Data Simplification for Ecological Network Inference

Abstract

Species-interaction networks are valuable ecological tools for abstracting ecosystem properties related to ecosystem services and resilience. Various approaches have been proposed for inference of existence and directionality of species-interactions from observational data. Nevertheless, contemporary literature recognises shortcomings which need to be overcome before species-interaction networks can be inferred reliably from existing observational data. Here, we present a new species-interaction network inference methodology that uses information on the identity, fitness, phylogenetic relatedness, and functional similarity of interaction partners and climatic suitability to establish whether co-occurrence patterns reflect possible interaction networks. Species-interaction inference is promising for macroecological studies for which direct observation of species-interactions is prohibitively laborious. However, while the availability and volume of observational records increase as the scale of analysis increases, the quality of information decreases. Consequently, we apply our methodology across several scales to investigate how this scale-driven data simplification affects our capability of inferring species-interactions from observational data.