Unraveling Turbulence through Ensemble Decomposition

Turbulence governs essentially all large-scale flows on our planet, including our atmosphere and oceans.

With a vast number of engineering applications in transportation technology, renewable energies, and more, turbulence has a direct impact on our lives.

However, developing predictive theories of turbulence, which ultimately all modeling applications rely on, remains one of the outstanding scientific challenges.

Moreover, while massive simulations on the largest supercomputers are nowadays an established tool, reaching realistically high Reynolds numbers remains prohibitive.

Already today, analyzing the sheer amount of peta-scale simulation data requires new paradigms for making meaningful progress. Fundamentally new approaches are needed to achieve a breakthrough.

UniTED will deliver such an approach by a unique, synergistic combination of data-driven theory and large-scale computations. How? Recently, I showed that the complex statistics of turbulence can be disentangled into much simpler sub-ensembles.

This significant reduction of complexity points toward exciting new theoretical pathways and novel computational methodologies, which I will explore in this project. In UniTED, we will (A) dissect the multi-scale structure of turbulence through massively parallel computations.

This will (B) provide the foundation for a statistical theory of turbulence which is based on a novel ensemble decomposition approach.

Combining (A) and (B), we will (C) develop a novel ensemble-based simulation approach, enabling unprecedented insights into turbulence at high Reynolds numbers.

We will then use this approach to (D) provide big data for modeling small-scale turbulence using physics-informed machine learning.

UniTED will boost our fundamental understanding of turbulence at very high Reynolds numbers and provide new modeling approaches in a breadth of fields such as computational engineering, the Earth sciences, renewable energy, and plasma physics.