Skip to main content

Reconstructing Earth’s mantle convection

Mid-Term Report Summary - AUGURY (Reconstructing Earth’s mantle convection)

The AUGURY team has pursued the goal of reconstructing the past motions of the Earth’s mantle and tectonics combining geological information and a new generation of 3D spherical mantle convection models. The focus is on the numerical aspects of the task. During the first half of the project, we first built a digital infrastructure for scientific computing, managing the digital heritage of the project in terms of calculation results and digital resources, and dissemination of the work ( In parallel, we made progresses on two different scientific directions: we improved our knowledge and competencies on models of mantle convection that self-consistently generate plate-like behavior, and we developed data assimilation tools to set the pace for tectonic and convection reconstructions of the Earth targeted for the second half of the project.

The aim to use mantle convection model with plate-like behavior as the basis of our data assimilation schemes for joint reconstructions of tectonics and convection imposes that we build knowledge about these models. This naturally leads to a better understanding of the dynamic links between mantle flow and tectonics. In 2014, with the group of Sydney, we proposed that the density contrast between continental lithosphere and mantle has a strong potential to initiate subduction, but mostly in the first 2 billion years of the Earth’s history. This paper is published in Nature. We also improved our knowledge about the organization of the plate layout on Earth. We showed that the size of the larger plates and the distribution of the smaller ones are determined by the dynamic feedback between the strength of the lithosphere and deep mantle flow. These results are published in Nature in 2016.

To prepare data assimilation, we had to estimate the forecasting power of convection models. Therefore, we measured the predictability limit of numerical models of mantle dynamics, like it is done for weather forecast for instance. This paper is published in Geochemistry Geophysics Geosystems in 2014. However, this limit is computed without considering model errors, therefore we investigated the impact of errors on rheological parameters in classical models of flow calculation in the mantle today. This is published in Earth and Planetary Science Letters in 2015. Unpublished tests led us to improve our parameterization of convection and to make a step forward, in particular for the modeling of subduction in global models. Using such parameterizations, we realized unprecedented tectonic forecasts experiments. These results are in review for PLOS One.

On the data assimilation side, we first had to evaluate the feasibility of reconstructing past mantle flow and tectonics from surface information only. For simplicity, we considered plate tectonic reconstructions as the data for the inversion (kinematics and seafloor ages) in the first place. Using a suboptimal sequential data assimilation scheme, we showed that the reconstruction problem could be successfully realized, in a paper published in Geophysical Journal International in 2016. This proof of concept led us to develop more sophisticated sequential data assimilation tools. Building on pre-existing data assimilation framework PDAF by Lars Nerger (Alfred Wegener Institute), we improved our scheme to be ready for applications to the Earth in the year to come. A manuscript is submitted to Nonlinear Processes in Geophysics.

A long-standing task of the project is to develop a variational data assimilation tool to transcend some of the limitations of the sequential methods we developed. Instead of directly coding the adjoint code, we preferred to use an automatically generated adjoint code for a more versatile and sustainable approach. With the FastOpt Company, we have achieved the automatic generation of the tangent linear model.
The first phase of code development and exploration of the model is now done, setting us ready for application to the Earth in the next phase of the project.