Binary Probabilistic Inference of Megathrust Segmentation in the Presence of Slow Slip Events: Insights from the Mexico Subduction Zone

Axel J. Periollat, Gareth J. Funning, & Roland Bürgmann

Submitted August 30, 2026, SCEC Contribution #15193, 2026 SCEC Annual Meeting Poster #TBD

Subduction megathrusts are frictionally heterogeneous, with locked patches – the sources of major earthquakes – embedded within broader stably-sliding and transitional domains. Identifying compact locked areas is challenging because GNSS velocities combine long-term loading with transient processes (slow slip events, postseismic deformation); which can obscure the underlying locking distribution. Classical coupling maps rely on regularized inversions that produce a single smooth estimate of the megathrust’s slip behavior, making fine-scale locking difficult to detect when GNSS coverage is sparse or transient deformation affects the velocity field.

We introduce a Metropolis-Hastings (MH) sampler that explores locking configurations on realistic megathrust geometries. Each interface elements are modeled as either locked or freely sliding, as the long-term plate convergence rate. The MH algorithm samples the binary model space, favoring configurations that better fit the GNSS observations. From the resulting posterior ensemble, mean and entropy fields respectively identify the robust locking state and the variability allowed by the data.

Applied to the Mexico subduction zone – a transient rich-margin with limited GNSS density– the method recovers a sharp along-strike segmentation. Compact, low-entropy locked patches coincide with historical rupture areas, while low-entropy and unlocked domains align with the location of known Guerrero slow-slip events and the downdip tremor band, indicating that posterior ambiguity reflects genuine physical variability rather than sampling noise. These results show that probabilistic binary inference can isolate localized locking and creeping domains even under sparse GNSS coverage, providing a clearer and more interpretable alternative to deterministic coupling maps.

Key Words
GNSS, Locking, Probabilistic, Asperity

Citation
Periollat, A. J., Funning, G. J., & Bürgmann, R. (2026, 08). Binary Probabilistic Inference of Megathrust Segmentation in the Presence of Slow Slip Events: Insights from the Mexico Subduction Zone. Poster Presentation at 2026 SCEC Annual Meeting.

Related Projects & Working Groups
Tectonic Geodesy