Measuring the fine-scale segmentation of the subduction megathrust in situ

The subduction megathrust and its frictional properties are central to controlling the short- and long-term dynamics of subduction zones. While the frictional properties are largely controlled by the 3D structure of the megathrust interface, our understanding of this structure is limited. Exhumed outcrops provide evidence for a complex mélange of ductile and brittle materials, which is segmented in a fractal manner. However, for active subduction zones, we lack direct evidence for such fine-scale structural segmentation as well as quantitative constraints on the segmentation structure.

Here, we use two high-resolution earthquake catalogs from the South American subduction margin to characterize the fine-scale segmentation in situ. We show that two overlying processes govern the fractal distribution of earthquake hypocenters. At short time scales, aftershock clustering dominates the earthquake distribution. At long time scales, averaging over many mainshock-aftershock sequences, the underlying structure shows. However, with typical catalog durations of only a few years, it is crucial to infer structure from short-term catalogs as well. We show, both theoretically and in our observational data, that even in short-term catalogs, structural constraints can be derived by looking at near-field interactions (< 100 m).

Based on our analysis, we find a fractal segmentation of the subduction interface in Northern Chile and Southern Peru with a fractal dimension D=1.6-1.7. This is consistent with the fractal distribution of brittle inclusions in exhumed outcrops. Notably, the fractal distribution is stable down to the hypocenter uncertainty (< 10 m), suggesting self-similarlity over several orders of length. We find an increase in fractal dimension with depth, suggesting a more uniform interface in the downdip region. This work provides a method to gain direct insights into the structure of the subduction interface and systematically quantify it. This way, we aim to connect structural observations to frictional properties and large scale dynamics.