Geodesy-based estimates of loading rates on faults beneath the Los Angeles basin with a new, computationally efficient method to model dislocations in 3D heterogeneous media

Tuesday, 15 December 2015: 15:10
2002 (Moscone West)
Christopher Rollins, California Institute of Technology, Pasadena, CA, United States, Donald F Argus, JPL, Pasadena, CA, United States, Jean-Philippe Avouac, California Institute of Technology, Geological and Planetary Sciences, Pasadena, CA, United States, Walter Landry, Caltech, Pasadena, CA, United States and Sylvain Barbot, Earth Observatory of Singapore, Singapore, Singapore
North-south compression across the Los Angeles basin is accommodated by slip on thrust faults beneath the basin that may present significant seismic hazard to Los Angeles. Previous geodesy-based efforts to constrain the distributions and rates of elastic strain accumulation on these faults [Argus et al 2005, 2012] have found that the elastic model used has a first-order impact on the inferred distribution of locking and creep, underlining the need to accurately incorporate the laterally heterogeneous elastic structure and complex fault geometries of the Los Angeles basin into this analysis. We are using Gamra [Landry and Barbot, in prep.], a newly developed adaptive-meshing finite-difference solver, to compute elastostatic Green’s functions that incorporate the full 3D regional elastic structure provided by the SCEC Community Velocity Model. Among preliminary results from benchmarks, forward models and inversions, we find that: 1) for a modeled creep source on the edge dislocation geometry from Argus et al [2005], the use of the SCEC CVM material model produces surface velocities in the hanging wall that are up to ~50% faster than those predicted in an elastic halfspace model; 2) in sensitivity-modulated inversions of the Argus et al [2005] GPS velocity field for slip on the same dislocation source, the use of the CVM deepens the inferred locking depth by ≥3 km compared to an elastic halfspace model; 3) when using finite-difference or finite-element models with Dirichlet boundary conditions (except for the free surface) for problems of this scale, it is necessary to set the boundaries at least ~100 km away from any slip source or data point to guarantee convergence within 5% of analytical solutions (a result which may be applicable to other static dislocation modeling problems and which may scale with the size of the area of interest). Here we will present finalized results from inversions of an updated GPS velocity field [Argus et al, AGU 2015] for the inferred distribution of locking and creep on 1) the Argus et al [2005] dislocation source and 2) major fault planes in the SCEC Community Fault Model, using the full 3D SCEC velocity model in addition to vertically stratified and elastic halfspace models.