The value of catch statistics and records of Guano Harvests for managing certain South African Fisheries

Doctoral Thesis


Permanent link to this Item
Journal Title
Link to Journal
Journal ISSN
Volume Title
Research into species inhabiting Benguela current region has been the southern part biased towards of those the of commercial importance and accessible than others. A those which are either more visible or crude evaluation of material flows and standing stocks, including the lesser studied species, was made. Estimates suggest that the productivity of juvenile hake C3 million metric tonnes per year or more, squid(> 0. 5 million metric tonnes per yearly, and midwater shoaling fish (together > 0. 5 million metric tonnes per year) is substantial compared to the total annual landing of marine fish off South Africa which is in the region of 0. 5 million metric tonnes. The consumption of anchovy and other epi-pelagic shoaling fish is dominated by piscivorous fish, particularly snoek. Snoek may consume as much as 0. 4 million metric tonnes of anchovy per year, out of a total anchovy production of in the region of 2 million metric .tonnes wet mass per year. Anchovies are the main fish consumers of plankton, with an estimated consumption of more than 20 million metric tonnes wet mass per year. Total annual plankton production is in excess of 800 million metric tonnes wet mass. About 32% of this enters higher trophic levels in the epi-pelagic region, mainly through zooplankton grazing. It is suspected that the bulk of production in the zooplankton and hake communities is recycled by omnivorous feeding in the case of zooplankton and cannibalism in the case of hake. The amount of cannibalised hake is about 10 times the amount taken by trawlers. The biomass, production, and consumption rates of apex predators such as seals, whales, dolphins and seabirds (sizes of the order of metres to 10' s of metres) is quantitatively unimportant in comparison with species consisting of smaller individuals (sizes of the order of 10's of centimetres). Due to the complexity of the food web, a conceptual model based on size may be preferable to one involving trophic levels. Overall, however, the complexity of the system, and the sketchiness of knowledge about it mitigates against the use of multispecies models for making useful predictions. A maximum likelihood estimation approach to Virtual Population ·Analysis ( VPAl is adopted, developed, and implemented. It is shown from results obtained using artificial fisheries data that the database conventionally considered adequate for VPA work, viz. the catch mass and age structure, does not lead to reliable estimates of stock abundances for recent years. Use of further data related to stock adundance, such as effort data in situations where catch per unit effort is proportional to abundance, is investigated. Results using simulated effort data show dramatic improvements in abundance estimator precision~ It is therefore argued that VPA should not be attempted in the absence of data complementary to the catch masses and age structures. With this in mind, VPA is applied to three stocks of commercial importance which occur off the South African coast 1 i ne: anchovy, pi 1 chard and hake. Results are presented which show that anchovy biomasses are determined with reasonable precision if catch rates can be assumed to be proportional to stock abundance, and if the selectivity functional form is known. However, it is argued that it is unlikely that catch rates are proportional to anchovy abundances. Furthermore, recent surveys of the anchovy stock size and habitat range suggest that the relationship between age and selectivity is different to that used in the analysis and is essentially unknown. Accordingly further analyses are performed, excluding the _ effort data, and replacing the initial one parameter selectivity function with unique selectivity parameters for each age class seen in the catch. Very imprecise biomass estimates are obtained when the likelihood function is formulated in this way, due to the now enhanced instability of the position of the maximum of the VPA likelihood function. Separate simulations illustrate that the bias in abundance estimates introduced by using the same age length key in all years for converting catch length distributions to catch age distributions may be as much as 50%. This, the aforementioned imprecision of VPA biomass estimates, unassessed but probable errors in the single age length key, which is used, and finally very recent ageing re-evaluation studies suggesting that the catch consists almost entirely of 0-year-olds, effectively rule out the possibility of obtaining management-orientated information from the anchovy VPA. Complementary data for the pilchard VPA is obtained by analysis of records of guano harvests which are made at a number of islands off the South African coastline. Evidence is presented suggesting that the harvest at Bird Island, Lambert• s Bay is approximately linearly related to pilchard recruitment and these guano harvests are thus included in the likelihood function for the pilchard VPA. The resultant biomass estimates are plausible in terms of the likely system carrying capacity for pelagic fish. The biomass trend does not reflect the major peak in biomass in 1960 which earlier work consistently suggested. An appropriate sum of squares _minimisation procedure is adapted from the LudwigWal ters' formulation for fitting the pilchard stock recruit function to the recruitment and spawning biomass data from the VPA. The surplus production curve is estimated by combining the estimated recruitment function parameters with a yield per recruit analysis. The results indicate a mean virgin spawning biomass of 2. 4 million metric tonnes, and an MSY spawning biomass level of 1. 0 million metric tonnes. The maximum sustainable annual catch, for a 20 -year scenario starting with the population at half the unexploited biomass and ensuring a 10% maximum risk of obtaining a stock size of less than 0.2 of the pristine size, is 187 thousand metric tonnes. The surplus production curve for the hake stock off the west coast of South Africa C ICSEAF Div. 1. 6> is estimated using the same methods as applied to pilchard. No substantial difference between the VPA results and those produced using dynamic catch effort models could be demonstrated - both VPA and the Schaefer model predict a maximum sustainable yield of about 140 thousand metric tonnes. Closer analysis shows however that the additional age structure information taken into account by the VPA has not led to better estimates than are provided by the catch mass and effort data alone. The suspected past practice of discarding small hake at sea potentially invalidates all sustainable yield calculations, even those based on the dynamic catch-effort models, by introducing spurious effects into the catch-based statistical records. The future in hake stock assessment may therefore lie with Casis the case for the pelagic species> direct survey techniques.