Factors influencing fish diets in reef food webs

The effects of environmental and anthropogenic factors on fish community structure are typically assessed by compartmentalizing fish species into functional groups. In many cases species are placed in specific trophic groups (e.g. higher carnivores, benthic carnivores, planktivores and herbivores) independent of their size. This assumes that species and individuals within species have similar roles within each trophic group, regardless of the stage of ontogenetic development. This simplistic approach contradicts Hutchinson's well-known paradigm, which suggests that each species possesses a unique role (niche) within the environment it inhabits. The studies presented in this thesis investigate the effects of no-take marine protected areas (MPAs) as well as other factors on fish community structure using this traditional approach (i.e. functional trophic group) and, alternatively, by using diet predictions generated from size- and taxon-based quantitative models. Prey predictions for individual fish have been summed to estimate total fish community consumption, providing a new tool for analysis of trophic pathways in ecological studies. These diet predictions represent an innovative approach for analysing the fish community as a continuum based on the estimated diets of fishes. The removal of large predators from a fish community theoretically should cause a trophic release, allowing the next lower trophic group to increase in size. As part of the work presented in this thesis, I test this assumption by investigating fish community structure using a global dataset with over 1,800 sites, including both no-take MPAs and open-access fishing sites. The trophic release theory predicts that lower trophic groups of smaller size will have lower biomass in MPAs relative to fished sites. As expected, larger (>30cm) fish for all four trophic groups had significantly higher biomasses in MPAs, but, in contrast to expectations, none of the small (<7.5cm) fish groups showed a significant decrease in biomass in MPAs. These results suggest that fishing affects the biomass of all fish trophic groups and size classes, and that this effect is stronger than trophic release through the near absence of larger predator fishes. A predictive dietary model was developed using a comprehensive data set for predator fish and their prey for Western Port, Australia. The predictive dietary model used k-nearest neighbour procedures to predict prey type, and linear models to predict prey size for fish classified at species level, size and location. We obtained reasonable predictions for prey type (mean percent overlap between predicted and actual prey types =77%) and prey size (`r^2` between predicted and observed prey size =93%) when the taxonomic identity of the fish and its size were included in the model. Contrary previous expectation, the most important predictor for prey type was the size of the fish, while the most important predictor for prey size was fish' taxonomic identity. Little loss of accuracy was detected when only family rather than species identity was used in the model. This predictive dietary model, developed for a single location, was applied to widely dispersed locations with differing species composition across southern Australia. At this wider geographical scale the model was robust enough to predict with moderate accuracy prey type (accuracy=67%) and prey size (`r^2` = 56%) using training data from Western Port. Predictions for prey type (accuracy=73%) and mean prey size (`r^2`=89%) substantially improved when data from all southern Australian sites rather than just Western Port were used to 'train' the model. Exclusion of site, habitat, ecoregion or province as factors resulted in little loss in accuracy. Accuracy of continental-scale predictions declined very little when family instead of species was used in the model. Application of our models in situations where family identity and size of the consumer fish are known provides a mechanism for broad-scale testing of influential ecological hypotheses dealing with community-level consumption and trophic structuring. These dietary predictions are appropriate for addressing key questions about the ecology of marine systems and human impacts and management interventions, as well as the effects of environmental variables. In an application of the dietary model, I assessed the effect of no-take MPA protection on fish community consumption for a set of 376 reef sites surveyed using underwater visual census methods in temperate Australia. I found that consumption was higher and prey was larger in MPAs relative to open access sites. This agreed with the coarse trophic group analysis. However, the diet predictive model approach showed that certain diet types were consumed in greater abundances than indicated by the trophic group approach. This finding has important ecological consequences. Our predictive diet model also estimates higher daily consumption of prey of larger size by bigger and more abundant fishes in MPAs, in relation to fished sites. I concluded that more complex ecological pathways appear to operate within no-take MPAs. Fish assemblages in no-take protected areas with sufficient surveillance and enforcement appear to include more habitat-modifying species, and higher numbers of fish species, than in fished locations, thereby contributing to complex ecological processes with possible implications in the habitat structure of the ecosystems.