Scientific visualization of large-scale vector fields with modern, so-called integration-based methods that rely on the analysis of particle trajectories is not feasible with current methods, since existing algorithms cannot make efficient use of parallel architectures such as clusters and supercomputers. This status leaves researchers in science and industry unable to visualize, analyze and understand the processes described by large vector field data from simulation or measurement. The project aims at developing a novel methodological framework for integration-based visualization that will provide visualization of largest-scale vector fields in current scientific applications. The novel methodology will allow the efficient use of parallel architectures for fast and interactive visualization of very large vector field data sets, which is not possible with current methods. The projects approaches will combine techniques from scientific visualization, parallel algorithms, applied mathematics, and software design. The resulting increased ability to study large vector fields will strongly impact fundamental scientific research in a large and interdisciplinary setting of scientific and industrial application areas that rely on vector field visualization. This includes research on technologies related to timely problems such as combustion, fusion, and aerodynamics.
Query-Driven Visualization (QDV) is a visualization-based discovery strategy that combines state-of-the-art methods from Scientific Data Management with modern visualization approaches to support rapid data analysis. By restricting computational and cognitive workload in visualization and interpretation to records defined to be significant by a scientist, fast visualization responses can answer intuitive questions about the data. Thus, query-driven techniques are ideal tools for data discovery and hypothesis testing. However, uncertainty in data and query can strongly and negatively influence a visualization result and hence the insight obtained from it. We provide a novel framework that generalizes QDV and is aimed at addressing deficiencies in existing methods. By modeling uncertainty at all levels of the query-driven process, we enable robust visualization techniques that incorporate the uncertain nature of the analysis process into the visualization result, thus enabling tools that will allow users to factor uncertainty into the conclusions drawn from data sets. The methods we develop are leverage multi-resolution data representations incorporating uncertainty information; this results in improved efficiency and parallel computation in answering queries over very large, high-dimensional data sets. (official project page)
Development of visualization techniques for isogeometric finite element data that is based on NURBS representation (in collaboration withthe Differential-Algebraic Systems group at TU Kaiserslautern).
Investigation of new ways to understand parameter fitting models in spectroscopy applications (together with the Molecular Biophysics group at TU Kaiserslautern).
Collaboration on the design and implementation of scalable visualization algorithms that allow the visualization of very large datasets on current and next-generation supercomputers (with Prof. Childs' group at the University of Oregon.).
Joint development of interaction techniques specifically aimed at understanding and analyzing ensemble data consisting of many simulation runs over a large parameter space (with IDAV, UC Davis).
Investigation of new algorithms for parallel particle integration that can serve as a basis for next-generation integration-based vector field visualization techniques (with D. Camp, Lawrence Berkeley National Laboratory).
Implementation of new visualization techniques into the open-source visualization software tool VisIt.
Lagrangian methods assume a central role in the analysis and visualization of vector fields resulting from simulation and measurement across many application domains. These methods provide key insight into vector field structures and dynamics, but are based on the expensive computation of integral curves. Applied to large-scale problems and data sets, they are burdened heavily by enormous computational cost. To improve this situation, we replace the on-demand and problem-specific computation of integral curves typically employed in vector field visualization techniques by a two-stage process consisting of an adaptive pre-computation of integral curve sets and methods for interpolation within these sets, effectively transferring the vector field representation to the Lagrangian domain. Hence, we isolate the computational burden into a pre-computation stage, and transfer the obtained Lagrangian representation of a vector field into efficient out-of-core data structures. Integration-based visualization algorithms can leverage fast interpolation of integral curves, whose approximative characteristics we examine in detail, from this pre-computed data. Based on this generic framework, we enhance existing integration-based visualization methods to become interactive, and provide a basis for research into fast and efficient and interactive vector field visualization for very large vector fields. We take advantage of the expected performance and efficiency of the proposed methods to develop novel visualization tools based on Lagrangian analysis of transport processes in vector fields. In order to increase the impact of our research, we integrate our results with an open-source visualization package to distribute our work to a large community of scientists and engineers. (official project page)