Seismic Sampling of the Mantle Under the South Pacific With a Drifting Array of Autonomous MERMAID Floats

An array of 50 floating seismic sensors has been deployed into the South Pacific to record hydroacoustic data in this vast and understudied region. Our drifting network spans roughly 6% of Earth's surface, encircles the South Pacific Superswell, and lies above the Pacific Large Low-Velocity Province (LLVP). The sensor is called MERMAID for Mobile Earthquake Recording in Marine Areas by Independent Divers. It is a fully autonomous oceangoing robot that records the ambient acoustic wavefield at 1500 m below the sea surface and, when triggered by its onboard detection algorithm, surfaces to transmit seismograms via satellite. With MERMAID we receive signals of global earthquakes within hours of their rupture from remote oceanic regions without any user intervention. Our array is named SPPIM--South Pacific Plume Imaging and Modeling--and it is returning tomographically-useful data that will refine images of the mantle under the South Pacific.

Our data are primarily high-frequency (around 1 Hz) P waves that have sampled novel ray paths at all depths in Earth: from the oceanic lithosphere below and between our stations, through subducting slabs under Tonga, into the deep mantle and across entire ocean basins, and even signals that have traversed Earth's core. We have developed methods for the high-precision picking of various phase arrivals in our data set and the estimation of their associated timing uncertainties.

Beyond showing in detail the quantity and quality of MERMAID seismograms we present, for the first time, our travel-time residuals computed against a three-dimensional model of P-wave velocities and smeared along their novel rays paths to give a geographic sense of the velocities anomalies that our instruments are recording. We find, very generally, that our residuals are delayed compared to the model, meaning that the signals recorded by MERMAID in the South Pacific are traversing regions of the mantle with lower-than-predicted wave speeds. Our residuals and their associated uncertainties will prove invaluable in future tomographic inversions and help to reveal the connection between deep-mantle features like the LLVP and surface features like the Superswell.