Main

Natural fluid seepage systems are now recognised as widely distributed in the marine environment1,2. Expelled fluids can be composed of water or hydrocarbons under liquid or gaseous form and combine into a complex 2 or 3 phase flow during their migration towards the seafloor. Hydrocarbon seepage is particularly important due to its frequent liquid and gas composition, including potent greenhouse gas (methane) that represents a strong societal concern3,4. At present, fluid expulsion fluxes from the ocean are subject to large uncertainties5,6, and the local/regional mechanisms that trigger fluid expulsions are poorly studied due to remote locations and the limited accessibility for in-situ measurements during activation/deactivation phases.

Satellites cover large areas and offer high revisit frequencies for remote monitoring of seeping activity (Radar7; or multispectral8). Remote sensing techniques have demonstrated that oil seepages at the sea surface are exclusively transient and subjected to uneven expulsion frequencies9,10,11. However, the mechanisms of the fluid expulsions are diverse. For example, earthquake-triggered fluid expulsions are well-documented for mud volcanoes in sub-aerial conditions12,13,14 and for methane in active margins15, with a typical response time of a few days14. In the marine domain, earthquakes have been proposed as possible trigger mechanisms16,17, but the role of earthquakes in oil seepage is unclear due to a dependence of weather conditions on oil detection, a diversity of potential triggering factors (e.g., tides or anthropogenic activities), and the difficulty to estimate the delay between triggering and seepage response delay. Therefore, with the volume of hydrocarbon seepage triggered by an earthquake being potentially significant, we need to define the role and to which extent earthquakes can impact seepage expulsions.

Here, we examine hydrocarbon seepage in the Cyprus Arc region following the 6 February 2023 Mw7.8 Kahramanmaraş, Türkiye earthquake sequence (Fig. 1a). The Cyprus Arc exhibits a complex historical deformation pattern, including subduction-related processes and differential strike-slip motion at the Eurasian-Anatolian plate border (25–30 mm.yr−1) and the African Plate-Eurasian Plate border (10 mm.yr−1 in the Cyprus Arc region18,19). Both the trans-tensive Latakia Ridge and the trans-pressive Cyprus Arc act as the deformation front between the African and the Anatolian plates20,21, separating the Tartus basin (Anatolian plate) from the Levantine basin (African plate; Fig. 1a, b). The latter basin has been a focal point for hydrocarbon resource exploration22.

Fig. 1: Geological and seismological context of the oil seep sites.
figure 1

a Tectonostratigraphic map (after35) of the Cyprus basin superimposed with the earthquake events detected during the Turkish earthquake of 2023. b Seafloor map of the seepage area, main geological structures, and basins. c Interpreted seismic section nearby (~200 m) the seepage area activated by the earthquake sequence26.

Full size image

Northwest of the Cyprus Arc, the Mw7.8 Kahramanmaraş, Türkiye earthquake nucleated on 6 February 2023 as a complex supershear event that propagated bilaterally along the main East Anatolian Fault and produced strong directivity of low frequency seismic radiation toward the Latakia ridge23 (Fig. 1a). Due to its strong directivity, the aftershock zone of the Kahramanmaraş earthquake was ~200 km wide, triggering several thousands of aftershocks on multiple fault segments23 including a Mw7.6 event in less than 24 h, 3 magnitude 6 events within 2 weeks, and more than 40 magnitude 5 events in the first 8 months (Fig. 1a). Their size and proximity to a region known for hydrocarbon seepage leads to the questions: can earthquakes induce hydrocarbon seepage, and if so, how? To address this question, we apply remote sensing techniques and borrow concepts from well-documented studies of fluid systems affected by external forcings13,14,24. We demonstrate that several earthquakes from the Türkiye sequence likely extended fluid pathways and mobilized fluids in multi-phase hydrocarbon systems of the Cyprus Arc.

Oil seepage temporality

SAR (synthetic-aperture radar) imagery demonstrates that the Cyprus Arc has been intermittently active with repetitive oil slicks on the sea surface since October 2014 until the Kahramanmaraş mainshock (Figs. 2b, 1a). Between 2014 and 6 Feburary 2023, oil slicks emanated from three different recurrent seep sites (see Oil Slick Origin in Methods9). The high variability in slick direction is associated with sea surface current variability of the 3 gyres (Fig. 1a) controlling water circulation in the area: the dominant anticyclonic Cyprus Eddy (CE) limited to the north by the topographic scarp of the Cyprus arc, and cyclonic Latakia and North Shikmona Eddies (LE, NSE;25). The number of natural oil slick detections varies from one seep to the other with a ratio between the number of detected slicks and SAR coverage varying from 0.1 to 13% (Fig. 2a). One site shows high expulsion frequencies (Site 1) compared to the two other sites (Sites 2 and 3). From a geological standpoint, the Oil Slick Origins (OSOs) are located along the Cyprus arc at the boundary between the Western Tartus Basin and the Levantine Basin.

Fig. 2: Natural oil seepage spatiotemporality monitored with satellite imagery.
figure 2

a Compilation map of seepage slick outlines detected using Sentinel-1 images before the earthquake (Oct. 2014–Jan. 2023). b Compilation map of seepage slick outlines detected after the earthquake (Feb. 2023–Dec. 2023). Occurrence rate (%) refers to the number of slicks detected per image. c Temporal evolution of seepage slick detection following the mainshock using a combination of multispectral and radar spatial imagery.

Full size image

After the Kahramanmaraş mainshock, satellite imagery reveals the recurrent presence of oil slicks originating from an area previously devoid of such phenomena (Site 4; Fig. 2b). While multiple satellite scenes were available from the day following the Kahramanmaraş mainshock (see Methods), the first manifestations of oil slicks were detected 9 days after the event (15th February 2023, Fig. 2c). From this date, this newly activated Site 4 manifested intermittent activity during 2023 and has the highest occurrence rate (22.9%). The water depth is 1680 m at the vertical projection of Site 4, and the bathymetric data reveals a disturbed area without showing any focused vent (Fig. 2b), either because the 50 m resolution is too sparse or because the vent did not exist at the time of seismic acquisition (BLAC campaign26).

After the mainshock, the detected oil area per image for Site 4 shows a progressive increase for a 3-month period before reaching a stable state (May–June, Fig. 3a). This period was followed by a progressive decrease in intensity (July to September) with Site 4 completely deactivating in November 2023.

Fig. 3: Evaluation of energy density and strain imparted at the oil seep sites during the Kahramanmaraş earthquake sequence.
figure 3

a Timeline of earthquake energy densities and seep productivity at seep site 4. Black line = 2-day cumulative energy density mean. Orange line = 10-day average of slick area per image. Red line = envelope of 10-day average of slick area per image. Red and black dots = days with/without satellite data, respectively. b Distance vs. magnitude for earthquakes (circles) in our database. Black outlined circles = earthquakes occurring between 6 February and 28 February 2023 with their dynamic strain29 shown in color. Squares = earthquakes known for triggering hydrological responses at sub-aerial mud volcanoes12 and in a California well28. Solid lines = constant energy densities76. The seep sites all lie outside of the static strain zone for earthquakes in our study, suggesting dynamic strain as a mechanism. See Methods for earthquake database as well as static strain zone, dynamic strain, and energy density calculations.

Full size image

Earthquake triggering potential and timing

Most earthquakes in this study occurred far from the seep sites (Fig. 3a). Site 4, closest to the Türkiye sequence, lies beyond a typical aftershock zone length (static strain regime, see Methods) of the earthquakes, suggesting a possible dynamic strain triggering mechanism (Fig. 3b). Coupling earthquake magnitude, distance, and expected energy density at Site 4 (see Methods;13), 4 earthquakes—the Mw7.8 Kahramanmaraş mainshock and its Mw7.6 Ekinözü, M6.6 Sakçagözü, and M6.4 Samandağ aftershocks—appear to be responsible for activating the seepage. Their energy densities are akin to earthquakes that have triggered sub-aerial mud volcanoes12,27 and a well in California28. Earthquake-generated dynamic strains calculated from estimated peak ground velocities (PGV) at Site 4 (see Methods29) demonstrate that the Mw7.8 mainshock and its Mw7.6, M6.6, and M6.4 aftershocks all generated dynamic strains greater than 10−5 (Fig. 3b), comparable to triggered mud volcano observations30.

The Mw7.8 mainshock and its Mw7.6, M6.6, and M6.4 aftershocks occurred early in the seep activation (Fig. 3a). The mainshock, the Mw7.6 aftershock, and the M6.6 aftershock nucleated on the same day (6th February 2023), 9 days before the first seepage observation (Fig. 2b) and generated energy densities >4 × 10−3 J.m−3. On February 17th, no seepage was observed in satellite data, and then, on February 20th, the M6.4 Samandağ aftershock nucleated ~220 km from seep Site 4, generating an energy density ~9 × 10−3 J.m−3. After the M6.4 aftershock, earthquake energy densities decreased, and hydrocarbon seepage continued in a much stronger state than previously observed (Fig. 3a).

Fluid system reactivity and resilience

The delay between the 3 earthquakes on February 6th and seep visibility at the ocean surface depends on 4 factors: (i) the distance from the sedimentary bypass system to the seafloor, (ii) the hydraulic diffusivity of the media through which the oil must travel30, (iii) upward ascent time of an oil droplet through the water column31,32 and (iv) sea surface slick spreading at the ocean’s surface33. Ascension velocity of an oil droplet can vary between 2 cm.s−1 and 50 cm.s−1 31,34. The seep sites having depths between 1680 m and 1820 m, ascension of an oil droplet ranges between 1 to 23 h, while oil spreading at the surface could be 5 h for a 4-km long slick (like the one on February 15th, Fig. 1c). The observed delay between the mainshock and seepage is much longer compared to the inferred time for oil to migrate in the water column and spread along the sea surface, indicating that oil was expelled from the seafloor at least a few days after the mainshock. This suggests the diffusivity of the oil is low in the sediment, and even if the source depth of the mobilized oil in the sediments is unknown, a considerable period may have been necessary to migrate from a productive series (suspected Pre-salt; Fig. 1c) to reach the seafloor.

In the Cyprus and Tartus basins, the hydrocarbon sealing system is favored by the deposition of thick evaporitic series (Mobile Units 1 and 2; ~300 m thick35; Fig. 1c) during the Messinian Salinity Crisis (MSC) that led to the Mediterranean Sea drying out36. In addition, the evaporites are overlain by Plio-Quaternary fine-grained sillici-clastic deposit succession in the basin (Unit 1,37). However, the Plio-Quaternary sedimentary overloading induces a dynamic stress on the mobile evaporitic unit, triggering diapiric-like salt tectonic movements. This induces significant lateral variations in the thickness of post-salt sedimentary layers (Fig. 1c). The transpressive nature of the faulting system induces a disconnection of the salt series, which creates a breach in the sealing system and can explain the positioning of the oil seepage sites detected in satellite images along the Cyprus arc (Fig. 1c). Mud volcanoes, crater pockmarks, and diverse types of mounds are visible on the seafloor map, attesting focused fluid flow within the basin38.

Seepage activity was verified for more than 8 years before the mainshock, and the substantial change in seepage activity was only detected after the mainshock and its first aftershocks without any on-going seepage activity. The oil seep Site 4 remained activated for six months afterward and, with a transient seepage mechanism, indicates the duration for which an earthquake could influence fluid seepage systems. Following the centerline method (see Methods39), we calculated the average flux for slick detections by evaluating the necessary spreading time assuming a fixed drifting velocity compared to slick length. When normalized by the occurrence rate, we obtained an annual oil output estimate for each site before and after the earthquake (Table 1).

Table 1 Data availability, slick detections, and oil seepage distribution per seep site before and after the earthquake
Full size table

The estimated seepage output shows opposite changes between Sites 1 and 4, which could be explained by modifications in geodynamic stresses between the sites following the earthquake. Even though Site 1 decreased in expelled output following the earthquake sequence, the total expelled volume was increased by 71.2% after the earthquake sequence due to the activation of Site 4. Hydrocarbon volume emitted from Site 4 represents 14.8% of the total volume of hydrocarbons detected since satellite imagery observation began in 2014.

Wave propagation mechanism and impact on fluid seep intensity

External forcings have been long considered as capable of modifying pore pressure and subsequently the temporal evolution of seepage activity (ocean tides40; Earth tides41), either by inducing mobilization or retention of hydrocarbons, similar to our observations (Table 1). It has also been shown that earthquakes with high dynamic strains, strong low-frequency components, and high energy densities can increase discharge by mobilizing water or gas bubbles trapped by capillary pressures in single-phase and 2-phase systems42,43,44. In addition, some high-frequency body waves of high amplitudes can mobilize oil droplets in restricted pathways45.

In our case, the Türkiye earthquake sequence occurred along the East Anatolian strike-slip fault, and therefore, the amplitudes of the transverse seismic waves are expected to be significantly larger than their radial counterparts. For some earthquakes in our study, the seismic wave amplitudes were above the dynamic range for the horizontal components of the nearest seismometers, which limited our ability to analyze the earthquakes’ transverse waveform characteristics. However, vertical component spectral analysis for the Mw7.6, M6.6, and M6.4 aftershocks indicate that these 3 events generated small body wave amplitudes and high-amplitude low-frequency waves near the seep sites, particularly during the expected surface wave arrival times (Figs. S3-S5).

Based on our theoretical calculations, the Mw7.8 Kahramanmaraş mainshock and its Mw7.6 Ekinözü, M6.6 Sakçagözü, and M6.4 Samandağ aftershocks generated both high dynamic strains and high energy densities (Fig. 3). Love and Rayleigh waves are visible in both the waveforms of the Mw7.6 and M6.6 earthquakes that occurred at distances >360 km to seep Site 4 (Figs. S3-S4). We presume that similar observations could be made for the Mw7.8 mainshock, which clipped the nearest seismometer. The M6.4 earthquake, which occurred much closer to the seep sites ( ~ 220 km), was also clipped by the nearby seismometer (Fig. 1) but produced what appears to be a strong Lg wave of amplitude greater than or equal to the Mw7.6 and M6.6 events and a smaller Rayleigh wave, typical of strike-slip events (Fig. S5). A comparison to a Cyprus land station at a similar distance (Fig. 1a) that was not clipped confirms the existence of the Lg wave for the M6.4 event and demonstrates that the Lg wave is comprised of strong, low frequencies of 0.2 to 1 Hz across 3 components (Fig. S6). The amplitudes of the Lg wave are much smaller at the land station compared to the offshore station, likely due to continental attenuation versus amplification by sedimentary cover. These findings demonstrate that the hydrocarbon seepage can be linked to the high dynamic strains, strong low-frequency motions amplified by a sedimentary layer, and high energy densities of these four earthquakes.

The question remains how exactly did these hydrocarbon systems respond to these four earthquakes? With no detected earthquake activity on faults near the seep sites indicating fault reactivation as a cause, we suspect that positive (dilational) strain cycles associated with the passive seismic waves of these larger, distant earthquakes increased permeability and first hydrocarbon movement, while negative (compressional) strain cycles ‘squeezed’ the hydrocarbons into the new openings (dilatancy migration46). Thus, fluid overpressure may have been induced by each strain cycle triggering the upward mobilization of hydrocarbon fluids towards the seafloor. As stated previously, in a two-phase system of gas/water such as for mud volcanoes, gas bubbles and particles can be dislodged by strong low-frequency seismic waves of earthquakes, resulting in increased fluid mobility42. In our case, we may be observing fluid mobilization in a 3-phase system of oil, gas, and water. Water mobilized by as little as 3 kPa can cause either a trapping or a mobilization of oil47, particularly if gas is also mobilized. Considering the timeline of hydrocarbon seepage and the earthquakes, we surmise that the strong Love waves of the Mw7.8 Kahramanmaraş mainshock and its Mw7.6 Ekinözü, M6.6 Sakçagözü aftershocks likely created fractures in the subsurface on 2023 February 6 at Site 4, where seepage had not occurred for more than 8 years. Subsequently, their weaker Rayleigh waves either (1) mobilized gas bubbles trapped by capillary pressures in the hydrocarbon system and displaced the oil or (2) injected water which then mobilized the gas and displaced the oil. These initial fractures coupled with gas or water/gas mobilization led to the weak seepage that did not appear in satellite data until February 15th (Fig. 2). On February 20th, a large amplitude Lg wave of the M6.4 earthquake radiated past the seep sites, creating additional fractures in the subsurface. Given that the hydrocarbon system had already been weakened by the Mw7.8, Mw7.6, and M6.6 earthquakes, this led to increased hydrocarbon seepage at Sites 2, 3, and 4. While the tectonics of our study area are of a thrusting nature and thrust faults are usually less responsive to remote triggering48, it has been shown that the Türkiye earthquake sequence triggered seismic activity on thrust faults in the Caucasian region49. Therefore, it is possible the Love waves of these predominantly strike-slip earthquakes generated undetectable fractures or induced a further propagation of hydrocarbon pathways towards the seafloor at the seep sites. On the other hand, seep Site 1 appears to have reduced its output of hydrocarbons following the earthquakes (Fig. 2). We attribute this reduction to partial trapping due to injected water that mobilized gas to high pressures and confined conduits, reducing the mobility of the oil47. However, metocean conditions influence the detection of oil slicks and therefore could potentially introduce biases, i.e., resulting in an apparent decrease in estimated seepage.

In the current geological context, we interpret that fluid expulsions are driven on a regular basis by slight pressure variations in the hydrocarbon system and can be affected by substantial variations in their intensity when punctual events impact pore pressures. This study thus presents a seismicity-fluid relationship in the marine domain whose mechanisms contrast with those known on land, where self-sealing mechanisms do not seem to prevail (Sloshing50). These theories, observations, and interpretations altogether highlight the complexity of triggered hydrocarbon seepage and suggest that the current state of a hydrocarbon system plays an important role in how it will respond to earthquakes.

Towards integration into the global carbon budget

In this study, we investigated the impact of earthquakes on 4 marine seep sites in the Levantine Basin along the Cyprus Arc. While we were only able to document the oil output following the earthquake, marine seeps are known to be one of several types of geological emissions of methane, which are considered as the second biggest producer of atmospheric methane contributing to the global carbon budget51. Despite the known contribution of geological emissions of methane to greenhouse gas, discovered marine seeps, such as in our study, remain vastly underreported with few studies reporting their geographical and geochemical data (e.g.,1,2). Therefore, as a first step towards integration into the global carbon budget, there is a need to centralize marine seep geographical and geochemical data.

Secondly, while offshore oil seepage is easily monitored via satellite imagery, gas seepage monitoring for carbon budget estimates is more difficult. Gas seepage monitoring using satellites is possible onshore (GHG SAT52,53), but electromagnetic absorption interference between water and atmospheric gas prevents reliable detection offshore. While sun glint reflection allows for gas monitoring offshore54, it requires a specific satellite acquisition configuration. Currently, direct gas monitoring from the seafloor is limited to in-situ measurements, remotely operated vehicles, gas samplers, or acoustic cameras55. Indirect methods may include using ocean-bottom seismometers for detection of characteristic short-duration events proposed to be related to fluids expelled from the seafloor56. However, the possible origins of short-duration events are diverse, including small tectonic events and biological activity, and discrimination by frequency content and waveforms still requires technological advancements57. In the water column, acoustic imaging provides a broad view but is temporally limited to the duration of a sea campaign58,59,60. On the other hand, autonomous underwater vehicles can investigate for longer periods but are limited by the distance of the sensor to the site. Although limited by its resolution and revisit period, satellite imagery currently offers the best long-term tracking of fluid expulsions over large distances with revisit periods of only a few days, enabling the identification of the link between seismicity and fluid expulsions over such distances.

While the need for estimating expulsed gas volumes offshore remains, this study shows that earthquakes can induce hydrocarbon seepage at sites never known to produce seepage and even at a great distance from seep sites. It remains, however, to quantify how much triggered seepage contributes to the global carbon budget. Although the 2023 Türkiye earthquake induced an increase in oil output in the environment, the associated volume remains insignificant compared to global-scale naturally expelled volumes ( ~ 0.04 % of the volume expelled from the 3 main oil-emitting provinces10,11,16). Therefore, our approach should be further deployed toward active seismological provinces showing frequent oil expulsions as evidenced from satellite imagery (Santa Barbara Basin61, Aegean Sea62, Caspian Sea16).

Methods

Satellite inventory of oil seepage slicks

The radar images were acquired from Sentinel 1A and Sentinel 1B platforms (Synthetic Aperture Radar; C-Band). We used the Interferometric Wide swath (IW) products, VV Level 1 - Ground Range Detection (GRD) processing level. We accessed the satellite data using DIAS (Data and Information Access Services) server of Copernicus. The processing steps consists of thermal noise removal, Calibration, noise level reduction, Multilook, Ellipsoid-Correction-GG, and finally Linear To dB conversion. The radar data used in this study covers the full Sentinel-1 data stack between launch (2014) until December 2023. Sentinel-2 multispectral data (MSI) were integrated using the visible range (B04, B03, B02) and oil spill index [(B03 + B04)/(B02)] to increase the revisit period between the mainshock and 31 December 2023. The revisit period over the study area is every 3–4 days. 771 Sentinel-1 and 129 Sentinel-2 scenes were interpreted (Fig. S1). Satellite images were analyzed in the QGIS mapping software. Seepage slick detection was based on the analysis of optical and radiometric anomalies associated with oil-covered areas and were manually delineated. While oil sicks detection and segmentation are now possible using automatic detection techniques and artificial intelligence63, these algorithms are subjected to false alarms (biogenic algal mats, pollutions, ocean fronts)64. In addition, manual delineation only requires visualisation accessible using remote visualisation portals while automatic detection requires much heavier data handling, including downloading, formatting, preprocessing, and validation. Considering that the surface of the study area is limited, the manual delineation method was chosen to perform slicks mapping. Diverging structures on the stack of slick outlines detected on satellites images highlights recurrent oil seeps due to current variability at the sea surface (Oil Slick Origin, OSO9). Spatial attributes (date, oil slick type, confidence index, surface area, and length) were computed for each slick. The confidence index displays the degree of discrimination of liquid hydrocarbons compared to the different look-alikes (bioturbations, algal blooms, pollutions65). The hydrocarbon occurrence rate (in %) was calculated from the ratio between the number of identified hydrocarbon slicks and the satellite data coverage.

Expelled oil volumes

The displacement of the slicks at the sea surface is calculated from the addition of sea surface current velocity and 3% of the wind speed65. The friction with surface water implies that only a small fraction of the wind (3% estimated) account for slicks displacement. The expelled volumes were computed assuming an average surface velocity and the centerline method39. The flow rate was calculated for each slick by considering sea surface velocity, as well as attribute values such as the area and time of spreading (obtained with the centerline) and the thickness of the slick (estimated at 0.145 µm66). The flow rate calculated for each oil slick was then used to calculate the average flow rate discharged by site over the entire observation period. This value was then normalised by the rate of occurrence considering the sites to be intermittent9,11,67,68, thus making it possible to obtain flow rate values per site averaged over the observation period.

Earthquake database

Multiple onshore and offshore fault structures were activated by the 2023 Kahramanmaraş, Türkiye earthquake sequence. There are, therefore, many potential seismicity sources that could be responsible for, in turn, activating hydrocarbon seeps offshore Cyprus. To explore all potential seismic sources, we acquired earthquake information from the Cyprus Geological Survey (CGS) and the Kandilli Observatory and Earthquake Research Institute (KOERI, Türkiye), the primary monitoring agencies of the study region. For each monitoring agency, we searched for earthquakes located between 31.330°E and 39°E and 33.674°N and 38.38°N and occurring between 1 January 2023 and 6 December 2023. For these criteria, 1000 and 36,027 earthquakes were reported respectively in the CGS and KOERI catalogs, as of 11 December 2023. Given these agencies cover different zones of our study region (a function of their seismic networks), we merged these earthquake catalogues into a single database and removed duplicate reported events, leaving 35,570 earthquakes for analysis (Fig. 3). We note that, using these same search criteria, we also searched for earthquakes from the International Seismological Centre (ISC), which acts as a global agent for collecting earthquake information from monitoring networks worldwide. However, less than 6000 earthquakes were reported to the ISC by other agencies. Therefore, we only used events reported by the regional agencies (CGS and KOERI) for our analyses.

Static strain regime

We consider the static strain zone of an earthquake as being up to twice an earthquake’s rupture length from its epicenter, which is generally thought to be the zone where a mainshock causes static strain change69. In Fig. 3, we approximate the static strain zone using a range of magnitudes between 0 and 8. For each magnitude, we estimate the expected rupture length (L) as \({{\boldsymbol{L}}}={{\boldsymbol{0}}}{{.}}{{\boldsymbol{01}}}{{\times }}{{{\boldsymbol{10}}}}^{\left({{\boldsymbol{0}}}{{.}}{{\boldsymbol{5}}}{\times }{{\boldsymbol{M}}}\right)}\), where M is magnitude and L is in kilometers70. While this model was developed using earthquakes from California, we assume it can be applied elsewhere. The estimated static strain zone of L to 2 L is shown as a gray box in Fig. 3 for comparison to the epicentral distances and magnitudes in our earthquake database. In Fig. 3, all ~35,000 earthquakes fall within the dynamic strain regime.

Dynamic strain

For the ~35,000 earthquakes in our database, we computed a theoretical dynamic strain to give a first approximation of their triggering potential at Site 4 using earthquake magnitude and distance to the seep site29. We first used the surface-wave magnitude (Ms) relation to estimate the displacement by an earthquake: \({{\mathbf{log }}}_{{\mathbf{10}}}{{\boldsymbol{A}}}_{{\mathbf{20}}}={{\boldsymbol{M}}}_{{\boldsymbol{s}}}-{{\mathbf{1.66}}}\,{{\mathbf{log }}}_{{\mathbf{10}}}{{\mathbf{\Delta }}}-{{\mathbf{2}}}{{\mathbf{\Delta }}}\), where A20 is displacement in micrometers and Δ is distance in degrees to the seep site. While not always the case, here we assumed Ms is the magnitude reported in the earthquake database for this first-order approximation. Using the determined displacement (A20), we then estimated the peak ground velocity (PGV): \({{\boldsymbol{PGV}}}=\frac{2\pi {{\mathbf{A}}}_{{\mathbf{20}}}}{{\mathbf{T}}}\), where T is period. In our case, we used T = 20 s for the expected larger seismic wave amplitudes. Finally, we calculated dynamic strain from PGV71 by \(\epsilon = \frac{{{\boldsymbol{PGV}}}}{{{\boldsymbol{v}}}_{{\boldsymbol{ph}}}}\), where vph is phase velocity and for which we assume a value of 3500 m.s−1 for a pressurizing Rayleigh wave capable of mobilizing fluids72. We compared theoretical PGVs to measured PGVs at an ocean-bottom seismometer ~20 km from Site 4. PGVs were measured from the vertical component of station CY602 (open access: http://ds.iris.edu/mda/IM/CY602/) for 42 larger earthquakes by first downloading their waveforms, removing the instrument response, and then measuring their peak velocity (Fig. S2). Theoretical PGVs at the seismometer were calculated as described above but with distance being the distance between the earthquake and the seismometer. As seen in Fig. S2, the theoretical approximation is appropriate, particularly for magnitude 5 and above, which is relevant to this work. Furthermore, this calculation assumes that all the earthquakes generate surface waves. At short distances, surface waves may not be developed, and here, we are assuming surface waves are the driving force for the hydrocarbon seepage. However, the continental crust that constitutes the Levantine Basin basement73, can trap body waves near the surface at short distances ( > 150 km) and generated crustal Lg and Rg waves74. These crustal surface waves are visible in all 3 components, travel at speeds of 3.5 km.s−1to 2.8 km.s−1, and contain strong, low frequency amplitudes in the 0.2 to 5 Hz range75. Furthermore, the Türkiye earthquake sequence largely produced strike-slip type events. As such, our theoretical PGVs which correspond with measured vertical PGVs very likely underestimate maximum PGV at Site 4, as strike-slip events produce larger transversal waves than vertical ones. We note that we show only theoretical dynamic strain in Fig. 3 because not all earthquakes were recorded or visible in the ocean-bottom seismometer data. Strains shown in Fig. 3 can be found in Table S1. PGVs and strains shown in Fig. S2 can be found in Table S2.

Energy densities

We estimated energy densities for the earthquakes using empirical relations for the geometrical and physical attenuation of seismic energy and the relationship between seismic energy and magnitude, similar to previous studies investigating earthquake-triggered hydrological responses76: \({{\mathbf{log }}}\, {{\boldsymbol{r}}} = {{\mathbf{0.48}}}{{\boldsymbol{M}}}-{{\mathbf{0.33}}}{{\mathbf{log }}}\left({{\boldsymbol{e}}}\left({{\boldsymbol{r}}}\right)\right)-{{\mathbf{1.4}}}\), where r is distance in kilometers, M is magnitude, and e(r) is energy density in J m−3 as a function of distance. We assume this equation can be applied to our region, despite being strictly valid for California77.