draft status of the u.s. petrale sole resource in 2012 star panel melissa haltuch 1, kotaro ono 2,...
TRANSCRIPT
Draft Status of the U.S. petrale sole resource in 2012
STAR Panel
Melissa Haltuch1, Kotaro Ono2, Juan Valero3
1NWFSC, Seattle2UW, SAFS, Seattle
3CAPAM, La Jolla
13 May 2013
Outline Introduction Data
– Fishery Independent– Biological– Fishery Dependent
Previous Modeling Responses to 2011 STAR panel General Model Description
Base Case Model– Sensitivity Analysis– Retrospective Analysis– Historical Assessment Analysis– Likelihood Profile– Harvest Projections
Introduction - Biology
Right eyed flounder
Gulf of Alaska to Baja California
Soft bottoms
550 m depth
No genetic work
Adults migrate seasonally
No strong indication of multiple stocks
Introduction - Fishery
1876 off of San Francisco, CA
1884-1885, established by 1937 in OR
Began about 1932, established by 1936 in
WA
Early concerns about stock depletion in the
1950s
Targeting of winter spawning aggregations
developed through the 1950s and 1960s
By 1980s winter catches exceeded summer
catches in many years
Introduction - Management
Assessment
Status Year(s) OFL ACL
1984-1999
Near Target 1983-2006
3200-2762
3200-2762
2005 Precautionary
2007-2009
3025-2811
2499-2433
2009-2011
Overfished 2010-2012
2751-1279
1200-1160 Management actions since the late 1990s
– Area closures– Trip limits– Gear modifications– IFQ
Triennial survey
1977 (excluded)– Depth: 91-457 meters
1980-1992– Depth: 55-366 meters
1995-2004 – Depth: 55-500 meters
Run by RACE until 2004 when run by FRAM Random trawls on systematic line transects
Triennial Survey Timing
June 15
August 15
July 15
September 15
140
160
180
200
220
240
260
1979 1984 1989 1994 1999 2004 2009
Year
Ju
lian
da
y
NWFSC survey
1999-2002
– surveyed 183-1280 meters
– Did not always go as far south
2003 through 2008
– 55-1280 meters
– 32.5° to 48.17°
Random trawl locations
Random vessels chosen each year
Survey differences
Triennial NWFSCSystematic random stations on equally spaced transects
Randomly selected within stratified random block
AK class commercial trawlers65-147 feet
WC commercial trawlers65-92 feet
High opening Nor’Eastern trawl
4 panel Aberdeen style trawl
250 feet net to doors 205 feet net to doors
Roller gear (121’ footrope) Continuous disk footrope (104’)
Bare wire bottom bridles 8” disks in wings
6’ X 9’ V-Door 5’ X 7’ V-Door
5” mesh, 3.5” codend, 1.25” liner
5.5” mesh, 5” cod end, 2” liner
30 minute tow 15 minute tow
3.0 knot towing speed 2.2 knot towing speed
Surveys
Triennial– Two time series: 1980-1992 and 1995-
2004
– Excluded the Conception area (S of 36°)
– Biomass and length frequencies
NWFSC– One time series: 2003-2012
– All areas
– Biomass, lengths, and age-at-length
Catch rates: Triennial
Catch rates: NWFSC
Density: NWFSC
Survey post-stratification
Post-stratify depth using fish length– Ontogenetic movement to deeper water– Lai et al (2005) used Bayesian change-
point analysis– Haltuch et al (2009) used a frequentist
approach and came to same result– Significant split just greater than 100 m
Mean fish length vs depth
50 100 150 200 250 300 350 400
1520
2530
3540
45
Triennal
Females
100 200 300 400
2030
4050
60
Males
100 200 300 400 500
2030
4050
NWFSC
Females
100 200 300 400 500
1520
2530
3540
45
Males
Depth (m)
Me
an
len
gth
(cm
)
Survey stratification for GLMM/GLM
Strata collapsed to satisfy condition of at least 3 positive observations in each year/area/depth stratum
Depths– Triennial (Early and Late): 55-100 meters and
100+ m– NWFSC: 55-100 m, 100-183 m, 183+ m
Areas – Triennial
Early– Shallow: Vancouver/Columbia, Eureka,
Monterey/Conception – Deep: Vancouver/Columbia/Eureka,
Monterey/Conception Late:
– Shallow: Vancouver/Columbia, Eureka, Monterey/Conception
– Deep: Vancouver/Columbia, Eureka, Monterey, Conception
– NWFSC Shallow/Middle – Vancouver/Columbia,
Eureka, Monterey, Conception Deep – Combined
Eureka-Columbia/Vancouver
-128 -126 -124 -122 -120 -118
3540
45
PSMFC areas
Longitude (°)
Latit
ude
(°)
1A
1B
1C
2A
2B
2C
3A
3S
-128 -126 -124 -122 -120 -11835
4045
INPFC areas
Longitude (°)
Latit
ude
(°)
CP
MT
EK
CL
VN
Model based biomass estimates
Delta-GLMM
– NWFSC: random year:vessel effects
– Triennial: NO random vessel effects
– Fixed effects: year, strata, depth, year:strata
Lognormal errors
MCMC’s to determine variability
Model selection – residual deviance
Survey Model Mean
Tri_early Gamma-Strata:YearFixed 7814.3
Tri_early Lognormal-Strata:YearFixed 7655.1
Tri_late Gamma-Strata:YearFixed 6903.1
Tri_late Lognormal-Strata:YearFixed 6840.5
NWFSC Gamma-Strata:YearFixed 19885.2
NWFSC Lognormal-Strata:YearFixed 19723.7
Model fit - lognormal
Survey biomass estimates
Triennial length frequencies
NWFSC length frequencies
NWFSCagefrequency
NWFSC survey length at age
Summary of survey data
NWFSC survey has higher catch rates than the triennial survey, resulting in larger biomass estimates
2004 triennial survey catch rates are on average higher than rest of triennial series
Trend in NWFSC survey peaks in 2004, declines through 2008, and increases after 2009– Smaller and younger fish observed in 2008-2010
Biological Data - Weight-Length
NWFSC Survey
Biological Data – Maturity at length
Oregon Washington 2002
Biological Data – Natural Mortality
1940s Catch Curve– M: 0.18-0.26– F: 0.19-0.21
Hoenig’s Method– 0.15 max age of 30
(female petrale sole live at least 30 years)
Hamel prior– M median: 0.206, SD: 0.16– F median: 0.151, SD: 0.206
Ageing Precision and Bias 3 Labs
– Cooperative Ageing Lab OR and CA commercial (1986-present), NWFSC Survey
– WDFW– CDFG (only samples pre ~1980s)
Surface ages – pre 1980s– OR 2001-2004
Combo method – OR 1981-1984, 1987-1988, 1991-1997 (reader issues)– WA ~1990 – 2009
Break and Burn – NWFSC survey– OR (1985-1986, 1989-1990, 1998-1999, 2007-present)– WA (2009-present)– CA (1986-present)
Ageing Error Methods
Punt et al. 2008; simulation tested Estimate ageing error assuming one reader is
unbiased – based on bomb radiocarbon age validation
Data pooled across reader Early surface age error estimate for pre-
1990s samples Sample sizes – 100’s of double and triple
reads Model selection – AIC
– Shape of bias, shape of error, minus age, and plus age
Ageing Error - Results
Pikitch Discard Data
WCGOPObserverLengths
Su
mm
er
Win
ter
WCGOP Observer Data 2002-2010– SummerSpatial distribution of observed catch
WCGOP Observer Data 2002-2010 – WinterSpatial distribution of observed catch
Fraction DiscardedDiscard/Total Catch
Fishing North winter North summer South winter South summerYear Mean SD Mean SD Mean SD Mean SD1985 0.0222 0.1103 0.0346 0.04191986 0.0215 0.1162 0.0343 0.04321987 0.027 0.1186 0.0315 0.045
2002 0.0077 0.0034 0.1856 0.0253 0.03720.024
4 0.0569 0.0158
2003 0.01 0.0064 0.1111 0.0252 0.00620.002
6 0.0325 0.0126
2004 0.0019 0.0008 0.0843 0.0244 0.05260.052
1 0.0343 0.0153
2005 0.0013 0.0009 0.0421 0.0112 0.00690.007
1 0.0122 0.0035
2006 0.0131 0.0073 0.078 0.0171 0.05980.044
6 0.036 0.0157
2007 0.0037 0.0015 0.1138 0.0232 0.01940.013
9 0.061 0.0209
2008 0.0275 0.0146 0.0502 0.0167 0.00990.005
6 0.0259 0.0147
2009 0.0253 0.0151 0.2018 0.0673 0.02210.014
7 0.0233 0.0082
2010 0.1971 0.0444 0.1037 0.0308 0.25840.071
7 0.0554 0.01192011 0.0017 0 0.037 0 0.0009 0 0.0411 02012 0.0006 0 0 0 0.0046 0 0 0
Commercial Length Comps – North
Sum
mer
Win
ter
Commercial Age Comps
Sum
mer
W
inte
r
Landings
2011 v. 2013Landings
CPUE standardization steps:
1. Data filtering
2. Identify the covariates to use/test
3. Build a regression model that best fits the
data
4. Create an index of abundance with some
credibility interval
1. Data Filtering
Spatial
Spatially defined fishing grounds
• 2003-2008
• Summer – May-October
• Shoreward of 75fm
• remove tows with flatfish catch rates in lower 10%
• Winter – Nov-Feb
• Seaward of 150fm
• Remove tows with petrale catch rate in lower 10%
Data quality
Remove
Tows outside EEZ
mid-water trawls
Tow duration ≤0.2 hours
Difference between map and logbook depths > 70 fm
Tows ≥ 300 fm (S); ≥ 400 fm (W)
Tow duration ≥ 4 hours (S); ≥ 6 hours (W)
Vessels < 5 years in fishery (sensitivity test)
Winter Nov-Dec data
Output: average CPUE (lbs/hr) by fishing trip
1. Data Filtering
Tow by tow data
Trip by trip data
SummerNorth 79752 18627
South 16405 6645
WinterNorth 10936 3984
South 2841 1307
2. Covariates to test
Models for each fleet separately:
North winter, North summer, South
winter, South summer
Time: year, bimonth
Space: spatial grid
Vessel effects: port, vessel ID, gear,
targeting
2013 model stratificatio
n
3. Model building
Build a regression model
- Data contains a lot of zero in addition to the
positive data delta (hurdle) model
- Mixed effect model with vessel as random
effect
Choose covariates through model
selection (AIC)
Check model assumptions
Changes from 2011
1. Summer data filtering corrected
2. Changed the “reference level” of the covariates during the index
standardization to be the mean (continuous) or most frequently
observed (categorical)• index of abundance can be interpreted as index per
“reference” unit
• calculate a confidence interval
3. Finer spatial stratification
4. Aggregate tow level data to trip level data
• Greater independence
5. fishing tactics covariates
6. Sensitivity to random vessel effects
7. WA and OR aggregated into North fleet
The 2011 best main effect models (determined after model selection)
Winter Summer
WA 43% 30%
OR 34% 52%
CA 40% 40%
Explained deviance
The 2011 best main effect models
+ removed tows outside EEZ
Winter Summer
WA 43% 27%
OR 34% 42%
CA 40% 40%
Explained deviance
The 2011 best main effect models
+ removed tows outside EEZ
+ change “reference” levels
Winter Summer
WA 43% 27%
OR 34% 42%
CA 40% 40%
Explained deviance
The 2011 best main effect models
+ removed tows outside EEZ
+ change “reference” levels
+ standardized the index by its geometric mean (change of scale)
Winter Summer
WA 43% 27%
OR 34% 42%
CA 40% 40%
Explained deviance
2011 best main effect models
+ removed tows outside EEZ
+ change “reference” levels
+ standardized the index by its geometric mean (change of scale)
2011 best main effect models
+ trip level data
Winter Summe
r
WA 34% 32%
OR 35% 47%
CA 34% 42%
Explained deviance
2011 best main effects models
+ trip level data
2013 best main effects models
+ trip level data
Winter Summer
WA 34% 35%
OR 34% 48%
CA 38% 44%
Explained deviance
2013 best main effects models
+ trip level data
The 2013 best main effects models
+ trip level data
+targeting covariates
Winter Summer
WA 66% 66%
OR 74% 71%
CA 72% 65%
Explained deviance
The 2013 best main effects models
+ trip level data
+targeting covariates
The 2013 best main effects models
+ trip level data
+targeting covariates
+ mixed effect model
Winter Summer
WA 44% 67%
OR 68% 71%
CA 74% 65%
Explained deviance
Model fits
Lognormal componentSummer OR Winter OR
Final index of abundance
2013 assessment has a different spatial set-up:North fishery (WA+OR) and the South fishery (CA)
Graphs are scaled so that the 2004 index = 1.
WinterSumme
r
North 78% 74%
South 74% 65%
Explained deviance
Final index of abundance with prediction interval
Conclusions
- New CPUE methods didn’t radically change the index of abundance
- Advantages of 2013 approach1. Data are more independent (compared to the tow
by tow data)
2. The spatial stratification is finer
3. Calculate the prediction interval around the standardized index of abundance
4. Some vessel behavior is taken into account through the use of targeting covariates
5. Model fit improved by including targeting covariates
DataSummary
Responses to 2011 STAR panel recommendations
Establish a formal framework and to conduct petrale sole assessments jointly with Canada. • A formal framework for joint stock assessment and management of
U.S-Canadian transboundary groundfish stocks does not exist. This stock assessment follows the PFMC terms of reference for groundfish stock assessments.
Conduct a formal review of all historical catch reconstructions and if possible stratify by month and area.• The PFMC is responsible for such reviews, resources not available. Document and review WCGOP discard estimates outside of the STAR panel process. • The WCGOP data have been documented but have not been reviewed
by the PFMC. Combine Washington and Oregon fleets in future assessments within a coast-wide model.• Washington and Oregon fleets have been combined, landings are
summarized by port. Update maturity and fecundity information.• Not updated.
Responses to 2011 STAR panel recommendations
SS3, investigate simpler, less structured models, to compare and contrast results.• Simple model comparisons show similar results to SS3 (J. Cope , pers.
comm). The length binning structure in the stock assessment should be evaluated.• The impact of changing the bin size from 2 cm to 1 cm bins was
explored. The residual patterns in the age-conditioned, length compositions from the surveys should be investigated and the potential for including time-varying growth, selectivity changes, or other possible solutions should be examined.• Options for better fitting all of the length and age data have been
explored via selectivity and fleet/model structure. A NMFS Fisheries and the Environment (FATE) funded project to investigate and conduct a meta-analysis of time-varying growth for California Current groundfish in underway.
MSE is recommended to examine the likely performance of new flatfish control rules.• The NWFSC has not had the resources available to conduct an MSE for
the PFMC flatfish control rule.
Changes from 2011 Model
SS-V3.24o
Landings summarized by port of landing rather than area of catch.
Combining the Washington and Oregon fleets into a single northern fleet.
Use of the Oregon historical landings reconstruction.
Specification of the male growth parameters to be directly estimated rather than estimated as an offset to the female growth parameters.
Use of an early, pre-1990s, age error matrix for surface ages.
Addition of data for 2011 and 2012.
Base model
Coast-wide model 12-month model with seasonal fleet structure 4 fleets Sex-specific Asymptotic selectivity Blocks on selectivity and retention Estimate growth Estimate sex-specific natural mortality
– Diffuse prior on M Estimate steepness with R. Myer’s prior Composition effective sample sizes tuned
Estimated parameters
ParameterNumber
estimatedNatural mortality (M, female) 1Natural mortality (M, male) 1
Stock and recruitmentLn(R0) 1Steepness (h) 1Ln(Early Recruitment Deviations): 1845-1958 114Ln(Main Recruitment Deviations): 1959-2009 51
IndicesLn(q) – NWFSC survey -Ln(q) – Triennial survey (early and late) -Ln (q) – North winter commercial CPUE 1Ln (q) – South winter commercial CPUE 1Beta (power) – North winter commercial CPUE 1Beta (power) – South winter commercial CPUE 1Extra SD – NWFSC survey 1Extra SD – Early Triennial 1Extra SD – Late Triennial 1Extra SD – North winter commercial CPUE 1Extra SD – South winter commercial CPUE 1
Estimated parametersParameter Number estimated
Fisheries Selectivity (asymptotic, sex specific, with retention curves)
Length at peak selectivity 4
Ascending width 4
Male parameters 1 and 2 8
Retention parameters 1, 2, and 3 12
Selectivity time block parameters (Peak) 20Retention time block parameters (Inflection, Slope, Asymptote)
36
Surveys Selectivity (asymptotic, sex specific) Length at peak selectivity 3
Ascending width 3
Male 1 parameters 1 and 2 6Individual growth
Length at age min 2
Length at age max 2
von Bertalanffy K 2
CV of length at age min 2
CV of length at age max 2
Total: 119 + 180 recruitment deviations =299 estimated parameters
Growth
Parameter Value
Females:
Length at Linf 60.32
von Bertalanffy K 0.13
CV of length at age min 0.18
CV of length at age max 0.03
Males:
Length at Linf 46.79
von Bertalanffy K 0.21
CV of length at age min 0.13
CV of length at age max 0.05
Survey and Productivity parameters
Parameter Value
Catchability, Power, Extra SD:
NWFSC survey catchability (q) 2.95
Triennial survey catchability (q) early, late 0.52; 0.73North winter commercial CPUE (Beta) 0.63 (0.15, 1.11)South winter commercial CPUE (Beta) -0.13 (-0.56, 0.3)Q_extraSD North Winter 0.082Q_extraSD South Winter 0.112Q_extraSD Triennial survey early 0.130Q_extraSD Triennial survey late 0.175Q_extraSD NWFSC -0.050 (-0.094, -0.006)
Productivity:
R0 9.82
Steepness (h) 0.84
Female natural mortality (M) 0.16
Male natural mortality (M) 0.18
Survey abundance
q=0.52 q=0.73 q=2.95
Survey selectivity
Survey lengths
Survey lengths
NWFSC survey agesFemale Male
NWFSC female survey age-at-length
NWFSC male survey age-at-length
Commercial CPUE
Time varying selectivity
Time varying retention
North fleet end year selectivity and retention
South fleet end year selectivity and retention
Commercial lengthsFemale Male
Commercial lengthsFemale Male
Commercial agesFemale Male
Commercial agesFemale Male
Discard fraction (Discard/Total catch)
Discard mean weight
Discard lengths
Discard lengths
Discard length residuals
Discard length residuals
Tuning: Sigma R and Lengths
Sigma R in = 0.40 Sigma R out = 0.35
Lengths Ages
Fleet
Variance Adjustmen
t MeaneffNMeaninputN
Variance Adjustment
MeaneffN
MeaninputN
Winter north 2.0 1.1 7 1.0Summer north 1.4 1.0 1.7 1.0Winter south 1.6 1.1 1.9 0.9Summer south 1.2 1.0 1.4 1.1Early Triennial 1.3 1.1Late Triennial 1.0 1.1NWFSC 1.0 0.9 0.3 0.7
Biomass trajectory
Spawning Biomass Depletion
Recruitment deviations
Fishing mortality
Spawning potential ratio
Management performance
Retrospectives
Model Retrospectives
Assessment Year Base 2011 2010 2009 2008 2007
SSB Unfished 31,998 31,809 32,007 32,433 31,828 32,196
2007 Depletion 0.143 0.144 0.146 0.153 0.151 0.160
2008 Depletion 0.134 0.135 0.138 0.148 0.147 0.162
2009 Depletion 0.129 0.130 0.135 0.148 0.150 0.172
2010 Depletion 0.126 0.127 0.132 0.150 0.162 0.183
2011 Depletion 0.156 0.157 0.164 0.189 0.209 0.219
2012 Depletion 0.206 0.209 0.220 0.246 0.261 0.254
2013 Depletion 0.275 0.280 0.287 0.305 0.302 0.280
Sensitivities
Red-4 fleets w/ North comps
Mirrored
Green-6 fleets
Sensitivities:Removal of 2012 survey data
Green-index
Yellow-Ages Red-Lengths
LikelihoodProfiles
LikelihoodProfiles
BetweenAssessmentModelComparison
1999 – Red 2005 – Green 2009 – Blue 2011 – Light Blue 2013 - Black
Decision Table State of nature Low Base case High M = 0.14 M = 0.16 M = 0.18
Relative probability 0.25 0.5 0.25
Management decision Year Catch
(mt) SB(mt) Depl SB (mt) Depl SB(mt) Depl
OFL
2015 3,944 11,546 0.334 11,921 0.373 12,361 0.4142016 3,914 11,568 0.335 11,847 0.370 12,198 0.4092017 3,722 11,063 0.320 11,241 0.351 11,492 0.3852018 3,485 10,422 0.302 10,526 0.329 10,702 0.3592019 3,279 9,862 0.286 9,934 0.310 10,070 0.3372020 3,133 9,446 0.274 9,523 0.298 9,651 0.3232021 3,043 9,161 0.265 9,264 0.290 9,407 0.3152022 2,990 8,967 0.260 9,108 0.285 9,273 0.3112023 2,958 8,828 0.256 9,007 0.281 9,197 0.3082024 2,935 8,721 0.253 8,934 0.279 9,142 0.306
Status quocatches
2015 2,592 11,546 0.334 11,921 0.373 12,361 0.4142016 2,592 12,336 0.357 12,611 0.394 12,946 0.4342017 2,592 12,661 0.367 12,814 0.400 13,021 0.4362018 2,592 12,750 0.369 12,791 0.400 12,884 0.4322019 2,592 12,744 0.369 12,700 0.397 12,709 0.4262020 2,592 12,705 0.368 12,605 0.394 12,559 0.4212021 2,592 12,659 0.367 12,527 0.391 12,451 0.4172022 2,592 12,616 0.365 12,469 0.390 12,379 0.4152023 2,592 12,579 0.364 12,428 0.388 12,332 0.4132024 2,592 12,547 0.363 12,397 0.387 12,300 0.412