Analyzing and Visualizing Multivariate Volumetric Scalar Data and Their Uncertainties

Abstract

Data sets from the real world and most scientific simulations are known to be imperfect, often incorporating uncertainty information. Exploration and analysis of such variable data can lead to inaccurate or even incorrect results and inferences. As a powerful tool to communicate the data with users, an effective visualization system should present and inform users of the uncertainty information existing in the data. While some research has been conducted on visualizing uncertainty in spatio-temporal data and univariate data, little work has been reported on multivariate data. In addition, there are two main disadvantages in the existing uncertainty visualization methods for volumetric data. First, they rely heavily on the human perceptual system to recognize the uncertainty information, lacking the capability to depict them quantitatively. Second, they often present large amounts of diverse information in a single display, which may result in visual clutter and occlusion. In this paper, we present our hybrid framework that combines both information visualization techniques and scientific visualization techniques together to allow users to interactively specify features of interest, quantitatively explore and analyze the multivariate volumetric data and their uncertainties as well as localize features in the 3D object space. In comparison with those existing methods, we argue that our method not only allows users to quantitatively visualize the uncertainties within multivariate volumetric data, but also yields a clearer data presentation and facilitates a greater level of visual data analysis. We demonstrate the effectiveness of our framework by reporting a case study from the OpenGGCM (Open Geospace General Circulation Model) simulation in space science application domain.