Prediction of species composition of plant communities in a rural landscape based on species traits

Abstract

AbstractThe species composition of a community is a subset of the regional species pool, and predicting the species composition of a community from ecological traits of organisms is an important objective in ecology. If such a prediction can be made feasible, we could assess the risk of invasion of locally new species (alien species and genetically modified species) into natural communities. We developed and tested statistical models to predict a community's species composition from ecological traits of the species pool. Various types of communities (forest, meadow, and weed communities) exist in a small area of traditional rural landscape in Japan, and have been maintained by human activities. These communities and the tracheophytes species pool in the 1‐km2 research area were considered. We used logistic regression and decision‐tree analysis to construct predictive models of community species composition based on plant traits, using the presence or absence of species in a community as the dependent variable and ecological traits as independent variables. Plant traits were grouped by cluster analysis, and the average in each trait group was used for model building to avoid multiple collinearity. Statistical prediction models were significant in all communities. About 60–75% of species composition could be predicted from the measured plant traits in forest communities, but 33–56% in the meadow and weed communities. Our results showed the possibility of predicting the species composition of plant communities from the ecological traits of the plant species together with the information on local species pool.