Approaches to Uncertainty Visualization

Alex Pang and Craig Wittenbrink and Suresh Lodha. “Approaches to Uncertainty Visualization“. In The Visual Computer, vol. 13, no. 8, pp 370–390, 1997.

This paper surveys techniques for presenting data together with uncertainty introduces as the data are derived, transformed, interpolated, and rendered. These uncertainty visualization techniques present data in such a manner that users are made aware of the locations and degree of uncertainty in their data so as to make more informed analyses and decision.

This research lies in the lack of methods that present uncertainty and data. The common underlying problem is visually mapping data and uncertainty together into a holistic view. The ultimate goal of uncertainty visualization is to provide users with visualizations that incorporate and reflect uncertainty information to aid in data analysis and decisions making. The authors define uncertainty to include statistical variation or spread, error and differences, minimum-mamixum range values, noise, or missing data. In this paper, 3 types of uncertainty are considered: statistical, error and range.

The sources of uncertainty, errors and ranges within data include:

• Uncertainty in acquisition: With instruments, there is an experimental variability whether the measurements are taken by a machine or by a scientist. The more times the measurements is taken, the more confident the measurement. But there will be a statistical variation in these measurements.
• Uncertainty in transformation: Data are rescaled, resampled, quantized prior or as part of the visualization stage. These transformations alter the data from its original form, and have the potential of introducing some uncertainty.
• Uncertainty in visualization: The rendering process introduces uncertainty arising form the data collecting process, algorithmic errors, and computational accuracy and precision.

The authors create a classification of uncertainty visualization techniques with five characteristics:

1. Value of datum and its associated value uncertainty (scalar, vector, tensor, multivariate)
2. Location of datum and its associated positional uncertainty (0D, 1D, 2D, 3D, time)
3. Extent of datum location and value (discrete or continuous)
4. Visualization extent (discrete or continuous)
5. Axes mapping defines visualization mapping (experimental or abstract)

The authors developed a variety of new uncertainty visualization methods. They are organized into a table showing general approach versus applications domain.

• Add glyphs: a glyph is a geometrically plotted specifier that encodes data values
• Add geometry: While glyphs do add geometry, they are placed at discrete locations. Adding geometry is used to denote a more continuous representation of data. Techniques include contour lines, isosurfaces, streamlines, and swept surfaces and volumes.
• Modify geometry: Geometry may be translated, scaled, rotated, or generally warped or distorted. They may also be displace, subdivided or refined.
• Modify attributes: uncertainty can be visualized by modifying attributes of geometrey in the rendered scene.
• Animation: Application to most applications, including comparison of animation data and techniques. Uncertainty information can be visualized by mapping them to animation parameters such as: speed or duration, motion blur, range or extend of motion.
• Sonification: Mapping uncertainty to sound.
• Psycho-visual approaches: stereo-pairs and subliminal messages

From the author’s exploration of uncertainty visualization techniques, they have found that continuous visualization extents are more challenging than discrete visualization techniques. This is based on a basic methodology that uses visual tests where users examine visualizations and decode the information within the graphics. The amount of errors between the user interpretation and the encoding is statistically evaluated to determine if the visualization is effective.

Similar papers include:
Visualizing Geospatial Information Uncertainty: What We Know and What We Need to Know and Visualizing Uncertainty in Geo-Spatial Data.
Relation to my thesis: the pipeline of the sources of uncertainty has similarities with my current categorization. The methodology to evaluate infoviz is based on visual tests in during which the user interpretation is evaluated. Interesting reference is the report which identifies four ways of expressing uncertainty:

Barry N Taylor and Chris E Kuyatt. Guidelines for evaluating and expressing the uncertainty of NIST measurement results. Technical report National Institute of Standards and Technology Technical Note. Gaithersburg MD January