Combining local earthquake tomography and petrological models to constrain wavespeeds in the subducting Nazca Plate

Intermediate-depth seismicity in Northern Chile shows a pattern which is distinct from a conventional double seismic zone (DSZ) setting. While two distinct seismicity planes are present in the updip part of the slab, there is a sharp transition to a highly seismogenic cluster of 2530 km thickness at 80-90 km depth, extinguishing the gap between the two seismicity planes. As seismic velocities can be used to constrain mineralogy and fluid content, characterizing seismic wavespeeds of this subduction zone segment using local earthquake tomography can provide important constraints on the mineralogical processes that produce the seismicity pattern seen here. We used the catalog of Sippl et al. (2018), which contains arrival time data from permanent stations of IPOC (Integrated Plate Boundary Observatory Chile), complemented by several temporary deployments spanning shorter time sequences. The catalog contains more than 100,000 earthquakes and 1,200,404 P- and 688,904 S-phase picks for the years 2007 to 2014. In order to use the best available picks for tomography, we limit our analysis to events that have more than 14 P-arrivals as well as more than 7 S-arrivals, leading to a total of 8883 events with 213,908 P- and 99466 S-arrivals. Parallelly, we also attempt to obtain an estimate of the possible mineral compositions at the depths and P-T conditions relevant to our study in the DSZ setting. For this, we assume a simple model where the upper plane of the DSZ is considered to be evolving from MORB-like composition and the lower plane of the DSZ from depleted-mantle composition. These global average compositions are then fed into Perple X (Connolly Kerrick, 1987) as starting compositions and pseudosections of possible mineral assemblages are constructed for P-T conditions significant to this study. We calculate the theoretical Vp and Vp/Vs values for those P-T conditions using the same software. We present 3D models of P- and S-wavespeeds from the Northern Chile forearc between about 20.4 S and 22.5 S, as well as images of ray coverage, relocated seismicity, and synthetic resolution tests. Tomography models for different choices of grid spacing and damping-smoothing parameters are compiled and compared to derive the optimal settings for the inversion. The seismic velocity distribution obtained through tomography is compared with the aforementioned theoretical wavespeeds to narrow down the range of possible reactions occurring at depth.