Research
Throughout history, Earth’s geography has sparked curiosity, prompting researchers to investigate the intricate processes that shape our planet. A compelling theory suggests that the Earth’s visible surface is not merely a random arrangement but reflects the equilibrium and dynamic interactions occurring within its interior. To better understand this relationship, scientists employ physical models that conceptualize the Earth as a highly viscous fluid. This perspective allows us to gain valuable insights into the complex parametrization of the Earth’s interior and impose additional constraints on geophysical inversion models.
However, the computational demands of running these simulations are substantial due to the need for high spatial and temporal resolution, as well as the inherent complexity of the Earth’s interior, which is often both heterogeneous and nonlinear. Conventional high-fidelity numerical schemes have demonstrated inefficiency in such tasks, underscoring the need for more innovative approaches.
This project aims to develop a novel reduced-order modeling framework that leverages the latest advancements in Scientific Machine Learning (SciML) to overcome these challenges. The framework effectively addresses fully nonlinear 3D Stokes flow problems, ensuring the required efficiency without compromising accuracy.
Host
SISSA
Expected Results
Greater conceptual and technical understanding of geodynamic and geological processes involving crustal flows and of the numerical methods to address them. Development of open-source software to numerically solve nonlinear Stokes equations as well as to build an associated reduced order model with a focus on their application in crustal flows.