Spatial and species compositional networks for inferring connectivity patterns in ecological communities

Aim Multiple spatial and non-spatial processes are involved in determining the complex patterns underlying the spatial variation of individual species and their assemblages. This complexity, and the logistical challenges involved in following dispersal for multiple species across multiple sites, make it challenging to infer the processes underlying metacommunity spatial heterogeneity. The goal of our paper is to present a robust quantitative framework for inferring spatial patterns across multiple ecological communities. Innovation Unlike numerous metapopulation studies that have inferred migration rates based on landscape connectivity metrics which take into account the spatial positioning of occupied and empty patches, metacommunity studies have relied on spatial predictors built without considering such information. Here, we introduce a novel method called the multi-species spatial network (MSSN) to detect and explain spatial variability in community assemblies using a graph-theoretical approach. The MSSN approach can be best described as a reconciliation between the spatial positioning of sites and their patterns of patch occupation. Main conclusions Our simulation and real data analyses showed that our MSSN approach was better at detecting spatial patterns within metacommunities than the commonly used MEM method (Moran's eigenvector maps). In addition, our proposed framework is also useful in estimating the levels of spatial connectivity for each local community. Finally, our framework is flexible enough to incorporate different types of functions, metrics and algorithms to detect complex spatial patterns.