An Anticipation Experiment for Plate Tectonics

Abstract

Although plate tectonics has pushed the frontiers of geosciences in the past 50 years, it has legitimate limitations, and among them we focus on both the absence of dynamics in the theory and the difficulty of reconstructing tectonics when data are sparse. In this manuscript, we propose an anticipation experiment, proposing a singular outlook on plate tectonics in the digital era. We hypothesize that mantle convection models producing self‐consistently plate‐like behavior will capture the essence of the self‐organization of plate boundaries. Such models exist today in a preliminary fashion, and we use them here to build a database of mid‐ocean ridge and trench configurations. To extract knowledge from it, we develop a machine learning framework based on Generative Adversarial Networks (GANs) that learns the regularities of the self‐organization in order to fill gaps of observations when working on reconstructing a plate configuration. The user provides the distribution of known ridges and trenches, the location of the region where observations lack, and our digital architecture proposes a horizontal divergence map from which missing plate boundaries are extracted. Our framework is able to prolongate and interpolate plate boundaries within an unresolved region but fails to retrieve a plate boundary that would be completely contained inside of it. The attempt we make is certainly too early because geodynamic models need improvement and a larger amount of geodynamic model outputs, as independent as possible, is required. However, this work suggests applying such an approach to expand the capabilities of plate tectonics is within reach.

Constraining the Range and Variation of Lithospheric Net Rotation Using Geodynamic Modeling

“…With the relatively small number of models we have for this study, the complexity means that we cannot predict or invert for LNR at any particular point in time. With a much larger number of simulations (probably tens of thousands more), we may be able to use statistical methods, such as machine learning tools, to find characteristics of mantle flow that can determine the magnitude of LNR (Atkins et al., 2016 ; Gillooly et al., 2019 ). This is still not particularly practical for absolute plate motion models, because we have limited information on the tectonic arrangement of continents in the geological past.…”

Deep learning for surrogate modeling of two-dimensional mantle convection

“…By using multiple parameters such as, e.g., mantle reference viscosity (related to Ra) and activation volume and activation energy of diffusion creep rheology (controlling the temperature and pressure dependence of the viscosity) as inputs to the FNN, it is possible to directly predict the thermal evolution of the entire horizontally averaged 1D temperature profile of the mantle in time with a mean accuracy of 99.7%. Another interesting study uses generative adversarial networks to reconstruct missing plate boundaries derived from horizontal divergence maps of a steady-state 3D convection model [28].…”