A few years ago my fascination with the internet started when i produced my own fisheries management website. Being a fisheries scientist at heart it was based mainly around capture fisheries and inland fisheries consultancy that at the time was my vocation. I wrote many pieces that are still on the net and viewed by around 70000 people a year. This piece, however, is worthy of a re-load as its relevance is still current. I apologise if it gets a little scientific in places, but please bear with it as it will give a great insight into just how difficult it is to calculate stock levels and issue quotas.
Accurate stock assessment is of growing importance as the human population and demand for fish increases. Continued technological advances in fishing fleets increase efficiency directly effecting natural fish stocks. Attempting to match natural stock fluctuation with fishing effort may help to avoid any further long term damage of exploited species; this is of great importance as fish provide vital contributions to food supplies and influence employment in coastal areas.
Various methods are applied to calculate estimates of recruitment, stock sizes, and age groups. It is apparent that stock assessment techniques are highly dependent on available data, whether long or short-term predictions are the aim, both strengths and weaknesses are influenced by the abundance of this information. For correct predictions many techniques require large inputs of unbiased data, therefore the strength of any stock biomass prediction will be influenced by the weakness of the available inputs; validating final modal estimates of a fishery.
Rose (1997) offers another view for these problems, indicating that fisheries scientists have lost track of their science by becoming 'keyboard ecologists' whom rarely, if ever work directly with real fisheries. This of course does not reflect on the collection of fisheries data but more the interpretation and wisdom required to gain results. This diversion may lead to incorrect long-term analysis; potentially undermining fishery techniques, therefore it is crucial that all stakeholders in a fishery increase an understanding and trust in stock assessment procedures (Anon 1998).
Cortes's (1998) study of shark stock assessment concluded no accurate results could be gained without increased collections of biological and fishery data, coinciding with a better understanding of stock recruitment relationships. Many fisheries are in dire straits due to data collection leaving them with a retrospective problem for stocks. This is pinpointed by Mohn (1999), who studied data on the East Scotian Shelf cod fishery. He concludes that failure to correct the problems encountered by traditional analysis techniques, leads to catch level advice twice or more the intended level. Stock and recruitment data sets should not be published or used unless estimates of error variance are shown; without this information Walters & Ludwig (1981) believe they are meaningless and misleading.
Data types can be split into two groups, dependent or independent of the fishery. Fishery dependent data comprises of four usable types, the total catch, amount of fishing, (the combination known as) catch per unit effort (CPUE) and age or size compositions.
Catch data is essential for most stock production models, inaccurate or biased collection can have damaging long term effects. Importantly, this data should be totalled over ages, fleets, and nations with longer term information helping to predict the past life of the fishery.
Problems arise with data collection as some fleet members find financial rewards in discarding initial catches, searching for larger or higher valued cohorts. When total catches are calculated these discarded fish that obviously made up part of the initial stock are rarely accounted for. The result can be higher biomass predictions, thus allocation of total allowable catches are higher than the available stock. This incorrect estimation of stocks can result in either continued over exploitation or economic hardship when quotas are cut. Some models include discard data, but as Mesnil (1996) points out, the general assumption is that all discards die even though there is proof that in some fisheries (usually shellfish) a significant fraction are able to survive. Either way the incorrect analysis of discarded fish will result in wrong estimates of the fishery.
The 'amount of fishing' or 'effort data' on its own is less important to fishery scientists unlike economists who study activity trends. The collection of days fished at sea and number of boats operating in conjunction with the previously mentioned catch data is much more valuable. This information known as catch per unit effort (CPUE), if based on age and composition, can be a very important factor in fisheries modelling. Effort based production models (PM) use this catch effort data with catches recorded in weight, in an attempt to estimate parameters as a stock production curve. They also assume that effort is closely related to fishing mortality (Kimura et al. 1984).
When writers such as Roff (1983) suggest the catch/effort data is only reliable to detect major fluctuations in population size and "attempts to determine equilibrium yields from catch/effort data are as likely to be successful as finding the pot of gold at the end of the rainbow", doubts over the strengths of models using CPUE data must arise. Collie & Sissenwine (1983) were more liberal in their views but still express difficulties in the standardising of CPUE data from commercial and recreational fisheries. They pinpoint the of 'constant catchability coefficients' and continued 'technical improvements' as the main areas for biased data.
Many models rely on accurate age, length, and age-length calculations of a stock. The difficult nature of collecting this data probably has the most influence hence the most accurate sources are usually from survey vessels. If capable industry collects this data, but usually samples are taken from the catch when landed; the two usual techniques of ageing these fish are via scale and otolith readings. Otolith ageing is less adaptable to a fishery wide sampling program than scale readings, due to the difficult and time-consuming nature of collection. Otolith morphology has also been shown to be an effective tool for stock discrimination in certain species (Freidland and Reddin 1994).
Independent data collection via fishing surveys helps limitations apparent from actual fishery dependent data. As with commercial methods, variable catches and weather conditions affect surveys; technical influences such as mesh size and vessel efficiency may not coincide with actual fleet averages. Changes in vessel efficiency or shifts in effort may not accurately reflect trends of abundance or fishing mortality. Therefore, in the determining of age structures, growth, mortality rates and historical trends, survey techniques may provide the only basis of data collection (Clark 1979). Simplistic assumptions that areas have been swept clean of fish, and common assumptions that trawls are giving unbiased samples both as to species and size of the local demersal fish abundance, may prove inaccurate and damaging in the long-term.
Acoustic methods can be used to either estimate population sizes of pelagic species directly or in conjunction with survey vessels when beam trawling for demersal fish. Engas & Vold Soldal (1992) believe trawl catch rates cannot be relied upon to provide representative estimates and any bias will therefore affect the equivalent acoustic estimates. This may be due to unrealistic requirements such as the confinement of a stock in an area small enough to be surveyed in a set time, at a required intensity, in mid-water not to close to the shore (MacLennan & Forbes 1987). Further assumptions are limited numbers of other species detected, the target strength of the species is accurately known, acceptable weather conditions and no response from the fish to the vessel. Fish densities estimated by horizontal beaming can be up to fifty times higher than vertical beaming due to boat avoidance creating large errors in final data (Kubecka & Wittingerova 1998). Although much interference may be apparent, Pope (1982) still believes the data gained may be valuable when setting precautionary catch levels.
Methods of tagging can be applied to gain imprecise levels of natural mortality (Shepherd 1988), main weaknesses being the sometimes over-expectant assumptions that need to be made. A fixed population with an equal capture rate and no change in catchability level, coinciding with no loss of marks or tags, all seems a little un-realistic. Peterson's closed population method cannot even test these assumptions whereas the Scnabel open population system can, but still with uncertainties. These methods are applied throughout the world indicating that they must work with certain species under perfect conditions. Uncertainties, as with many other techniques do not seem unique, but the methodology of tagging does seem to have greater assumptions than any other method applied to fisheries.
The choice of assessment type will depend on the biology of the species, the time scale required, the area and purpose of the assessment and any specific goal of the fisheries manager. This choice may be difficult as stock production models used for long-term management are frequently no better in the forecasting of the following years CPUE than is the previous years CPUE (Stocker & Hilborn 1981). Long-term assessments aid strategic decisions by managers, information such as maximum sustainable yield (MSY) can be estimated and relationships between stock and recruits can be found. Short-term assessments can reveal information on the likely catch in the next or following years (CPUE), as well as consequences of recruitment in the near future. The latter relates to the suggestion and tactics of long-term strategy. Describing of uncertainties in these strategies is of great importance to managers when weighing the benefits and losses of different techniques. Rosenberg & Restrepo (1994) suggest methods of analysing and assessing risk in management strategies implying that every possible analysis of risk should be undertaken.
Hilborn (1992) pinpoints three dominant approaches to fisheries stock assessment: the investigation into catch at age data, uninvolved models of biomass dynamics, and examination of length-frequency data. He suggests that these methods ignore what is known about the biology of the fish and tend to rely on single types of data. This point is of importance as natural mortality, assumed in many modals may increase via predation or reduced food sources causing large errors in calculations.
Age based methods such as Virtual population analysis (VPA), require catches recorded in numbers at age on an individual cohort basis to solve the exponential form of the catch equations (Kimura et al. 1984). The dependency of knowing the catch at age in numbers is a downfall as age data is costly and technically difficult to obtain. VPA or simpler cohort analysis needs data from various other sources, any of which could be bias or incorrect.
Catch in weight, natural and fishing mortality, weights at cohort, and proportion of mature fish are all required for cohort analysis. Although these methods are the most commonly applied to stock assessment, the large variety of necessary information will have any final say on the weakness of this technique. Interestingly, Agnew et al. (1998) believes that cohorts of certain species have differing dynamics, and therefore should be considered as different stocks. This would render total stock calculation models redundant, with very few other options available to fishery scientists this opinion seems to be alone.
Some typical problems arising with these methods includes the missing of year data, changes in survey techniques and age determination methodologies; Richards et al. (1997) suggests some graphical techniques to portray these uncertainties. The errors encountered in age structure data can to some extent be cancelled by using mean age calculations in the assessment models (Richards & Schnute 1998). These of course are statistical problems that may be lost in complicated calculations. Important and essentially undetectable problems arise with discard levels, the guessing of terminal fishing mortality, and predation mortality (Christensen 1996). The statistical problems can be corrected with the application of more accurate data collection, but these biological influences need highly intensive studies before a complete understanding can be hoped for. The effects of various percentage errors in the population of a year class, due to incorrect values of fishing mortality are shown in figure 1.
Figure 1: This graph plots percentage error of Ni (population of year class at the ith birthday), against cumulative fishing mortality. The under estimation of Ft (fishing mortality at the last age of a year class to which catch data are available) will result in guesses of Nt that are to small, overestimation has the reverse effect. Interestingly, as the cumulative fishing mortality increases errors in both Ft and Ni decrease. If the cumulative fishing mortality is greater than 2 and Ft can be estimated within the given range many users will find errors in Ni and Fi small enough to work with. Accurate estimations of Ni and Fi require careful choices of Ft if the cumulative fishing mortality is small. This case may arise when numbers of recruits to a year class is guessed from catches of partially recruited age groups. Similar graphs allow fishery scientists to produce the error range of their calculations that will aid assessment of their value. Source: Pope (1983)
When age data is sparse or the species cannot easily be aged, length based assessments are an alternative. Chen's (1997) comparison between age and length structured yield-per-recruit models showed length structured techniques better incorporated information observed from fisheries, but age structured gave more precise and conservative estimates of yield-per-recruit. This is the main reason why age structured models are chosen from the conservation perspective in fisheries management. The obvious difference between age and length is that age is a linear measure of time whereas length is non-linear. This makes data interpretations more difficult, more assumptions of growth reductions due to age must be made. Assumptions removed from a model increase accuracy, this is why age methods are preferred if feasible.
A potential strength of fishery science will be the adoption of multi-species models to fisheries that currently utilise single species methods. These models, although essential for future management purposes, seem unreliable and more imprecise than the currently used methods. They require more data that could lead to inaccurate assumptions, thus leaving fisheries in a worse state.
The key area that multi-species models address is predation. It is often assumed that fishing mortality alone is responsible for the variation in fish survival, but in some fisheries, losses to predation can exceed losses to fisheries (Bax 1998). This could indicate that assumptions of natural mortality in single species models are drastically misleading. Mertz & Myers (1997) point out that if bad estimations of natural mortality are used in calculations of cohort strength derived from catch data, the accuracy may be greatly corrupted. Pereiro (1995) supports that where species are not linked to a specific substratum natural mortality will always predominate over fishing mortality thus fishing mortality is not the subsidiary factor. Either way the addition of accurate natural mortality estimations into models must be welcomed.
This review has shown some major problems encountered when estimating populations from a fishery. Strengths seem sparse, maybe the biggest being that these techniques are the only available methods for estimating stock dynamics. Assessment techniques have strengths over each other and it is imperative the correct method is paired to its purpose.
Weaknesses seemed over bearing and many writers have tried to remove errors from previously presented models resulting in a claim that theirs is now the most accurate. Until data collection methods have improved there will always be inaccuracies in results. The addition of computer programs should aid time-consuming calculations allowing scientists to return to the field of study to uncover new methods of improving the currently used stock assessment techniques.
Anon. (1998). Improving fish stock assessments: Report of the committee on fish stock assessment methods. Ocean studies board, Commission on geosciences, environment, and resources, National research council. http/www.fishingnj.org/artasess.htm [on-line].
Agnew, D.J., Baranowski, R., Beddington, J.R., desClers, S., and Nolan, C.P. (1998). Approaches to assening stocks of Loligo gahi around the Falkland Islands. Fisheries Research, 35, 3, 155-169.
Bax, N.J. (1998). The significance and prediction of predation in marine fisheries. ICES J. Mar. Sci., 55, 6, 997-1030.
Chen, Y. (1997). A comparison study of age- and length-structured yield-per-recruit modals. Aquatic living resources, 10, 5, 271-280.
Christensen, V. (1996). Virtual population reality. Reviews in fish biology and fisheries, 6, 243-247. Clark, S.H. (1979). Application of bottom-trawl survey data to fish stock assessment. Fisheries, 4, 3, 9-15.
Collie, S.J. and Sissenwine, M.P. (1983). Estimating population size from relative abundance data measured with error. Can. J. Fish. Aquat. Sci., 40, 11, 1871-1879.
Cortes, E. (1998). Demographic analysis as an aid in shark stock assessment and management. Fisheries Research, 39, 2, 199-208.
Engas, A. and Vold Soldal, A. (1992). Diurnal variations in bottom rawl catches of cod and haddock and their influence on abundance indicies. ICES J. Mar. Sci., 49, 89-95.
Freidland, K.D. and Reddin, D.G. (1994). Use of otolith morphology in stock discriminations of Atlantic salmon. Can. J. Fish. Aquat. Sci., 51, 1, 91-98.
Hilborn, R. (1992). Current and future trends in fisheries stock assessment and management. South African journal of marine science, 12, 975-988.
Kimura, D.K., Balsiger, J.W., and Ito, D.H. (1984). Generalized stock reduction analysis. Can. J. Fish. Aquat. Sci., 41, 9, 1325-1333.
Kubecka, R. and Wittingerova, M. (1998). Horizontal beaming as a crucial component of acoustic fish assessment in freshwater reservoirs. Fisheries Research, 35, 1-2, 99-106.
MacLennan, D.N. and Forbes, S.T. (1987). Acoustic methods for fish stock estimation. In Bailey, R.S. and Parrish, B.B. (Eds.). Developments in fisheries research in Scotland, pp 40-55. Fishing News Books Ltd, Farnham.
Mertz, G. and Myers, R.A. (1997). Influence of errors in natural mortality estimates in cohort analysis. Can. J. Fish. Aquat. Sci., 54, 7, 1608-1612.
Mesnil, B. (1996). When discards survive: Accounting for survival of discards in fisheries assessments. Aquatic living services, 9, 3, 209-215.
Mohn, R. (1999). The retrospective problem in sequential population analysis: An investigation using cod fishery and simulated data. ICES J. Mar. Sci., 56, 4, 473-488.
Pereiro, J.A. (1995). Assessment and management of fish populations - A critical view. Scientia Marina, 59, 3-4, 653-660.
Pope, J.G. (1982). Background to scientific advice on fisheries management. MAFF Directorate of fisheies research labortory leaflet, 54, 27 pp.
Pope, J.G. (1983). An investigaton of the accuracy of virtual population analysis using cohort analysis. In Cushing, D.H. (Ed.). Key papers on fish populations, pp. 292-301. IRL Press, Oxford.
Richards, L.J. and Schnute, J.T. (1998). Modal complexity and catch-age analysis. Can. J. Fish. Aquat. Sci., 55, 4, 949-957.
Richards, L.J., Schnute, J.T., and Olsen, N. (1997). Visualizing catch-age analysis: a case study. Can. J. Fish. Aquat. Sci., 54, 7, 1646-1658.
Roff, D.A. (1983). Analysis of catch/effort data: A comparison of three methods. Can. J. Fish. Aquat. Sci., 40, 9, 1496-1506.
Rose, G.A. (1997). The trouble with fisheries science. Reviews in fish biology and fisheries, 7, 365-370.
Rosenberg, A.A. and Restrepo, V.R. (1994). Uncertainty and risk-evaluation in stock assessment advice for US marine fisheries. Can. J. Fish. Aquat. Sci., 51, 12, 2715-2720.
Shepherd, J.G. (1988). Fish stock assessments and their data requirements. In Gulland, J.A. (Ed.). Fish population dynamics, pp. 35-62. John Wiley & Sons Ltd, Chichester.
Stocker, M. and Hilborn, R. (1981). Short-term forcasting in marine fish stocks. Can. J. Fish. Aquat. Sci., 38, 1247-1254.
Walters, C.J. and Ludwig, D. (1981). Effects of measurement errors on the assessment of stock-recruitment relationships. Can. J. Fish. Aquat. Sci., 38, 704-710.