Tunas in hot water: forecasts of population trends … · Tunas in hot water: forecasts of population trends for ... (sashimi) but low volume of ... Spatial dynamics of tuna populations

Tunas in hot water:forecasts of population trends for

two species of tuna under a scenario of Climate Change

Patrick Lehodey

Marine Ecosystems Modeling and Monitoring by SatellitesCLS, Satellite Oceanography Division, Toulouse, France

International Symposium, Gijón, Spain, May 19 - 23, 2008

Contact: [email protected]

Effects of Climate Change on the World's Oceans

PFRP SPC

Co-authors and collaborators

Tunas in hot water

Inna Senina, Beatriz Calmettes- CLS, Toulouse, France

John Sibert- Pelagic Fisheries Research Program, JIMAR, UH, Honolulu, USA

Laurent Bopp- Laboratoire des Sciences du Climat et de l'Environnement (LSCE), Gif-sur-Yvette, France

John Hampton- Oceanic Fisheries Programme, SPC, Noumea, New Caledonia

Raghu Murtugudde- ESSIC, University of Maryland, USA

GLOBEC-CLIOTOP Working Group 4

Tunas in hot waterEffects of Climate Change on the World's Oceans, Gijon, Spain, 19-23 May

PFRP SPC

Outline

• What is a tuna?

• End-to-end ecosystem modeling (SEAPODYM, briefly)

• Parameters optimization and hindcast simulations

• Confidence and uncertainty

• Forecast: and the winner is …

• So what?

Tunas in hot water

Tunas in hot water

Effects of Climate Change on the World's Oceans, Gijon, Spain, 19-23 May

PFRP SPC

Tunas

Skipjack (Katsuwonus pelamis) Bigeye (Thunnus obesus)vs

← Total annual catch →

← Fishing gears →

← Economical value →

← Status of stocks →

~2 million tonnes

Surface (purse seine)

Low (canneries) but high volume of catch

Approaching full exploitation

~0.5 million tonnes

Longline + surface

High (sashimi) but low volume of catch

Fully exploited to over-fished.

Fisheries

- 60

- 50

- 40

- 30

- 20

- 10

0

10

20

30

40

50

60- 20- 100102030405060708090100110120130140150160170180- 170- 160- 150- 140- 130- 120- 110- 100- 90- 80- 70- 60- 50- 40- 30- 20

- 20- 100102030405060708090100110120130140150160170180- 170- 160- 150- 140- 130- 120- 110- 100- 90- 80- 70- 60- 50- 40- 30- 20- 60

- 50

- 40

- 30

- 20

- 10

0

10

20

30

40

50

60

Blue = skipjack; Red = bigeye


PFRP SPC

Tunas


← Lifespan →

← Max size / weight →

← Age at maturity →

← Fecundity →

← Natural mortality →

← Food →

4 yrs +

75 cm / 20 kg

10-12 months

Very high

~0.4 per month

Micronekton

12 yrs (+)

180 cm / 225 kg

2.5 years

Very high

~0.1(-) per month

Micronekton

Biology


PFRP SPC

Tunas


← Thermal habitat →

← Oxygen tolerance →

← Vertical habitat →

← Spatial distribution →

Warm! 20 – 30 oC

Low! >3-4 ml l-1

0-200 m

Tropical

Extended! 10-30 oC

Good! >1.5 ml l-1

0-1000 m

Tropical to sub-temperate

Ecology


PFRP SPC

SEAPODYM: General scheme

TADR tuna model

Eqns for 0-3 month old juveniles:spawning, foraging, passive transport,

survival, mortality, cannibalism

Predictions

Six forage components:epi-pelagicMigrant and non-migrant meso-pelagicMigrant, non-migrant and highly-migrant bathypelagic

Pole-and-line:tropical and sub-tropical gears

Purse seine: associated and

unassociated fleets

Longline: Shallow, deep, …

3-layer data:Temperature,

currents,Oxygen

PredictionsTuna spatial distributions, catch andlength frequencies time series

Primary Production, Euphotic depth

Eqns for 1-X quarter old adults:recruitment, foraging, migrations,

ageing, natural and fishing mortality

Bio-physical environment Mid-trophic sub-model Fishing data

Optimization:Preliminary sensitivity analysis

Constructing cost function according to data distributionminimization, parameter estimation and errors.

Estimates of model parametersReference:

Senina, I., Sibert, J., Lehodey, P. (in press). Parameter estimation for basin-scale ecosystem-linked population models of large pelagic predators: Application to skipjack tuna.


PFRP SPC

Spatial Ecosystem And Populations Dynamic Model

SEAPODYM

day night

Epipelagic

Mesopelagic

Bathypelagic

Migrant bathypelagic

Migrant mesopelagic

Highly-migrant bathypelagic

acoustic backscatter transect kindly from R. Kloser, CSIRO

day

nightsunset, sunrise

Epipelagic layerT, U, V

surface1 2 3 4 5 6

Mesopelagic layerT, U, V

BathypelagicLayerT, U, V

Day length= f(Lat, date)PPE

En’

References:

Lehodey et al 1998. Fish. Oceanog.

Lehodey, P., 2001. Prog. Oceanog.

Lehodey, Murtugudde, Senina. (submitted). Scaling laws for modelling mid-trophic functional groups and their control on predator dynamics.


PFRP SPC

Prey: Micronekton

Epipelagic micronektonic group

6 functional groups in 3 vertical layers

SEAPODYM

Predators : tunas

Age-structured PopulationGrowth

mortalityby cohort

Feeding Habitat(accessibility to 6 functional

prey groups based on physical constraints of

species and age)

Spawning Habitat

Seasonal switch

Movement toward

feeding grounds

Spatial dynamics of tuna populations are based on habitats definition

Eulerian approach:

movement is modelled from

Advection-diffusion

equations

Movement toward

spawning grounds

mortality

Spawning success

Stock/R (if mature and season)References:Bertignac et al., 1998. Fish. Oceanog., 7, 326-334Lehodey et al., 2003. Fish. Oceanog., 12, 483-494Lehodey, Senina, Murtugudde, submitted.


PFRP SPC

Parameters optimization

Data

PFRP SPC


Fishing data:Fishing data:

•• SpatiallySpatially--disaggregated monthly catch data disaggregated monthly catch data SkipjackSkipjack = 4 purse= 4 purse--seine and 2 poleseine and 2 pole--andand--line fisheries line fisheries (original data on 1(original data on 1--deg resolution)deg resolution)BigeyeBigeye = idem skipjack + 15 long= idem skipjack + 15 long--line fisheries (original data line fisheries (original data on 5on 5--deg resolution)deg resolution)

•• Quarterly length frequencies data for each fisheryQuarterly length frequencies data for each fishery

BioBio--physical input dataset:physical input dataset:

• coupled physical-biogeochemical reanalysis from Earth Science System Interdisciplinary Center, University of Maryland, USA (R. Murtugudde)

Note: with fixed definition of vertical layers (0-100m, 100-400m, 400-1000m)

Time period for optimization experiment:Time period for optimization experiment:

• 1984-2004: 2 deg x 2 deg x month


Method

PFRP SPC


Model predictions: total catch by fleet and size (month x 2 deg)

The task of finding the optimal parameterization of the numerical model by fitting its prediction to observation consists in maximizing the likelihood function (or commonly, minimizing negative log-likelihood).

Numerical estimationQuasi-Newton minimization of negative log of joint likelihoodDerivatives of likelihood function with respect to all estimatedparameters computed by adjoint methods

Senina, I., Sibert, J., Lehodey, P. (in press). Parameter estimation for basin-scale ecosystem linked population models of large pelagic predators: Application to skipjack tuna. Prog. Oceanog.

θ Skipjack Bigeye

βp 0.296 ± 0.0018 0.073 ± 0.0005

Mmax 0.5* 0.25 ± 0.003βs -0.044 ± 0.0015 -0.097 ± 0.008Α 31* 80.6 ± 0.008σ0 3.5* 0.82 ± 0.012Τ0 30.5 ± 0.0047 26.2 ± 0.013α 0.1* 0.63 ± 0.02BHa 0.5* 0.0045 ± 6e-4σa 2.62 ± 0.0015 2.16 ± 0.004Τa 26* 13 ± 0.004Ο 3.86 ± 0.0009 0.46 ± 0.0006Dmax 0.4 ± 0.005 0.22 ± 0.002Vmax 1.3 ± 0.006 0.32 ± 0.002

Fishing parameters: catchabilities & selectivities


Estimated parameters

PFRP SPC


Nat

ural

m

orta

lity

Spaw

ning

habi

tat

Adu

lt ha

bita

t

Movem

ent

∧

± st.dev; * fixed

Hindcast simulations

Pacific skipjack

PFRP SPC


OptimizationHindcast

Time series of skipjack biomassTime series of skipjack biomass is lagged by 8 months backwardis lagged by 8 months backward

* MULTIFANMULTIFAN--CL: statistical stockCL: statistical stock--recruitment modelrecruitment model

Hindcast simulations

Pacific bigeye stock

Predicted distribution of adult biomass and observed (circles) longline catch

WCPO adult bigeye

0.20

0.25

0.30

0.35

0.40

0.45

1965 1970 1975 1980 1985 1990 1995 20000.15

0.20

0.25

0.30

0.35

0.40

0.45

Bio

mas

s (1

06 m

t)

EPO adult bigeye

0.28

0.30

0.32

0.34

0.36

0.38

1965 1970 1975 1980 1985 1990 1995 20000.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80Bi

omas

s (1

06 m

t)

optimization hindcast optimization hindcast

WCPO EPO Black curve: SEAPODYM

Red curve: Statistical Stock-recruitment model (MULTIFAN-CL)

PFRP SPC


Confidence / uncertainty

Environmental forcing data set

PFRP SPC


La Niña El Niño

Obs

erva

tion

Sate

llite

Pr

edic

tion

mod

el

Micronekton

PFRP SPC


ADCP data (Mc Phaden, Radenac et al.)

800

820

840

860

880

900

920

940

960

06-0

0

10-0

0

02-0

1

06-0

1

10-0

1

02-0

2

06-0

2

10-0

2

02-0

3

06-0

3

10-0

3

02-0

4

06-0

4

10-0

4

0,60,70,80,91,01,11,21,31,41,51,6Obs. ADCP night total

Model Pred. epipelagic night

800

820

840

860

880

900

920

940

960

06-0

0

10-0

0

02-0

1

06-0

1

10-0

1

02-0

2

06-0

2

10-0

2

02-0

3

06-0

3

10-0

3

02-0

4

06-0

4

10-0

4

0

20

40

60

80

100

120Obs. ADCP night totalModel Pred. PPtot

800

850

900

950

1000

1050

01/0

6/00

01/0

9/00

01/1

2/00

01/0

3/01

01/0

6/01

01/0

9/01

01/1

2/01

01/0

3/02

01/0

6/02

01/0

9/02

01/1

2/02

01/0

3/03

01/0

6/03

01/0

9/03

1,00

1,20

1,401,60

1,80

2,00

2,20

2,40

2,60Obs. ADCP night totalModel Pred. epipelagic night

800

850

900

950

1000

1050

01/0

6/00

01/0

9/00

01/1

2/00

01/0

3/01

01/0

6/01

01/0

9/01

01/1

2/01

01/0

3/02

01/0

6/02

01/0

9/02

01/1

2/02

01/0

3/03

01/0

6/03

01/0

9/03

020406080100120140160180200Obs. ADCP night total

Model Pred. PPtot

820

840

860

880

900

920

940

960

980

01/0

9/96

01/1

2/96

01/0

3/97

01/0

6/97

01/0

9/97

01/1

2/97

01/0

3/98

01/0

6/98

01/0

9/98

01/1

2/98

01/0

3/99

01/0

6/99

01/0

9/99

1,5

2,0

2,5

3,0

3,5

4,0Obs. ADCP night totalModel Pred. epipelagic night

820

840

860

880

900

920

940

960

980

01/0

9/96

01/1

2/96

01/0

3/97

01/0

6/97

01/0

9/97

01/1

2/97

01/0

3/98

01/0

6/98

01/0

9/98

01/1

2/98

01/0

3/99

01/0

6/99

01/0

9/99

0

50

100

150

200

250

300

Obs. ADCP night totalModel Pred. PPtot

165E

170W

140W


• Micronekton (tuna forage) predicted time series are in phase with ADCP data

03/98

03/00

• In all cases, fluctuations of ADCP time series are shifted by several months relatively to the primary production


Coherence in parameter estimates

PFRP SPC


θ Skipjack Bigeye

βp 0.296 ± 0.0018 0.073 ± 0.0005

Mmax 0.5* 0.25 ± 0.003βs -0.044 ± 0.0015 -0.097 ± 0.008Α 31* 80.6 ± 0.008

Nat

ural

m

orta

lity

0 20 40 60 80 100 120

0.0

0.1

0.2

0.3

0.4

Natural mortality

age (month)

mor

talit

y co

effic

ient

thre

shol

d ag

e of

sen

esce

nce

Habitat =0

Habitat =1

0.00.1

0.20.30.4

0.50.6

0.70.8

0 10 20 30 40Age (quarter)

Mor

talit

y

SKJ

BET

Natural mortality rates

In average, converging with independent studies

!! Values at early stages are correlated with spawning coefficient (lack of data)

θ Skipjack Bigeyeσ0 3.5* 0.82 ± 0.012Τ0 30.5 ± 0.0047 26.2 ± 0.013

σa 1.62 ± 0.0015 2.16 ± 0.004Τa 26* 13 ± 0.004Ο 3.86 ± 0.0009 0.46 ± 0.0006

Thermal habitat and oxygen tolerance

PFRP SPC


Spaw

ning

habi

tat

Adu

lt ha

bita

t

Bigeye

0.0

0.5

1.0

5 15 25oC

Inde

x

1st cohort 2yr4 yr last cohort

∧

0.0

0.5

1.0

0 1 2 3 4 5Oxygen (ml/l)

Inde

x

skipjack bigeyeOxygen:ok for bigeyeok for skipjack

Temperature:ok for bigeye

!! difficult for skipjack



Skipjack

0.0

0.5

1.0

5 15 25oC

Inde

x

1st cohort 2yr4 yr last cohort

Confidence / uncertaintyPFRP SPC



larvae survival leading to recruits

The number of larvae recruited in each cell of the grid at each time step is the product of a Beverton-Holt relationship coefficient linking the number of larvae to the density of mature fish and the spawning index IsIs: combines the effect of temperature and a measure of the trade-off ratio between food (~PP) and predators (micronekton) of larvae

General agreement with existing (limited) data

! Difficulties for skipjack in parameters optimization

Confidence / uncertaintyPFRP SPC



Movements Skipjack

Observed movement from skipack tagging data

(Lehodey et al. 1997, Nature)

Half-time…PFRP SPC


Are you confident with this model ?

Well… reasonably. OK. Tell me my

future.

Forecast: IPSL Simulation

Ocean Reanalysis

Atmospheric reanalysisheat flux, wind

Ocean General circulation Model

Ocean Biogeochemical (Nutrient-Phyto-Zoo-

Detritus) Model

Climate Simulationvs

Ocean General circulation

Model (OPA)

Ocean Biogeochemical (NPZD) Model

Sea-ice model (LIM)

Atmospheric model (LMDZ)

Realistic seasonal, inter-annual and decadal variability in time and space (e.g. reproduce El Niño / La Niña years, PDO

shifts, NAO, ...)

Land surface Model

(ORCHIDEE)

Atmospheric CO2 concentration

Coupler (OASIS)

Realistic seasonal variability

Realistic internal frequency at inter-annual and decadal scale, but not necessarily

coinciding with observed variability

3D-forcing

1D-forcing

PFRP SPC


PFRP SPC


Nino3Nino5

STN0

200

400

600

800

1000

Sep

-97

Sep

-98

Sep

-99

Sep

-00

Sep

-01

Sep

-02

Sep

-03

Sep

-04

Sep

-05

Sep

-06

Sep

-07

PP

PP-sat SubTrop N PP-IPSL SubTrop N

0

200

400

600

800

Sep

-97

Sep

-98

Sep

-99

Sep

-00

Sep

-01

Sep

-02

Sep

-03

Sep

-04

Sep

-05

Sep

-06

Sep

-07

PP

PP-sat Nino3 PP-IPSL Nino3

100

150

200

250

300

Sep

-97

Sep

-98

Sep

-99

Sep

-00

Sep

-01

Sep

-02

Sep

-03

Sep

-04

Sep

-05

Sep

-06

Sep

-07

PP

PP-sat Nino5 PP-IPSL Nino5

Ex: Primary production

Comparison with primary production deduced from Satellite data (Behrenfeld and Falkowski, 1997)

Ref.: Schneider, Bopp et al., 2007. Spatio-temporal variability of marine primary and export production in three global coupled climate carbon cycle models. Biogeosciences Discussion, 4: 1877–1921


PFRP SPC


Changes in forcing fields

Levitus seasonal climatology for oxygen IPSL predicted monthly oxygen outputs

ESSIC IPSL

Constant depth of layers Definition of vertical layers by euphotic depth


Revising parameters optimization

PFRP SPC


θ F ESSIC IPSL

βp 0.073 0.083

Mmax 0.25 0.3

βs -0.097 -0.11

Α 80.6 75.8

σ0 0.82 0.21Τ0 26.2 26.5

α 0.63 0.69

BHa 0.0045 0.009

σa 2.16 2.74

Τa 13 9.94Ο 0.46 1.05Dmax 0.22 0.11

Vmax 0.32 0.4

Nat

ural

m

orta

lity

Spaw

ning

habi

tat

Adu

lt ha

bita

t

Movem

ent

Fishing parameters

∧ 5.463.86 Ο

0.951.3 Vmax

0.270.4 c

26*26*Τa

2.12.62 σa

0.5*0.5*BHa

0.1*0.1*α

29.530.5 Τ0

3.5*3.5*σ0

31*31*Α

-0.1-0.044 βs

0.5*0.5*Mmax

0.1 0.296 βp

IPSLESSICFθ

Nat

ural

m

orta

lity

Spaw

ning

habi

tat

Adu

lt ha

bita

t

Movem

ent

Fishing parameters

∧

Bigeye Skipjack


The estimated biology reflects the environment

Bigeye has a long life span and extended habitat strongly influenced by seasonal variability, facilitating the revision of parameters with CC simulation

Skipjack has a short life-span, with equatorial core habitat influenced by ENSO variability: it is much more difficult to revise parameters optimisation with CC simulation

PFRP SPC


BIGEYE: Global average predicted distribution

Total biomass 1990-1999

For bigeye, model gives superficially realistic predictions in other oceans using parameters estimated from Pacific populations


PFRP SPC


SKIPJACK: Global average predicted distribution


For skipjack, Pacific parameterization gives less convincing “visual correlation” (based on catch distribution) in other Oceans



PFRP SPC


Larvae distribution


Trends

2000

2050

2099

Big

eye

1950

2000

2050

2099

1950

Skip

jack

Spawning sites (black) and larvae

distribution (between dotted

lines). From ICCAT report (1999)

PFRP SPC


Adults biomass distribution

Trends

2000

2050

2099

Big

eye

1950

2000

2050

2099

1950

Skip

jack

Bet SkjE Pac.

108-142

282-439

W Pac.

117-134

1136-1370

Ind. 115-135

422-489

Atl. 76-103

115-160

W Pac

E Pac

W Pac

E PacRange of annual catch (‘000 t) by ocean 2000-04

(FAO,2007)

Tropical Indian O. area ~ ½trop. W. Pac. area

1E+02

3E+02

5E+02

7E+02

9E+02

1950

1960

1970

1980

1990

2000

2010

2020

2030

2040

2050

2060

2070

2080

2090

2100

Indian Pacific Atlantic

PFRP SPC


And the winner is…

1E+05

3E+05

5E+05

7E+05

1950

1960

1970

1980

1990

2000

2010

2020

2030

2040

2050

2060

2070

2080

2090

2100

Indian Pacific AtlanticBiomass of larvae by Ocean

skip

jack

Big

eye

Biomass of adult by Ocean

1E+04

2E+04

3E+04

4E+04

5E+04

6E+04

1950

1960

1970

1980

1990

2000

2010

2020

2030

2040

2050

2060

2070

2080

2090

2100


1E+06

3E+06

5E+06

7E+06

1950

1960

1970

1980

1990

2000

2010

2020

2030

2040

2050

2060

2070

2080

2090

2100


… The Atlantic Ocean?

1E+02

3E+02

5E+02

7E+02

9E+02

1950

1960

1970

1980

1990

2000

2010

2020

2030

2040

2050

2060

2070

2080

2090

2100


PFRP SPC


By the way…

1E+05

3E+05

5E+05

7E+05

1950

1960

1970

1980

1990

2000

2010

2020

2030

2040

2050

2060

2070

2080

2090

2100

Indian Pacific AtlanticBiomass of larvae by Ocean

skip

jack

Big

eye

Biomass of adult by Ocean

1E+04

2E+04

3E+04

4E+04

5E+04

6E+04

1950

1960

1970

1980

1990

2000

2010

2020

2030

2040

2050

2060

2070

2080

2090

2100


1E+06

3E+06

5E+06

7E+06

1950

1960

1970

1980

1990

2000

2010

2020

2030

2040

2050

2060

2070

2080

2090

2100


Did you notice this?

Has climate warming already affected the pelagic ecosystem and tuna population dynamics?

PFRP SPC


Conclusions

Summary

UncertaintySingle reanalysis and some changes in forcing fields of climate change scenarioDifficulties in optimization of some parameters for skipjackVery few estimate of accuracy of forage componentsInteraction between species?

ConfidenceRigorous optimization leads to meaningful biological parameters

Many mechanisms in the model based on relative rather than absolute parameterization

Predictions in agreement with independent stock assessment estimates (MFCL) including for hindcast simulations

Limited and coherent changes in parameter estimates between reanalysis and CC simulations

Realistic distribution in three oceans with one single (Pacific) parameterization

Coherent changes in habitats and dynamic

So what?

Modeling• End-to-end ecosystem modelling is possible by avoiding excessive reductionism.

• Optimized parametrization is key issue to investigate the top-down & bottom-up control of marine ecosystems.

• SEAPODYM will find application in evaluation of tuna fishery management

Climate change simulation1. We got different (complex) responses by species and Ocean.

2. Despite general increase in larvae abundance, adult biomass decrease in some places but not in others: e.g., increase in the Atlantic, abrupt decrease of bigeye in the Indian and western Pacific… Which mechanisms? thresholds?

3. Tuna populations may have been already impacted by CC. Any evidence of that?

4. Large pelagic highly migratory animals are sentinelle species of the CC. We should monitor these animals carefully.

PFRP SPC


Conclusions

Next steps

PFRP SPC


Conclusions

Compare fishery and climate effects during historical period in Atlantic and Indian O.

Apply to yellowfin tuna

Optimization using other reanalysis and simulation products

Optimization with multi-species for testing species-interaction (need parallel code)

Verify forage components; implement MAAS project (CLIOTOP: next meeting in Bergen, Norway, June 24)

Forecast fishing effects into the future (needs model of fishery dynamics, topic for CLIOTOP WG5)

CLIOTOP web site: http://web.pml.ac.uk/globec/structure/regional/cliotop/cliotop.htm

PFRP SPCThe End

Thank you

Hmm…The Atlantic?

Why not, I’ ve never been there before!


Documents

Tunas in hot water: forecasts of population trends … · Tunas in hot water: forecasts of population trends for ... (sashimi) but low volume of ... Spatial dynamics of tuna populations