An Updated Map of Tectonic Regionalization for Africa with Application to Gravity Inversion

Abstract:

Recent advancements in inversion of satellite gravity data suggest that laterally variable density contrasts associated with different geological regions help to determine the crustal architecture. Tectonic regionalization is predestined to map these different tectonic units. One way to obtain their lateral geometry is to apply k-means clustering to a seismic velocity model. This groups regions with similar vertical velocity profiles together. The k-means algorithm has been previously applied to the global tomography model SL2013 and it was found that the velocity structure naturally clusters into six distinct tectonic categories.

In this contribution, we focus on the regionalization of the recent seismic tomography model AF2019 for the African continent. We find that the number of clusters, which is the crucial variable when implementing the k-means algorithm, cannot directly be taken from the global tomographic model. Rather, the number of clusters should be adjusted according to the seismic tomography. We test several methods to identify the optimal number of clusters that define the different tectonic regimes in Africa.

As an alternative to the traditional k-means clustering, we show how trans-dimensional clustering, based on the reversible jump Monte Carlo Markov Chain approach, can be used to gain additional insight into the required model complexity and the number of clusters.

Our regionalization maps are designed to represent the mantle lithosphere only. Together with an extended seismic database we use the updated tectonic regionalization as independent constraints for gravity inversion to image the Moho depth underneath the African continent. This helps to differentiate the individual cratons, which might have undergone a different tectonic history. In general, we foresee the updated maps of tectonic regionalization as basis for interdisciplinary studies of the lithospheric architecture of the African continent.