Abstract

Numerical models of mantle convection have become increasingly more sophisticated in the last two decades and are growing in complexity.Unfortunately, the increase in resolution and the incorporation of new features based on empirical relationships, parametrizations and long-range extrapolations limit the predictive power of these models, while at the same time allowing for an easy fit of individual geophysical and geological datasets. This calls for a more thorough testing of geodynamic models and a shift in their use, from scenario calculations to explicit modeling of Earth's dynamics. A promising pathway to test geodynamic models relies on geodynamic retrodictions, that is reconstructions of past mantle flow states obtained using presently available information. They link explicitly assumptions and inferences from various fields of the Earth Sciences in a comprehensive Earth model based on the physics of mantle convection, thus highlighting their interrelations. Moreover, reconstructing the past history of mantle convection allows for a test of the underlying model against the geological record. The emphasis on the cross-disciplinary testing of model results against observations marks the transition from numerical to observational geodynamics, of which this dissertation can be seen as a primer. Using the Atlantic realm as a case study, we collect a number of geophysical observations - from seismic structure to the history of plate motions, from epeirogenic movements to gravity anomalies - and bring them to bear. We first analyse them using theoretical considerations and analytical solutions, to gain some fundamental insight and formulate clear geophysical hypotheses. We then proceed to formally solve the geodynamic inverse problem, obtaining the first retrodictions of geodynamically plausible mantle flow in the Atlantic region.