Statistical Analysis of GPS Vertical Uplift Rates in Southern California
Thursday, 18 December 2014
Variations in crustal surface velocities obtained from GPS stations provide key constraints on physical models that predict surface deformation in response to earthquake cycle loading processes. Vertical GPS velocities, however, are highly susceptible to short scale (<10’s km) variations in both magnitude and direction induced by local changes in water-storage, pore pressure, precipitation, and water runoff. These short-wavelength spatial variations both dominate and contaminate vertical GPS velocity measurements and often mask coherent long-wavelength deformation signals. Because of these complications, vertical GPS velocities, like those provided by EarthScope’s Plate Boundary Observatory (PBO), have traditionally been omitted from crustal deformation models. Here we attempt to overcome these obstacles by first eliminating GPS velocities influenced by non-tectonic deformation sources based on high-resolution InSAR data. Second, we employ model selection, a statistical technique that provides an objective and robust estimate of the velocity field that best describes the regional signal without overfitting the highly variable short-wavelength noise. Spline-based interpolation techniques are also used to corroborate these models. We compare these results to published physical models that simulate 3D viscoelastic earthquake cycle deformation and find that the statistical PBO vertical velocity model is in good agreement (0.55 mm/yr residual) with physical model predictions of vertical deformation in Southern California. We also utilize sources of disagreement as a tool for improving our physical model and to further inspect non-tectonic sources of deformation. Moreover, these results suggest that vertical GPS velocities can be used as additional physical model constraints, leading to a better understanding of faulting parameters that are critical to seismic hazard analyses.