Simultaneous equations modelling of communities with interacting species networks

To understand community assembly, ecologists have long sought to extract the signal of biotic interactions from species co-occurrence patterns. These efforts face multiple difficulties such as confounding environmental effects, confounding indirect interactions between multiple species and asymmetry of interactions. To address these problems, we propose Simultaneous Community Equations Modelling (SCEM) as a framework to explicitly account for asymmetric interaction networks in community models. SCEM uses a system of equations to model the occurrence of each species as a function of measured and unmeasured (latent) environmental predictors, and the occurrence of potentially all the other species in the community. Biotic interactions most supported by the data are identified using heuristic optimization of a parsimony criterion, implemented as a Genetic Algorithm. Extensive simulations show that SCEM can recover interaction network topologies in virtual communities. We present a software to implement SCEM and illustrate its application with a case study.