Development of methods of regional classification

The major task of this work package was to extend the existing methods or to develop new methods for the automatic numerical and spatial classification, which were aimed at the specific needs of the Alps and their communes. The result was that the classical classification algorithms could be used in a first step to classify the communes of the Alps into about 10 different types. Thus, the extension of classical classification algorithms for large data sets was necessary. The communes were classified based on their structural characteristics. In a second step the spatial consideration and analysis of the communes of the different types should follow. By interpreting these generated types, the individual communes could then be pooled to bigger spatial units regarding the several structural similarities and dissimilarities.

One of the major objectives for the next period was especially the interesting question, whether the so called probabilistic methods for classification or the methods of fuzzy clustering could produce "good" solutions within this special framework. Therefore, it should be analyzed, which approach of the named methods solved the present problem in the best way.

Classification and description of structural and environmental data

Based on analysis of relevant literature and discussions with experts on mountain agriculture, indicators which characterize agriculture and its economic and ecological background for each alpine commune were selected. These indicators were divided into socio-economic indicators (working persons, residential population, population shifts), describing the general economic surrounding of agriculture, natural resources indicators (altitude, climate, exposition, land use, slope, geology) and agricultural indicators describing farm structure (labour force, livestock, type of farming). 77 indicators were selected all of which were defined uniformly and were available for each alpine commune. The indicators were mainly taken from the national surveys on the structure of agricultural holdings and other national statistics on the community level.

The natural data set was more difficult to acquire since investigations dealing with the entire alpine region were rare. Based on a Digital Elevation Model spatial derivations, as there were exposition and slope, were yielded. In order to obtain other parameters, such as climatic conditions, analogue maps were digitized and analyzed. Digital information on land use was purchased from the European Satellite Data Centre which provides data captured by the CORINE program of the European Union. The elaborated set of indicators gives an overall description of mountain agriculture itself, the natural conditions for farming and the socio-economic surroundings.

Assessment of both quality and quantity of this data in view of the mutual effects

Not all structural data could be used immediately for the cluster analysis. The main part of the data needed to be linked to and calculated with each other to provide the highest information content for the use in the cluster analysis. In discussions among all partners algorithm were established in how to connect that data that was used for the cluster analysis. For a better overview indicators were combined in homogeneous groups and given different weight in relation to each other according to their importance for the creation of model regions by the cluster analysis. Redundancy of indicators was detected and described. For the cluster analysis independent indicators were necessary to provide a broad descriptive basis. As a result, some indicators were eliminated, especially redundant ones.

In the course of this work package a comprehensive amount of indicators was reduced to about 43. At the same time mountain agriculture should still be represented in all its aspects covering all different forms of alpine farming from the dominant permanent crops in some parts of the south to the predominantly dairy farming of northern regions.

Definition of a set of representative types of agricultural structures

By means of the structural data of each alpine commune different types of communes were to be defined. As method, the numerical "spatial" classification was applied.

Selection of model regions for these structural types

Model regions, which were typical for each structural type, were selected in all alpine countries.