Exploring Ensemble Visualization

Abstract

An ensemble is a collection of related datasets. Each dataset, or member, of an ensemble is normally large, multidimensional, and spatio-temporal. Ensembles are used extensively by scientists and mathematicians, for example, by executing a simulation repeatedly with slightly different input parameters and saving the results in an ensemble to see how parameter choices affect the simulation. To draw inferences from an ensemble, scientists need to compare data both within and between ensemble members. We propose two techniques to support ensemble exploration and comparison: a pairwise sequential animation method that visualizes locally neighboring members simultaneously, and a screen door tinting method that visualizes subsets of members using screen space subdivision. We demonstrate the capabilities of both techniques, first using synthetic data, then with simulation data of heavy ion collisions in high-energy physics. Results show that both techniques are capable of supporting meaningful comparisons of ensemble data.

Figure 1

Multidimensional visualization examples: (a) SDDS visualization of an MRS scan of a brain, red spheres represent cholin, orange represents creatin, yellow represents glutamine, green represents n-acetylaspartate, low levels of all four metabolites (small spheres) are seen within a grey tumor region; (b) attribute block visualization of Nevada vote totals for Clinton and Obama during the 2008 Democratic presidential primary election, showing counties with a near-even split of votes (near-constant color), and counties where one candidate was favored over the other (a checkerboard pattern)

Figure 2

Three simulated ensemble members whose volumes represent an apple, a banana, and a pear, combined into a single ensemble visualization: (a) ensemble visualization; (b) apple volume; (c) banana volume; (d) pear volume

Figure 3

The same three volumes as in Figure 2, but with visibility controlled using a sinusoid visibility function: (a) ensemble visualization with size visibility; (b) ensemble visualization with opacity visibility; (c) sinusoid visibility functions, ensemble in (a) shown at t = 0; (d) sequential visibility function, ensemble in (b) shown at $t=\frac{1}{3}$

Figure 4

The same three simulated ensemble members—apple, banana, and pear—as in Figure 2, with an apple reference member and differences between members visualized using a yellow tint for banana, an orange tint for pear, and grey for regions in the apple that do not overlap either comparison member

Figure 5

Two apple member volumes with different attribute value distributions: (a–d) visualized with size controlling visibility, color identifying a glyph's parent member, and size representing attribute values; (e–h) visualized with opacity controlling visibility, shape identifying a glyph's parent member, and color representing attribute values; (i) visualized using screen door tinting, differences between the two members are seen as red regions, and similarities as white regions

Figure 6

A pear and an apple member volumes containing common attribute values within their overlap: (a–d) visualized with size controlling visibility, color identifying a glyph's parent member, and size representing attribute values; (e–h) visualized with opacity controlling visibility, shape identifying a glyph's parent member, and color representing attribute values; (i) visualized using screen door tinting, similarities between the two members are seen as white regions, areas in the apple that do not overlap the pear are seen as grey regions

Figure 7

RHIC simulation ensemble with four members: (a–d) isosurface visualizations of the members’ surfaces; (e–h) visualization of temperatures in a slice through the center of each member, blue for cold to red for hot

Figure 8

Eight frames from the RHIC ensemble, size controls visibility, color identifies a glyph's parent member, and size represents temperature values

Figure 9

Eight frames from the RHIC ensemble, opacity controls visibility, shape identifies a glyph's parent member, and color represents temperature values

Figure 10

RHIC ensemble visualized with screen door tinting, the first member is the reference member, blue tint shows differences with the second member, green tint shows differences with the third member, and red tint shows differences with the fourth member