Abstract1. With global change modifying species assemblages, our success in predicting ecosystem-level consequences of these new communities will depend, in part, on our ability to understand biotic interactions. Current food web theory considers interactions between numerous species simultaneously, but descriptive models are unable to predict interactions between newly co-occurring species. Incorporating proxies such as functional traits and phylogeny into models could help infer predator/prey interactions.2. Here, we used trait matching between predator feeding traits and prey vulnerability traits, along with phylogeny (used as a proxy for chemical defence and other traits difficult to document), to infer predatory interactions using ground beetles as model organisms.3. A feeding experiment was conducted involving 20 ground beetle and 115 prey species to determine which pair of species did or did not interact. Eight predator and four prey functional traits were measured directly on specimens. Then, using a modelling approach based on the matching-centrality formalism, we evaluated 511 predictive ecological models that tested different combinations of all predator and prey functional traits, and phylogenetic information.4. The most parsimonious model accurately predicted 81% of the observed realized and unrealized interactions, using phylogenetic information and the trait-matches predator biting force/prey cuticular toughness and predator/prey body size ratio.The best trait-based models predicted correctly >80% which species interact (realized interactions), but predict <58% of which species did not interact (unrealized interactions). Adding a phylogenetic term representing the evolutionary distance within each trophic level increased the ability to predict which species did not interact to >75%.5. The matching of predator biting force and prey cuticular toughness demonstrated a better predictive power than the commonly used predator/prey body size ratio.Our novel model combining both functional traits and phylogeny extends beyond existing descriptive approaches and could represent a valuable tool to predict consumer/resource interactions of newly introduced species and to resolve cryptic food webs. K E Y W O R D Sfood web, functional traits, matching-centrality formalism, networks, trophic interactions

The characterization of ecological communities with functional traits allows to consider simultaneously the ability of a species to survive and reproduce in an environment, its interactions with other species and its effects on the ecosystem. Functional traits have been studied mainly by plant ecologists, but are increasingly common in the study of other taxa including arthropods. Arthropods represent a group for which a functional trait approach could be highly profitable because of their high diversity, abundance, ubiquity and role in many important ecological processes. This review synthesizes two decades of functional trait research on terrestrial arthropods. We show that while the approach has gained popularity, particularly in the last decade, the absence of clearly postulated hypotheses is a recurrent problem limiting generalization. Furthermore, studied traits are often poorly related to studied functions. To address these problems, we propose a step-by-step protocol to postulate clear hypotheses prior to trait selection and emphasize the need for a common set of more generalizable traits in future studies. Extending the functional trait approach to arthropods opens the door to improving our understanding of interspecific interactions and potential links between response and effect traits. We present the concept of trait-matching with several examples of arthropod traits known to be effective predictors of consumer-resource interactions. The development of a successful functional trait approach for terrestrial arthropods will necessitate an understanding of relevant traits, standardized measurement protocols and open access databases to share this information. Such progress will provide ecologists with a new set of tools to answer broad questions in several fields including the study of community assembly, ecological networks and multitrophic functionality.

The functional trait approach proposes that relating traits of organisms within a community to variation in abiotic and biotic characteristics of their environment will provide insight on the mechanisms of community assembly. As traits at a given trophic level might act as filters for the selection of traits at another trophic level, we hypothesized that traits of consumers and of their resources covary in space. We evaluated complementary predictions about top‐down (negative) and bottom‐up (positive) trait covariation in a detrital food web. Additionally, we tested whether positive trait covariation was better explained by the Resource Concentration Hypothesis (i.e., most commonly represented trait values attract abundant consumers) or the Resource Specialization Hypothesis (i.e., resource diversity increases niche availability for the consumers). Macroarthopods were collected with pitfall traps over two summers in three forested sites of southern Quebec in 110 plots that varied in tree species composition. Six feeding traits of consumers (detritivores and predators) and six palatability traits of their resources (leaf litter and prey) were matched to assess spatial covariation. Trait matches included consumer biting force/resource toughness, detritivore mandibular gape/leaf thickness, predator/prey body size ratio, etc. Our results demonstrate for the first time a covariation between feeding traits of detritivores and palatability traits of leaf litter (31–34%), and between feeding traits of litter‐dwelling predators and palatability traits of potential prey (38–44%). The observed positive covariation supports both the Resource Concentration Hypothesis and Resource Specialization Hypothesis. Spatial covariation of consumer and resource traits provides a new tool to partially predict the structure of the detrital food web. Nonetheless, top‐down regulation remains difficult to confirm. Further research on top‐down processes will be undoubtedly necessary to refine our capacity to interpret the effect of biotic interactions on co‐distribution.

No abstract