Statistical Modelling in the Northern Prawn Fishery
The NPF opens up a variety of challenging statistical
problems that are currently being addressed by the Environmetrics team.
These include
-
The species split problem. Logbook
records for prawn catches often only record the total catch for a
species group, but this species group is a collection of several good
biological species, the progress of each of which has to be assessed
separately to guard against differential stock mortality. Thus Tiger
prawns, for example, are really two species of prawns: Grooved Tiger
prawns (Penaeus semisulcatus) and Brown Tiger prawns (P.
esculentus). The logbooks themselves can give us no information on
the separate catch of each species, so survey data has to be used to
build up a model for imputing a species split in the logbook data and
estimating its accuracy. The survey data gives sex and carapace data
on individual animals either sampled from actual commercial catches or
from special purpose trawls on research vessels, stretching over 25
years.
The logbook data is only recorded spatially to the
nearest 6' by 6' grid square. The classical method of species split merely
uses the accumulated survey data for the relevant grid square and its
nearest neighbours to infer a split proportion. Statistical modelling
using generalized additive models with some interaction terms are proving
more effective in estimating the split proportion and are showing that
such things as time of year are important variables: the species split
proportions vary regularly throughout the year in response to different
movements of species, largely according to their breeding cycles. The new
methods are also raising the possibility of a long-term trend, with Brown
Tigers gradually declining overall as a proportion of the joint stock.
This has not yet been credibly proven, but the possibility has been
established and the modelling methods are both showing the urgent need for
more survey data to investigate it further as well as giving good guidance
on the most effective design for such surveys.
-
The Fishing Power Analysis project .
Statistical modelling plays a crucial role in this project in trying
to isolate the effect of gear and technology on prawn catches while
keeping prawn abundance locally constant. This will involve either
random effects modelling of fishing catches to cater for such
unmeasurable effects as skipper skill and fleet collaboration (which
induces correlations between catches of different vessels). The
project is at an early stage of development but already the results
are beginning to show some unexpected features.
-
The marine surrogates project . This
project is concerned with optimally using data collected over many
years to build up a picture of the major benthic communities in the
NPF, with the ultimate aim being the identification of surrogate
variables that will allow rapid but accurate determination of areas
important for management and conservation. The study involves
correspondence analysis (and some original extensions of it) as well
as spatial modelling and generalized linear modelling.
Contact: Allan Adolphson
Ph: (02)9325 3261 Fax: (02) 9325 3200
|
To
top