Africa's Crustal Architecture Inferred From Probabilistic and Perturbational Inversion of Ambient Noise: ADAMA

Africa's continental crust hosts a variety of geologic terrains and is crucial for understanding the evolution of its longest-lived cratons. However, few of its seismological models are yet to incorporate the largest continent-wide noise dispersion data sets. Here, we report on new insights into Africa's crustal architecture obtained using a new data set and model assessment product, ADAMA, which comprises a large ensemble of short-period surface wave dispersion measurements: 5-40 s. We construct a continent-wide model of Africa's Crust Evaluated with ADAMA's Rayleigh Phase maps (ACE-ADAMA-RP). Dispersion maps, and uncertainties, are obtained with a probabilistic approach. This model update, and a crustal taxonomy derived from unsupervised machine learning, reveals that the architecture of Africa's crust can be classified into two main types: primitive (C1: faster velocities with little gradients) and modified (C2-C4: slower velocities in the shallow crust with more pronounced gradients). The Archean shields are "primitive," showing little variation or secular evolution. The basins, orogens, and continental margins are "modified" and retain imprints of surface deformation. The crustal taxonomy is obtained without a-priori geological information and differs from previous classification schemes. While most of our reported features are robust, probabilistic modeling suggests caution in the quantitative interpretations where illumination is compromised by low-quality measurements, sparse coverage or both. Future extension of our approach to other complementary seismological and geophysical data sets-for example, multimode earthquake dispersion, receiver functions, gravity, and mineral physics, will enable continent-wide lithospheric modeling that extends resolution to the upper mantle. The rocks that constitute Africa's crust record the history of different geological periods. We produce a map, for the entire continent, of how fast shear waves travel within these rocks. We obtain this map from ambient noise surface wave vibrations. The ambient noise surface waves are generated from ocean and atmospheric waves that couple with the solid Earth. There are two types: Rayleigh and Love waves and they travel at different speeds for different wavelengths. This property is called dispersion and it is used to tell how fast the shear wave speeds travel within the subsurface rocks. Constructing the final map from ambient noise surface waves requires the solution of a computational imaging problem. We solve the most challenging computational task with a probabilistic approach-using random sampling-and this enables us to also construct associated error maps. The new maps of Africa's crust show new features that have important implications for subsurface geology of the continent. A continent-wide s-velocity model of Africa's crust is constructed using probabilistic modeling of the largest catalog of dispersion measurementsA crustal taxonomy, derived with unsupervised machine learning, reveals that Africa's crust is one-third primitive and two-thirds modifiedArchean shields are primitive and show no secular evolution; basins, orogens, and margins are modified, retaining imprints of deformation