlled by seismic activities along the Ganos Fault. The erosional truncations observed along the slopes and deepening of the E-W trending basin were controlled by the dominant tectonic regime in the region. Moreover, the tectonic deformations also affect the external shapes of the sedimentary units and their internal reflections.
The marine terraces observed in the study area are located at shallower depths compared to those along the northern shelf, possibly due to regional tectonic uplift.
Another finding regards the discrepancies observed between the global sea-level curves and the water-level changes in this semi-closed inland sea, indicating that water-level changes in the Sea of Marmara are unique and not fully dependent on the eustatic oscillations of the oceans and the adjacent seas. Although we have provided some dating data representing the key reflectors, deeper cores and dating at frequent intervals are needed to reveal more detailed sea level changes.