Joint Species Distribution Models - a summary for macroecologists

A long-standing criticism of species distribution models (SDMs) is that both single-species and stacked SDMs fail to capture the influence of species on each other (biotic interactions), which arguably limits their ability to infer fundamental niches and produce accurate climate change projections. Joint species distribution models (jSDMs) appear to solve this problem by introducing a species-species covariance term on top of a stacked SDM. However, if and when this covariance can be interpreted as a biotic interaction has been the subject of intense debate. In this talk, I will take stock of the state-of-the-art in jSDM modeling for macroecology. I will review the assumptions and technical implementations of current jSDM software, summarize what information can be extracted from fitted models, and explain when they can provide better predictions than simple stacked SDMs. I will argue that some early statements about jSDMs, particularly about their ability to correct for biotic interactions, were probably overly optimistic. Nevertheless, jSDMs are currently the most complete tool for statistical analysis of spatial community data, and as such, they have the potential to complement SDMs as a standard approach for macroecological analyses.