Influence of selectivity and size composition misfit on the scaling of population estimates and possible solutions: an example with north Pacific albacore

Steven L. H. Teo and Kevin R. PinerSouthwest Fisheries Science Center

CAPAM Selectivity Workshop11-14 March

What is the Problem?

• Highly migratory species move around a lot!

• Regional fisheries• Many HMS assessments do not

model movement due to lack of consistent tagging data

• Assume well-mixed stock and differences in selex used as proxies for movements

• But selex processes modeled as less variable in time and space than actual movements

What is the Problem?

• May cause important misfit to size compositions

• Influence recruitment and population scaling

• Similar to mis-specified time-varying selex

• In addition, mis-specified selex of one regional fishery can be strongly linked to selex of other fisheries catching

Albacore non-example

1965 1970 1975 1980 1985 1990 1995 2000 2005 20100

100000

200000

300000

400000

500000

600000

0.0010.010.025

Spaw

ning

Bio

mas

s (10

00 t)

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 110

10.2

10.4

10.6

10.8

11

11.2

Size composition weighting

LN_R

0

• 16 fleets, 8 fisheries dependent indices, and conditional age-at-length

• Estimate growth• Spawning biomass scaled

strongly with weighting of size composition data

• Size composition weighted to 0.01

• R0 vs Weighting plot• R0 profile plots

Albacore Piner PlotsSize composition weighting of 0.01

Albacore Piner PlotsSize composition weighting of 1.0

Plan BWestern Pacific Eastern Pacific

• Operating model based on Piner et al. 2009 SS model of Pacific bluefin tuna with 2 box annual movement (no tagging data)

• Somewhat funky movement model • All fish move back to western Pacific after every year

and no fish move to eastern Pacific after age 4• Created synthetic data set from the model (expected

data without obs errors)

Age 1-4

Age 1-5

7 fisheries4 longline indices1 age-0 index

2 fisheries (1 selex)3 purse seine indicesAge-based logistic selex

SS Estimation Model

• Estimated dynamics using SS model with no movement• Compare different methods of dealing with the selectivity and

misfit1. Estimate EPO selex (with and without time blocks)2. Fix selex with previous model run and don’t fit to lencomp

data (with and without time blocks)3. Downweight lencomp data4. Annual time-varying selex5. Calc average selex from annual time-varying selex and

don’t fit to data (with and without time blocks) 6. Kitakado the EPO selex (with and without time blocks)7. Kitakado the EPO selex with time-varying selex to help

with convergence

Eastern Pacific Size data

Selectivity of Eastern Pacific PS

0 50 100 150 200 250 3000.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Est Selex no tblk

Est Selex tblk52-93

Est Selex tblk94-07

Time-Varying Selectivity

Eastern Pacific Size Comp Fits

Estimate selectivity with no time block

Eastern Pacific Size Comp Fits

Recruitment

1950 1960 1970 1980 1990 2000 20100

5000

10000

15000

20000

25000

30000

35000

40000

True Modelest True ModelEst F8Fix F8Dwt F8TimeVary F8Avg F8 tblkKK F8 tblkKK F8 rndwlkKK F8Est F8 tblkFix F8 tblk

Recr

uitm

ent (

1000

fish

)

Spawning Biomass

19521955

19581961

19641967

19701973

19761979

19821985

19881991

19941997

20002003

20060

20000

40000

60000

80000

100000

120000

140000

160000

180000

True ModelEst SlxFix SlxTimeVary Slx

Spaw

ning

Bio

mas

s (1

000

t)

SSB with Timeblocks in Selectivity

19521956

19601964

19681972

19761980

19841988

19921996

20002004

0

20000

40000

60000

80000

100000

120000

140000

160000

180000

Operating ModelEst SlxFix SlxTimeVary SlxEst Slx tblkFix Slx tblk

Spaw

ning

Bio

mas

s (1

000

t)

SSB using Kitakado Method

19521955

19581961

19641967

19701973

19761979

19821985

19881991

19941997

20002003

20060

20000

40000

60000

80000

100000

120000

140000

160000

180000

True ModelEst SlxTimeVary SlxEst Slx tblkKK F8KK F8 tblk

Spaw

ning

Bio

mas

s (1

000

t)

Prelim Results & To Do List

• See Felipe’s talk• R0 profile (aka Piner Plots) useful in

understanding scaling influences in model• Do time-varying selex at least once to understand

how selex might be changing – try non-parametric selex

• Create a new synthetic data set with a simpler model (fewer fisheries & indices) with varying amounts of movement

• Include observation and other process errors• How does that affect management?

Averaging Blocks of Selectivity from Time-varying Selectivity

• Fit time-varying selex model

• Average the annual selex for wanted time blocks

• Use selex24.xls and solver to get selex parms

• Use parms in fixed selex for fishery and don’t fit to size comp data