Community ecology in the age of multivariate multiscale spatial analysis

作者：Stéphane Dray, Raphaël Pélissier, Pierre Couteron, Marie‐Josée Fortin, Pierre Legendre, Pedro R. Peres‐Neto, Edwige Bellier, Roger Bivand, F. Guillaume Blanchet, Miquel De Cáceres, Anne‐Béatrice Dufour, Einar Heegaard, Thibaut Jombart, François Munoz, Jari Oksanen, Jean Thioulouse, Helene H. Wagner · 发表于：Ecological Monographs · 年份：2012 · DOI：10.1890/11-1183.1 · 被引用次数：686 · 研究领域：Ecology and Vegetation Dynamics Studies、Land Use and Ecosystem Services、Plant and animal studies

Species spatial distributions are the result of population demography, behavioral traits, and species interactions in spatially heterogeneous environmental conditions. Hence the composition of species assemblages is an integrative response variable, and its variability can be explained by the complex interplay among several structuring factors. The thorough analysis of spatial variation in species assemblages may help infer processes shaping ecological communities. We suggest that ecological studies would benefit from the combined use of the classical statistical models of community composition data, such as constrained or unconstrained multivariate analyses of site‐by‐species abundance tables, with rapidly emerging and diversifying methods of spatial pattern analysis. Doing so allows one to deal with spatially explicit ecological models of beta diversity in a biogeographic context through the multiscale analysis of spatial patterns in original species data tables, including spatial characterization of fitted or residual variation from environmental models. We summarize here the recent progress for specifying spatial features through spatial weighting matrices and spatial eigenfunctions in order to define spatially constrained or scale‐explicit multivariate analyses. Through a worked example on tropical tree communities, we also show the potential of the overall approach to identify significant residual spatial patterns that could arise from the omission of important unmeasur...