Unravelling and interpreting plate motions in the presence of finite-rotation noise
IPGP - Îlot Cuvier
Séminaires Géosciences Marines
Australian National University, Canberra
Plate motions, which shape Earth’s surface through time, arise from the dynamic balance of shallow- and deep-rooted geological processes. These kinematics are of paramount importance to make inferences on Earth’s tectonic history, mantle/lithosphere interactions and dynamic topography among others. Over the last few years much effort has gone into high-temporal-resolution reconstructions of plate motions inferred from ocean-floor finite-rotations. However, measurements feature substantial noise arising mainly from the challenge of identifying precisely ocean-floor magnetic lineations, as well as from uncertainties in geomagnetic reversal timescales. It is therefore standard practise to smooth these reconstructions, typically by averaging over 2 to 5 Myr-long intervals. This, however, comes at the price of i) downgrading the native resolution of costly measurements and ii) having to choose one among several smoothing methods that do not yield a unique result. At the native resolution of reconstructions, however, the scenario arising is puzzling, as plate motions appear to vary erratically and significantly over short periods of less than 1 Myr. This equally undermines our ability to make geodynamic inferences, because the rates at which plate torques need to be built to explain these kinematics exceed – in some cases by far – the most optimistic estimates based, for instance, on the temporal evolution of lithospheric slabs into Earth’s mantle. I will focus on the longest and most temporally-resolved records of ocean-floor spreading ever produced to date, to show that the largest kinematic changes inferred actually relate to data-noise. I will also demonstrate how to overcome this limitation by resorting to Bayesian Inference, which drastically reduces noise without loss of temporal resolution. Changes in the temporal trends of plate motions occur on timescales no shorter than a few million years, yielding simpler kinematic patterns and more plausible dynamics. Finally, I will show how our inferences on driving and resisting forces upon plates indeed benefit from the reduction of noise through Bayesian Inference.