m.b. santos, g.j. pierce, i. riveiro , j.m. cabanas, r. gonzález-quirós & c. porteiro

20
Cycles and trends in the Iberian sardine (S. pilchardus) stock and catch series and their relationship with the environment M.B. Santos, G.J. Pierce, I. Riveiro, J.M.Cabanas, R. González-Quirós & C. Porteiro Sardine and climate

Upload: more

Post on 10-Jan-2016

30 views

Category:

Documents


0 download

DESCRIPTION

Sardine and climate. Cycles and trends in the Iberian sardine ( S. pilchardus ) stock and catch series and their relationship with the environment. M.B. Santos, G.J. Pierce, I. Riveiro , J.M. Cabanas, R. González-Quirós & C. Porteiro. VIIIb. VIIIc. BAY OF BISCAY. CANTABRIAN SEA. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

Cycles and trends in the Iberian sardine (S. pilchardus) stock and catch series and their relationship with the environment

M.B. Santos, G.J. Pierce, I. Riveiro, J.M.Cabanas, R. González-Quirós & C. Porteiro

Sardine and climate

Page 2: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

CANTABRIAN SEABAY OF BISCAY

GULF OF CADIZ

IXa

VIIIcVIIIb

VIIIc-EastVIIIc-West

IXa-North

IXa-Central North

IXa-Central South

IXa-South Portugal IXa-South

Cadiz

Sardine and climate

• Single stock, delimited by Spanish-French border and Strait of Gibraltar

• Supports important fishery in Spain and Portugal

• Sardine has rapid growth rate, short generation time, long spawning season; females produce high number of eggs

Iberian sardine

Page 3: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

Sardine and climate

Iberian sardine

Year

SS

B, '

00

0 to

nn

es

1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

02

00

40

06

00

Year

F, y

ea

r-1

1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

0.0

0.1

0.2

0.3

0.4

Year

Re

cru

itme

nt,

mill

ion

s

1978 1981 1984 1987 1990 1993 1996 1999 2002 2005 2008

05

00

01

00

00

15

00

02

00

00

spaly10

WG2009

WG2008

• High importance of recruitment in overall population dynamics

• Periods of consecutive low recruitments (+ high F) have led to “crises” in the fishery

SSB

F

R

• Single stock, delimited by Spanish-French border and Strait of Gibraltar

• Supports important fishery in Spain and Portugal

• Sardine has rapid growth rate, short generation time, long spawning season; females produce high number of eggs

Page 4: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

Sardine and climate

López-Jamar et al, 1995Negative correlation: R v

upwelling

Dickson et al, 1988Galician upwelling v

catches

Roy et al., 1995Wind strength v R

Previous studiesMany studies highlight apparent environmental relationships

Page 5: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

Sardine and climate

Stock, catch + environmental variables

R, SSB series: 1978-2009 (32 y)

Landings in each area 1948-2009 (62 y)

Sun spots

Upwelling , IPC indexes, SST, Wind, AT, CLO, etc

NAO, NAO winter, AMO, GULF, EA

and a series of global, regional and local environmental variables

Posiciones variables

VIIIc

IXa

VIIIb

40

41

42

43

44

10 5 0

10 5 0

40

41

42

43

44

España

Po

rtug

al

Golfo de Vizcaya

Page 6: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

6

Sardine and climate

Selection of explanatory variables

Dynamic Factor Analysis (a dimension-reduction technique) to try to identify

common trends in the EVs.

jan

feb

mar

apr

may

jun

jul

aug

sep

oct

nov

dec

Time

20 40 60

Fitt

ed v

alue

s

50

100

150

Monthly. Best model: 1 common trend. Average value used for analysis.

Exploration of collinearity in EVs + variable selection

Relationships between EVs: are they correlated?

Page 7: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

Time series = trend + cycle and/or AC + residual

Sardine and climate

Modelling approach

• Model each RV as function of EVs: select best EVs (GAM)

• Quantify AC. If AC persists in model residuals, use GAMM

• Decompose RVs, EVs into simple trends and residuals (GLM v time)

• Compare RV trends and residuals with EV trends and residuals (which components appear to be driving the relationship?)

Page 8: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

Trends in Recruitment + SSB

Sardine and climate

Exploration of time series

Possibility of linear + simple polynomial relationships with time - GLMs

0

5000000

10000000

15000000

20000000

25000000

Linear trend in R: decreasing from high values in the 80s to low values in the 90s and now

No trend in SSB

0

100000

200000

300000

400000

500000

600000

700000

800000

SSBR

Page 9: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

Cycles in Recruitment + SSB

Sardine and climate

Exploration of time series

0.25 cycle / y = 1 cycle every 4 years

Spectral analysis (note: we are also detrending the data because we want to concentrate on the cycles)

0.1 cycle / y = 1 cycle every 10 years (short time series, only 32 years)

SSBR

Page 10: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

(Partial) AC in Recruitment + SSB

Sardine and climate

Exploration of time series

Recruitment:PAC significant at time lag 1

Partial autocorrelograms

SSB:PAC very significant at time lag 1

Page 11: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

All response variables

Sardine and climate

Exploration of time series

Variable Cycle(years)

Trend PAC (lag, years)

R* 4 Linear ↓ 1

SSB 10 None - 1

L_Total* 20? Cubic ∩ 1

L_VIIIcW* ≥5 Cubic ↘ 1

L_IXaN 20? Quadratic ∩ 1, 4

L_IXaCN* 20? Cubic ↘ 1 (3)

L_IXa 20? Quadratic ∩ 1

L_VIIIc 20? Cubic ∩ 1, 2, 16

* Log transformed

Page 12: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

Explanatory variables

Sardine and climate

Exploration of time series

Variable Cycle(years)

Trend PAC (lag, years)

Sunspots 10 Linear ↓ 1, 2, 3

NAO 4 None - No AC

AMO 11 Cubic ↗ 1, 9

EA 4 Linear ↑ 1, 2

Upwelling Unclear None - 1, 2, 3

IPC 15 Linear ↓ No AC

W40350_W 3 Linear ↑ 1, 2, 3

SST40350_W Unclear None - 1

AT40350_W Unclear Quadratic ∩ 1

CLU40350_W 20 Cubic ↘ 1

Page 13: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

13

Sardine and climate

Results: L_IXaCN

Explanatory Variable %DE in singleEV model

P (final model) Df (final model)

SST42350_W W42350_Wavspots

10.1%21.2%16.1%

0.00100.03990.0023

2.21

2.1

• Landings are the most complex series to model as they will a priori contain stock, fishing and environment effects

• AC persists in model residuals• GAMM with AR1 variance structure is an “improvement”• …but AC persists and all environmental effects then become non-

significant

Final GAM for L_IXaCN (%DE=45.6%, AIC = 20.47)

GAM/GAMM : which EVs best explain Landings?

Page 14: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

14

Sardine and climate

Results: SSB

Explanatory Variable %DE in singleEV model

P (final model) Df (final model)

CLU40350_W AT40350_WAMO

26.0%30.0%33.9%

0.00100.03990.0023

12.12.8

No autocorrelation (confirmed by comparing AIC of best model with/without an AR1 variance structure)

Final GAM for SSB (%DE=68.3%, AIC = 819.18)

GAM/GAMM : which EVs best explain SSB?

Page 15: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

15

Sardine and climate

Results: Recruitment

Explanatory Variable %DE in singleEV model

P (final model) Df (final model)

W40350_W SST40350_WavspotsNAO

20.2%35.3%25.2%3.2%

0.02600.00020.02730.1451

111

2.84

No autocorrelation (confirmed by comparing AIC of best model with/without an AR1 variance structure)

Final GAM for LogR (%DE=64.6%, AIC = -17.44)

GAM/GAMM: which EVs best explain R?

Variable Trend Cycle

LogR Linear ↓ 4 y

W40350_W SST40350_WavspotsNAO

Linear ↑None -Linear ↓None -

3 yUnclear

10 y4 y

Page 16: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

16

1. GLM Extract trend and residuals (noise) from both LogR and Wind strength2. GAM LogR as a function of W, W trend and W noise

3. GAM LogR noise as function of W noise

LogR v trend + noise in winter wind strength (W40350_W)

Sardine and climate

Results: effect of wind strength

Log R v W(%DE=20.2P=0.0143)

Log R v W trend(%DE=39.2P=0.0001)

Log R v W noise(%DE=0.1P=0.870)

Log R noise v W noise(%DE=0.1P=0.837)

Page 17: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

17

1. GLM trend and residuals (noise) from LogR (no significant trend in SST40350_W)2. GAM LogR as function of SST

3. GAM LogR trend and LogR noise as function of SST

LogR v trend + noise in winter SST (SST40350_W)

Sardine and climate

Results: effect of SST

Log R v SST(%DE=35.3P=0.0115)

Log R noise v SST(%DE=16.3P=0.0219)

Log R trend v SST(%DE=13.0P=0.289)

Page 18: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

18

1. GLM Extract trend and residuals (noise) from both LogR and Sun2. GAM LogR as a function of Sun, Sun trend and Sun noise

3. GAM LogR noise as function of Sun noise

LogR v trend + noise in average number of sunspots (avspots)

Sardine and climate

Results: effect of sunspots

LogR v Sun(%DE=25.2P=0.0034)

LogR v Sun trend(%DE=39.2P=0.0001)

LogR v Sun noise(%DE=4.1P=0.267)

LogR noise v Sun noise(%DE=8.0P=0.116)

Page 19: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

19

• In short-lived fish, environmental relationships can be an important component of stock and fishery dynamics

• Iberian sardine R, SSB, catch series all show “environmental” effects:

- wind strength, SST, AT, NAO, AMO, sunspots,

• GAMM sometimes permits removal of AC (and may not be needed)

• Need to investigate the nature of the relationships to understand mechanisms; separating trends and noise is useful guide

• Short time series remain a limitation (e.g. to detect cycles)

• Relationships for R:

- wind, sunspots: effects due to opposite/similar linear trends

- SST effect relates more to short-term variation around trend

Sardine and climate

Conclusions

Page 20: M.B.  Santos,  G.J.  Pierce, I.  Riveiro ,  J.M. Cabanas,  R.  González-Quirós  & C. Porteiro

We would like to thank all our Portuguese and Spanish colleagues working on sardine, all the crew and scientists in the acoustic surveys, everyone who collected the landings data, and Alain Zuur (Highland Statistics) for statistical advice

Xunta de Galicia, Programa de Recursos Humanos

Plan Nacional de I + D + I, Proyecto CTM 2010- 16053 (LOng-Term variability OF small-PELagic fishes at the North Iberian shelf ecosystem)

Sardine and climate

Acknowledgements