In this paper, we explore the use of novel neural-network architectures to distinguish natural seepages from artificial slicks in Synthetic Aperture Radar (SAR) images. They exploit a distinctive property of natural seepages, which is their temporal recurrence in the same geographical area. This information can be captured in different SAR images acquired at the same location over time, but not necessarily at a regular time-frequency. The proposed neural-network architectures are then built as specific block layers, which efficiently treat the unordered temporal information, followed by more conventional neural-network layers, which are widely used for image classification. Different block layers for unordered temporal information are compared on Sentinel-1 images acquired in the Aegean Sea. Following data augmentation steps, our dataset contains a consistent subset of images gathered among 16000 time-patches. We demonstrate that non-linear time-based block layers and block layers that avoid information bottlenecks are most efficient to discriminate between natural and artificial oil spills. Compared with standard neural-networks which use single time-frames, the integration of unordered temporal information increases the overall accuracy (OA) from 82% to 92% on our dataset, which demonstrates the effectiveness of the proposed approach.
Understanding the mechanisms controlling hydrocarbon seepage in marine environments is crucial due to the toxicity of liquid components and the contribution of the gaseous fraction to carbon budgets. While natural seepage is common, the role of earthquakes remains poorly understood. Here, we show that major earthquakes along the East Anatolian Fault in 2023 triggered hydrocarbon expulsion hundreds of kilometers away. Satellite monitoring revealed that nine days after the 7.8 moment magnitude mainshock, oil seepage began at a new site, leading to a 71% increase in the emitted oil volume over six months. Four events had energy densities and dynamic strains similar to known cases of earthquake-triggered hydrological responses. Our results suggest that low-frequency waves and high dynamic strains likely extended pathways and mobilized fluids in three-phase systems (oil, gas, water). Earthquake-driven hydrocarbon seepage should be monitored near active seismic zones to better account for greenhouse gas emissions in carbon budgets.
15 October 2026 in ISPRS Open Journal of Photogrammetry and Remote Sensing
The Lesser Antilles subduction zone is one of very few regions where old oceanic crust formed at slow spreading rates is being subducted. Crust accreted at slow spreading ridges differs from crust formed at higher rates, mainly in its higher content of material originating from the upper mantle, which is mostly hydrated to form serpentinites. The water stored in these serpentinites is released in the subduction process and then migrates upwards, towards the seafloor, where it forms fluid escape features, such as mud-volcanoes and pockmarks. Seismic reflection profiles from offshore Antigua and Barbuda image high amplitude seismic reflectors extending from the top of the downgoing crust roughly 15 km down into the mantle. They possibly originate from low-angle detachments related to exhumation of mantle material at the slow spreading Mid-Atlantic Ridge. As serpentinite rheology differs from that of basalt and gabbro, and because the amount of fluids from dewatering serpentinite is significantly higher than from mafic crust, the rheological properties of the plate interface and the margin are likely to influence slip behavior.During the Manta-Ray cruise in 2022, bathymetric and seismic data were acquired in the Lesser Antilles region with the objective to study the influence of subduction of this ultramafic basement on the tectonic deformation, fluid circulation and seismogenesis. At the accretionary prism fluid extrusion sites have been identified in the bathymetric data and their structures are finely imaged by high-resolution seismic data. Mud-volcano structures are linked to deeper faults and individual mud flows are imaged with a high enough a resolution to be described. In the region of the previously imaged deep reflectors, a 3D seismic experiment was conducted during which 23 deep sounding seismic profiles were acquired, crossing 75 ocean-bottom seismometers and using a 5000 cu-inch airgun array and a 6 km long seismic streamer. The extension of the reflectors east of the trench was proven and initial interpretation of the seismic data clearly shows the existence of several fault families dipping in different directions. In this area also numerous fluid extrusion features were identified which might have formed during accretion at the Mid-Atlantic ridge.During the LAVAS project, starting in 2025, we will further study the relationship between the hydration the downgoing plate and seismicity in the Lesser Antilles. In the scope of the project passive seismic data will be acquired using a sailing vessel along the different fluid extrusion sites to record signals from fluids leaving the seafloor. A submarine glider survey along these sites will help to identify possible methane accumulation in the water column. Satellite images will be used to identify natural hydrocarbons leaking from the seafloor and arriving at the sea surface. In an associated pedagogic project seismometers will be installed in schools on the islands of Guadeloupe and Saint Martin.