Exploring fault geometry uncertainties in finite-slip inversion with multiple moment tensor inversion

Finite-fault source inversions are often performed with an assumed fault geometry. Green’s functions calculated using a fixed strike, dip and rake can introduce hard-to-quantify errors in the inversions if the true fault geometry deviates from the model assumptions. For mega-earthquakes with large rupture areas, it is important to consider fault curvature and other non-planar fault effects. To accommodate uncertainties in the fault geometry, we propose to parameterize the fault as multiple moment-tensor sources within a 3D grid of possible source locations. The grid is defined with respect to an assumed initial fault plane, spanning the likely 2D rupture extent and a small range in the direction to normal to the fault to accommodate errors in assumed fault location or orientation. Instead of determining the slip-rate history, a moment-rate function is solved. There are six unknowns per point at a given frequency and the best-fitting double-couple (i.e., strike and dip) can be extracted from the results. If we assume the earthquake occurs within a few fault planes, the unknowns to be solved should be spatially sparse. Recently emerging tools such as compressive sensing (CS) can be used to deal with the problem. For a case study, we will analyze the 2013 Okhotsk Mw 8.3 earthquake and hope to understand the uncertainty limits caused by fault geometry in the finite fault modeling.