The point is that consistent theories need to be developed 
that describe the structure and character of geo information. 
That can deliver models for its classification, qualification, 
time-dependency, generalisation and selection aggregation 
etc. 
Without this it will be difficult to proceed responsibility in 
achieving the three strategies proposed in Fig.2 which are 
today espoused by many organisations. 
SOME STRUCTURAL AND SEMANTIC ASPECTS OF 
GEO-INFORMATION. 
The previous discussion explained why from a management 
perspective a geo-information theory is needed. If we 
interprete the requirements that have been formulated this 
theory should deal a.o. with the structural (syntactic) and 
semantic aspects of geo-information, with the implementation 
in the logical datamodels developed in computer-science and 
the theory should deal with the uncertainty aspects of geo- 
information, see [Molenaar 1991a]. The further discussions 
in this paper will emphasise the syntactic aspects and their 
relationship to semantic modelling in GIS. That will help us 
to understand why data definition should always be 
embedded in a particular users context. In many cases it will 
be difficult to transfer data from one context to another 
without data transformations, which will then be called 
context transformations [Molenaar 1991a and 1991b]. The 
topics of logical data modelling and uncertainty will be 
referred to only shortly. 
In GIS there are two important methods for terrain 
description. The first method is to link values of some 
thematic attribute to positions. E.g. terrain heights are given 
either in randomly distributed points or in a regular grid. 
Other examples are the observations of ground water depth 
or soil characteristics etc. 
The other method is to identify terrain objects which have 
thematic and geometric characteristics. A representation in 
an information system will consist of an object identifier 
(e.g. a name, or a number) which is linked to a set of 
thematic data and to a set of geometric data as in fig. 3. 
object 
identifier 
Fig. 3. Information structure for representation of terrain 
objects. 
    
   
This basic structure has been applied in many information 
systems for cadastre, urban management, utilities and many 
other applications.In most cases the thematic aspects play a 
dominant role in the object definitions. 
That is why a geo-information theory should emphasise the 
thematic context of the object definitions and provide a 
structural framework for dealing with these thematic aspects. 
In this respect there is not much difference with information 
models for administrative databases. An important specific 
aspect of geo-information theory is the link between the 
750 
thematic and the geometric object descriptions. A more 
detailed description of object hierarchies will be helpful to 
understand the problems met in spatial data modelling, the 
concepts presented here have been discussed in more detail 
in [Molenaar 1991b]. 
TERRAIN OBJECT CLASSES 
In most applications the terrain objects will be grouped in 
several distinct classes, according to their thematic aspects, a 
list of attributes is connected to each class. Terrain objects 
belonging to one class inherit the attribute structure from the 
class. This means that each object of the class has a list 
containing a value for each attribute of the class attribute 
list 
[ class | | attribute list | 
  
  
E object sd attribute values = 
Fig. 4 Class structure of objects 
superclass 
sc.attr.j , sc.attr.j 
sc.attr.j, values 
  
  
   
  
  
   
sc.attr.j, values 
c.attr. ‚values 
Fig. 5 Class and superclass structure of objects. 
      
When two or more classes do have common attributes, then 
a superclass can be defined with a list containing these 
common attributes, these will then be called super class 
attributes. The classes at the next lower level will be 
subordinated to these superclasses. To each class a list of 
class attributes will be linked, in general these lists will be 
different for different classes. The terrain objects are then 
subordinated to these classes. With these observations we 
find the class hierarchical structure of fig. 5. 
It is possible to add more hierarchical levels to the structure 
of fig. 5. Each level inherits the attribute structures of the 
next higher level and propagates it possibly with an 
extension to the next lower level. At the lowest level in the 
hierarchy are the terrain objects, at this level the attribute 
structure is not extended anymore, here the inherited 
attributes are evaluated. 
If the classes are disjoint then the terrain objects will get 
their attribute structure only through one inheritance line in 
the hierarchy, i.e. they have a unique thematic description. 
The 
class 
obje 
class 
certa 
not, 
them 
one 
elem 
AGC 
The 
exist 
aggre 
class 
An a 
be bi 
objec 
and 
facto: 
From 
fig. 6 
house 
Fig. € 
The c 
towns 
build 
res 
dis 
Fig. 7 
The ag 
sense 
objects 
upwarc 
values 
The f: 
compo: 
be agg 
numbei 
total nu