of Longline Catchability Models in Assessments of Pacific Highly Migratory
reports (PDF): FY
The proposed study will concentrate on two aspects that largely affect
longline catchability - the vertical distribution of hooks and the vertical
distribution of a species catchability based on depth and habitat. Emphasis
will be given to applying statistical models to the Japanese longline
fishery given the long history (>50 yrs) of catch and effort data and
the Hawaii-based longline fishery that can be analyzed at a variety
of spatio-temporal scales and have additional information on gear configuration
from logbook and scientific observer programs. The species of interest
include tunas (bigeye, yellowfin and albacore), billfishes (blue and
striped marlin, swordfish) and sharks (primarily blue shark).
The major objectives are:
1. Quantifying longline gear depth and variability
2. Reconsideration of environmental parameters that affect longline
3. Improvements to the statHBS and public domain release
4. Depth and habitat catchability comparisons and model validations.
Objective 1. Quantifying longline gear depth and variability
Central to each standardization method is the vertical distribution
of longline hooks. Longline maximum fishing depth is a key parameter
allowing the estimation of the vertical distribution of longline hooks.
However, for a given fishing operation, maximum fishing depth is variable
amongst different baskets along the mainline. This variability is a
primary consequence of ocean current effects. Thus, for a given fishing
operation, maximum fishing depth corresponds to a distribution of depth
values observed at the middle of the mainline between buoys. In this
context, data from monitored longline experiments carried out in the
central south Pacific within the ECOTAP project (1995-1999) will be
considered to develop an empirical model for estimating maximum fishing
depth distributions according to fishing tactic information and ocean
currents from output of Ocean General Circulation Models.
Objective 2. Reconsideration of environmental parameters that
affect longline catchability
Predicted fields of key variables (e.g. ambient temperature, oxygen,
thermocline gradient, and deep scattering layer) will be produced and
efforts will concentrate on several focus areas with extensive archival
tag data (e.g. Coral Sea, Hawaii and eastern Pacific Ocean). Specific
focus areas may lead to a better determination of environmental preferences
or suggest different formulations for use in the statHBS or GLM standardizations.
Ocean dynamics occur on a variety of spatial and temporal scales.
Analyses will be initially conducted with environmental data on spatial
scales that correspond to fishery data such as 1º and 5º per month data.
Objective 3. Improvements to the statHBS and public domain release
The statHBS model currently exists as undocumented AD Model Builder
code. The model will be re-written and improvements will include an
interface with R software for increased diagnostic capabilities. Environmental
data are required for the model and acquisition of such data could follow
similar web-based protocols as implemented in the current Kalman filter
geolocation package with SST (Nielsen et al. in press). A composite
GLM and statHBS model is also envisaged whereby environmental information
are incorporated into a GLM approach with covariates.
The gear component can be parameterized in the statHBS framework. The
species composition of the catch has changed over time in the longline
fishery due to gear changes and the abundance of individual species.
One concept is to take a multispecies approach in the statHBS estimation
process as vertical distribution in catchability differs among species.
For example, gear and habitat or depth preferences could be estimated
for species that have demonstrated niche differences, such as yellowfin
(mixed layer, upper thermocline), albacore (upper thermocline) and bigeye
tuna (lower thermocline). Gear attributes could be parameterized and
bounded by the qualitative results from Objective 1. The habitat preference
component of the model will be further developed to include covariates
that may influence habitat preference (e.g. ocean bio-ecological area,
Objective 4. Depth and habitat catchability comparisons and model
Analytical comparisons are required to address the depth versus habitat
argument. At least three methods of estimating catchability are envisaged:
1) habitat in the statHBS framework, 2) deterministic catchability with
depth (coefficients from Ward and Myers 2005) and 3) statistically estimating
catchability with depth (statHBS configured with depth instead of habitat).
Values of the objective functions will be compared along with temporal
and spatial residual analyses.
Longline monitoring studies can further address the depth versus habitat
argument and may provide validation for a particular approach. Gear
depth and shoaling were documented on 599 longline sets (266 tuna, 333
swordfish) in the Hawaii-based fishery (Bigelow et al. 2006). Sets were
monitored with time-depth recorders (TDRs) and scientific observers
recorded the hook number and size of each fish caught, consequently
the actual vertical distribution is known for ~20,000 individuals representing
13 WPRFMC Pacific Pelagic Management Utilization Species. Models based
on vertically distributing a species by depth or habitat could be compared
with the detailed hook monitoring.
The robustness of various standardization methods could be assessed
by applying standardization models to a simulated fish population. This
approach was taken in the Atlantic Ocean by the ICCAT Working Group
on Assessment Methods. A simulator was developed to generate Japanese
catch and effort data for blue and white marlin from 1956 to 1995 (Goodyear
2003, 2005a). While the project is ongoing, recent results indicate
that evaluations of GLM and deterministic habitat-based models using
simulated longline data failed to identify a useful CPUE standardization
methodology (Goodyear 2005b). This unsatisfactory result could be a
consequence of problems with the simulator, the assumptions or data
used in the simulations, or the standardization methods themselves.
This proposed project would consider simulations for the Pacific Ocean,
but careful planning would be necessary because the methodological considerations
are complex. The simulator code has been provided by P. Goodyear to
two of the PIs should the project consider the use of model testing
through a simulator or operational model.
funding for this 2-year project estimated to be available mid-2006.
K., Musyl, M.K., Poisson, F., Kleiber, P., 2006. Pelagic longline gear
depth and shoaling. Fish. Res. 77:173-183.
• Goodyear, C.P., 2003. SEEPA - A data simulator for testing alternative
longline CPUE standardization methods. SCRS/2003/032.
• Goodyear, C.P., 2005a. Simulated Japanese longline CPUE for blue and
white marlin. SCRS/2005/032.
• Goodyear, C.P., 2005b. Performance diagnostics for the longline CPUE
• Nielsen, A., Bigelow, K., Musyl, M., Sibert. J., In Press. Improving
light based geo-location by including sea surface temperature. Fish.
• Ward, P., and Myers, R.A. 2005. Inferring the depth distribution of
catchability for pelagic fishes and correcting for variations in the
depth of longline fishing gear. Can. J. Fish. Aquat. Sci. 62: