Geoscience Reference
In-Depth Information
(even obvious) in the display. These patterns must then be recognised and recorded (or further
analysed) by the researcher. Involving the researcher intimately in the process of discovery is
definitely not a new idea(!) but here the traditional responsibilities and roles are shifted away
from the system providing the results of analysis in statistical or numerical form, along with
confidence intervals or error bars. Thus, we can see that visualisation is by its very nature more
subjective -
what is seen depends on how one looks and on who is looking
. And before - dear
reader - you dismiss this as a weakness, first, try on the idea that it might actually be a strength.
If you are a geographical researcher, you can probably bring a great deal of expertise to bear on
data exploration that would be difficult to express or operationalise statistically. Visualisation
does not force you to try.
5.2 GEOVISUALISATION FOR DISCOVERY
In this section, the need for better discovery methods is highlighted, and the history and genesis of
GeoViz is briefly outlined, followed by the introduction and definition of some of the key terms that
recur throughout the later sections.
Strong motivation is provided for the development of specifically geo-capable visualisation
systems by the increasing amounts of data becoming available for many types of application
right across geography and the wider geosciences. New families of geo-physical and thematic
instruments, demographic surveys, crowdsourced information and disease registries are creating
datasets with large numbers of data dimensions and with high thematic, spatial and temporal res-
olution, from which exploration into trend and change detection is needed. Furthermore, improve-
ments in data interoperability, ontology alignment and linked geographical data (Janowicz et al.,
2012) generate opportunities to combine existing datasets in new ways, again creating more data
complexity (Spira and Goldes, 2007). For example, a disease incidence study using census and
medical data
might
require 30 or more variables including case data, administrative boundaries
and demographic surfaces (e.g. Guo et al., 2003), whereas a land cover change mapping exercise
might use multichannel multi-date, remote sensing imagery supplemented with ground surveys,
elevation surfaces and aerial photographs. Such augmented datasets give rise to new challenges
relating to data discovery, some of which can be addressed by using the enhanced functionality
and finer control that visualisation environments offer to enable the discovery of useful patterns
and trends.
Many traditional forms of spatial analysis can become prohibitively complex or unreliable when
dealing with such large and diverse multivariate datasets. To a lesser extent, the same can be said
of machine learning techniques; they can become difficult to configure and slow to converge as the
complexity of the analysis task increases by adding in data dimensions. GeoViz offers the capabil-
ity to view many multiple, related data themes concurrently, without recourse to statistical sum-
marisation or data reduction - both of which may mask or even remove the very trends in the data
that might be of greatest interest to us. To quickly illustrate the point, Figure 5.2 shows more tradi-
tional statistical plots, mixed with a time series and a map, to view a complex dataset that contains
demographics, ethnicity and cancer incidence data (for five cancer types) over a 5-year period. The
combination of displays allows spatial, temporal and statistical trends and patterns to be tracked
concurrently.
GeoViz has emerged over the last 20 years as an active and coherent sub-discipline within
GIScience that - as its name suggests - focuses on the portrayal of geographical information in
visual form. It essentially tracks the more mainstream information visualisation (InfoViz) and
scientific visualisation (SciViz) disciplines, though with the specific inclusion of maps and dis-
plays of spatial data. Both InfoViz and SciViz draw from a rich and diverse research literature
including computer graphics, statistics and cartography, but mixed with a healthy dose of cogni-
tive science and psychometric research to help us understand and utilise the human visual system
to good effect.
