CPUE analysis methods, progress and plans for 2011 Simon Hoyle
Introduction Methods used in 2010 Work with operational data Background information about problems in CPUE and development needs Plans for 2011 development Plans for 2011 assessments
Data JP DWLL aggregated data, 1952-present
Spatial stratification Fig. 3
JP effort
JP catch
Nominal CPUE
JP spatial strata
Standardization Linear model on log(catch), normally distributed error assumed Terms for time, location (5° lat x long), HBF (set depth), effort (hooks),
Bigeye 2010 approach vs nominal
Yellowfin 2010 approach vs nominal
Regional scaling CPUE index a measure of density. Doesn’t account for size of each region and magnitude of CPUE and therefore abundance.
Nominal CPUE varies among regions (YFT)
But CPUE GLM indices are normalized to average 1
Regional re-weighting Widest spatial distribution of JP LL fleet, before shift in fishing effort to BET
Sum the 5° indices by region to get relative abundance
Region specific CPUE index = relative abundance between regions i.e. region scaling factors. Regional scaling – weighting each region in the model
Then adjust the average CPUE indices for the same period to the right level
Regional scaling assumptions Assume constant LL q’s between regions. Assumes equivalent vertical distribution over entire model domain. Both assumptions are unreliable
15 N 5 N 5 S 15 S Longitude Depth m
Regional scaling BigeyeYellowfin BigeyeYellowfin
Japanese LL effort distribution (1*1 deg) Number of hooks
Summary Assessments critically dependent on JP LL CPUE data: regional structure, regional scaling, relative abundance. Assume constant catchability (q) over history of fishery (for standardised CPUE). Contraction of fishing range by JP fleet.
Analyses of Japanese operational CPUE data for bigeye tuna Simon Hoyle, Hiroshi Shono, Hiroaki Okamoto, & Adam Langley WCPFC-SC SA-WP-02_Bigeye_Operational_CPUE
Origin of this work Japan (NRIFSF) and SPC arranged collaboration, to use set-by-set longline data compiled from logsheets submitted by Japanese longline fishermen, with aims of: – standardizing Japanese longline CPUE of bigeye tuna; and – Estimating the historical trend of Japanese longline catchability of bigeye tuna.
Some poorly understood observations Standardized indices in equatorial areas decline less than nominal Inconsistencies with other assessment data – Stable or increasing bigeye longline CPUE – Yellowfin CPUE decline greater than expected
Need Primary need is to understand the processes driving CPUE trends For understanding, operational data are far more useful than aggregate data
Analyses carried out Data preparation & summaries Catchability analyses Regional scaling Comparisons of indices with those estimated from aggregated data
Data summaries 5x5 squares fished through timeCPUE by species through time
Summaries unique to operational data Proportion of sets with zero catch (by species, time, and region)
Fishing patterns in region 3 North of 10N, effort since 1995 has caught more albacore and less bigeye This pattern is associated with individual vessels with low bigeye catch rates
Equatorial modelFull region 3
Region 3 models Different CPUE trends in equatorial region and north of 10N – Probably because of arrival of more albacore, and the albacore-targeting fleet, after 1995 – i.e. targeting (not abundance) affecting CPUE Much higher bigeye catch rates (& abundance) south of 10N Indices should focus on core area
Regional scaling BET: yellow is high CPUE
Catchability analyses Results indicated two important influences on catchability – Changes in individual vessels through time – Changes in targeting through time
Changes in individual vessels Equatorial model Red CPUE series includes vessel effect Difference about 19% over 30 years
HBF effects HBF effects region 1 to 4 Equatorial regions have higher CPUE at lower HBF (shallower sets) R1 R2 R3 R4
Data aggregation Long-term trend Short-term trends
Conclusions in 2010 Very encouraging collaboration on operational CPUE of Japanese longline data Further development high priority for stock assessments – Primary need is to understand the processes that affect CPUE Changes in targeting and changes associated with individual vessels both affect catchability through time For 2010, indices for region 3 focused on the equatorial area 0-10N
Background about problem areas Data weighting and aggregation Targeting changes and vessel behavior Regional scaling ‘Filling the gaps’ – assumptions about unfished areas
Weighting & aggregation Data weighting affects results (Campbell 2009) – Important where CPUE varies in space, and effort concentration changes – Bigeye CPUE varies in space, effort is contracting & concentrating into high CPUE areas Operational index trends positively biased – More weight to areas with more sets Aggregated indices also need investigation – Same weight to all strata, but less fished areas have fewer strata – Increasing concentration at scales smaller than 5x5 Change approach to weighting
‘Filling the gaps’ Exclusion of fleet from EEZ waters. Contraction of JP fleet. GLM lat/long variable generally can deal with this. However, potential biases in index if spatial (lat/long) differences in CPUE trend, esp. in case of some cells not being included in analysis. Missing cells – assumed same CPUE trend. May not be valid.
Changes in targeting – YFT vs BET Increasing focus of longliners on bigeye tuna through time Region 3 YFT catch rates have declined, while BET have not We cannot estimate BET/YFT targeting – YFT and BET CPUE positively correlated at the set level – HBF is not a consistent target indicator Targeting substantially affects CPUE We need to understand the processes better
Results from Langley 2007 Effort appears more aggregated – Collaboration between vessels – Better information Increasing proportion of sets in locations with high BET CPUE Increasing probability of moving after sets with low BET CPUE, but not low YFT
Generalize fishing process: individual trip Fishing location “cluster” Movement (60+ km) Intermediate set Entry Exit
Higher BET CPUE within a cluster (a). Lower rate of decline in BET CPUE within cluster (b). More sets within a cluster (c). Increasing proportion of sets in locations with high BET CPUE = “hyperstability”
Plans for 2011 Development: Operational data analyses in Shimizu, Japan (if possible) – Find approach to deal with data weighting issue – Catchability analyses for YFT – Comparative analyses with oceanographic data – Derive indices , for both BET & YFT Assessments 2011 – Use operational data indices – TW indices (regions 4 and 6)? Regional scaling?