A7 - Visualization of coherence and variation in meteorological dynamics
Principal investigators: Prof. Dr. Filip Sadlo, Prof. Dr. Rüdiger Westermann
Other researcher: Dr. Marc Rautenhaus (Postdoc), Michael Kern (PhD), Theresa Diefenbach (Master), Karsten Hanser (PhD)
Former researcher: Ole Klein (PhD)
In this project, we cover various needs of the W2W Collaborative Research Center in the analysis of spatio-temporal coherence and its variation in weather ensemble simulations. We will develop novel visualization techniques that provide interactive visualization of these aspects with focus on the 4D space-time structure of the individual fields and their ensembles.
We will develop new definitions of and efficient computational schemes for coherence indicator fields, Eulerian and Lagrangian, for both scalar and vector fields, and all also for ensembles. On the basis of these fields, new feature definitions relevant for meteorological research will be developed and combined with established definitions for warm conveyor belts, cyclones, jet-streams, and precipitation and convection features. To address error growth, we will build on our previous research on transport predictability in terms of Lagrangian coherent structures and the finite-time Lyapunov exponent, on transport of heat due to advection-diffusion, and on vorticity transport. Based on that, we will derive new techniques for specific visualization of error growth in weather forecast, also in the context of Rossby wave trains. While the resulting visualization techniques target a quantitative visualization of error growth, we will also derive new feature definitions in this context, involving space-time representation and further parametric domains, such as turbulence, uncertainty, and cloud properties, and develop efficient schemes to extract them.
Simultaneously, we will extend both the quantitative approaches and the feature-based approaches to ensemble simulation, developing techniques to extract and visualize the spatial and in particular the temporal variability of features. From this, spatio-temporal probabilities of feature occurrence will be derived, and the contribution of selected features to a probability region will be analyzed quantitatively.
Our research aims at providing visual depictions of the spatio-temporal variability of features, as well as means for identifying similarities in the shape and structure of these features. On the one hand, this requires to develop new visual abstractions for sets of dynamic surface- and volume-based structures from which the variability can be concluded. On the other hand, feature-based similarity metrics need to be developed and applied to group features and to determine most representative occurrences for selected groups.
To further improve our understanding of the complex dynamics of meteorological processes, visualization techniques must be able to show the spatio-temporal interrelations between different features. This will require the development of new visual representations to track ensembles of motion paths and their variations in space and time. In close collaboration with meteorologists, we will employ and adapt the visualizations to depict errors between meteorological forecast fields and observations.
One PhD position is available within this project. See the job description here.