Complex fault growth controls 3-D rift geometry: Insights from deep learning of seismic reflection data from the entire northern North Sea rift

Understanding how normal faults grow is critical to an accurate assessment of seismic hazards, for successful exploration of natural (including low-carbon) resources and for safe subsurface carbon storage. Our current knowledge of fault growth is, in large parts, derived from seismic reflection data of continental rifts and margins. These seismic datasets do however suffer from limited data coverage and resolution. In addition, detailed fault mapping in increasingly large seismic reflection data requires a large amount of expertise and time from interpreters. Here we map faults across the entire northern North Sea rift using a combination of supervised deep learning and broadband 3-D seismic reflection data. This approach requires us to interpret <0.1% of the data for training and allows us to extract almost 8000 individual normal faults across a 161 km wide (E-W), 266 km long (N-S) and 20 km deep volume. We find that rift faults form incredibly complex networks revealing a previously-unrecognised variability in terms of fault length, density and strike. For instance, while we observe up to 75.9 km long faults extending from the Stord Basin and Bjørgvin Arch in the south into the Uer and Lomre Terrace to the north, most faults (>90%) are closely spaced (< 5 km) and relatively short (<10 km long). Moreover, these faults show a large range of strikes varying from NW-SE to NE-SW with two dominant fault strikes (NE-SW & NW-SE) almost perpendicular to each other. This observation is difficult to reconcile with previous studies on the extension directions during rifting of the northern North Sea. While previous studies suggest that pre-existing shear zones control faulting in the northern North Sea, we only observe faults aligning with the southern parts of the Lomre shear zone and the eastern parts of the Ninian shear zones, but none of the other eight previously mapped shear zones. Instead we think that these variations in fault strike could occur naturally through the complex evolution of fault networks. As such our innovative approach allows us to map faults across the entire northern North Sea revealing complex networks, which challenge many conventional views of fault growth during continental rifting.