Automatic geological mapping from geodynamic models using offline Lagrangian tracers and Radial Basis Function interpolation

A major symptom of the dynamic evolution of planets is their geological map. Rock types express the imprint of the deep and surface processes. Structures often represent the tectonic styles of planets. Fossil content provides clues on past life evolution and environments. Although geologic maps are among the most used tool to infer planetary evolution, we still miss quantitative systemic links between planetary scale processes and the structure and nature of the maps. Also, the nature of rocks at the surface determines the chemistry of atmosphere-ocean composition.

The development of Earth system models that integrate geodynamic models open the possibility of generating synthetic geological mapping, a major step to link maps to dynamic evolution. Global geodynamic, climate and surface processes models provide analogous synthetic data. However, tracking these quantities through space and time and building progressively a geological map remains challenging, primarily due to the enormous quantity of information to proceed and interpret.

Here, we present an automated approach to produce geological maps from geodynamic models based on “offline” Lagrangian tracers. Tracers are introduced during post-processing and advected using a fifth-order Dormand-Prince scheme, while physical quantities are interpolated using radial basis functions (RBF). The mesh-free nature of RBF interpolation provides large flexibility in handling data of variable spatial resolution and discretization. This is particularly advantageous for coupling geodynamic models with surface processes, climate or biological evolution models, each relying on distinct spatial representation.

Like geologists combine observations to produce geological maps and interpret geodynamic systems, we integrate synthetic data to produce synthetic geological maps evolving over time. The developed software is parallel and HPC-ready, enabling efficient processing of large-scale models. Beyond geological mapping, the framework is fully generic and can interpolate and advect arbitrary fields defined on arbitrary discretizations, making it a versatile tool for multi-physics Earth system modelling.