Inversion of Geodetic Data from Cascadia: A Comparison of Dislocation and Finite Plate Model Predictions

We use GPS velocities, surface tilt rates, and uplift rates to estimate the parameters controlling fault locking and rigid block rotation in the Cascadia subduction zone. We use both a traditional elastic half-space dislocation (EHSD) model and a finite elastic plate (FP) model in our inversions. For the EHSD model we solve for slip (or slip deficit) rates on the fault, and for the FP model we solve for stress rates on the fault and along the base of the plate. For both models we also solve for one or more rotation poles representing motion of rigid blocks within the overriding plate with respect to North America. The inversion is linear, subject to the nonlinear constraint that the slip (or shear stress) rates along the fault are parallel to the computed convergence direction between the subducting Juan de Fuca plate and the overriding block. Additional linear constraints may also be placed on the models, such as a positivity constraint for the computed quantities along the fault, and a constraint that the computed velocities approach zero at distances greater than 700 km from the trench, where we presently have no data. We performed inversions assuming either zero, one, or two rotation poles. For either model type, there is not a significant penalty for assuming a single pole rather than separate poles for points north and south of the Olympic-Wallowa lineament. There is a significant penalty when no rotation poles are assumed, although the penalty is smaller for FP models. For a wide range of model types, the EHSD single-pole models predict rotation poles that lie between 45 and 48N, 116 and 118W, and with rotation rates ranging from 0.7 to 0.9 rad/ma clockwise. There is greater variability among the FP models. Since the EHSD models utilize Green's functions that are much more localized than those for the FP models, these models must account for nearly all of the velocity field far from the trench in terms of block rotations, thus providing tighter constraints on the pole parameters.