Geography Reference
In-Depth Information
include scores or even hundreds of variables. In the innovative 3D seafloor visualizations
created by Schmidt
et al.
(2004), for example, only five variables were simultaneously dis-
played, and in the iconographic visualizations created by Gee
et al
. (1998) and Healey (1998),
only a similar number of data variables were encoded by means of icon geometry and colour.
Claims (e.g. by Wright, 1997) that certain 3D data visualization techniques are capable of
displaying hundreds of variables are therefore little more than marketing hype.
Given this problem, it becomes necessary to think the unthinkable, and to consider setting
aside the representational spatial framework in which much of our data is gathered. Indeed,
relevant 2D and 3D visualization techniques that are stripped of a geographical frame of
reference may provide powerful insights into complex data in ways that cannot be provided
by conventional spatial visualizations.
Three broad strategies have been adopted. The first is to retain the geographical seman-
tics of the real world in the 3D display space, but to switch between visual representations
of selected (small) subsets of variables. Most of the discussion in this chapter so far has
focused on this strategy. A second strategy is to abandon geographical semantics in the
3D display space, and to use visual data mining or visual analytics techniques (de Olivera
and Levkowitz, 2003) to display information for multiple variables in 2D displays. Such
visualizations include parallel coordinates (Inselberg and Dimsdale, 1989), pixel-based dis-
plays (Keim, 2000), 'dimensional anchors' (Hoffman, Grinstein and Pinkney, 2000), and
an increasingly large number of other multivariate data visualization techniques (Wong
and Bergeron, 1997). Where the number of variables or dimensions is extremely large, 2D
projections may be made from multidimensional data into 2D spaces, or components may
be derived from the original data by various data reduction techniques, and displayed in
2D display space (e.g., Yang
et al.
, 2004, 2007). By coupling data reduction techniques to
data visualization, visualization continues its tradition of helping the analyst to steer the
data mining in potentially fruitful directions (Keim and Kriegel, 1994; Keim, 2002), but
operating within statistical space rather than geographical space.
A third strategy is to combine the spatial and non-spatial approaches in a multiple linked
views environment. For example, parallel coordinates displays have been incorporated into
several software systems designed for the analysis of spatial data, along with 2D maps
and other forms of tabular and graphical display (e.g. Stolte, Tang and Hanrahan, 2002;
Andrienko and Andrienko, 2003; Guo, 2003; Guo, Chen and MacEachren, 2006; Marsh,
Dykes and Attilakou, 2006). (This multiple linked views approach has also been developed
in other research fields, e.g. Gresh
et al.
, 2000.) A notable feature of these hybrid systems is
that 3D spatial representations are notable by their absence; almost without exception, they
only incorporate 2D maps and 2D statistical graphics.
This suggests two interesting possibilities. The first is that 2D display techniques may
be more effective than 3D techniques for routine data interpretation purposes, even for
spatial data. The second is that analysts looking for comprehensive data visualization soft-
ware should not expect to find them solely among the offerings of current GIS and desktop
mapping software vendors, whose focus is mainly on spatial data management and analysis.
Most of the innovative data visualization software of the past two decades has emerged
from the non-spatial sciences, and especially the information visualization community. By
interfacing effective modular software from these sources to standard 2D mapping tools
developed within geography, it may be possible to acquire the most effective toolkit for
both spatial and non-spatial data analysis. This has been the motivation for the author's
own experimental 3D data visualization software, which imports data from MapInfo and
Search WWH ::
Custom Search