Content uploaded by Christophe Menkes
Author content
All content in this area was uploaded by Christophe Menkes on Nov 11, 2019
Content may be subject to copyright.
Glob Change Biol. 2019;00:1–13. wileyonlinelibrary.com/journal/gcb
|
1
© 2019 John Wiley & Sons Ltd
Received:20Decem ber2018
|
Revised:2 0September2019
|
Accepted:30September2019
DOI : 10.1111/gcb .14858
PRIMARY RESE ARCH ARTICLE
Trends in tuna carbon isotopes suggest global changes in
pelagic phytoplankton communities
Anne Lorrain1 | Heidi Pethybridge2 | Nicolas Cassar1,3 | Aurore Receveur4 |
Valérie Allain4 | Nathalie Bodin5,6 | Laurent Bopp7 | C. Anela Choy8 |
Leanne Duffy9 | Brian Fry10 | Nicolas Goñi11 | Brittany S. Graham12 |
Alistair J. Hobday2 | John M. Logan13 | Frederic Ménard14 |
Christophe E. Menkes15 | Robert J. Olson9 | Dan E. Pagendam16 | David Point17 |
Andrew T. Revill2 | Christopher J. Somes18 | Jock W. Young2
1IRD,CNRS ,Ifremer,LEMAR,UnivBrest,Plouzané ,France
2CSIROOceansandAtmosphere,Hobart ,Tas.,Australia
3Divisio nofEarthandOce anSciences,Ni cholasS chooloft heEnvironment ,DukeUni versit y,Durham,NC,USA
4Pacifi cCommunity,Ocea nicFish eriesProgram me,Nou méa,Fra nce
5IRD,FishingPort,Vic toria,Mahe,RepublicofS eychelles
6SeychellesFish ingAuth orit y(SFA),Vic toria ,Mahe,Re publicofSeychelles
7Labor atoiredeMétéorologieD ynamique(LMD),I nstitutPierr e‐SimonL aplace(IPSL),EcoleNor maleSupérieu re/PSLRe s.Univ.,CNRS,
EcolePoly technique,S orbon neUniversité,P aris,France
8IntegrativeOceanographyDivision,ScrippsInstitutionofOceanography,Unive rsityofCalifornia,SanDiego,LaJoll a,CA ,USA
9Inter‐Ameri canTropic alTunaCommis sion,(I ATTC),LaJolla,CA,USA
10AustralianR iversInstitu te,GriffithUniver sity,Nath an,Qld ,Australia
11AZ TI,MarineResearch,Pasaia ,Gipuzkoa,Spai n
12Nationa lInstituteofWaterandAtmos phericResearch,Ltd.(NIWA),Wellington,NewZeala nd
13Massa chuset tsDivisionofMarineFisheries,NewBedford,MA,US A
14AixMarseill eUniver sity,Universit yofToulon,CNRS,IRD,MIO,UM110,Mars eille,Fr ance
15IRD,UMRENT ROPIE,N ouméacedex,NewCale donia
16CSIRO,ComputationalInformatics,Brisb ane,Qld,Aust ralia
17Obser vatoireMidi‐P yréné es,GE T,UMRCNRS5563/IRD234,Université́PaulSabat ierToulouse3,Toulouse,France
18GEOMARH elmholtzCentreforOceanResearchKiel,Kie l,Germany
Correspondence
AnneLorrain,IRD,CNRS ,Ifrem er,LEMAR,
UnivBre st,F‐29280Plouzan é,France .
Email:anne.lorrain@ird.fr
Funding information
LabexMER,Grant/AwardNumber:ANR‐10‐
LABX‐19;FrenchGover nment
Abstract
Considerable uncertainty remains over how increasing atmospheric CO2 and
anthropoge nic climate changes are af fecting ope n‐ocean marine ecos ystems from
phytoplanktontotoppredators.Biologicaltimeseriesdataarethusurgentlyneeded
forthe world's oceans.Here,weusethecarbonstableisotopecompositionoftuna
toprovide afirstinsight intothe existence ofglobal trends in complex ecosystem
dynamicsand changes in the oceanic carbon cycle. From 2000to 2015, consider‐
abledeclinesinδ13Cvaluesof0.8‰–2.5‰wereobservedacrossthreetunaspecies
sampled globally,with moresubstantial changesin the Pacific Ocean comparedto
theAtlanticandIndianOceans.TunarecordednotonlytheSuesseffect,thatis,fossil
fuel‐derivedandisotopicallylightcarbonbeingincorporatedintomarineecosystems,
2
|
LORRAIN et AL .
1 | INTRODUCTION
Ov e rthe p ast5 0 yea r s , 90%o fthe h e ata s soci a tedw i thg l o b alw a r m‐
ing,and30%ofthefossilfuelcarbonemissionshavebeenabsorbed
bytheoceans(LeQuereetal.,2018).Suchprocessesarepredicted
toseverelyimpactmarine biota (Poloczanska et al., 2016)through
enhancedoceanstratificationandacidification.Unfortunately,there
are large un certaint ies on how oceanic e cosystems have ch anged
or may change i n the future. Fo r example, th e current gene ration
of eart h system mod els simulates a wi de range of futu re changes
inglobaloceannetprimaryproductivity(NPP),withbothincreases
anddecreasesofupto20%by2100(Boppetal.,2013;Kwiatkowski
etal., 2017),highlightinglargediscrepancies inthetrends ofsimu‐
latedNPP.Onlyalimitednumberofempiricaldatasetsrecordtrends
inthephytoplanktoncommunitycompositionorphysiology(Gregg
&Rousseaux,2014;Gregg,Rousseaux,&Franz,2017;Rousseaux&
Gregg,2015).Themagnificationofrelativechangesin phytoplank‐
tondynamicsacrosstrophiclevelshasrarelybeeninvestigatedwith
fewavailableempiricalmethodscapableofquantifyingecosystem‐
level resp onses. Biolo gical time ser ies dataset s are imperati ve for
understandingpast responsesofthe world'soceans andfor quan‐
titati ng uncert ainty in fu ture climate p rojectio ns (Bonan & Do ney,
2018).
Carbonstableisotopes(δ13Cvaluesor13C/12C)havebeenused
to reconstruct the oceanic carbon cycle using direct measure‐
mentsormarinearchives(e.g.,marinesediments,corals)frompa‐
leoclimatestothecurrentanthropogenicperturbation(Ehleringer,
Buchman n, & Flanagan , 2000 ; Freeman & Hayes, 1992; Keel ing,
2017;Wuetal.,2018).SincetheIndustrialRevolution,therisein
atmosphericCO2hasbeenaccompaniedbyadecreaseinthecar‐
bonisotope ratioof atmospheric CO2, known as the Suessef fect
(Keeling, 1979).Thisdecreaseisat tributedtotheatmosphericre‐
leaseofisotopically lightcarbonfromfossilfuelcombustion.Due
totheoceanic uptake ofthis 13C‐depletedCO2,the oceanicδ13 C
value of dissolved inorganic carbon (δ13CDIC) is decreasing (Quay,
Sonnerup,Munro, & Sweeney,2016;Quay et al., 2007).Changes
in δ13 CDICvaluesarerecordedinphytoplanktonδ13Cvaluesafter
accounting for an isotopic fractionation factor associated with
photosynthesis (defined as εp). Isotopic fractionation is depen‐
dent on seawater characteristics, phytoplankton composition,
and physio logy. The primar y factors that are b elieved to affec t
the isotopic values of phytoplankton are: (a) the concentration
and δ13Cvaluesofdissolved CO2([CO2]aq;Fr y,1996;Laws,Popp,
Bidigar e, Kennicutt, & M acko, 1995;P opp et al., 1998); (b) phy‐
toplankton community composition and cell morphology (Popp
etal.,1998);and(c)cellulargrowthrate(Bidigareetal.,1997;Fry,
1996).Secondary physiologicaltraits (e.g., decreases in bicarbon‐
ateu pta keo ri nca r bon ‐co nce nt r ati ngm ech ani sma c tiv ity)ca nal so
impact isotopicvalues, but are difficult tomodel (Cassar,Laws,&
Popp,2006).
Carbonisotopic changes at the baseoffood webs are trans‐
ferredtohighertrophiclevelswit hvaluesincreasingslightly(ty p‐
ically 0.5‰–1‰)with each trophic transfer (Fr y,2006; Graham
etal.,2010).Metabolically activetissuesof consumers (e.g., fish
muscle) integrate the stable isotope values of thisbase through
their diet (C herel & Hobso n, 2007; Graham e t al., 2010). While
nitrogenisotope (δ15N)values are commonly usedtoinvestigate
changes in trophic levels, δ13C values provide information on
animal die ts and on spat ial variation s at the base of food w ebs
(Cherel&Hobson,2007;MacKenzie,Longmore,Preece,Lucas,&
Trueman,2014;Trueman,MacKenzie,&Palmer,2012).Historical
studies focusing on baseline changes haveex amined accretion‐
arybioarchivesthatsufferlittledegradationafterformationsuch
askeratin baleenplates, feathers or teethdentin of marinecon‐
sumers thatreflectthefood theyingest and therefore,theδ13C
valuesofphytoplank ton(Jaeger& Cherel,2011;Newsomeetal.,
2007;Schell,20 01).Thesestudiesdemonstratetheutilityofiso‐
tope meas urements to rec onstruct p ast and present o cean pri‐
mary productivity,andprovideevidenceof past climatechanges
atregional scales(Hobson,Sinclair,York,Thomason,& Merrick,
2004;Newsomeetal.,2007;Schell,20 01).Finally,metabolically
inert but inorganic accretionary str uctures (e.g., bivalve shells,
co ralsk ele to n s,s cl e ro spon ge s ,o rfi sho to lit hs)ca nal sor ef l ect the
δ13CDICvalueof th ee nv ir on me nt ac ro ssth ei rl ifet im e(Frailee ta l. ,
but also recorded profound changes at the base of marine food webs. Wesuggest
a global shif t in phytoplankton community structure, for example, a reduction in
13C‐richphytoplankton suchas diatoms,and/ora changeinphytoplankton physiol‐
ogyduringthisperiod,althoughthisdoesnotruleoutotherconcomitantchangesat
higherlevelsinthefoodwebs.Ourstudyestablishestunaδ13Cvaluesasacandidate
essentialoceanvariabletoassesscomplexecosystemresponsestoclimatechangeat
regionaltoglobalscalesandoverdecadaltimescales.Finally,thistimeserieswillbe
invaluableincalibratingandvalidatingglobalearthsystemmodelstoprojectchanges
inmarinebiota.
KEYWORDS
albacoretuna,AtlanticOcean,bigeyetuna,biogeochemicalcycles,carboncycle,IndianOcean,
PacificOcean,phy toplankton,Suesseffect,yellowfintuna
|
3
LORRA IN et AL .
2016;Swart,2010)astheyusuallyprecipitateinequilibriumwith
seawater, although vital effects can complicate environmental
reconstruction (Lorrain, 2004;McConnaughey &Gillikin, 2008).
Similarly,δ13C values of metabolically active tissues may reflect
trendsinphysio‐chemicalprocesses (CO2 and δ13Caqvalues)and
biologicalprocesses(phytoplanktonδ13Cvalues).
Theaimofourstudy wastoassesstrendsinatimeseriesof
stableisotopevaluesofmetabolicallyactivetunatissues,andto
test if thi s could be used to d etect ecos ystem‐level re sponses
over deca dal time scales at r egional to global sc ales. For this
purpos e, we analyze d δ13C and δ15Nvalues ofmusclesamples
from thre e species of tuna (ye llowfin tun a, Thunnus albacares;
bigeye tuna,T. obe su s;and albacore tuna,T. alalunga) collected
throughout tropical, subtropical, and temperate oceans from
200 0 to 2015 (n=4,477;Figure1).Eachofthesespecieshas
different vertical and foraging distributions (from surface to
mesopelagic depths)(Olson et al., 2016). Therefore, our study
seekstoresolvebroadhorizontalandverticalspatialpatternsin
oce anicf oo dwebs .A stunaar ewide lydistributedandha rves te d
globall y (Majkowski, 20 07), they are goo d candidate s to study
how obser ved and suspec ted changes in phys ical and biologi‐
calprocessesat globalandoceanbasinscales maybereflected
inconsumerδ13C values. We developed a theoretical modelto
decompose the observed temporal changes in consumer δ13C
valuesintoputativecausal contributors. The model accounted
for(a)known temporaltrends infossilfuel–derivedcarbon (the
Suesseffect)andCO2availability;(b)possiblechangesinphyto‐
planktondynamicsincludingcommunitycompositionandgrowth
rates;and(c)potentialchanges inthetrophicfractionation fac‐
tor.Our study,which focusesoncarbonbutdraws on nitrogen
isotopes to a ssess potenti al changes in tun a trophic posi tions,
suggestslarge‐scaleshifts inphytoplankton communitiesfrom
2000to2015.
2 | MATERIALS AND METHODS
2.1 | Tuna carbon isotope data
We assemble d a global dat abase using pu blished an d unpublishe d
regional carbon isotope studies resulting in 4,477 records from
2000to 2015. Det ails on isotopic methods andpredatorsampling
are provid ed in Pethybridge e t al. (2018) who analy zed the same
globaldatasetbutforδ15Nvalues.Asforδ15Nandotherglobalcom‐
pilation studies (Birdetal., 2018), we assumedthat the agreement
between δ13C values generated ac ross different lab oratories was
<0.2‰–0.3‰.Tunasize(forkleng th,incm)wasmeasuredforeach
individual. Tuna were sam pled from three oce an basins (Atlantic,
Indian,andPacificOceans)withalbacoretunaoccupyingmoretem‐
perate waterscomparedto thetropicalyellowfinandbigeyetunas
(Figure1;Olsonetal.,2016).
The Pacific Ocean had the most extensivesampling with 2,504
individualsandnogapfrom2000to2015,exceptforalbacorewhere
datawerenotavailablefor4years.IntheIndianandAtlanticOceans,
datawere more scattered(Table S1). Pelagictuna are mobile preda‐
torsandthestablecarbonisotopiccompositionoftheirmuscletissue
representsanintegrationoftheirforagingenvironmentoverapproxi‐
mately6monthsto1year(Houssardetal.,2017).Tunamuscletissue
δ13Cvalueswerecorrectedforlipidsinallsampleseitherwithchemi‐
calextractionorusingamassbalanceequationforelevatedlipidcon‐
tentsamples(C/N>3.5)withparametersderivedfromAtlanticbluefin
tuna(T. thynnus)muscle(Loganetal.,2008).
2.2 | Temporal trends in tuna δ13C values
Time series analyses based on multiple linear regression analysis,
performedusingthe R‐3.2.4 software (R Development CoreTeam,
2016) and the nlme package (Pinheiro et al., 2018), wereused to
examineandtestforsignificantlineartrendsintunacarbonisotope
values.Toensurethat tuna length(size)did nothaveany effecton
potentialtemporaltrends,aninteractionbetweensizeandyearwas
testedandwasnotfoundtobesignificantforthePacificorAtlantic
Ocean.Fortheglobaldataset,wetestedamodelwithtuna sizein‐
cluded,byspeciesandoceanbasins,andthenfittedamodelexplain‐
ingtheresidualsofthis firstmodelasafunctionofyear.Theslopes
weresimilar to those obtainedwithout theeffectofsizeincluded,
meaningthattheadditionofsizedoesnotchangetheobservedpat‐
ternsand that this factor has a small impact on temporal trendsin
δ13Cvalues.Wefinallytestedforthreevariables:year(quantitative),
ocean with three levels (Atlantic, Indian, and Pacific Oceans), and
tuna species with threelevels (albacore, bigeye, andyellowfin). All
combinationsweretestedandthefinalmodelwaschosenusingthe
Akaikeinformationcriterion.Weaddedanautocorrelationstructure:
FIGURE 1 Mapofglobalstudyarea
andlocationsof4,477samplesfor
threetunaspecies.Theblacksquare
delineatestheNewCaledonia–Fijiregion
usedforafocusedspatialandtemporal
analysis
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
●
●
●
●
●
●
●
●
40ºS
20ºS
0
20ºN
40ºN
60ºN
60ºE 100ºE 140ºE 180º 140ºW 100ºW 60°W 20ºW
Tuna species
●
●
●
Albacore
Bigeye
Yellowfin
4
|
LORRAIN et AL .
a one‐degree autoregressive integrated moving average (ARIMA;
Pinheiro et al., 2018)via the gls function fitted by groups of tuna
sampled at the same date and position.Autocorrelation structures
onresidualswerecheckedwith anautocorrelationfunction. Finally,
toaccountforpossiblespatialbiases(a)inyearsofsamplingaccord‐
ingtolocations(FigureS1) or(b)due tobaselineisotopicvariations
acrossspace(McMahon,LingHamady,&Thorrold,2013),tunaδ13C
trendswerealsodeterminedatasmallerspatialscalebyconsidering
oneregionwheresufficientdatawereavailableperyear,thatis,New
Caledonia and Fiji (see Figure1 for selected region area) andwith
similarisotopevaluesat thebaseofthe foodweb (Houssardet al.,
2017; Magozzi, Yool, Vander Zanden, Wunder, & Trueman, 2017).
Furthermore,all New Caledonia and Fiji samples were analyzed in
thesamelaboratoryandtimeperiod.
2.3 | Modeling the factors influencing tuna
δ13C values
Wedevelopedatheoretical model to explain thepotentialeffects
of various f actors and p rocesses know n to explain tren ds in tuna
δ13Cvalues.First,weconsideredtheisotopevalueofphytoplankton
(δ13Cp)that has beenshown to be driven by the magnitude of car‐
bonisotopicfractionationduringphotosynthesis(εp)andtheisotope
valueofCO2(δ13Caq,thatis,theSuesseffect),withεpdependenton
thecarbonisotopefractionationassociatedwithcarbonfixation(εf)
andthespecificgrowthrate(µ;Lawsetal.,1995).
with
wherebisaconstant(mM/day)reflectingthedegreeofdependence
offractionation on theCO2concentration,and isbelievedto vary
between species and as a function of growthconditions (Bidigare
etal.,1997;Cullen,Rosenthal,&Falkowski,2001).Whiletheparam‐
etervaluesarearbitrary,theyarewithintherangeofvaluesreported
intheliterature(Table1).Aninitialvalueof120wasusedforbwhich
isconsistentwiththerangeofvalues(52.6–137.9)fromPoppetal.
(1998)forEmiliana huxleyi(εp = 24.6–137.9 µ/CO2)themostcommon
species of coccolithophoreglobally (Beardall & Raven, 2013), and
Phaeodactylum tricornutum(εp = 25.5–52.6 µ/CO2),adiatommostly
usedinlaboratorystudiesbutnotrepresentativeoftheglobalocean.
Fortheintrinsicfractionationduringphotosynthesisbytheenzyme
Rubisco(εf),weused avalueof25‰,whichhas beenestimated to
rangebetween~22‰ and 30‰ depending on species (e.g., Popp
etal.,1998),withvaluesaslowas11‰fortheRubiscoofthecocco‐
lithophoreE. huxleyi(Boller,Thomas,Cavanaugh,&Scott,2011).The
valueof25‰forεfhasoftenbeenproposed(Bidigareetal.,1997)in
studiestryingtounderstandtemporaltrendsinmarinechronologies
(Schell,2001).
(1)
δ
13Cp=
1
(
1+𝜀p
1,000 )
(δ13Caq −𝜀p)
,
(2)
𝜀
p=𝜀f−
b
𝜇
CO
2
,
Factors (x)
Starting
value or
equation
used
Sensitivity
(d
𝛅
13C
tuna
𝛅13C
tuna /
dx
x
)
Imposed
change
%
Tun a
Δδ13C,
‰,
16‐years
% Change explained
NC‐Fiji PO AO IO
δ13Caq Quayetal.
(2016)
−0.08 NA 0.30 14 12 22 38
CO2Cassaretal.
(200 6)
0.21 NA 0.06 3 2 4 8
Bidigare
(1997 )
NA 0.27 13 11 20 35
Growth
rate(µ)
0.3 −0.2 −5 0.20 10 814 26
−10 0.30 14 12 22 38
−15 0.48 23 19 35 62
Carbon
fixation
fractionation
factor(εf)
25 1.53 +1 0. 24 12 10 17 31
+2 0.48 23 19 35 61
+3 0.72 35 29 52 92
+5 1.20 58 48 87 154
bFactor 120 −0.2 −2 0.10 5 4 7 13
−4 0.20 10 814 26
−8 0.40 19 16 29 51
−10 0.50 24 20 36 64
Trophic
fractionation
factor(εfc)
4−0.23 −5 0.20 10 814 26
−10 0.40 19 16 29 51
−15 0.59 29 24 43 76
TABLE 1 Parameterandtimeseries
datausedtorunvariousscenariosof
imposedchangesinphytoplankton
dynamic s,andtheeffectsonthedifferent
spatialareasexamined(NewCaledonia–
Fiji[NC‐Fiji]region,PacificOcean[PO],
AtlanticOcean[AO],andIndianOcean
[IO]).Forexample,δ13Caqexplained
12%ofthetunaδ13CdecreaseΔδ13C, in
16yearsinthePO,whileanimposed5%
changeincarbonfixationfractionation
factor(εf)explained48%ofΔδ13C,this
factorhavingthelargestsensitivity
(1.53)comparedtoallfactorstested(see
Section2formoredetails)
|
5
LORRA IN et AL .
While precise estimates ofεf or b are not available, this param‐
etrization provides a quantitativedemonstration of how even small
changesinphytoplanktoncommunitycompositionorphysiologymay
influence tuna muscle δ13C values, and hence emergent signals of
change(Table1).Growthrateµwas setat0.3 day−1asitis the me‐
dianvalueatStationALOHA(Lawset al., 2013)inthe central North
Pacific(Hawaii OceanTime‐series) and was alsousedbyseveral au‐
thors(Cullenetal.,2001).Rangesforµfrom0.1to1day−1havebeen
reported in the literature(Boyd et al., 2013; Laws et al., 2013). We
proposedadecreaseingrowthrateofupto15%overthe2000–2015
studyperiod(from0.30to0.26),whichisonthehighendofobserved
modernchanges(Greggetal.,2017)andpredicteddecreasesforthe
future(Kwiatkowskietal.,2017).Theδ13Caq valuesweretakenfrom
StationALOHA(Quayetal.,2016).
Changesintunaδ13Cvaluescaninturnbedescribedbythefol‐
lowingequation:
whereεfcistheoverallfractionationassociatedwithtrophictrans‐
fersandisconsideredtobelow(~0.5‰to1.8‰pertrophiclevel),
thereforeweused4‰fortunathatareconsideredtobeatatrophic
positionof~4(Olsonetal.,2016).Asacomparison,Birdetal.(2018)
found an average difference of4.6‰ between phytoplank ton and
sharksonaglobalscale.
CombiningEquation(2)withEquation(3)leadsto:
We used two dif ferent parametrizations developed in the litera‐
ture on the effect of the concentrationof CO2aqontheisotopeval‐
ues of phytoplankton (Bidigare et al., 1997; Cassar et al., 2006).
Indeed, while the first parametrization (Bidigare et al., 1997; Laws
et al., 1995) provides quantitative intuition for the dependence of
tuna δ13C value s on phytoplank ton physiology, it do es not account
for additional factors influencing the phytoplankton isotope val‐
ues, including changes in the carbon source (bicarbonate vs. CO2),
deactivation of carbon‐concentrating mechanisms in response
to increased CO2 availability, and changes in the growth condi‐
tions (nutrient vs. light limitation). For completeness, we also show
the predictions based on the parametrization presented in Cassar
et al. (20 06). The sensiti vity of tuna δ13C va lues to each fa ctor was
assessedbycalculatingtheratioofthepercentagechangeoftunaδ13C
values to p ercentage chan ge of each facto r. T he slope (‘m’) of eac h
curveisrelatedtotheratioofthepercentagechangeoftunaδ13C val‐
ues to perce ntage change of e ach factor a ccording to the fol lowing
equation:
Thissensitivityanalysisexamines the influence of one parameter at
a time on tuna isotopic composition with assumed initialvalues for
eachparameter basedonliteraturevalues.Toreinforcethis analysis,
weconductedaBayesianapproachthattakesintoaccounttheuncer‐
taintyofall parameterssimultaneouslytoexplaintuna isotopiccom‐
position with ranges and uncertainties taken from literature values
(AnalysisS1).Toreconstructtheobservedtrendsintunaδ13Cvalues,
anumberofscenarioswereruntosimulatepercentagechangesinthe
phytoplanktonparametersµ, b, and εf or εfc(Table1).Thesescenarios
wereusedtoresolvecompetinghypothesesfortheobservedpatterns.
3 | RESULTS
3.1 | Trends in δ13C values
Overtheentirerecord(Figure1),individualtunaδ13Cvaluesranged
from−19.9‰to−12.9‰.Mean annualδ13 Cvalues decreased by
0.8‰ to 2.5‰ within species and ocean basins from 20 00 to
2015(Table2;Figure 2). These negative trendsweresignificant,
withsimilarobservedslopesforeachtunaspeciesbyoceanbasin
(p<.0001;Table2).Thelargestdecreasewasobservedforthe
Pacific O cean and the l owest in the In dian Ocean, wi th the de‐
creaseintheAtlanticOceanbeingintermediate.Forreference,we
showedaglobaldecreasingtrendacknowledgingthatmostofour
observationswerefromthePacific(56%)(Table2;Figure2;1.8‰
decreasefrom2000to2015),whileobservationsfromtheIndian
andAtlanticOceanseachcomprised22%ofthedata(Table2).
In the regi on of New Caledo nia and Fiji, whe re we have the
most complete record to derive a temporal trend(Figure S1),the
same decreasing patternintunaδ13Cvalues was observedforall
threespecies(2.1‰decreasebetween200 0and2015;FigureS2)
asoverthe broader Pacific region (2.5‰ decrease).No temporal
changesinδ15Nvalu es wereo bser ve dfort hethreetunas pe ci esin
theNewCaledonia andFijiregions (FigureS3), suggestingnosig‐
nificanttunatrophicpositionchangesovertherecord.Someweak
trendsinδ15Nvalueswere foundfor some speciesandoceanba‐
si ns , whi chcou l da r is efr o mt h ei nte r act i onwit hot h ercon fou ndi n g
factor s such as tuna s ize and locat ion (Figure S 4). Twolikel y ex‐
planationsfortheobservedtunaδ13Ctrendsarediscussedbelow.
3.2 | Accounting for the observed Suess effect and
CO2 availability
Reported declines in δ13Caq values at Station ALOHA during the
2000–2015period(−0.3‰) explained 14%of thedecrease in tuna
Δδ13CvaluesobservedinourNewCaledonia–Fijiregion(Figure3a).
Assumingsimilar Suess effects inthe other ocean basins,12%,
22%, and 3 8% of the decreas e in tuna δ13C value s in the Pacific ,
Atlantic,andIndianOceans,respectively,canbeexplainedbyδ13Caq
(Table1).
An incre ase in the conce ntration of CO2aq obs erved at St ation
ALOHA also leadsto a decrease in tuna δ13C values by increasing
carbonisotopicfractionationduringphotosynthesisεp(Equation2).
(3)
δ
13Ctuna =
1
(
1+𝜀fc
1,000
)
(δ13Cp−𝜀fc )
,
(4)
δ
13Ctuna =1
1+𝜀fc
1,000
1
1+𝜀f−b𝜇
CO2
1,000
δ13Caq −𝜀f−b𝜇
CO2
−𝜀fc
.
(5)
(
dδ13Ctuna
δ13Ctuna
/
dx
x
)
=mx
δ13Ctuna
.
6
|
LORRAIN et AL .
Depend ing on the par ametrizat ion used to acco unt for this ef fect
(Bidigareetal., 1997;C assar etal., 2006), 2%–35%ofthe trendin
tunaδ13Cvalues can be explained bychanges in CO2availabilityin
the New Caledonia–Fijiregion (Figure 3b). Usingthe largerdegree
ofchangeinresponsetoCO2availability(Bidigareetal.,1997),per‐
centages of chang e similar to thos e from the Sues s effect c an be
explainedacrossthedifferentoceanbasins.Theadditiveimpactsof
CO2availabilit yandtheSuesseffectinexplainingtunaΔδ13Cvalues
are23%, 41%,and73%inthePacific,Atlantic,and IndianOceans,
respectively. IftheCO2availabilityef fectreportedbyCassar et al.
(2006)isused,inadditiontotheSuesseffect,thenonly14%,26%,
and 46% of the Δδ13C decrease inthe Pacific, Atlantic,and Indian
Oceans,respectively,canbeexplained.
3.3 | Hypothesized changes in phytoplankton
dynamics
According to theoretical models (see Section 2), the then unex‐
plainedtemporalchangesintunaδ13Cvalues(~27%–86%)mustbe
relatedto (a)a decreasein phytoplanktoncellulargrowthrates (µ)
orphysiology(e.g.,carbon‐concentrating mechanism activity);and/
or (b) potential changes in phytoplankton communities (through
changesinspecies‐dependentparametersb and εf)orinthetrophic
fractionation factor εfc. Based both on the sensitivityanalysis and
theBayesian inference,variations in the growthrate(µ) and in the
trophicfractionationfactor(εfc)haveasmalleffect ontunaisotope
values(Table1;AnalysisS1).Asanexample,an imposedsubstantial
15%decreaseinthe growthrate µover16yearsresultedinonlya
0.48‰decreaseintuna δ13Cvalues(Figure 3c;around20% ofthe
totaldecrease in theNewCaledonia–Fiji regionandin the broader
PacificOcean).This effectislargerin theAtlantic(35%)andIndian
(62%)Oceans.However, morereasonable declines of 5% and 10%
ofµover16yearsresultedinsmallerdecreasesof0.2‰and0.3‰
intunaδ13Cvaluesfrom2000to2015,respectively(Table1),which
explained 8%–38% of this overall signal in various ocean basins.
Variationsinthetrophicfractionationfactorεfcwereofsimilarorder
(Table1)withlargedecreasesofthisparameterneededtoexplainthe
tunaδ13Cpattern.
FIGURE 2 Timeseriesoftunamuscle
tissueδ13Cvalues(‰)withobservations
dividedbyoceanbasin.Theshadedarea
alongthelinearfitcorrespondstoa95%
confidenceinterval
Albacore Bigeye Yellowfin All oceans and species
2002 2006 2010 2014 2002 2006 2010 2014 2002 2006 2010 2014 2002 2006 2010 2014
−20
−18
−16
−14
Year
δ
13
C
Ocean
●
●
●
Atlantic
Indian
Pacific
Ocean basin/
region
Tun a
species
Intercept
(in 2000) Slope
Slope
stand‐
ard error r2
Temporal
change
(2000–
2015,
in ‰) n
Atlantic Albacore −17. 4 −0.092 0.0085 69.1 −1 . 3 8 608
Bigeye −16 . 6 126
Yel lo w fin −1 6 .6 256
Indian Albacore −16 . 8 −0.052 0.0077 69. 1 −0.78 24 8
Bigeye −16 . 1 237
Yel lo w fin −1 6 .1 498
Pacific Albacore −15 . 2 −0.166 0.0048 69. 1 −2.4 9 878
Bigeye −14 . 5 645
Yel lo w fin −14. 5 981
New
Cale donia–
Fiji
Albacore −15 . 3 −0 .138 0.0095 42.1 −2.0 7 364
Bigeye −14 . 8 120
Yel lo w fin −14. 9 331
Global All −15 . 4 −0.120 0.0057 22.1 −1 . 8 0 4,477
TABLE 2 Regressionanalysisoutput
includingtheslopeandinterceptforeach
tunaspeciesandoceanbasin.Onlyone
valueisshownwhensimilarforseveral
species
|
7
LORRA IN et AL .
Thecarbonfixationfractionationfactor(εf)andbvaluescanvary
widely among phytoplankton species (Popp et al., 1998). The sen‐
sitivit y analysis and the Bayesian model (that takes into account a
largerangeofvaluesfortheseparameters)showedthatthecarbon
fixationfractionationfactorεfhadthelar ges teffectont het unaiso‐
topevalues compared toallother factors(Table1;Analysis S1).As
anexample,wearbitrarilysetthechangesto5%forεfand10%forb
(Figure3d)toreflecttheirdifferentialimpactontunaΔδ13Cvalues.
Thissmall5% increasein εfresultedinalargedecreaseintunacar‐
bonisotopevaluesof1.2‰(i.e.,~50%ofthetunaΔδ13CintheNew
Caledonia–FijiregionandthePacific Ocean)(Figure3d; Table1).In
comparison,a 10%declinein‘b’values only causeda0.5‰decline
in tuna δ13C val ues, which ex plained 20%–36% in the P acific and
AtlanticOceans,and64%intheIndianOcean.
AfteraccountingfortheSuessandCO2availabilityeffects,sev‐
eral permutations for the parameters reflecting productivity (µ)
and spec ies composit ion (with chan ges in εf and b comb ined) may
accountfor the remaining Δδ13C changes. The combination of the
Suess ef fect, th e effect of inc reasing [CO2]aq on εp, and a chang e
of5%–10%inspecies‐specificparameters(10%forband5%forεf)
with no cha nge in produc tivity or t he trophic fr actionat ion factor
εfc,produced a 2.1‰decrease in tuna δ13C values,consistentwith
theobservedchangeinthePacificOcean(Figure3d).InthePacific
Ocean, w here we have the most ro bust dataset , changes in phy‐
toplank ton paramete rs seem to have occur red, unless we a ssume
that growth rates have changed by >70%. If we assume that no
changesingrowthratesandεfchaveoccurredandusetheBidigare
etal. (1997)parametrization, thenmore than 60% of Δδ13C has to
beexplainedbya changeinspecies compositioninthePacific and
AtlanticOceans,againstonly27%intheIndianOcean(Table1).The
use of the Cassaret al. (2006) parametrizationimplieseven larger
changes in s pecies comp osition. Aver aging all tu na species an d all
ocean basins,and both parametrizations used to calculate carbon
fractionationfromphytoplankton,theglobaltrendintunaδ13C val‐
ues(Δδ13 C)can,forexample,beexplainedby(a)theobservedSuess
effectand increasesinCO2aq(upto26%);(b)a5%decreaseinpro‐
ductivity (11%),a 10% decrease in thetrophicfractionation factor
(17%);and(c)imposedchangesof5%inspecies‐specificparameters
FIGURE 3 Predicted(colorline)versusobserved(blackline)changesintunamuscleδ13 Cvalues(Δδ13C,‰)intheNewCaledonia–Fiji
regionofthePacificOceanasafunctionofvariousprocesses.(a)TheSuesseffect.(b)IncreaseinCO2aqinseawaterundertwoscenarios
basedondifferentparametrizationsintheliterature(Bidigareetal.1997,Cassaretal.,2006;seeSection2fordet ails).(c)Adecreasein
phytoplanktoncellulargrowthrateof15%.(d)Achangeof5%forthecarbonfixationfractionationfactorεfand10%fortheconstantbused
tocalculatecarbonisotopefractionationduringphotosynthesis(seeSection2fordetails),andalsoallfactorsconsideredtogether,except
growthrate(blueline=Suesseffect+CO2aqfromBidigareetal.,1997+b + εf)
(c) Growth rate effect (d) Phytoplankton species (or physiological) effects
(a)
Suess effect
(b)
CO2 availability effect
2000 2005 2010 2015 2000 2005 2010 2015
−2.0
−1.5
−1.0
−0.5
0.0
−2.0
−1.5
−1.0
−0.5
0.0
Year
Observed tuna change
Observed tuna change Observed tuna change
Observed tuna change
All factors but growth rate
b
Growth rate
δ
13Caq effect
from Bidigare et al. (1997)
ε
f
from Cassar et al. (2006)
∆ δ C
13 tuna (per mil)
8
|
LORRAIN et AL .
indicatingashiftinspeciescomposition(46%;Figure4).Whilethisis
onepote ntials ce nario,inpa r ti nfor me da ndco ns tra in ed byobse rva‐
tionsintheliterature(Greggetal.,2017;Rousseaux&Gregg,2015),
thereisamultitudeofpermutationsthatmayfittheobservedtrend.
Neve rtheless,changesint hec ar bonfixat ionfractionationfactor(εf)
had the largest effects in the simulations using bothmodeling ap‐
proaches,and better accountedforthe observed tunatrends than
changesinproductivity,thetrophicfractionationfactor(εfc),oreven
theknownSuesseffect.
4 | DISCUSSION
Ouranalysisrevealedthatchangesinthebiologicalcomponentofthe
marinecarboncyclecanbetracedinthetissuesofmarinetoppreda‐
tors.Weobservedsubstantialandwidespreaddeclinesintunamuscle
δ13Cvalues(by0.8‰–2.5‰)inthreetunaspeciesacrossthreeocean
basins.Suchatrendovera16‐yearperiodhasneverbeenrecordedin
metabolictissuesofamarinepredator.Theuseoftwoseparatemod‐
elingapproaches(sensitivityanalysisandBayesianinference)revealed
thattheparameterlinked to phytoplankton carbon fractionation (εf)
hadthelargestinfluenceontheobservedtemporaltrendintunamus‐
cle δ13Cvalues.Our calculationsthensuggest thatupto 60% ofthe
decrease in tuna δ13C values seems tobe due to a change in phy‐
toplankton parameters in the PacificOcean, compared to only 27%
inthe western Indian Ocean.Whileourmost robust dataset isfrom
the Pacific Ocean, the same decreasingpattern in tuna δ13C values
in all ocean basins (Pacific,Atlantic, and western Indian) suggests a
widespreadshiftinmarineplanktoncommunitiesorachangeintheir
physiology,butdoesnotexcludeotherfactorsthatmayactinsynergy
(e.g.,achangeinproductivityorinthetrophicfractionationfactor).
Previouslyreportedtemporal changesinδ13C valuesaregen‐
erallyattributedtotheSuesseffectorchangesinmarineproduc‐
tivity in various organisms and ecosystems (Fraile et al., 2016;
Newsome et al., 2007; Schell, 2001). For example, Schell (2001)
foundasignificantlong‐termdeclineinδ13Cvaluesininertbaleen
plates (~2.7‰) over a 30 year p eriod (bet ween 1965 and 1997)
attributingthisdeclinetoa~3 0%–40%declineinprimar yprodu c‐
tivit y in the Bering S ea. Cullen et al . (2001) propos ed that part
ofthe decrease observed by Schell (2001)was due to theSuess
effectanddue to the influence of changesinCO2 concentration
onph y to p la n kto np hys iol ogy (a ssh ow nin Fig ure3b and de s cri bed
herein). In co ntrast, t he Suess ef fect is relat ively small over o ur
timeperiod(0.3‰;Figure3a)andonlyexplains~12%–20%ofthe
observeddecreaseintunamuscleδ13Cvalues(intheAtlanticand
PacificOceans).Similarly,increasingCO2concentrationsonlyex‐
plain asmall percentage (~2%–18%; Table1)of theobservedde‐
creaseintunamuscleδ13Cvalues,usingthe Bidigareetal.(1997)
orCassaretal.(20 06)par ametrizationsforth eireffec tonthecar‐
bonfixationfactorfractionation.
AfteraccountingfortheobservedSuesseffectandchangesin
[CO2]aq availabili ty, our model w as used to expl ore how change s
inphytoplankton growth rates and species composition can fur‐
therreconstructourobserveddeclinesintunamuscleδ13Cvalues.
Thesensitivity,Bayesian,andscenarioanalysisdemonstratedthat
relativelysmallchangesinphytoplanktoncommunitycomposition
canleadtolargedeclinesintunaδ13Cval ues(Tabl e1;AnalysisS1),
while lar ger changes in pr oductivi ty or the trop hic fract ionation
factor w ould be need ed. In our st udy, a 15%d ecrease in p hyto‐
plankton productivity cannot explain the decline we observed
intuna muscle δ13C values, evenwhen combined with the Suess
effectand thecumulativeeffect of increasing [CO2]aq on εp. T he
changeinproductivitythatwetestedinthisstudyisatthehigher
endofpreviouslyreporteddeclines(typicallyrangingbetween0%
and 1.4% year−1attheregionaltooceanbasinscale;Behrenfeld,
2006; Gregg & Rousseaux, 2014; Joo, 2015).Other studieshave
sh ow nnore cen tt ren dsi ng lob alp rim ar ypro duc t ivi ty(Gre ggeta l.,
2017;Rousse au x&G re gg ,2017)andso mereg ionalincre aseshave
FIGURE 4 Synthesisofthepotential
effectsofvariousfactorsonthetuna
δ13Ctemporaltrend(Δδ13C).Different
combinationsarepossible(seetext
formoredetails)TunaillustrationLes
Hata©SPC
|
9
LORRA IN et AL .
been rep orted historic ally (e.g., Pacific Ocean; Karl, Bidig are, &
Letelier,2001).Eveninmodelstudies,projectionsofglobalmarine
NPParehighlyuncertainwit hrelati vechangesbe tween−20%and
+20%overthe21stcentury(Kwiatkowskietal.,2017).
The rate ofdecrease observed inour tuna δ13Cvaluesrequires
concomitantchanges in phytoplankton species‐specific parameters
(εf and b).Young,B ruggeman,R ickaby,Erez,an dConte( 2013)alr ea dy
repor ted evidenc e of change in the bi ological c arbon isoto pic frac‐
tionation byphy toplankton(εp) with significant increases bet ween
1960and2010,inparticularinthesubtropics,wherethischangewas
thehighestcomparedtootherregions.Thechangeinbiologicalfrac‐
tionationestimatedthrough theirmodel (i.e.,amaximumof0.4‰in
16years)istwotofi vetimeslowerth anourob ser vationsintuna,de‐
pendingontheoceanbasin.Theirstudyisbasedonacompilationof
particulateorganiccarbon(POC)datafromseveraltransectsmostly
from theAtlantic Ocean with few data in our Pacific region, which
could explainthe differences between their model and our obser‐
vations. H owever,t hey showed a time s eries of δ13C POC (δ13
POC)
values in the North Atlanticoff Bermudawith a 2‰decreasefrom
1980to2007(i.e., ~1.2‰decreasein16years).Thisresultissimilar
tot he1.4‰decreaseintunaval ue sweobser vedinth eAtlanticfr om
2000to2015.
Suppor tforourhypothesisofashiftinmarineplanktoncom‐
munitiesalreadyexists(McMahon,McCarthy,Sherwood,Larsen,
&G u i l d e r s on , 2 0 15 ; P o l ov i n a , Ho w e l l ,& A b e c a ss i s , 2 0 0 8 ;P o l o vi n a
& Woodworth, 2012; Rousseaux & Gregg, 2015). Diatoms are
predicted to decrease in abundance in response to increased
seawaterstratificationwithareporteddeclineof1.22%year−1 in
the Nor th Pacific ( Rousseaux & G regg, 2015). Su ch a reducti on
intheabundanceofdiatoms, a 13C‐rich carbon sourceinmarine
food webs (Fry & Wainright, 1991), is expected to decrease the
δ13Cvaluesofconsumers,and this diatom contributionhas been
emphasize d in a recent mode l of phytoplan kton δ13C variations
in the glob al ocean (Magoz zi et al., 2017). Tuerena et al . (2019)
alsorecentlyfoundthatcellsizewastheprimarydeterminantof
δ13
POC in the Sout h Atlantic subtropical conver gence zone and
predicted that carbon isotopic fractionation will increase in the
future, l eading to lower δ13
POCthatmaypropagatethroughthe
food web. A decrease in the abundance of coccolithophores,
another 13C‐rich carb on source, migh t also explain so me of the
tuna δ13C trend. However,Rivero‐Calle, Gnanadesikan, Castillo,
Balch, and Guikema (2015) found an increase in the occurrence
ofcoccolithophoresintheNor thAtlantic,butdataare notavail‐
ableataglobalscale.Oceanbasindifferencesfoundinourstudy
in the temp oral slopes in tu na δ13C values bet ween the Indian
Ocean (0. 8‰) and the Atlanti c and Pacific O ceans (from 1.4‰
to2.5‰) could be due to a combination of severalfactors. The
magnitu de of the Suess e ffect may v ary region ally (Quay et al. ,
2016). Changesinphytoplanktoncommunitiesorphysiologyare
also depe ndent on regio nal‐scale p rocesses (Gr egg et al., 2017;
Siegel et al ., 2013). The use of spat ially resolve d models of the
ocean 13Ccyclewouldhelptounderstandtheregionaldifferences
(Tagliabue&Bopp,2008).
5 | CAVEATS AND LIMITATIONS
Phytoplanktonhavemanystrategiestotakeupcarbonasafunction
of growth co nditions t hat could aff ect frac tionatio n. For exampl e,
adecreaseinbicarbonate uptake or carbon‐concentratingmecha‐
nismactivityingeneralwouldbepredictedtoincreasetheapparent
fractionation.Ourpredictionsshouldthereforebeinterpretedwith
cautionasisotopicfractionationis not a single function of µ/CO2,
even within a single phytoplankton species (Cassar & Laws, 2007;
Cassaretal.,2006).Furthermore,otherpermutationsofεf or b may
fitthe observeddecrease. However,thisparametrizationtogether
with the Bayesian inference demonstrates how small changes in
phytoplanktoncommunitycompositionorphysiologymayinfluence
tunamuscleδ13Cvalues.
We also note that r egional variat ions cannot b e captured by
the time series of δ13CaqatStationALOHA(Hawaii,Pacific).
Long‐term declines of δ13Caqvaluesduetothecombinationof
the Sues s effect , vertical m ixing, and p rimary pro duction (re sid‐
ual carbon pool af ter POC production) have been documented
at other monitoring stations, with varying effects according to
regionandlatitude,inparticular in the southern regions (King &
Howard,2004).However,bothinstrumentalandproxyrecordsof
δ13Caqindicateaconsistentaveragedecreaseperyearof0.027‰
atfivePacificstationsfromHawaiitoAmericanSamoasince1980
(corresponding to an approximately0.4‰ decline in our 16 year
peri od ;Wu,2018) .Fur therm ore,Gru beretal.(1999) co mpare dt he
δ13Caqtrendsinseveraloceanicregi onsandfoundthatt hehighest
decreas e of 0.025‰ was in the s ubtropica l gyres (Be rmuda and
Hawaii) an d the lowest in t he equatoria l upwelling reg ion of the
Pacific (0.015‰),with the Indian Oceandisplayingadecreaseof
0.020‰ p er year. Therefore, t he predic ted ranges in al l oceanic
reg ionso f0.2‰–0 .4‰de cr easeoverou rs tu dyperiodof16y ea rs
are too smal l to explain the 1. 8‰ average decline in t una δ13C
values.
Other factors, related to food web ordietary processes, could
alsoinfluencethetunaδ13Ctrend.Sizedifferencesinsampledtuna
through timecould introduce abias, but no consistent relationship
betwee n tuna size and δ13C and δ15N val ues or any size changes
with time among tuna species were obser ved (Figures S5and S6).
Whiledecadalshiftsinthedietofyellowfintunahavebeenrecorded
in the eas tern Pacif ic Ocean from t he 1990s to the 20 00s (Ols on
et al., 2014), the sim ilar δ13C slopes obse rved for the thre e tuna
species in our study seem inconsistent with changes in foraging
location or diet. A shift in the tuna foragingrange or timing could
also be argued to explain the observed decrease in tuna δ13C val‐
uesastherearelargespatialandintraannualvariabilitiesintheδ13C
values of phy toplankton (Magozzi et al., 2017). Similarly,if all tuna
foraged deeperonmoremesopelagicpreythathadlowerδ13C val‐
ues than su rface prey (δ13CDIC is known to decreas e with depth;
Quayetal., 2003),tunaδ13Cvalueswoulddecrease.However,the
slopeofthisdecreasewouldvaryamongspeciesgiventhatyellowfin
tunamainlyinhabitsurfacewaters while bigeye tuna mostlyforage
inmesopelagicwaters(Olsonetal.,2016).Adecreaseinthetrophic
10
|
LORRAIN et AL .
fractionationfactorεfcwouldreducetunaδ13Cvaluesthroughtime
(Table 1; Figure 4). Var ious processes in cluding a change in foo d
chain leng th, food web s tructu re, quality of f ood, or tuna met ab‐
olism (Barnes, Sweeting, Jennings, Barry, & Polunin, 20 07) could
alter εfc. A change in the overall trophic fractionation factor could
thereforeoccuratmultiplelevelsofthefoodweb,drivenornot by
changesatthebaseofthefoodweb.Toourknowledge,thereareno
dataavailableintheliteraturetoexplorethisfurther.However,while
wecannot rule out thepossibility that changes in food web struc‐
ture are negated by changes in source 15N (e.g., denitrification vs.
N2fixation;Deutschetal.,2014;Somesetal.,2010),wedidnotsee
temporalchangesintunaδ15Nvaluesatglobaloroceanbasinscales,
suggestinglittlechangeinfoodchainlengthorstructure.
Finally,ourdatasethassomelimitationsinherenttothesampling
andweacknowledgethatourmostrobustanalysisisforthePacific
Ocean tha t covers a large are a with many ind ividuals by yea r and
species( TableS1).MoredataoverbroadreachesoftheAtlanticand
IndianOceans areneededtoprovide robustestimatesofbiological
changesintheseoceans.
6 | CONCLUDING REMARKS
Weshowedthatδ13Cvaluesofmetabolicallyactivetissuesofmobile
marinepredatorslikelyreflectrecentchangesatthebaseofmarine
food webs.Wedetectedasubstantial worldwide decreaseintuna
δ13Cva luesov erth e2 000–2015p eriodwhichcanberelatedto va ri ‐
ousprocessesknowntoinfluenceoceancarboncyclingintheglobal
oceans. Ouranalysissuggests thatphytoplanktonspecies(e.g.,dia‐
toms)th atun de rgoa la rgerfract ionatio nofcar bo nd uringphotos yn ‐
thesis(andthushavehigherδ13Cvalues)havebeendecreasingover
recent dec ades or that the se phytoplank ton communities al tered
their physiologies. Whilewe cannot rule out a widespreaddecline
inphy toplankton productivity,we showedthatevena large (>15%)
decline wouldhaveasmall impact on tuna δ13C values and cannot
fullyexplaintheobservedglobaltrend.Whilerecognizingthatacon‐
comitantshift at higher levelsoffood webs (change in the trophic
fractionation factor or in tuna diet or physiology) could occur and
thatmoretunacarbonisotopedataareneededfromtheAtlanticand
IndianOceans,thepresentstudyexpandsourunderstandingofthe
mainf actor st ha taff ectth eisotopicva luesof to ppre da to rs an dp ro ‐
videsaframeworktointerpretandmodelcarboncyclingatregional
to global sc ales. New ob servatio nal or modele d data that prov ide
estimatesofperiodicchanges in marine plankton communitieswill
enab leourmo deltoprovidees tim atesoftheothercontributingfa c‐
tors. Fi nally, the framewo rk presented h ere, through t he study of
tuna car bon and nitr ogen isotope s values, co uld suppor t develop‐
mentofauseful essentialocean variable(EOV)forimplementation
within aglobal oceanobservingsystem to documentcomplex eco‐
systemchangesatregionaltoglobalscalesandoverrelativelyshort
timescales(decadestocenturies).Theuseofpredatorisotopesasan
EOVwouldcomplementregionaleffortstoacquireinsitumeasure‐
mentsofplanktonabundanceanddiversity(Miloslavichetal.,2018).
ACKNOWLEDGEMENTS
We thank the observer programs, the many observers and re‐
searche rs who collec ted the sample s in each of the ocean b asins
and the Pa cific Marine S pecimen Tis sue Bank. Thi s work contrib‐
utes to CLIOTOP WG 3‐Task team 01. N .C. was suppo rted by the
“Laboratoire d'Excellence” LabexMER (ANR‐10‐LABX‐19) and co‐
funded byagrant fromtheFrenchGovernmentundertheprogram
“Investissement sd'Avenir.”Wethank P.Quaywhokindlyprovided
datafromStationALOHA,E.A.Lawsforhelpfuldiscussionsonthe
manuscript,andthethreereviewersfortheirthoroughreviews.
CONFLICT OF INTEREST
Theauthorsdeclarenocompetinginterests.
AUTHOR CONTRIBUTIONS
A.L .,A.R., N.C.,andH.P.analyzed thedataand interpreted the re‐
sults with thehelpof B.F.A.R .andD.E.P.performedthestatistical
analysis.A .L.,N.C., and H.P.wrotethemanuscriptwiththehelpof
A.J.H.andA.R.Allauthorscontributedtoandprovidedfeedbackon
variousdraftsofthepaper.
ORCID
Anne Lorrain https://orcid.org/0000‐0002‐1289‐2072
Heidi Pethybridge https://orcid.org/0000‐0002‐7291‐5766
Nicolas Cassar https://orcid.org/0000‐0003‐0100‐3783
Aurore Receveur https://orcid.org/0000‐0003‐0675‐4172
Valérie Allain https://orcid.org/0000‐0002‐9874‐3077
Nathalie Bodin https://orcid.org/0000‐0001‐8464‐0213
Laurent Bopp https://orcid.org/0000‐0003‐4732‐4953
C. Anela Choy https://orcid.org/0000‐0002‐0305‐1159
Alistair J. Hobday https://orcid.org/0000‐0002‐3194‐8326
John M. Logan https://orcid.org/0000‐0002‐0590‐7678
Frederic Ménard https://orcid.org/0000‐0003‐1162‐660X
Christophe E. Menkes https://orcid.org/0000‐0002‐1457‐9696
Dan E. Pagendam https://orcid.org/0000‐0002‐8347‐4767
David Point https://orcid.org/0000‐0002‐5218‐7781
Andrew T. Revill https://orcid.org/0000‐0003‐2486‐5976
Christopher J. Somes https://orcid.org/0000‐0003‐2635‐7617
REFERENCES
Barnes,C., Sweeting, C. J., Jennings,S., Barr y,J. T., &Polunin ,N.V.C.
(2007).Effectoftemperatureandrationsizeoncar bonandnitrogen
stableisotopetrophicfractionation.Functional Ecology, 21, 356–362.
https ://doi.org/10 .1111/j.1365‐24 35. 20 0 6.012 24. x
Beardall, J., & Raven, J. A. (2013). Calcification and ocean acidifica‐
tion: New i nsights fr om the cocco lithoph ore Emiliania huxleyi. New
Phytologist, 199,1–3.https://doi.org/10.1111/nph.12297
|
11
LORRA IN et AL .
Behrenfeld,M.J.,O'Malley,R.T.,Siegel,D.A.,McClain,C.R.,Sarmiento,
J.L., Feldman, G . C., … B oss, E.S. (2006). Climate‐driventrendsin
contemporaryoceanproductivity.Nature, 444,752–755.http s://doi.
org /10.1038/nat ure 05317
Bidigare,R .R ., Fluegge,A., Freeman,K .H.,Hanson,K .L .,Hayes, J.M.,
Hollander,D.,…Millero,F.J. (1997).Consistentfractionationof13C
innatureandinthelaboratory:Growth‐rateeffectsinsomehapto‐
phyte algae. Global Biogeochemical Cycles, 11,279–292.https://doi.
org /10.1029/96gb0 393 9
Bird, C . S., Veríssimo, A ., Magozzi, S ., Abrantes , K. G., Ag uilar, A., Al‐
Reasi, H. , … Bouchoucha, M . (2018). A global per spective on th e
trophicgeographyofsharks.Nature Ecolog y & Evolution, 2, 299–305.
htt ps://doi.o rg /10.103 8/s41559‐ 017‐0 432‐z
Boller,A .J., Thomas,P.J.,C avanaugh,C .M.,&Scott, K.M.(2011).Low
stable carbon isotope fractionation by coccolithophore RubisCO.
Geochimica et Cosmochimica Acta, 75, 7200–7207. https://doi.
org/10.1016/j.gca.2011.08.031
Bonan, G.B.,&Doney,S.C.(2018).Climate,ecosys tems, andplanetary
futures:ThechallengetopredictlifeinEarthsystemmodels.Science,
359,eaam8328.https://doi.org/10.1126/science.aam8328
Bopp,L.,Resplandy,L.,Orr,J.C.,Doney,S.C.,Dunne, J.P.,Gehlen, M.,
…Tjiputra,J.(2013).Multiplestressors of ocean ecosystemsinthe
21st centu ry: Projections with C MIP5 models. Biogeosciences, 10,
6225–6245.ht tps://doi.org/10.5194/bg‐10 ‐6225‐2013
Boyd,P.W.,Rynearson,T.A.,Armstrong,E.A.,Fu,F.,Hayashi,K.,Hu,Z.,
…Thomas,M.K .(2013).Marine phytoplanktontemperature versus
growthresponsesfrompolartotropicalwaters–