The spatial pattern of a forest ecosystem

Statistical analysis of stands of trees as a whole need suitable methods of spatial statistics. Obviously, trees within a stand affect development and survival of their neighbours. They interact and therefore have to be considered as a system of dependent random variates from an unknown stochastic process. One such statistical model which considers the spatial dependence among trees in a forest and their characteristics is a marked point process. The 'points', called events in spatial statistics, are the tree positions and the 'marks' are tree characteristics such as crown lengths or tree species. A minimal prerequisite for any serious attempt to model an observed pattern is to test the hypothesis of complete spatial randomness (CSR). Concerning the fitting of parametric models to spatial point patterns, a class of models which seems potentially useful for describing the present type of data is the class of marked Gibbs (pairwise interaction) point processes. Essentially, these processes characterise the interaction between events by some parametrically specified function of distance. In this paper several statistical methods to test CSR are described and marked Gibbs processes are used to fit a model in two different forest ecosystems. (C) 1998 Elsevier Science B.V. All rights reserved.