Variation in species composition explained by spatially autocorrelated environmental variables
David Zelený1, Li-Wan Chang2, Chang-Fu Hsieh3
1 Department of Botany and Zoology, Masaryk University, Czech Republic
2 Technical Service Division, Taiwan Forestry Research Institute, Taiwan
3 Institute of Ecology and Evolutionary Biology, National Taiwan University, Taiwan
in Modelling Species and Ecosystems
Variation explained by environmental variables in multivariate methods such as constrained ordination is often used to quantify the link between environment and species composition (concept of R2 and adjusted R2). In the presence of spatial autocorrelation, explained variation is biased toward higher values (the more the variable is autocorrelated, the more variation it explains). Here, we introduce concept of explained variation adjusted for the effect of spatial autocorrelation. We define it as the amount of variation explained by one or several real spatially autocorrelated variables additionally to the variation explained by the same number of randomly generated variables with the same level of spatial autocorrelation. Using artificial data, we compare performance of several methods generating random variables with certain level of spatial autocorrelation (such as toroidal shift, dual-tree complex wavelet transform). Using vegetation dataset from 25 ha subtropical forest dynamics plot in Taiwan as a case study, we study the relationship between species composition and environmental variables, quantified by explained variation adjusted for the effect of spatial autocorrelation.
Keywords: constrained ordination, forest dynamics plot, R2, random spatially autocorrelation variables