Figure 4 - Perception properties of the visual 
variables (source: BOS, 1984, p.23) 
  
  
  
  
  
Position | Form | Orient, | Colour | Texture | Value | size 
Associative * * * * 0 - - 
Selective - - 0 ++ + + + 
Ordered - - - - 0 ++ + 
Quantitative - - - - - - ++ 
  
  
  
  
RELIABILITY DIAGRAM 
E 
/, " Figure 5 - Typical 
NAM example of a reliability 
diagram which may be 
found on a topographic 
map. The visual variable 
orientation has been 
used to show the 
various data sources 
: = (Fiji 1:50 000, DOS, 
AIR PHOTOGRAPHY 1964). 
a JULY 1954 
b JUNE 1952 
OTHER MATERIAL 
c ADMIRALTY CHART 2691 
d ADMIRALTY CHART 905 
    
Figure 6 - The representation of attribute 
accuracy by means of the visual 
variable value in a choropleth map 
ACCURACY DUERLAY 
  
Soil Classification Accuracy Overlay 
Figure 7 - Error ellipses used to show positional 
inaccuracy (the larger the ellipse, the 
greater the inaccuracy) (Source 
RICHARDUS, 1974, p429) 
  
612 
It was in this context that in 1991 an internally 
funded ITC project was established for the 
creation of an "Uncertainty Subsystem" for ILWIS. 
3.1 The ILVIS "Uncertainty Subsystem" 
The general concept of the "Uncertainty Subsystem" 
is that for any information generation operation 
in ILWIS there will be a near parallel production 
of information describing the quality of that 
generated information at the GIS user's request, 
as shown in Figure 8. This will require a means of 
storing control points and their quality 
statistics, positional and attribute data quality 
for all database objects, for propagating error 
through the selected GIS processing models, and 
finally for displaying the quality of the 
generated information in an appropriate manner as 
discussed in section 2 of this paper. This paper 
deals mainly with the last of these (displaying 
the quality of the generated information), but 
other students and colleagues are working on 
different aspects of the "Uncertainty Subsystem". 
4. A LAND REALLOCATION PROJECT TO EXAMINE 
THE DISPLAY OF INFORMATION QUALITY IN GIS 
To test our approach to the display of quality 
information, data and processing models from an 
ongoing land reallocation project located near our 
institute were examined. Land reallocation is 
performed when agricultural land holdings in an 
area have become highly partitioned as a result of 
inheritance; the holdings are consolidated, with 
the owner being guaranteed a holding of the same 
value. The determination of a holding’s value 
involves several valuation submodels - one of 
which determines the holding's grazing 
suitability, land parcel by land parcel. This 
grazing suitability model, treated as a GIS 
processing model in which the quality of the input 
data and generated information is to be displayed, 
is considered here. 
4.1 Grazing Suitability Model 
  
As a processing model the grazing suitability 
model is Boolean or logical [DRUMMOND and RAMLAL, 
1992] and uses three sets of information [RAMLAL, 
1991] to provide Grazing Suitability (3 classes): 
1. soil drainage status (5 classes); 
2. soil moisture supply capacity (5 classes); and 
3. topsoil bearing capacity (3 classes), 
The model was checked [MARSMAN and DE GRUIJTER, 
1986] and found to provide correct grazing 
suitability predictions in 95% of cases. The model 
is shown in tabular form: 
  
Drainage 
Status 1 2 3 4 5 
Bearing 
Capacity | 12 | 12 [12 } 12 3 | 2 3 
Moisture 
Supply 
Capacity 
1 1:1 1-1 11 123 l2 3 
2 11 11 11 123 |; 23 
31221] 221} 227} 223133 
4#{ 33 | 33133] 333) 33 
51, 33 |. 33}; 33] 333133 
  
  
  
  
  
  
  
  
and can be explained by the following examples: 
  
3 a EUER pe ni An A