Multi-Species Foraging Associations in the Hauraki Gulf

Abstract

Multi-species foraging associations (MSFAs) are characterised by the joining of two or more species to feed on ephemeral prey patches. Demographic analysis provides valuable insight into species interactions in MSFAs, with compositions and affinities highlighting the different costs and benefits of associating, such as kleptoparasitism, predation risk, and shifting prey availability through the scattering/dispersal and herding behaviours of other predators. Here, we use a combination of demographic and activity budget analysis tools to investigate MSFA dynamics (n = 179) in the Hauraki Gulf surveyed between 2019 and 2021. Within MSFAs, eight unique clusters of species were found, with associations forming on the basis of prey type and movement and foraging ecology. Species exhibited distinct social affinities, reflecting foraging ecologies, seasonal prey availability and predator movements and migrations. Using activity budget analysis tools and group focal follows, we found that cluster influenced the proportion of time groups spent in various behavioural states, with planktivores, and planktivore cluster participants, spending more time actively foraging than piscivores and piscivore and multitrophic cluster participants. Using these activity budgets, we provide novel insights into the influence of predator composition (cluster) and season on group behaviours. To further this work, we developed a proof-of-concept artificial intelligence (AI) tool to automate the detection and classification of animals and behaviours from imagery with human-like accuracy. Resulting YoloV4 and ResNet-50 models indicate that species can be detected to a mean average precision of 93.6 - 99.6%, and behavioural classifiers to 30.7% - 97.1%. Model accuracy improved with image resolution, but not number of classifier layers, with accuracy further increasing with the number of trained objects (animals), iterations and epochs. AI models were thus proven to be useful for future analysis of behavioural data from drone footage. Overall, this research provides a baseline of information and understanding of the complexity of MSFA dynamics with the current ocean state. As large marine animals reflect cumulative stressors in their environment, fine-scale drivers of MSFAs warrant further research, with long-term monitoring of MSFAs providing insight into the effects of future climate and human-mediated environmental changes in the Hauraki Gulf Marine Park.