Incorporating fault mechanics into inversions of aftershock data for the regional remote stress, with application to the 1992 Landers, California earthquake

- New algorithm to invert aftershocks for remote stress incorporates fault mechanics.
- Algorithm performs well when aftershock slip directions are considered.
- Complete earthquake solution provided: fault slip, stress, and deformation fields
- Inversion results stable for sliding friction values on main faults less than 0.5.
- Algorithm accommodates additional datasets (GPS, InSAR, fractures, etc).

We present a new stress inversion algorithm that accounts for the physics relating the remote stress, slip along complex faults, and aftershock focal mechanisms, in a linear-elastic, heterogeneous, isotropic whole- or half-space. For each new remote stress, the solution of the simulation is obtained by the superposition of three pre-calculated solutions, leading to a constant time evaluation. Consequently, the full three-dimensional boundary element method model need not be recomputed and is independent of the structural complexity of the underlying model. Using a synthetic model, we evaluate several different measures of fit, or cost functions, between aftershocks and model results. Cost functions that account for aftershock slip direction provide good constraint on the remote stress, while functions that evaluate only nodal plane orientations do not. Inversion results are stable for values of friction ≤ 0.5 on mainshock faults. We demonstrate the technique by recovering the remote stress regime at the time of the 1992 M 7.3 Landers, California earthquake from its aftershocks and find that the algorithm performs well relative to methods that invert earthquakes occurring prior to the Landers mainshock. In the mechanical inversion, incorporating fault structures is necessary, but small differences in fault geometries do not impact these inversion results. Each inversion provides a complete solution for an earthquake as output, including fault slip and the stress and deformation fields around the fault(s). This allows for many additional datasets to be used as input, including fault surface slip, GPS data, InSAR data, and/or secondary fracture orientations.