Spatio-phylogenetic multispecies distribution models

Abstract

Summary: Ecologists increasingly consider phylogenetic relatedness in both community composition and spatial arrangements in communities. Here we considered both the phylogenetic correlation between multiple species and the spatial correlation induced by unobserved spatial heterogeneity on multiple plots. For this analysis, we introduced phylogenetic spatial generalised linear mixed models (PSGLMMs), which are an extension of phylogenetic generalised linear mixed models (PGLMMs). We used the framework of generalised linear array models to simultaneously model species and plot dimension. Such models have the potential to explain the correlation of the phylogenetic relationship of the observed species and of the spatial proximity of the plots, or both. We proposed model selection strategies based on proper scores and empirically evaluated them in a case study using bird count data. In our analysis, we focused on two special cases: the community composition model and the environmental sensitivity model. Our simulation study indicated that it might be difficult to correctly identify phylogenetic signals when the phylogenetic correlation is rather low and when studying presence-absence or count data of rare or pervasive species.