Geography Reference
In-Depth Information
of dynamic processes may exist only at specific temporal scales. The spherical surface of the
Earth cannot be accurately represented on a flat piece of paper or on a computer screen. Other
activities in the cartographic abstraction process lead to intentional and unintentional bias,
and models in general, whether visual or computational, and no matter how sophisticated,
can only approximate reality. Therefore, a fundamental and persistent discrepancy exists
between geographic data and the 'reality' they are meant to represent.
In many applications and for many audiences, providing information enabling the un-
derstanding of this gap between reality and representation is critical. This chapter highlights
research on communicating uncertainty and spatial data quality to GIS users, specifically fo-
cusing on uncertainty visualization. Providing information about uncertainty is considered
by many users to be either irrelevant or detrimental for successful data communication and
insight generation. Herein, we argue that uncertainty in geographic data should be made
usable
through innovative visualization techniques. In many cases, representing uncertainty
has a positive effect in the visualization process. We provide a framework for increasing the
usability of information about data uncertainty for decision-making.
14.2 The complexity of uncertainty
Uncertainty is, of course, a broad interdisciplinary subject, the infinite nuances of which
cannot be explored in detail here (MacEachren
et al.
, 2005; Devillers and Jeansoulin, 2006;
Goovaerts, 2006). Any discussion of uncertainty, however, should begin with a closer look
at the many different terms in the literature that relate to, and are sometimes used inter-
changeably with, uncertainty.
Uncertainty broadly refers to incompleteness in knowledge, unknown or unknowable in-
formation about the discrepancy between an actual value and its representation in language,
mathematics, databases or other forms of expression. Uncertainty can take many forms. The
form of uncertainty that is relevant to a given problem or situation may change depending
on the user of the data and the purpose of the data's use.
There are only a few forms of uncertainty that have been the focus of uncertainty visualiza-
tion research. MacEachren, Brewer and Pickle (1998) represented the reliability (accuracy)
of death rates for a map of mortality data. Cliburn
et al
. (2002) visualized the variability
(error) in results of a water balance model. In their evaluation of fuzzy classifications, Blenk-
insopp
et al
. (2000) were faced with uncertainty in the form of vagueness and ambiguity.
Clearly by these examples, uncertainty is expressed in many different forms, many of which
are summarized in Table 1.
The term
data quality
refers to the degree to which a dataset fulfils or conforms to specific
requirements. It includes a range of attributes such as completeness, consistency and lineage
(MacEachren, 1992; Evans, 1997; Veregin, 2005; Devillers and Jeansoulin, 2006). Data quality
implies both an objective and subjective evaluation of the data. For example, for a particular
study, a researcher might use a dataset that contains relevant information, but that was
created for another study two years earlier. This dataset is comprehensive, with a large
number of samples; the data source and producer are known, and all the data processing
that was completed is well documented. In general, the dataset would be considered to have
an objectively high level of quality. However, the time difference would render the same
dataset as of subjectively low quality. Thus the criteria used to define quality may be very
different for those producing and those using the data.
Search WWH ::
Custom Search