Figure 13. (a) 3- month vertical prediction of an afterslip-only model. The coseismic fault width is 120km. (b) 3-month vertical prediction of a viscoelastic relaxation only model. The viscosity structure is 50km LAB, having a cold nose to the wedge and the wedge Maxwell viscosity is 5*1018 Pa-s. The error ellipse show 95% confidence. The region of the rupture areas is > 1m slip. Dashed light grey lines outlines the depth contours from the Slab2 model (Hayes et al., 2018). The white barbed line shows the plate boundary between the Pacific plate and the North American plate.
Figure S7 show the predicted vertical signal of viscoelasticonly models based on different assumed viscosity structures, and Figure S8 show the predicted vertical signal of aftersliponly models based on different fault widths. The observed displacement at site AC40 is about 1 cm uplift, and all of our afterslip models with different fault widths assumed predict about 1cm subsidence; this require a ~2cm uplift contribution from viscoelastic relaxation at site AC40. Our maximum viscoelastic relaxation contribution model used above (40km LAB + 1019 Pa-s wedge viscosity and no cold nose) only predicts ~0.5cm uplift at site AC40, but a different combination of model parameters might produce a larger uplift.
Figure S7 and Figure S1 show that by increasing the LAB depth and adding a cold nose to the wedge, the model predicts more uplift and a smaller horizontal displacement at site AC40. Thus, a cold nose to the wedge and a deeper LAB than used in our models is needed in order to get a predicted ~2cm uplift from viscoelastic relaxation at site AC40, although this also would require a lower mantle wedge viscosity. Together with afterslip, such a model can predict the observed vertical displacement with a Maxwell viscosity of the mantle wedge of 5*10 Pa-s. Thus, we consider a reasonable model with a 50km lithosphere thickness, a cold nose to the wedge. Because of a tradeoff between the lithosphere thickness (and cold nose geometry) and asthenospheric viscosity, a model with a lower viscosity can give about the same horizontal displacements as the models discussed earlier, although the vertical displacements and also far-field displacements will differ.
We computed the horizontal displacement of the viscoelastic relaxation based on this viscosity structure and subtract that from the GPS observation. We use the residu to search for the best-fit afterslip model based on the 10km coseismic fault. Finally, we add together the vertical prediction of viscoelasticonly model and afterslip model and compare that to the observation. Figure 14 shows the horizontal viscoelastic relaxation only prediction, the afterslip horizontal fit and the total vertical fit. The model is a reasonable fit to both the vertical and horizontal observations, but needs to be tested with a longer time span of data and data from sites farther from the rupture, and the parameter tradeoffs explored more thoroughly. In particular, there are strong tradeoffs between the geometry of the cold nose, lithospheric thickness, and wedge viscosity. These will be easier to explore at later times when the afterslip contribution is smaller.
(a) (b) (c)