Glyphs for Visualizing Uncertainty in Vector Fields

 

 

The authors present types of inherent uncertainty that are often ignored in visualization of environmental data. They consider 2 techniques for visualizing uncertainty along with the reported data: the overloading approach (color, transparency, etc.)  and a new nonoverloading approach that they term verity visualization. This visualization technique involves the representation of data via glyphs that display both the data and its associated uncertainty.

 

Types of Uncertainty

Source errors : happen during data acquisition (ex – sensor array)

Process errors : happen during data transformation (ex – quantization, interpolation)

Use errors : happen during visualization (ex – direct volume rendering)

 

Techniques to Visualize data and uncertainty

Glyphs: represent data through visual properties such as color, shape, size, orientation

Graphic Attributes: color, transparency, line width

Overlaying : sandwich layers

Side by side comparison data with uncertainty information

Pseudocoloring

Animation

Blurring

Sonification

Verity Visualization

 

 

Visualizing data with uncertainty

Overloading : mapping uncertainties to additional visualization parameters such as color

                      user may be able to confuse the overloaded graphics specifier with the data that is mapped to it

 

Verity Visualization : uncertainty information is integrated with the data into a visualization graphic, called a glyph

                                    holistic approach -uncertainty is not treated as an add-on

 

 

Example – Wind Vectors and Ocean Currents in Environmental Data

Need to display magnitude, bearing as well as both these uncertainties

May be displayed as an overloaded  graphic, utilizing psudocoloring or transparency

 

 

Design of Glyphs for Vector Fields

Must address dense displays

Must represent magnitude and bearing

Must represent data and uncertainty in different coordinate frames

 

 

 

Expressing Uncertainty in Bearing

Loop around arrow head

Pie swept area

Ellipsoidal target area

 

Expressing Uncertainty in Magnitude

Loop around arrow head

Encoded in the body of the glyph

Range of pie swept area

Multiple arrow heads

 

 

 

 

 

 

 

The Design the Authors Present

Vector magnitude represented by either glyph area or glyph length. See comparison.

Magnitude Uncertainty represented by extra arrowheads depicting the range of possible magnitudes

Vector bearing represented by angular rotation of arrow

Bearing uncertainty represented by angle swept by the glyph

 

 

 

 

In use in a map:

 

 

 

 

 

 

 

Compare to arrow glyphs (uncertainty not depicted):

 

 

 

Compare to arrow glyphs that are pseudocolored to depict uncertainty :

Color depicts magnitude uncertainty
Mapping from black to white for bearing uncertainty

Assessing the Usefulness of the New Glyph Design

Quantitative measures:  larger amounts of data with economy of expression

Information per unit of space

Information per unit of ink

Number of multifunctioning graphical elements

Able to express high data density

 

Qualitative measures: can users accurately decode information from the glyphs?

Decoding tests were administered to determine interpretation accuracy after subjects were trained in use of new glyphs

Subject were asked to identify magnitude, bearing, uncertainty magnitude and/or uncertainty bearing

Subjects were naval officers ursuing masters degrees in Meteorology

Response times were not included in the test

Results of comparison of standard arrows to new glyphs:

   mean magnitude decoding error (scale of 1 to 10)

      arrow : 0.989 (about 10%)

      new glyph : 1.46 (about 15%)

   angular errors were statistically equivalent

   errors in uncertainty information (new glyph only) are comparable to those for bearing and magnitude