Environmental correlates of the fine scale understory variation in oak or hornbeam dominated forests in the Czech Republic
| Authors | |
|---|---|
| Year of publication | 2014 |
| Type | Appeared in Conference without Proceedings |
| MU Faculty or unit | |
| Citation | |
| Description | To analyze a fine scale environmental and community heterogeneity is a somewhat daunting task, mainly due to a great sampling effort required and an almost inevitable spatial dependence between samples imposed by their close spatial proximity. Spatial autocorrelation in data, in turn, leads to inflation of Type I error rates, which makes the tests of significance invalid. This might be particularly serious when identifying environmental variables associated with ecological patterns. In this study we address a relatively simple question: which environmental variables are the best predictors of fine scale community structure within the forest understory when controlled for spatial autocorrelation of model residuals? Taking an advantage of our regular sampling design, we adopted the eigenvector-based spatial filtering approach, a method convenient for its applicability in the familiar multiple regression framework. We collected data in three oak and hornbeam dominated forest sites located in south-eastern part of the CzechRepublic. At each site we established a set of one hundred 2 × 2 m plots covering an extent of 1 ha. In each plot, we recorded species composition of herb understory and measured a set of environmental variables characterizing topographical, soil and light conditions. To deal with spatial autocorrelation of both species composition and environmental variables, a set of spatial filters (Moran’s eigenvector maps) was generated using eigenvector decomposition of connectivity matrix and incorporated into partial RDA models along with environmental predictors. The three sites differed both in the amount of variation explained by environmental factors (R2adj = 0.23–0.02) and the relevant environmental factors themselves. The most important variables were soil pH, phosphorus, topography and light. When accounted for the spatial dependence, some of the selected variables lost their significance. However, the models combining more variables remained significant. |
| Related projects: |