close

Вход

Забыли?

вход по аккаунту

?

The power of ecosystem monitoring.

код для вставкиСкачать
AQUATIC CONSERVATION: MARINE AND FRESHWATER ECOSYSTEMS
Aquatic Conserv: Mar. Freshw. Ecosyst. 17: S79–S92 (2008)
Published online in Wiley InterScience
(www.interscience.wiley.com) DOI: 10.1002/aqc.909
The power of ecosystem monitoring
KEITH REID*, JOHN P. CROXALL and EUGENE J. MURPHY
British Antarctic Survey, Natural Environment Research Council, High Cross, Madingley Road,
Cambridge, CB3 0ET, UK
ABSTRACT
1. Implementing an ecosystem approach to fisheries management requires an effective ecosystem
monitoring programme, the utility of which depends upon its ability (measured by the statistical
power) to detect effects that trigger management action.
2. Using data from a long-term ecosystem monitoring programme of the predators of Antarctic
krill Euphausia superba at South Georgia together with a krill population model to simulate natural
and fisheries induced variability in krill abundance, the power to detect the effects of different levels
of fishing was examined.
3. The power to detect the effects of fishing using either the krill population or a combined
predator response index was low (20–40% power after 20 years with the probability of a type I error
ðaÞ ¼ 0:05). The power increased to >50% when a was increased to 0.2 when the ability to detect
change was greater with the predator response index than using the krill population itself.
4. The results indicate that although this monitoring programme has a proven ability to detect the
effects of natural variability in krill abundance, its ability to detect the effects of fishing may be limited
if there is a requirement for statistical significance at the 95% level. A situation where changing a
produces a marked increase in statistical power, and the difference in the relative ecological costs of
making type I and type II errors is likely to be high, may require a more flexible approach to choosing
significance levels required to trigger management action.
5. Although long-term monitoring provides a wealth of basic ecological information it is essential
to evaluate, the ability to detect specific changes in order that management action is not delayed
because of an inability to detect an effect rather than the lack of an effect of the fishery.
Copyright # 2008 John Wiley & Sons, Ltd.
Received 29 August 2007; Accepted 10 September 2007
KEY WORDS:
ecosystem monitoring; fisheries; Antarctic krill Euphausia superba; statistical power
*Correspondence to: K. Reid, British Antarctic Survey, Natural Environment Research Council, High Cross, Madingley Road,
Cambridge, CB3 0ET, UK. E-mail: k.reid@bas.ac.uk
Copyright # 2008 John Wiley & Sons, Ltd.
S80
K. REID ET AL.
INTRODUCTION
Understanding the consequences of fisheries on both target and non-target components of the ecosystem is
a central theme in ecosystem-based approaches to fisheries management (Garcia and Cochrane, 2005). As
with any management scheme, the success of an ecosystem approach can only be determined by a process of
monitoring and feedback. Monitoring programmes that are able to detect changes in the ecosystem brought
about by fishing are an increasingly important tool in the management of marine ecosystems (Greenstreet
and Rogers, 2006). In the Southern Ocean there is particular interest in the management of the fishery for
Antarctic krill Euphausia superba, given its central role in the marine foodweb (Laws, 1977; Croxall et al.,
1985; Murphy et al., 1998). Data from ecosystem monitoring programmes in the Southern Ocean have
revealed the extent to which natural fluctuations in the abundance of krill impact the reproductive
performance of a range of upper-trophic level species (Croxall et al., 1999; Reid and Croxall, 2001; Croxall,
2006).
The Commission for the Conservation of Marine Living Resources (CCAMLR), the body that manages
fisheries in the Southern Ocean, has adopted an ecosystem-based approach to managing krill fisheries
and the catch limit for krill is set to take account of the population demographics of krill as well as an
allowance for the potential consequences of that fishery on dependent and related species (Constable et al.,
2000). The fishery for Antarctic krill operates in a remote region and is relatively limited both in spatial
extent and with respect to the current precautionary catch limit. Nevertheless the potential for expansion
(Nicol and Endo, 1999; Croxall and Nicol, 2004) and the central ecological role of Antarctic krill in the
Antarctic marine ecosystem (Laws, 1977; Croxall et al., 1985) requires the development of a management
framework that provides for an orderly development of the fishery. An essential part of this approach is an
ecosystem monitoring programme that is able to detect the effects of krill fishing on dependent species
(Agnew, 1997).
CCAMLR takes a precautionary approach to setting the catch limit for krill by using a Monte-Carlo
simulation model of the krill population, which incorporates uncertainty in recruitment and mortality to
produce a suite of krill population trajectories at different levels of fishing intensity (Constable and de la
Mare, 1996). The model is then used to determine the proportion of the pre-exploitation biomass that can
be caught each year according to a three-part decision process. The first part of this process relates to the
maintenance of the krill population and is to determine the level of fishing, as a proportion of the preexploitation biomass, that produces a 10% probability of the spawning stock biomass dropping below 20%
of the pre-exploitation level at any point during a 20 year harvesting period. The second part relates to the
need to accommodate the requirements of krill predators and is the level of fishing that reduces the median
spawning stock biomass to 75% of the pre-exploitation level at the end of a 20 year harvesting period. The
final step is then to calculate the krill catch limit by multiplying the lower of the two values from step one
and step two by the pre-exploitation biomass (Constable et al., 2000).
CCAMLR uses a management approach that produces a precautionary catch limit for krill and
implements the CCAMLR monitoring programme (CEMP), intended, at least in part, to determine
whether this level of catch has a detectable impact on dependent species. Clearly, a crucial part of this
management approach is to determine the ability of that monitoring programme to detect the type and
magnitude of changes that might occur as a result of fishing.
Evaluating whether a monitoring scheme can detect changes arising as a result of fishing might be
considered as a process of using the monitoring data to test the null hypothesis that fishing, at a particular
catch level, has no negative impact on dependent species. In designing and implementing a monitoring
scheme that has feedback into the management process it is essential for that scheme to have the power to
detect any changes in the performance of indicator species that would trigger a management response. For
such a management scheme to be precautionary the probability of incorrectly accepting the null hypothesis
(making a type II error) should be as low as possible. The power (defined as 1 } the probability of making a
Copyright # 2008 John Wiley & Sons, Ltd.
Aquatic Conserv: Mar. Freshw. Ecosyst. 17: S79–S92 (2008)
DOI: 10.1002/aqc
THE POWER OF ECOSYSTEM MONITORING
S81
type II error) of the statistical procedure used to detect such changes arising from fishing will therefore be a
function of:
1. the chosen level of probability (a) of incorrectly rejecting the null hypothesis (a type I error);
2. the level of change that needs to be detected (the effect-size); and
3. the variability in the process/indicator being measured.
Given the research emphasis placed on the causes and consequences of variability in the abundance of the krill
population (Priddle et al., 1988; Murphy et al., 1998; Croxall et al., 1999; Reid et al., 1999), the analyses in this
paper will concentrate on the consequences of that variability on the power of a monitoring scheme to detect
fishery-induced changes. This can be explored using a simulation model of a krill population that varies over
time through variation in recruitment and mortality; the latter being adjusted to simulate the effects of fishing.
The consequential effects of different levels of fishing on a suite of monitoring parameters derived from uppertrophic level predators can then be examined using parameter estimates from the British Antarctic Survey
(BAS) long-term ecosystem monitoring programme at South Georgia where measures of both krill abundance
and predator performance are available (Reid et al., 2005). The aims of this simulation are to:
(i) consider the potential response of predator-derived indicators to different levels of krill fishing;
(ii) examine the power to detect changes in both the krill population and in the response of predators
given different scenarios of variability in the krill population; and
(iii) consider the influence of the choice of different levels of a on the power to detect the effects of fishing.
METHODS
Krill population model and stock projection
The formulation of the krill model follows that of Butterworth et al. (1994) and Constable and de la Mare
(1996). The number of individuals N in each of seven age classes a in time period t of year y where t>1 was
Nða;y;tÞ ¼ Nða;y;t1Þ eðMða;t1Þ Fða;t1Þ Þ
ð1Þ
and the number at t ¼ 1 was
Nð1;y;1Þ ¼ R
Nða;y;1Þ ¼ Nða;y1;tðmaxÞ Þ eðMða;t1Þ Fða;t1Þ Þ
ð2Þ
where M(a,t) is the natural mortality of age class a at time t. The number of recruits R that enter the
% and a coefficient of variability (Rcv)
youngest age class on day 1 in each year y where specified by a mean R
% and standard deviation from
and were drawn from a log-normal distribution with the mean ¼ ln(R)
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
d ¼ logð1 þ Rcv2 Þ
ð3Þ
F(a,t) is the fishing mortality of age class a at time t, where F(a,t) is
Fða;tÞ ¼ sða;tÞ FðmaxÞ
ð4Þ
in which s(a,t) is a function that describes the size-selectivity of the fishery and F(max) is the mortality on a
fully selected age-class. The size-selectivity function s(a,t) is based on total length at age such that,
8
lða;tÞ 5l1
>
< 0;
ð5Þ
sða;tÞ ¼ lða;tÞ l1 =ðl2 l1 Þ; l1 4lða;tÞ 4l2
>
:
1;
lða;tÞ > l2
Copyright # 2008 John Wiley & Sons, Ltd.
Aquatic Conserv: Mar. Freshw. Ecosyst. 17: S79–S92 (2008)
DOI: 10.1002/aqc
S82
K. REID ET AL.
where lða;tÞ is a growth function which gives the length of individuals of age a at time t and l1 and l2 are
constants that specify the range over which selectivity changes from 0 to 1 (following Constable and de la
Mare, 1996, Equation (4)).
The biomass in each age class a in time period t of year y
Bða;y;tÞ ¼
7
X
ð6Þ
Nða;y;tÞ oða;tÞ
a¼1
where oða;tÞ is the mass of an individual krill of age a at time t such that
(
½l1 ð1 eðkðaþð4t=360ÞÞÞ Þ3 ; 05t490
oða;tÞ ¼
oða;t1Þ ;
t591
ð7Þ
where k is a constant and l1 is the asymptotic length of krill, assuming that growth occurs for the first 90
days of the year (360 days).
The stock was projected in fixed time steps (t) in each year for 40 years during which fishing occurred in
years 20–40. During the period when fishing occurred the required yield Y was a fixed proportion of the
median biomass on day 100 of years 1–20. The fishing mortality F(y) required to produce the yield was
found by minimizing the function
YðyÞ ¼ Bða;y;1Þ eMFðyÞ Bða;y;1Þ eM
ð8Þ
with respect to F(y). This was done iteratively using the fzero function in Matlab (www.mathworks.co.uk)
for each age class of krill based on the fishery selection matrix F(a,t) in order to account for the differences in
mass at age and size selection by the fishery. As recruitment to the krill population at South Georgia is not
generally considered to be the result of local spawning events, the model contained no stock–recruitment
relationship.
In order to consider the effects of variability in the krill population the krill model was run using growth
and mortality values that reflected different regions within the Scotia Sea and at South Georgia following
the rationale in Murphy and Reid (2001). In order to perform this simulation for two krill populations, the
growth constant k in Equation (7) and the natural mortality M in Equation (1) were set to 0.4 and 0.6,
respectively, for the low growth/low mortality population and to 0.7 and 1.5 for the high growth/high
mortality population.
Predator response functions
The predator response vectors (PR) were derived on the basis of a significant non-linear relationship with krill
density (K) using an asymptotic exponential function that describes a Holling type II functional response
(Reid et al., 2005) and are shown in Table 1. For each model projection the resultant krill time-series was used
as the independent variable to produce a m n predator response matrix where m is the number of years and n
the number of predator response vectors listed in Table 1. The predator response matrix was then used to
generate a combined standardized index (CSI) value for each year following the methods of de la Mare and
Constable (2000), and Boyd and Murray (2001). This process required that each of the response vectors was
standardized to a mean ¼ 0 and sd ¼ 1 producing a sum I for each year y such that
Iy ¼ a0 xy
ð9Þ
where x is a vector of values for all response vectors in year y and a is an identity vector of the same
dimensions as x that takes a value of 1 for those vectors where observations exist and of 0 for missing data.
The variance V of Iy is given by
Vy ¼ aSa0
Copyright # 2008 John Wiley & Sons, Ltd.
ð10Þ
Aquatic Conserv: Mar. Freshw. Ecosyst. 17: S79–S92 (2008)
DOI: 10.1002/aqc
S83
THE POWER OF ECOSYSTEM MONITORING
Table 1. The species, response vectors and parameters from the Holling Type II response to krill variability used in the simulation
model
Species
Response vector
n
a
b
R2
F
Antarctic fur seal
Antarctic fur seal
Antarctic fur seal
Antarctic fur seal
Antarctic fur seal
Gentoo penguin
Gentoo penguin
Gentoo penguin
Macaroni penguin
Macaroni penguin
Macaroni penguin
Black-browed albatross
Foraging trip duration
Male growth deviate
Female growth deviate
Male weaning mass
Female weaning mass
Breeding success
Meal mass
% krill mass
Fledging mass
Meal mass
% krill mass
Breeding success
15
12
12
14
14
14
13
13
12
13
13
15
1.8
9.0
8.0
14.3
11.9
6.2
698.1
3.4
3321.8
415.6
7.2
7.6
5.7
10.8
9.2
3.3
4.2
13.6
9.9
9.8
2.8
9.4
16.3
7.2
0.482
0.513
0.504
0.403
0.569
0.445
0.308
0.397
0.418
0.343
0.468
0.461
12.10
10.48
10.14
8.10
15.85
9.55
4.96
7.26
7.18
5.76
9.13
11.16
where S is the covariance matrix of the standardized response vectors; hence the CSI in year y was
Iy
CSIy ¼ pffiffiffiffiffiffi
ð11Þ
Vy
Detecting the response to fishing
In order to determine the duration of monitoring required to produce a detectable response in the krill
population and in the performance of predators, the two simulated time-series were projected over time
periods of 2 to 20 years before and after fishing (i.e. time projected symmetrically forwards and backwards
from the point at which fishing began). For each time interval the null hypothesis that the mean krill
population size or CSI prior to fishing was greater than that during fishing was tested using a one-tailed t
test. In order to take account of the potentially auto-correlated nature of the time-series data, which inflates
the value of the t statistic, a correction factor was subtracted such that
sffiffiffiffiffiffiffiffiffiffiffi
1þr
t0 ¼ t ð12Þ
1r
where r is the correlation between consecutive observations (following Stewart-Oaten et al., 1986). Each
simulation produced a value of t[a][df] for each increment in the duration of the sampling period from 2 to 20
years; the process was repeated 1000 times and the frequency with which t[a][df ]>t0 [c.crit][df ] was recorded for each
time increment to determine the power curves as a function of sampling period. This process was repeated for
values of g between 0 and 0.5 to produce power curves at the different levels of fishing intensity. In order to
examine the influence of the choice of a this process was then repeated for differing levels of a from 0.05 – 0.2.
RESULTS
Krill population model
Arising from a krill population with low growth/low mortality (Figure 1(a)) the time-series for zero fishing
(g ¼ 0) and fishing at g ¼ 0:01 were indistinguishable, whereas levels of fishing between g ¼ 0:05 and 0.3
produced time series that did not overlap; the time series for fishing at g ¼ 0:4 and 0.5 showed considerable
Copyright # 2008 John Wiley & Sons, Ltd.
Aquatic Conserv: Mar. Freshw. Ecosyst. 17: S79–S92 (2008)
DOI: 10.1002/aqc
S84
K. REID ET AL.
1.0
0.9
Krill units
0.8
0.7
0
0.01
0.05
0.1
0.2
0.3
0.4
0.5
0.6
0.5
0.4
0
10
(a)
20
30
40
Years
1.00
0.95
0.90
Krill units
0.85
0.80
0
0.01
0.05
0.1
0.2
0.3
0.4
0.5
0.75
0.70
0.65
0.60
0
(b)
10
20
Years
30
40
Figure 1. Simulated time-series of krill population with different levels of fishing where fishing begins in year 20 for a krill population
with (a) low growth and mortality and (b) high growth and mortality. Each trajectory is the mean of 100 model runs.
overlap. In a population with high growth/high mortality there was much greater variability in the
individual time-series and considerable overlap particularly between g ¼ 0 and 0.1 and also between g ¼ 0:3
and 0.5 (Figure 1(b)).
Copyright # 2008 John Wiley & Sons, Ltd.
Aquatic Conserv: Mar. Freshw. Ecosyst. 17: S79–S92 (2008)
DOI: 10.1002/aqc
S85
THE POWER OF ECOSYSTEM MONITORING
Predator response
There was considerable overlap between adjacent time series of predator response to fishing at all levels of g
arising from the krill population with low growth and mortality (Figure 2(a)); the CSIs ranged between 0.5
0.6
0.4
0.2
CSI
0.0
-0.2
0
0.01
0.05
0.1
0.2
0.3
0.4
0.5
-0.4
-0.6
-0.8
-1.0
0
10
20
Years
10
20
Years
(a)
30
40
0.6
0.4
0.2
CSI
0.0
-0.2
0
0.01
0.05
0.1
0.2
0.3
0.4
0.5
-0.4
-0.6
-0.8
0
(b)
30
40
Figure 2. Simulated time-series of predator response CSI arising from a krill population with different levels of fishing where fishing
begins in year 20 for a krill population with (a) low growth and mortality and (b) high growth and mortality. Each trajectory is the
mean of 100 model runs.
Copyright # 2008 John Wiley & Sons, Ltd.
Aquatic Conserv: Mar. Freshw. Ecosyst. 17: S79–S92 (2008)
DOI: 10.1002/aqc
S86
K. REID ET AL.
and 0.8. In contrast the CSIs arising from the high krill growth and mortality scenario varied between 0.3
and 0.4; indeed there were occasions where the CSI time series for zero fishing had a lower value than the
series where g ¼ 0:5: As the effects of fishing increased relative to the effects of natural variation, with
increasing values of g, the standardization process (to an overall mean ¼ 0 and sd ¼ 1) results in the CSI
during the pre-fishing parts of the time series showing both an increased mean and reduced variability
(Figure 2(a)).
Power analysis
The power analysis for the krill population indicated that the power to detect a change, at a ¼ 0:05; in the
krill population with low growth and mortality was 46% when g ¼ 0:5; increasing to 56% after 20 years
(Figure 3(a)). The power to detect a change in the krill population with high growth and mortality, for
g ¼ 0:5 and a ¼ 0:05; was 26% after 10 years, rising to 42% after 20 years (Figure 3(b)).
By comparison the analysis for the resultant CSI time series indicated a power of 28% after 10 years
(g ¼ 0:5; a ¼ 0:05) for the low krill growth and mortality scenario, rising to 54% after 20 years (Figure
4(a)). In the high krill growth and mortality scenario the respective powers to detect changes after 10 and 20
years were 23% and 45% (Figure 4(b)).
Analysis of the power curves analogous to those in Figures 3 and 4, but produced using a ¼ 0:05; 0.10
and 0.20, indicate that for a value of g of 0.2 (i.e. a value of g that is approximately twice that used in the
current management of the krill fishery in the Scotia Sea; see CCAMLR (2000)) the power to detect change
is, with one exception, above 50% only when a ¼ 0:20 (Table 1). The power to detect change from both
krill and predator CSI increases as a is increased; moreover the CSI showed a greater rate of increase in
power with increasing a, such that when a ¼ 0:20 the power to detect change was greater with the CSI than
with the krill population (Table 2).
DISCUSSION
Using one of the longest and most methodologically consistent ecosystem monitoring data-series from the
Southern Ocean, it is shown that the power to detect the effects of a fishery for Antarctic krill on the
reproductive performance of krill-dependent species is likely to be rather low. However, it is important to
recognize that this does not mean that such a fishery will not have any effects, but that the available
monitoring tools have a very low probability of detecting them over time scales of 10–20 years using
conventional statistical approaches and confidence levels. While this may be disappointing, there are a
number of reasons why it should not be surprising.
The monitoring programme at Bird Island was established some 10 years prior to the inception of CEMP
with the aim of collecting long-term data on the biology and ecology of key upper-trophic level species in
the Antarctic foodweb (Croxall and Prince, 1979; Croxall et al., 1985, 1988). This monitoring revealed
distinct variability in the reproductive performance of krill predators in response to natural variability in
krill abundance and there was a natural assumption that these predators had the potential to be used as
indicators of the effects of krill fishing.
During the process of development of CEMP there was no power analysis of the monitoring programme
to determine what would be required to detect the effects of fishing. That such an analysis was not
undertaken was not an oversight but simply reflected that the basic data for such an analysis were not
available. The ability to undertake prospective power analysis of a monitoring programme depends upon
assumptions about feasible sample sizes and variances as well as the effect sizes thought to be biologically
important. These assumptions in turn depend upon the level of knowledge of the system that is to be
monitored. Indeed it is only now that the biological data collected by the monitoring programme, and
Copyright # 2008 John Wiley & Sons, Ltd.
Aquatic Conserv: Mar. Freshw. Ecosyst. 17: S79–S92 (2008)
DOI: 10.1002/aqc
S87
THE POWER OF ECOSYSTEM MONITORING
1.0
0
0.01
0.05
0.1
0.2
0.3
0.4
0.5
0.8
Power
0.6
0.4
0.2
0.0
0
5
(a)
10
Years
15
20
10
Years
15
20
1.0
0
0.01
0.05
0.1
0.2
0.3
0.4
0.5
0.8
Power
0.6
0.4
0.2
0.0
0
(b)
5
Figure 3. Power curves arising from the monitoring of the krill population as a function of time since the start of fishing for a krill
population with (a) low growth and mortality and (b) high growth and mortality (see methods for details).
related science programmes, are available, that we have the ecological understanding and data required for
the retrospective power analysis.
To suggest that the monitoring programme at Bird Island may have limited power to detect changes
arising as a result of a krill fishery might have implications associated with future investment in the
programme. However, it is believed that such risks are outweighed by avoiding a misapprehension over the
Copyright # 2008 John Wiley & Sons, Ltd.
Aquatic Conserv: Mar. Freshw. Ecosyst. 17: S79–S92 (2008)
DOI: 10.1002/aqc
S88
K. REID ET AL.
1.0
0
0.01
0.05
0.1
0.2
0.3
0.4
0.5
0.8
Power
0.6
0.4
0.2
0.0
0
5
(a)
10
Years
15
20
10
Years
15
20
1.0
0
0.01
0.05
0.1
0.2
0.3
0.4
0.5
0.8
Power
0.6
0.4
0.2
0.0
0
(b)
5
Figure 4. Power curves arising from the krill predators CSI as a function of time since the start of fishing for a krill population with (a)
low growth and mortality and (b) high growth and mortality (see methods for details).
ability to detect change. If the current monitoring programme was used to provide information on the
impact of fishing on dependent species, fishing could be allowed to continue, or even increase, on the basis
of the inability to detect an effect (at a 95% confidence level) rather than the absence of such an effect of the
fishery. Therefore, identifying and addressing the capabilities/limitations of monitoring programmes that
Copyright # 2008 John Wiley & Sons, Ltd.
Aquatic Conserv: Mar. Freshw. Ecosyst. 17: S79–S92 (2008)
DOI: 10.1002/aqc
S89
THE POWER OF ECOSYSTEM MONITORING
Table 2. The power to detect the effect of a fishery at different levels of a using g ¼ 0:2 in krill populations with a differing levels of krill
variability
Time period
Low G þ M
High G þ M
krill
CSI
krill
CSI
10 years
20 years
a
a
0.05
0.1
0.2
0.05
0.1
0.2
0.24
0.13
0.14
0.12
0.41
0.32
0.23
0.25
0.66
0.70
0.42
0.48
0.35
0.23
0.20
0.22
0.53
0.50
0.32
0.38
0.76
0.84
0.54
0.59
have a role in decision making is critical to their use in management (Legg and Nagy, 2006; Taylor et al.,
2007).
In considering the purpose of such monitoring it is important to recognize that we are only considering
here the competitive effects of a fishery. Population monitoring can also detect important direct impacts of
fisheries, e.g. the unsustainable declines in populations of albatrosses at Bird Island arising from incidental
mortality associated with long-line fishing (Croxall et al., 1998; Prince et al., 1998). Despite the lack of any
competition between the Patagonian toothfish Dissostichus eleginoides fishery and albatrosses there was a
considerable impact on the albatross populations. Recently a quantifiable rate of incidental mortality of
Antarctic fur seals in the krill fishery has become apparent (Hooper et al., 2005); thus there is the potential
for direct as well as indirect effects of the krill fishery on krill-dependent species.
Increasing the power of monitoring
One of the defining characteristics of the krill population in the Southern Ocean is the high level of inter-annual
variability in abundance and distribution. This is especially so at South Georgia where the amplitude of
variability is much greater than elsewhere in the Scotia Sea (Brierley et al., 1998). This is probably a
consequence of episodic recruitment of krill into an adult population that has relatively high rates of growth
and mortality (Reid et al., 2002). A consequence of this high level of variability in krill abundance is the
resulting large variability in the reproductive success of krill-dependent predators, in particular the occurrence
of years of extremely low krill abundance associated with almost complete breeding failure (Croxall et al.,
1999). This extreme range of responses of krill predators is a key factor in our ability to parameterize the nonlinear relationships between krill abundance and reproductive performance of predators in that region (Reid
et al., 2005). However, while this natural variability is key to understanding the form of the relationship
between krill abundance and predator performance, there is a large ‘cost’ in terms of the reduction in the
power to detect change brought about by fishing. Nevertheless the non-linear response of predators to changes
in the abundance of krill acts as a filter of the variability in krill, since the high values for krill are constrained
by the asymptotic value for the predator response index. This means that values of krill abundance above the
asymptotic value results in the same CSI value and there is a consequential reduction in the overall variability
in the CSI. As there is a reduction in power associated with greater variability this may provide the potential
for increased power to detect a negative effect of krill fishing using the CSI compared to the direct
measurement of the krill population. However, we recognize that the use of a deterministic relationship
between krill abundance and predator response is a limitation of the current study.
Variability in the abundance of krill represents the sum of the effects of the environmental processes
influencing krill recruitment, the natural growth and mortality of the adult population and the effect of
fishing. Therefore, in order to detect the effects of krill fishing, there is a need to identify and to quantify the
Copyright # 2008 John Wiley & Sons, Ltd.
Aquatic Conserv: Mar. Freshw. Ecosyst. 17: S79–S92 (2008)
DOI: 10.1002/aqc
S90
K. REID ET AL.
component of variability that arises from fishing. The role of changes in the physical environment on
variability in krill populations is becoming better understood and recent analysis of long-term monitoring
data has provided new insights into the role of environmental processes, mediated through krill abundance,
on krill-dependent predators (Forcada et al., 2006). Indeed it is the data derived from the long-term
monitoring of penguins and seals that has provided the most compelling linkages between large-scale
environmental forces, such as El Nino, and changes in reproductive output of krill-dependent species
(Fraser and Hofmann, 2003; Trathan et al., 2006; Murphy et al., 2007). Developing a better understanding
of the role of environmental processes on variability in krill abundance would, in effect, allow the
environment to be included as a covariate in the analysis of monitoring data. This would effectively control
for the environmentally driven component of the overall variability and increase the power to detect change
arising specifically from the effects of the krill fishery. In this analysis we have used values of recruitment
drawn from a defined statistical distribution, however, improved understanding of the environmental
drivers of recruitment, including the potential for autocorrelation (Murphy et al., 2007), could lead to a
more informed parameterization.
Choosing a
The lessons from the history of commercial fishing, and a recognition of the uncertainties in our
understanding of ecosystem function, make the precautionary approach a compelling option. However,
using the ‘conventional’ approach to statistical analysis (i.e. a ¼ 0:05), to detect a difference before and
after fishing requires us to be 95% certain that there is an effect of fishing before the null hypothesis of no
effect is rejected. It may be more appropriate to consider the choice of a in the context of the relative costs
of making type I or type II errors, rather than on the basis of statistical dogma (Field et al., 2004). In
essence, the cost of a type I error may be to implement action that adversely effects the operation of the
fishery when such action may not have been required. In contrast, the cost of a type II error is to allow
fishing to continue while negative impacts are occurring that may only become detectable when they are
sufficiently severe to be potentially beyond the point at which remedial action is likely to be effective
(Peterman, 1990). Indeed, Maxwell and Jennings (2005) considered that increasing the risk of a type I
error by choosing a ¼ 0:20 would be appropriate in dealing with detecting declines in populations
of fish species where there was a risk of extinction if the decline remained undetected and action was not
taken.
In our case using a ¼ 0:20 is likely substantially to increase the ability to detect a change in some
indicator that might be used to trigger management action. In such situations it may be necessary to ensure
that the management response is proportionate (i.e. that it takes into account the level of uncertainty).
For example, if, using a ¼ 0:20; an effect was detected for three consecutive years, then management
action would be implemented, whereas if the same effect was detected with a ¼ 0:05 in any one year then
that management action would be implemented immediately. The potential for a more objective approach
to scalable management decisions, based on a risk evaluation of the consequences of type I and type II
errors, has been suggested by a range of studies (Mapstone, 1995; Field et al., 2004). However, a
clear strategy for applying such evaluations, from which levels of a and power can be chosen, remains
elusive.
Monitoring for the future
For fisheries management systems that seek to take account of the effects of fishing on dependent species,
the problem of a high risk of failing to detect such an effect when one is present, thereby compromising
management objectives, is a critical problem. The CEMP data provide an excellent illustration of the
likelihood of this problem happening and how it might be addressed in ways relevant to precautionary
Copyright # 2008 John Wiley & Sons, Ltd.
Aquatic Conserv: Mar. Freshw. Ecosyst. 17: S79–S92 (2008)
DOI: 10.1002/aqc
THE POWER OF ECOSYSTEM MONITORING
S91
approaches in many other ecosystems. However, considerable further complementary work is needed,
including within the CCAMLR context, in order to:
(i) interpret ‘to detect the effects of fishing’ in an operationally useful way, e.g. by re-defining it as ‘ to
detect when the effects of fishing on process x have exceeded the agreed limit/reference point’. This
would enable the design of the monitoring programme to be prospectively evaluated to ensure that it
provided sufficient statistical power to detect the required change;
(ii) acknowledge and accept an appropriate trade-off between the specificity of the aims of the monitoring
programme and the recognition of the uncertainty in ecosystem function. Such a process would need
to include an assessment of the costs associated with type I and type II errors and consequently the
value of a that is appropriate the level of proof and risk involved;
(iii) provide policy-makers with realistic expectations for the ability of monitoring programmes to deliver
data relevant to a specific management objective. This must include advice on how to use the
information from monitoring in ways that recognize the different levels of proof and risk associated
with decisions to initiate particular types of management action.
ACKNOWLEDGEMENTS
We would like to thank all those who have contributed to the long-term monitoring at Bird Island, Peter Rothery for
statistical advice and Roger Hewitt and one anonymous referee for helpful comments on the manuscript.
REFERENCES
Agnew DJ. 1997. The CCAMLR ecosystem monitoring programme. Antarctic Science 9: 235–242.
Boyd IL, Murray AWA. 2001. Monitoring a marine ecosystem using responses of upper trophic level predators. Journal
of Animal Ecology 70: 747–760.
Brierley AS, Demer DA, Hewitt RP, Watkins JL. 1998. Concordance of inter-annual fluctuations in densities of krill
around South Georgia and Elephant Islands: biological evidence of same-year teleconnections across the Scotia Sea.
Marine Biology 134: 675–681.
Butterworth DS, Gluckman GR, Thomson RB, Chalis S, Hiramatsu K, Agnew DJ. 1994. Further computations of the
consequences of setting the annual catch limit to a fraction of the estimate of krill biomass from a survey. CCAMLR
Science 1: 81–106.
CCAMLR. 2000. Report of the 19th Meeting of the Scientific Committee. Hobart, Australia.
Constable AJ, de la Mare WK.1996. A generalised model for evaluating yield and the long-term status of fish stocks
under conditions of uncertainty. CCAMLR Science 3: 31–54.
Constable AJ, de la Mare WK, Agnew DJ, Everson I, Miller D. 2000. Managing fisheries to conserve the Antarctic
marine ecosystem: practical implementation of the Convention on the Conservation of Antarctic Marine Living
Resources (CCAMLR). ICES Journal of Marine Science 57: 778–791.
Croxall JP. 2006. Monitoring predator-prey interactions using multiple predator species: the South Georgia experience.
In Managing Marine Ecosystems, Boyd IL, Wanless S, Camphuysen CJ (eds). Cambridge University Press:
Cambridge, UK.
Croxall JP, Nicol S. 2004. Management of Southern Ocean fisheries: global forces and future sustainability. Antarctic
Science 16: 569–584.
Croxall JP, Prince PA. 1979. Antarctic seabird and seal monitoring studies. Polar Record 19: 573–595.
Croxall JP, Prince PA, Ricketts C. 1985. Relationships between prey life-cycles and the extent, nature and timing of seal
and seabird predation in the Scotia Sea. In Antarctic Nutrient Cycles and Food Webs, Siegfried WR, Condy PR, Laws
RM (eds). Springer: Berlin; 516–533.
Croxall JP, McCann TS, Prince PA, Rothery P. 1988. Reproductive performance of seabirds and seals at South
Georgia and Signy Island, South Orkney Islands, 1976–1987: implications for Southern Ocean monitoring studies. In
Antarctic Ocean and Resources Variability, Sahrhage D (ed.). Springer: Berlin; 261–285.
Croxall JP, Prince PA, Rothery P, Wood AG. 1998. Population changes in albatrosses at South Georgia. In Albatross
Biology and Conservation, Robertson G, Gales R (eds). Surrey Beatty & Sons: Chipping Norton, Australia.
Copyright # 2008 John Wiley & Sons, Ltd.
Aquatic Conserv: Mar. Freshw. Ecosyst. 17: S79–S92 (2008)
DOI: 10.1002/aqc
S92
K. REID ET AL.
Croxall J, Reid K, Prince P. 1999. Diet, provisioning and productivity responses of marine predators to differences in
availability of Antarctic krill. Marine Ecology Progress Series 177: 115–131.
de la Mare WK, Constable AJ. 2000. Utilising data from ecosystem monitoring for managing fisheries: development of statistical
summaries of indices arising from the CCAMLR ecosystem monitoring programme. CCAMLR Science 7: 101–117.
Forcada J, Trathan PN, Reid K, Murphy EJ, Croxall JP. 2006. Contrasting population changes in sympatric penguin
species in association with climate warming. Global Change Biology 12: 411–423.
Field SA, Tyre AJ, Jonzen N, Rhodes JR, Possingham HP. 2004. Minimizing the cost of environmental management
decisions by optimizing statistical thresholds. Ecology Letters 7: 669–675.
Fraser WR, Hoffmann EE. 2003. A predator’s perspective on causal links between climate change, physical forcing and
ecosystem response. Marine Ecology Progress Series 265: 1–15.
Garcia SM, Cochrane KL. 2005. Ecosystem approach to fisheries: a review of implementation guidelines. ICES Journal
of Marine Science 62: 311–318.
Greenstreet SPR, Rogers SI. 2006. Indicators of the health of the North Sea fish community: identifying reference levels
for an ecosystem approach to management. ICES Journal of Marine Science 63: 573–593.
Hooper J, Clark JM, Charman C, Agnew DJ. 2005. Seal mitigation measures on trawl vessels fishing for krill in
CCAMLR subArea 48.3. CCAMLR Science 12: 195–205.
Laws RM.1977. The significance of vertebrates in the Antarctic marine ecosystem. In Adaptations within Antarctic
ecosystems, Llano GA (ed.). Smithsonian Institution: Washington, DC; 411–438.
Legg CJ, Nagy L. 2006. Why most conservation monitoring is, but need not be, a waste of time. Journal of
Environmental Management 78: 194–199.
Mapstone BD.1995. Scalable decision rules for environmental-impact studies } effect size, type-I, and type-II Errors.
Ecological Applications 5: 401–410.
Maxwell D, Jennings S. 2005. Power of monitoring programmes to detect decline and recovery of rare and vulnerable
fish. Journal of Applied Ecology 42: 25–37.
Murphy EJ, Reid K. 2001. Modelling Southern Ocean krill population dynamics: biological processes generating
fluctuations in the South Georgia ecosystem. Marine Ecology Progress Series 217: 175–189.
Murphy EJ, Watkins JL, Reid K, Trathan PN, Everson I, Croxall JP, Priddle J, Brandon MA, Brierley AS, Hofmann
E. 1998. Interannual variability of the South Georgia marine ecosystem: biological and physical sources of variation
in the abundance of krill. Fisheries Oceanography 7: 381–390.
Murphy EJ, Watkins JL, Trathan PN, Reid K, Meredith MP, Thorpe SE, Johnston NM, Clarke A, Tarling GA, Collins
MA et al. 2007. Spatial and temporal operation of the Scotia Sea ecosystem: a review of large-scale links in a krill centred
food web. Philosophical Transactions of the Royal Society of London Series B } Biological Sciences 236: 113–148.
Nicol S, Endo Y. 1999. Krill fisheries: development, management and ecosystem implications. Aquatic Living Resources
12: 105–120.
Peterman RM.1990. Statistical power analysis can improve fisheries research and management. Canadian Journal of
Fisheries and Aquatic Sciences 47: 2–15.
Priddle J, Croxall JP, Everson I, Heywood RB, Murphy EJ, Prince PA, Sear CB. 1988. Large-scale fluctuations in
distribution and abundance of krill } a discussion of possible causes. In Antarctic Ocean and Resources Variability,
Sahrhage D (ed.). Springer: Berlin; 169–182.
Prince PA, Croxall JP, Trathan PN, Wood AG. 1998. The pelagic distribution of South Georgia albatrosses and their
relationships with fisheries. In Albatross Biology and Conservation, Robertson G, Gales R (eds). Surrey Beatty &
Sons: Chipping Norton, Australia; 137–167.
Reid K, Croxall JP. 2001. Environmental response of upper trophic-level predators reveals a system change in an
Antarctic marine ecosystem. Proceedings of the Royal Society. London Series B 268: 377–384.
Reid K, Watkins J, Croxall J, Murphy E. 1999. Krill population dynamics at South Georgia 1991–1997, based on data
from predators and nets. Marine Ecology Progress Series 117: 103–114.
Reid K, Murphy EJ, Loeb V, Hewitt RP. 2002. Krill population dynamics in the Scotia Sea: variability in growth and
mortality within a single population. Journal of Marine Systems 36: 1–10.
Reid K, Croxall JP, Briggs DR, Murphy EJ. 2005. Antarctic ecosystem monitoring: quantifying the response of
ecosystem indicators to variability in Antarctic krill. ICES Journal of Marine Science 62: 366–373.
Stewart-Oaten A, Murdoch WW, Parker KR. 1986. Environmental-Impact Assessment - Pseudoreplication in Time.
Ecology 67: 929–940.
Taylor BL, Martinez M, Gerrodette T, Barlow J, Hrovat YN. 2007. Lessons from monitoring trends in abundance of
marine mammals. Marine Mammal Science 23: 157–175.
Trathan PN, Murphy EJ, Forcada J, Croxall JP, Reid K, Thorpe SE. 2006. Physical forcing in the southwest Atlantic:
ecosystem control. In Managing Marine Ecosystems, Boyd SL, Wanless S, Camphuysen CJ (eds). Cambridge
University Press: Cambridge, UK.
Copyright # 2008 John Wiley & Sons, Ltd.
Aquatic Conserv: Mar. Freshw. Ecosyst. 17: S79–S92 (2008)
DOI: 10.1002/aqc
Документ
Категория
Без категории
Просмотров
5
Размер файла
355 Кб
Теги
monitoring, ecosystems, power
1/--страниц
Пожаловаться на содержимое документа