Knowledge 
— |nformation ^: 
   
——$ 
M 
N 
Data Knowledge > 
  
   
eBpe|Wouy 
Knowledge 
  
  
PO(MOLS 
< | 
[TUA 
  
Information 
  
Knowledge 
P 
a e 
oe 
NSS : 
Con 
jie 
Knowledge 
  
Figure 1 : Relation between data, information and knowledge. 
e graphic transforms : information describing the graphical 
representation of the objects under a certain theme. These 
are symbol types, line styles, fill modes etc. 
The data in a GIS are represented in the following logical data 
models (R. Bill and D. Fritsch, 1991). Hierarchical or CODA- 
SYL models are often used to store and maintain the geometry 
and topology in a GIS. Relational data models, the standard in 
commercial databases, are more and more used to handle the at- 
tribute data in a GIS. In some cases also the geometry and topol- 
ogy is stored and treated in relational form. Object-oriented data 
models is one of the most often used keywords in GIS, which is in 
most cases misused if one follows the definition of computer sci- 
entists. Object-oriented data models will be of importance in the 
future for GIS. Today there are only a limited number of products 
available on the market which are partially object-oriented. The 
term 'object-oriented' should, therefore, be illustrated in more 
detail following the computer-sciences (for further references see 
R. Bill and D. Fritsch (1992)). They are separating three types of 
object-oriented data models : the structural, the behavioural and 
the fully object-oriented systems mainly differing with respect to 
flexibility. Structural object-oriented systems are able to handle 
complex objects as an atom in a data base having predefined 
operators for these data types. Behavioural object-oriented sys- 
tems support the definition of application specific data types and 
operators, so called methods. Fully object-oriented systems are 
combining the properties of both types of data models. An ob- 
ject consists of data and methods. The object itself is a small 
world in its own (called encapsulation). The object structure and 
behaviour may only be manipulated with the methods belonging 
to the object. A message is send via a channel to manipulate 
the encapsulated object which is illustrated in figure 2. Object- 
oriented systems are having close relations to some concepts of 
knowledge representation, which we will discuss afterwards. 
2 KNOWLEDGE INTEGRATION IN 
GIS 
2.1 Types of knowledge 
Various types of knowledge are illustrated in figure 3, which 
should not be seen as disjunct sets but as correlative forms of 
knowledge. In addition the figure tries to illustrate what types 
736 
  
of knowledge are implicitely or explicitely describable in a GIS. 
In particular the implicit form of knowledge is written down in 
the data model and the procedural flow of the GIS. GIS related 
examples for the various form of knowledge are given in detail in 
R. Bill (1991). Beside the contradictionary knowledge all other 
forms of knowledge could (and should) be integrated in a GIS. 
Today the assumption of completeness and exactness in the GIS- 
data model and the GIS-procedures are neglecting all types of 
uncertain and incomplete knowledge. 
2.2 Architecture of knowledge-based systems 
In a knowledge-based system the knowledge base is the major 
part of the system; there the rules of type ’If - then’ are stored. In 
GIS-products one can find knowledge about rule-based relations 
and prototypical knowledge integrated in a procedural manner. 
The integration of fuzzy knowledge would be usefull for certain 
applications. Figure 4 shows the design of a knowledge-based sys- 
tem compared to a conventional program system: the last may 
stand for the typical GIS-product. One can easily see the differ- 
ence lying in the flexibility of the problem solving strategy and 
in the domain specific knowledge which are integrated parts of a 
knowledge-based system. This would make it much more flexible 
to design an application-specific GIS. But the problem today is 
that the architecture of GIS-products is not very flexible. GIS to- 
day completely belong to the group of conventional programming 
systems. T'his leads to the current situation of knowledge integra- 
tion in GIS, where the knowledge part is completely done outside 
the GIS. The GIS is seen as a data capture and maintainance tool 
with the ability of visualizing results. To make use of the devel- 
opment of knowledge-based systems the procedural design of a 
GIS needs to be replaced with the inference-mechanism of an 
expert system, heuristical problem solving approaches instead of 
fix-coded procedural ones are to be integrated. 
2.3 Knowledge acquisition 
Usually in an application domain, which intends to make use of 
GIS technology, the whole expertise i.e. the knowledge about the 
application is with the application specialists. Thus, the carrier 
of the knowledge is not the GIS developper and, in some cases 
also, not the GIS user. The problem is to transfer the knowledge 
of the application specialists to the one who is responsible for