Post on 02-Jan-2016
transcript
Spatial issues in WCPO stock assessments
(bigeye and yellowfin tuna)
Simon HoyleSPC
Introduction
• Overview of WCPO fisheries and stock assessments
• Summary of spatial issues• Deeper examination of several issues
WCPO
Complex fisheries
• Fishing methods – longline, purse seine, pole and line, other
• Fleets– JP DW LL, JP offshore LL, Korea, China, US, …
• Species– Skipjack, bigeye, yellowfin, albacore, swordfish, …
• Spatial– Oceanography, thermocline depth, seasonal changes,
convergence zones, seamounts, migrations, age effects, EEZs,
Spatial stratification
– bigeyetuna
– yellowfintuna
30
S1
0S
10
N3
0N
120E 150E 180 150W 120W
R1 R2
R3 R4
R5 R6
LPS Z
Max=39395Max=39395Max=39395Max=39395
Legend
30
S1
0S
10
N3
0N
120E 150E 180 150W 120W
R1 R2
R3 R4
R5 R6
LP
S ZMax=348801Max=348801Max=348801Max=348801
Legend
Model structure• 6 regions. Qtrly age classes (e.g. 40 bigeye).• Input data
– Catch, CPUE, size frequency, tag recapture• Population dynamics
– Recruitment: overall regional proportion, temporal trend, regional deviates, B-H SRR.
– Growth (estimated).– Age-specific natural mortality (fixed).– Movement dynamics (estimated).
• Fitting to data– LL CPUE index – shared catchability and selectivity among
regions.– Size frequency, given selectivity by fishery– Tagging data
Selected spatial issues
• Regionalization of model– Choosing the right regions
• Regional scaling of CPUE• Estimating movement among regions• Spatial effects not included in the model
Criteria for defining regions• Simplicity (“less is more”)• Homogeneous abundance trend (CPUE) in
region*• Homogeneous size data in a region*• Biogeography• Enough data from region to reliably index
abundance and size• Specific management issues (e.g. ID)• Consistency between assessments – fishery
interactions
Spatial heterogeneity – YFT CPUE trendsJP LL standardised CPUE.
GLMs include 5*5 lat/long, HPB, bet_cpue, and no_hooks.
Consistent local variation accounted for by lat-long GLM effects
Different trends require separate regions
Spatial CPUE variation
• Yellowfin in region 3, one band of latitude per plot.
• Less pronounced decline north of 10N
• Steep decline in far west (of 120E), arguably deserves separate region
17.5
7.5 2.5
-2.5 -7.5
12.5
Spatial heterogeneity - Size• Most size data 10 lat * 20 long blocks.• Compare trends in median length/weight
between 10*20 blocks.• Minimal size data from far west (of 120E)
120 140 160 180 200 220
-40
-20
020
1950
120 140 160 180 200 220-4
0-2
00
20
1960
120 140 160 180 200 220
-40
-20
020
1970
120 140 160 180 200 220
-40
-20
020
1980
120 140 160 180 200 220
-40
-20
020
1990
120 140 160 180 200 220
-40
-20
020
1950
120 140 160 180 200 220
-40
-20
020
1960
120 140 160 180 200 220
-40
-20
020
1970
120 140 160 180 200 220
-40
-20
020
1980
120 140 160 180 200 220
-40
-20
020
1990
Decadal trends in median yellowfin size by 10*20 lat/long cell.
Red fish smallYellow fish large
JP LL size data.
Consistent local effects may require separate fisheries
Different trends may require separate regions
weight length
LL 3 fishery subdivision – different sizes
30
S1
0S
10
N3
0N
120E 150E 180
R1 R2
R3 R4
R5 R6
00N 120E 00N 140E
10N 120E 10N 140E
10S 120E 10S 140E
LL ALL 3 fishery (yft) changed to exclude area approx. PNG waters.
New fishery with different selectivity, catchability. High catch in the 1950-60s.
Trends arguably different for 10S, 0N, and 10N (as for CPUE, but trends imply different things)
Region and fishery definitions
• Rule of thumb– If population trajectories differ, separate regions– If fish sizes differ, separate fisheries
• Pragmatic choice– Does it affect the results?
• May need more (smaller) regions where fish & effort are concentrated
• More data makes smaller regions possible
Scaling up size data
• Catch at size varies with – Time (cohorts moving through fishery)– Fishing method (even to vessel level)– Set (fish school by size)– Space (consistent local effects)
• Assume constant selectivity for fishery• But effort moves around region through time,
which affects size– Sampling is ad hoc, may not be representative of catch
Scaled size data
Size data
Catch
0 50 100 150
0.0
00
.02
0.0
4
WCPO scheme for scaling up size data
• Catch data (numbers of fish) from a region are aggregated at the same spatial resolution as the size data (usually 20° longitude, 10° latitude cells).
• Check size data are available from all cells that cumulatively account for 70% of the total quarterly catch from the region. If not, reject the size data from that quarter.
• Check there are at least 20 fish sampled from each of the main cells fished and at least a total of 50 fish sampled per quarter. If not, reject the size data from that quarter.
• Combine the sample data from each cell weighted by the catch in each cell (number of fish).
• Scale the overall weighted sample to the total number of fish measured in the quarter.
‘Representative’ size data• Size data changes may drive population biomass estimates• Model ‘assumes’ size changes within fishery reflect changes in
population + sampling error• Should size data reflect the catch or the population? • Length frequency data can strongly affect population trend (e.g. sth
Pacific albacore assessment)• Standard approach is to reflect catch, but this may be problematic if
there is significant size variation in space• If this occurs,
– Downweight size data sample size to reflect heterogeneity– Define more fisheries
• ‘Standardize’ size data?– Also an issue for other effects on selectivity, such as LL set depth and gear type
Regional CPUE scaling
30S
20S
10S
010
N20
N30
N
120E 140E 160E 180 160W 140W 120W 100W 80W
1 2 7
3 4 8
5 6
0
20
40
60
80
100
120
1 4 7 10 13 16 19 22 25 28 31 34 37Time
CP
UE
Region specific CPUE index
= relative abundance between regions i.e. region scaling factors.
( , ) ( ) ( )ln( ) other parametersu v u vCATCH aYRQTR bLATLONG
Area weighted GLM index
CPUE indices comparable between regions and reflect relative biomass in each region.
1. GLM model for each region.Data aggregated 5*5 lat/long, HBF, month. YR/qtr index.
2. Region scalar.
Sum coefficients within region (at HBF=5).YR/QTR index multiplied by region scalar.
( , , , ) , , ,ln LL HBF y qtr LATLONG R HBF R y R qtrCPUE a b c d
30S
20S
10S
010
N20
N30
N
120E 140E 160E 180 160W 140W 120W 100W 80W
Legend0 3.5
Relative YFT CPUE – from WCPO GLM.
Region scaling factors.
0
0.2
0.4
0.6
0.8
1
1.2
1 2 3 4 5 6
Region
Sc
alin
g f
acto
r
YFT area weighted CPUE indices - WCPO
0
0.5
1
1.5
2
2.5
3
3.5
4
1952 1955 1958 1961 1964 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003
Re
lati
ve
ab
un
da
nc
e
Region 6
Region 5
Region 4
Region 3
Region 2
Region 1
Spatial variation in biology
• Reproduction– Maturity– Sex ratio– Fecundity at length
• Growth
Reproductive parameters may vary in space (e.g. bigeye maturity)
• Model assumes same reproductive output at age for all females. – Affects ‘spawning biomass’ reference points– Affects other ref pts when steepness < 1
L50= 102.4 cm
L50=135 cm
L50=105 cm
Yellowfin growth (within WCPO)
0 5 10 15 20 25
050
100
150
200
Age class
Leng
th (
FL)
cm
2006 MFCLRegion 3 MFCLLehodey & Leroy 1999Yamanaka 1990
0 5 10 15 20 25
050
100
150
200
Age class
Leng
th (
FL)
cm
2006 MFCLRegion 3 MFCLID/PH tagsPNG tagsother tags
Region 3 growth estimated by using only region 3 data
Slower initial growth within region 3 (western equatorial) compared to overall growth estimated by MFCL WCPO model.
WCPO growth estimates strongly influenced by Region 1 size data.
Change growth; change fixed M-at-age (length-based).
Spatial variation in yft growth0
5010
015
020
0
1 5 10 15 20 25
Age (quarters)
Leng
th (
cm)
Base case2006 SARegion 3 growthInitial values
Improved fit to the length frequency data from region 3 small fish fisheries when apply region 3 growth to WCPO model.
Model pars/outputs – biomass (yellowfin)
Total and adult biomass.
R3 and R4 account for most of the biomass.
010
2030
4050
6070
Region 1
050
100
200
300 Region 2
050
100
150
200
Region 3
010
020
030
040
0
Region 4
010
2030
Region 5
010
2030
40
1950 1960 1970 1980 1990 2000
Region 6
020
040
060
080
0
1950 1960 1970 1980 1990 2000
WCPO
To
tal b
iom
ass
(1
00
0s
mt)
TotalAdult
Model pars/outputs
(bigeye) – recruitment
R3 and R4 account for most of the recruitment.
Increase in R3 recruitment from early 1990s.
Most recent recruitment approximates long-term average.
0.0
0.5
1.0
1.5
Region 1
02
46
810
12
Region 2
05
1015
Region 3
02
46
810
12
Region 4
0.0
0.2
0.4
0.6
0.8
1.0
Region 5
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1950 1960 1970 1980 1990 2000
Region 6
010
2030
4050
1950 1960 1970 1980 1990 2000
WCPO
Re
cru
itme
nt (
mill
ion
s o
f fis
h)
Model pars/outputs
(bigeye) – fishery impact
020
4060
8010
0
Region 1
050
100
200
300 Region 2
010
030
050
070
0 Region 3
010
020
030
040
0
Region 4
05
1015
2025
3035
Region 5
010
2030
1950 1960 1970 1980 1990 2000
Region 6
020
060
010
00
1950 1960 1970 1980 1990 2000
WCPO
Fished biomass
Unfished biomass
To
tal b
iom
ass
(1
00
0s
mt)
Estimating movement
• Biological issues– Food, oceanography, spawning
• Effects and consequences– Fishing pressure variation– Biomass trends
• Data– Tagging data directly inform movement– Length frequency and CPUE data also affect
estimates
Movement model
• Instantaneous movement at start of quarter• Between regions that share common
boundary• Parameter indicates prop. in A that move to B• Usually 4 seasonal movements each way
across each boundary pair• Other time effects not modelled• Age effects usually not estimable
Bigeye movement – tagging data(> 1000 n. miles)
Model pars/outputs – movement (bigeye)R1 R2
R3 R4
R5 R6
Quarter 1
R1 R2
R3 R4
R5 R6
Quarter 2
R1 R2
R3 R4
R5 R6
Quarter 3
R1 R2
R3 R4
R5 R6
Quarter 4
Max. 4% per quarter
Reg 1 Reg 2 Reg 3 Reg 4 Reg 5 Reg 6
0.0
0.2
0.4
0.6
0.8
1.0
Pro
po
rtio
n o
f bio
ma
ss b
y so
urc
e r
eg
ion
Movement issues
• Movements are difficult to estimate and data are not very informative– Implausible estimates seen – e.g. albacore
assessment– Model uses movement and recruitment to account
for lack of fit in age-based selectivity estimates. • May be better to use biologically reasonable
diffusion rates as prior distributions– Would also permit age-based movement estimates– Oceanography needs to be included
Conclusions
• Multiple spatial issues • Work in progress…