Graph-based techniques for visual analytics of scientific data sets

Imagine a typical workflow for a climate scientist exploring a 3D volumetric data set produced from a simulation. The data set is time-varying and multivariate, containing scalar and vector quantities. The scientist first plays back the entire time sequence of the precipitation scalar field and spots some features of interest. She wants to extract those features and track them over time. However, the features are difficult to select or pick directly in 3D. Visually tracking multiple features over time is very difficult as she has to rely on her memory to tie together their connections. Furthermore, she also integrates a set of flow lines from the wind velocity vector field and needs to examine their relationships. Visual exploration of these flow lines poses quite a challenge due to the lack of capability to observe them and their spatial relationships in an occlusion-free and controllable fashion. Moving forward, she even hopes to investigate these scalar and vector quantities simultaneously, for instance, identifying regions of high pressure and low precipitation and studying flow lines passing through these regions only, or exploring flow lines that connect attracting spiral saddles and repelling spiral saddles. This becomes increasingly impossible without a new solution for data abstraction and relationship exploration. As the size and complexity of scientific data continue to grow, these challenges will only become more severe. © 2018 IEEE.