A7 - Visualization of coherence and variation in meteorological dynamics
We develop novel techniques to analyze meteorological features, the underlying physical mechanisms, and the numerical approaches used for their simulation. We focus on individual solutions as well as ensembles, and on Eulerian representation as well as Lagrangian finite-time transport. Our research aims at improving the robustness of feature extraction algorithms, by employing multi-scale, multi-physics field representations, as well as coherence-based concepts and clustering techniques. We further provide techniques that help to understand the physical processes that lead to meteorological phenomena such as heat waves and warm conveyor belts.
On the feature level, we go beyond Phase 1 by considering ensembles of features with vastly different geometry and topology, conveying their variability, and determining major trends and structural uncertainties. To shed light on the relationships between features and the characteristics of the underlying fields, we combine feature- and location-based uncertainty visualization. This is in particular challenging, because it requires the simultaneous visualization of feature and data distributions, and their tracking over time.
On the physics level, we develop visualization concepts that capture the interplay of finite-time advective dynamics and diabatic processes. This will, on the one hand, enable us to support a deeper understanding of the formation of heat waves and cloud systems with respect to advection, and provide insights into their spatio-temporal organisation. On the other hand, it will reveal fluxes of enthalpy across different scales in space and time, as well as relate features to each other in a physical manner.
Our third main field of research does not address the analysis of data—instead we aim at providing methods to analyze the operation of the simulation techniques that generate them. Data assimilation is a prominent tool in meteorological forecasting, but the interplay of the model, the prediction, and the measured data to be assimilated is very complex—in terms of spatio-temporal causalities, model design, and optimization of costly measurements. We address Kalman ensemble filters, with a focus on the Kalman gain, to study this interplay. Such analysis will help assess the importance of measured data—and more particular—the consistence between the prediction model and the obtained measurements, and also provide a basis for improving both the model and the acquisition of the data to assimilate.