Signature of ocean warming in global fisheries catch.pdf

7/28/2019 Signature of ocean warming in global fisheries catch.pdf

1/5

LETTERdoi:10.1038/nature12156

Signature of ocean warming in global fisheries catchWilliam W. L. Cheung 1, Reg Watson 2 & Daniel Pauly 3

Marine fishes and invertebrates respond to oceanwarming throughdistribution shifts, generally to higher latitudes and deeper waters.Consequently, fisheries should be affected by tropicalization of catch 14 (increasing dominance of warm-water species). However,a signature of such climate-change effects on global fisheries catchhasso farnot been detected.Herewe reportsuchan index,the meantemperature of thecatch (MTC), that is calculated from the averageinferred temperature preference of exploited species weighted by their annual catch. Our results show that, after accounting for theeffects of fishing and large-scale oceanographic variability, globalMTC increased at a rate of 0.19 degrees Celsius per decade between1970 and 2006, and non-tropical MTC increased at a rate of 0.23

degreesCelsiusper decade.In tropicalareas,MTCincreased initially because of the reduction in the proportion of subtropical speciescatches, but subsequently stabilized as scope for further tropicaliza-tion of communities became limited. Changes in MTC in 52 largemarineecosystems, covering themajority of the worlds coastal andshelf areas, are significantly and positively related to regionalchanges in sea surface temperature 5 . This study shows that oceanwarming has already affected global fisheries in the past four dec-ades, highlighting the immediate need to develop adaptation plansto minimize the effect of such warming on the economy and foodsecurity of coastal communities, particularly in tropical regions 6,7 .

Assessing the effects of climate change on marine fisheries is oneof the challenges for sustainable management of marine ecosystems.Marine biota respond to ocean warming through changes in distribu-

tions andabundance3,4,8

, phenology 9

andbodysize10

, leadingtoalterationofcommunitystructure 4,8 andtrophic interactions 11, andultimatelyaffect-ing fisheries7. Studies assessing the potential effects of climate changeon globalfisheries under scenarios of greenhousegasemissions predicta large-scale redistribution of maximum fisheries catch potential 12,13

and increased vulnerability of manycoastal fisheries to climatechange,particularly in the tropics 6. Climate change effects on some fisherieshave been detected14,15 . For example, the rapid increase in catches of red mullet ( Mullus barbatus), a warm-water species, around the UK issuggested to be relatedto ocean warming 15 . However, a signatureof theeffect of climate change on global fisheries has so far not been demon-strated. Because marine fisheries contribute to the economy and foodsecurity of many coastal communities, fisheries responses to climatechange need to be better understood to inform the development of effective management and adaptation policies7.

Shifts in distributions of exploited stocks areexpected to affect theiravailabilityto fisheries. Spatialdistributionsof marine fishesand inver-

tebrates are strongly dependent on the relationship between physio-logical optima and limits under different temperatures, oxygen levelsand other biotic and abiotic conditions 16,17 . Organisms living in tem-peratures outside their thermal optima experience reduced aerobicscope, negatively affecting their growth and reproduction, and ulti-mately reducing their abundance 18 . Incontrast, some species may findthat areas previously unsuitable for their survival have become morefavourable. Consequently, the distribution margins and centroids of many marine fishes and invertebrates shift following changes in oceanconditions 16 , with a tropicalization of species compositions occurring as the ocean warms1. Concomitantly, species with limited dispersalpotential or a narrow range of temperature tolerance, such as in semi-enclosed seas and tropics, will decrease in overall abundance18,19 .

Werepresent the temperaturepreferenceof speciesin fisheries catch

by an index called mean temperature of the catch (MTC) (Fig. 1). Thisglobal index builds on previous use of temperature preferences toevaluate potential effects of climate change on fisheries locally 14,20 . Wepropose thatoceanwarmingleadsto increasedcatches ofwarmer-waterspecies and decreased catches of colder-water species, resulting in an

1 Changing Ocean Research Unit, Fisheries Centre, The University of British Columbia, Vancouver, British ColumbiaV6T 1Z4, Canada. 2 Institute for Marine and Antarctic Studies, University of Tasmania,Taroona, Tasmania 7001, Australia. 3 Sea Around Us Project, The University of British Columbia, Vancouver, British Columbia V6T 1Z4, Canada.

T r o p

i c a

l

T e m p e r a

t e

S u

b t r o p

i c a

l

Temperature ( C)

M T C ( C )

M T C ( C )

M T C ( C )

MTC

MTC MTC MTC

Spp. 1Spp. 2Spp. 3

Spp. 4

Multi-decadal ocean warming (SST)

0 5 10 15 20 25 30 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Year

Oceantemperature

Latitude Decade 1 Decade 2 Decade 3

0 5 10 15 20 25 30

0 5 10 15 20 25 30

5 10 15 20 25 300

5 10 15 20 25 300

5 10 15 20 25 300

10 15 20 25 300 5

Figure 1 | Changes in catch speciescomposition in relation to oceanwarming and the resulting changesin MTC. Species distributions arerelated to ocean temperature(coloured bars) and temperaturepreferences of the exploited species(grey curves). Increase and decreasein abundance due to ocean warming are indicated by green curves and thereduction in area under the grey curves, respectively. The verticalblack and red arrows represent MTCin the initial and subsequent decades,respectively. D MTC represents thedifference in MTC relative to theinitial decade. Species localextinction and invasion because of warming are indicated by red andgreen dotted curves, respectively.Theexpected changes in MTC over timeare shown on the right.

1 6 M AY 2 0 1 3 | V O L 4 9 7 | N AT U R E | 3 6 5Macmillan Publishers Limited. All rights reserved201 3
http://www.nature.com/doifinder/10.1038/nature12156http://www.nature.com/doifinder/10.1038/nature12156


2/5

increase in MTC. We further propose that the general relationshipbetween MTC and sea surface temperature (SST) is modified in thetropics, where catch compositions consist of both tropical and sub-tropical species. Warming would cause the initial decrease in the pro-portionof catchesof subtropicalspecies, resultingin increases in MTCbecause of the poleward shift in the equatorial range boundary of subtropical species. Such initial increases in MTC would then stabilizeas catches become dominated by tropical species, but the scope forfurther tropicalization is limited 16 . Further warming, to levels that

exceed the temperature preference and tolerance of tropical species,is then expectedto reducetheirabundance 12 without changesin MTC.

We calculated MTC as the average of the inferred temperaturepreference of exploited species (990 species in total) weighted by theirannual catch between 1970 and 2006, for 52 large marine ecosystems(LMEs) that account for most of the worlds fisheries (Methods). Toattribute changes in MTC to ocean warming, we used a generalizedadditive mixed model (GAMM) with an ensemble of different combi-nations of factors to account for the potential effects of large-scale

*

Rate of change in SST ( C yr 1 )

0.04 0.02 0.02 0.04 0.06 0.080

1.0

0.5

2.0

a

b

c

2.0

0.5

1.0

1.0

0.0

1.5

1.0

0.0

1.0

13

1

13

13

02

2

0.0

1.01.5

0.01.5

0.40.0

0.4

0.8

0.0

0.6

0.4

0.20.6

0.4

0.20.6

0.4

0.0

0.3

1.0

0.5

1.5

1.00.00.6

0.40.0

0.4

0.2

0.4

0.0

0.2

0.20.0

0.3

0.3

0.0

2

10

2

20

2

20

13

1

0.2

0.2

0.50.5

0.0

1.0

1.5

0.5

3

1

1

12

1

1

131

13 10

2

2

0

224

02

0.2

0.2

0.2

0.00.6

0.0

0.6

0.6

0.4

0.0

1.5

1.5

0.0

10

2

1

1.2

0.2

0.4

1.0

0.5

1.0

0.0

1

21

1

21

0.2

0.0

0.0

24

20

24

20

0.0

0.2

0.2

0.0

0.3

Figure 2 | Changes in MTC andSST of 52 LMEs between 1970 and2006. Changeovertime( x axis;yearsbetween 1970 and 2006) in MTCanomalies relative to the mean of thetime series ( y axis; u C). Rate of SSTchange is shown by the colour scale.a , Eurasia; b, the Americas; c, Asiaand Oceania. Red lines represent

tropical LMEs. To highlight thenonlinear trends of MTC anomaliesin this figure, each MTC time serieswas fitted with a spline smoothing function using GAMM: grey line,mean; shaded area, 95% confidenceinterval.

RESEARCH LETTER

3 6 6 | N AT U R E | V O L 4 9 7 | 1 6 M AY 2 0 1 3Macmillan Publishers Limited. All rights reserved201 3


3/5

oceanographic variability and fishing effort, both of which are knownto affect fisheries strongly 21,22 . In addition, we repeated the analysisusing a subset of 698 species with their temperature preference pre-dicted from the above and an alternative approach to species distri-bution modelling (Supplementary Information).

Overall, theMTC in the52 LMEs from theensembleGAMManalysisincreased at an average rate of 0.19u C per decade between 1970 and2006 (Fig. 2), andthe MTCin non-tropical LMEs increasedat a rate of 0.23 u C per decade. Specifically, MTC increased consistently in LMEsin the northeast Pacific Ocean (0.48 u C per decade) and the northeastAtlantic Ocean (0.49 u C per decade), where SST increased by 0.20 and0.26 u C per decade, respectively. For each LME, and on the basis of theresults fromthe ensemble GAMMs, we ranked the relative importanceof SST, fishing effort and large-scale oceanographic indices as factorsaccounting for the changes in MTC over time. The results suggestedthat there was no significant difference between these factors, in termsofthefrequency distribution ofLMEs withdifferent rankings(KruskalWallis rank-sumtest, P . 0.05; Supplementary Fig.1). However, therewere regionaldifferences in their importance(SupplementaryTable1).

MTC changes in the52 LMEs were significantly related to therateof SST changesbetween1970 and 2006 (P , 0.05; Fig. 3 and Supplemen-tary Table 2). The relationships remained significant when the bestmodels (lowest Akaike information criteria) were selected instead of the model ensemble (Supplementary Table 2) and when the twenty-fifthand seventy-fifth percentilesof the temperaturepreferenceprofiles

ofthe species, insteadofthe median values, wereusedin calculatingtheMTC(SupplementaryFig. 2).Using thealternativespecies distributionmodelling approach to calculate species temperature preference hasno significant effect on the relationship between changes in MTC andSST (P . 0.1; Methods and Supplementary Information).

TropicalLMEs showed, overall, an asymptoticpattern ofMTCchange,with a reduction in the proportion of subtropical species in catches(Fig. 3). Average MTC from 14 tropical LMEs increased rapidly between 1970 and 1980 by around 0.6u C, and subsequently stabilizedat around 26 u C (Fig. 3). Moreover, the average temperature tolerancerange, calculated from thedifferencein temperature betweenthe twenty-fifth and seventy-fifth percentilesof the temperaturepreferenceprofileof each species (Supplementary Information), decreased significantly

during the period. Subtropical species had wider temperature tole-rances than did tropical species17 (Supplementary Fig. 3), providing further evidence to support the hypothesis that fisheries catches in thetropics are becoming inimicalto subtropicalspecies. Moreover,averageSST change in the tropical LMEs increased consistently, at a rate of 0.14 u C per decade.

A number of factors indicate that MTC is a valid proxy to examinechanges in composition of catches in a region in relation to the tem-perature preference of the exploited animals. First, there was no dif-ference (analysis of covariance (ANCOVA), P . 0.1; Supplementary Table 4) between using catch data and scientific bottom trawl survey data in the calculated rate of change in MTC in North Sea (Fig. 4 andMethods). Because the scientific survey data set can be viewed as areliable indicator of relative abundance of animals in the ocean, ourresults therefore support the use of catch data to detect climate changesignature in fisheries. However, because the availability of survey-based data was limited to a few well-studied regions23 , we were unableto use such data to analyse global fisheries. The relationship betweenthe thermal preference of the species and the environmental temper-ature also corroborates the results of regional-scalestudies using survey data14,24 .

Second, fishing efforts in many LMEs have been increasing con-tinuously since the1970s. This coincided with increases in SST, result-ing in strong correlation between changes in SST and fishing effort insome LMEs. However, there is no evidence that fishing systematically alters MTC. Specifically, significant but weak relationships betweenmaximum body size (positively related to vulnerability to fishing, ingeneral) and the temperature preference of exploited species is foundin only 19 LMEs, with the majority (13) of them showing a positiverelationship,suggesting that the increasing MTCtrend wasnota result

0.15

0.10

0.05

0.00

0.05

0.10

0.15

0.01 0.01 0.03 0.05

R a

t e o

f c

h a n g

e i n M T C ( C y r

1 )

Rate of change in SST ( C yr 1 )

1:1 line

0.6

0.4

0.2

0.0

0.2

0.4

0.6

25.0

25.2

25.4

25.6

25.8

26.0

26.2

26.4b

a

1970 1980 1990 2000

M T C ( C )

S S T an

om

al y

( C )

Year

MTC

SST

Figure 3 | Relationship between rates of change in MTC and SST between1970and2006 in52 LMEs. a , Rate of change of MTC was calculated from theensemble GAMM in each LME(slope of linearregression, P , 0.005,R2 5 0.19

(coefficient of determination)). The black line shows the mean and the grey lines delineate the 95% confidence interval.b, Changes in MTC (red) and SSTanomalies (grey) in tropical LMEs. The dashed lines are fitted with asymptoticand linear models for the MTC and SST anomalies, respectively.

1.5

1.0

0.5

0.0

0.5

1.0

1.5

1970 1980 1990 2000Year

From survey data

From catch

M T C a n o m a

l y ( C )

Figure 4 | Comparison of MTC calculated from fisheries catch data andrelative abundance calculated from scientific survey data. MTC is expressedas an anomaly relative to the mean of the time series, and relative abundance isexpressedrelative tothe average between 1970 and2006. Thetypeof data used hadno effect on the rate of change in MTC (ANCOVA, P . 0.1).

LETTER RESEARCH

1 6 M AY 2 0 1 3 | V O L 4 9 7 | N AT U R E | 3 6 7Macmillan Publishers Limited. All rights reserved201 3


4/5

of the depletion of large fish by fishing that was reported by many fisheries25 (Supplementary Information).

Third, our findings are not an artefact of the quality of fisheries andenvironmentaldata, whichvarybetween regions. In four of fiveareas inwhich misreporting of the level of catch was previously estimated, theestimated rates of changein MTC were not significantly different fromthosecalculated from data sets withoutcorrectionfor misreporting. Forthe only area that showed a significant difference in rate of MTC

change, MTCs calculated from both data sets show significant increaseandthequalitative trend thusremains robust. Serious over-reportingof fisheriescatches byChinasince themid 1980shave been documented 26 .Recalculating MTC in the South China Sea, the East China Sea and theYellow Sea using catch data from 19701985 increased the goodnessof fit of its significant relationship with SST. The use of different SSTclimatology to calculatespecies temperaturepreference, moreover,hadno significant effect on the findings (Supplementary Table 3).

Although fisheries catch statistics are usually not reported at stock level, an analysis using simulated species and stock distributions, tem-perature preferencesand,subsequently, MTCchanges showedthat theexistence of stock structure andstock-specific temperaturepreferencesdid not significantly bias the rate of change of MTC calculated fromaggregated species-level data (P . 0.1; Supplementary Information).

Potential phenotypicand evolutionary responsesof species to warming,if occurring, would reduce the rate of change in MTC. The accuracy of theSST data setvaried between regions as a result of differentdensitiesofmeasurements.Given these uncertainties, thefactthat ouranalysis isable to detect a signature ofclimatechangein globalfisheries highlightsthe robustness of the underlying relationship.

Overall, ourresults suggest that changein thecomposition of marinefisheriescatch is significantlyrelated to temperaturechange in theocean,with an increasing dominance of catches of warmer waters species athigher latitudes anda decrease in theproportion of catches of subtropicalspecies in the tropics. Such changes in catch composition have directimplicationsforcoastal fishing communities, particularly thosein tropicaldeveloping countries, which tend to be socioeconomically vulnerable totheeffects of climate change6. Continuedwarmingin the tropics to a level

that exceeds the thermal tolerance of tropical species may largely reducecatch potential in this region 12. This highlights the need to prioritizeresources to develop adaptation plans immediately to minimize impactson the economy and food security of tropical coastal communities.

METHODS SUMMARYThemappedfisheries landingsdataweresourcedfromthe SeaAroundUs project 27 ,which utilized a rangeof sourcesincludingthe Food andAgricultureOrganizations(FAO)fisheriesdatabase supplementedby regional datasets (SupplementaryInfor-mation). We inferred the thermal preference of each species on the basis of itsmodelled distribution 28,29 (Supplementary Table 5 and Supplementary Information).The MTC was computed fromthe average inferred temperature preference of 990species of exploited fishes and invertebrates weighted by their annual catch:

MTCyr ~P ni T iC i, yrP n

iC i, yr

1

Here C i ,yr is catch of species i in a specific region in year yr, T i is the mediantemperature preference of species i and n is the total number of species. Similarly,MTC was calculated using survey data in the North Sea from equation (1) with C i ,yrreplaced by the relative abundance for species i in year yr. Relative abundance datawereobtainedfromtheICESInternationalBottomTrawlSurvey.A totalof 55 speciescommon to both the catch database and the survey data set were included in thisanalysis.We tested theeffectof the useof catch or surveydata in estimating changesin MTC over time using ANCOVA, accounting for temporal autocorrelation.

Weusedthe GAMM(R packageMuMIn)witha full model,MTC 5 s(Effort) 1s(OI) 1 Year,where sisa splinesmoothingfunction (Supplementary Information)with a temporalautocorrelation term (R functioncorAR1), allowingfor a lagof upto three years in the response of MTC to these two variables and the time effectexpressed by Year. To estimate the average rate of change in MTC from thecoefficients of this time effect, we did not include a spline smoothing functionfor Year. Fishing effort data were sourced from the Sea Around Us project30

(Supplementary Information).

Full Methods and any associated references are available in the online version ofthe paper .

Received 13 October 2012; accepted 5 April 2013.

1. Cheung, W. W. L. et al. Climate change induced tropicalization of marinecommunities in Western Australia. Mar. Freshw. Res. 63, 415427 (2012).

2. Wernberg, T. etal. An extreme climaticeventalters marine ecosystem structure ina global biodiversity hotspot. Nature Clim. Change 3, 7882 (2013).

3. Perry, A. L., Low, P. J., Ellis, J. R. & Reynolds, J. D. Climate change and distributionshifts in marine fishes. Science 308, 19121915 (2005).

4. Dulvy,N.K.etal. Climatechangeand deepening of theNorthSea fishassemblage:a biotic indicator of warming seas. J. Appl. Ecol.45, 10291039 (2008).

5. Belkin, I. M. Rapid warming of large marine ecosystems. Prog. Oceanogr. 81,207213 (2009).

6. Allison, E.H.et al. Vulnerability of national economies to the impacts of climatechange onfisheries. Fish Fish. 10, 173196 (2009).

7. Sumaila,U.R.,Cheung,W.W. L.,Lam,V. W.Y.,Pauly,D.& Herrick,S.Climatechangeimpacts on the biophysics and economics of world fisheries. NatureClim. Change1, 449456 (2011).

8. Simpson,S. D. etal. Continental shelf-wide responseof a fishassemblageto rapidwarming of the sea. Curr. Biol.21, 15651570 (2011).

9. Edwards, M. & Richardson, A. J. Impact of climate change on marine pelagicphenology and trophic mismatch. Nature 430, 881884 (2004).

10. Cheung, W. W. L. et al. Shrinking of fishes exacerbates impacts of global oceanchanges on marine ecosystems. Nature Clim. Change 3, 254258 (2013).

11. Harley, C. D. G. Climate change, keystone predation,and biodiversityloss. Science334, 11241127 (2011).

12. Cheung, W. W. L. et al. Large-scale redistribution of maximum fisheries catch

potential in the global ocean under climate change. Glob. Change Biol. 16, 2435(2010).13. Blanchard, J. et al. Potential consequences of climate change for primary

production and fish production in large marine ecosystems. Phil. Trans. R. Soc. B367, 29792989 (2012).

14. Pinsky,M. & Fogarty,M. Lagged social-ecological responses to climate and rangeshifts in fisheries. Clim. Change 115, 883891 (2012).

15. Cheung, W. W. L., Pinnegar, J., Merino, G., Jones, M. C. & Barange,M. Review ofclimate change impacts on marine fisheries in theUK and Ireland. Aquat.Conserv.Mar. Freshwat. Ecosyst. 22, 368388 (2012).

16. Sunday,J. M., Bates, A. E. & Dulvy, N. K. Global analysis of thermal tolerance andlatitude in ectotherms. Proc. R. Soc. Lond. B278, 18231830 (2011).

17. Po rtner, H. O. & Farrell, A. P. Physiology and climate change. Science 322,690692 (2008).

18. Po rtner,H. O. & Knust,R. Climatechangeaffects marinefishesthrough theoxygenlimitation of thermal tolerance. Science 315, 9597 (2007).

19. Ben Rais Lasram, F. et al. The Mediterranean Sea as a cul-de-sac for endemicfishes facing climate change. Glob. Change Biol. 16, 32333245 (2010).

20. Collie,J. S.,Wood,A. D. & Jeffries, H. P. Long-term shifts in thespecies composition

of a coastal fish community. Can. J. Fish. Aquat. Sci.65, 13521365 (2008).21. McGowan, J. A., Cayan, D. R. & Dorman, L. M. Climate-ocean variability andecosystem response in the northeast Pacific. Science 281, 210217 (1998).

22. Pauly,D. etal. Towardssustainability inworldfisheries. Nature 418, 689695(2002).23. Costello, C. etal. Statusand solutions forthe worldsunassessedfisheries. Science ,

(2012).24. Howell, P. & Auster, P. J. Phaseshift in an estuarine finfishcommunityassociated

with warming temperatures. Mar. Coast. Fish. 4, 481495 (2012).25. Cheung, W. W. L., Watson, R., Morato, T., Pitcher, T. J. & Pauly, D. Intrinsic

vulnerability in the global fish catch. Mar. Ecol. Prog. Ser. 333, 112 (2007).26. Watson,R. & Pauly,D. Systematicdistortions inworldfisheriescatchtrends. Nature

414, 534536 (2001).27. Watson, R., Kitchingman, A., Gelchu,A. & Pauly,D. Mapping global fisheries:

sharpening our focus. Fish Fish. 5, 168177 (2004).28. Jones,M., Dye,S.,Pinnegar,J., Warren,R. & Cheung,W. W. L. Modelling commercial

fish distributions: prediction and assessment using different approaches. Ecol.Modell. 225, 133145 (2012).

29. Rockmann, C., Dickey-Collas, M., Payne,M. R. & van Hal, R. Realized habitats ofearly-stage North Sea herring: looking for signals of environmental change. ICES J. Mar. Sci.68, 537546 (2011).

30. Watson, R. A. et al. Global marine yield halved as fishing intensity redoubles. FishFish. (2012).

Supplementary Information is available in the online version of the paper .

Acknowledgements W.W.L.C. acknowledges funding support from the NationalGeographic Society and the Natural Sciences and Engineering Research Council ofCanada. R.W. and D.P. were supported by the Pew CharitableTrust through the SeaAround Us project. We are grateful to S. Pauly and D. Palomares for reviewing themanuscript and providing the Aquamap distributions from FishBase, respectively.

Author Contributions W.W.L.C. and D.P. designed the study. W.W.L.C. conducted theanalysis. R.W.and D.P.providedthe fisheriescatchand effortdata fromthe SeaAroundUs project. All authors contributed to the writing of the manuscript.

Author Information Reprints and permissions information is available atwww.nature.com/reprints . The authors declare no competing financial interests.Readersare welcome to commenton the online version of thepaper . Correspondenceand requests for materials should be addressed to W.W.L.C.([email protected]) .

RESEARCH LETTER

3 6 8 | N AT U R E | V O L 4 9 7 | 1 6 M AY 2 0 1 3Macmillan Publishers Limited. All rights reserved201 3
http://www.nature.com/doifinder/10.1038/nature12156http://www.nature.com/doifinder/10.1038/nature12156http://www.nature.com/doifinder/10.1038/nature12156http://www.nature.com/reprintshttp://www.nature.com/doifinder/10.1038/nature12156mailto:[email protected]:[email protected]://www.nature.com/doifinder/10.1038/nature12156http://www.nature.com/reprintshttp://www.nature.com/doifinder/10.1038/nature12156http://www.nature.com/doifinder/10.1038/nature12156http://www.nature.com/doifinder/10.1038/nature12156


5/5

METHODSGlobal catch and effort data. We use the terms fisheries catch and fisherieslandings interchangeably, but strictly these both refer here to landings; the weightof marine fishes and invertebrates that were caught and retained. The mappedfisheries landings and effort data were sourced from the Sea Around Us project27

(Supplementary Information). Landings data were not used to estimate temper-ature preference ofthe species, andtemperature data were notusedto infer speciesdistributions. Effort data were standardized and collated on the basis of fishing boat engine power (watts) and fishing days30 (Supplementary Information).Inferring species thermal preference. We inferred the thermal preference of each species from its modelled distribution (Supplementary Table 5). First, wemodelled the present (19702000) distribution of each species using an algorithmdevelopedby theSea AroundUs project anddocumented inref. 28.The algorithmestimated the relative abundance (on a 309 3 309 latitudelongitude grid) of aspecies in each spatial cell. Input parameters for each species considered in themodel included the species maximum and minimum depth limits, northern andsouthern latitudinal range limits, an index of association to major habitat types(seamounts, estuaries, inshore, offshore, continental shelf, continental slope andthe abyssal) and known occurrence boundaries. For pelagic species, seasonal(summer and winter) distributions wereconsidered. The parametervalues of eachspecies were derived from data in online databases, mainly FishBase (http://www.fishbase.org ) and SeaLifeBase (http://www.sealifebase.org ). Catch and tem-perature data were not used to model the current species distributions. Eachmodelled species distribution was normalized and overlaid over the SST climato-logy from the Hadley Centre SST data set for 1970200029 . The predicted speciesdistributionis considered to be representative of thedistributionof relativeabund-ance. The temperature preference profile at SST bini ( pi) of each species wascalculated from the total relative abundance, K i , and range area, Ai , at SST bini(Supplementary Information): pi 5 (K i/ Ai)/P i (K i/ Ai).

The median and the twenty-fifth and seventy-fifth percentiles of the temper-ature preference were calculated. The modelled species distribution and temper-ature preference represents an average of the species life stages. The analysis wasrepeated with alternative estimate climatologies for 19511960 and 19912000.The choice of climatology has no significant effect (ANCOVA, P . 0.1) on therelationship between rate of change in MTC and rate of change in SST(Supplementary Table 3). The analysis is also repeated using an alternativeapproach to species distribution modelling (Supplementary Information).Calculation of MTC. The MTC was computed from the average inferred tem-perature preference of 990 species of exploited fishes and invertebrates weightedby their annual catch:

MTCyr ~P ni T iC i, yrP ni C i, yr

1

Here C i ,yr is catch of species i in a specific region in year yr, T i is the mediantemperature preference of species i and n is the total number of species.

In the North Sea LME, MTC was calculated from equation (1) with C i ,yrreplaced by the relative abundance for species i in year yr. Relative abundancedata were obtained from the ICES International Bottom Trawl Survey. A total of

55 species that coexist in both the catch database and the survey data set wereincluded in this analysis. We tested whether the changes in MTC calculated fromcatchdata and scientific survey datawere different, using ANCOVA and account-ing for temporal autocorrelation.Large-scale oceanographic indices. Indices representing large-scale oceano-graphic conditions for each of the six ocean basins were included in this study (Supplementary Information). For LMEs in the Pacific Ocean, their relationshipwith the Pacific decadal oscillation was used. For LMEs in the North Atlantic andSouth Atlantic, the North Atlantic oscillation index and, respectively, the dipoleindex were used. For LMEs in the Indian Ocean, the Indian Ocean dipole index was used. For LMEs in the Arctic, the summer sea-ice extent was used. For LMEsin the Southern Ocean, the Antarctic Oscillation index was used.Accounting for the effects of fishing and large-scale oceanographic indices.For changes in MTC in each LME, we accounted for the effect of changes in totalfishing effort (E ) and the corresponding large-scale oceanographic index (OI). Wethenuseda GAMM (R package MuMIn)witha full model,MTC 5 s(Effort 1 s(OI)1 Year, where s represents a spline smoothing function (Supplementary Infor-mation), allowingfor a lagof upto three years inthe response of MTC to these two variablesand the time effect Year, and includingan autoregressiveterm to accountfortemporal autocorrelation. To estimate theaveragerateof change inMTC fromthe coefficients of the time effect, we did not include a spline smoothing functionfor the variable Year. These models were analysed using a multi-model analysisframework (R package MuMIn). There were significant autocorrelations of 1-yrlag in the MTC time series for all LMEs (P , 0.05). Rates of MTC changes (coef-ficientof Year) wereestimated using a multi-modelensemblemean withweighting (W ) calculated from W i 5 exp[(AICmin 2 AICi)/2], where AICmin is the bestmodel (that with minimum Akaike information criterion) and AIC i was basedon alternative model i. In addition, the rates of MTC change were also estimatedfrom the model with the lowest Akaike information criterion. The relationshipbetween the rates of change in MTC and SST remained significant with thesealternative estimates.

LETTER RESEARCH

ll bl h d ll h d201
http://www.fishbase.org/http://www.fishbase.org/http://www.sealifebase.org/http://www.sealifebase.org/http://www.fishbase.org/http://www.fishbase.org/

Documents

Signature of ocean warming in global fisheries catch.pdf