Constraints on the geometry of the subducted Gorda Plate with converted phases generated by local earthquakes

The largest slip in great megathrust earthquakes often occurs in the 10–30 km depth range, yet seismic imaging of the material properties in this region has proven difficult. We utilize a dense onshore-offshore passive seismic dataset from the southernmost Cascadia subduction zone where seismicity in the mantle of the subducted Gorda Plate produces S-to-P and P-to-S conversions generated within a few km of the plate interface. These conversions typically occur in the 10–20 km depth range at either the top or bottom of a ∼5 km thick layer with a high Vp/Vs that we infer to be primarily the subducted crust. We use their arrival times and amplitudes to infer the location of the top and bottom of the subducted crust as well as the velocity contrasts across these discontinuities. Comparing with both the Slab1.0 and the updated Slab2 interface models, the Slab2 model is generally consistent with the converted phases, while the Slab1.0 model is 1–2 km deeper in the 2–20 km depth range and ∼6–8 km too deep in the 10–20 km depth range between 40.25°N and 40.4°N. Comparing the amplitudes of the converted phases to synthetics for simplified velocity structures, the amplitude of the converted phases requires models containing a ∼5 km thick zone with at least a ∼10%–20% reduction in S wave velocity. Thus, the plate boundary is likely contained within or at the top of this low velocity zone, which potentially indicates a significant porosity and fluid content within the seismogenic zone.