Juan de Fuca subduction zone from a mixture of tomography and waveform modeling

Seismic tomography images of the upper mantle structures beneath the Pacific Northwestern United States display a maze of high‐velocity anomalies, many of which produce distorted waveforms evident in the USArray observations indicative of the Juan de Fuca (JdF) slab. The inferred location of the slab agrees quite well with existing contour lines defining the slab's upper interface. Synthetic waveforms generated from a recent tomography image fit teleseismic travel times quite well and also some of the waveform distortions. Regional earthquake data, however, require substantial changes to the tomographic velocities. By modeling regional waveforms of the 2008 Nevada earthquake, we find that the uppermost mantle of the 1D reference model AK135, the reference velocity model used for most tomographic studies, is too fast for the western United States. Here, we replace AK135 with mT7, a modification of an older Basin‐and‐Range model T7. We present two hybrid velocity structures satisfying the waveform data based on modified tomographic images and conventional slab wisdom. We derive P and SH velocity structures down to 660 km along two cross sections through the JdF slab. Our results indicate that the JdF slab is subducted to a depth of 250 km beneath the Seattle region, and terminates at a shallower depth beneath Portland region of Oregon to the south. The slab is about 60 km thick and has a P velocity increase of 5% with respect to mT7. In order to fit waveform complexities of teleseismic Gulf of Mexico and South American events, a slab‐like high‐velocity anomaly with velocity increases of 3% for P and 7% for SH is inferred just above the 660 discontinuity beneath Nevada.