Recent Geological and Geophysical Research on the Marmara Fault, Turkey

 Introduction

The Marmara Fault is the submerged portion of the North Anatolian Fault Zone (NAFZ) running under the Sea of Marmara just south of Istanbul. It is a right-lateral strike-slip fault with a high slip rate (~20–25 mm/yr) – comparable to California’s San Andreas Fault​. No major rupture has occurred on the central Marmara section since 1766, leaving a ~150–160 km “seismic gap” between the 1912 Ganos (Mw 7.3) and 1999 İzmit (Mw 7.4) earthquake rupture zones​. This long quiescence indicates significant strain accumulation and an overdue large earthquake, making the Marmara Fault one of the most closely studied seismic hazards in the world​. In recent years, extensive geological and geophysical research – including seismic imaging, monitoring of fault slip, stress modeling, crustal deformation measurements, and earthquake hazard simulations – has greatly improved our understanding of this fault’s behavior and risks.

 

Seismic Imaging of Fault Structure

Advanced seismic imaging has revealed the Marmara Fault’s subsurface geometry and physical properties in unprecedented detail. Researchers have integrated marine seismic surveys, deep seismic reflection profiles, gravity modeling, and even machine-learning techniques to map the fault zone. Key findings include:

Fault Geometry and Segmentation: A 2024 3-D seismic interpretation (using AI algorithms on 3-D reflection data) delineated the fault’s complex structure in the western Marmara Basin​. This study identified multiple strike-slip fault segments at depth, linked by extensional and contractional step-overs. The extensional step-overs correspond to the pull-apart basins (e.g. Tekirdağ and Central basins), while a restraining step-over underlies the Western High uplift​. The fault zone spans a broad ~20 km width and is segmented by these structural bends. Such segmentation is thought to influence where earthquake ruptures can initiate or terminate, acting as potential barriers​.

Crustal Structure and Moho Uplift: The same 3-D study combined seismic and gravity data to show that the crust is thinned beneath the Marmara Fault. The Moho discontinuity rises to ~24 km depth under the western Marmara Basin – about 6 km shallower than the regional average – forming an E–W trending “Moho ridge” directly beneath the fault trace​. This localized uplift of the Moho beneath the pull-apart basins indicates intense crustal stretching from ongoing fault motion. Modeling suggests the along-strike changes in crustal thickness (strong crust vs. thinned crust) create persistent earthquake rupture barriers and weak zones in the upper crust​.

Tomographic Velocity Imaging: Three-dimensional seismic tomography of the Marmara region has identified distinct high- and low-velocity anomalies associated with different fault segments. For example, a 2020 seismic tomographic study using ocean-bottom seismometers mapped two vertical low-velocity, high Vp/Vs “shear zones” beneath the main fault strands​. These ~10 km-wide zones extend from ~8 km depth to the deep crust along the Western Marmara fault segment and beneath the eastern Çınarcık Basin, respectively, and are interpreted as the deforming shear zone of the fault at depth​. They likely contain fractured, fluid-saturated rocks – consistent with a fault zone that can accommodate aseismic slip (creep) in places​. In between these zones, the tomographic images show a 50 km-long central section with relatively high seismic velocities and low Vp/Vs ratio, indicative of strong, locked crust​ (this central segment corresponds to the seismic gap off Istanbul). Small high-velocity patches within the fault zone were also imaged, representing strong asperities (stuck patches) at depth with little microseismicity​.

Implications of Imaging: The seismic imaging results confirm a highly segmented fault with variations in material properties along strike. Notably, the central Marmara segment appears highly locked (high-velocity, intact crust and absence of microearthquakes), whereas the western and eastern segments show more evidence of distributed deformation and fluids. Researchers infer that the locked central section may act as a seismic barrier under current stress conditions, impeding rupture propagation – until stress builds high enough to overcome it​. Paradoxically, if a rupture does break through this strong patch, the relatively uniform, elastic properties there could allow an earthquake to propagate at supershear speeds​ (faster than shear-wave velocity), potentially increasing shaking intensity. These structural insights are crucial for assessing how future ruptures might unfold along the Marmara Fault.

 

Fault Slip Behavior: Creeping vs. Locked Sections

Intense monitoring has focused on how different parts of the Marmara Fault are slipping (or not slipping) between big earthquakes. A combination of microseismic studies, repeating earthquake analysis, and geodetic observations has revealed a mixed behavior: some portions of the fault creep aseismically at depth, while others are fully locked, steadily accumulating strain. Major findings on fault slip behavior include:

Repeating Earthquakes and Variable Creep: Using a 15-year catalog of micro-earthquakes, researchers have identified numerous “repeaters” – small earthquakes that recur on the same patch – which serve as indicators of aseismic creep on the fault. Analysis published in 2023 (Becker et al.) showed that the fraction of slip occurring as creep varies significantly along-strike of the Main Marmara Fault​. Overall, the western Marmara segments (toward the Tekirdağ side) exhibit higher creep rates (a larger portion of the plate motion is accommodated aseismically), whereas the central-eastern segments (offshore Istanbul) appear mostly locked with little repetitive seismicity​. This along-fault variation in creep was quantified by summing the seismic slip from repeaters and comparing it to the long-term slip rate. Some patches were found to be nearly fully creeping, while others were nearly fully locked​. Intriguingly, one sequence of repeaters showed accelerated activity after a nearby Mw 5.2 earthquake, suggesting that moderate seismic events can locally increase the creep rate (perhaps by perturbing stress on a creeping patch)​. These findings demonstrate that the Marmara Fault’s behavior is not uniform – it includes both steadily creeping zones and stuck zones, which has direct implications for earthquake potential.

Locked “Asperities” Identified: Consistent with the seismic imaging, areas with little or no repeater activity correspond to locked asperities. The central Marmara section beneath the Princes’ Islands stands out as a locked patch: it lacks repeaters and produces very few microquakes, and geodesy confirms it is not slipping (see next section). The tomographic anomalies also mark this zone as high-strength. By contrast, creeping sections correlate with clusters of small repeaters and seismic swarms, often near fault branch intersections or basin edges. Geophysical models link this pattern to crustal heterogeneity – stronger crustal blocks can stick and accumulate stress, while adjacent weaker or fluid-rich zones creep. For example, where the fault bends or encounters a strong crust (due to buried high-density rocks), creep diminishes and an aseismic gap in seismicity appears​. Between these bends, in structurally “softer” sections, creep is more prevalent​. This heterogeneity in fault locking was also inferred by thermal-rheological modeling in 2021, which showed that mechanically strong crust (coupled to the mantle) coincides with the fault segments that stay locked, whereas decoupled weaker crustal domains promote creeping behavior​. In short, the Marmara Fault is mechanically segmented into creeping vs. locked patches by the variations in crustal strength along its length.

Transient Slow Slip Events: Besides steady creep, scientists have detected occasional slow slip events (SSEs) on the Marmara Fault system. One notable SSE in the eastern Sea of Marmara (near the Armutlu Peninsula) in 2016 was identified through a combination of strainmeter, seismicity, and GPS data​. During this months-long slow slip, the local microearthquake activity showed a distinctive evolution: initially a swarm of small quakes accompanied the onset of slow slip, and later a quiet period followed as the fault silently slipped. A 2022 study provided direct evidence that a slow-slip transient can modulate earthquake activity in this region – as the slow slip progressed, the frequency-magnitude distribution of earthquakes (the “b-value”) and their clustering behavior changed, indicating the fault’s stress state was being transiently altered​. Once the slow slip event decayed, the seismicity patterns gradually returned to normal​. These observations from the Armutlu area highlight that aseismic slip and seismic slip are closely linked. Even moderate slow slip episodes can redistribute stress and either trigger or dampen local seismicity. Monitoring such transients is now a key part of Marmara fault research, as they may precede or influence larger quakes.

Figure: Segments of the Main Marmara Fault and their inferred slip behavior. Colors along the fault indicate sections inferred to be creeping (blue), transitional or partially creeping (yellow), and locked (red), based on microseismic “repeating earthquake” analysis​. The stars mark the epicenters of recent moderate earthquakes (2019 Mw 5.7 and 2025 Mw 6.2), with focal mechanisms illustrating their faulting style. The 2019 event occurred on a minor reverse-fault splay, whereas the 2025 quake ruptured the main strike-slip fault. These moderate shocks did not release the strain on the major locked patch (red) offshore Istanbul, which remains a concern for a future large earthquake. (Credit: Temblor, after Becker et al. 2023)

 

Crustal Deformation and Strain Accumulation

Geodetic measurements of crustal deformation around the Marmara Sea provide critical insight into how stress is accumulating on the fault. In the past few years, new onshore and offshore observation techniques have refined estimates of strain build-up:

GPS and InSAR Observations: Continuous GPS networks in northwest Turkey show that the Marmara region is deforming at roughly 20–25 mm/yr across the fault zone, consistent with the long-term slip rate​. This secular strain rate, measured on both sides of the Sea of Marmara, indicates that the Anatolian plate is steadily moving westward relative to Eurasia, while the Marmara Fault locks in the interseismic period. InSAR (satellite radar interferometry) has been used in some studies to map ground deformation, but onshore GPS and campaign measurements remain the primary data on land. These onshore data alone, however, cannot resolve exactly how slip is distributed on the submerged fault – which led scientists to deploy offshore instruments in recent experiments.

Seafloor Geodetic Monitoring: A breakthrough came from direct offshore monitoring of the fault. In an international effort, geoscientists installed a network of seafloor acoustic transponders across the central Marmara Fault (southwest of Istanbul) to measure any relative motion on the seafloor with millimeter precision​. This project (reported in 2019) observed the seafloor for 2.5 years and detected no significant creep or displacement across the fault segment​. The absence of measurable slip, together with the lack of microearthquakes there, confirms that the central offshore segment is fully locked from the near-surface to at least a few kilometers depth​. In fact, the fault appears locked down into the crystalline basement, implying the entire seismogenic zone is accumulating strain​. From the geodetic data, researchers calculated that since its last major rupture in 1766, this segment has accumulated a slip deficit of at least ~4 meters, equivalent to what would be released in an earthquake of magnitude on the order of 7.1–7.4​. This direct measurement of strain accumulation offshore was a milestone – it unambiguously demonstrates that a large earthquake’s worth of strain energy is pent up beneath the Sea of Marmara.

Strain Distribution and Locked Zone Width: The geodetic and seismic data together suggest that the Marmara Fault’s locked zone is quite wide (spanning the entire fault width in the central part). Some sections of the fault may creep at depth below ~15–20 km (as inferred for portions of the NAF in eastern Turkey), but in Marmara the plate interface is relatively shallow and likely fully coupled through the upper crust. The 2019 Nature Communications study noted complete locking to at least 3 km depth on the central segment, but presumably the lock extends much deeper​. Ongoing efforts are aimed at measuring deeper creep if any – for instance, deploying longer-duration seafloor instruments and utilizing fiber-optic cables on the seafloor for strain sensing (an experimental technique). On land, dense GPS arrays around the Marmara Sea (including stations on the Princes Islands) continue to monitor how strain is partitioned. These measurements will indicate if any acceleration in strain occurs as stress builds toward failure.

Regional Stress Field: An important context for crustal deformation is the regional tectonic stress regime. Recent stress models (e.g. the World Stress Map project updates) show a consistent strike-slip stress orientation in the Marmara region, with maximum horizontal stress oriented NW–SE, conducive to right-lateral slip on the E–W trending fault. There is evidence that stress may concentrate at the fault’s geometric complexities – for example, the ends of the Marmara seismic gap (near the 1912 and 1999 rupture zones) could be areas of stress concentration. Additionally, a 2024 Science study by Ergintav et al. revealed that even distant large earthquakes can subtly perturb the stress and strain in Marmara: the great February 2023 Kahramanmaraş earthquakes (~800 km away in SE Türkiye) caused measurable but unexpected far-field crustal deformations up to the Marmara region, indicating the elastic lithosphere responded over long distances​. While these displacements were very small, they highlight that the Marmara Fault is part of a broader neotectonic framework that can be influenced by large stress changes elsewhere. Overall, continuous monitoring of crustal deformation is critical for detecting any precursory changes as the Marmara Fault continues to load.

 

Earthquake Hazard and Modeling

With Istanbul’s population at risk, considerable research in recent years has focused on modeling the Marmara Fault’s earthquake hazard – from rupture scenarios to ground motion predictions:

Earthquake Scenario Modeling: A central question is whether the entire 150+ km seismic gap could rupture in one great earthquake or break in smaller segments. Recent geological evidence of past events (and segmentation models) suggest that ruptures may be limited by the structural segments. Historical quakes in 1766, for instance, occurred in two large events (separated by a few months) rather than one giant rupture, hinting the Marmara fault might not break its whole length at once. Nonetheless, the worst-case scenario would be a multi-segment rupture through the central Marmara, producing an earthquake possibly in the mid-7s magnitude (Mw ~7.4±)​. Probabilistic models have been updated as new data emerges. An oft-cited estimate by Parsons et al. (2004) put the probability of a M≥7 quake on the Marmara Fault at ~40% within 30 years (starting 2004)​ – meaning that by 2025 the probability is even higher. While such numbers carry large uncertainties, they underscore a significant hazard. The timing of moderate events adds nuance: for example, the Mw 5.7 earthquake in September 2019 and the Mw 6.2 on 23 April 2025 ruptured peripheral portions of the fault system, but neither relieved the main stress on the central gap. The 2025 event broke only a ~20 km section of the fault (typical for its size)​, adjacent to but not penetrating the locked patch. These moderate shocks may slightly adjust stress distribution, but the consensus is that the seismic gap still holds essentially all of its earthquake potential.

Figure: Map of the Marmara Fault zone showing major historical earthquake ruptures (in blue) and the seismic gap offshore Istanbul. The April 23, 2025 Mw 6.2 earthquake (yellow star) ruptured the central part of the gap but only over ~20 km, leaving most of the segment unbroken. Dashed lines indicate inferred rupture extents of the largest historical events in 1509 and 1766 (both mid-M7 earthquakes) which likely covered much of the gap​. The accumulated slip deficit since 1766 (~4 m) corresponds to an earthquake of roughly magnitude 7.1–7.4​. The Marmara Fault’s slip rate (~23 mm/yr) means strain has been building rapidly over the past 250 years​.

 

Tectonic Stress and Rupture Dynamics: Modeling studies in 2021–2022 have highlighted how 3-D fault structure might control rupture behavior. As noted, mechanical segmentation could cause a large quake to break in pieces. One scenario is that a rupture initiates in the western Marmara (or eastern) segment – which have abundant microseismicity signaling critical stress – and then stops or slows at the central barrier if it cannot immediately overcome it​. Alternatively, if the rupture does propagate through, it might do so in a high-speed fashion (supershear) given the homogeneous properties of the central segment​. Such a scenario could exacerbate ground shaking in its path. These possibilities are being tested with dynamic rupture simulations and physical models. Researchers are incorporating the detailed structural data (velocity variations, asperity locations from seismic imaging) into physics-based earthquake simulations to predict how a magnitude 7+ rupture would unfold and what ground motions it would generate in Istanbul. Early results confirm that the central locking can indeed act as an obstacle – large ruptures may preferentially nucleate at the edges of the locked patch and sometimes may not break through it completely. This aligns with the idea of an upper bound on rupture length (perhaps two ~M7 ruptures rather than one ~M7.5).

Ground Motion and Directivity: A crucial aspect of hazard is how severe the ground shaking could be in Istanbul and surrounding areas. Recent studies have zeroed in on rupture directivity – the tendency of earthquake ruptures to propagate in a certain direction, which can focus seismic energy. In 2025, Chen et al. analyzed 31 moderate earthquakes (M3.5–5.7) under the Sea of Marmara and found a dominant eastward rupture propagation on the western Main Marmara Fault​. The median rupture direction was about 85° east of north, essentially aligned along the fault toward Istanbul​. This means seismic waves from those ruptures were preferentially beamed eastward. The study concludes that if a large Marmara earthquake nucleates in the western part of the seismic gap, it would likely rupture eastward and thereby direct a disproportionately high amount of energy toward the Istanbul metro area​. In effect, Istanbul could experience stronger shaking than if the rupture went the opposite way. This finding is vitally important for risk models – it suggests that beyond just earthquake magnitude, the rupture directionality should be accounted for in probabilistic seismic hazard assessments. Ground motion prediction equations for the Marmara region may be updated to include this directivity effect. Moreover, near-fault seismic arrays (like a borehole array on the Princes Islands) are continuously recording local quakes to further refine directivity and attenuation characteristics.

Earthquake Hazard Maps and Preparedness: All the new data are feeding into updated hazard models for Turkey. The national earthquake hazard map (last updated in 2018) already recognizes the Marmara region as a high hazard zone, but researchers are working to incorporate recent findings (e.g. locked/creeping segmentation, possible multi-segment rupture scenarios, directivity) into more sophisticated models. International projects and monitoring initiatives have been pivotal in this effort. The MARsite project (Marmara Supersite, an EU initiative) and GONAF (Geophysical Observatory of the North Anatolian Fault, a German–Turkish project) established dense monitoring networks – including ocean-bottom seismometers, borehole strainmeters and seismometers, and continuous GPS – which have collected the rich datasets underpinning many of the findings summarized above. As a result, the Marmara Fault is now one of the best-instrumented offshore fault systems. This multidisciplinary approach, combining geology, geophysics, and advanced modeling, is yielding a much clearer picture of the seismic hazard. In summary, the earthquake risk to Istanbul remains serious: an overdue ~M7+ earthquake is likely in the not-too-distant future​, and new research underscores that such an event could produce intense shaking in the city due to the fault’s geometry and rupture behavior. Ongoing studies of tectonic stress accumulation, fault slip processes, and ground motion simulations are crucial for refining hazard estimates and informing preparedness measures in the Marmara region.

 

References

Chen, X., Martínez-Garzón, P., et al. (2025). Rupture Directivity of Moderate Earthquakes Along the Main Marmara Fault Suggests Larger Ground Motion Towards Istanbul. Geophysical Research Letters, 52(1), e2024GL111460​gfzpublic.gfz-potsdam.de​phys.org.

Yılmaz, Ö., et al. (2024). 3-D Seismic Delineation of the North Anatolian Fault System Shear Zone in the Western Marmara Basin, Turkey. Tectonophysics (in press)​papers.ssrn.com​papers.ssrn.com.

Tarancloğlu, A., Özalaybey, S., & Kocaoğlu, A. H. (2020). Three-dimensional seismic tomographic imaging beneath the Sea of Marmara: Evidence for locked and creeping sections of the Main Marmara Fault. Geophysical Journal International, 223(2), 1172–1187​research.itu.edu.tr​research.itu.edu.tr.

Gholamrezaie, E., et al. (2021). Lithospheric strength variations and seismotectonic segmentation below the Sea of Marmara. Tectonophysics, 815, 228999​archimer.ifremer.fr​archimer.ifremer.fr.

Becker, D., Bohnhoff, M., et al. (2023). Variation of fault creep along the overdue Istanbul–Marmara seismic gap in NW Türkiye. Geophysical Research Letters, 50(6), e2022GL101471​files.scec.org​files.scec.org.

Bocchini, G. M., et al. (2022). Direct evidence of a slow-slip transient modulating the spatiotemporal and frequency-magnitude earthquake distribution: Insights from the Armutlu Peninsula, NW Turkey. Geophysical Research Letters, 49, e2022GL099077​gfzpublic.gfz-potsdam.de​gfzpublic.gfz-potsdam.de.

Lange, D., et al. (2019). Interseismic strain build-up on the submarine North Anatolian Fault offshore Istanbul. Nature Communications, 10, 3006​nature.com​nature.com.

Stein, R. S., & Sevilgen, V. (2025). A M6.2 quake strikes the Marmara Fault at site of large historic earthquakes near Istanbul. Temblor.net (April 23, 2025)​temblor.net​temblor.net.