Previous Page Table of Contents Next Page


by G.P. Kirkwood

Renewable Resources Assessment Group

Centre for Environmental Technology

Imperial College of Science, Technology and Medicine

8 Prince's Gardens, London, SW7 1NA, UK.

and A.D.M. Smith

Division of Fisheries, CSIRO

GPO Box 1538, Hobart, Tasmania 7001



A fishery management strategy has three essential components: data pertaining to the fishery (e.g. catch at age data), a method for analyzing the data to produce a stock assessment (e.g. VPA), and a decision rule taking the output from the assessment and translating it into a specification of a technical management measure (e.g. an F0.1 TAC). Even though a management strategy may incorporate elements that are intended to make it precautionary, so that it may be deemed precautionary in principle, it does not necessarily follow that it will actually be precautionary in practice. This can arise through deficiencies or uncertainties in any one of the components of the strategy. In this paper, we discuss how the degree of precaution in a management strategy can be assessed quantitatively.

The first step in evaluation of a management strategy is to identify appropriate performance criteria for determining how well it meets the management objectives for the fishery. In a precautionary setting, these will include, but not be restricted to, its performance in maintaining the stock (or stocks) above critical biological reference points, such as spawning stock threshold levels. Evaluation then proceeds by repeatedly simulating the application of the management strategy to a fishery, where the underlying dynamics of the stocks are governed by a variety of specified operating models, and the simulated data for use in the assessment part of the strategy have specified statistical properties. The operating models will normally be much more complex than the dynamics models implicitly or explicitly assumed in the assessment method, and the different operating models used should reflect the full range of plausible hypotheses about the true dynamics of the stocks. The degree of precaution in the management strategy can then be assessed through examination of how well it meets the performance criteria.

Use of this methodology is illustrated with two examples. The first describes the approach adopted by the International Whaling Commission (IWC) during the development of its Revised Management Procedure. The second describes how these methods have been used in an ecosystem setting by the Commission for the Conservation of Antarctic Marine Living Resources (CCAMLR) for determining precautionary catch limits.


Recent reviews by FAO of the state of the world's fishery resources have amply shown that most stocks either are fully or over-exploited, and that a number of fisheries have collapsed, either biologically or economically. Much of the blame for the parlous state of the fisheries must be laid at the door of either inefficient or over-optimistic management. Rectifying the situation requires modification of current management practices, in the direction of making them more cautious.

Developed originally in the early 1970's in West German legislation as the “Vorsorgeprinzip” in the context of controlling pollution, the precautionary principle has been widely endorsed in a number of international fora. While there is no single accepted definition of the precautionary principle, it generally is taken to have two main attributes: (i) where there are risks of serious or irreversible environmental damage, regulatory action to alleviate this risk is required even in the absence of full scientific certainty that it will occur; and (ii) reversal of the burden of proof, that burden being placed on those who contend that there will be little or no environmental impact.

Although the applicability of this principle to potential pollution problems is fairly clear-cut, its strict applicability to fishery management has been rather more controversial. As discussed by Garcia (1994), the precautionary approach introduced in the UNCED RIO declaration softened the strong requirements of the precautionary principle by recognising differences in local capabilities and the need for alleviatory measures to be cost-effective. These ideas were developed further by Garcia (1994), who proposed a set of characteristics that a precautionary approach to management should have.

One of the tasks of this meeting is to amplify these ideas and to further develop practical guidelines for a precautionary approach to management. These, of necessity, will take the form of general principles. However, even if they are followed, and a management strategy may be deemed to be precautionary in principle, there remains the question of whether it is actually precautionary in practice. In this paper, we address the issue of how the degree of precaution in a management strategy can be assessed. For the most part, we restrict our attention to precaution in respect to biological conservation. In the discussion we suggest a wider view of what it means to be precautionary.

The structure of the paper is as follows. In section 2, we discuss a number of general issues related to the identification and selection of criteria for assessing the degree of precaution. Section 3 outlines the characteristics of management strategies. In section 4, the procedures needed to evaluate the performance of a management strategy are described. Section 5 contains two practical examples. The first of these briefly describes how the procedures for evaluating the performance of a management strategy were applied by the International Whaling Commission (IWC) in the development of its Revised Management Procedure. In the second, the approach taken by the Commission for the Conservation of Antarctic Marine Living Resources (CCAMLR) to ecosystem management is briefly outlined. In the final discussion, we draw some general conclusions from the approaches we have presented, suggest a wider view on precautionary management, and briefly allude to how this wider approach could be evaluated.


In seeking criteria and methods for assessing how well a fishery management strategy performs in managing a fishery, it is essential to look first to the management objectives, for it was to meet these objectives that the management strategy should have been designed. It follows immediately, therefore, that the first place to seek evidence on the precautionary nature of a management strategy is in the management objectives themselves. Making any judgement at this level requires at the very least that the objectives are explicitly stated. Objective assessment of performance of strategies in meeting the objectives requires further that the objectives are also stated or subsequently interpreted in a sufficiently quantitative way that the extent to which they are being met can be determined unambiguously.

In practice, it is difficult enough to extract from managers any explicit statement of objectives, and in most cases where they are available, the objectives take the form of motherhood statements such as “ensure the long term conservation of the stocks”. No analysis of objectives like these can discern whether a management strategy designed to meet them is precautionary or not. The first step, therefore, in seeking to encourage a precautionary approach to management is to advocate incorporation of an explicit recognition of the approach into revised management objectives. An example of objectives that contain aspects of a precautionary approach is found in the convention governing the operations of the Commission for the Conservation of Antarctic Marine Living Resources (CCAMLR), where the ideas of reversibility over fixed time scales and biological reference points are explicit. The approach taken by CCAMLR is discussed further in a later section of this paper. Note that mere insertion of the words “precautionary approach” into a statement of objectives is not sufficient to ensure precautionary management.

Evaluation of how well a management strategy performs in meeting the management objectives and how precautionary it is also requires the identification of performance criteria. These can be either qualitative or quantitative. Considering first qualitative criteria, one useful approach is to assess the management strategies against a checklist of desirable and undesirable properties. For example, on the positive side a properly precautionary management strategy must at least incorporate continuing collection of appropriate monitoring data and it must have sufficient flexibility to allow a quick reaction to any signs that something is going wrong. On the negative side, in most cases a strategy that controls a fishery solely by setting total allowable catches and that has no mechanisms for controlling fishing capacity is not precautionary. A very useful first attempt to develop practical guidelines for precautionary management was included in Garcia (1994), and one of the aims of this meeting is to progress further with that approach.

While it is possible to progress quite a long way in defining characteristics of a precautionary fishery management strategy by using qualitative criteria, the extent to which they actually succeed in being precautionary can only be properly assessed quantitatively, and this requires identification of quantitative criteria. A good example of this process is that adopted by the International Whaling Commission (IWC) during the development of its revised management procedure. As discussed further below, this included quantitative interpretation of general management objectives and identification of a range of statistics that allowed measurement of the performance of a management strategy in meeting each objective in a simulation study.

A key step in identifying performance criteria for assessing degrees of precaution is to establish both targets and threshold levels. For many years, there was only one management target: MSY. Fortunately, this target, expressed in terms of catches, has largely gone out of fashion (Larkin, 1977), although its ill-defined cousin, the “optimal” sustainable yield, apparently still persists. Nearly all management objectives retain the goal of seeking high sustainable catches, and the extent to which this is achieved is obviously one criterion for judging the performance of a management strategy.

Recent attention, however, has focused more on thresholds or biological reference points, usually expressed in terms of population levels, below which there is an unacceptably high risk of stock collapse. With these, an additional management constraint is imposed, in that exploitation is allowed only to the extent that the probability of the population falling below the threshold levels is acceptably low. Identification of biological reference points has received considerable attention in the recent literature, and it has been the subject of a major international conference (Smith et al, 1993). The connection between biological reference points and management strategies has been further discussed by Mace (1994), and in the context of a precautionary approach to management, the topic is discussed in detail in a separate background paper (Rosenberg and Restrepo, 1995).

A familiar application of this approach has been to set a threshold level on spawning stock biomass. For example, an oft-quoted rule of thumb sets this threshold at 20% of the unexploited spawning stock biomass. The degree of precaution may then be assessed in terms of the probability that the spawning stock falls below the 20% threshold in a fixed period. One may also look for specific remedial action to be triggered by a precautionary management strategy (e.g. immediate cessation of catches or at least substantial reductions in them) if the threshold is inadvertently crossed. The difficulty with this approach when it is applied across a range of species or stocks is that it can be largely arbitrary. For some species, the 20% threshold may involve far too much risk of stock collapse, while for others, it may be more cautious than is necessary, thereby leading to lower catches than could safely have been taken. It is far better, of course, if the threshold can be determined directly from historical data for the stock being managed, e.g. from historical stock-recruitment data.

Intimately associated with the idea of thresholds is that of risk1. Continuing with the theme of a spawning stock threshold, obviously the lower the probability of the spawning stock falling to below a critical threshold, the lower is the risk of stock collapse and the greater is the degree of precaution. It is important, however, to realise that the probability of exceeding a threshold is not the same as the risk of stock collapse, other than in the exceptional case when the threshold corresponds to a point of critical depensation, below which stock collapse is certain.

This distinction becomes important when a threshold is explicitly included in a management strategy, such that the size of (say) the spawning stock is assessed regularly, with pre-determined remedial actions to be triggered if the threshold is crossed. In such cases, the true risk of stock collapse can only be determined properly on a stock-by-stock and a strategy-by-strategy basis, since the risk depends heavily on the dynamics of the stock, the variability in recruitment, the information content of the past and likely future data, the types of triggering action to be taken, and the timescales involved.

The approach adopted by the IWC in its revised management procedure provides an interesting example of this: the threshold for stock numbers was set at the essentially arbitrary level of 54%, and catch limits were to be set to zero by the management strategy whenever the stock numbers were assessed to fall below this threshold. Simulation studies were then used to determine that there was an acceptably low probability of inadvertently continuing to allow catches when the stock was actually below the threshold but assessed to be above it. With such a high threshold, this approach clearly is strongly precautionary, but it is also highly conservative. That may be appropriate for whales, where the present climate of opinion would probably favour no commercial catches to be taken at all, but it would be much less so for many fish stocks. Balancing risks, thresholds and commercially acceptable catch levels is rather more difficult.

This last comment leads to our final point in this discussion of criteria for assessing degrees of precaution. In many fisheries, biological thresholds will be set at relatively low stock levels, and there may be little argument that if these thresholds are crossed, then drastic action is needed to allow the population to recover. What is less clear, however, is the role that precaution should play when a stock is assessed to be overexploited, in the sense that higher catches could be taken sustainably if the stock were allowed to increase, but where the stock is not at so low a level that a biological threshold has been crossed. In the earlier case, the interests of the stock should obviously be paramount, but in the latter case there is a need also to take account of the welfare of the fishermen, who would inevitably suffer in the short term from the reduction in catches needed to achieve long-term increases. Many management strategies attempt to take partial account of this by incorporating restrictions on the speed at which allowable catches can change. These additional restrictions, however, themselves have implications for the risks associated with breaching a threshold: if only small changes are allowed each year, it may not be possible to react quickly enough in the face of a declining stock to prevent it from reaching unacceptably low levels.

1 In this paper, we use the term “risk” to denote the probability of something bad happening. It is not used in its technical sense in a decision theoretic framework as meaning expected loss. Detailed discussion of risk in its technical sense is given in Huppert (1995)


In identifying, and later evaluating, fishery management strategies, the first problem is to decide how broadly or how narrowly to define the problem. In most western industrialized fisheries, there is a management system already in place with a well, defined and highly regulated fishing fleet, regular collection of fishery data, and so on. In many of these fisheries, the fishery management problem is conceived quite narrowly, generally revolving around issues of limiting fishing mortality (e.g. via quotas or other technical measures) to protect stocks while at the same time attempting to maximize economic returns from fishing. At the other extreme perhaps are artisanal fisheries in developing countries, where the fishery management problem is inextricably linked with much broader socioeconomic processes and changes in the country as a whole. In this paper, the main focus will be on the former situation rather than the latter, though application of the methods and approaches discussed here to the broader socioeconomic problems will be discussed briefly in general terms.

Taking for the moment the narrower focus on typical western industrialized fisheries, a management strategy or management procedure may be thought of as a set of rules that specifies the technical measures that will be adopted (over time) in managing the fishery. Technical measures here refer to the sets of management regulations on the fishery, such as limits on entry to the fishery, catch quotas, gear restrictions, size limits, and closed seasons and areas. An example of a management strategy for a quota managed fishery might be to set the catch quota in each year to try to achieve a reference fishing mortality rate (e.g. F0.1). This in turn requires information on stock abundance and selectivity, which might be obtained from analysis of catch at age data, using for example virtual population analysis (VPA).

The preceding example illustrates that there are usually three ingredients to a management strategy, namely data (e.g. catch at age), a model or method to analyse the data (e.g. VPA), and a decision rule for taking the output of the data analysis and translating it into the specification of a technical measure (in this case application of F0.1 to derive a catch quota)3. Another example might be use of growth and mortality data in a yield per recruit analysis to set a minimum size limit. Thus, the general pattern is

2 Throughout the literature, the terms “management strategy” and “management procedure” have been used synonymously. In this paper, while we mainly use the term “management strategy”, we revert to the alternative when not to do so would be awkward (e.g. when discussing the IWC's revised management procedure)

3 Some management strategies effectively use only data and a decision rule, with no formal assessment model in between. An example is the management procedure for South African anchovy described by Butterworth and Bergh (1993), where results from a biomass survey are fed directly into the decision rule

strategy = data + model + decision rule -- > technical measure

Choice of different combinations of data, model and decision rule quickly generates a potentially very large number of strategies for managing fisheries. In practice, most of the variety has come from developments of stock assessment methods that use different combinations of models and data, and rather less attention has been paid to developing new decision rules. The latter may be categorized in various ways and a brief typology follows, with a particular eye on issues relating to precautionary management.

One categorization of decision rules is by the extent to which they use future information. Those that do not use future information may be called non-feedback or non-adaptive strategies. They use data up to the present combined with a model to generate decisions into the future that will not be altered by future data or analysis. For example Francis (1992) considered a series of TAC scenarios for management of an orange roughy stock over 10 years4. Each scenario involved a fixed schedule of TACs over the period considered. By virtue of their inability to take account of future data or analyses, such non-feedback strategies are intrinsically non precautionary. They are also quite rare in modern fishery management.

Strategies based on decision rules that use future information are called feedback or adaptive strategies. In the above example, incorporation of feedback would involve simply setting a TAC for the next year, then updating the assessment with the new information collected during that year and setting the following year's TAC on the basis of the revised assessment, and so on into the future. In general, such feedback strategies ought to be more precautionary than non-feedback strategies, to the extent that they will be better able to respond to changing circumstances (such as evidence of overfishing).

The class of feedback or adaptive strategies may be further categorized by whether they take active account of future learning. The above example fits into the category of passive adaptive strategies, in that while it is adaptive to the extent to which assessments are updated as new information comes in, no specific account is taken in the current year that further learning opportunities will arise in later years. Walters (1986) has dealt at considerable length with active adaptive or experimental fishery management strategies. These strategies recognize that in many circumstances there is a relationship between the way a stock is managed and how well its dynamics are understood. There can therefore be a (longer term) value in deliberately “perturbing” the stock away from its current level (which may appear to be optimal given current information) to test its productive potential at either a higher or lower stock size. In most instances such strategies have been evaluated with respect to their performance in maximizing long term catches, and not in relation to risk of overfishing. Clearly, strategies that call for increased exploitation rates in the short term will be less precautionary. However, experimental strategies need not in general be less precautionary, and in the longer term they may lead to better performance with regard to overfishing.

An important subset of control laws in fisheries management are those that are stock-size-dependent. These strategies prescribe catch quotas as a function of current estimates of stock size. Three examples of such control laws are the constant catch, constant harvest rate and constant escapement strategies (Hilborn and Walters, 1992). In terms of maximizing catches over time, the constant escapement strategy can be shown to be optimal under some restricted sets of conditions, and it generally performs better with regard to minimizing the risk of low stock sizes. However these performance attributes come at the expense of very high inter-temporal fluctuations in catch, which can lead to poor economic performance. The range of strategies based on the fishing mortality rate (e.g. F0.1, Fmed etc.) which have been proposed and used in fisheries management fall into the class of constant harvest rate strategies. The comparative performance of these types of control laws in the presence of uncertainty has been examined further by Frederick and Peterman (1995), who found that appropriate adjustments to the harvest policies to take account of uncertainties in estimates of stock abundance and biological parameters can vary widely with the stock and harvest policy considered.

4 Although these 10 year fixed schedules were used for evaluation purposes, in this case there was no intention that they were to be set in place without the ability to revise them based on future information

More complex variants of the stock-size-dependent strategies are easily established. One variant is to apply one of the basic strategies described above, but to limit either the absolute or proportional changes in catch quota from one period to the next. As noted earlier, to the extent that this limits the flexibility to reduce catches quickly in the face of declines in stock size, such variants might be regarded as inherently less precautionary.

Another variant is to make the decision (the catch limit to be applied) a function not only of the current estimated stock size, but also of the degree of uncertainty in that estimate. A strategy that is definitely precautionary is one that deliberately sets lower catch limits in the face of higher uncertainty. Note, however, that the appropriate extent of downward adjustment of catch quotas in the face of uncertainty will vary from case to case (Frederick and Peterman, 1995).


As noted above, there are several prerequisites to a quantitative assessment of how precautionary a particular management strategy might be. This assessment is no different in kind to an assessment of how well a particular strategy achieves any of its objectives. The steps involved include (i) obtaining a clear statement of objectives, (ii) identifying one or more performance criteria related to each objective, and (iii) clearly specifying the management strategy to be evaluated. The specification of the management strategy must include the stock assessment method to be used (a combination of model and data) and the control law or decision rule that specifies the technical measure to be chosen at each decision point. Once this information is available, the performance of the management strategy can then be evaluated.

The only really reliable way of evaluating a fishery management strategy is to apply it in practice to the real fishery for which it was intended. Where, however, there is a range of potential strategies to be evaluated, for example, this obviously is completely impractical. The only alternative is to seek a “laboratory world” in which to test the strategies. This usually involves developing a computer simulation model of the fish stock and fishery, which not only allows the consequences of management actions specified by the management strategy to be determined, but also simulates the “data” used in the stock assessment part of the management strategy. This underlying simulation model used for the evaluation is often called the operating model.

The operating model should not be confused with the dynamics model implicitly or explicitly contained in the stock assessment. Normally, the operating model will be far more complex than the stock assessment model, as it is designed to mimic the real world as far as possible and to allow examination of the consequences of failures of the assumptions in the stock assessment model. Thorough evaluation of a management strategy will usually require testing it with a range of operating models in order to cover the full extent of uncertainties about the real world in which the management strategy will be applied. Inevitably, this will mean that in some cases the dynamics of the simulated fish stock and the associated biological parameters may be quite different from those assumed by the stock assessment.

The issue of uncertainty is intimately linked to the notion of precaution and risk. In predicting the consequences of a management strategy using the above approach, a number of sources of uncertainty must be recognized and dealt with. One of these uncertainties is data error or observation error. All fisheries data, be they catch, effort, survey abundance, or age and length data, have measurement error associated with them. Random error in such data is relatively easy to deal with, through specification of an appropriate observation equation. Systematic error or bias is generally harder to detect and deal with. In contrast to measurement error in data, process error is used to represent uncertainty in the underlying dynamics of the resource. For example, recruitment is frequently represented as a stochastic process, although an underlying relationship with parental stock size is often assumed.

Both observation and process error have been considered fairly extensively in the fisheries literature in relation to parameter estimation and stock assessment. Much less attention has been paid to the more general issue of model uncertainty. This may take the form of uncertainty about parameters of the model, or more generally of uncertainty in the form of functional relationships or major structural assumptions of the model.

Sainsbury (1991) has gone further than most in explicitly dealing with model uncertainty in the context of evaluating fishery management strategies. The example he investigated involved the assessment and management of a tropical multi-species fishery where the exploitation history had resulted in considerable changes in species composition of the catch over time. Sainsbury identified four alternative hypotheses or models that could account for the observations, and then designed and evaluated a number of experimental management regimes to distinguish between the alternative models. The evaluation dealt not only with model uncertainty, but also with process and observation error and parameter estimation within each model. The experimental approach used spatial contrasts in the fishery effected through spatial closures.

One of the principal benefits of evaluating management strategies using simulation testing lies in the ability to quantify the sensitivity of the results to the various sources of uncertainty identified above. This generally involves testing the performance of the strategy in a series of robustness trials. This approach is more fully described in section 5.1 below, and further examples may be found in Punt (1992) and Butterworth and Bergh (1993). The robustness trials are essentially alternative specifications of the operating model. However, one of the judgements to be made in using this approach to evaluate the performance of management strategies lies in the selection of an appropriate set of robustness trials. This is because it will almost always be possible to identify some hypothesis for an underlying “reality” that will cause any management strategy to fail. The question then becomes one of assessing how credible such a hypothesis might be.

Punt and Smith (1995) offer some qualitative guidelines for the selection of hypotheses for testing robustness. They note first that hypotheses should be framed in a way that allows them to be tested through data analysis, at least in principle. They then suggest four criteria for ranking a set of alternative hypotheses:

  1. How well do the data support each hypothesis for the species and region under consideration?

  2. How well do the data support each hypothesis for a similar species or another region?

  3. How well do the data support each hypothesis for any species?

  4. How strong or appropriate is the theoretical basis for each hypothesis?

In general, those hypotheses that are supported higher up the list should be given higher weight or credibility. Clearly there is still scope for subjective judgement in this approach. For example, what criteria are used to judge similarity between species (life history characteristics, taxonomic proximity, functional ecological similarity)? Nevertheless, the idea of a hierarchy of support for ranking hypotheses seems to be a useful step forward.

The output from an evaluation of management strategies using simulation testing can take various forms, but it is often usefully summarized as a decision table (see, for example, Hilborn and Peterman, 1995). For the evaluation of a single harvest strategy, such a table might present the performance for each criterion against each of a set of selected robustness trials. For those performance criteria that seek to measure or represent precautionary performance, this provides a measure of how well the strategy achieves precautionary objectives (for example, how often critical thresholds are exceeded). The performance across robustness trials also helps to identify critical uncertainties.

Where several alternative management strategies are being considered, an alternative form for a decision table presents a summary (across robustness trials) of performance for each strategy against each key objective. This may be a much more useful table for decision makers as it allows a ranking of alternative strategies for any one objective (such as a conservation objective) and therefore provides a basis for decision making irrespective of the absolute level of performance. More particularly, it also provides a basis for quantifying the tradeoff to be made between alternative and conflicting objectives. This will be particularly important in developing a practical approach to precaution, where often there will be a clear tradeoff between conservation and economic objectives. This approach allows a quantitative assessment of that tradeoff. Decision makers can assign their own subjective rankings to different objectives, and choose the strategy that best (or adequately) addresses their primary objectives. Note that this approach (evaluating alternative strategies against a range of objectives) is quite different from an optimization approach that seeks the single strategy that performs best for an explicitly weighted set of objectives.


In this section we discuss two examples. The first, dealing with the revised management procedure of the IWC, illustrates how the methodology of the previous section has been applied to evaluate the performance of a set of alternative management strategies. The second, rather different, example shows how precautionary aspects have been explicitly incorporated into the management objectives of CCAMLR and describes the approach that body has taken towards setting precautionary catch limits.

The revised management procedure of the IWC

Before its decision in 1982 to declare a moratorium on commercial whaling, the catch limits set by the IWC were based on stock assessments, developed by its scientific committee, which were very similar in nature to standard fishery assessments at the time. In essence, for each stock all the available data were used to obtain best estimates of current and historical stock sizes and of the productivity of the stock. Catch limits were then set with the aim of keeping the stock at or above the level at which the MSY could be taken, or moving it towards that level. One of the major reasons for deciding to impose the moratorium was the difficulty experienced by the scientific committee in reaching consensus on the status of stocks, given the prevailing uncertainties in the data and in their interpretation.

During the late 1980's and early 1990's, the scientific committee of the IWC developed a revised management procedure designed to resolve these difficulties. The development process involved a reexamination of management objectives, taking a realistic view of the uncertainties inherent in current and likely future data and in the baleen whale dynamics, and a very thorough testing of the robustness of proposed procedures to these uncertainties. Although a precautionary approach was not explicitly considered, the way in which the revised management procedure was developed and tested gives an ideal illustration of the methodology described in the preceding section of this paper.

The first step in the process involved identification and quantification of the IWC's management objectives. After much discussion, the following (brief) statement of its objectives was agreed:

  1. stability of catch limits;
  2. acceptably low risk of stock depletion to below 54% of carrying capacity;
  3. making possible the highest continuing yield from the stock.

The IWC agreed that a management procedure must first satisfy objective (ii). Subject to that, it was then free to maximise catches under (iii), while performing satisfactorily in (i). A stock assessed to be below 54% of its carrying capacity (i.e. below the protection level in the previous IWC management procedure) should have a zero catch limit. Acceptable risk was then to be judged in terms of the likelihood of inadvertently setting non-zero catch limits when the stock was actually below the protection level, but was assessed to be above it.

For a revised management procedure to be acceptable, it must be able to meet the above objectives, regardless of existing and continuing uncertainties in the basic data, stock structure and dynamics of whale populations. Whether or not a procedure is robust to these uncertainties can only be decided by examining its performance across a wide range of plausible situations. Development of a revised management procedure therefore proceeded on two fronts: identification and refinement of potential procedures, and specification of means for testing the performance and robustness of these procedures in meeting the objectives. The procedures themselves are not relevant to this paper; interested readers are referred to IWC (1992, pp 93–103), in which descriptions are provided of the five procedures as they stood at the time a choice was made between them. Rather, we shall concentrate on the methods used to test their performance.

Simulation trials of management procedures

Since experimental application to management of actual whale stocks was out of the question, the approach taken was to simulate the management of whale stocks. An initially unexploited whale population was set up and subjected to a series of historical catches before the onset of management. The dynamics of the simulated stock were governed by specified operating models. Estimates of abundance, which in addition to the historical catches formed the primary input data source for whale stock assessments, were simulated so that they had the same nature and properties believed to occur in real data of those types. Computer programs implementing potential management procedures were then applied to this simulated stock.

A lengthy series of computer-based trials was then conducted. Each trial examined management of a simulated whale stock over a 100 year period. This was repeated at least 100 times for each trial scenario. Summary statistics monitoring the performance of the procedure in relation to the three management objectives were collected for each trial.

Two categories of trials were identified: “base case” trials and “robustness” trials. The base case trials consisted of a short series of relatively mild trials. These examined the ability of procedures to manage both unexploited, moderately depleted and heavily depleted stocks of whales with different levels of productivity, in cases where the dynamics of the stocks followed conventional models and the abundance data were unbiased. The scenarios covered in these trials were typical of those now often examined in sensitivity tests carried out in association with fish stock assessments, though such tests rarely include future application of a management procedure.

The key to the performance evaluation lay in the additional robustness trials, which examined a much wider range of possible departures from assumptions than is normally considered in sensitivity tests. Each trial was repeated for a selected subset of the base case scenarios on initial abundance and stock productivity. The robustness trials included the following (for a full list, see IWC 1993, p. 224a).

  1. Incorrect assumptions about the dynamics of the true stock. This formed the largest category. Cases examined included widely differing forms of density dependent responses, differing biological parameters, trends and cycles in the carrying capacity of the population, and cyclic changes in productivity.

  2. A wide range of initial abundance levels.

  3. Upward and downward bias in the abundance data, and trends in that bias, as well as differing frequencies of collection and levels of precision.

  4. Uncertain or inaccurate catch histories before exploitation, and long periods of protection before management starts.

  5. Irregular episodic events (e.g. occasional occurrence of epidemics).

  6. Deterioration of the environment, with declining trends in both carrying capacity and productivity.

A further set of trials examining interactions between a subset of these factors (those that were most important on their own) was also carried out.

Statistics for evaluation of performance in meeting management objectives

For each trial, statistics allowing evaluation of the performance of management procedures in meeting the three management objectives were collected. The primary statistics and the management objectives to which they referred were:

Objective (i): the average inter-annual variability in catch limits;

Objective (ii): percentiles of the lowest population size during the 100 years of management;

Objective (iii): percentiles of the total catch over 100 years, and of a measure of “continuing” catch, which in most cases was the average catch over the final 10 years.

Similar statistics were also collected for the final population size after 100 years. This was effectively used as a proxy for a target population size.

To assess the probability of whaling being inadvertently allowed when stock levels were below the protection level of 54% of carrying capacity, two further sets of statistics were used. These were percentiles of the “realised protection level”, which was the lowest population size in a trial at which non-zero catches were set, and of the relative degree of recovery, which compared the time it took to recover to the protection level under the management procedure being tested with the time it would have taken had zero catches been set. These last two statistics, or slight modifications of them, would be ideal for evaluating performance in the presence of a biological reference point or threshold. They are complementary in nature, because while a management procedure may inadvertently set non-zero catch limits at stock levels below the protection level, they may be so small that any delay in stock recovery is also very minor.

Selection of the revised management procedure

By 1991, the development and testing process for management procedures applicable to single stocks of baleen whales was completed. At its 1991 meeting, the Scientific Committee reviewed the large set of performance statistics on the trials of each of the five procedures (IWC, 1992b). All procedures considered were found to have performed satisfactorily on the simulation trials. The best performing procedure (developed by Dr Justin Cooke) was subsequently adopted by the IWC.

Implications for assessing degrees of precaution

The two key features of the process adopted by the IWC were that all elements of the management strategy were tested simultaneously and that robustness was examined to a much wider range of uncertainties than is normally considered.

The results of the trials showed clear interactions between the precision and quantity of data and the degree of conservatism needed to meet the objectives. These proved to be quite nonlinear, further amplifying the findings of Frederick and Peterman (1995). A valuable aspect of the best-performing procedure was that it incorporated a mechanism for automatically adjusting the catch limit in line with the apparent precision of the assessment. This is not a new suggestion, but the important role it played in ensuring good performance suggests that this may be a design feature that should be included among the characteristics of a precautionary management strategy.

The equivalent of the stock assessment method used in the best-performing management strategy involved fitting a simplified production model by Bayes-like techniques. By itself, this carries no particular connotations for other fisheries, since whales have rather different dynamics to fish, but in this case it was found that increasing the apparent realism of the underlying dynamics of the model would not necessarily improve the performance (cf. Ludwig and Walters, 1985). This is good news for fisheries for which data availability is relatively low, since it provides an example where robust precautionary management can be achieved without having to rely on the data-hungry types of stock assessment typically used for temperate western industrialised fisheries.

The results of the robustness trials strongly emphasised the distinction brought out in an earlier section between a strategy that was precautionary by design and one that was precautionary in performance. Both the final and earlier versions of each of the five potential whale management procedures were precautionary by design. They clearly differed, however, in the degree to which they exhibited precautionary (conservative) performance. Futhermore, this difference in performance itself varied across the robustness trials. In particular, most performed relatively well when faced with the base case trials, in which the dynamics and the data satisfied most of the usual assumptions made in previous whale assessments. Not surprisingly, much greater differences were observed in robustness trials where the assumed properties of the data differed substantially from what was expected, and where the underlying dynamics was quite different from that implicitly assumed by the procedures. The clear lesson was that the true degree of precaution of a management strategy cannot be determined just from an analysis of the management objectives and the structure of the management strategy alone.

Precaution in an ecosystem setting - CCAMLR

Two important steps identified for increasing the precautionary nature of fishery management are incorporation of explicit precautionary elements into the management objectives for a fishery, and taking proper account of the ecosystems affected by capture fisheries. In our second example, we briefly outline the steps taken by CCAMLR to address these issues. In particular, we highlight the incorporation of precautionary aspects into the CCAMLR convention and the ways these have been translated into management strategies, particularly for management of lower trophic level species.

The CCAMLR convention is unique, in that it explicitly attempts to address ecosystem management. Paragraph 3 of Article II of the convention states:

“Any harvesting and associated activities in the area to which this Convention applies shall be conducted in accordance with the provisions of this Convention and with the following principles of conservation:

  1. prevention of decrease in the size of any harvested population to levels below those which ensure its stable recruitment. For this purpose its size should not be allowed to fall below a level close to that which ensures the greatest net annual recruitment;

  2. maintenance of ecological relationships between harvested, dependent and related populations of Antarctic marine living resources and the restoration of depleted populations to the levels defined in sub-paragraph (a) above; and

  3. prevention of changes or minimization of the risk of changes in the marine ecosystem which are not potentially reversible over two or three decades, taking into account the state of available knowledge of the direct and indirect impact of harvesting, the effect of the introduction of alien species, the effects of associated activities on the marine ecosystem and of the effects of environmental changes, with the aim of making possible the sustained conservation of Antarctic marine living resources.”

When interpreting this Article, it should be noted that Article II of the Convention includes rational use among the meanings of the term “conservation”.

This statement of objectives explicitly includes the idea of biological reference points, the concepts of risk and of reversibility of changes over a specific time span, and a requirement to take account of the state of available knowledge in assessing risks and reversibility. It furthermore requires account to be taken of effects of harvesting on both the population being harvested and on dependent and related populations. By any measure, these objectives have strongly precautionary aspects, though the term “precautionary” does not appear specifically.

The need to address conservation of the whole Antarctic marine ecosystem, rather than of just the species that would be harvested directly, arose because of the nature of the Antarctic ecosystem and of the potential fisheries. Around the time the CCAMLR Convention was being negotiated, there was a popular view that a potentially huge harvestable surplus of krill (Euphausia superba) existed in the Antarctic, resulting from the heavy overexploitation of baleen whale stocks there. Consequently, there was strong interest in developing a substantial krill fishery.

Fortunately, the more extreme versions of the krill surplus theory soon lost currency and wiser views prevailed (e.g. May et al, 1979). Krill, near the base of the Antarctic food chain, is the key species in the ecosystem on which nearly all other species are either dependent or related. Clearly, if a major krill fishery did develop, there was a considerable risk that it could have a substantial effect on these other species.

One of the major thrusts in response to this by CCAMLR has been the setting up of a comprehensive ecosystem monitoring programme, concentrating on key krill predators, to which most member governments contribute. In this programme, selected biological parameters are monitored using standardised methods at sites around the Antarctic. A number of species of penguins, flying birds and seals are monitored in this programme. Individual member governments also conduct research programmes aimed at evaluating and improving the utility of the biological parameters being monitored, and providing the background information needed to interpret changes in the monitored parameters.

Such monitoring programmes take considerable time to set up, and often quite long time series are needed before any apparent changes can be properly interpreted. It is therefore perhaps fortunate that technical and other marketing difficulties have so far delayed the anticipated development of a large krill fishery. Current krill catch levels are believed to be much lower than those that may have deleterious effects on the ecosystem. Despite this, CCAMLR has taken the step of imposing a series of precautionary catch limits for krill, which are much larger than current catch levels, in preparation for any future increase in krill fishing.

The precautionary catch limits for krill were based on application of a krill management strategy. This strategy incorporates an explicit single species biological reference point and an additional ecosystem constraint, with precautionary TACs being determined using simulation studies, based on a krill yield model, that were similar to those conducted by the IWC described above. The management strategy is designed for use with previously unexploited (or very lightly exploited) stocks, for which an estimate of pre-exploitation biomass is available. Details of the computational and simulation methods in the krill yield model are given in Butterworth et al (1994). This approach has also recently been adapted by Constable and de la Mare (1994) to calculate precautionary TACs for the myctophid Electrona carlsbergi.

In the management strategy, if B0 is the estimated pre-exploitation biomass, then the precautionary TAC is set as:

TAC = &agr; B0

The value of &agr; to be used is the minimum of &agr;1 and &agr;2, where, based on simulation studies using plausible operating models,

  1. &agr;1 is the value such that under a constant TAC of &agr;1 B0, the probability of the spawning biomass falling below 20% of its pre-exploitation level over a 20 year period is 0.1; and

  2. &agr;2 is the value such that under a constant TAC of &agr;2 B0, the median escapement over a 20 year period is 75% of B0.

The first constraint is the now-common single species constraint on the probability of falling below a biological reference point in a given time span. The second is quite different; it is aimed to leave at least some of the prey for other predators. The biological reasoning for this is as follows. A standard single species production model that completely ignores the interests of the prey, such as the Schaefer model, suggests that the population level at which MSY can be taken is around 50% of the pre-exploitation level, so that the “optimal” single species escapement from the fishery would be 50% of B0. If all the prey were to be reserved for the predators, than the appropriate escapement from the fishery would be 100% of B0. The figure chosen, 75% is halfway between these.

Clearly, the 75% figure chosen is largely arbitrary and the biological underpinnings are not strong. As further information is accumulated on the dynamics of both the prey and predator species, the ecosystem constraint will be refined. However, the principle by which account can be taken explicitly of dependent species seems a very good one and well worthy of consideration under the umbrella of a precautionary approach to management of harvested prey species in a marine ecosystem.


A strong distinction has been made in this paper between management strategies that are “precautionary in principle” and “precautionary in practice”. The former can often be judged quite qualitatively, for example using guidelines similar to those outlined by Garcia (1994). For a management strategy to be precautionary in practice, it is probably necessary for it to be precautionary in principle, but it is definitely not sufficient.

The results of simulation testing of baleen whale management strategies described earlier provided very clear examples of this distinction. In addition, recent simulation testing of management strategies for developing fisheries has further emphasised the distinction between precaution in principle and in practice (R.I.C.C. Francis, A.D.M. Smith and S.E. Wayte, unpublished data). This study evaluated a feedback management strategy for a new fishery which was deliberately precautionary in nature. In particular, an explicit element of the feedback decision rule was to select, as the maximum current catch, that catch which resulted in less than a 10% chance of reducing the stock to less than 20% of virgin biomass across a series of projected future catch trajectories. Despite the clear precautionary nature of this element in the decision rule, simulation tests revealed that this strategy resulted in frequent reductions in stock below the 20% biomass threshold. This decision rule clearly did not meet its own (inbuilt) performance criteria.

As noted in section 3, most quantitative evaluations of fishery management strategies to date have been concerned with quota management systems. One aspect of such systems not often considered in such evaluations is the allocation of the catch limit for the stock (equivalent to a TAC) among the fishing fleet(s) and the likely effects of different ways of doing this. Arguably, this is at least as important as the other processes leading up to it. Few would argue, for example, that a completely open access fishery management policy was likely to be precautionary. While it is not axiomatic that open access will lead to overcapacity in the fishing sector, it very frequently seems to be the case. As overcapacity increases, this is likely to lead to increased pressure on the managers to raise catch limits, or at least not to reduce them in cases where that becomes necessary.

A number of alternative technical management measures to open access have been proposed and used in different fisheries around the world, ranging from limited licensing to individual transferable quotas, with varying degrees of success. What is extremely difficult, however, is to disentangle the contributions to this success of the management measure per se from the degree of precaution (or otherwise) of the remainder of the management strategy. In principle, this can be addressed by simulation, only now the behaviour of the fishermen needs also to be included in the operating models. If one adds the further complication of imperfect compliance with management measures (cf. Rosenberg and Brault, 1993), the task of carrying out a full quantitative assessment of the degree of precaution of a management strategy becomes truly formidable.

Evaluation of management procedures using the methods described in this paper is undeniably computationally intensive, though it is only for complex industrialised fisheries that really extensive research on the scale of that conducted by the IWC would normally be necessary or appropriate. It is equally important to realise that, no matter how exhaustive the evaluations by simulation are, there can still be no guarantee that the management strategy studied will turn out in real application to exhibit the same degree of precaution as the simulation studies suggest. That accepted, however, current evidence suggests that even only modestly sized evaluations can provide valuable insights into the degree of precaution in a proposed management strategy that are very difficult to obtain by other means.

Much of the discussion of the precautionary approach to management has focused on a narrow definition of precaution, where the aim is essentially to prevent overfishing leading to stock collapse. A broader view of precaution sees the aim being to maintain a flexible, resilient fishery system, where that system is taken to include the fish stock, the ecosystem of which it is a part, the fishing fleet, and the management agency which regulates it. This view is closely tied to the notion of reversibility, where for example the management system is sufficiently flexible that previous decisions can be reversed without undue delay or cost (both economic and political), where fishermen do not become economically locked into a position where they cannot afford to reduce effort even temporarily, and where the biological system is maintained in a state where irreversible changes are not triggered by overfishing.

This broader definition makes it clear that the precautionary approach does include non-biological considerations. This may be particularly true for artisanal fisheries in developing countries, where the fish may provide the only source of protein and employment for substantial communities. In such circumstances, precautionary management should also be concerned with conserving the fishery and fishermen, as well as the fish stock. This should be reflected by incorporation of socioeconomic aims, as well as biological aims, in the management objectives for that fishery. In principle, methods for assessing the degree of socioeconomic precaution similar to those described above can be developed, but they do require that the operating models for the fishery system have a much wider scope than most that have been used to date.


Butterworth, D.S. and M.O. Bergh. 1993. The development of a management procedure for the South African anchovy resource. Pp 83–99 in S.J. Smith, J.J. Hunt and D. Rivard (eds.) Risk evaluation and biological reference points for fisheries management. Canadian Special Publication of Fisheries and Aquatic Sciences 120. National Research Council and Department of Fisheries and Oceans. Ottawa.

Butterworth, D.S., Gluckman, G.R., Thomson, R.B., Chalis, S., Hiramatsu, K. and D.J. Agnew. 1994. Further computations of the consequences of setting the annual krill catch limit to a fixed fraction of the estimate of krill biomass from a survey. CCAMLR Science 1: 81–106.

Constable, A.J and W.K. de la Mare. 1994. Revised estimates of yield for Electrona carlsbergi based on a generalised version of the CCAMLR krill yield model. Paper WG-FSA-94/21 presented to the CCAMLR Scientific Committee, October 1994.

Francis, R.I.C.C. 1992. The use of risk analysis to assess fishery management strategies: a case study using orange roughy (Hoplostethus atlanticus) on the Chatham Rise, New Zealand. Can. J. Fish. Aquat. Sci. 49: 922–930.

Frederick, S.W. and R.M. Peterman. 1995. Choosing fisheries harvest policies: when does uncertainty matter? Can. J. Fish. Aquat.Sci. 52 (2): 291–306.

Garcia, S.M. 1994. The precautionary approach to fisheries with reference to straddling fish stocks and highly migratory fish stocks. FAO Fisheries Circular No. 871.23 pp.

Hilborn, R. and R.M. Peterman. 1995. The development of scientific advice with incomplete information in the context of the precautionary approach. Background paper TCPA/BP 2 to this conference.

Hilborn, R. and C.J. Walters. 1992. Quantitative fisheries stock assessment. Choice, dynamics and uncertainty. Chapman and Hall, New York.

Huppert, D.D. 1995. Risk assessment, economics and precautionary fishery management. Background paper TCPA/BP 3 to this conference.

IWC. 1992. Report of the Scientific Committee. Rep. int. whal. Commn 42: 51–267.

IWC. 1993, Report of the fifth comprehensive assessment workshop on revised management procedures. Rep. int. whal. Commn 43: 229–240.

Larkin, P.A. 1977. An epitaph for the concept of maximum sustainable yield. Trans. Am.Fish.Soc. 106: 1–11.

Ludwig, D. and C.J. Walters. 1985. Are age structured models appropriate for catch-effort data? Can. J. Fish. Aquat. Sci. 42: 1066–1072.

Mace, P.M. 1994. Relationships between common biological reference points used as thresholds and targets for fisheries management. Can.J.Fish.Aquat.Sci. 51: 110–122.

May, R.M., Beddington, J.R., Clark, C.W., Holt, S.J. and R.M. Laws. 1979. Management of multispecies fisheries. Science 205: 267–277.

Punt, A.E. 1992. Selecting management methodologies for marine resources, with an illustration for Southern African hake. S.Afr. J. Mar. Sci. 12: 943–958.

Punt, A.E. and A.D.M. Smith. 1995. Some comments on ranking alternative hypotheses when selecting management procedures. Paper SC/47/Mg4 presented to the IWC Scientific Committee, June 1995.

Rosenberg, A.A. and S. Brault. 1993. Choosing a management strategy for stock rebuilding when control is uncertain. Pp 242–249 in S.J. Smith, J.J. Hunt and D. Rivard (eds.) Risk evaluation and biological reference points for fisheries management. Canadian Special Publication of Fisheries and Aquatic Sciences 120. National Research Council and Department of Fisheries and Oceans. Ottawa.

Rosenberg, A.A. and V.R. Restrepo. 1995. Precautionary management reference points and management strategies. Background paper TCPA/BP 4 to this conference.

Sainsbury, K.J. 1991. Application of an experimental approach to management of a tropical multispecies fishery with highly uncertain dynamics. ICES Mar.Sci.Symp., 193: 301–320.

Smith, S.J., Hunt, J.J., and D. Rivard (Eds). 1993. Risk evaluation and biological reference points for fisheries management. Canadian Special Publication of Fisheries and Aquatic Sciences 120. National Research Council and Department of Fisheries and Oceans. Ottawa.

Walters, C.J. 1986. Adaptive management of renewable resources. Macmillan, New York.

Previous Page Top of Page Next Page