← Back to Sessions

Towards Objective Models of Locking on Partially Creeping Faults and Subduction Zones

Description: 

Accurate modeling of earthquake scenarios and assessment of seismic hazards requires accurate estimates of the dimensions of likely earthquake sources. In cases where faults are partially creeping, as in most continental creeping faults and most subduction interfaces, dimensions of the fault area that is fully locked can be much smaller than the total fault area. Inverse models based on inversion of geodetic data are one way to estimate locking, but tend to be underdetermined and need strong regularization which can reduce resolution. In addition, most geodetic data are land-based, and provide weak constraints on locking models.

Boundary element models offer a means of adding additional constraints to such models through introducing additional physics. Elements on the fault surface can be 'locked' (backslipped or stationary with respect to a driving dislocation) or 'creeping' (freely sliding to accommodate stress). Stress shielding – the phenomenon where creep rates are reduced due to proximity of locked areas – adds some sensitivity to the location of locked elements, even where resolution of inverse models may be low.

I present a method for determining objectively the distribution of such locked elements, employing a modified Metropolis-Hastings algorithm with no intervention from the user. I will show two examples – the partially creeping Hayward fault in California, constrained by InSAR and GNSS data, and the Kamchatka subduction zone, constrained by GNSS velocities. In both cases I run the algorithm for 1 million iterations. After a short ‘burn-in’ period, each iteration produces an alternative plausible model of fault locking. The ensembles of models obtained for each case reveal details of fault behavior – including robust determination of areas where locking does not occur (<2% of models show locking) and the locations of likely asperities (locked in >70% of models), as well as areas where some locking may occur, but with poor constraints on location (locked in 20-50% of models). I identify additional structure in the model ensemble by applying hierarchical clustering based on Jaccard dissimilarity.

Session: Opportunities and Challenges in Source Modeling for Seismic Hazard Analysis

Type: Oral

Date: 4/20/2023

Presentation Time: 08:15 AM (local time)

Presenting Author: Gareth J. Funning

Student Presenter: No

Invited Presentation: 

Authors