2.  THE DEVELOPMENT AND DIVERSITY OF REFERENCE POINTS

2.1  MANAGEMENT OBJECTIVES AND THE CONCEPT OF A ‘REFERENCE POINT’

Reference points begin as conceptual criteria which capture in broad terms the management objective for the fishery. To implement fishery management it must be possible to convert the conceptual Reference Point into a Technical Reference point, which can be calculated or quantified on the basis of biological or economic characteristics of the fishery (Fig. 1). For example, when the objective is to maximise yield, MSY has frequently been used as a conceptual reference point. The concept of MSY has been interpreted in various ways, ranging from its strict technical meaning as the peak of the surplus production curve, or the point of maximum surplus reproduction on a stock recruitment curve, to its more literal interpretation as the maximum constant yield that can be taken year after year, as described by Sissenwine (1978) and Annala (1993).

Figure 1:   The sequence of development of conceptual and technical reference points incorporating scientific models and societal goals for fisheries management.

The objectives of fishery management are generally more diverse than a simple maximisation of yield. They often include considerations of foreign exchange, employment, contribution to disadvantaged rural areas, profit, inter alia. The concept of an overall objective that incorporates all important factors for a fishery was reflected in the 1958 United Nations Oceans Convention in Geneva, where the term “Optimal Sustainable Yield” emerged. Optimal Yield (OY) has since been variously defined as allowing for inputs of “…economic, social and biological values […] rather than being limited to maximizing net profits or maximizing sustainable yield” (Wallace 1975). Since it has no single technical definition, we do not consider OY to be a technical reference point, but a state which may result when a series of criteria are satisfied which effectively ensure that the fishery remains within a safe and productive area.

Smith et al. (1993) identify the lack of clearly defined management objectives as one of the main impediments to establishing and adhering to Reference Points. As described above there may be many societal objectives in managing a fishery, and each may correspond to the interests of a particular user group. Thus the stakeholders in a fishery need to agree on the management objectives for the fishery. In order to reach agreement on a conceptual reference point users must understand the relationships between the objectives, and the characteristics of the fishery: they must be able to appreciate the trade-offs among the various possible reference points in real, even if only relative, terms: whether expressed as fishing mortality rates, catch rates, mean fish sizes, etc. Various means of simplifying these relationships must be explored, in order to facilitate the participation of all users (e.g. Fig. 2).

Figure 2:   Some relative ranges for the fishing mortality rates corresponding to different societal objectives for marine resource use in the context of a multispecies surplus production model for reef fisheries (after Mahon 1992)

Thus far, there are very few instances in which multiple objectives have been formally incorporated into a management strategy, much less related to a single technical reference point. Healey (1984) proposed an analytical approach to determining optimum yield based on multi-attribute utility analysis. The methodology, which quantifies and weights the objectives of the users appears to be a reasonable way of making agreed upon decisions when there are multiple objectives. However, there are few instances in which we know of a multi-criterion approach to decision making being formally applied in fisheries management: two of these are the Yucatan Shelf octopus fishery (Diaz-de Leon and Seijo 1992), and a chinook salmon fishery in Alaska (Merrit and Criddle 1993).

Another conceptual reference point is the point beyond which ‘overfishing’ is said to occur. In the USA, plans require that a definition, or technical reference point be provided for overfishing for each stock. In the ICES area, this conceptual point is referred to as the Minimum Biologically Acceptable Level, MBAL, for the fishery. For implementation purposes ‘overfishing’ and MBALs must be technically defined, and this has been done in several ways as will be described later.

For the purposes of this paper, a Reference Point will be defined as a conventional value, derived from technical analysis, which represents a state of the fishery or population, and whose characteristics are believed to be useful for the management of the unit stock. Defining a reference point as a conventional value reflects that in practical terms they may frequently assume arbitrary values and are often specified without variance terms. It is worth noting here that all of the model-based reference points, and the parameters they are derived from, are only known approximately, often with a poorly-defined level of error.

2.2 THE UNDERLYING POPULATION MODELS

The technical reference points used in fishery management are largely based on biometric or econometric models, and hence on mathematical conceptualization of fish populations, and can be difficult to assimilate for the non-technical reader. Consequently, throughout this paper we will try to support our arguments with graphical summaries where possible. Annex I presents a few key concepts. Fuller treatment of the basic methods can be found in standard texts on biological and economic assessment of fisheries (e.g. Ricker 1975, Gulland 1983, Clarke 1985, Hilborn and Walters 1992).

The relationship between fishing mortality (F), stock biomass (B) and yield can provide the basis for discussion of most reference points (Fig. 3). F and B are the most basic Reference Variables. Reference Points on these variables are set using various criteria to be discussed later, e.g. the F which, if applied over a number of years, produces an average yield equivalent to MSY; the F which maximizes the average yield per recruit; the biomass which will produce a desired level of recruitment. Conventional fishery management seeks to control F or sustain B at levels which correspond to target values, using a variety of methods (Fig. 3). Fishing mortality and natural mortality (M) due to predation, disease, etc., combine to equal the total mortality rate (Z=F+M) for the population.

Figure 3:   The main population, reference and control variables used in defining biological reference points. In addition to the three primary measures of the state of an exploited population, fishing mortality rate (F), biomass, (B) and yield (Y), whose interrelationship is specified in a catch equation, other secondary measures shown may also be used as reference variables.

Other variables which directly influence, are related to, or are indicators of the basic Reference Variables, can also be used as Reference Variables. For example, fishing mortality and fishing effort (f) are related by the catchability coefficient (q), which is usually assumed to be constant, such that F = q.f. If this assumption is true, the catch rate, or catch per unit effort (CPUE) is an indicator of the population biomass (Fig. 3). In this paper all reference points will be expressed in terms of the primary variables. The current status of the stock according to any reference variable is designated by the subscript now, e.g. F_now, B_now, Y_now.

The models underlying MSY (and many of the other reference points described later) were originally equilibrium models, which implied that the yield represented by the production curve is that resulting when the corresponding standard effort has been applied for the years necessary for equilibrium to be reached, the ‘return time’ (Beddington and May 1977). However, it is erroneous to assume that a given level of fishing effort/mortality allows a certain surplus yield to be maintained indefinitely, irrespective of environmental conditions (Hilborn 1979).

Although analytical models incorporating growth and mortality rates, age at first capture, etc., are widely used in ‘developed’ country fisheries, the data required to estimate current age-structure of the fish population are not available for many small, or tropical stocks, or are labour and technology intensive. Where assessments have been performed, many of them still depend on low precision approaches using sparse or inaccurate data, making management by target reference points problematical, and precautionary approaches to avoiding stock collapse mandatory, even if this apparently involves forgoing some immediate benefits in terms of yield.

In summary, to manage the fishery, it is necessary to ensure that all Conceptual Reference Points to be used are represented by one or more Technical Reference Points, for which the methodology of derivation and measurement is clearly specified (Fig. 1). To borrow terminology from logical framework analysis (CEC 1993), these reference points must also have a means of verification (MOV), and an objectively verifiable indicator (OVI). These must be clearly defined and agreed upon in advance, so that they can be acted upon without the necessity for negotiation. If we have learned anything from the historical performance of fisheries management it is that it is more important that the basis for fishery management action be clear and indisputable than that it should claim to be precise and accurate.

2.3  REFERENCE POINTS AS TARGETS OR LIMITS

The many technical reference points which have been proposed for rational exploitation of fishery resources can, in terms of their use, be placed in two categories: Target Reference Points (TRPs) and Limit Reference Points (LRPs). Traditionally, Target Reference Points have been considered as indicators of a stock status which is a desirable targets for management. It has been assumed that managing a fishery corresponds to adjusting the inputs to, or outputs from, a fishery until one or more of the primary or secondary variables correspond to the TRP chosen (which is, of course, TRP). As previously mentioned, MSY has most often been used in this sense. TRP management requires active monitoring and continual readjustment of management measures on an appropriate (usually annual) time-scale. It also requires attention to the effect of a variety of sources of uncertainty on the estimates of the TRP and of the stock status.

In 1987, the ICES Advisory Committee on Fishery Management (ACFM) report noted that “biological reference points are intended to provide guidance concerning management, and that no biological reference point can serve as a universal target”. In order to protect the resource and the fishing industry against long-term damage, it is important to define and agree on a ‘red area’ where the continuity of resource production is in danger, and immediate action is needed, such as a substantial reduction in fishing effort/mortality, or in the extreme case, closure of the fishery for a period of time (ICES 1988). Reference points which indicate when such a danger area is about to be entered can be referred to as threshold reference points (e.g. Quinn et al. 1990) or (to avoid confusion with the acronym Target RP), as limit reference points (LRPs).

In the United States, fisheries are managed through management plans which are “… required to specify, to the maximum extent possible, an objective and measurable definition of overfishing for each stock or stock complex covered by that FMP, and to provide an analysis of how the definition was determined and how it relates to reproductive potential” (Mace and Sissenwine 1993).

An LRP may either correspond to some minimum condition (e.g. a dangerously low spawning biomass) or some maximum condition (a high rate of decline in stock size, or a high mortality rate) at which point a management response which has been negotiated earlier with the participants in the fishery, is automatically triggered. For new fisheries, or those in developing countries where the information required to use the mathematical fisheries models if often not available, qualitative or semi-quantitative criteria also can be used directly as LRPs. Even when there is adequate information for the definition of sophisticated LRPs, but there are broader ecological concerns about the sustainability of benefits due to the possible impacts of exploitation on the ecosystem, it may be desirable to set LRPs using a precautionary approach (Garcia 1994).

Definitions:

A Target Reference Point indicates to a state of a fishing and/or resource which is considered to be desirable and at which management action, whether during development or stock rebuilding, should aim.

A Limit Reference Point indicates a state of a fishery and/or a resource which is considered to be undesirable and which management action should avoid.

2.4  TARGET REFERENCE POINTS (TRPs)

2.4.1  Maximum Sustainable Yield Criteria: F_msy and 2/3 F_msy

The 1982 Convention specifies only one technical Reference Point, notably the Maximum Sustainable Yield (MSY), a descriptive term for the highest point of the curve describing the relationship between the annual standard fishing effort applied by all fleets, and the yield that should result if that effort level were maintained until equilibrium were reached. This is, at first sight, an obvious target for management of a single species fishery and was widely used for this purpose by fishery commissions in the 1960s and 19770s. Subsequent developments in the theory, and perhaps more so, practical experience in fishery management, have cast doubts on the usefulness of MSY as a safe TRP (e.g. Larkin 1977). This sparked the search for alternative Reference Points, as summarised below.

Maximum Sustainable Yield as the underlying Reference Point for fisheries considerations in the 1982 Convention is the basis for defining ‘surplus yield’ as that part of the MSY not currently being taken by a coastal State. According to the 1982 Convention, any shortfall of national harvests below MSY within an EEZ, may be considered ‘surplus to the needs’ of the coastal State. This consideration has contributed to the continued use by many countries of MSY as a target. However, viewed in the wider context of management objectives, the inappropriateness of this interpretation of ‘surplus yield’ has been noted by many coastal States, since catching this surplus by other fishing states would negatively impact catch rates in fisheries of the coastal State. Several States have concluded that Reference Points based on fishing rates significantly below F_msy levels are more favourable economically and ecologically to their fishing industry and resources. A more appropriate interpretation sees the removal of the MSY as an option which may rarely be used: the stock is held at a level of biomass such that the MSY could but not necessarily should, be harvested, without endangering the stock, implying that stock levels are kept adequately high to permit the MSY to be harvested ‘at least once’ (although as noted by Doubleday 1976, there is no guarantee that, as conventionally defined, it can be harvested year after year).

The MSY, and its equivalent levels of standard fishing effort/fishing mortality (f_msy/F_msy), was first formulated for the symmetrical Schaefer or logistic model, (e.g. Fig. 4). The reference point is ‘model based’, depending on one of the several published production model formulations (e.g. Fox 1970; Pella and Tomlinson 1969), and requires statistical fitting of historical catch and standard effort data on which a considerable literature exists (see Hilborn and Walters 1992, Polachek et al. 1993). The effort level, F_msy at which MSY occurs, can be converted to a fishing mortality, F_msy, if the catchability coefficient q is known.

Figure 4:   The equilibrium Schaefer model, showing the relationship between MSY, Z, and the fishing rate corresponding to the Maximum Biological Production (MBP). Note that F_mbp < F_msy, and that there is no yield when Z ≤ M.

Choosing any model-based Reference Point, implies that the underlying mathematical model that reflects most faithfully the fish population dynamics is agreed upon. The question may, however, be less one of picking the Reference Point with the most robust theoretical underpinnings, and more one of picking an Reference Point that provides conservative advice under conditions of uncertainty. From this perspective, MSY has not performed well as a TRP (see e.g. Doubleday 1976, Larkin 1977, Sissenwine 1978, Garcia 1984). High initial catch rates inevitably result in a substantial overshoot of equilibrium F_msy conditions. There are then serious economic problems in reducing effort and/or fleet size or access downwards to a much lower equilibrium MSY level in later years. This has led to serious criticism of production models which assume equilibrium in forecasting short-term yield (Hilborn and Walters 1992).

There have been few explicit estimates of the accuracy with which MSY conditions have been achieved, but inspection of many production models suggests that an accuracy of better than ± 20% of the standard effort yielding MSY would be unusual. The estimation of MSY by statistical fitting of the model to historical data assumes that past conditions have a similar probability of recurring in the future. However, in a series of years with poor recruitment, a fishing mortality of F_msy produces a yield well below that indicated by the model which is assumed to have been fitted from data for years with more ‘normal’ levels of recruitment. An attempt to harvest the statistically predicted MSY in these ‘poor’ years would require fishing above, and possibly well above, F_msy. Thus the use of the word ‘sustainable’ for an MSY obtained in the conventional way is inappropriate, since “in the presence of fluctuations in production, attempts to remove the MSY yield each year from a stock leads to disaster” (Doubleday 1976).

Owing to uncertainty as to the actual status of the stock with respect to this TRP, it is obvious that a fishery believed to be operating in the region of F_msy is always either overfishing or underfishing with respect to this benchmark, and often significantly so, but the biological production responses of the resource to overfishing and underfishing are not necessarily symmetrical. Overfishing leads to fewer age groups in the fishery, hence increasing the dependence of overall yield on occasional good year classes, as well as leading to declining mean sizes and catch rates. Relatively constant year-to-year recruitment is the exception rather than the rule (Hennemuth et al. 1980), and reduced or more variable recruitment with reduced spawning stock size is accompanied by increased dependence of the fishery on newly-maturing age-classes. This in turn will result in an increase in environmentally-induced variation in stock biomass. From theoretical considerations, Beddington and May (1977) noted that once F_msy has been exceeded, stocks will fluctuate more severely, and their return time to equilibrium will increase markedly.

The appropriate type of production model for a particular fishery can only be known after overfishing has occurred, and the total effort that provides MSY has been exceeded, thus revealing the form of the relationship. Total yield may then drop (implying a dome shaped model), or reach a plateau (as is often the case in tropical shrimp fisheries). Controlled overfishing strategies have even been proposed as one way of better locating MSY conditions and, as noted above, these often occur despite a stated objective of MSY management. Because the type of model to use, and the level of effort or fishing mortality which approximately corresponds to MSY may be only roughly known under the best of circumstances, other more conservative production model based TRPs have been defined.

Following the Marginal Yield concept of Gulland, Doubleday (1976) postulated that fishing at the effort level that corresponded to 2/3 of the effort needed to produce MSY would allow a very large fraction (about 80%) of the MSY to be harvested with a significantly reduced risk of stock collapse. This target, although safer than F_msy, has been criticised, in our view unfairly, as arbitrary, empirical and insensitive to changes in recruitment. It may be useful to note that this approach to setting reference levels for F may be generalized for other commonly used population models (Table 1) and can be regarded as a precautionary approach to using the results of production modelling. This table illustrates that reducing effort levels below those yielding MSY does not result in a correspondingly large reduction in the long-term equilibrium yield once the stock has recovered from the previously heavier exploitation rate.

Sissenwine (1978) has pointed out that for little-studied stocks, MSY has often erroneously been equated with the Maximum Average Yield (MAY). This latter quantity has occasionally been used as a TRP, but gives dangerous weight to the early years of the fishery when catches were high as the virgin stock was being fished down. Again, the use of MAY-based reference points would seem to call for a suitable degree of precaution, as described in the previous paragraph.

A more rational interpretation of MSY for a stock subject to wide variations in recruitment, would be the yield which could be removed in perpetuity from the resource with an accepted low probability of endangering it (Sissenwine 1978). ICES (1993a) has recently been considering this interpretation as an approach to stable long-term management of fisheries. Similar interpretations of MSY are in current use in management of New Zealand fisheries (Annala 1993). The first is a static interpretation called Maximum Constant Yield (MCY) which is defined as “The maximum constant catch that is estimated to be sustainable, with an acceptable level of risk, at all future levels of biomass”. This interpretation is radically different from MSY as normally derived, since it implies much lower levels of fishing mortality (F_mcy) and catch than at F_msy. The second interpretation is a dynamic one called Current Annual Yield (CAY), which is defined as the one year catch calculated by applying a reference fishing mortality, F_ref, to an estimate of the fishable biomass present during the next fishing year. F_ref is the level of instantaneous fishing mortality that, if applied every year, would, within an acceptable level of risk, maximise the average catch from the fishery (F_ref is often set equal to F_0.1) In the New Zealand fishery assessment process, Maximum Average Yield (MAY) is the long-term average of CAYs, and is higher than MCY since the CAYs closely track the variation in fishable biomass (Annala 1993).

In conclusion, it is now evident from the variety of technical, conceptual, and practical difficulties associated with the use of F_msy as a TRP, that it should be used more as an LRP as discussed in the following Section.

2.4.2  Yield per recruit criteria: e.g. F_max, F_0.1.

The fact that production models combine all aspects of population productivity, recruitment, growth and mortality and ignore details such as age/size at recruitment, has led to the use of analytical, age-structured, models based on detailed population dynamics. These are particularly useful when there are several fleet components exploiting different age groups, and when gear regulations affecting age/size at first capture may be an important management tool.

The early theory of population dynamics of exploited fishing stocks emphasised the calculation of F_max, the level of fishing mortality for a given size at first capture, which maximizes the average yield from each recruit entering the fishery. The yield-per-recruit analysis uses information on average individual growth, natural mortality and vulnerability to fishing. This was one of the earliest benchmarks for fisheries management, and as for MSY, suffered from a number of failures as a TRP.

The use of yield-per-recruit as a reference variable does not take into account the effect of fishing mortality on the proportion of mature fish left in the population and hence its reproductive potential. Although generalizing about the relative performance of Reference Points developed from production models and analytical models can be hazardous, there seems little doubt that F_max is usually greater than F_msy, and that fishing at this rate over an extended period of time is liable to deplete the spawning stock and reduce future recruitment (e.g. Clarke 1991). Although there may be good reasons for eliminating the use of F_max as a TRP, it could be considered as an upper limit for F, i.e. as an LRP for the stock

For many species there is no clear maximum to the curve of yield-per-recruit against F. The fishing mortality level F_0.1 proposed as a conservative TRP by Gulland and Boerema (1973) does not require that there be a maximum, being an arbitrary criterion based on the slope of the yield per recruit curve at the origin. F_0.1 is the fishing mortality rate at which the slope of the yield per recruit curve as a function of fishing mortality is 10% of its value at the origin (Fig. 5). In South Africa, an even more restrictive criterion is used, notably F_0.2.

Figure 5:   Illustrating the method of defining F_0.1, given a known relationship between fishing mortality rate and yield per recruit, as the point on the Y/R curve at which the slope of (a line tangential to) the curve is 1/10th the slope of (a line tangential to) the curve at the origin.

The F_0.1 measure, although arbitrary, is in a sense a bioeconomic criterion, in that a marginal yield of less than 10% was felt to be close to the point at which most fisheries administrators would consider further increases in fishing mortality or effort to be no longer economically worthwhile. This measure has been widely used in many fisheries of the Northwest Atlantic (e.g. Rivard and Maguire 1993, Hildén 1993). F-based strategies have been followed off eastern Canada for more than a decade, and F_0.1 is often used in establishing catch quotas.

Knowledge of the correct catch is essential to estimating current F-values under quota control, but there are known to be substantial problems with the accuracy of commercial catch reporting. This has affected stock assessments and has been especially pronounced where there is fleet overcapacity. With under-reporting there is a high probability that target F values will be exceeded. This, and not just the changes in F_0.1 that occur with changing fishing pattern and input values for M (Jakobsen 1992), may be the main explanation for declines in several stocks managed under F_0.1 criteria.

2.4.3  TRPs based on the size of fish caught

Yield-per-recruit analysis indicates the mean age/size of fish in the catch that provides the maximum yield-per-recruit for a given set of population parameters and given fishing mortality level. When the data required to estimate an optimal level of fishing mortality are not available, the mean size of fish in the catch can be used in conjunction with other data as a ‘proxy’ TRP.

The use of mean size of fish as a TRP may be based on yield-per-recruit analysis or may consider the recruitment ogive (partial recruitment) in relation to the size at first maturity. A rational target would be to aim for an exploitation rate such that the average size of fish caught is equal to, or greater than, the average size at maturity. Thus, at least 50% of individuals would have an opportunity to reproduce. For iteroparous species the relationship of this target to a target %SSB (see below) would depend on M, which determines the average number of years a mature fish can be expected to spawn in an unfished population before dying of natural causes. Caddy and Die (in press) substitute the average length at maturity into Beverton and Holt's equation relating Z to mean size in the catch to estimate a corresponding reference value for total mortality Z^*. This may be useful when survey data from which to estimate Z are available, but catch data from which to estimate the mean size are not.

2.4.4   Reference points based on the natural mortality rate, M.

New fisheries usually develop in the absence of adequate assessment information, and management has to proceed on the basis of available information. A cautious approach may result in underexploitation, but will not necessarily lead to a long-term loss of yield. In the 1960s and 1970s, many new fisheries developed in different parts of the world for which the only data on stock status were one or more exploratory survey estimates of biomass. In an attempt to provide some basis for fleet and fishery development, Gulland (1973) proposed a simple empirical formula for the MSY in terms of the virgin biomass B₀ and the natural mortality rate, M: notably, MSY = 0.5MB₀ (a reformulation of the second yield equation in Annex I). This follows the symmetrical Schaefer yield model in assuming that MSY will occur at half the virgin stock size B₀, and that at MSY, the fishing mortality and natural mortality rates will be equal. Later, the equation was generalized to MSY = x.M.B₀ with the value of x being related to the stock characteristics, and variations were proposed to accommodate situations in which there was already some fishing mortality (Gulland 1983). Garcia et al. (1989) proposed several estimators for MSY when historical data series are not available.

There is little empirical evidence that F_msy = M for many stocks. Beddington and Cooke (1983) suggested that x is generally smaller than 0.5, whereas for tropical penaeids Garcia and LeReste (1981) suggested that values x = 0.32 to 0.44 are appropriate. From a set of 11 stocks, Caddy and Csirke (1983) found x values ranging from by 0.33 to at least 4. The lowest values were for short-lived shrimp and sardine populations, and the highest were for two northern demersal finfish; apical predators with low natural mortality rates. From an analysis of several stocks of small pelagic fishes, Patterson (1992) found that only low exploitation rates, corresponding to no more than x = 0.33, are sustainable. For new fisheries in New Zealand, a conservative approach is used, where MCY = 0.25F_0.1B₀ (Annala 1993). These benchmarks, though very approximate, may be the only ones immediately available for setting ‘Precautionary’ Reference Points for many stocks off developing coastal countries.

2.4.5  Reference points based on the total mortality rate, Z_mbp, Z^*

Since partitioning the overall mortality into components due to fishing, predation, etc., is often problematical, there may be advantages in expressing Reference Points in terms of the overall mortality Z experienced by the stock due to all causes of death. Virgin populations are dominated by large, old individuals, whose contribution to biological production (growth, yield, plus deaths due to predation) is lower than that of younger individuals which gradually dominate population as exploitation intensifies. Thus we can postulate that there is a mortality level, Z_mbp, at which the Maximal Biological Production is obtained from the stock (Caddy and Csirke 1983).

For the Schaefer model, Z_mbp and F_mbp correspond to a fishing mortality which is consistently below F_msy, being progressively more so for species low in the food chain with high natural mortality rates (Fig. 4). Simulations show that it is difficult to produce excessively high fishing mortalities using this TRP (Caddy and Die in press). There should be little risk of environmentally induced collapse when the stock is at its maximum productive capacity.

2.4.6  TRPs derived from recruitment considerations

In addition to size-based reproductive TRPs (Section 2.4.3) due to the frequently demonstrated dependence of recruitment on the spawning stock size, there is a potential role for TRPs which ensure that the spawning capacity of the stocks to reproduce will be conserved. TRPs based on recruitment considerations may be derived from stock-recruitment (S-R) relationships, or from an extension of yield per recruit analysis which incorporates age/size at maturity in calculating the spawning biomass per recruit (SPR) at various levels of F. Recently, these two types of analysis have been linked to calculate the stock biomass levels associated with various SPR levels. The targets may be stated in terms of a stock biomass or spawning stock biomass that is expected to yield the desired recruits, or in terms of the fishing mortality level which is expected to result in these biomass or SPR levels.

Early approaches to stock-recruitment analysis involved fitting various types of curves to time series of data on stock and recruitment. In all S-R relationships, the spawning biomass corresponding to the Maximum Surplus Reproduction, B_msr, occurs at some level intermediate between a high and a very low stock size (Ricker 1975) (Fig. 6). In theory, for any stock size greater than B_msr, there is a level of fishing mortality, F_msr, that would allow the B_msr to survive and reproduce in that year. In practice, given natural variation in stock sizes from year to year, this level of fishing mortality would have to be changed annually to achieve a ‘constant escapement’ of B_msr. Thus, the fishing rate that would allow constant escapement must be calculated annually. This may be a useful strategy for managing salmon which can be counted during their upriver spawning migration, but is likely to exceed the information levels for most widely dispersed stocks in open sea systems. A further problem with direct use of the S-R relationship is in determining the correct stock recruitment model to use under conditions of high recruitment variability.

Figure 6:   Illustrating (after Ricker 1975) a generalized relationship between spawning stock size and the number of recruits (progeny). The line A-B corresponds to the stock size or biomass of spawners for this specific relationship where the surplus of progeny over parental stock size is at a maximum: the point at Maximum Surplus Reproduction.

The considerable variability in recruitment data, and other methodological problems (Walters and Ludwig 1981) has made it difficult to describe statistically significant, and biologically meaningful relationships between spawning stock size and number of recruits for many stocks. This problem has led to other approaches to the use of S-R data to generate Reference Points. Evans and Rice (1988) propose methods of predicting recruitment directly from observations on stock and recruitment without the mediation of a functional relationship. Getz and Swartzman (1981) propose an approach in which the S-R scattergram would be divided into stock biomass and recruitment ranges. A transition matrix indicating the probability of each recruitment range resulting from each stock biomass range, can be used as a guide to setting a target biomass range for fishery management, and the extent to which it may be necessary to avoid low biomass levels. A major problem with S-R analysis is that a relatively long time-series spanning a range of stock sizes (e.g. Myers et al. 1994) is needed to produce a reliable stock-recruit curve. This is rarely available for setting TRPs for newly exploited or little studied stocks.

The calculation of ‘spawning biomass per recruit’ (SPR) is an extension to yield-per-recruit analysis which can be carried out in the absence of historical data, if information on maturity/fecundity at size/age is available (Gabriel et al. 1989). Mace and Sissenwine (1993) explain the derivation and application of Reference Points based on SPR calculations. Unlike yield-per-recruit which shows a maximum with increasing F, SPR decreases monotonically. SPR is usually expressed as a percentage of the SPR under unfished conditions (i.e. at virgin spawning biomass, B₀) and is variously designated as (%SSBR or %SPR). The F which produces any particular %SPR is designated F_%spr or just F_%.

Reference Points based on SPR or %SPR have only recently been defined based on the relationship between SPR and the survival ratios (R/S) obtained from pairs of stock-recruitment observations (Fig. 7). For any F level there is a corresponding straight line through the origin of the S-R scatterplot. The slope of this line is the inverse of the SPR which corresponds to the F level. The S-R values and plot can then be used to select a survival ratio for use as a Reference Point. This can be translated back into SPR values and projected onto the F scale to determine the corresponding F level. The reference level F_rep (Sissenwine and Shepherd 1987), also referred to as F_med by ICES (1993b), corresponds to the line representing an average survival ratio, S/R = 1, at which the stock replaces itself. At this level of F, SR would be expected to be > 1 in 50% of the years, i.e. corresponds to the F where recruitment in half of the years more than balances losses due to mortality.

Figure 7:   Illustrating the definition of F_low, F_med and F_high and their relationship to spawning stock biomass per recruit (SPR) (redrawn from Jakobsen 1992)

Heavy fishing and depressed stocks in the North Atlantic in recent years have led fishery scientists to emphasize spawning stock considerations in advice to management bodies. Thus the use of recruitment-based TRPs derived from SPR and S-R data has been pioneered in the ICES area (ICES 1984). ICES routinely estimates three reference levels of F in this way, F_high, F_med and F_low F_low and F_high bracket F_med, and are similarly defined t .brb.o leave 90% and 10% of the data points for recruitment above the line through the origin corresponding to that level of fishing mortality (Fig 7). Assuming that S-R relationships continue as in the past, they have the following properties (ICES 1991; Jakobsen 1992):

F_low — low probability of stock decline, and some likelihood of stock increase,
F_med — likely that current stock levels will be sustained,
F_high - likely that fishing at this level will result in stock declines.

More recently, ICES has tended to view F_med as a limit Reference Point since at F levels higher than F_med stocks can be expected to decline (see Section 5.6).

These measures appear less vulnerable to the consequences of assuming an incorrect value of M than F_max and F_0.1 levels (Jakobsen 1992). F_med fell close to F_max and F_msy fo .brb.r Georges Bank haddock (Gabriel et al. 1989).

Simulations showed that for northern demersal stocks a yield of at least 75% of MSY is possible as long as the spawning biomass is maintained in the range 20–60% of the unfished level; irrespective of the spawner-recruit relationship (Clarke 1991). Relative spawning biomass can be maintained in this range by choosing an F value that will reduce spawning biomass per recruit to about 35% of the unfished level. This F value is usually very close to F_0.1 (Clarke 1991). Variation in recruitment calls for a slightly higher target level of SBR, around 40%, particularly if there is serial correlation in recruitment (Clark 1993). Thompson's (1992) analysis of uncertainty in the stock recruitment relationship supports this finding, and suggests the intuitive conclusion that F should be constrained when the S-R relationship is uncertain.

In a recent comparative study, %SPR was found to be positively correlated with natural mortality and negatively correlated with various indices of size: thus cod and most flatfish require low levels of %SPR, but some pelagics require values as high as 40–60% for consistent stock replacement. Although these conclusions agree with those in the earlier section on M based Reference Points, it is probably dangerous to use them outside their geographic region of origin, since the data upon which this generalization is based are mainly for fishery resources in higher latitudes. Nonetheless, the use of %SPR criteria is not as information-demanding as other reproductive criteria, and has broad potential in the developing fisheries context.

Mace (1994) observed that TRPs and LRPs are highly dependent on the degree of density dependence in the S-R relationship. She recommended that when the S-R relationship is unknown, F_40% be adopted as a target fishing mortality, but that it be adjusted to accommodate any known or assumed degree of density dependence in the S-R relationship. This corresponds to a recruitment of about 50% of that expected from a virgin stock. For recruitment-based TRPs where biomass is stated in relation to virgin biomass, the latter is estimated from the intersection of the S-R curve or mean recruitment with the replacement line corresponding to F=0, the unfished condition.

2.4.7  TRPs derived from economic considerations - the optimal fishing effort, F_may

Normal market forces are believed to maximize economic benefits to participants (Gordon 1954), but in open access fisheries, the individual efforts of private agents (fishermen), each working to improve their individual economic situation, does not ‘guide the net sum of private activities towards the common good’. In fact, recent analyses of global fishery trends have revealed that the high level of over-investment in fleets is the major causal factor for overfishing within and outside EEZs (FAO 1992a, b). Combined with restrictions on fisheries within EEZs, this has motivated movement into largely unrestricted fisheries beyond 200 miles. The global sum of fishing subsidies was estimated at about US $54,000 million per year (FAO 1992b). The development of effective management criteria therefore would potentially release substantial global financial resources, as well as reducing adverse impacts on stocks.

There is an extensive literature on fisheries economic theory in which the Gordon-Schaefer equilibrium production model is central (Gordon 1954, Schaefer 1957, Clark 1983). This theory holds that there is an economic TRP, the Maximum Economic Yield (MEY), which occurs at the effort level yielding the greatest margin of revenue over cost from the resource (Fig. 8). For a linear cost curve, this inevitably occurs to the left of MSY on the fishing effort axis. Since F_mey occurs at lower levels of effort than F_msy, the use of this economic Target Reference Point is less likely to result in biological overfishing than the use of F_msy.

As a TRP, F_mey is responsive to any changes in the economic environment which affect either the value of fish, or the cost of fishing. It may also be dependent on changes in fish abundance, if market price increases with declining abundance and is independent of the availability of similar resources elsewhere. Subsidies or external economic considerations such as fuel taxes will also affect the location of an economic Reference Point (e.g. Panayotou 1988).

The effect of supply on fish prices may, under certain circumstances, result in higher total profit, or profit per unit catch, when total catch is reduced. This characteristic may be a consideration in setting target fishing levels or catches but is least likely to be effective in situations where fish prices are set by global markets, e.g. the tuna fishery for the canning industry.

The value of a unit weight of the landed catch may vary with the size of individual fish, and in multispecies fisheries with species composition. Both fish size and species composition are functions of fishing mortality, and based on purely economic criteria, may be used as target reference points. Even if the actual target F cannot be estimated, in theory, F could be adjusted in increments until the desirable target catch characteristics are achieved.

In considering TRPs based on economic criteria, it is important to be aware of the effect which the practice of discounting could have on reference points. In evaluating investment projects, including resource management, economists discount the future value of any commodity. Discount rates may be in the order of 10%. In the case of a fishery where the population growth rate does not exceed the discount rate, then a strict application of economic theory would suggest that in the absence of other considerations (such as an economic value placed on recreational use of resources) the whole stock should be harvested now, and the proceeds of their sale invested. Long-lived species with slow growth rates, such as whales, clearly fall into this category. The blatant contradiction between this common economic approach, and the concept of sustainability, constitutes an unresolved paradox (Hilborn and Walters 1992).

Figure 8:   The Graham-Schaefer equilibrium production curve relating yield or revenue to effective fishing effort, showing three Reference Points: MEY, MSY and the bioeconomic equilibrium point E. These occur at progressively higher levels of fishing effort.

2.5  LIMIT OR THRESHOLD REFERENCE POINTS (LRPs)

2.5.1  F_msy as an LRP

The use of F_msy as a LRP rather than a TRP could provide flexibility in choosing a more cautious F-based TRP that has useful management characteristics (McGarvey and Caddy in press). This is illustrated in Fig. 9. It is necessary to have information on the variability associated with the estimate of current fishing mortality, F_now. While such information is infrequently reported in the literature, it seems unlikely for well studied fisheries that the coefficient of variation for F will be less than 15–30%. For poorly studied fisheries it will probably be much higher. In the hypothetical example shown in Fig. 9, FMSY = 0.6 is assumed to be a LRP. The actual fishing mortality rate FNOW that will be exerted in the current season is not precisely known, but two cumulative probability distributions are shown, one corresponding to a high level of precision? (c.v. = variance/mean = 20%); the other to a low level (c.v. = 40%). The means of both distributions are positioned relative to each other along the X-axis such that the points for which there is a 10% probability that FNOW > = FMSY = 0.6, coincide. It is clear that for this situation to prevail, the centre or mean of the distribution of FNOW which is imprecisely known, must be located at a lower fishing mortality rate than when more accurate statistical information is available.

In simple terms, this example is intended to show that the collection of accurate and complete statistics which allows the death rate due to fishing to be calculated with a higher precision, permits a higher fishing rate to be maintained with the same risk of overshoot, than if data collection is given a low priority. This illustrates clearly the economic value of a good system of data collection in a precautionary management system.

Figure 9:   (caption) illustrating two situations: one (the upper curve) where the precision of forecasting the current fishing mortality rate is relatively high, compared with that for a low precision (lower curve). If both have an equal 10% probability of exceeding FMSY at FNOW = 0.6, and entering the RED AREA, the target fishing mortality with the lower precision must be set more cautiously than if, due to improved statistics, the current fishing mortality is better known.

A more elaborate use of MSY as a LRP was incorporated in the New Management Procedure developed by the International Whaling Commission (Garrod and Horwood 1979), where a maximum harvest of 90% of MSY (set at 60% of the unexploited stock level) was agreed to. This was to be reduced progressively by 10% for each 1% shortfall of the stock below the level required to produce MSY. Thus, as soon as stock size dropped below 90% of the MSY level, there was a threshold at which the stock entered a fully protected category. This example also illustrates one other essential feature of a LRP-based management system: the pre-negotiation of future automatic management responses once the system enters an agreed endangered state.

2.5.2  LRPs derived from stock recruitment considerations

The guidelines for US fishery management plans state that although some types of overfishing (growth, localised and pulse) may be permissible, management must guard against recruitment overfishing. Mace and Sissenwine (1993) noted that 60% of the definitions of overfishing to date had been based on spawning stock biomass/recruit (SSB/R) analysis, with typical values ranging from 20–35% of virgin stock levels. In response to the above guidelines and for MBALS in ICES assessments, there has been considerable recent activity aimed at developing various methods of calculating recruitment based LRPs and at evaluating their relationships to various TRPs (Mace and Sissenwine 1993, Clark 1993, Goodyear 1993, ICES 1993b, Mace 1994, Myers et al. 1994). Most of these LRPs are variations of the TRPs discussed in the previous section, and are derived in a similar fashion.

An extreme LRP for spawning stock biomass is F_T, which is based on the slope of the S-R relationship at the origin (Mace and Sissenwine 1993). When F > F_T, effective stock extinction is assured. One proposed way of estimating F_T is to use the 90th percentile of observed survival ratios (S/R); the same as the F_high of ICES. However, the authors noted that if recruitment is highly variable, and the majority of S-R observations are at a low stock size, this approach will probably overestimate F_T. In fact, if the S-R scattergram consists only of points on the ascending linear part of the curve, F_T will be more closely approximated by the 50th percentile of observed survival ratios, which is the same as F_med of ICES. Even for a stock in which stock recruitment data cover the full range of stock sizes, given that stock decline would be expected at sustained F levels in excess of F_med, it appears that F_med may the most rational recruitment-based LRP for most stocks.

From a theoretical analysis of biological reference points for stocks with a wide range of life history characteristics, Mace (1994) recommends that when the S-R relationship is known, TRPs should be estimated directly. She points out that neither the slope at the origin of the S-R relationship nor the SSB which would be expected to provide 50% of the maximum recruitment (R_max) are likely to be conservative LRPs, and “…should probably be treated as absolute boundaries not to be crossed”. In view of this she recommends a target F which should be as close as possible to F_msy subject to the constraint that the probability that stock biomass will fall below 100.T.B₀ should be no greater than 0.05.

Myers et al. (1994) pursued the issue of defining LRPs based on conservation of spawning stock biomass using S-R data from 74 stocks for which there were > 20 years of data. They evaluated eight methods for estimating the critical spawning stock biomass (%SSB). Of six methods which relied on fitted S-R relationships, two estimated the point where expected recruitment would be 50% of its maximum value. The remaining four estimated the critical point as 20% of estimated virgin biomass. They concluded that there was no one single method for estimating critical spawning levels for all stocks. However, they proposed a number of simple criteria to determine if an LRP estimated from S-R data using any of the above criteria is sensible. These are based on the relative slopes of the log-transformed S-R points above and below the estimated LRP:

-   If both slopes are positive, and the slope above the LRP is less than that below it, the LRP is sensible,

-   If both slopes are positive, and the slope above the LRP is greater than that below it, the LRP is probably set at too low a biomass,

-  If both slopes are negative, the LRP is probably conservative,

-   If the slope above the LRP is positive and that below it is negative, the data should be considered uninterpretable.

Another simple rule is that recruitment below the LRP should be on average lower than above the LRP (ICES 1993b).

The above methods depend on the availability of S-R data. In the absence of information on stock and recruitment, practical management advice has been based on generalisations from examination of a large number of exploited stocks. A survey of 91 stock and recruitment data sets for Europe and North America suggest that for stocks considered to have average resilience, a biomass level of 20% of the unfished level should be considered a recruitment-based LRP. In the case of little known stocks, the LRP should be set at 30% of the unfished biomass level. The theoretical analysis by Mace (1994) supported these recommendations and suggested that these results may be applicable to stocks outside the North Atlantic.

2.5.3  Other biological LRPs

If the age at first capture falls below the age at first maturity there is a risk of recruitment overfishing. If control of fishing effort is unreliable, one Reference Point that could be used would be to require fishing to take only individuals at and above the size of first maturity, without discarding or damaging undersized individuals.

Die and Caddy (in press) suggested other possible warning signals which could be adopted as LRPs in the absence of adequate information or more precise analyses, as is often the case with fisheries in developing countries. These include: (a) when total mortality Z rises above some agreed value, such as that corresponding to Z_mbp or Z^* for the stock (see section 2.4); (b) when the proportion of mature individuals in the stock falls below some agreed percentage of that for the virgin stock; and (c) when annual recruitment remains poor for a predetermined number of years in a row. Other robust indices which are often associated with low stock size and hence reduced intraspecific competition, are increases in weight-at-age and reduced size at maturity, but by the time these biological indicators have changed significantly, overfishing may already be severe. Figure 10 illustrates the use of survey data to monitor stock status in relation to an agreed LRP. This may be particularly useful when it is difficult to obtain representative samples from the fishery.

Figure 10:   Illustrating the use of a survey CPUE or Z value as an agreed upon LRP. Surveys may be the only approach to monitoring fishery status relative to and RP when little or no fishery data are available.

Zheng et al. (1993) compare and evaluate threshold estimation methods for pollock and herring in the Bering Sea. Most are variations of the LRPs dealt with above, but their treatment includes some different approaches to estimating the TRP, including the incorporation of zero production thresholds and depensatory production into the traditional surplus production model.

2.5.4  LRPs derived from economic considerations

It is generally acknowledged (e.g. Panayotou 1988) that on the curve of revenue versus fishing effort, the point of ‘economic equilibrium’ is an ‘attractor’, albeit extremely undesirable, for an open access fishery in which net earnings from the fishery equal the costs of fishing (Gordon 1954). Beyond this level of effort the whole fishery is operating at a loss. This is also the point at which the cost of managing the fishery for mortality rate will theoretically be zero. The effort level corresponding to this point (E in Fig. 8), is artificially increased when subsidies reduce the cost of fishing (see FAO 1992b).

In situations where management is impossible, or where the State(s) in question cannot afford any form of management, the unsubsidised point of economic equilibrium could be adopted as an LRP. It would be achieved by the removal of any subsidies supporting the fishery sector.

Since catch per unit effort is often assumed to be proportional to biomass (CPUE = q.B), and revenue is proportional to CPUE, the revenue per unit effort (RPUE) is a potential economic reference variable. This may be particularly useful in some fisheries for highly migratory resources where survey methods are difficult to implement. The point at which RPUE is equal to the cost per unit effort of fishing is a variation of the LRP suggested above. It is presumably axiomatic that a fishing operation that does not generate rent but contributes to dangerously depleting the stock is difficult to justify. However, it will be necessary to separate low CPUE due to availability (e.g. at the beginning and end of the local fishing season for a migratory resource) from that due to low stock size.

2.6  REFERENCE POINTS FOR NEW OR DEVELOPING FISHERIES

Establishing Reference Points for new or developing fisheries requires special consideration if overcapitalisation is to be avoided. Reference points will usually be derived from exploratory or survey biomass estimates, as described in section 2.4.4 above, with considerable uncertainty as to the appropriate values of x to use. Precautionary or probing strategies are suggested, but which restrict fisheries to fishing intensities well below the likely MSY levels revealed by exploratory fishing (e.g. Annala 1993). Intensive data gathering should be an objective in a developing fishery and provision of data should be a requirement for licensing any ‘pilot scale’ fishery. Although there is the need to acquire good information from the fishery at low stock sizes for future fitting of fishery models, it must be recognized that as exploitation intensifies, the behaviour of the population, including the S-R relationship, may not be adequately described by data from the period when the exploitation rate was low.

2.7  REFERENCE POINTS FOR STOCK REBUILDING

Considering the overexploited condition of many marine fish stocks (FAO 1994), stock rebuilding towards long-term TRPs must be a priority for management. Rebuilding requires that effort be reduced to permit the accumulation of surplus production. This means that the fishing industry must accept a short-term loss in revenue in return for the expectation of higher yields-per-unit effort in the long term (e.g. Overholtz et al. 1993). Appropriate targets levels of F for rebuilding will depend on the extent of overexploitation and on the economic impacts of the action, but may need to be considerably lower than those which can be sustained at long-term target stock sizes.

Since stock rebuilding generally requires several years, fishing intensity need to be reduced continuously for the required period. For relatively long-lived species such as cod and haddock, Rosenberg and Brault (1991) showed that rebuilding over moderate time spans (say 5 years) is less economically destructive than short, sharp reductions in fishing mortality (2-year rebuilding scenarios), but that longer rebuilding periods are likely to be too long to see signs of effective recovery. In the case of short-lived stocks, the rebuilding time is likely to be correspondingly shorter, however. For many stocks which are currently heavily exploited, larger-than-normal cohorts make up a progressively larger part of the annual yield, but may not occur very frequently. Keying in on the protection of these larger-than-normal cohorts may be the most rapid way to rebuild a stock.

For stock rebuilding, F must be below F_med, the level at which the stock replaces itself (Mace and Sissenwine 1993). For extremely depressed stocks, F_low, the level at which recruitment is expected to exceed replacement level in 90% of the years, may be the most appropriate strategy. In any case, the rebuilding target level of F will be an arbitrary level which depends on the desired rate of rebuilding. As with other reference points, it must be agreed upon prior to implementation and sustained in the face of short-term market requirements. In rebuilding it may be reasonable to allow F to progressively approach the target F as stock biomass increases. Owing to the inevitable pressure from the fishing industry to increase effort at high catch rates and to the dependence of most current management systems on short-term decision making, the schedule for increasing F towards the long-term target level in relation to the reference points chosen, must also be agreed to in advance by participants.

Reference points can be envisioned for recovering stocks that would signal various stages in the recovery process and indicate that the recovery or enhancement plan is working. These recovery reference points can be in terms of various states of the population. For example: biomass could be indicated as a percentage of the rebuilt target biomass; for multi-age stocks, an age structure that includes an increased number of age classes; expansion of the area of or number of, localities occupied by the species; the point at which the requirements of predators species have been satisfied.

2.8  PRECAUTIONARY REFERENCE POINTS

In some fora there has been reference to “Precautionary Reference Points”, where the intention appears to indicate that such reference points, whether Limit or Target Reference Points, are to be used in a precautionary fashion.

Garcia (1994) provides a discussion of the issues involved in applying the Precautionary Principle to fisheries. His definition of this Principle is:

“Accepting that, in order to protect a marine area from possibly damaging effects of the most dangerous fishing practices and gears, a precautionary approach is necessary which may require action to control fishing activities even before a causal link has been established by absolutely clear scientific evidence.

States accept the principle of safeguarding the marine ecosystem by reducing dangerous fishing practices, by the use of the best technology available and other appropriate means. This applies especially when there is reason to assume that certain damage or harmful effects on the living resources are likely to be caused by such fishing practices and technologies, even where there is no scientific evidence to prove a causal link between practices and effects (the principle of precautionary action).”

These concepts and guidelines for their implementation have been further developed in the Technical Consultation on the Precautionary Approach to Capture Fisheries (FAO/Govt. of Sweden 1995). Their deliberations were predicated on the definition of precaution as: “Caution exercised beforehand to provide against mischief and secure good results - prudent foresight.”

The term precautionary reference points does not refer to how the reference point was developed, or its technical basis, but to how it is used as a component for precautionary management strategy. This seems to be how the term is being used in the ‘United Nations Conference on Straddling Fish Stocks and Highly Migratory Fish Stocks’ (United Nations 1995).

Annex 2 of this last-cited report “Guidelines for Application of Precautionary Reference Points in Conservation and Management of Straddling Fish Stocks and Highly Migratory Fish Stocks” is of particular relevance, and is reproduced in full below:

2.9  REFERENCE POINTS FOR HIGHLY MIGRATORY RESOURCES

From a technical point of view, reference points for straddling and highly migratory stocks do not differ from those for shared stocks (Gulland 1980, Caddy 1982), or for those occurring entirely within an EEZ. However, the feasibility of application of individual reference points may differ, due to the multijurisdictional nature of the resources, rather than their biological characteristics. The variety of types of shared/migratory resources is described by Caddy (1982). Here, we consider the most difficult situation, that of sequential exploitation along a migratory route.

This particular case has been the subject of intense international negotiation at the United Nations Conference on Straddling Fish Stocks and Highly Migratory Fish Stocks in 1994–95. An excerpt from Article 7 of the Chairman's report is given in the following box:

Complex management arrangements may be needed to deal with highly migratory resources in which multiple fisheries occur sequentially at different locations on the overall migratory route (Fig. 11). Such local fisheries are usually seasonal, and often too short to allow declines in catch rate and/or size to be unambiguously attributed to fishing as opposed to changing regional availability or migration. At each locality, availability to fishing and age composition of the catch may differ. Under these circumstances, there is no ready alternative to pooling catch data and performing a global assessment. One possibility is to use an escapement or gauntlet model for management (Paulik and Greenough 1966).

A practical consideration for sequential fisheries on a common stock, is that fishing locations may differ in suitability in relation to a size-based Reference Point such as the optimal size at first capture I_c based on yield-per-recruit or spawning biomass-per-recruit analyses for the whole stock (Fig. 11). Consequently, the sacrifices needed to either achieve an optimal yield-per-recruit, or to protect the spawning stock or juveniles from overfishing, are not shared equally for all participants. They often depend for their success on the actions of one or a few coastal States where these critical life history stages occur. Under these circumstances, the overall yield from the population will be suboptimal if all participating States are obliged to harvest the stock exclusively within their EEZs, and if only a few (e.g. juvenile) age classes are available in any given EEZ (Caddy 1982). The optimal solution from a yield-per-recruit perspective would be to agree to limit harvesting to seasons/areas where the size frequency, catch rates and market prices are optimal. This would require access to these areas for all parties, or compensation for those parties prepared to forego fishing for suboptimal sizes within their own EEZs.

In a sequential fishery for a highly migratory species, the best overall Reference Point is one which ensures that a target spawning biomass survives all fisheries to replace the stock. From a simple example, it is clear that this can be achieved by many different combinations of national allocations, which all result in the same cumulative risk of death prior to spawning (Table 2). If the mechanism proposed in the last paragraph is rejected in favour of sub-optimal harvesting within each jurisdiction, the vector of mortalities-at-age, and the corresponding allocations, should still be negotiated by participants in relation to one or more of the overall stock reference points discussed earlier.

Figure 11:   An idealized gauntlet fishery for a highly migratory species crossing 3 jurisdictions during its life history. The size at first recruitment to the fishery l_r, occurs in jurisdiction 1; the optimum size at first capture (Ic_opt) which gives maximum Y/R occurs in jurisdiction 2, and spawning occurs in jurisdiction 3.

Economic reference points may not always be practical as reference points for management of straddling stocks, and even less so for highly migratory resources, because each national fleet may have a different economic optimum, depending on its costs, earnings and national market prices. In general, F_mey is not easily defined in fisheries involving several fleet components with different gears and fishing practices.

Similarly, for straddling and highly migratory stocks, it can be extremely difficult to estimate a yield-per-recruit based TRP if fleet-specific vectors of F-at-age for an exploited resource differ among jurisdictions, and the relative effort of the fleets changes from year to year.

2.10  MULTISPECIES AND ECOSYSTEM CONSIDERATIONS IN SETTING REFERENCE POINTS

The need for multispecies and ecosystem perspectives in fisheries management has been frequently noted (Mercer 1982, Sugihara et al. 1984). For fisheries management, these considerations may be seen as comprising the following categories:

Technical interactions -- The technical problem of managing sets of species which are harvested together, regardless to whether there are any biological interactions among them;

Species interactions -- Primarily the effects of predation and competition on the population responses of the species for which management advice is being provided;

Ecosystem interactions¹ -- The effect which the reduction of biomass of exploited species may have on other organisms in the ecosystem of which they are part.

Fisheries scientists have recognized the potential impacts of all of these relationships on the probability of success of management based on single species models, and have devoted considerable effort over the past two decades to developing practicable solutions to these problems. Formal incorporation of these considerations into management advice has, however, been difficult to achieve.

2.10.1   Technical interactions

The major problem with harvesting several species from an assemblage using a single unselective gear, such as trawl or fish trap, is that each species will have different life-history characteristics, and consequently different responses to exploitation. Thus an overall F-based Reference Point will overexploit some species and underexploit others. Scaling speciesspecific values of F-based LRPs for different trophic levels according to their relative rates of natural mortality, remains a theoretical possibility (Caddy and Sharp 1986), but would be difficult to implement for fishing gears such as bottom trawls or fish traps that are relatively unselective by species. For such unselective types of gears, a precautionary approach for all species being exploited risks ecosystem exploitation being defined in terms of the species with the least resistance to harvesting. Thus if harvesting is to be optimized for the individual species of an ecosystem, developing more selective modes of harvesting is a high priority. This can be approached through the use of more selective gear, or through knowledge of the spatial distribution of the resource and deployment of effort accordingly.

¹   Clearly, species interactions play a leading role in ecosystem responses to exploitation. Nonetheless, the latter two categories are intended to distinguish between population level and ecosystem level phenomena.

When there are separate assessments for individual species, as in demersal trawl fisheries in the north Atlantic, the species TACs are rarely in proportion to the relative catch rates in the trawl. Optimization of the concurrent harvesting by trawl fisheries on jointly-occurring groundfish species, has been approached using information on the spatial variability in composition of the catch to deploy fishing effort among species assemblage areas, in such a way as to optimize harvesting of the combined quotas (Murawski et al. 1983, Murawski and Finn 1988). In this case, this target reference point is the sum of quotas of all species. In theory, the target can only be achieved if all quotas are met at exactly the same time, otherwise the fishery closes when the first quota is met, and the remainder of the other quotas remain unharvested. In reality, the fishery compensates for this by discarding, and one major objective of quota management is aimed primarily at reducing this undesirable feature of multispecies fisheries.

There is increasing concern about the direct physical effect of fishing activities on marine habitats (ICES 1993b, EEC 1994). As these physical effects become better understood, these concerns could give rise to further fishery reference points which may define and limit the permissible extent of physical damage. For example, trawling and dredging directly affect benthic habitats and communities. Thus fishing may be limited to a level in which the total area trawled in any year does not exceed some proportion of the total trawlable area based perhaps on the observed rate of regeneration of the habitat or community. At present there are extreme examples in which the use of fishing gear and practices which damage habitats such as coral reefs are prohibited, i.e. are part of a set of management rules.

Direct effects of fishing on non-target species are also of concern (EEC 1994). Sea birds, turtles and mammals are primary examples which are leading to the imposition of limits on fishing. In the case of turtle by-catch in shrimp trawlers, it is recommended to use turtle excluder devices (TEDs) which increase fishing costs, and reduce shrimp catches (Gibbons-Fly et al. 1994) but may provide the only ecologically acceptable option. The most notable example is the tuna purse seine-dolphin interaction in the east-central Pacific in which catches are limited by an allowable number of dolphin kills annually. This has dramatically affected tuna fishing practices and management procedures (see Annex 3).

2.10.2  Species interactions

The 1982 Convention (UN 1983) considers the potential impact that fishing one resource may have on others. These kinds of impacts are likely to be most pronounced for species that are competitors, predators, or prey of the target species, or are taken as bycatch. Reference points that explicitly recognize and quantify these specific types of interactions, have not been routinely applied in many fisheries. The information requirements generally go beyond the level of knowledge presently available for most marine ecosystems. However, there are various instances in which interaction of coexisting species are considered in setting TRPs.

Predator prey situations have long been of concern to fishery scientists (Clepper 1979). Pauly (1979) emphasises the trade-offs in attempting to exploit both predator and prey. In some situations management has provided for the food requirements of a predator when the prey is harvested. For capelin off eastern Canada a catch LRP of no more than 10% of the spawning biomass has been used. This relatively low rate was selected arbitrarily based on capelin's position in the food chain as a prey species, assuming that there was a relatively high M due to predation by marine mammals and cod. In particular, there was concern that excessive harvesting could negatively impact the production of cod (Shelton et al. 1993). Capelin management in Norway explicitly uses an estimate of the amount of capelin required by cod (Anon 1993). Even in apparently simple two-species systems, the complexities of species interactions may diminish the effectiveness of multispecies approaches. For example, management may provide for a high prey biomass of a small pelagic fish, sprat, for cod, but not allow for the fact that large sprat prey on cod eggs! Nonetheless, it is clear that setting Reference Points for exploitation of prey species should, where possible, consider predators, whether they are exploited or not.

More formal, model-based approaches to taking species interactions into account have included attempts to link single species models together by including terms for the interactions, notably predation. The multispecies model for the North Sea, and subsequent multispecies VPAs for that area are notable examples. A multispecies VPA models the tradeoffs between exploiters of different ecosystem components, but is extremely data intensive. At this time there are few systems for which the data for this potential management approach are available. Even for the Northwest Atlantic (a well studied area), participants at a workshop on ‘Inclusion of fishery interactions in management advice’ agreed that “Large multispecies models…were beyond the scope of the resources and manpower of…[fishery] laboratories” (Mahon 1985). There is no evidence that this situation has changed, and most attempts to incorporate multispecies approaches into fisheries assessment have focused on interactions between two or more components of an ecosystem, and the fleets exploiting them.

An alternative approach to dealing with the effect of species interactions on yield from an assemblage of coexisting species is to combine them into a single analysis which assumes that the behaviour of the combined biomass will be similar to that of a single species. This approach has been used to estimate MSY for coral reef fisheries, in which there are too many species to attempt single species analyses (Medley et al. 1993). It has also been used as a second, limiting, tier to single species management, because when the exploited assemblage includes both prey and predator species, there is reason to expect that the sum of yields from single species assessments will overestimate the combined yield. Therefore, a combined analysis was also used by ICNAF in a two-tiered approach to management, in which the sum of single species quotas was constrained by the yield estimated from the combined stocks (see Mahon 1985).

2.10.3  Ecosystem interactions

A further area of concern regarding sustainability of fishery production is that the significant reduction of biomass of several, possibly keystone, species from an ecosystem will bring about changes in the system which may be precipitous and possibly irreversible. Many ecosystems have exhibited significant changes in response to exploitation, although this has often been confounded with environmental changes and with pollution. Notable examples are the demersal assemblages off the north eastern USA and in the Gulf of Thailand (Saila 1993). Similarly, several studies have noted the effects of fishing on the species composition of coral reef fish assemblages (Medley et al. 1993). The observation that species composition in exploited assemblages tends to shift towards a predominance of smaller, more resilient, but often less valuable species could provide the basis of a diversity-based target reference point for management of multispecies fisheries.

Quantitative descriptors of fishery induced trends in exploited assemblages could be used as reference variables. Saila (1993) suggests multivariate trend analysis methods which can be used to detect trends. Participants could agree upon a desired mix of species based on their size and value, and adjust fishing effort in an attempt to achieve the desired mix. This would constitute a multispecies TRP. The proportion of one or more desirable or undesirable species in the catch could be used to set limit reference points in terms of their minimum proportion in the catch.

The question of precipitous (sometimes referred to as catastrophic) changes in ecosystems has frequently been raised. A major question is whether there are multiple stable states in ecosystems, such that when exploited, the system would move from one stable state to another, but would not move back when exploitation ceased. One such example for northern warm-water lakes is the transition from assemblages dominated by percids to those dominated by smaller, less desirable species, as a function of lake productivity and exploitation (Kerr 1976). On the basis of the information provided by Kerr, it should be possible to propose an LRP for exploitation of such lakes. The large number of northern lakes provided the sample size required to identify this transition point. The marine environment does not usually provide similar opportunities for sample replication. Thus one must either infer from other systems, or learn from experience, by which time it may be too late. However, if there is the expectation that the system to be managed may exhibit a change in states, precautionary LRPs should be adopted.

Another approach to incorporating ecological principles into fishery management involves use of the biomass size-spectrum of living organisms as an indicator of the effect of exploitation (Dickie et al. 1987, Platt and Denman 1978, Caddy and Sharp 1986). Ecosystem characteristics of immediate interest to fishery managers are directly related to the sizespectrum due to inter and intra-specific scaling of physiological processes. Reference points could be based on this ecosystem characteristic, and though less precise than conventional F based reference points, would be more easily monitored.

From a practical management perspective, there is limited management experience with deliberate manipulation of the relative biomass of ecosystem components. Such changes affect the equity of fleets fishing different resources, and requires negotiation between users of different components of the food web (see e.g. Brander and Bennett 1989) prior to selecting species-specific Reference Points for the ecosystem component in question. A current example of an unresolved contention of this type is the tuna purse seine-dolphin interaction in the east-central Pacific where there is disagreement between the ‘users’ of these two interacting resources as to the ideal overall harvest rates (see Annex III).

Concern about ecosystem responses to exploitation extends beyond the effects on fisheries outputs, to include aspects of overall ecosystem health, stability and biodiversity (United Nations 1992, Chapter 17, Norse 1993, Gimbel 1994). In attempting to encompass these concerns, Sherman (1994) takes a broad ecological view of sustainability of biomass yields from marine ecosystems. He promotes the use of the Large Marine Ecosystem (LME) concept as a context for management of renewable resources. In this context, management decisions would have to be viewed in the light of their expected impacts on the entire ecosystem. He cites several LMEs in which an holistic approach to management is being attempted (Yellow Sea, Benguela Current, Great Barrier Reef, NW Australian Shelf and Antarctic marine ecosystems) or is being developed (Black Sea, Barents Sea, North Sea and North California Current ecosystems) (see Sherman et al. 1993 for basic descriptions of these systems). One praiseworthy attempt at whole ecosystem management is found in the Convention on the Conservation of Antarctic Marine Living Resources (see CCAMLR 1993), but despite the clauses in the Convention, many Antarctic finfish resources are severely depleted, largely due to a lack of means to control access.

The CCAMLR convention (Article IIc) explicitly requires a management response to ‘potentially irreversible changes to the ecosystem as a whole’ due to a wide range of possible causal factors. This leads to discussion on what actions are irreversible how to tell, how to tell when an irreversible change has occurred, what elements of the ecosystem are controllable, and to what extent.

For the most part, initiatives towards ecosystem level management are at the stage of developing conceptual reference points or guidelines, and/or defining the research and monitoring needed to address the major questions (e.g. Holling 1993, Apollonio 1994, Sherman 1994). Indeed, there are significant challenges in providing descriptions of ecosystem health, even without reference to the impacts of exploitation (GESAMP 1994). Converting these conceptual approaches to target or limit management reference points which can be routinely used in management decisions will take time and negotiation (e.g. Norse 1993). There are also a number of institutional issues which must be addressed. At both national and international levels, many of the important components of ecosystems cut across issues of health, trade, tourism, transportation, inter alia and are beyond the current terms of reference of institutions solely charged with marine resource management. Nonetheless, as these concepts evolve towards applicability, they can still provide a context within which fishery managers can attempt to understand the effects of their decisions on the systems they are managing (Apollonio 1994).

2.11  AN OVERVIEW OF REFERENCE POINTS

At the time when they proposed F_0.1 as a TRP, Gulland and Boerema (1973) noted that there was no theoretical model for setting annual catch quotas that combines the desirable features of being readily understandable to decision makers; describing and realistically predicting to an acceptable degree of precision the events within the fish stock; and being applicable to a specific fishery without great demands in data and analysis. It is not clear that the situation is greatly improved today.

The population model-derived reference points are technically complex and require considerable quantities of data, usually collected over many years (Table 3). Other less complex Reference Points have also been considered. Many of these are based on population models, or on generalisations from the application of models to many different types of populations. Some, such as mean sizes in the catch relative to size at maturity, can be easily observed with minimal technical expertise and cost. Others must be based on the best available information borrowed from similar fish stocks in other areas. In the absence of more data and sophisticated analyses, such Precautionary Reference Points should be adopted, at least as interim measures, until the necessary assessments have been carried out which allow more precise reference points to be developed.

Limit or threshold reference points indicate that an undesirable stock status is unacceptably close, and that immediate management action is urgently needed. The incorporation of an LRP into a management strategy requires that its responses be pre-established by negotiation among the participants in the fishery and to be acted upon immediately the agreed-upon indices or assessments indicate that the LRP has been reached.

Reference Points related to multispecies or ecosystem considerations are still in their formative stages, but can still provide a valuable basis for management. The challenge is to formulate some of these concepts in terms of Reference Points which can be widely agreed upon and adopted as conventions.