For more than two centuries a silent structural threat has been building beneath the Sea of Marmara. A one hundred fifty kilometre section of the Main Marmara Fault has not ruptured since 1766, and new modelling work shows that it has remained locked while the surrounding fault system continues to shift. This locked segment sits directly offshore Istanbul, a city where millions of residents live within range of intense shaking if the fault breaks in a single event. The new study provides a detailed view of how this section has stored strain across generations and why it has resisted the break patterns that define the rest of the regional fault network. The findings suggest that the most hazardous part of the system is not the one that ruptures often. It is the one that has refused to rupture at all.
The research team created a three dimensional physical model of the Main Marmara Fault that incorporates sediment thickness, crustal heat flow, and realistic frictional properties for both basement rock and basin sediments. These factors determine how easily a fault can slip and how far a rupture can propagate once it begins. The Sea of Marmara region contains thick sedimentary basins and a high heat flow environment that influence the behaviour of the fault in ways that have not been fully captured in earlier hazard assessments. The simulations recreate more than ten thousand years of earthquake cycles and reveal how small differences in lithology and temperature can change the entire rupture pattern of a fault that crosses a major population centre.
One of the most important findings is the behaviour of the locked segment beneath the central part of the sea. This section does not break in the simulations that include realistic sediment and temperature variations. Instead it functions as a physical barrier that blocks the passage of ruptures. Western segments release strain in magnitude seven events that resemble the 1766 and 1912 earthquakes. Eastern segments generate smaller to moderate events with varied rupture lengths. The central section between them resists failure in most scenarios and remains stuck while stress accumulates. This is the same portion that has lacked a major historical rupture for more than two centuries. The simulations show that this is not a coincidence. It is the natural outcome of its unique mechanical conditions.
The locked zone is defined by deep sediment layers that extend several kilometres beneath the sea floor. These sediments reduce the frictional strength of the shallow crust and create a transition zone where stable sliding dominates instead of sudden rupture. At greater depths, high temperatures reduce the seismogenic thickness and narrow the zone where earthquakes can occur. The combination of these factors creates a bottleneck in the crust where ruptures entering from either side lose momentum. Instead of breaking through, they stop at the boundary and leave the central section unruptured. The surrounding regions keep producing earthquakes of various sizes while the locked zone becomes a long term reservoir of stress.
This behaviour is consistent with the historical record. The Main Marmara Fault has generated several damaging earthquakes over the last five hundred years, including the 1509, 1766, 1894, 1912, and 1999 events. These earthquakes broke different parts of the system but none of them ruptured the central locked segment beneath the Sea of Marmara. The study confirms that this section has unique mechanical controls that limit its ability to rupture under normal conditions. That pattern has persisted for centuries according to both historical observations and the modelled earthquake cycles.
The simulations also show the consequences of this persistent segmentation. A locked zone that does not release strain builds stress until rare conditions allow a rupture to propagate through it. These rare conditions require deeper thermal weakening or specific nucleation patterns that overcome the mechanical barrier created by the sediments and heat anomaly. When these conditions occur in the model, the central zone breaks in events reaching magnitude seven point three. These events do not happen often. When they do, they produce long rupture lengths and strong ground motion directed toward Istanbul due to the geometry of the fault. The rarity of such events does not reduce their significance. A low frequency, high impact rupture directly offshore represents the scenario with the highest potential damage.
The researchers tested multiple thermal models and sediment configurations to evaluate how sensitive the fault is to changes in strength and temperature. When the sediment layer is removed from the model the fault behaves differently. Ruptures are able to travel across the region more easily and full length earthquakes become more common. When temperature variations are flattened the seismogenic zone becomes thicker and allows ruptures to connect across the full fault. Both of these simplified scenarios produce larger earthquakes more frequently but do not match the historical record. The locked behaviour of the central segment only appears when realistic geological conditions are included. This confirms that the region’s structural complexity governs its long term earthquake cycle.
Creep behaviour is another important factor identified in the study. Shallow creep is observed along portions of the Tekirdag and Central Basin segments. This creep is consistent with measurements from long term geodetic monitoring and microseismic activity recorded over the last two decades. Creep does not eliminate hazard. It redistributes strain within the fault and influences where ruptures can initiate. The central locked zone is surrounded by creeping sections that accommodate slow slip in the upper crust while deeper portions remain fully locked. This configuration creates sharp mechanical contrasts that further reinforce the boundary between rupturing and non rupturing zones.
The study also evaluates the possibility that dynamic weakening processes could allow ruptures to cross the locked segment under specific conditions. Thermal pressurization of pore fluids is one such process. When frictional heating occurs during slip it increases fluid pressure within the fault zone and lowers the effective normal stress. This can temporarily weaken the fault and allow a rupture to travel through sections that would otherwise stop it. In simulations where thermal pressurization is activated, some ruptures succeed in propagating through the locked zone. These events reach magnitudes of seven point three and produce longer rupture paths with higher slip. They remain uncommon but they demonstrate that the mechanical barrier is not absolute. If deeper conditions align, the locked region can participate in a larger event.
For Istanbul the primary concern is not the maximum magnitude. It is the location of the rupture relative to the city and the depth of energy release. A magnitude seven centred on the Prince Islands or Kumburgaz segments would direct severe shaking toward Istanbul. The locked zone sits closer to the urban coastline than the segments that produced the 1912 and 1999 earthquakes. Any rupture that begins near or breaks through this region would deliver strong ground motion at short distances. The study does not assign a probability to this outcome but it identifies the structural conditions that would make it possible.
The locked segment has been building strain since the eighteenth century. The simulations show that the region is part of a long term seismic supercycle where periods of partial rupture alternate with rare, larger events. The system behaves in a way that allows stress to accumulate over centuries rather than decades. This makes the central fault segment one of the most important seismic hazards in the eastern Mediterranean. It does not release energy frequently. It stores it for extended periods. That long term storage creates a threat that develops slowly and quietly without the visible surface evidence associated with more active fault systems.
The study highlights the need for hazard assessments that incorporate realistic geological complexity. Simple models of fault behaviour tend to overestimate the likelihood of long ruptures and underestimate the diversity of rupture modes. The Marmara region contains variations in temperature, rock type, sediment thickness, and fault geometry that all influence how ruptures propagate. A hazard model that fails to include these factors will not reflect the true behaviour of the system. The new research shows that accurate modelling requires a full understanding of the region’s structural and thermal conditions.
Istanbul’s future exposure is shaped by these deep processes. The locked zone beneath the Sea of Marmara has resisted rupture for more than two hundred and fifty years. It remains locked in every realistic simulation that the researchers performed. It sits next to creeping segments that release strain without reducing the load within the central portion. It is part of a long term cycle that allows rare but significant earthquakes to occur when deeper conditions align. The study confirms that the central Main Marmara Fault is structurally capable of producing the next major event affecting Istanbul. It also confirms that the region has not experienced such an event since the eighteenth century.
The findings do not predict when the locked zone will break. They identify the mechanical conditions that control its behaviour and demonstrate that it is accumulating strain without release. For a region with extensive urban development along the coastline this presents a clear risk that must be acknowledged in future planning. The locked section beneath the Sea of Marmara is not dormant. It is active in a way that is difficult to observe directly but is revealed through geological and mechanical modelling. It continues to store the energy of centuries. It remains the structural feature most capable of producing the next damaging earthquake affecting Istanbul.
Source:
Guvercin, S.E., and Barbot, S. (2025). Persistent rupture segmentation of the Main Marmara Fault. Communications Earth & Environment, 6, 1028.
https://doi.org/10.1038/s43247-025-03007-4
