One fundamental aspect here that is left out of much of the technical documentation on RPs is that in both the Code and UN Agreement, and also emphasized in the guidelines for applying the precautionary approach (FAO, 1995c), it is understood that reference points are only relevant if placed in their management context as part of a Fisheries Control Law that must include industry and management to be effective and must take into account risk at all levels of implementation of a Control Law (Table 3). This requires cooperation to include some or all of the following elements. Each partner in the management system needs to play a coordinated role, as illustrated in simplistic form by the following table.
Table 3. Relative roles of fishery scientists, managers and the fishing industry in management by Reference Point/Control Law
|
Scientists |
Management |
Industry |
|
- Define options for safe targets for exploitation (TRPs) and how to measure them |
Consider how to monitor landings, expenditure of fishing effort, changes in fishing power and area/season of fishing |
Provide information on landings, biological data and areas/times as condition of licensing |
|
Define upper limits for harvest/exploitation rate/surviving biomass which ensures adequate spawning |
Consider feasibility of reducing fishing effort drastically when stock is depleted; develop recovery plans |
Negotiate with management on actions to be taken when stock reaches LRPs |
|
Suggest options for decision rules within which LRPs and TRPs will play a key role |
Agree on a decision rule and on TRPs and LRPs and consider their implications for MCS and for stock recovery strategies as needed |
Negotiate with management on practicalities of the fishermen's response when key points in decision rule and LRPs are reached |
It is clear that TRPs and LRPs alone do not lead to a responsible or precautionary management response, and Figure 1 illustrates that a high degree of rigour is appropriate in setting LRPs only if the management response is designed to be rapid and effective in controlling potential overfishing. Reference points can only be effective if appropriate management responses (and, by the same token, agreement with the fishing industry) are prenegotiated and effectively implemented. If this is not the case, what may be needed are a number of empirical, but easily understood, indicators of potentially dangerous conditions, which can be phrased as easily measured quantities, directly relevant to the fishery in question.
Relevant in this respect is that, within ICES, the precautionary basis for advice given to ACFM will be that, for a given stock, "the probability of exceeding the limit reference point will be no greater than 5% in any given year" (Serchuk et al., 1997). The implication of this, of course, given that current data from the fishery (e.g. catches, survey biomasses) have an intrinsic coefficient of variance probably of the order of 20-30%, is that, in order to rigorously avoid the zone close to the LRP, the TRP must be set at a low level, or one or more other controls must be in place to detect and respond to declines in stock levels even before the LRP is reached.
One issue that comes out of recent intensive studies by ICES, NAFO and other organizations is that the search for practical limit reference points has generated an unprecedented amount of new research on how to set the limits to exploitation. The danger, however, is that this search become disconnected from the management process as a whole and that the systems view of a fishery be lost. Views were even expressed in the relevant scientific reports that the precautionary approach is not a matter of concern to scientists but solely relevant to managers. Certainly, scientists cannot preempt the role of managers and should not bias their analyses in the search for results that are more 'safe' for the fishery as a whole. At the same time, the fishery is an objective system, and someone needs to consider its performance in toto, and this is a legitimate role for systems scientists in cooperation with the other players or stockholders. In doing this, it will also be necessary for scientists to calculate the risks of the overall management system not meeting its goals, given a series of likely eventualities. Defining LRPs alone will not achieve this 'Systems Science Goal'.
With this in mind, it may be worth illustrating several alternative ways in which an LRP-based system might operate. Figures 3 and 4 show decision rules adopted by ICES and NAFO and Table 4 gives some of the reference points proposed for inclusion in these rules. One might suppose that following procedures similar to those proposed by the IWC in the late 1980s that, if the fishery and resource are well understood, such 'harvest- control laws' could be elaborated in even more detail, negotiated with industry and set to function largely independently of annual assessments in the form of 'fishery management procedures'. Such an approach has been described for South African fisheries (e.g. Butterworth et al., 1997); the 'traffic-light' approach described later offers a different approach to the same problem.
Table 4. Some biological reference points*
*(from Report of the Comprehensive Fishery Evaluation Working Group, ICES Headquarters, 25 June -4 July 1997. Advisory Committee on Fishery Management, ICES CM 1997/Assess: 1 5)
F01: fishing mortality rate at which the slope of the yield per recruit curve as a function of fishing mortality is 10% of its value near the origin.
Fmax: fishing mortality rate which corresponds to the maximum yield per recruit as a function of fishing mortality.
Flow: fishing mortality rate on an equilibrium population with a SSB/R equal to the inverse of the 10th percentile of the observed R/SSB.
Fmed: fishing mortality rate on an equilibrium population with a SSB/R equal to the inverse of the median observed R/SSB.
Fhigh: fishing mortality rate on an equilibrium population with a SSB/R equal to the inverse of the 90th percentile of the observed R/SSB.
Fx%: fishing mortality rate on an equilibrium population with a SSB/R of the SSB/R for the corresponding unfished population.
BMSY: biomass corresponding to maximum sustainable yield as estimated from a production model.
FMSY: fishing mortality rate which corresponds to the maximum sustainable yield as estimated by a production model.
Fcrash: fishing mortality which corresponds to the upper intersection of the yield and fishing mortality relationship with the fishing mortality axis as estimated by an F-based production model.
FLOSS: the replacement line corresponding to the Lowest Observed Spawning Stock (LOSS).
B50% R: the level of spawning stock at which average recruitment is one half of the maximum of the underlying stock-recruitment relationships.
B90% R' 90% Surv: level of spawning stock corresponding to the intersection of the 90th percentile of observed survival rate (R/S) and the 90th percentile of the recruitment observations.
Estimates of these quantities may be conditional on the stock-recruitment relationships assumed.
Figure 6. Excerpt from Caddy and McGarvey (1966) assuming FMSY as the LPR. Above: Illustrating variance in the current best estimate of fishing mortality, FNOW Below: Corresponding levels of F (TRP) for different values of the c.v. and the agreed overshoot of F (LRP).
Estimates of target fishing mortality (FNOW) in relation to (a) the risk (probability P) that current F exceeds the limit reference point FMSY = 0.6 (fishing mortality at maximum sustainable yield) and (b) the coefficient of variationa of F (CVF), calculated for normal and log-normal distributions of uncertainty in F
|
P |
aCVF |
||||
|
|
0.05 |
0.10 |
0.25 |
0.50 |
1.00 |
|
|
Normal distribution of uncertainty in F |
||||
|
30% |
0.58 |
0.57 |
0.53 |
0.48 |
0.39 |
|
20% |
0.55 |
0.55 |
0.50 |
0.42 |
0.33 |
|
10% |
0.53 |
0.53 |
0.45 |
0.37 |
0.26 |
|
5% |
0.52 |
0.52 |
0.43 |
0.33 |
b-0.04 |
|
|
Lognormal distribution of uncertainly in F |
||||
|
30% |
0.58 |
0.57 |
0.53 |
0.48 |
0.39 |
|
20% |
0.58 |
0.55 |
0.50 |
0.42 |
0.33 |
|
10% |
0.56 |
0.53 |
0.45 |
0.37 |
0.26 |
|
5% |
0.55 |
0.51 |
0.40 |
0.28 |
0.15 |
a CVF = SD of the normal or lognormal distribution of F)/(mean of the respective distribution), or s /FNOW.b Predicted target F-values with normal distributions of uncertainty in F are in error and diverge from those predicted by the lognormal distribution but only when uncertainty is high and acceptable risk of mortality overshoot (P) is very low.
As noted, two categories of RPs are now defined: the reference point may refer to a feature of the underlying model providing it (e.g. FMSY F0.1, FCRASH. etc.), which are referred to in Table 2 as 'technical' reference points, or the RP 'acronym' may simply refer to the role of the reference point in the decision rule adopted (e.g. FPA, BBUF, FLIM). These may be defined in terms of one or more of the above model-based RPs or be based on other empirical criteria that are of importance to the fishery in question.
What is clear from examination of recent ICES and NAFO reports is that almost all RPs currently in use are based in some way or other on the availability of age-structured data and on information on stock and recruitment accumulated over a significant period of time. Where production models are used offering estimates of MSY or FCRASH, For ICES these are largely derived from the Shepherd (1982) model which creates a general production function from stock-recruit data.
This dependence on stock recruit information was the problem facing the NAFO meeting in March 1998 since for many NAFO stocks of fish or invertebrates age data and time series are not readily available. In a sense, then, these NAFO stocks (such as cephalopods and shrimp) face some of the same problems as most tuna fisheries and almost all developing country fisheries in defining precautionary reference points. For these, one may have to seek other less rigorous approaches to setting reference points. This is where it is again essential to consider how reference points are going to be used by managers and understood by fishermen. In other words, a highly technical RP or control law will be difficult to explain to industry and will still need to accumulate practical experience 'on the grounds', while a less precise or 'empirical' RP may be effective if it is understood and receives consensus from industry and still leads to reproducible results.
It is even possible that relatively arbitrary value(s) may be adopted for RPs which attempt to reproduce situations similar to those remembered by the industry in earlier favourable years in the fishery (e.g. B1965 or B1970, or values for RPs may be used that appear to be precautionary from experience in this fishery or elsewhere. In fact, it can be noted that the setting of 'pairs' of LRPs and TRPs using formulations such as [FPA = FLIM.exp-2S) in Table 2 is an arbitrary but necessary procedure; the 'distance' between the TRP and LRP depends on the value of S which so far has not usually been defined but is typically taken as 0.2-0.3 (ICES sources).
The point is that, however defined, RPs will need to be subject to some 'tuning' as part of a fishery management system and will probably have to be modified in light of practical experience. Relevant experience here was gained in South African fisheries, where, even though in an early version of a management system, a prenegotiated management response had been agreed to with industry, but experience soon led to its modification in the light of its application.
It is of particular interest to note that, where there is inadequate data for establishing a harvest control law, it may be still possible to set a single LRP that corresponds to serious but not catastrophic conditions and then pick a TRP based on estimates of variance and probability of overshoot. Such an approach was suggested by Caddy and McGarvey (1996) and may provide a useful tool for guessing a suitable value for a TRP, given a pre-established LRP. Although, as noted by Caddy and McGarvey (1996), the variance of fishing mortality is difficult to estimate, reasonable orders of magnitude can usually be guessed at by those familiar with a fishery. As an example, given a value of CVF = 0.25 and FLIM = 0.6, this provides a value for FNOW = FPA of the order of F = 0.4-0.43, with an acceptable probability of overshoot of 5%. A similar value could be derived from the Appendix and the table in Figure 6, from Caddy and McGarvey (1996). As noted, such an estimate requires that the risk of overshoot of the LRP acceptable to managers be also specified, a factor which is not generally taken into account in most LRP-based systems.
