Exploring the Invisible : FINDING STRUCTURE IN SCIENTIFIC DATA

Sammanfattning: In this thesis, I present contributions towards the aim of understanding flow-related scientific data sets by communicating relation, properties, and structure. The individual papers are contributions to three different areas. First, real-world visualization challenges with domain specific tasks. The individual applications are ranging from analyzing transport behavior in a centrifugal pump, to visualization of the impact of volcano eruptions and their atmospheric aftermath, and studying circulation dynamics and eddy movements in the ocean currents of the Red Sea. Although the three individual publications target different domains, they share common demands. Furthermore, the experience shows that combining and adapting different visualization techniques to support experts is essential for these scenarios. Second, technical visualization research with a strong focus on geometry-based, interactive, and explorative techniques. In this area a new type of particle system and a novel geometry-based flow visualization technique based on evolutionary algorithms are presented. With both approaches, areas of interest can be highlighted in a semi-automatic fashion by facilitating user-defined importance measures. Lastly, a method for decoupling definition and tracking of features. Here, the development of a fast but flexible method for defining and tracking cyclonic features in pressure fields using a solid and robust mathematical basis is presented. The initial theoretical work is discussed in context of its practical applications by pointing to relevant follow-up publications.The experience from real-world visualization tasks shows that understanding and gaining insight of scientific data with the help of visualization is an interactive, explorative, and non-linear process. Here, different methods must be combined and adapted such that they complement each other. Through this practice, relation, properties, and structure can be revealed, and a mental model can be created.From the real-world visualization challenges and the contributions in research, demands on techniques and their embedding in a visualization toolkit can be derived. Here, the ideal software is flexible, adaptable, and allows for interactive exploration. Furthermore, the process is benefiting from a semi-automatic approach guiding the domain expert during analysis. These aspects are used as guidelines for the implementation and development work associated with the contributions of this thesis and are presented in a dedicated Chapter.