Understanding tens of thousands of pockmarks and domes using machine learning (Gulf of Lions, NW Mediterranean Sea).

A broad spectrum of seabed deformation structures, including pockmarks and domes, related to various processes, have been observed on continental margins worldwide. This study provides the first regional-scale inventory and quantitative characterization of pockmarks and domes potentially linked to subsurface fluid-escape processes on the continental shelf and upper slope in the Gulf of Lions (NW Mediterranean). Using high-resolution multibeam bathymetry and seismic reflection data, approximately 29,000 pockmarks and domes were identified. Morphometric and seismic analyses reveal distinct groups that differ in size, shape, and spatial organization. Among them a widespread population of domes rooted on the Holocene maximum flooding surface, which represent ~ 85% of the occurrences. The remaining 15% are pockmarks, divided into five main subtypes. Spatial association of domes and pockmarks suggests a genetic relationship between dome formation and subsequent collapse (pockmarks). Machine learning-based spatial modeling extends the estimated total to ~ 80,000 between 10 and 1000 m water depth, including zones not covered by swath-bathymetric surveys. Among various controlling factors, sediment thickness, water depth, and grain-size variability exert primary controls on the distribution of pockmarks and domes. These results demonstrate that seabed deformations in the Gulf of Lions are spatially organized, reflecting the interplay between fluid-related processes, stratigraphic architecture, and sediment properties.