The development of an inventory and evaluation methodology for non-timber functions of forests in the frame of management inventories was the aim of the theses. The first objective was to determine the information needed for managing non-timber functions, define appropriate variables and the way to inventory them. A second objective was to find a combination of methods of multivariate statistics that can integrate site specific information and provide for evaluation of multiple observations.
An overview of the literature regarding inventories of non-timber functions and ecological land and site classification resulted in two groups of factors that are needed to describe forest ecosystems and their functions. The first group, named external factors, includes major environmental components (rock formation, soil, climate, vegetation etc.) and human interventions that can alter the effects of environmental factors. The external factors determine ecosystem potentials, that is, the capacity of ecosystem to fulfill a function. The second group – internal factors – comprises attributes of the forest stand. They describe the spatial horizontal (coverage) and vertical (stand-storeys) distribution of vegetation within a stand. The evaluation of the internal factors determines the actual conditions of forest stand with respect to a function, that is, the degree at which a stand fulfills a function. Both groups of factors have been organized in two systems which provide the basis to derive appropriate variables to be inventoried.
Due to the fact that each variable can have a different meaning with respect to different functions, a significance to the variables values should be assigned for each function considered. This has been achieved through research of opinion among experts and scaling of their judgements, using an interval scale with four classes.
The majority of environmental and stand related factors are spatially referenced. But the boundaries of one factor seldom coincide with the boundaries of another. Because a function is the combined expression of all factors, a reference land unit is needed. GIS offers a good solution to this problem. Intersection of the forest area with a grid network can create land units of desired dimensions. A guide for the optimum grid size gives the spatial variability of the factors to be inventoried. In the testing area land units have been created with the help of Arc/Info GIS, using a grid with 1 Km2 cell size for the external factors, while for the internal ones the cell size was selected to agree with the spacing of the grid network of the timber inventory (250 x 250m).
Existing maps and data archives are the main sources of information for the external factors. Aerial photographs and other remote sensing products constitute a gut alternative, also for some of the internal factors. But the majority of the variables of the internal factors are to be inventoried in the field through systematic sampling.
For the evaluation of both external and internal factors with respect to a non-timber function, two methods have been tested. The first one, has been developed in the frame of an EU research project and involves research of experts opinion to derive weights for the factors. The weights estimate the relative importance of each factor in the formulation of a function. The sum of the weighted factor values provides for evaluation of the function. The greater the sum is, the higher the ecosystem capacity (external factors) or the better the conditions of a stand (internal factors) with respect to the function are.
The second method comprises a combination of multivariate analyses. In particular, factor analysis and cluster analysis have been considered. Factor analysis ordinates variables according to the constellation of their correlations. The aim is to derive a model that can explain interrelationships among variables and reduce the number of variables considered for further analysis.
In cluster analysis, the set of variables that describe characteristics of land units is used to derive similarities among land units and construct groups (clusters) in such a way that units in one cluster are the most similar while units in different clusters are most dissimilar. Hierarchical agglomerative clustering of land units has been applied. In a preliminary examination of various clustering algorithms, Ward-linkage resulted in plausible and better interpretable classification of land units.
The testing of the inventory and evaluation methodology has been done in the forest of Thessaloniki, Greece for the non-timber functions water percolation and protection against erosion. Nineteen and eleven variables of the external and internal factors respectively have been derived from the respective systems. Their inventoried values have been scaled with respect to the functions considered and compiled for the 49 land units of the external and the 142 land units of the internal factors.
The weighting technique has been firstly applied for the evaluation of the functions. The evaluation of the external functions showed that the estimation of relative importance (weights) of the factors was incorrect. All land units have been classified of in one class of function’s capacity, although they have presented great variability. Similar were the evaluation results of the internal factors. The majority of land units have been classified in one class of function’s fulfillment and only a small number of them in a second class. The weighting technique induces a reduction of variability at successive weighting steps and leads to a homogenization of land units. Consequently, it provides no basis to derive appropriate management prescriptions or silvicultural measures to be taken in favor of the non-timber functions.
The implementation of the methods of multivariate statistics, on the other hand, have given plausible and useful results. For both functions and groups of factors considered, the models derived from factor analysis seem to far-reaching pass to the actual environmental conditions in the testing area. The high factor loadings have shown that the extracted
factors clearly comprehend important patterns of variability among the variables. Similarly, cluster analysis has resulted in fairly interpretable classification of land units into a small number of clusters suitable to forest management planning. The similarity of land units within clusters allows for each cluster to be perceived as a zone of homogeneous environmental conditions for which common management prescriptions and silvicultural treatments can be drawn.
Conclusively, it can be said the application of multivariate analyses come out with useful to forest management results. The evaluation of the external factors provides management prescriptions that can be easily integrated into forest management plans. The same holds for the silvicultural treatments derived from the evaluation of the internal factors. Both can be easily presented on a map of desired scale and be available to the practical forester.