Assimilating Intra-Plate Lava Geochemistry into Adjoint Reconstructions of Earth’s Mantle Evolution

Reconstructing the thermo-chemical evolution of Earth’s mantle across geological time is a central challenge in the geosciences. Addressing this problem increasingly relies on adjoint-based approaches, which cast mantle convection modelling as an inverse problem and enable the systematic assimilation of observational data into time-dependent simulations. Such methods underpin emerging efforts to build a digital twin of Earth’s mantle: a dynamic, physics-based representation constrained by diverse geological and geophysical observations.

To date, adjoint geodynamic inversions have primarily relied on constraints that act at the beginning or end of model evolution, or at Earth’s surface only, such as plate motions, geodesy, or seismic tomography. However, these datasets provide limited leverage on the evolving thermal and chemical structure of the mantle through time. Intra-plate volcanic lavas offer an underexploited observational constraint, as their major- and trace-element geochemistry records the pressure, temperature, and composition of mantle melting at the time of eruption, providing direct insight into past lithospheric thickness, plume excess temperature, and mantle source heterogeneity.

Here, we present an integrated framework for assimilating geochemical information from ocean island basalts into adjoint models of mantle convection using the Geoscientific ADjoint Optimisation PlaTform (G-ADOPT). Using simulation-informed inversions of rare earth element concentrations, we demonstrate the power of geochemical data to recover the thermal structure of plume melting regions, including lithospheric thickness and plume excess temperature. We then use synthetic experiments to show how these geochemically derived constraints on melting conditions can be incorporated into adjoint reconstructions, substantially improving recovery of mantle temperature fields and flow trajectories relative to inversions based on surface or boundary constraints alone.

By explicitly linking geochemical observables to mantle thermal structure and flow, this approach reduces non-uniqueness in time-dependent inversions and strengthens the ability of adjoint models to retrodict mantle evolution. More broadly, it highlights the transformative potential of integrating geochemistry into data-assimilative geodynamic frameworks and represents a key step toward a fully constrained digital twin of Earth’s interior.