Covariate-informed latent interaction models: Addressing geographic & taxonomic bias in predicting bird-plant interactions.

Climate change and reductions in natural habitats necessitate that we better understand species' interactivity and how biological communities respond to environmental changes. However, ecological studies of species' interactions are limited by geographic and taxonomic bias which can lead to severe under-representation of certain species and distort our understanding of inter-species interactions. We illustrate that ignoring these biases can result in poor performance. We develop a model for predicting species' interactions that (a) accounts for errors in the recorded interaction networks, (b) addresses the geographic and taxonomic bias of existing studies, (c) is based on latent factors to increase flexibility and borrow information across species, (d) incorporates covariates in a flexible manner to inform the latent factors, and (e) uses a meta-analysis data set from 166 individual studies. We focus on interactions among 242 birds and 511 plants in the Brazilian Atlantic Forest, and identify 5% of pairs of species with an unrecorded interaction, but posterior probability of existing that is over 80%. Finally, we develop a permutation-based variable importance procedure and identify that a bird's body mass and a plant's fruit diameter are most important in driving the presence and detection of species interactions, with a multiplicative relationship.