Paul Bouch and Dave Reid - PowerPoint Presentation

Slide1

Paul Bouch and Dave Reid

Stock Assessment Models for Data Poor Scenarios: A Comparison of Spict (Stochastic Surplus Production Model in Continuous Time) and CMSY (Catch-MSY) 

Two EBFM research projects -

FishKOSM and ProByFish

Multi species interactions

Mixed fisheries

MSY management targetsData poor stocksEcosystem/food web models

FishKOSM

-

Fisheries Knowledge for Optimal Sustainable Management

ProByFish

- Protecting bycaught species in mixed fisheries

Populating the Models

Food Webs

Establish Biomass and fishing pressures

Fisheries and Biomass reference points and ranges

Celtic Sea >200 species, 12 assessed, 13 monitoredNeed MSY proxies for data limited stocks

Data Limited Stocks

Time series may be limiting in terms of quality and quantity

No age or size data

One option - Surplus production models

Require time series of catch data and maybe an estimate of biomassUse abundance indices IF available?

Surplus Production Models - Schaefer

Surplus production is the production from a stock that is greater than the losses due to natural mortality

.

The surplus production can be fished without affecting the stock size, and there is a biomass that maximises this yield (MSY)

Bt

= current biomass

r = intrinsic growth rate

k = unexploited stock size

Ct = catch

CatchMSY

CatchMSY

Data Requirements

Uses catch time series

Schaefer model – symmetrical surplus production curve

Prior

Information Ranges

Requires estimate of species resilience to create a prior for r

Uses stock biomass trend.

Expert

opinion or

default

rules based on when in the

max and min catches in a time

series

Estimate

for unexploited stock size (k). Can be expert defined or

default

rules based on max catches and the species resilience

FISH and FISHERIES, 2013,

14

, 504-514

CatchMSY Method

Monte Carlo Method

Choose pairs of values within the predetermined ranges for r and k, and a starting biomass

An r-k pair is viable if the simulation does not lead to stock collapse and the biomass falls within the predicted biomass ranges.

From the viable pairs, the most likely r-k values are calculatedCalculate FMSY, fisheries reference points and current fishing mortality and stock biomass

BSM – Bayesian Schaefer Model

An

elaboration

to CMSY

As well as catch time series it also requires an abundance index (CPUE or Survey Index)Uses the prior ranges created with CMSY rulesBayesian state-space implementation of the Schaefer modelUses Schaefer model with the assumption that the surplus production curve is symmetrical

Stochastic Surplus Production Model in Continuous

Time (SPICT)

Utilised

in ICES 2018 assessments (Plaice 7f-g, Megrim 6b)

Catch data and biomass index (CPUE or Survey Index)Can use multiple abundance indicesShape of the production curve is not fixed although it can beCan accommodate the use of priors for r, k and q

So how do they compare?

Twenty data rich stocks from Celtic, Irish and North Seas

Using catch and abundance indices from 2017/2018 ICES assessments

Compare ICES

MSY reference points with those generated by CMSY, BSM and SPICTComparing visually and using calculating errorsARE (Absolute Relative Error) and MSE (Mean Squared Error)

Celtic Sea

Irish Sea

CMSY, BSM and SPICT Comparison - FMSY

Statistically, SPICT and CMSY perform similarly well

-

BSM

weakerCMSY shows poor agreement for FMSY > 0.3SPICT does better for FMSY > 0.3, but more variable than CMSY and wider confidence limits

CI Match

F

MSY

ARE mean

F

MSY

ARE median

F

MSY

MSE mean

F

MSY

MSE median

BSM

14

33.549

30.634

0.014

0.005

CMSY

14

23.787

17.372

0.014

0.002

SPICT

18

27.806

19.503

0.010

0.002

CMSY, BSM and SPICT Comparison – F/FMSY

CMSY

most accurately matches up with ICES assessment

SPICT often estimates F/FMSY to be lower than 0.05, which would be surprising for a commercially exploited stock.

CI Match

F/F

MSY

ARE mean

F/F

MSY

ARE median

F/F

MSY

MSE mean

F/F

MSY

MSE median

BSM

14

134.285

50.688

4.409

0.359

CMSY

16

32.709

31.801

0.537

0.145

SPICT

14

72.245

94.906

0.671

0.161

Outcome?

SPICT and CMSY perform equally well in estimating FMSY, but CMSY does far better at estimating current stock biomass and fishing pressure

BSM performs the worst of the three methodologies

CMSY which uses default rules (or expert judgement) to provide ranges performs more robustly

More information, in the form of abundance indices, does not seem to improve the assessmentWhy?

Effect of using abundance indices – Cod

West of Scotland

6 indices used in the ICES assessment

If SPICT uses them all: FMSY is 0.17

If we use each one individually, FMSY values range from 0.17 to 0.3

ICES

Fmsy

F/FMSY from multiple indices

F/FMSY ranges from 0.004 to 3.2

Abundance

indices

seem to add variability – especially for current fishing mortality and stock biomassMaybe explain why CMSY outperforms SPICT and BSM?

Conclusions

CMSY outperforms SPICT and BSM using most metrics

Abundance indices, even when from data rich stocks, appear to not help the assessment

For

these projects, we want to provide provisional reference points as a starting point for ecosystem modelsData will be very limited in quantity and qualityCatchMSY provides the most promising and robust assessment technique for

the projects

Thanks

The work was funded by the Irish Government’s Department of Agriculture, Food and the Marine’s

Competitive Research Funding Programmes for

FishKOSMAnd: The European Maritime and Fisheries Fund for

ProByFish

Surplus Production Models – Pella & Tomlinson

N defines the shape of the surplus production curveIf n=2 then it is symmetrical and identical to the top equation (Schaefer)

Modification to the Schaefer Model

CMSY, BSM and SPICT Comparison – Stock Biomass

The current biomass for each stock is best estimated for by CMSYFor several stocks, SPICT struggles to estimate the biomass with some very large results and huge confidence intervals

CI Match

Biomass

ARE mean

Biomass

ARE median

Biomass

MSE mean

Biomass

MSE median

BSM

5

68.114

74.957

5545899

1129.902

CMSY

6

84.982

58.025

1011252

625.597

SPICT

10

1051.044

67.406

43119634

9204.476

Building Ecosystem Models

Size based model using the GADGET frameworkEcopath and Ecosim

Calculate MSY within a food web contextTest harvest control rules using MSE