Professor Louis Moresi

I am trying to understand the thermal-mechanical evolution of the Earth through geological time. This includes the fundamental question of how convective heat loss from the deep Earth is expressed mechanically as plate tectonics, the role of the continents in modulating this expression, the interaction between processes on this planetary scale with instabilities at the lithospheric scale — whether rheological or mechanical in nature — the influence of surface processes and atmospheric feedbacks on the solid Earth, and the signatures of all these processes that we can expect to find through geophysical observation of the present-day Earth, and in the long-term geological record.

Much of the complexity in the surface expression of mantle flow can be attributed to the non-linear nature of the constitutive laws. The Earth, on geological timescales, behaves as a non-linear viscoelastic fluid with a finite strength due to small-scale processes such as faulting and ductile shear localization which can be treated through the theory of plasticity. The underlying processes which we treat in this manner introduce a significant dependence on the stress, strain and thermal histories of the fluid representation. Plasticity is typically a phenomenological description of the material response to stresses and is cast in terms of the stress state of the material, from which the motions follow; in the Earth, stresses are less well known than the kinematics, and much is likely to be learned from plasticity models which are founded on the kinematics of surface motions.

In practice this spans a range of problems: the dynamics of subduction zones and the forces that drive plate motions; the origin and distribution of intraplate earthquakes, and how large earthquakes deform a viscoelastic planet; the deformation of continental crust at the basin scale and its coupling to erosion, sedimentation and landscape evolution; and the comparative evolution of other planets — Venus and Mars — where the same physics produces very different outcomes.

Questions like these usually need computational methods that do not exist when the question is first asked. My own contribution has been to particle-in-cell finite element methods for large-deformation, history-dependent materials, and to scalable parallel solvers for the elliptic problems that arise from realistic Earth-like rheology — viscoelasticity, anisotropy and strain-softening plasticity. That work is embodied in a lineage of open-source codes — CITCOM, CitcomS, Ellipsis and the Underworld platform — which run from the desktop to high-performance computing and cloud infrastructure. Underworld3 is designed to be mathematically self-describing: the equations a model actually solves can be recovered from the code itself.

I am correspondingly interested in what AI changes about computational geoscience: how large language models and agentic tools are reshaping the way research software is written, reviewed and documented; how correctness, reproducibility and comprehensibility are maintained when they do; and how self-describing mathematical code serves human and machine readers at once. This carries directly into education, where I teach data science and AI-assisted programming to Earth science students and have brought AI-assisted workflows into curriculum design.

Underlying all of it is a commitment to open, reproducible computational geoscience — source code released alongside the paper, benchmarks that can be re-run, and shared community infrastructure through the Computational Infrastructure for Geodynamics, AuScope and NASA's Science Explorer — so that the tools outlast any individual project.