Analyse der Beziehungen zwischen Standort und Vegetation mit der logistischen Regression - eine Studie in Nadelwäldern des Schwarzwaldes
Relationships between vegetation and environment were examined in coniferous forests of the southeastern part of the "Mittlere Schwarzwald" (Middle Black Forest, southwestern Germany, Fig. 1). One hundred plots were located by systematic random sampling, with minimal requirement on their homogeneity. The environmental variables investigated were: geomorphology, climate, geology, site history, soil and humus layer characteristics, nutrient contents of Hylocomium splendens, stand age and structure, and relative photosynthetically active radiation. In all, 43 variables are summarized in Tabs. 11-19. Some parameters are displayed in Figs. K1-K19, which show the systematic sampling grid; each plot is represented by one quadrat. By means of correlation and principal component analyses (Tabs. 8-10), the relations of 58 site factors were analysed. Differences within the study area as well as the significance regarding vegetation are discussed. On the whole, the climate is of a cool and humid montane type, the soils are very acidic and poor in nutrients. There is a marked west-east gradient within the investigation area with an ecocline of climate, geology, geomorphology, pedological factors and air pollution. The second important complex of factors, humus layer characteristics and the corresponding nutrient supply, is characterized by localized variation. The vegetation was classified by a numerical method (TWINSPAN) and Braun-Blanquet tablework (Tabs. A-D). All vegetation types are exactly defined by differences of differential species groups. Even though they represent special anthropogenic types of the eastern Black Forest, they can be assigned to the forest associations Galio-, Luzulo- and Vaccinio-Abietetum. Vegetation types are displayed in Fig. K20, the distribution of some plant species in Figs. K21-K40. By means of logistic regression analysis, relationships of environmental factors to 23 plant species and 9 vegetation types are studied. Statistical results, features of individual sample plots and evidence from the literature were considered in setting up statistically and ecologically significant regression models (Tab. E). Using these, the probability of occurrence of selected plant species or vegetation types can be estimated if some site factors are known. In this way the vegetational properties of stands can be "predicted". In addition to the regression parameters, the ecological behaviour of the vegetation is characterized by "Wso-values" (the 50 % probability of occurrence) and species (or vegetation type) response curves (Figs. 2-32). The nutrient supply is of particular importance, but other environmental variables play important roles, too. The application of logistic regression analysis in the context of vegetation ecology is discussed in detail. It has proved to be a valuable tool if certain precautions are followed.