The Precautionary Principle in North Pacific Groundfish Management
By Grant Thompson
Alaska Fisheries Science Center
REFM Division
7600 Sand Point Way NE
Seattle WA, 98115
Introduction
During the past 10 years, the
precautionary principle has emerged as an increasingly popular theory which has has been
applied to the areas of environmental law and resource management on both a national and
international level. The precautionary principle was first referred to in an official
setting at the Second International Conference on the Protection of the North Sea, held in
London in 1987. Regulation of marine pollution was the subject, and the precautionary
principle was advanced in an attempt to shift the burden of proof from the regulatory
authority to the polluter. Basically, the precautionary principle holds that the existence
of scientific uncertainty regarding the precise effects of human activities on the natural
environment constitutes legitimate grounds for constraining such activities rather than
for pursuing them.
Recent years have witnessed a number of
calls for extension of the precautionary principle to fishery management. At the
international level, such calls have been featured in several agreements developed under
the auspices of the United Nations (U.N.), including the Code of Conduct for Responsible
Fisheries developed by the U.N. Food and Agriculture Organization (the FAO Code of
Conduct), the Rio Declaration of the U.N. Conference on Environment and Development (the
Rio Declaration), and the U.N. Convention on the Law of the Sea Relating to the
Conservation and Management of Straddling Stocks and Highly Migratory Fish Stocks (the
Straddling Stocks Agreement). Interestingly, although each of these agreements advocates
the precautionary principle’s use, none of them provides an operational definition of
the term as it applies or should apply to fishery management. In an effort to begin
filling this void
In an effort to address this problem, a
Technical Consultation on the Precautionary Approach to Capture Fisheries was convened in
1995 by the Government of Sweden in cooperation with the FAO (the FAO Technical
Consultation). While the FAO Technical Consultation succeeded in providing some broad
insights into what a precautionary approach to fishery management might look like, it too
stopped short of giving an operational definition.
At the national level, the Magnuson
Fishery Conservation and Management Act (MFCMA) has guided marine fishery management in
the United States since 1976. Although the MFCMA does not mention the precautionary
principle specifically, it contains provisions which seem to bear directly on this
principle. For example, National Standard 1 of the MFCMA mandates both the prevention of
overfishing and the achievement of optimum yield. The 602 Guidelines,
published in 1989 as the National Oceanic and Atmospheric Administration’s official
interpretation of National Standard 1, require each fishery management plan
(FMP) to
specify an "objective and measurable definition of overfishing" incorporating
"appropriate consideration of risk" and a delineation of "management
measures necessary to prevent overfishing." In the four years following publication
of the 602 Guidelines, more than 100 definitions of overfishing were submitted and
approved for use in FMPs across the country, including an overfishing definition for the
groundfish fisheries of the North Pacific.
To ensure that the various definitions
established under the 602 Guidelines were adequate to prevent
overfishing, the
National Marine Fisheries Service convened a special panel (the Overfishing Definitions
Review Panel) in 1993 to review these definitions. The report of the Overfishing
Definitions Review Panel, published in 1994, contained several general recommendations
that bear on the application of the precautionary principle to fishery management. In
addition, the report made specific recommendations for modifying the overfishing
definition used in the groundfish fisheries of the North Pacific. At about the same time,
the Scientific and Statistical Committee of the North Pacific Fishery Management Council
(NPFMC) made its own suggestions for modifying the North Pacific groundfish overfishing
definition. As a result of the suggestions made by the Overfishing Definitions Review
Panel and the Scientific and Statistical Committee, the NPFMC revised its definition of
overfishing for the North Pacific groundfish fisheries in June 1996. The revised
definition provides a clear exposition of the precautionary principle as it relates to
fishery management. This article presents three questions fundamental to developing a
precautionary approach to fisheries management and a set of possible answers based on the
revised definition of overfishing for the North Pacific groundfish fisheries.
Three
Questions to be Answered
- What should be the relationship between
intended catch targets and absolute catch limits? On the one hand, the two
concepts could be synonymous. Assuming that an optimal harvest level exists, it could be
argued that any catch, no matter how small, in excess of that optimal level should not be
tolerated, meaning that the optimal level is not only the intended target but also an
absolute upper limit. However, such an argument is impractical because it implies an
impossible level of precision in the management process. Instead, it should be assumed
that there is almost no chance of the actual catch matching the target exactly;
rather it will be off by some amount, either plus or minus. If it is understood that the
target harvest level will sometimes be overshot by small amounts due to random chance, it
makes sense to draw a distinction between an intended catch target and an absolute catch
limit, the former being an amount to which management is trying to come as close as
possible, the latter implying a cap on the permissible amount by which the target can be
exceeded accidentally without jeopardizing the stock’s long-term productive capacity.
The intended target is associated with the optimum level of harvest, while the absolute
limit is associated with the boundary of the danger zone. Harvesting at a rate greater
than that corresponding to the target would be expected to result in suboptimal fishery
performance over the long term, but would not be expected to do irreversible damage to the
stock’s innate productive capacity provided the harvest rate corresponding to the
absolute catch limit is not exceeded.
The FAO Code of Conduct suggests that nations should determine "stock specific target
reference points" and "stock specific limit reference points"
(emphasis added). Likewise, the Overfishing Definitions Review Panel report states that
"it is important to make distinctions between the management targets and overfishing
definition thresholds." However, neither report is very specific as to how these
objectives should be accomplished.
- What should be the relationship between
stock size and catch? In other words, is it appropriate to harvest a constant proportion
of the stock regardless of the stock’s size, or should the harvest rate change in the
event that the stock becomes depleted? Much of the literature on optimal harvesting argues
in favor of a constant harvest rate independent of stock size. However, it seems unlikely
that such an argument can be considered valid in the case where a stock size becomes
extremely small, for even if a particular stock could be safely harvested at a certain
rate across a large range of stock sizes, few would suggest maintaining the same harvest
rate in the event that the stock was on the brink of extinction.
The Overfishing Definitions Review Panel report suggests that harvest rates should be
defined "using a combination of a maximum fishing mortality rate, a precautionary
biomass level below which the maximum allowable fishing mortality rate is reduced, and an
absolute minimum biomass threshold." However, the values associated with these rates
and levels are not specified.
- What should be the management response to a
given level of uncertainty surrounding estimates of key population parameters? Scientists
are rarely, if ever, certain regarding the precise long-term effects of a particular
fishing level on a given fish stock. Scientists may have data and analyses which permit a
description of the most likely effects, but such descriptions are inevitably
associated with a potentially high level of uncertainty. If the level of uncertainty is
great, any particular level of fishing could be too low, thus foregoing harvests available
in the short term, or too high, thus diminishing the level of harvests achievable in the
long term. In some parts of the country, fishery managers have had difficulty rejecting
the claims of resource users who feel that a lack of scientific certainty diminishes the
Government’s right to constrain harvests. So, if the amount of uncertainty regarding
a stock’s productivity happens to increase, should the target catch level change as
well, and if so, by how much and in which direction?
The FAO Code of Conduct advises that "in implementing the precautionary approach,
States should take into account, inter alia, uncertainties relating to the size and
productivity of the stocks." Likewise, the FAO Technical Consultation concluded that
"a precautionary approach to fishery management would implicitly account for
uncertainty by being more conservative." Again, however, specifics are lacking.
New Policy in the North
Pacific
Management of groundfish in the
U.S. EEZ (Exclusive Economic Zone) portion of the North Pacific (the eastern Bering Sea,
Aleutian Islands region, and Gulf of Alaska) has been characterized by a deliberately
conservative approach since passage of the MFCMA. During the first several years of
management under the MFCMA, the mechanisms for maintaining this conservative approach were
largely informal. For example, the groundfish FMPs lacked an objective and measurable
definition of overfishing. Further, target catch levels were typically based on acceptable
biological catch (ABC), which was defined only loosely in the groundfish
FMPs. Responding
to publication of the 602 Guidelines, the NPFMC addressed the first of these two
problems in 1990 when it adopted an objective and measurable definition of the overfishing
level (OFL) for the North Pacific groundfish fisheries. The OFL definition provided an
absolute upper limit on the amount of fish that could be harvested in any given year.
However, the relationship between this upper limit and ABC remained somewhat nebulous. In
June 1996 the NPFMC moved to address shortcomings of the existing OFL definition as well
as ambiguities in the relationship between ABC and OFL when it approved a pair of
amendments to redefine both ABC and OFL in the FMPs for North Pacific
groundfish. The new
definitions encompass a set of tiers corresponding to the types of data or parameter
estimates that might be available for the various stocks covered by the
FMPs. The most
fully developed tiers are those nearest the top of the hierarchy, that is, those
applicable to stocks for which assessment information is the most complete, though not
necessarily the most precise. The remainder of this article focuses on how the new
definitions of ABC and OFL on the top tier in the hierarchy relate to the precautionary
principle. In general, the top tier deals with the three previously posed questions as
follows.
What should be the relationship between
intended catch targets and absolute catch limits? Answer: Intended target catches (ABC)
should be well below the levels at which the stock’s long-term productive capacity
might be jeopardized (OFL).
What should be the relationship between
stock size and catch? Answer: Depleted stocks should be harvested at a lower relative rate
than healthy stocks.
What should be the management response to
a given level of uncertainty surrounding estimates of key population parameters? Answer:
Greater uncertainty regarding a stock’s productivity should correspond to greater
caution in setting the target catch rate.
Intended Target Catch
Well Below Absolute Catch Limit
The new ABC/OFL definitions
keep catch targets below catch limits by distinguishing between the ABC, or the intended
target, and the OFL, or the absolute limit. An explicit buffer is imposed between the two
quantities so that inadvertantly overshooting the ABC level for Species X by a small
amount does not automatically close all other fisheries that might take small amounts of
Species X as unavoidable bycatch. It should also be noted that the explicit buffer imposed
between ABC and OFL is a minimum buffer, allowing the NPFMC to set a larger buffer for any
particular species in any particular year if it wishes. This flexibility is provided by
defining the OFL harvest rate as an equality and the ABC harvest rate as an inequality.
The new definition does not allow the OFL harvest rate to vary from the formula specified
in the FMP, whereas the ABC harvest rate is expressed as an upper bound only, thereby
allowing the NPFMC the option of setting a lower target harvest rate and thus a larger
buffer.
Depleted Stocks
Harvested at Lower Rates than Healthy Stocks
The new ABC/OFL definitions
treat depleted stocks more cautiously than healthy stocks by tying the two harvest rates
explicitly to stock size. The precise relationships are illustrated for a hypothetical
stock in Figure 1. When the stock is above the biomass level associated with maximum
sustainable yield (BMSY), neither the ABC nor the OFL harvest rate
varies with stock size. However, should the stock fall below BMSY, both
the ABC and OFL harvest rates decrease linearly as a function of stock size, down to a
value of zero at some very low abundance level (typically 5% of BMSY).
Although the absolute magnitudes of the ABC and OFL rates vary, the ratio between them
remains constant.
Greater Uncertainty
Corresponds to Greater Caution
Before addressing how the new ABC/OFL definitions treat uncertainty, it is helpful to discuss the topic of uncertainty
in general. First, if the values of the parameters governing stock dynamics such as
population growth rate and carrying capacity could be known with certainty, it would be
fairly easy to compute the value of the harvest rate that maximizes sustainable yield, FMSY.
However, because their measurements are always subject to error, parameter values are
never known with certainty, so the best that can be hoped for in practice is to estimate
the relative plausibility of alternative values for FMSY. For example,
it might be possible to determine for a particular stock that there is only a 5% chance of
FMSY being smaller than about 0.10, that there is only a 5% chance of FMSY
being greater than about 0.35, and that the most likely value of FMSY is
about 0.16. These probabilities can be expressed in the form of the curve shown in Figure
2. Such a curve is called a probability density function or PDF. Given a
PDF, it is easy
to compute an average or expected value for FMSY. The expected value for
the curve shown in Figure 2 is 0.20. The expected value, which describes the center of
gravity of the PDF, is also called the arithmetic mean. For example, the curves shown in
Figure 3 represent four different PDFs, all with an arithmetic mean of 0.20 (the PDF whose
peak is furthest to the right is the same curve shown in Figure 2). In a sense, each curve
in Figure 3 balances at the arithmetic mean of 0.20.
If the value of FMSY is
known with a great deal of precision, the PDF will be tightly clustered around the
arithmetic mean, whereas if the value of FMSY is known with little
precision, the PDF will be much more spread out, indicating a relatively high probability
that the true value of FMSY is quite different from the arithmetic mean.
The four PDFs in Figure 3, for example, correspond to four different levels of
uncertainty. As the level of uncertainty increases, the curve becomes broader and the peak
of the curve moves to the left.
One measure of the amount of uncertainty
associated with a PDF is the coefficient of variation or CV. The CV measures, on a
relative scale, the average amount by which the true value might differ from the
arithmetic mean. The curve shown in Figure 2 has a CV of 40%. The curves shown in Figure
3, moving from right to left in order of the location of the peak, have CVs of 40%, 60%,
80%, and 100%, respectively. The higher the CV, the higher the level of uncertainty.
To insure that greater uncertainty
regarding a stock’s productivity corresponds to greater caution in setting target
harvest levels, the new ABC/OFL definitions use the information in a PDF such as those
shown in Figure 3 to establish the minimum buffer between the ABC and OFL harvest rates.
The new definition accomplishes this by setting the OFL harvest rate at the arithmetic
mean of the PDF while capping the ABC harvest rate at the harmonic mean. The
difference between these two means can be summarized as follows: The arithmetic mean gives
the expected value of the points along the horizontal axis, while the harmonic mean gives
the reciprocal of the expected value of the reciprocals of the points along the horizontal
axis. It can be demonstrated that the harmonic mean of the FMSY PDF is
the optimal harvest rate from the viewpoint of risk-averse decision making, at least
within the context of one type of mathematical model used in fishery stock assessment. Two
more general properties of the harmonic mean are that it is always less than the
arithmetic mean and that the ratio between the harmonic and arithmetic means decreases as
the level of uncertainty increases. For example, the harmonic means of the four PDFs in
Figure 3 (all of which have an arithmetic mean of 0.20) behave as described in the table
below:
|
Table 1.
The harmonic means of the four probability density functions (pdfs) in Figure 3.
|
| Coefficient of
variation: |
0.400 |
0.600 |
0.800 |
1.000 |
| Harmonic mean: |
0.172 |
0.147 |
0.122 |
0.100 |
| Ratio (harmonic
mean to arithmetic mean): |
0.862 |
0.735 |
0.610 |
0.500 |
A convenient rule of thumb for computing
the ratio between the harmonic and arithmetic means is
This rule is exact for certain types of PDF, but is only approximate for others (and then only for relatively small CV values,
say, CVs of less than 50%). The above rule of thumb is illustrated in Figure 4, with the
special cases of CV=0.5 and CV=1.0 highlighted.
Conclusion
The new ABC/OFL definitions for
North Pacific groundfish constitute a significant step toward translating the
precautionary principle into practical and easily interpretable terms. By clearly
separating intended catch targets from absolute catch limits, by lowering harvest rates
for depleted stocks, and by requiring greater caution in the presence of uncertainty, the
new definitions provide a framework for realizing National Standard 1 of the Magnuson
Fishery Conservation and Management Act: prevention of overfishing while achieving optimum
yield.
Definitions of
Statistical Terms
Probability density function (PDF): A description of the probability associated with different values of a
variable. For example, in a coin flip the probability of tossing heads is 50% and the
probability of tossing "tails" is 50%. As another example, in tossing a
six-sided die the probability of tossing a "1" is 16.667% and the probability of
tossing something other than a "1" is 83.333%. The probabilities in a PDF must
always sum to 100%.
Arithmetic mean: If X is a
random variable, the arithmetic mean is the average value of X. For example,
consider a game of chance based on a coin flip, where the random variable X denotes
the prize associated with the game. The player gets $72 if he or she tosses heads and $24
if he or she tosses tails. The arithmetic mean prize for this game is
|
 |
|
As another example, consider a game of
chance based on the toss of a six-sided die, where again the random variable X
denotes the prize associated with the game. The player gets $72 if he or she tosses a
"1" and $24 if he or she tosses anything else. The arithmetic mean prize
associated with this game is
Harmonic mean: If X is a random variable, the harmonic mean is 1 over the
average value of 1/X. For example, consider the game of chance based on a coin flip
described above. The harmonic mean prize associated with this game is
As another example, consider the game of
chance based on the toss of a six-sided die. The harmonic mean prize associated with this
game is
|
 |
|
Note that the harmonic mean is less than the arithmetic mean in both of these examples
($36 versus $48 for the coin flip and $27 versus $32 for the die toss). For all practical
purposes, this relationship always holds (i.e., the harmonic mean is always less than the
arithmetic mean). Thus, if the random variable X represents a fishing mortality
rate, the harmonic mean is a more conservative (i.e., lower) rate than the arithmetic
mean.
Coefficient of Variation (CV): For
a random variable X, the coefficient of variation is the standard deviation of X
divided by the arithmetic mean of X. The standard deviation, in turn, is a measure
of the average amount by which the various possible values of X differ from the
arithmetic mean. A bit more precisely, the standard deviation is the square root of the
average squared difference between the various possible values of X and the
arithmetic mean. For the coin flip example (above), the CV is given by
while for the die toss example (above), the
CV is given by
|
