| Analyzing time-evolving phenomena is an important task in various scientific fields. State-of-the-art scientific computing technologies allow accurate numerical modeling of multiple properties of such phenomena, producing time-varying multivariate volume (TVMV) data. Such data need to be analyzed in spatial, temporal, and variable domains. For instance, hurricane simulation data allow us to understand the spatial and temporal correlation between temperature, pressure, and wind speed when hurricanes move over land. Scientific visualization has proven very effective in gleaning insight into such data. This dissertation creates a unified solution to visualize and analyze TVMV data. A data packing technique enhances the overall system usability by lowering the data transfer cost between storage units. This solution elegantly adapts to the various data analysis cases, while taking into account the user's intent. A multivariate volume rendering technique enables scientists to analyze the spatial correlations between multiple variables in a single imagery. A tri-space user interface, which accommodates the rendering and packing techniques, allows the user to easily and freely specify features of interest. The interface reveals complex relations hidden in multi-dimensional data through tightly linked spatial, temporal, and variable views. Finally, an animation support, AniViz, enables scientists to create expressive visualization animations, while they are exploring the data. The key element used in AniViz is templates, which abstract each frequently used animation effect. An expressive animation is created by combining multiple templates. A visualization system based on these designs can dramatically enhances scientists' productivity, facilitating new scientific discoveries. |
