[1] The long-term average velocity field of the western United States is computed with a kinematic finite element code. Community data sets include fault traces, geologic offset rates, geodetic velocities, principal stress directions, and Euler poles. There is an irreducible minimum amount of distributed permanent deformation, which accommodates one third of Pacific-North America relative motion in California. Much of this may be due to slip on faults not included in the model. All data sets are fit at a common RMS level of 1.8 datum standard deviations. Experiments with alternate weights, fault sets, and Euler poles define a suite of acceptable community models. In pseudoprospective tests, fault offset rates are compared to 126 additional published rates not used in the computation: 44% are consistent; another 48% have discrepancies under 1 mm a À1 , and 8% have larger discrepancies. Updated models are then computed. Novel predictions include dextral slip at 2-3 mm a À1 in the Brothers fault zone, two alternative solutions for the Mendocino triple junction, slower slip on some trains of the San Andreas fault than in recent hazard models, and clockwise rotation of some domains in the eastern California shear zone. Long-term seismicity is computed by assigning each fault and finite element the seismicity parameters (coupled thickness, corner magnitude, and spectral slope) of the most comparable type of plate boundary. This long-term seismicity forecast is retrospectively compared to instrumental seismicity. The western United States has been 37% below its long-term average seismicity during 1977-2008, primarily because of (temporary) reduced activity in the Cascadia subduction zone and San Andreas fault system.
Motivation[2] There are at least two reasons to pursue a unified kinematic model of ongoing deformation in each of the world's orogens: (1) Dynamic theory and modeling (which involve rheology, stress equilibrium, and driving forces) will be more nearly correct when they develop from a good kinematic description of what is actually happening. (2) Any complete kinematic model can be converted to a long-term seismicity forecast, from which seismic hazard maps and seismic risk statistics can be computed for guidance of public policy and personal choices.[3] This paper contributes to both goals. By computing minimum rates of distributed permanent deformation (between model fault traces), I will show that this distributed deformation accommodates a significant fraction of relative plate motion in California, and that kinematic or dynamic models with purely elastic microplates separated by a small number of plate boundary faults are not appropriate. By converting the preferred model to a long-term seismicity forecast which is independent of historical seismicity, I highlight regions in which future seismicity will probably be greater than historical seismicity. A subsidiary goal is to illustrate a process for mapping of long-term seismicity which is rule-based, objective, and transparent, while providing a mecha...