the 
nent 
] its 
ss to 
ying 
the 
OWS 
1ake 
The 
nple 
ities 
u gh 
d to 
odel 
text, 
The 
'oad 
and 
vork 
1s. 
een 
New 
n its 
OWS 
| the 
ures 
reat 
n in 
sult 
nore 
ting 
ntic 
989 
703. 
tial 
ium 
68. 
on : 
ohn 
ion. 
phic 
onal 
and, 
RESPONSIBILITY, UNCERTAINTY AND FITNESS-FOR-USE: 
IMPLICATIONS FOR GEO-INFORMATION THEORY. 
Allan J. Brimicombe. 
Centre of Land & Engineering Surveying, 
Hong Kong Polytechnic, 
Hong Kong. 
ABSTRACT: 
significant limitations for imprecise spatial data. A general model is proposed as a basis 
for developing workable Solutions. An example of how the model operates is given. 
Implications are that uncertainty measures should be embedded within GIS data rather than 
the use of global measures. Users must take responsibility for assessing fitness-for-use 
of their data for the particular context of an analysis. 
KEY WORDS: Data uncertainty, Propagation, Theoretical model. 
THE ISSUE OF UNCERTAINTY IN GIS 
Recognition of the Issue 
Geographic Information Systems (GIS), as a tool, is 
proving successful in apparently reducing practical 
problems of organizing and integrating spatial data 
and for carrying out complex spatial analyses. 
Initial concerns for developing and implementing 
the technology on the one hand and the pressing 
need to convert both spatial and non-spatial data 
from analog to digital formats on the other, has 
largely resulted in the need to address data and 
analytical integrity being overlooked. One suspects 
that in the euphoria of automated overlay analysis, 
earlier warnings such as those of MacDougall (1975) 
concerning the integrity of the manual overlay 
process were either forgotten or deemed no longer 
relevant. It may even be that operators and 
managers from non-spatial disciplines were unaware 
of this dimension to their work, other than the 
need to address blunders. 
As users have become more familiar with the 
technology and as GIS has been accorded a more 
sophisticated role such as through incorporation 
into decision support systems, the quality of 
results portrayed in the graphical output has been 
brought increasingly into question. Data quality in 
GIS began to be discussed in the literature in the 
early 1980’s (e.g. Chrisman, 1982; Mead, 1982; 
Blakemore, 1984: Newcomer & Szajgin, 1984; Vitek et 
al., 1984). However it was not until the late 
1980's that issue seems to have become of more 
widespread concern (17 papers in 1987, 2 papers in 
1988, 14 papers in 1989). A bibliography has been 
published (Veregin, 1989) and a book devoted to the 
Subject (Goodchild & Gopal, 1989), but although the 
issue is now regularly featured in conferences and 
texts, the number of papers would appear to be in 
decline. The problem has been identified and 
defined, but few solutions have emerged. Not 
surprisingly then, vendors do not seem to be taking 
much action on the issue. 
The. Nature of the Issue 
Uncertainty can be used as a global term to 
encompass any facet of the data, its manipulation 
or its presentation which may raise concern, doubt 
759 
or skepticism in the mind of the user as to the 
validity of the results or intended message. 
Theoretically this definition would also ‘include 
mishandling of the data through improper analysis, 
inappropriate or erroneous use of GIS functions, 
poor cartographic technique and so on. However, it 
is not the purpose of this paper to address that 
aspect of the issue. Mostly it is concerned with 
the inaccuracies, inexactness or inadequacies 
inherent in most spatial data sets and how these 
may propagate to possibly invalidate the 
informational output of the system. Lack of 
appropriate visualization techniques to portray 
data variability is also likely to exacerbate the 
issue (Beard et al., 1991). 
Figure 1 summarizes the possible sources of 
uncertainty. GIS is a veritable minefield in this 
regard (see for example Burrough, 1986; Walsh et 
al., 1987 and Maguire et al., 1991 for general 
discussions). The end effect results in two broad 
categories of uncertainty: the position of objects 
(in 2, 3 or 4 dimensions) and the characterization 
(description and/or additional dimensionality) of 
these objects (Table 1). If we also include here 
the location and correct assignation of any null 
Space, then the above two categories will include 
concerns over the completeness of the data. 
Notwithstanding Figure 1, the main cause for 
concern rests with the ability of  Giss to 
adequately record, manipulate and display the 
natural variation that exists in most spatial data. 
Hence  Burrough's statement that "many soil 
scientists and geographers know from field 
experience that carefully drawn boundaries and 
contour lines on maps are elegant 
misrepresentations of changes that are often 
gradual, vague or fuzzy" (Burrough, 1986). 
The. Effect 
Despite increased awareness on the part of users, 
canned demonstrations continue to overly impress 
(Congalton, 1990). Complex manipulations of data, 
seductive graphics, and implicit faith in the 
computer easily turns an old adage on its head - 
garbage in, gospel out! The author believes that 
this situation has developed due to the absence of 
an adequate theory or general model for handling 
uncertainty in GIS.