|Title of PhD thesis:|| ||Visual Data Mining and Visualization Design for Climate Research|
|Reviewers:|| ||Heidrun Schumann, University of Rostock, Germany|
| ||Helwig Hauser, University of Bergen, Norwegen|
| ||Holger Theisel, Otto-von-Guericke-University of Magdeburg, Germany|
The combination of visualization and automated mining methods – so-called visual data mining – is an important means of analyzing large data sets. However, there is a gap between the available methods in sophisticated software systems and their use in special domains. Based on the example of climate data this thesis shows how to close this gap by tightly coupling interactive visualization and automatic analysis methods as well as by providing a high degree of user interaction. The main focus of this work is on how visualization techniques can be enhanced and adapted to the heterogeneous, multi-variate data sets used in climate research. To this end, the possible fields of application of visualization techniques are systematically investigated and new techniques are proposed, especially for spatial and temporal data analysis and for comparative visualization. Furthermore, innovative techniques displaying results of cluster and principal component analysis are provided. Moreover, a general procedure to apply the proposed visual data mining techniques to the whole process of modeling, simulation and model evaluation is designed and illustrated. Finally, to simplify the application of the proposed techniques, a selection and parameterization mechanism for visualization techniques is designed. The important factors influencing this mechanism (metadata, goals of analysis) are specified, and methods for their collection and management offering a high degree of user support are proposed.