Understanding the interplay between processes operating at large and small spatiotemporal scales in shaping biotic interactions within biological communities remains challenging. Recent studies illustrate how phenotypic specialization, species life-history traits and/or resource partitioning recurrently underlie the structure of mutualistic interactions in terrestrial ecosystems along large latitudinal gradients of biodiversity. However, we know considerably less about how local processes interact with large-scale patterns of biodiversity in modulating biotic interactions in the marine realm. Considering agonistic behaviour as a proxy for contest competition, we empirically investigate whether the structure of reef fish agonistic interactions is conserved across a 34,000-km longitudinal gradient of biodiversity. By sampling coral reefs using standardized remote underwater video, we found recurrent patterns of fish agonistic behaviour in disparate communities distributed across five biogeographic provinces of the Pacific and Atlantic oceans. While the sheer number of species increases with regional richness, the number of aggressive disputes at the habitat scale is similar across communities. We then combined generalized linear models and network theory to reveal that, the emergent structure of local agonistic networks was not modular but instead recurrently display a nested structure, with a core of highly interactive site-attached herbivores of the Pomacentridae family. Therefore, despite the increase in the number of species involved in agonistic interactions toward speciose communities, the network structure is conserved along the longitudinal richness gradient because local disputes are mostly driven by closely-related, functionally-similar species. These findings suggest that evolutionary and local processes interact in modulating reef fish agonistic behaviour and that fine-scale niche-partitioning can structure the ecological networks in marine ecosystems.

We surveyed seven tropical reefs over a 34.000 km longitudinal gradient of biodiversity: four at the Atlantic (Rocas Atoll, Abrolhos, Curaçao, Yucatán) and three at the Indo-Pacific Ocean (Moorea, Fiji, Bali). We designed a standardised study to record data on reef fish behaviour using Remote Underwater Video (RUV) sampling. We carried out the same sampling effort in all localities to control for possible influence of sampling on the relationships between agonistic behaviour and species richness across communities. Our standardised sampling effort (87.5h of footage) consisted of 50 video samples per locality. A remote video sample consisted of a 15-min shot by a high-definition camera, fixed and focused over a 2m² reef area previously chosen at random by a free diver. To encompass the local habitat heterogeneity, we sampled three different sheltered shallow reefs with depths between 2–10 m in each locality.

We identified all reeef fish species and recorded all agonistic interaction events between them occurring in the central 10 min of each video sample (the first and last 2.5 min were discarded to remove any potential disturbance by the observers). We considered an agonistic interaction as a chasing event between individual fish—a behaviour that is well-documented as representative of disputes for food resources, territory and/or mating opportunities. We focused on species larger than 5 cm in total length because identification of smaller individuals in remote video analysis can be unprecise.

We investigated the distribution of agonistic interactions among reef fishes across localities using the network formalism. For each locality, we combined the data from all 50 video samples to create an asymmetrical binary adjacency matrix in which elements a_ij=1 represented the agonistic behaviour of reef fish species itowards the speciesj, a_ji=1 the behaviour of jtowards i, and a_ij=a_ji=0 when species never interacted. These one-mode directed networks depicted species as nodes colour-coded by taxonomic families connected by links indicating the direction of the agonistic interaction (outward=actor, inward links=receptor). 

We related the number of agonistic interactions performed by species of all localities to key descriptors of reef fish agonistic behaviour—taxonomic family, trophic groups, and home range. Since our study focuses on macroecological scale patterns, we classified fish species into broad functional groups. We considered five trophic groups based on their primary diets—herbivores, carnivores, omnivores, planktivores, and invertebrate feeders—and three categories of home range—site-attached (less than 5m²), small home range (within reef) and mobile (between reefs).

We provide a table of reef fish richness recorded across spatial scales and seven adjacency matrices of reef fish agonistic behaviour from seven tropical reefs in the Atlantic and Indo-Pacific Oceans: Rocas Atoll, Abrolhos, Curaçao, Yucatán, Moorea, Fiji and Bali. The species richness table comprises the total number of reef fish species observed in the remove underwater videos, and number of species observed interacting agonistically across spatial scales and localities. The total number of species in all video samples  (local scale - 100m²)  and the average of the total number of reef fish species registered on each video sample( habitat scale - 2m²) were detected for each locality. The seven adjacency matrices describe the one-mode directed networks of agonistic interactions among reef fish species on each locality (across all 50 video samples, local scale - 100m²), whose scientific names are given in the rows and columns.