Knowledge-based systems as decision support tools in an ecosystem approach to fisheries: Comparing a fuzzy-logic and a rule-based approach

In an ecosystem approach to fisheries (EAF), management must draw on information of widely different types, and information addressing various scales. Knowledge-based systems assist in the decision-making process by summarising this information in a logical, transparent and reproducible way. Both rule-based Boolean and fuzzy-logic models have been used successfully as knowledge-based decision support tools. This study compares two such systems relevant to fisheries management in an EAF developed for the southern Benguela. The first is a rule-based system for the prediction of anchovy recruitment and the second is a fuzzy-logic tool to monitor implementation of an EAF in the sardine fishery. We construct a fuzzy-logic counterpart to the rule-based model, and a rule-based counterpart to the fuzzy-logic model, compare their results, and include feedback from potential users of these two decision support tools in our evaluation of the two approaches. With respect to the model objectives, no method clearly outperformed the other. The advantages of numerically processing continuous variables, and interpreting the final output, as in fuzzy-logic models, can be weighed up against the advantages of using a few, qualitative, easy-to-understand categories as in rule-based models. The natural language used in rule-based implementations is easily understood by, and communicated among, users of these systems. Users unfamiliar with fuzzy-set theory must “trust” the logic of the model. Graphical visualization of intermediate and end results is an important advantage of any system. Applying the two approaches in parallel improved our understanding of the model as well as of the underlying problems. Even for complex problems, small knowledge-based systems such as the ones explored here are worth developing and using. Their strengths lie in (i) synthesis of the problem in a logical and transparent framework, (ii) helping scientists to deliberate how to apply their science to transdisciplinary issues that are not purely scientific, and (iii) representing vehicles for delivering state-of-the-art science to those who need to use it. Possible applications of this approach for ecosystems of the Humboldt Current are discussed.     

An analysis of yield variability from three harvesting strategies in the South African anchovy fishery,under conditions of randomly fluctuating recruitment success

An age-structured population model was used to examine the variability of yield created by randomly fluctuating recruitment success in the South African anchovy Engraulis capensis stock. Three different harvesting strategies were examined: (1) constant age-specific pattern of fishing mortality, (2) constant quota with effort limitation and (3) annual quota adjustments by means of an F _msy procedure. Variable recruitment was generated by a stochastic stock-recruit relationship, and mean yields, mean spawner biomasses and probability distributions of yield were calculated at each exploitation level after 150 projections from the population structure of anchovy in 1981. Under conditions of constant fishing mortality, the variability of yield increased continuously as the exploitation rate was increased. The maximum average yield (MAY) is the stochastic equivalent of the maximum sustainable yield (MSY) of deterministic models, but an attempt to sustain this MAY by a constant quota resulted in a very high risk of stock collapse. An Fmsy policy based on pre-season adult biomass resulted in more variable yields than were obtained at the equivalent constant fishing mortality at age. Stability of yield therefore demanded the acceptance of lower average yields than could be attained in the long term by quota adjustment. It was considered that the South African purse-seine industry could not cope with the wide fluctuations of yield necessary to attain MAY. Specific management policies sufficiently robust to withstand both fluctuating recruitment success and the uncertainties in the parameters of working population models would be required.     

Modelling the effect of food availability on recruitment success of Cape anchovy ichthyoplankton in the southern Benguela upwelling system

Cape anchovy Engraulis encrasicolus adapted its reproductive strategies to the southern Benguela system by spawning over the Agulhas Bank, an area of low productivity that is located upstream of the predominant upwelling system. Frontal jet currents transport eggs and larvae toward the west coast of South Africa, where recruitment takes place. To characterise the recruitment dynamics of Cape anchovy ichthyoplankton, we used an individual-based model forced by a coupled hydrodynamic–biogeochemical model. The results show the importance of food (especially diatoms and copepods) dynamics on the spatial and temporal patterns of recruitment success, and also confirm the importance of the spawning area, timing and water depth on the recruitment success of Cape anchovy larvae.     

Revised estimates of abundance of South African sardine and anchovy from acoustic surveys adjusting for echosounder saturation in earlier surveys and attenuation effects for sardine

Hydro-acoustic surveys have been used to provide annual estimates of May recruitment and November spawner biomass of the South African sardine Sardinops sagax and anchovy Engraulis encrasicolus resources since 1984. These time-series of abundance estimates form the backbone of the assessment of these resources, and consequently the management of the South African sardine and anchovy is critically dependent on them. Upgrades to survey equipment over time have resulted in recent surveys providing more accurate estimates of abundance, yet in order to maintain comparability across the full time-series, estimates of biomass mimicking the old equipment were used for a number of years. In this paper we develop a method to revise the earlier part of the time-series to correct for receiver saturation in the older generation SIMRAD EK400 and EKS-38 echo sounders and to account for attenuation in dense sardine schools. This is applied to provide a revised time-series of biomass estimates for the South African sardine and anchovy resources with associated variance–covariance matrices. Furthermore, the time-series presented here are based on updated acoustic target strength estimates, making this the most reliable time-series currently available for both resources.     

Top predators in the Benguela ecosystem — implications of their trophic position

Many top predators in the Benguela ecosystem feed on prey species targeted by commercial fisheries. Their roles as indicators of the state of exploited prey resources, as competitors with commercial fisheries for resources, and as susceptible to impact from commercial fishing on those resources are briefly considered. Trends in the occurrence of anchovy Engraulis capensis and pilchard Sardinops ocellatus in the diet of Cape gannets Morus capensis off South Africa's west coast are related significantly to survey estimates of the abundance of these fish species, and they provide useful confirmation of those estimates. In the 1980s, anchovy decreased in the diet of Cape gannets, but pilchard increased. In both the northern and southern portions of the Benguela system, groundfish were thought to eat most (66–73%) of the total quantity of cephalopods and vertebrates consumed by predators and man in the 1980s. South African fur seals Arctocephalus pusillus pusillus, predatory pelagic fish and man removed roughly equal amounts, with squids, seabirds and cetaceans having a smaller impact. In the 1980s, man and seals removed about two million tons live mass more than in 1930. Indices of the rate of natural mortality of anchovy and pilchard attributable to Cape gannets are not related to biomass of the prey species. That for anchovy was high in 1989 when a poor anchovy year-class was formed. Decreased abundance of anchovy led to poor breeding by Cape cormorants Phalacrocorax capensis in 1989 and 1990. A model linking the Cape cormorant population with anchovy is used to explore the impact of possible exploitation strategies for anchovy on Cape cormorants.     

Assessment and management of South African marine resources during the period of the Benguela Ecology Programme: key lessons and future directions

Three important developments in resource assessment and management during the period of the Benguela Ecology Programme are discussed. The first is the recognition of the limited information content for assessment purposes of the available data for many of the Benguela's major pelagic and demersal resources. This has meant that management measures for these resources could attempt no more than to maintain the status quo, rather than to achieve "optimal" utilization. Secondly, a post mortem on the commercial extinction of the kingklip suggests a need to consider Bayesian methods as a means to incorporate additional information into stock assessments. Finally, changes to the management approaches for the hake and anchovy fisheries are argued to have been particular successes, notwithstanding the recent heavily reduced Total Allowable Catch (TAC) for the latter resource. "Management procedures" have been developed and simulation-tested for these two resources. The implications of this approach for the future basis, whereby TACs are set for South Africa's fisheries, are discussed.     

The role of acoustic surveys in the assessment of pelagic fish resources on the South African continental shelf

Acoustic data on the abundance and distribution of anchovy Engraulis capensis, pilchard Sardinops ocellatus and round herring Etrumeus whiteheadi on the South African continental shelf have been collected from 21 echo-integrator surveys between 1984 and 1991. Most effort has been concentrated on estimating adult biomass of anchovy and pilchard in November (spring) and anchovy recruitment in autumn. Distribution maps from all surveys are presented and the biomass estimates considered most reliable documented. A series of distribution maps tracing movements of three anchovy year-classes over a four-year period is presented to illustrate the usefulness of the surveys in migration studies. The major findings of the survey programme have been that anchovy are generally considerably more abundant and widespread than was thought to be the case prior to the surveys, that the pilchard resource has recovered substantially in recent years, and that the round herring resource, about which little was known prior to the surveys, is probably of the same order of magnitude as the anchovy resource and is probably underexploited. The anchovy and pilchard resources are currently managed through procedures based largely on the acoustic estimates of biomass and their estimated precision. The role of these estimates in the management procedures is discussed in some detail.     

The potential use of predictions of recruitment success in the management of the South African anchovy resource

In the existing management procedure for the South African anchovy Engraulis capensis resource, an initial Total Allowable Catch (TAC) is set at the start of the fishing season in January, before the year's recruitment of 0-year-old fish, an important component of the catch, is known. Recruitment is therefore assumed to be equal to the observed median. This TAC may be revised in May or June, after the actual recruitment has been estimated. This procedure incorporates a risk that, if recruitment is below the median, the stock could be adversely depleted before the mid-year recruitment survey. A simulation approach is used in this paper to assess the possible benefits, in terms of average annual catch and interannual fluctuations in catch, of three different methods of predicting recruitment earlier in the season, if such methods were available. The results show that the average annual catch could be increased theoretically by up to 48 per cent if a very precise prediction (CV = 0,1) could be made at the start of the fishing season. In practice, more modest improvements of 21 per cent by prerecruit surveys undertaken in March each year, or 16 per cent by means of a two-class ordinal prediction with 70 per cent success, may be attainable. However, these improvements would result in increased interannual fluctuations in catch, a situation which may be detrimental to the fishing industry.     

Ocean robotics in support of fisheries research and management

South Africa’s small-pelagic fishery is a socio-economically important component of the country’s commercial fisheries sector, second in value only to the demersal trawl fishery. Management of this sector relies on infrequent hydro-acoustic surveys, which provide measures of anchovy Engraulis encrasicolus and sardine Sardinops sagax biomass used in the assessments of stock status and in the development of management plans for the sustainable utilisation of these resources. We demonstrate how technological capabilities in ocean robotics at the Council for Scientific and Industrial Research (CSIR) could augment the current resource-intensive hydro-acoustic ship-based survey programme and create opportunities for expanding its spatial and temporal resolution. We successfully implement and demonstrate an autonomous wave glider, fitted with a hydro-acoustic sensor and compare the data to a collocated ‘traditional’ ship-based acoustics survey. In the future these autonomous systems approaches could be seen as a means to lessen the cost burden of the ship-based survey, while at the same time with the added advantage of continuous collection over much wider spatial and temporal domains. This could enable a more reflexive stock management approach taking into account the seasonal characteristics of the fishery and its ecosystem. Gliders thus have potential to increase dramatically the quantity of information available to fisheries managers, thereby reducing uncertainty and contributing to improved management of valuable fish resources. They are likely to contribute to improved knowledge of the ecology of small pelagic fish species off the coast of South Africa in a changing climate and should potentially also permit the collection of biomass data for other marine resources currently not routinely monitored.     

Towards rational harvesting of the South African anchovy considering survey imprecision and recruitment variability

Alternative harvesting schemes for the South African anchovy resource are considered. For this preliminary study, strategies in which quotas are simply related to the pre-catch survey estimates of spawning biomass and recruitment are investigated, using a Monte Carlo simulation approach. The schemes evaluated are the strategies of a constant catch, a constant proportion of the recruitment estimate, and a variant on constant escapement. The relationship between modal recruitment and spawning biomass is assumed known, and the values of parameters such as natural mortality and mass-at-age are initially taken to be exact. True recruitment is assumed log-normally distributed about its modal value. The annual survey estimates of recruitment and spawning biomass are generated with typical measurement errors included. The results show the cost (expressed as a loss of potential annual yield) of imperfect survey information, natural recruitment variability, and constraints on year-to-year catch variability. The analysis is extended to integrate the outcome of specific harvesting strategies over the current estimates of the probability distributions for the model parameters, assuming that these are not updated as harvesting is continued. The final results suggest that the mean annual yield of the South African anchovy resource could be increased by about 40 per cent through use of a more flexible harvesting strategy than is the current practice, without at the same time increasing the risk of collapse.     

A decade of illegal fishing in Table Mountain National Park (2000–2009): trends in the illicit harvest of abalone Haliotis midae and West Coast rock lobster Jasus lalandii

Illegal fishing activities are reported to be on the increase in South Africa, including in its marine protected areas (MPAs). Research is presented on the nature and the scale of illegal fishing in Table Mountain National Park (TMNP) by analysing the numbers of abalone Haliotis midae and West Coast rock lobster Jasus lalandii confiscated from fishers operating in the park's marine protected area between 2000 and 2009. Data were collected from offence logbooks maintained by South African National Parks rangers and managers, the South African Police Services, and interviews with alleged or self-confessed illegal fishers. The research findings indicate that the annual numbers of illegally fished abalone and rock lobsters have increased significantly over time. Spatial analysis suggests that confiscations of abalone occur predominantly on the east coast of the park, whereas higher confiscations of illegally fished rock lobsters occur on the west coast. It is clear from this research that new and more efficient approaches will need to be designed and implemented to minimise illegal fishing in the TMNP MPA. Context-specific conservation targets that acknowledge and integrate social as well as developmental needs are required, and may be essential for limiting biodiversity loss in the longer term, which will ultimately ensure the success of fisheries management and conservation in TMNP.     

Evidence of deep genetic divergence between populations of an important recreational fishery species,Lichia amia L. 1758, around southern Africa

Lichia amia is an important coastal recreational fishery species with a cosmopolitan distribution in the eastern Atlantic. In southern Africa, it is distributed from southern Angola to northern KwaZulu-Natal in South Africa. A recent biological survey revealed differences between Angolan and South African individuals suggesting that they may represent separate stocks. As fishery management decisions should be based on accurate knowledge of population structure, an initial survey of the genetic substructuring of L. amia was conducted on samples collected in southern Angola and South Africa. Assessment of genetic diversity and population structure was conducted using a fragment of the mitochondrial Control Region. Obtained genetic diversity levels were within the expected range for marine teleosts (h = 0.867, π = 0.007), and two deeply divergent (southern Angolan and South African) populations were identified (average φ _ST = 0.78) across the Benguela Current system. These results suggest that Angolan and South African L. amia should be managed as two independent stocks. This work represents the first assessment of population genetic substructuring across the Benguela Current of a valuable coastal fishery resource, and has application for future fisheries management.     

Are African penguins worth saving? The ecotourism value of the Boulders Beach colony

Populations of the African penguin Spheniscus demersus have decreased dramatically over the past century, due in part to competition for food with commercial fisheries, and the species is now endangered as a result. Economic arguments are used to favour fisheries over the needs of penguins, but penguins have direct value to the South African economy thanks to penguin-based tourism at several breeding colonies. We estimated the value of African penguins at the most important tourist site for African penguins: Boulders on the Cape Peninsula, South Africa. As a mainland colony close to Cape Town, Boulders is accessible to large numbers of visitors; gate revenues in 2009/2010 alone were R14.5 million (US$2 million). A zonal travel-cost analysis revealed an average consumer surplus among Cape Town residents of some R20 per visit. Penguin-based tourism thus forms an integral part of the R25 billion Western Cape tourism sector. Given that the future survival of this valuable resource is dependent, among others, on the availability of sufficient prey, competition with South Africa's purse-seine fishing fleet should be limited through management strategies that lessen the potential effects of fishing on prey available at the local scale.     

Analysis of bycatch in the South African midwater trawl fishery for horse mackerel Trachurus capensis based on observer data

The South African midwater trawl fishery targets adult horse mackerel Trachurus capensis. The bulk of the catch is taken by a single freezer-trawler, the biggest fishing vessel operating in South African waters. As fishing takes place off the south coast in ecologically sensitive areas, there are concerns about the potential impacts of this fishing operation on non-target species. Fishing behaviour and bycatch of this fishery from 2004 to 2014 were investigated by analysing observer records with regard to catch composition, volume and temporal and spatial patterns. The midwater trawl fishery was estimated to have caught 25 415 tonnes annually, with a bycatch of 6.9% of the total catch, by weight. There are species overlaps with various fisheries, namely the demersal trawl, small-pelagic, line, shark longline and squid fisheries, yet the total bycatch estimates from this fishery are generally small relative to catches taken in the target fisheries. Bycatch species with the highest average annual catches were chub mackerel Scomber japonicus, redeye roundherring Etrumeus whiteheadi, ribbonfish Lepidopus caudatus and hake Merluccius spp. Large-fauna bycatch species included sunfish Mola mola as well as a number of CITES II- and IUCN-listed species, such as Cape fur seal Arctocephalus pusillus, dusky shark Carcharhinus obscurus, smooth hammerhead shark Sphyrna zygaena and thresher sharks Alopias spp. The 97.9% observer coverage is high and the 6.9% bycatch rate low compared to other South African fisheries; however, due to the large size of the individual hauls (average of 46.3 t), the average sampling rate of 1.56% is low. Our analyses suggest that bycatch in the South African midwater trawl fishery has been lower than in other South African fisheries and similar fisheries elsewhere, but due to the combination of high catch volumes and low sampling rates, estimation errors for rare species are high and there is a substantial risk of incidental unmonitored bycatch of rare large fauna and aggregations of small fauna. This could be mitigated by spatio-temporal management of this fishery, to avoid fishing in high-risk areas, and the introduction of an electronic monitoring programme.     

A knowledge base for management of the capital-intensive fishery for small pelagic fish off South Africa

As a contribution to South Africa's move towards an ecosystem approach to fisheries management, this study explores the existence of common perceptions about South Africa's pelagic fishery between resource users and scientists. It represents a collaborative research effort of social and natural scientists. A brief overview is given of the southern Benguela upwelling ecosystem and small pelagic fish resources, the fishery, and management of the fishery. Stakeholder knowledge and views were determined by conducting open-ended qualitative local knowledge interviews. Candidate indicators to address five major issues raised in the interviews were selected: length-at-50% maturity, total mortality, exploitation rate, proportion of bycatch, mean length of catch, and centre of gravity of catches. The indicator approach is shown to be a useful tool to manage the South African small pelagic fishery, and can be made compatible with existing management approaches. The foundation of a good adaptive fisheries management system is a data collection system that enables multi-disciplinary analysis and provides a basis on which decisions can be made.     

Diet of the South African fur seal (1974–1985) and an assessment of competition with fisheries in southern Africa

Almost 1 000 South African fur seals Arctocephalus pusillus pusillus were collected at sea (inshore and offshore) for examination of stomach contents and other biological parameters. The main prey of seals above the age of one year was teleost fish, of which at least 28 species were identified. Overall, 74,6 per cent of the diet was teleost fish, 16,7 per cent cephalopods, 4,8 per cent crustaceans such as shrimps, prawns, amphipods and isopods and 3,2 percent rock lobster. Within this overall dietary pattern there were marked regional differences. The importance of individual fish species in the diet was investigated from the numbers and the sizes of otoliths occurring in the stomachs, from which the identity and size classes of the prey species could be calculated. Again there were regional differences in the importance of different prey species. On the west coast of South Africa the most important species were anchovy and Cape hakes (each 23% of total diet), whereas on the south coast the important species were anchovy (17%), horse mackerel (14%), pilchard (12%) and Cape hake (14%). Off South West Africa the two dominant species were pelagic goby (52%) and horse mackerel (23%), anchovy and Cape hake constituting only 4 and 5 per cent respectively of total diet. The total annual consumption of anchovy and Cape hake by seals in South African waters was calculated at 125 000 and 120 000 tons respectively.     

Exploring the spatial distribution patterns of South African Cape hakes using generalised additive models

We developed delta generalised additive models (GAMs) to predict the spatial distribution of different size classes of South African hakes, Merluccius capensis and M. paradoxus, using demersal trawl survey data and geographical (latitude and longitude) and environmental features (depth, temperature, bottom dissolved oxygen and sediment type). Our approach consists of fitting, for each hake size class, two independent models, a binomial GAM and a quasi-Poisson GAM, whose predictions are then combined using the delta method. Delta GAMs were validated using an iterative cross-validation procedure, and their predictions were then employed to produce distribution maps for the southern Benguela. Delta GAM predictions confirmed existing knowledge about the spatial distribution patterns of South African hakes, and brought new insights into the factors influencing the presence/absence and abundance of these species. Our GAM approach can be used to produce distribution maps for spatially explicit ecosystem models of the southern Benguela in a rigorous and objective way. Ecosystem models are critical features of the ecosystem approach to fisheries, and distribution maps constructed using our GAM approach will enable a reliable allocation of species biomasses in spatially explicit ecosystem models, which will increase trust in the spatial overlaps and, therefore, the trophic interactions predicted by these models.     

Climate variability and pelagic fisheries in northern Chile

A time series analysis of long-term climate variability in northern Chile (18°21′–24°00′S) shows anomalies associated with the El Niño events and the longer warm period observed since 1976, followed by a cooling trend since mid 1980s. The succession of pelagic fisheries, anchovy (Engraulis ringens) and sardine (Sardinops sagax), occurring in this fishing zone was analyzed taking into account the landings, the CPUE abundance index, the fishing effort, and the environmental variables. The anchovy production model is a negative linear function of fishing effort and turbulence. For sardine, the production model is a negative linear function of fishing effort and a quadratic function of the sea surface temperature.An analysis of the relationship between recruitment, adult biomass and the environment shows that the annual recruitment of anchovy increases with turbulence intensity until wind speed reaches a value of 5.46 m s⁻¹, decreasing for higher values. For sardine, the recruitment increases with turbulence intensity until 5.63 m s⁻¹, stabilizing thereafter.It is deduced that the climatic variations associated to the El Niño events affect the abundance of coastal pelagic fishes, without forgetting the most likely effects upon its distribution and the fishing effort. However, it is the long-term variability that mainly affects the fishing activity.     

Stock assessment and management implications of horse mackerel (Trachurus japonicus) in Korean waters, based on the relationships between recruitment and the ocean environment

This study presents an example of horse mackerel (Trachurus japonicus) stock to demonstrate that marine environmental factors are important in stock assessment for the new Korean Total Allowable Catch (TAC)-based fisheries management system. The estimated survival rate (S) of horse mackerel ranged from 0.25 to 0.36. The instantaneous coefficient of natural mortality (M) was 0.48/year, and the age at first capture was 0.83 year. Annual biomass of horse mackerel in Korean waters was estimated by a biomass-based cohort analysis using annual catch in weight at age during 1965–1995. Yield-per-recruit and spawning biomass-per-recruit were estimated under various harvest strategies at F_max, F_0.1, F_30% and F_40%. A method for estimating acceptable biological catch (ABC) is proposed for dealing with the large differences in the quality and quantity of information and data available. Using recruitment of horse mackerel estimated from various spawner–recruitment relationship models combined with salinity, volume transport, and zooplankton biomass as environmental factors, the ABC under the best information available was estimated to range from 3100 to 3800 mt.     

Stock-environment recruitment analysis for Namibian Cape hake Merluccius capensis

The factors modulating recruitment success of Cape hake Merluccius capensis in Namibian waters are still unresolved. In this study, we used generalised additive models, regression tree analysis and the conventional Ricker model to examine the effect of environmental indices and spawning stock biomass (SSB) on hake recruitment success for the period 1984–2012. Results indicated that upwelling strength explained 51% of the recruitment variability, whereas SSB had no significant influence. The effect of SSB on recruitment only became significant when combined with upwelling strength, explaining 89% of the recruitment variability. SSB influenced recruitment during periods of strong upwelling. Optimal conditions for hake recruitment were associated with moderate upwelling strength. Low and high upwelling intensities reduced recruitment success. Our results are consistent with those of other studies suggesting a significant influence of environmental conditions on recruitment at a low spawning stock level. Our study highlights the importance of assessing the combined non-linear effects of both biotic and abiotic factors on hake recruitment.