The objective of this thesis is to update the Mori-Butterworth (2006) model of the krill-predator dynamics of the Antarctic ecosystem. Their analysis aimed to determine whether predator-prey interactions alone could broadly explain the observed population trends of the species considered in their model. In this thesis, the Antarctic ecosystem is outlined brie y and details are given of the main krill-eating predators including whales, seals, fish and penguins, together with an historical record of the human harvesting in the region. The abundances and per capita krill consumption of the krill-predators are calculated and used to determine the main krill-predators to be used in the updated model developed. These predators are found to be the blue, fin, humpback and minke whales and crabeater and Antarctic fur seals. The three main ship surveys (IDCR/SOWER, JARPA and JSV) used to estimate whale abundance, and the abundance estimation method itself (called distance sampling), are summarised. Updated estimates of abundance and trends are listed for the main krill-predators. Updated estimates for the biological parameters needed for the ecosystem model are also reported, and include some differences in approaches to those adopted for the Mori-Butterworth model. The background to the hypothesis of a krill-surplus during the mid-20th century is discussed as well as the effects of environmental change in the context of possible causes of the population changes of the main krill-feeding predators over the last century. Key features of the results of the updated model are the inclusion of a depensatory effect for Antarctic fur seals in the krill and predator dynamics, and the imposition of bounds on Ka (the carrying capacity of krill in Region a, in the absence of its predators); these lead to a better fit overall. A particular difference in results compared to those from the Mori-Butterworth model is more oscillatory behaviour in the trajectories for krill and some of its main predators. This likely results from the different approach to modelling natural mortality for krill and warrants further investigation. That may in turn resolve a key mismatch in the model which predicts minke oscillations in the Indo-Pacific region to be out of phase with results from a SCAA assessment of these whales. A number of other areas for suggested future research are listed. The updated model presented in this thesis requires further development before it might be considered sufficiently reliable for providing advice for the regulation and implementation of suitable conservation and harvesting strategies in the Antarctic.

Accurate interpretation of recent variability observed in fish populations, species compositions, distribution ranges or ecological indicators, depends on knowledge of their past dynamics and historical states. The onset of systematic fisheries data collection typically lagged decades or even centuries behind modern fishing exploits. As a consequence, pre-disturbed reference points and descriptions of subsequent change are rarely available. A remarkably detailed set of historical trawl survey data from South Africa provided such a rare opportunity. Government-funded exploration of the Agulhas Bank fishery potential resulted in meticulously-documented trawl survey data from 1897-1904, when prior human impacts on those resources were negligible. Although they used less effective technology, the information recorded and methods used were similar to modern surveys. This thesis investigated change in demersal fish fauna of the Agulhas Bank and documents comparisons between historical trawl surveys and modern re-enactments at the same locations. In comparing trawl survey catches over multiple decades or among different periods, unquantified changes in fishing power pose a key challenge. The shape, size, materials, mesh sizes and speed at which trawl nets are dragged, interact with the behaviour, size and shape of fish, influencing fishing performance. To accurately compare current catch rates with those of historical trawl surveys, the same trawl gear and methods were carefully replicated in repeat surveys at three sites. An investigation of literature and photographs of the original vessel and equipment were conducted to support the construction of a replica 'Granton' otter trawl net. The net was composed of Manila hemp with a headline length of 27 m (90 ft) and was attached to flat wooden trawl boards. The historical towing speed was estimated as 1.34 m s⁻¹ (2.6 knots). Three parts of the shallow Agulhas Bank that were surveyed 111 years prior, were re-surveyed in 2015. Species composition was contrasted between the historical and re-survey periods by way of unconstrained ordination, permutational multivariate analysis of variance and tests of the homogeneity of multivariate group dispersions. Taxa discerning between periods were identified with similarity percentage analyses. Changes of standardised catch between periods were tested for 27 taxa, using a non-parametric bootstrap approach. Proportions among size-classes, recorded for three taxa, were tested using Fisher's exact test. Results revealed a substantially transformed demersal catch assemblage, where the period effect explained almost half of the measured variance among samples. These changes included the disappearance or heavy depletion of kob (Argyrosomus spp., absent in re-surveys), panga (Pterogymnus laniarius; 2.4% of historical catch abundance) and east coast sole (Austroglossus pectoralis; 4.6% of historical catch abundance), which had jointly contributed 70-84% of historical catch composition. Average re-survey catches were largely made up of gurnards (Chelidonichthys spp.; 3 792% of historical abundance), horse mackerel (Trachurus trachurus; 4 738% of historical abundance), spiny dogfish (Squalus spp.; 3 121% of historical abundance), hake (Merluccius capensis; 558% of historical abundance) and white sea catfish (Galeichthys feliceps; 13 863% of historical abundance). Analysis of available length information confirmed the expectation that fish sizes (specifically M. capensis and A. pectoralis) had declined. This implies that comparisons by weight would be more severe for declined abundances and less severe for those that increased, relative to contrasts of numerical abundance. Habitat preferences as well as geographic and depth distribution appeared to separate the taxa that increased from those that declined. These factors, together with reproductive and growth characteristics, as well as indirect trophic impacts, likely shaped the responses of demersal fauna to fishing and other human impacts during the 111 years between trawl surveys. An assessment of distribution changes of 44 common demersal taxa was undertaken. These analyses were restricted to the last 30 years of trawl survey data as the units and spatio-temporal resolution of prior data were incompatible. Standardised catches were used from annual spring and autumn south coast trawl surveys conducted by the government fisheries department. Geostatistical delta-generalised linear mixed models were used to predict species distribution functions, which were used to calculate annual estimates of latitude/longitude centres of gravity and effective areas occupied by each population. Average trends over the study period (1986-2016) were assessed using a Bayesian state-space model. Of nine species found to have a trend in average location, six moved westward or south-westward, while three moved eastward or north-eastward. Two species showed a trend of contracting spatial extent and one showed an expansion. Across the entire assemblage combined, there was a significant contraction in extent and a westward shift in average location. These assemblage-wide average trends are interpreted to be driven by climate forcing. Fishing impacts are expected to have contributed to the eastward movement in centre of gravity for kob, lesser sandshark (Rhinobatos annulatus) and white stumpnose (Rhabdosargus globiceps). Interpretation of these distribution shifts is hampered by a lack of knowledge on subsurface hydrographic trends on the Agulhas Bank, which is identified as a research priority. My research revealed substantial change in demersal fish communities on South Africa's Agulhas Bank and adds novel insight to the history underlying current states of demersal ecosystems. Valuable additions include a) documentation of the extent to which demersal assemblages have transformed during the last century at three representative inshore sites; b) estimates of current abundances relative to pre-disturbed baselines at those sites, which c) highlights drastic local depletion for certain taxa and d) reveals substantial abundance increases of certain species during the post-industrial fishing period; e) novel evidence of distribution trends in south coast demersal species; and f) identification of trends in the average distribution of the demersal fish assemblage, suspected to be climate-forced. Long-term comparisons, using minimally-disturbed baselines, revealed drastic transformation of the fish assemblage during a century of industrialisation, which points to trawling-induced alteration of benthic habitats and substantial changes in ecosystem structure. Besides the provision of novel historical context for current and future studies and decision-making, this work counters the erosive nature of shifting baselines in South Africa's marine environment.

An understanding of the population structure of commercially exploited species is essential for effective fisheries management. Fish stocks are typically identified based on the observation of differences in genetic and/or phenotypic characteristics between fish from discrete units. Recent deliberations on the management of the South African purse-seine sardine fishery recognises the likely existence of two stocks, one on the west and another on the south coasts, following studies that have documented spatial variability in several phenotypic characteristics of this species around South Africa. Those studies typically examined spatial variability in a single characteristic (e.g. gill raker morphology and meristics, body shape, otolith shape, vertebral count, parasite loads), but the application of multiple stock identification methods to the same individual fish has been recommended in order to maximize the likelihood of correctly inferring and identifying fish stocks. This study seeks to assess the population structure of South African sardine Sardinops sagax using a combination of stock identification methods including meristic, morphometric and parasite markers. Most of the sardine samples used in this study were obtained off the west and south coasts of South Africa during pelagic surveys conducted in 2013 and 2014 by Fisheries Branch of the Department of Agriculture, Forestry and Fisheries, with some additional samples obtained from commercial fisheries. Fish caught from the west and south coasts are presumed to be part of the putative western and southern stocks respectively. The variables examined included body shape, gill arch length, gill raker spacing, number of gill rakers, otolith shape indices (otolith circularity and form factor), the number of vertebrae, and the abundance of a 'tetracotyle' type metacercarian parasite found in the eyes of sardine; these were first analysed individually and then collectively (excluding body shape data) in a multivariate analysis to test their effectiveness in discriminating between sardine from the putative western and southern sardine stocks. Analysis using geometric morphometrics revealed significant differences in body shape between sardine from the putative western and southern stocks. Results of GLM analyses indicated that gill arch length, the number of gill rakers, otolith form factor and circularity, and parasite abundance were effective univariate discriminators of sardine stocks. No significant difference in gill raker spacing and the number of vertebrae was found between individuals from the two stocks. A Stock Identification Index (SDI) of 0.75 derived from the univariate analyses, as well as results of the multivariate analysis of data provided strong evidence for the existence of two mixing stocks, therefore, supporting the two-stock hypothesis. The five multivariate classification models used in the study showed varying degree of allocation success. Overall classification accuracy ranged from a low 47% in the Linear Discriminant Analysis model to highs of 82% and 91% in the Classification Tree Analysis and Random Forest models, respectively. Both CTA and RF revealed the combination of variables with the strongest spatial discriminatory power to be the number of vertebrae and abundance of the 'tetracotyle' type metacercarian parasite. These results agree with those of previous studies and further support the inclusion of sardine population structure into management strategies for the purse-seine fishery for South African sardine. Given the importance of the 'tetracotyle' type metacercaria in the discrimination of South African western and southern sardine stocks, further studies were undertaken to identify and describe this parasite using light and scanning microscopy, as well as molecular tools. Such information will assist in the definitive identification of the first intermediate host of the parasite, presently hypothesized to be a sub-tidal gastropod occurring off the west coast only. Definitive identification of the first intermediate host (or hosts) will enable the full life cycle of this parasite to be determined and the parasite endemic area to be identified; a critical knowledge gap in application of the parasite biotag approach to stock discrimination using this 'tetracotyle' type metacercarian parasite. Metacercariae collected from fresh sardine sampled from a commercial landing in Gans Baai, South Africa, were manually excysted, relaxed in warm water, fixed in 70% ethanol and stained with haematoxylin for light microscopy. The metacercarial body is oval-shaped, measuring 762 - 967 x 512 - 677μm. It is divided by transverse folds into a forebody, midbody and hindbody. Diagnostic features include the unique, large excretory bladder lobes situated on the lateral sides of the body; two large pseudosuckers in the anterior part of the midbody; an acetabulum which is larger than the oral sucker; and a large lobulated holdfast organ in the posterior half of the midbody. These features are those of the metacercariae of the genus Cardiocephaloides, confirming the previous hypothesis regarding the genus of this digenean biotag. Analysis of partial 28S rDNA region sequence data showed that the metacercariae and the adult Cardiocephaloides found in the African penguin Spheniscus demersus are of the same species, likely C. physalis. This study is the first documentation of the morphological and molecular characterization of the stock-discriminating Cardiocephaloides metacercaria found in the eyes of Sardinops sagax in South Africa.