A Holistic Approach to Modeling Contemporary Crustal Deformation in the Western U.S. by Comparing and Integrating Geodetic, Geologic and Seismologic Data

Interseismic strain accumulation in the crust is an important indicator of future seismic hazard. Strain rate estimation methods are, however, non-unique when based on geodetic velocities from a heterogeneous GPS network and results depend on the modeling methodology used. We present a new suite of strain rate models for the entire western U.S. (WUS) based on various combinations of geodetic, geologic, and seismologic data. These data sets each represent different aspects of the deformation field and each have their own pros and cons, including those related to their spatial distribution. Our main data set of thousands of horizontal GPS velocities is a combination of those 1) estimated by us based from time-series we produced for all continuous and semi-continuous stations, 2) estimated by us from publicly available time-series of USGS campaign stations, 3) published in the literature for which we do not have data (i.e., non-USGS campaign stations). Prior to the velocity estimation we correct the time-series for visco-elastic postseismic deformation and remove common-mode components using a newly developed algorithm. We correct the velocity field for the interseismic strain accumulation due to the locked Cascadia Fault, which would otherwise leave a strain rate signal in the continental crust that does not reflect local deformation. Geodetic strain rates are modeled with both the MELD algorithm [Kreemer et al. 2018; 2020] and the method of Haines and Holt [1993], with the former being the a priori estimate for the latter. The latest set of fault slip rates from the USGS fault and fold database are considered in combination with the GPS velocities, either as additional data or a priori constraints. Additionally, we evaluate the intensity and style of seismic deformation by estimating a map of Gutenberg-Richter a-values and moment tensor values, respectfully. The former was done based on a declustered ComCat catalog while estimating regional variation in completeness magnitude and b-value. We compiled all available focal mechanisms and used a robust imaging approach to estimate moment tensors throughout the WUS. We present a suite of models representing different combinations of the data types. Ultimately, the preferred model is that which, when converting to a seismicity forecast, does best in a prospective forecast.