Geoscience Reference
In-Depth Information
compare predictive capabilities of both types of model ('linear' autoregressive
moving average models versus 'non-linear' artifi cial neural network models) in rela-
tion to Pacifi c halibut ( Hippoglossus stenolepis ) catch per unit effort (CPUE), based
on daily catches from May-September 1998-2003. The autoregressive moving
average models produced population forecasts with explained variance levels ( R 2 )
of between 9.0-39.8 percent and standard errors of prediction of around 50 percent.
The artifi cial neural network models produced explained variance levels ( R 2 ) of
between 37.2-91.0 percent and standard errors of prediction of around 16 percent.
In this case, the non-linear models were far more effective in predicting halibut
CPUE, although there were notable differences between the explained variances of
the different non-linear models. The diffi culties faced by both types of model in
explaining variance levels were probably due to limitations in the catch data used
to calibrate and validate the models, which is a common problem in predictive
modelling of resources. Nevertheless, investigations such as those conducted by
Czerwinski et al. (2007) indicate that incorporating non-linearity in catch and
population forecasting is essential, as fi sheries may be particularly prone to rapid
threshold transitions in population dynamics, and fi shing to MSY can potentially
lead to sudden stock collapses.
Pitchford et al. (2007) take this principle further and present a theoretical system
where catch limits are set based on a simple deterministic model constructed using
differential linear equations. Based on such a model, MSY is a fi xed value and is
sustainable, and so no further evaluation is needed. They then introduce elements
of stochasticity and non-linearity to the recruitment and catch variability within the
system (including human error), demonstrating that using a fi xed MSY value would
eventually lead to population collapse. They then consider two further management
techniques: (i) harvest control, whereby if the population falls below a critical value
(e.g., half the total carrying capacity), the yield is reduced to allow some recovery;
and (ii) marine protected areas (MPA) wherein yield is taken from only half of the
fi sh population (essentially leaving a protected area of ocean which is not harvested).
This effectively guarantees some population survival, creates a buffer against vari-
ability, and allows continued resource harvest (fi gure 25.7). Testing the validity of
these sorts of models using empirical data is important for the future of this industry,
as is further research to investigate the signifi cance of the many biotic and abiotic
interactions in fi shery ecosystems (e.g., Gertseva and Gertsev, 2006). Above all,
management should allow for the maintenance of sustainable populations and
resilient ecosystems.
Managing Variability
As noted above, variability is essential to ecosystem functioning and resilience and
should not be ignored by resource managers. Indeed, it can now be regarded as an
indicator of ecosystem health and can be a positive aspect of management. This is
evidenced by the concept of 'natural variability', which can be best defi ned as 'the
ecological conditions, and the spatial and temporal variation in these conditions,
that are relatively unaffected by people, within a period of time and geographical
area appropriate to an expressed goal' (Landres et al., 1999, p. 1180). This concept
is an explicit acknowledgement that an understanding of ecosystem variability prior
to anthropogenic disturbance is essential for sustainable resource management. The
concept of natural variability has two main principles: (i) that past conditions and
Search WWH ::




Custom Search