|dc.description.abstract||In the last decade, the number of seismic stations deployed globally increased dramatically, allowing to construct tomographic models with increasing resolution. Seismic data coverage however, is not homogeneous across the globe, and many regions are sampled by a distribution of ray paths that is both lower than average and uneven. In this work, we compute regional-scale tomographic models from massive, global waveform datasets, that we optimise for Africa, South America and the South- and North Atlantic Oceans, where coverage is highly heterogeneous.
To maximise coverage in the study areas, we assemble a very large dataset of both regional and global teleseismic waveforms, retrieving all freely available data in the Atlantic Ocean and surrounding continents. We then invert the waveforms using the Automated Multimode Inversion (AMI) of S-, Multiple S- and Surface waves. AMI produces a set independent linear equations with uncorrelated uncertainties for each source-receiver path, describing the path-average P- and S-wave velocity structure and dispersion curves within approximate sensitivity kernels. We then combine all the equations in a linear system and solve it for the 3D distribution of P- and S-wave velocities and 2-psi azimuthal anisotropy in the crust, upper mantle and transition zone. Similarly, we combine all phase velocity dispersion curves and invert them to produce 2D phase velocity maps independently at different periods. Finally, we exploit the mutual consistency of our very large dataset to automatically identify and remove the least consistent measurements from both our 3D models and phase velocity maps. In order to obtain the best possible models, for each study area we compute a different model that is tuned to yield the best results in the region. In South America and the South Atlantic Ocean, we parametrise our 3D model SA2109 on a ~300 km triangular grid and fine tune its regularisation with the aid of spike tests to yield robust results across the area. In South America, we image a cratonic lithosphere that is more complex than previously proposed and identify the boundaries of the cratonic cores of the Paran? and Parna?ba cratons; we also retrieve clear images of the Nazca slab, the Pampean slab gap, the subduction of the Chile Rise and the southernmost end of the subduction. In the South Atlantic Ocean, we image the low velocities of the Mid-Ocean Ridge (MOR) and a number of hotspots.
Using age-averaging techniques, we are able to identify the different cooling of the oceanic lithosphere in different oceanic basins, as well as the seismic signature of the Tristan da Cunha hotspot, masked by the MOR in tomographic images.
In Africa, our 3D model AF2109 reveals the presence of many smaller cratonic cores within the previously proposed boundaries of the West African and Congo cratons. Comparing the model with the outlines of cratons from field geology, we find that under the crust of the Angola, Tanzania and southwestern Kaapvaal shields, the craton lacks its lithospheric roots. By using a global dataset of diamondiferous kimberlites, we assess that in these areas, thick cratonic lithosphere was once present. Comparing the location and age of the kimberlites with our tomographic images, we infer that the Angola, Tanzania and Kaapvaal cratons lost a large portion of their cratonic lithosphere in the past 200 Million years (M.y). By combining the age of the youngest diamondiferous kimberlites in the area with Large Igneous Provinces and reconstructing the location of the cratonic erosion during the past 150 M.y., we infer that the erosion of the cratonic lithosphere followed the impact of mantle plumes on the cratons. In the Northeast Atlantic, we compute the 3D tomographic model NAT2019, which we parametrise over a very dense ~120 km grid. To tackle the very uneven coverage in the area, the model is regularised with 3D-varying coefficients that change in concert with data sampling. The model reveals the presence of a large, low-velocity anomaly, rooted under eastern Greenland in the transition zone, that rises, upwards and eastwards, towards Iceland. At 50-100 km depth, the low velocity anomalies distributes along the Reykjanes and Kolbeinsey Ridges. We interpret the low-velocity body as the Iceland Plume, captured by the MOR. Our model shows thinner lithosphere in the western part of the Northeast Atlantic, under the Greenland Plate. This is consistent both with the location of the ascent of the plume we image and the distribution of seamounts, more numerous on the Greenland Plate. Finally, we use the phase velocity dispersion curves produced by AMI to compute a set of 98 Love and Rayleigh phase velocity maps on a grid with average 225 km spacing that allows to extract densely spaced, robust dispersion curves anywhere on the globe. The data coverage at most periods is global, and the phase velocities, although computed independently for each period, vary smoothly across the produced maps. Compared to previous datasets, our maps sample the presence of strong lateral heterogeneities in the scale of few hundreds of kilometres. At short and intermediate periods, where the range of the sampled depths is narrowest, we can match the results to known geological structures such as cratons, orogenic belts and mid-ocean ridges. We then extract dispersion curves at all points of the model grid, that show, when clustered together, distinct trends for the oceanic and continental lithosphere. We tested the resolution of our dispersion measurements at a regional scale in northwestern Mongolia; at short periods, our dispersion curves sample the different structure between the Hangai Dome and the neighbouring Lake Region, ~500 km apart, with lower velocities underneath the Hangai Dome, in agreement with its higher elevation.||en