.ades 
been 
lised 
1s of 
GIS. 
sets 
tion 
patial 
, and 
Based 
paper 
ig GIS 
Ferent 
1S Now 
GIS. 
ation 
| and 
both 
| the 
'ibute 
idling 
| the 
led it 
. with 
Isers' 
lality 
aphic 
hould 
for 
er a 
on of 
till) 
ILWIS 
l be 
the 
stem" 
ssion 
stem" 
these 
e the 
aphic 
ality 
should be communicated very effectively by 
cartographic means (i.e. maps). So, in a GIS, we 
can assume that maps will be much more suitable 
for conveying quality information to the users of 
a GIS’s products than any other graphic, textual 
or numerical means. The latter will have to be 
used only if there are no regional (spatial) 
variations in quality or if these variations are 
or cannot be specified, as is often the case for 
information on completeness or logical 
consistency. If quality information on positional 
and attribute accuracy (and possibly on lineage) 
is available for individual mapping units and 
there is some regional variation in it, this 
information can be represented cartographically. 
2.1 The design of cartographic symbols 
  
In a GIS quality information can be stored, 
analysed, processed and presented just like any 
other attribute information related to point, line 
and area features. As far as the cartographic 
representation is concerned, the same symbol 
design principles apply as for other aspects of 
information. A systematic approach to symbol 
design, based on and an improvement of the early 
works of Bertin [BERTIN, 1981 and 1983], has been 
presented by Bos [BOS, 1984], see Figure 1. In 
such a systematic approach one of the first steps 
is to determine the measurement level of the 
information to be portrayed, and this will be 
either qualitative (nominal), ordered (ordinal) or 
quantitative (interval or ratio). The meaning of 
the information, established by determining the 
measurement level, should be represented by the 
so-called visual variable having the corresponding 
perception property (see Figure 2). 
There are seven visual variables (position, form, 
orientation, colour, texture, value and size), see 
Figure 3, each with its own perception properties 
(see Figure 4). The perception property of a 
visual variable may be regarded as responsible for 
transferring a certain meaning or concept to map 
users’ minds whilst they are perceiving the 
cartographic symbols which are differentiated by 
that particular visual variable. 
In a map, the visual variable position is a 
special case in that it is always applied to the 
symbols occuring in the map. But also, usually, 
the visual variable position is combined with one 
or more of the other six visual variables to 
represent other aspects of information related to 
the point, line or area features being portrayed. 
Several aspects of information can be represented 
at the same time in the same map by reserving at 
least one visual variable for each individual 
aspect. For example, in a map showing factories by 
point symbols, their numbers of employees can be 
represented by the visual variable size 
(differently sized point symbols) and the nature 
of manufacturing industries with the visual 
variable form (differently shaped symbols). 
Following this approach a visual variable used to 
represent differences in one aspect of information 
cannot, anymore, be used to represent differences 
in another aspect of information related to the 
same features (e.g. the visual variable size 
-which was used to represent the numbers of 
employees - cannot be used anymore to show the 
annual production figures in the same symbols 
representing the factories.) 
609 
2.2 The design of symbols to represent quality 
information 
From the foregoing, quality information can thus 
be treated and cartographically represented in the 
same way as any other attribute information (also 
see CLAPHAM and BEARD, 1992). In this treatment 
first of all the measurement level of the 
information has to be established. Information on 
(regionally different) data sources (for example 
part of a data set’s lineage information) could be 
considered to be of a qualitative nature and thus 
should be represented by means of a visual 
variable with an associative perception property, 
for instance form or orientation (see Figure 5). 
Attribute accuracy information, for instance a 
probability percentage, could be of an ordered 
nature, to be represented by means of a visual 
variable with an ordered perception property, such 
as (lightness) value (see Figure 6). A difference 
in currency of the data is also not regarded as 
quantitative information (but ordered instead) as 
it is not realistic to think of data being "twice 
as recent", etc. Absolute quantitative quality 
information (measured on a ratio scale) should be 
represented by means of a visual variable with a 
quantitative perception property, size being the 
only one. Considering for example, absolute 
positional discrepancies measured in meters in 
various directions and represented by error 
ellipses (see Figure 7), the generation of such 
hard- or softcopy cartographic displays of quality 
information is only possible if the GIS provides 
the required presentation software (modules). This 
could be in the form of a cartographic design 
expert shell to assist the non-cartographic 
GIS-user [MULLER & WANG ZESHEN, 1990]. Such a 
cartographic expert system may guide the user 
through the systematic symbol design approach 
referred to above. 
2.3 Integrated or separate cartographic quality 
information displays? 
  
In the examples given in Figures 5, 6 and 7, the 
quality is the only aspect of information 
portrayed by the symbols. Such analytical, 
mono-thematic map displays may be easily generated 
from a GIS. Usually, however, the GIS user wants 
to relate and compare the quality data to the 
other geographic information to which they belong. 
That is, it seems to be more useful if the map 
with the quality data appears as a separate window 
next to the map showing the related geographical 
information or to represent the quality data with 
a separate visual variable in the same map 
together with the geographic data to which they 
belong and which are represented by another visual 
variable. In this context, it is possible to think 
in terms of "quality overlays", which may be 
"switched on and off" (or "toggled") on-screen by 
the GIS-user at vill. The ease of so doing is a 
great advantage of the computerised GIS 
environment compared to eavironments when only 
paper maps could be produced by means of slow 
manual drawing techniques. 
However, in a GIS environment one should not think 
merely of physical overlays, or separate layers or 
levels of data, but of a genuine combination of 
visual variables in one set of symbols. If, for 
instance, in a soil map different soil classes are 
represented by means of the application of the 
visual variable colour (i.e. hue) to the mapping 
units, it would be appropriate to represent the 
related ordered reliability information by means 
of application of the visual variable value (i.e. 
lightness (in some software packages referred to