A short review of precautionary reference points and some proposals for their use in data-poor situations (1998)
 (introduction...) Introduction Some General Considerations The Role of RPs and Harvest Laws in Management Objectives Underlying Formulation of Management and Reference Points A Brief Classification of LRPs and their Applicability in Information-limited Situations Functioning of an LRP-based System Specific Problems and Approaches for Tuna Fisheries Reference Points from Size-and-age Data? Diagnosis of Areas Where LRPs are Required Multispecies RPs? Spatial Information and Reference Points Rotating Closures and Gauntlet Models Discussion References Annex: Alternative Closed-form Solution for FNOW

### Functioning of an LRP-based System

It is not always clear from the scientific literature how the LRPs specified are expected to be used. Two possibilities seem to be allowed for here, which can perhaps be referred to as 'targeted' or 'non-targeted'.

1. In the first case, a TRP is still specified, and the fishery is managed so as to try and hit this target, with the proviso that, from precautionary considerations, statistical analyses and modelling are aimed at ensuring that the possibility of overshoots is minimized.

In an earlier paper (Caddy and McGarvey, 1996), the strategy was proposed of setting the LRP first, based on, for example, known MSY conditions or biological fundamentals. A TRP is then estimated as a secondary quantity, based on the degree of uncertainty in the position of the fishery so that a degree of guesswork may be needed, and an acceptable probability of overshoot is agreed to by management. It is recognized that estimating variance is often problematical for control variables such as fishing mortality.

This admittedly approximate approach can allow a rapid solution for the TRP using a hand calculator and the formula in the Appendix from Caddy and McGarvey (1996) can be used if a normal distribution of error is believed to be a reasonable approximation. Alternatively, a mathematical package such as MAPLE gives solutions for a lognormal distribution of error around FNOW the current best estimate of fishing mortality on the stock, using other formulations reported by the above authors. The authors note that the normal distribution, though theoretically less acceptable, gives very similar results if the acceptable probability of overshoot is of 10-15% or more (which seems reasonable in practice, otherwise the fishery would have to be run at very low levels of target yield), making this approach a relatively simple way of obtaining an approximate solution for the TRP, given an LRP.

2. A 'non-targeted' or 'feedback-control' approach is, in theory, possible, whereby the fishery is allowed to operate at the economically optimal level of effort with no specific stock assessment target defined. This conclusion is subject to the proviso that, on triggering one or more LRPs, the fishing effort is then reduced significantly the following season. The level of drop in effort or quota must then be sufficient and of sufficient duration to ensure that the resource has a chance to recover to a level above that which applied before the management correction was imposed. In two recent papers (Caddy, 1998, and Caddy, in press), the idea was floated of a possible participatory approach to LRP management, using several LRPs in a 'traffic-light' control rule. This resembles theoretically fuzzy-logic controllers in smoothed feed-back systems, A variable response is implied such as that the severity of management correction increases as the number of LRPs turn from 'green' to 'red' on a 'management board' (similar to the public notice boards used for fire risk warnings in forests) (Fig. 10).

In the 'non-targeted' approaches (and probably in the targeted ones also, given that it is difficult to allow for regime shifts that change production levels over decadal and longer scales), the fishery is likely to go through a series of oscillations (see hypothetical example in Figure 11). This example also illustrates that corrections may needed to be applied over a series of years before a negative resource trend is reversed, and this makes such a management system always incipiently similar to a stock recovery strategy. Such oscillations may be more marked, if there are delays in applying controls before inevitably more severe management reactions to declining harvests are imposed by necessity. In such cases, the oscillations are due to 'runaway' overshoots in effort (or declines in spawning biomass), as opposed to undetected regime shifts which change recruitment and/or the effective fishing power (due to changes in aggregation patterns). What will also be required is to recognize that such an oscillatory system is probably inevitable. We should also note (as for stock recovery strategies) that corrective action should continue to be applied until a markedly higher biomass is reached than when they were initiated, otherwise the fishery risks becoming 'stuck' at close to the LRP level.

Figure 10. Using a 'basket' of LRPs to judge a fishery (left) from a change in a series of traffic 'lights' which change from green (bottom light) to red (top lights) as the LRPs are infringed. The number of red lights lit is accumulated and dictates the severity of the management response (right), which continues until the red lights turn green again.