Geoscience Reference
In-Depth Information
description of SciViz systems). The scenes themselves are often more subjective and more complex
to engineer, but they offer the tantalising possibility of combining many variables together into an
integrated whole. Such an approach is perhaps more in keeping with integrated theories of percep-
tion, as described by Treisman (1986b). Senay and Ignatius (1991, 1994) describe some different
mechanisms by which data may be combined within a scene while maintaining overall effective-
ness of the complete visualisation. Compositional approaches to visualisation recognise that some
of the symbols and surfaces used are capable of encoding many visual variables simultaneously,
for example, the Chernoff faces and landscape visualisations described earlier. Spare capacity can
be used to encode further data to the unassigned visual variables. The resulting scenes are usually
more complex (richer) and can appear rather confusing at first as the observer becomes oriented to
the visual encodings being used.
All visual paradigms have cognitive limitations, caused by the compromises used to display a
complex multivariate dataset in a limited space and in a manner that encourages discovery. If we
separate data into different layers, to avoid over-cluttering in any one layer, then we also may sepa-
rate out the components of interesting patterns - making them more difficult to observe. When addi-
tional layers or displays of data are required, then the user's focus of attention must
shift
between
these layers or displays in order to assess their interrelationships (to
see
pattern or structure). This
attention shifting is undesirable as it leads to a weakening of the overall stimulus at any given point
in the scene, since it is now divided among
n
layers.
5.3.3 a
niMation
and
the
u
Se
of
i
nteractorS
Animation techniques provide a powerful and visually effective means of studying the relationships
between objects or patterns of interest and their defining data (Keller and Keller, 1993). Movement
has been shown to have a high visual impact, and its detection in humans uses significantly differ-
ent neural pathways to the perception of
retinal
variables (see Section 5.4.1). Animation is therefore
highly complementary to techniques based around shape, colour and position.
The link between animated cartography and visualisation is described by Dorling (1992) and
Andrienko et al. (2009), and a comprehensive account of the different ways that temporal varia-
tion can encode geographical data is provided by MacEachren (1994). Another possibility, geared
towards visual data exploration, is to use animation to explore possible connections between data
artefacts and investigate relationships. Examples include the movement of surfaces, one through
another, the projection of
specific artefacts
between graphs or surfaces and the animation of sta-
tistical operations applied progressively to the data (see Section 5.6). To be useful for exploratory
analysis, these techniques must facilitate perception of the structural and positional relationships
between specific regions in the data.
5.4 PERCEPTUAL AND COGNITIVE ISSUES
Useful and effective visualisations seldom occur by chance. Just as with statistical and machine
learning approaches to analysis, we must understand the guiding principles, the methodology
and how these fit into and shape the process of science (in this case, we concentrate on discov-
ery as the major science activity). So, in order to use GeoViz effectively in research, we need to
understand
•
The effectiveness of different kinds of visual variables and their appropriateness for dif-
ferent tasks
•
The various steps in the process or workflow to produce a visualisation
•
The role(s) visualisation plays in the science/discovery process
Each of these themes is covered in the following subsection.
