Mathurin D. Wamba, Frederik J. Simons , Jessica C. E. Irving 

Seismic tomography is a principal method for studying deep mantle plume structure. Imaging Earth's wavespeed anomalies is conditioned by
seismic wave sampling, and the uneven distribution of receiving stations worldwide leaves several candidate plumes beneath various
hotspots across the globe poorly resolved. We regionally evaluate two full-waveform global tomography wavespeed models, GLAD-M25 and
SEMUCB-WM1, focusing on the mantle below the Pacific Ocean in the region of the South Pacific Superswell. This area contains multiple hotspots which may be anchored in the Large Low Shear-Velocity Province at the base of the mantle. The two models show similarities and differences in the target region. With a goal of guiding subsequent iterations in the GLAD model suite, we assess the quality of GLAD-M25 in the target region relative to its global performance using a regional partition of the seismic waveform data used in its construction. We evaluate synthetic waveforms calculated using the spectral-element method, based on how well they fit the data according to a variety of criteria measured across multiple seismic phases, wave
types, and frequency bands. The distributions of travel-time anomalies that remain in GLAD-M25 are wider regionally than globally, suggesting
comparatively insufficiently resolved seismic velocity structure in the region of interest. This will motivate regionally focused inversions based on a subset of the global data set, and the addition
of data sampling new corridors, especially using ocean sensors. We compare GLAD-M25 and SEMUCB-WM1 by cross-validation with a new,
independent, data set. Our results reveal that short- and long-wavelength structure is captured differently by the two models. Global models use misfit criteria that may strive for balance between portions of the data set, but could leave
important regional domains underserved. Our results lead us to recommend focusing future model iteration and data addition on and
around the Pacific Superswell to better constrain seismic velocity structure in this area of significant geodynamic complexity.