Assessing change of vegetation as a multistate system is a central issue in vegetation ecology (Wildi & Orloci 2007). One could of course argue that time is but another attribute in a sample, as explained in Section 2.3.2, such that no specific treatment would be indicated. However, time has some unique properties. For one thing it is one-dimensional, unlike space where direction is an issue and a decision may be needed when assessing order. Time always proceeds in the same direction, and even more importantly, it is transient. Once an event has taken place, there is no backtracking as can be done in space. This has consequences for investigating change in environmental systems, as discussed in detail by Green (1979). His hierarchical scheme of impact studies is reproduced in Figure 9.1. The most urgent
question in the investigation of an impact is whether reference plots exist. Undoubtedly, striving for a reference is worthwhile, because once the impact has taken place there is no way to reverse the process. One may succeed in protecting plots from an impact, but plots cannot be sheltered from time passing by. Some more philosophical implications of time dependence are discussed in Legendre & Legendre (1998).
In the first part of this chapter I postpone the question of reference, concentrating on the temporal change of sampling units only. This is classical time-series analysis and in the literature of vegetation ecology it is usually found under the buzzword 'succession' (e.g. Maarel 2005). The ideas presented below focus on the specific problem of change in multivariate data space of high dimension, as is the case in releve data.
Was this article helpful?