202 
efficient method to provide a framework for the evaluation of other areas. But depending on 
criteria, weights, resolution and modelling rules, various spatial and statistical distribution of 
diversity indices result (for a detailed description see Blaschke 1995). Neither hard- and 
software nor GIS-technology are limiting factors for a widespread operational use, but the lack 
of (more) standardised procedures for modelling and evaluating like Habitat Suitability Indices 
(HSI) and their adaptation to finer scales. More and more of the underlying technical problems 
are solved. After almost three decades of development of GIS attention has moved from 
primitive algorithms and data structures to much more complex problems in various 
applications (Goodchild 1992). 
The range of environmental data concerning structural diversity is very wide and the modelling 
possibilities have expanded due to technological advances in GIS. This means, not only 
primary data sources such as field surveys and remotely sensed data can be used, but also 
"invisible“ or generated information dealing with functional ecological units (catchments, 
barriers, buffering zones, network structures). Using such "second-order“ information built on 
a higher level of abstraction for this purpose requires caution to the rules of abstraction as well 
as on the study area bias and the prediction bias (Verbylla & Chang 1994). 
Practically, combining second-order information, we are acting on a third level of abstraction, 
because every map is a model of reality and maps of primary surveys are actually models and 
abstractions themselves. What does this mean in the given context? It is tried to use not too 
complex data layers in order to enable the greatest possible visibility for the model. For a 
simple illustration of this approach only two original data layers are used here: vegetation and 
structure of vegetation. Each of them is surveyed independently. 
Map algebra: the horizontal perspective 
Analysis within Geographic Information Systems has been systematically described by Berry 
(1987) and Tomlin (1990), who defined it as "map algebra“ operations ranging from simple 
reclassifications, overlays, measurements and neighbourhoods to more sophisticated modelling 
functions. This concept is currently well adopted to some raster-based GIS modules, but not 
standardised. This means, that the "generic“ language of Berry and Tomlin is implemented and 
adopted in various ways. As a first step in this case study the vector based layers are rasterised 
because of availability of software solutions. This is a very simple but important procedure. 
Depending on the pixel size, information might be lost or not. In this case study fine reso 
lutions of 4 and 5 meters are chosen. Regarding the hardware possibilities today, this 
resolution seems to be processable up to a landscape level, even 10 7 to 10 9 pixel will result. 
The following steps are a simple descriptive analysis using two data layers. The term 
"modelling“ is not used, because the author feels, it is restricted to more statistical analysis of 
spatial data with extended demands on the data and the models. Nevertheless, the ability to 
handle and compare multiple features and their attributes on the same geographical location 
renders GIS as ideally suited to develop and simulate environmental scenarios and to answer 
questions like "what-if‘. 
In a second step neighbourhood relations are calculated. In various systems the required tools 
for measuring the variation of values within a defined moving window are available and named 
similar (focaldiversity or focal variety). But as this simple neighbourhood analysis without 
taking account to directions and qualities of borderlines shows, there are a lot of different ways 
how to do it.