Tectonic and lithospheric controls on arctic hydrothermal vent systems constrained by deep-learning enabled microseismicity catalogs

As the Earth’s tectonic plates diverge, new oceanic crust forms along mid-ocean ridges, but the mechanisms of crustal accretion change fundamentally with varying spreading rates. At ultraslow spreading ridges (<20 mm/yr), limited melt supply produces a highly heterogeneous lithosphere, where melt-rich volcanic centers alternate with non-volcanic segments shaped by tectonic extension and long-lived detachment faults. Despite their sparse melt supply, these ridges host abundant hydrothermal vent fields, demonstrating that vigorous hydrothermal circulation can persist even where tectonic processes dominate. However, the subseafloor processes influencing these systems remain largely unresolved, particularly along the sparsely surveyed Arctic mid-ocean ridges (AMOR). Clarifying the interplay of tectonic faulting, crustal permeability structure, and transient magmatic activity is therefore essential for understanding what controls hydrothermal systems at ultraslow spreading ridges.

Microearthquakes provide key constraints on active faulting and brittle deformation, making them an effective tool for assessing tectonic controls on hydrothermal systems. This thesis analyzes data from two year-long ocean bottom seismometer (OBS) deployments around two Arctic hydrothermal vent fields: Loki’s Castle, the first visually confirmed vent field along the AMOR, and Aurora, the first visually confirmed vent field beneath the perennial Arctic sea ice. Both sites host >300 °C black smoker vents on basaltic mounds but are situated in contrasting tectonic settings. Loki’s Castle lies on an axial volcanic ridge where the Mohns and Knipovich Ridge intersect, in an area of pronounced across-axis asymmetry expressed by magmatic accretion in the east and tectonic accretion in the west. Aurora is situated at the intersection of the magmatically robust Western Volcanic Zone (WVZ) and the non-volcanically spreading Lena Trough, and despite its position on a basaltic mound, its vent fluid chemistry indicates interaction with ultramafic rocks at depth.

To process the large number of microearthquakes recorded from both OBS deployments, I developed a workflow that applies modern deep-learning algorithms to OBS datasets, producing robust earthquake catalogs despite the limited station coverage and the noisy marine environment. While results show that automatically compiled catalogs can reproduce the main seismicity patterns of a manually compiled reference subset, careful evaluation of available algorithms is necessary to ensure reliable results. The amount of required manual re-evaluation can then be tailored to the dataset and scientific goals.

Application of this workflow to the two OBS datasets revealed fundamentally different relationships between tectonic deformation and hydrothermal circulation. At Loki’s Castle, seismicity is concentrated on the western rift valley flank along an emerging detachment fault, indicating that plate divergence is accommodated away from the vent field atop the axial volcanic ridge. This spatial decoupling suggests that the hydrothermal system is rather sustained by a localized magmatic heat source beneath the ridge, while the emerging detachment fault more likely only influences the broader fluid pathways. At Aurora, in contrast, seismicity is distributed beneath the vent field, and the lack of evidence for a localized heat source directly below implies that circulation is maintained by a persistent, fracture-controlled permeability system within a transitioning spreading environment. The microseismicity delineates a shift from magmatic to non-volcanic spreading from the WVZ toward Lena Trough, and a thickening aseismic layer below the vent field likely reflects ultramafic lithologies at depth. Although both sites exhibit basalt-hosted high-temperature venting, they are sustained by fundamentally different subsurface processes controlled by their local tectonic settings, melt supply, and lithospheric structure.

More broadly, by applying modern deep-learning methods to the first long-term OBS datasets from hydrothermal vent fields along ultraslow spreading ridges, this thesis advances our understanding of how lithospheric architecture and spreading style influence hydrothermal systems in melt-poor and tectonically dominated environments.