We present a seismic model of the African Plate, made with the technique of full-waveform inversion. The purpose of our model is to become a foundation for future use and research, such as quantitative geodynamic interpretations, earthquake-induced ground motion predictions, and earthquake source inversion. Starting from the first-generation Collaborative Seismic Earth Model (CSEM), we invert seismograms filtered to a minimum period of 35 s and compute gradients of the misfit function with respect to the model parameters using the adjoint state method. In contrast to the conventional FWI approach, we use dynamically changing data subsets (mini-batches) of the complete dataset to compute approximate gradients at each iteration. This approach has three significant advantages: (1) it reduces computational costs for model updates and the inversion, (2) it enables the use of larger datasets without increasing iteration costs, and (3) it makes it trivial to assimilate new data since we can add it to the complete dataset without changing the misfit function, thereby enabling “evolutionary FWI". We perform 130 mini-batch iterations and invert data from 397 unique earthquakes and 184,356 unique source-receiver pairs at the cost of approximately 10 full-data iterations. We clearly image tectonic features such as the Afar triple junction. Particularly interesting are the low-velocity zones below the Hoggar, Aïr, and Tibesti Mountains, pronounced more than in earlier works. Finally, we introduce a new strategy to assess model uncertainty. We deliberately perturb the final model, perform additional mini-batch iterations, and compare the result with the original final model. This test uses actual seismic data instead of artificially generated synthetic data and requires no assumptions about the linearity of the inverse problem.
