Modelling the impact of the South African small pelagic fishery on African penguin dynamics William Michael Lewin Robinson Thesis presented for the degree of Doctor of Philosophy in the Department of Mathematics and Applied Mathematics University of Cape Town November 2013 Supervisors: Douglas S. Butterworth and E´va E. Plag´anyi Declaration I hereby declare that this thesis has not been submitted to this or any other university for a degree, either in the same or different form. I know the meaning of plagiarism and declare that all of the work in the thesis, save for that which is properly acknowledged, is my own. In that context, it is to be noted that the general linear models presented in Chapter 3 are an extension of an approach first proposed by Branda˜o and Butterworth (2007). iii Contents Declaration iii Abstract ix Acknowledgements xi Abbreviations xiii 1 Introduction 1 1.1 The ecosystem approach to fisheries with a focus on impacts on seabirds 2 1.1.1 Concern over global fishing intensity 2 1.1.2 Ecosystem considerations in fisheries 4 1.1.3 Implementations of EAF/EBFM 5 1.1.4 The importance of forage fish 7 1.1.5 Current trends in fisheries management 8 1.1.6 Ecosystem modelling 10 1.1.7 Seabirds as ecological indicators 11 1.2 Review of approaches to modelling seabird–fishery interactions 13 1.2.1 CCAMLR Ecosystem Monitoring Programme 14 1.2.2 Other seabird–fishery interaction studies 15 1.3 African penguins 16 1.3.1 Population trends 16 1.3.2 Prey abundance and distribution 18 1.3.3 Predation 18 1.3.4 Oil pollution 20 1.3.5 Other influences on penguin dynamics 21 v 1.4 The South African fishery for small pelagics 22 1.4.1 Recent prey distribution shift 24 1.5 The management procedure approach 25 1.5.1 Description of the MP approach 26 1.5.2 Management of the South African small pelagic fishery 29 1.6 Past approaches to assessing African penguin–small pelagic fishery interactions 32 1.7 Thesis objectives and outline 33 2 Data 35 2.1 Moult and breeder counts 36 2.1.1 Double Gaussian method 38 2.1.2 Results 40 2.1.3 Discussion 41 2.2 Re-sightings of banded birds 43 2.3 Major oil spills 46 2.3.1 Apollo Sea 46 2.3.2 Treasure 47 2.4 Pelagic biomass 48 2.5 Data series for GLM analyses 50 2.5.1 Fledging success 50 2.5.2 Breeders per adult moulter 51 2.5.3 Active and potential nests 52 2.5.4 Survival 52 2.5.5 Foraging trip parameters 53 2.5.6 Sardine and anchovy catches 53 2.5.7 Sardine and anchovy biomass 54 3 GLMs relating penguin demographics and foraging behaviour to forage fish abundances 75 3.1 Fishery closures experiment 76 3.2 Part A—GLM analyses conducted in 2010 77 3.2.1 Response and explanatory variables 78 vi 3.2.2 Results 82 3.2.3 Analysis 82 3.2.4 Discussion 84 3.3 Part B—GLMs including updated data 85 3.3.1 Response and explanatory variables 85 3.3.2 Results 87 3.3.3 Analysis 88 3.3.4 Discussion 89 4 Penguin–fish interaction model 125 4.1 Basic dynamics 126 4.1.1 Annual adult mortality 128 4.1.2 Age at first breeding 129 4.2 Population model 129 4.2.1 Adult mortality 131 4.2.2 Biomass–mortality relationship 132 4.2.3 Reproductive success 133 4.2.4 Immigration 135 4.3 Model fitting 135 4.3.1 Fitting to moult counts 135 4.3.2 Fitting to tag data 136 4.3.3 Objective function minimized 138 4.3.4 Parameter estimation 138 4.4 Results and discussion 140 5 Penguin population projections 159 5.1 Linkage with the small pelagic OMP 159 5.2 Future sardine biomass spatial distribution 161 5.3 Projected penguin numbers 162 5.4 Results 164 5.5 Sensitivity tests 165 5.6 Discussion 165 vii 5.6.1 Base case results of linkage with the pelagic OMP 165 5.6.2 Sensitivity tests 167 6 Conclusions 175 6.1 Thesis summary 175 6.1.1 Moult count aggregation 175 6.1.2 GLMs for penguin demographics and foraging behaviour 175 6.1.3 Penguin–fishery interaction model 176 6.1.4 Penguin population projections 179 6.2 Future research 181 6.2.1 Moult count aggregation 181 6.2.2 GLMs for penguin demographics and foraging behaviour 181 6.2.3 Penguin–fishery interaction model 182 6.3 Concluding remarks 184 References 187 viii Abstract Under an ecosystem approach to fisheries, managers aim not only to achieve a good yield from targeted stocks, but also to preserve the functioning of the whole ecosystem. The population demographics of predators such as penguins that are dependent on forage fish for food can give an indication of the health of the ecosystem in which they live. The rapid decline in the African penguin Spheniscus demersus population in the twenty-first century prompted the re-classification of the IUCN conservation status of the species as “en- dangered” in 2010. Reasons suggested for the decline include scarcity of quality accessible prey, predation, and oil spills. The South African fishery for small pelagics catches predominantly sardine Sardinops sagax and anchovy Engraulis encrasicolus, which are the two most important prey species of African penguins. This project aims to quantitatively assess the past and future impact of the fishery on penguin population dynamics and abundance trends. The working group which provides scientific advice on the management of the sardine-anchovy fishery convened a task team to conduct a feasibility study to evaluate the power of an experiment to detect an effect on penguin demographics and foraging behaviour of fishing restrictions around penguin breeding colonies. The colonies considered include Dassen Island and Robben Island off the west coast of South Africa and St Croix Island and Bird Island in Algoa Bay. The extent of relationships between penguin demographic parameters at these colonies and extractions of forage fish was explored with general linear models, taking account of fish abundance through biomass estimates from surveys. Under the assumption that prey densities at neighbouring islands should vary in approximate synchrony, the differential effect of fishing was evaluated. Results indicate (perhaps surprisingly) that penguin reproduction is better when catches are larger. Furthermore, the CVs of the standard deviations of the residuals of the response variables were estimated. ix These estimates are necessary to determine whether an island closure experiment will have the power to detect an effect of fishing within a reasonable period. A model of the Robben Island penguin population dynamics was developed. Penguin adult annual mortality was related to the sardine 1+ biomass observed west of Cape Agulhas in the November hydroacoustic survey for small pelagic fish. Data from moult counts and re- sightings of tagged penguins were integrated in a rigorous and statistically defensible manner. Results suggest that the rapid growth of the colony during the 1990s was driven primarily by immigration. Furthermore, the sardine–penguin mortality relationship predicts that on average penguin survival decreases only when the sardine 1+ survey biomass west of Cape Agulhas is less than approximately one-quarter of the maximum level observed (1343 thousand tonnes in 2003). The population model was used to calculate ten-year projections of penguin numbers. Adult survival rates were based on future sardine abundances predicted under the sardine and anchovy operational management procedure (formula for setting total allowable catches) currently being tested and its associated base case operating model. Projections were compared to equivalent scenarios without fishing. Results indicate that the effect of fishing is likely to be rather small, especially when compared to uncertainties in the dynamics related to the variable spatial distribution of the sardine population. The coupling of the penguin population model to the sardine operating model provides an objective assessment of the impact of fishing on Robben Island penguins. This is an example of how a fishery can be managed considering not only single-species goals but also ecosystem-related effects. Such quantitative consideration and fitting to data of a non-target predator is rarely found in fisheries management. This work advances implementation of an ecosystem approach to fisheries, which is important in the context of increasing pressure worldwide to account for the foraging needs of dependent predators. x