Abstract Constraining the effective rheology of major faults contributes to improving our understanding of the physics of plate boundary deformation. Geodetic observations over the earthquake cycle are often used to estimate key rheological parameters, assuming specific laboratory-informed classes of viscous or frictional rheological models. However, differentiating between various rheological model classes using only observations of a single earthquake (coseismic and postseismic deformation) is difficult—especially in the presence of coarse spatiotemporal sampling, inherent observational noise, and non-uniqueness of the inverted properties. In this study, we present a framework to estimate key rheological parameters of a subduction zone plate interface using simulations of sequences of earthquakes and aseismic slip, constrained by pre- and postseismic surface displacement timeseries. Our simplified forward model consists of a two-dimensional subduction zone, represented by a discretized planar fault or narrow shear zone, divided into a locked, shallow region (“asperity”) experiencing periodically imposed coseismic events, and a stress-driven creeping section governed by power-law viscoelasticity or rate-dependent friction. Our inverse model fits the rheological parameters of the interface to surface displacement timeseries in a Bayesian probabilistic way. We validate that our proposed framework can successfully recover depth-dependent profiles of effective viscosity using a synthetic dataset of pre- and postseismic observations. Our first set of numerical experiments show that our framework is only mildly sensitive to uncertainties in the rupture history or assumed coseismic slip, making it robust enough to be applied to real observations of subduction zones. Our second set of tests considers the similarities of surface displacement timeseries between synthetic models that model the plate interface either as a shear zone described by power-law viscosity, or a surface described by rate-dependent friction. Here, we find that the ability to fit surface observations using functional or mechanical models assuming frictional behavior does not constitute sufficient evidence to actually infer frictional behavior at depth, as the surface expressions are virtually indistinguishable from deformation generated from models with depth-variable power-law viscous behavior. Based on our numerical experiments, we conclude that studies that aim to infer the mechanical behavior and rheological properties at depth in subduction zones should consider the surface expression from time periods representative of the entire seismic cycle. Graphical Abstract
Read full abstract