Visualization: FlowZoom: Feature-based, multi-resolution in-situ sampling of very large turbulence simulations

Turbulence simulations have become so large, that the sheer size is an obstacle for even saving the data in full spatio-temporal resolution. This prevents the analysis of interesting phenomena at smaller temporal and spatial scales. This project targets a systematic assessment of small and large-scale phenomena in very large turbulence simulations. This includes the definition of tractable feature descriptors, development of robust detection and tracking strategies for flow phenomena of interest.

We worked on tracking and comparing the temporal development of features in time-dependent scalar fields. Based on a robust tracking, we compare their tracks on a spatio-temporal level. This means that we take their development into account, such as their growing or shrinking in size and intensity, and distinguish features with different developments. For example, this allows us to reveal different types of vortices created due to periodic vortex shedding in flows. Most importantly, our work automates the cumbersome process of finding and tagging similarities in large, time-dependent data sets.