Locating Fully Locked Asperities Along the South America Subduction Megathrust: A New Physical Interseismic Inversion Approach in a Bayesian Framework

Abstract

The largest earthquakes in subduction zones occur where significant interseismic slip deficit has accumulated on the plate interface. Slip deficit accumulates most quickly in mechanically locked regions, and these also cause the regions around them to accumulate slip deficit; therefore, large earthquakes are typically expected to rupture in and around locked areas. The locations and dimensions of these locked zones have been difficult to resolve using standard techniques and available data sets. We develop a new statistical interseismic inversion approach that incorporates the physical interactions between nearby fault areas to directly determine the distribution of locking on the subduction plate interface (simultaneously with rigid forearc motions) from interseismic surface velocities. Because we include physical prior information in the inversion procedure, this approach reduces uncertainties in the rate of slip deficit accumulation, even in locations (such as near the trench) where kinematic inversions of onshore data have relatively low resolution. Applying the inversion to the South America subduction zone, we find that the pattern of locking and corresponding slip deficit rates correlate well with recent and historical large earthquake ruptures. Locked patch dimensions are <40 km and account for no more than 30% of the area of the plate interface. The small size of the imaged locked zones is a natural outcome of our physical assumptions and implies that mechanical locking is caused by correspondingly small geological features. Despite their small dimensions, locked zones generate substantial slip deficit on the surrounding plate interface, consistent with the slip patterns of large megathrust earthquakes.

Energy‐Based Scenarios for Great Thrust‐Type Earthquakes in the Nankai Trough Subduction Zone, Southwest Japan, Using an Interseismic Slip‐Deficit Model

“…The estimation errors for the stress accumulation distribution were amplified relative to those of the slip-deficit rates, because the stress distribution was calculated as the spatial derivative of the displacement field. For instance, it is possible that the actual mechanical coupling zones are much smaller than those estimated in this study (Herman & Govers, 2020). However, the shear stress change rate distribution in this study was related spatially to very-low-frequency earthquakes at the shallow plate boundary along the Nankai trough (Takemura et al, 2019), which would reflect the heterogeneity of the effective frictional strength.…”

A Transient in Surface Motions Dominated by Deep Afterslip Subsequent to a Shallow Supershear Earthquake: The 2018 Mw7.5 Palu Case

“…Seismological data shown in Figures 1 and 2 for this research are included in this paper (and its supporting information files) (Supendi et al, 2020). Software for computing fault slip and surface displacements and performing the search is available online (https://doi.org/10.5281/ zenodo.3894137) and we follow the methodology outlined in Herman and Govers (2020). We use the PS-GRN/PSCMP software of R. Wang et al (2006) to compute viscous relaxation, which is available online (https://git.pyrocko.org/pyrocko/fomosto-psgrn-pscmp/).…”

“…We use the Metropolis‐Hastings algorithm (following Herman and Govers (2020)) to sample from the posterior probability distribution as we search the parameter space. A resultant, large ensemble of models provides a distribution of plausible model input parameters given the data; we combine the results of several, separate, randomly initialized search chains, after disregarding the first 50,000 iterations per chain that we treat as burn‐in.…”

Regional and Local Patterns of Upper‐Plate Deformation in Cascadia: The Importance of the Down‐Dip Extent of Locking Relative to Upper‐Plate Strength Contrasts

“…(2018) term the regions on the subduction interface surrounding the boundaries of such asperities to be “pseudo‐coupled”, because they cannot freely slip at the full plate motion rate (and thus they can and do accumulate slip deficit) as a result of their proximity to locked portions (asperities) of the interface. This relationship between fully locked asperities and the adjacent unlocked regions likely explains much of the inferred partial or apparent coupling within models – regions on the interface that appear to have a slip deficit accumulation rate less that of the full plate motion rate (Herman & Govers, 2020). Although kinematic inversions approximate the effects of pseudo‐coupling with spatially varying slip deficit magnitudes (full in regions of coupling and lower values in ‘pseudo‐coupled’ regions), as shown in the results of Figure 4 such models do not easily allow one to distinguish between areas of mechanical coupling and pseudo‐coupling.…”