J Appl Ecol. 2019;56:855–865. wileyonlinelibrary.com/journal/jpe | 855

Received:31August2018  |  Accepted:6November2018DOI: 10.1111/1365-2664.13327

R E S E A R C H A R T I C L E

Global reconstruction of life- history strategies: A case study using tunas

Cat Horswill1,2  | Holly K. Kindsvater3  | Maria José Juan-Jordá4,5  |  Nicholas K. Dulvy5  | Marc Mangel6,7 | Jason Matthiopoulos1

1Institute of Biodiversity, Animal Health & Comparative Medicine, University of Glasgow, Glasgow, UK; 2Department of Zoology, University of Cambridge, Cambridge, UK; 3DepartmentofEcology,Evolution,andNaturalResources,RutgersUniversity,NewBrunswick,NewJersey;4AZTI,Pasaia,Gipuzkoa,Spain;5EarthtoOceanResearchGroup,DepartmentofBiologicalSciences,SimonFraserUniversity,Burnaby,BritishColumbia,Canada;6Theoretical Ecology Group,DepartmentofBiology,UniversityofBergen,Bergen,Norwayand7InstituteofMarineSciences,DepartmentofAppliedMathematics,UniversityofCalifornia,SantaCruz,California

CorrespondenceCat HorswillEmail: catrsw@gmail.com

Funding informationNaturalEnvironmentResearchCouncil,Grant/AwardNumber:NE/P004180/1;NationalScienceFoundation,Grant/AwardNumber:DEB-1555729andDEB-1556779;MarineAllianceforScienceandTechnologyScotland,Grant/AwardNumber:SG411;MarieCurieInternationalOutgoingFellowship

Handling Editor: Robert Arlinghaus

Abstract1. Measuring the demographic parameters of exploited populations is central to pre-dictingtheirvulnerabilityandextinctionrisk.However,currentratesofpopula-tion decline and species loss greatly outpace our ability to empirically monitor all populations that are potentially threatened.

2. The scale of this problem cannot be addressed through additional data collection alone, and therefore it is a common practice to conduct population assessments based on surrogate data collected from similar species. However, this approach introduces biases and imprecisions that are difficult to quantify. Recent develop-ments in hierarchical modelling have enabled missing values to be reconstructed basedonthecorrelationsbetweenavailablelife-historydata,linkingsimilarspe-cies based on phylogeny and environmental conditions.

3. However, these methods cannot resolve life-history variability among populations or species that are closely placed spatially or taxonomically. Here, theoretically motivated constraints that align with life-history theory offer a new avenue for addressing this problem. We describe a Bayesian hierarchical approach that com-bines fragmented, multispecies and multi-population data with established life-history theory, in order to objectively determine similarity between populations based on trait correlations (life-history trade-offs) obtained from model fitting.

4. We reconstruct 59 unobserved life-history parameters for 23 populations of tuna that sustain some of the world's most valuable fisheries. Testing by cross-valida-tion across different scenarios indicated that life-histories were accurately recon-structed when information was available for other populations of the same species. The reconstruction of several traits was also accurate for species repre-sented by a single population, although credible intervals increased dramatically.

5. Synthesis and applications. The described Bayesian hierarchical method provides access to life-history traits that are difficult to measure directly and reconstructs

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, providedtheoriginalworkisproperlycited.© 2018 The Authors. Journal of Applied EcologypublishedbyJohnWiley&SonsLtdonbehalfofBritishEcologicalSociety

856  |    Journal of Applied Ecology HORSWILL et aL.

1  | INTRODUC TION

Monitoringapopulation'sstatusandriskofextinctionrequiresde-tailed knowledge of life-history characteristics, including survival,growth and reproduction (Pearson etal., 2014; Salguero-Gómezet al., 2016; Winemiller, 2005). However, our understanding of these traits is fragmented, error- prone and incomplete for many species (Myers,Mittermeier,Mittermeier, da Fonseca,&Kent, 2000). Forexample, more than 10% of the 87,967 animals, plants and fungi as-sessedbytheInternationalUnionforConservationofNature(IUCN)are classified as Data Deficient, such that their status and vulnera-bility cannotbedetermined (IUCN,2017).Consequently, identify-ing reliableways todetect declines and assess the extinction riskofdata-limitedpopulationshasbeenhighlightedasakeyecologicalproblem facing policymakers (Kindsvater etal., 2018; Sutherlandet al., 2006).

One approach of assessing data- limited populations involves judiciously filling- in missing values of life- history parameters with surrogate data from similar species or other populations of the same species. However, using surrogate parameters concurrently with ob-served life- history traits neglects the trade- offs, that is, budgetary compromises, that connect the different aspects of a population's demography, constraining the range of possible life- history strat-egies thatcanevolve (Stearns,1992).Thebiasesand imprecisionsgenerated by employing this approach are difficult to quantify, cre-ating a considerable margin for parameter misuse and population mismanagement. The process of using surrogate life- history values to overcome missing data has become increasingly formalized with the widespread availability of meta-analyses for key parameters(e.g.Denney,Jennings,&Reynolds,2002;Hutchings,Myers,Garcia,Lucifora, & Kuparinen, 2012; Thorson, Taylor, Stewart, & Punt,2014). Predictive approaches have also been developed to estimate missing values across large groups of species based on correlations between available life-history data (Freckleton& Jetz, 2009; Jetz& Freckleton, 2015; Thorson,Munch, Cope, &Gao, 2017). Thesestudies predict life- history parameters by grouping species based on phylogeny and fine- scale environmental determinants of life- history traits, such as temperature.

Commercial fisheries exploit populations at rates that greatly out-pace our ability to accurately assess them. Populations with missing life- history data are particularly problematic in this context because, by definition, the data necessary to develop evidence- based man-agement and policy action are incomplete. At present, less than 20% of the global fish catch originates from populations that are managed by formal population assessment (Costello, 2012), and therefore, threats from over- exploitation are rarely identified before species havesufferedlargepopulationdeclines(Burgess,Polasky,&Tilman,2013). Correlative approaches have been used to estimate specific life- history traits for data- limited fish species (Beverton, 1992; Hordyk,Ono,Sainsbury,Loneragan,&Prince,2014;Nadon&Ault,2016), incorporating phylogenetic and environmental determinants of traits to access large numbers of species (Thorson et al., 2017). However, phylogeny becomes less informative when considering population- level strategies, as opposed to species- level strategies, andwhenexaminingspecieswithinthesamegenus.Furthermore,populations that inhabit large geographic areas and migrate over broad latitudinal gradients preclude a straightforward assignment of fine- scale environmental determinants for life- history strategy. Predictive approaches informed by theoretically motivated con-straints that align with life- history theory (Kindsvater et al., 2018) offer a new avenue for providing informative distinctions between populations that are closely placed taxonomically and spatially.

Thesevenprincipalmarket tunas (skipjacktuna–Katsuwonus pelamis, yellowfin tuna – Thunnus albacares, bigeye tuna – Thunnus obesus, albacore tuna – Thunnus alalunga, Pacific bluefin tuna – Thunnus orientalis, Atlantic bluefin tuna – Thunnus thynnus, Southernbluefintuna–Thunnus maccoyii) are categorized into 23 highly migratory stocks, or populations, with ocean-scale distri-butions. These populations sustain some of the largest and most valuable fisheries in the world (Juan-Jordá, Mosqueira, Cooper,Freire,&Dulvy,2011),totallingasalesvalueofUS41billion(FAO,2016;MacFadyen,2016)andaccountingforapproximately9%ofglobal fish catches (Costello, 2012; FAO, 2014). However, thesepopulations have also experienced long- term declines in adult biomass since the1950s (Juan-Jordá etal., 2011), such that 43%are currently overfished with biomass levels below management

missing life-history information useful for assessing populations and species that are directly or indirectly affected by human exploitation of natural resources. The method is particularly useful for examining populations that are spatially or taxo-nomically similar, and the reconstructed life-history strategies described for the principalmarkettunashaveimmediateapplicationtotheworld-widemanagementof these fisheries.

K E Y W O R D S

Bayesian imputation, data limited, demography, fecundity, life-history theory, missing data, principalmarkettuna,Scombridae

     |  857Journal of Applied EcologyHORSWILL et aL.

recommendations (ISSF, 2018). Furthermore, population-specificlife- history data are missing for more than a third (36%) of maturity and fecundity parameters (Figure1; see Supporting InformationTableS1).Forexample, ageofmaturityandannual fecundityarepoorly resolved, and are only reported for 10 and two of the 23 populations respectively. Consequently, population assessments often rely on age of maturity estimates from other populations, and spawningstockbiomassisgenerallyusedasanindexoffecunditywhenfittingstockrecruitmentmodels(e.g.ISC,2016,Rice,Harley,Davies, & Hampton, 2014).

Here, we develop a Bayesian hierarchical approach that con-siders a multi- population, multispecies dataset and reconstructs a full life- history strategy for the 23 populations of principal market tuna that are recognized globally. We employ a multi-variate approach that determines the direction and strength of life- history trade- offs among traits in order to resolve the simi-laritybetweenspecies.Thisframeworkprovidesaccesstotraitsthat are difficult to measure directly, such as age of maturity and annual fecundity. Furthermore, by consideringmultiple sourcesof information and including error in the observed data, it also generates a measure of uncertainty associated with each param-eter. The model outlined here is applied to tunas; however, the approach could be applied to a wide range of taxa and we explore the predictive capacity of the method through a series of cross- validation experiments.

2  | MATERIAL S AND METHODS

2.1 | Life- history traits for the principal market tunas

We considered a seven trait life- history strategy for the principal markettunas.DataonsixofthesetraitswerecollatedfromJuan-Jordá,Mosqueira, Freire, Ferrer-Jordá, andDulvy (2016).As theprincipalmarket tunas are sexually dimorphic,we removed traitestimates based on males only. Traits included adult survival (Φ) derived from maximum age (Tmax; Equation 1); age of maturity (A, age that 50% of the sampled individuals are mature); spawning fre-quency (S, number of days between each batch of eggs, oocytes, being produced); spawning duration (D, proportion of the year that the population is actively spawning); absolute batch fecundity (F, mean number of oocytes produced per individual per batch); and somatic growth rate (K) from the von Bertalanffy growth function. Three additional studies on batch fecundity published after the original dataset was compiled were also included in order to bol-ster populations that were missing data across the four fecundity traits(SupportingInformationTableS1).Toresolveafullsetoffe-cundity parameters, information on the seventh life- history trait, annual fecundity per capita (V, mean number of oocytes produced per individual per year), was also included for the two populations withpublishedestimates(Figure1;SupportingInformationTable

F IGURE  1 Availability of data for each life- history trait for the seven species ofprincipalmarkettunasgroupedbypopulation,speciesandhabitat.Shadedblocksindicatethatlife-historydataareavailable in the original dataset for that population. Each column reflects a major ocean region and population; 1. Pacific Ocean, 2. Atlantic Ocean, 3. Indian Ocean, 4.MediterraneanSeaand5.SouthernOcean, where two populations are present within one region, its location is indicated abovethebar:N=north,S=south,W=westandE=east.FishpicturesfromDianeRomePeebles.Specieslistedinascending order of growth rates (k,Juan-Jordáetal.,2016).Population-specificsurvival rates are derived from maximum age (Tmax,Equation1).SupportingInformationTableS1showsnumberofstudies for each population per trait

Skipjack

Yellowfin

AlbacoreTr

opic

al s

peci

esTe

mpe

rate

spe

cies

2AtlanticOcean

4 / 5Other

1PacificOcean

3IndianOcean

W EW E

W E

W E

Bigeye

4.

5.

W E

1. Maximum age (Tmax)2. Maturity (A)3. Spawning freq. (S)4. Spawning duration (D)5. Batch fecundity (F)6. Somatic growth (K)7. Annual fecundity (V)

N S N S

Southern Bluefin

Pacific Bluefin

Atlantic Bluefin

858  |    Journal of Applied Ecology HORSWILL et aL.

S1). Finally, as broad similarities in life-history strategy can bedrawnbetweenspeciesoftunathatinhabitthesamebiome(Juan-Jordá,Mosqueira,Freire,&Dulvy,2013,2015),eachpopulationwas assigned a binary habitat variable (H) delineating tropical or temperatepreference(Figure1).Topromoteconvergenceacrosslife- history parameters with different scales (Congdon, 2003), each trait was standardized to a mean of 0 and a standard devia-tion of 1.

Rates of adult survival (Ф) were based on the following estima-tion of natural mortality (M) from maximum age (Tmax), where 4.3 is an estimated constant (Kenchington, 2014):

Mortality estimates that are based on maximum age report-edly perform better than those based on other parameters, such as body size (Then et al., 2015). Using a maximum age estimator also provided the largest sample size for rates of adult survival from the starting dataset, and allowed the estimation of this trait to be independent from metrics of body growth that would be used si-multaneouslytoimputemortalityinlateranalysis.Naturalmortal-ity is more commonly included as a demographic trait in fisheries assessment, compared to survival. However, we include the con-verted value of adult survival in order to demonstrate the wider application of this method to alternative systems where viability assessments are based on this trait. Varying amounts of life- history informationwereavailable foreachpopulation (Juan-Jordáetal.,2013;Figure1;SupportingInformationTableS1),andfewstudiesprovided information on more than one or two life- history traits. Thus, examination of trade- offs at the individual- study level was not possible.

2.2 | Model of life- history traits

The Bayesian hierarchical model comprised a demographic com-ponent with parameters describing species- specific relationships between life- history traits, and an observation component that integrated observed data for each population with trait- specific uncertainty associated with measurement error and process vari-ability in the data (e.g. spatial and temporal variation). The code anddata file for themodelareprovided inAppendicesS1andS2oftheSupportingInformation,andkeymodelparametersarelistedinSupporting InformationTableS2.Allmodelswere implementedinJAGS (v.4.3.0)via the“jagsUI” library (v 1.4.9) for program r (v. 3.4.1).Models were fitted by running threeMonte CarloMarkovchains (MCMC) for 1.5 × 107 iterations and retaining every 200th step in order to increase the effective MCMC sample size for the same amount of computer memory. The first 5,000 MCMC draws were removed as burn- in, and each chain was initialized at differ-ent points in the parameter space. Convergence of the chains was confirmedusingtheBrooks–Gelman–Rubindiagnostictool(allval-ues r̂≤1.01)andtheeffectivesamplesizeforeachparameter(werequired this to be >500).

The demographic component of the model (Equation 2) con-tained a function for each life-history trait. Each trait was modelled in relation to a binary habitat term (H) delineating tropical and tem-perate species, and two random- effect terms (γ and ω) that repre-sent the correlated residuals (i.e. trade- offs, Charnov, 1993) between traits at the species-and population-level, respectively:

Here, subscripts i and j denote population and species re-spectively. Lower case symbols (ϕ, a, s, d, f, v, k) represent sim-ulated life- history traits that are modelled on the linear scale; that is, logit for survival and spawning duration, log for all other variables. The trait- specific coefficients for the effect of habitat (Equation 2: ψϕ, ψa, ψs, ψd, ψf, ψv, ψk) were assigned from normal priors centred on zero: N(0,0.1). Initial model development also included replacing somatic growth with asymptotic body size (L∞) from the von Bertalanffy growth function, as well as maximum ob-served body size (Lmax). All populations had available data points for somatic growth rate, asymptotic body size and maximum ob-servedbodysize.TheBrooks–Gelman–Rubindiagnostictoolandthe effective sample size of model parameters indicated poorer mixing of the MCMC in these latter models, compared to a model including somatic growth (Supporting InformationAppendixS3).Consequently, all methods and results relate to the model that included growth rate.

To capture the trade- offs between the seven life- history traits at the between- and within- species level, we modelled two random- effect terms (γ and ω respectively) using multivariate nor-mal distributions. The covariance structures (Σi and Σj: Equation 3) parameterizing these distributions allow the correlations, or life- history trade- offs, that connect the different aspects of a popula-tion’s demography to emerge during model fitting:

The mean of the multivariate normal distributions (μ, Equation 3) hadnormalpriors.Standardizingtheinputdatatoameanofzeroandastandard deviation of one meant that all traits were able to receive priors centredonzero.Forthespecies-level term(γ), being centred on zero reflectsthestandardizedtraitmean.Forthepopulation-levelterm(ω), being centred on zero allows the population mean to operate as a classic random effect. The standard deviation in the species- level priors was set to 0.25 (i.e. a precision of 15) to allow flexibility, and this was reduced to a standard deviation of 0.14 for the population-level (i.e. a precision of 50). The variance–covariance matrices in the multivariate normal

(1)M=4.3

Tmax

,!=e−M

(2)

!i,j="!Hi+#!,j+$!,i

ai,j="aHi+#a,j+$a,i

si,j="sHi+#s,j+$s,i

di,j="dHi+#d,j+$d,i

fi,j="fHi+#f,j+$f,i

vi,j="vHi+#v,j+$v,i

ki,j="kHi+#k,j+$k,i

(3)(

!i)

∼MVN("1,Σi), "1= ("1,#,"1,a,"1,s,"1,d,"1,f,"1,v,"1,k),"1,∗ ∼N(0,15)(

$j

)

∼MVN(

"2,Σj

)

, "2= ("2,#,"2,a,"2,s,"2,d,"2,f,"2,v,"2,k),"2,∗ ∼N(0,50)

     |  859Journal of Applied EcologyHORSWILL et aL.

distributions (Σi and Σj, Equation 3) were assigned inverse Wishart prior distributions with an identity matrix for the scale matrix (Ω) and eight degrees of freedom (df):

This equation is repeated in the model, such that c is either spe-cies (i) or population ( j). Setting thedf parameter for the Wishart prior to one more than the dimensions of the variance–covariance matrix (n =numberoftraits)achievesauniformpriordistributionontheindividualcorrelationparameters;thatis,anequallylikelyprob-abilitybetween−1and1(Gelman&Hill,2006).Ascalingparameter(ξ) assigned from a uniform prior: U(0,5), was also incorporated to overcome constraints on the scale parameters associated with an inverse- Wishart model (Gelman & Hill, 2006).

Observed life- history data for each trait (Φ, A, S, D, F, V, K) were incorporated through an observation component that reflected a combination of measurement error and process variability in the data (e.g. spatial and temporal variation). Upper case notation is used to denote observed, population- specific data. Data were included at the study level (p) for each species (i) and population ( j) :

Variance was assumed to be Gaussian, where the recon-structed population values (ϕi,j, ai,j, si,j, di,j, fi,j, vi,j, ki,j)weretakenfromEquation 2, and the precision (τϕ, τa, τs, τd, τf, τv, τk) was set from a normal distribution centred on 20: N(20,0.1). This assumes a rea-sonably high precision on the reconstructed value for data- limited traits; however, for data- rich traits, this prior distribution will allow the precision to travel into the tails, that is, become high or low in line with the observed variance in the data. Thus, in addition to im-puting missing values, posterior distributions for the observed data points werealsoobtained.Thisacknowledges thatevenobserved life-historydata are a sample from a theoretical population of possible trait- specific values whose mean and credible intervals can be obtained. Although this exercise is similar to the traditional approach of sampling distributions, the joined inference across all data prevents the resulting parameter posteri-ors from being restricted by small sample sizes.

2.3 | Model validation

We cross- validated the predictive capacity of the life- history model by examining whether the model was able to accurately predict the observed life- history from a single, readily available trait: somatic growth. Model performance was examined under two scenarios: (a)

when other populations of that species have life- history data, and (b)whenthespeciesisrepresentedbyasinglepopulation.Forthefirst scenario, we removed the life- history data for a population with data for all seven life- history traits, retaining somatic growth rate only; here, we chose the southern Pacific population of albacore tuna. This species is represented by five additional populations in the model (Figure1;Supporting InformationTableS1).For thesecondscenario, we removed the life- history data for a species represented by a single population, retaining somatic growth rate only; here, we choseSouthernbluefintuna(Figure1;SupportingInformationTableS1).Wealsorananadditionalsixmodelsimulationstoexaminehowthepredictedlife-historyofSouthernbluefintunachangedwiththeaddition of each life- history trait, alongside growth rate.

The significance of the habitat term in the demographic functions was examined based on the posterior credible interval of the coef-ficient parameters (Equation 2: ψϕ, ψa, ψs, ψd, ψf, ψv, ψk), as well as a comparison of the imputed traits and model fit obtained from the full model versus one where insignificant habitat terms were removed. Parameters where the 97.5 posterior credible interval spanned zero weredeemed insignificant.Formanyspecies,especially fish, infor-mationonfecunditymaynotbeknown.Therefore,wealsoexaminedthe dependence of the model on the inclusion of the four fecundity metrics by running the model with a reduced set of traits: survival, age of maturity and somatic growth. We compared the imputed traits and associated credible intervals from this model with the output from the full model that also included metrics of fecundity.

3  | RESULTS

3.1 | Hierarchical modelling of life- history trade- offs

The vectors of predicted, or reconstructed, life- history traits (ϕ,a,s,d,f,v,k,Equation2)arepresented inFigures2and3.Pearsoncorrelation coefficients between traits were estimated using median posterior values. Correlations between life- history traits became, on average, 14% stronger following reconstruction (Figure2), and alloriginal data points remained within the 95% credible intervals of the reconstructed values. Correlations between all traits and sur-vival, age of maturity and somatic growth were consistently strong (−0.3≥ρ≥0.3, Figure2B), with the exception of age of maturityand annual fecundity (ρ=0,Figure2B–I).Correlationsbetweenthefour fecundity traits were predominantly weaker (−0.2≤ρ≤ 0.2, Figure2B–f,i,m,n),withtheexceptionofannualfecundityandbatchfecundity (ρ=−0.7,Figure2B–o),andbatchfecundityandspawningduration (ρ=−0.4,Figure2B–j).

Thereconstructedlife-historystrategies(SupportingInformationTableS3)of theprincipalmarket tunas appeared along a thermal-growth gradient, whereby tropical species had faster rates of so-matic growth compared to species that spend a large portion of their annualcycle intemperatewaters (Figure2). Inaddition, therewasa gradual increase in survival rates from the fast- growing tropical species to the slower growing temperate species (Figure3A). The

(4)

Ωc= In

Qc∼InvWishart(Ωc,n+1)

Σc=Diag(!c)QcDiag(!c)

(5)

logit(!p,i,j)∼N("i,j,#")

log(Ap,i,j)∼N(ai,j,#a)

log(Sp,i,j)∼N(si,j,#s)

logit(Dp,i,j)∼N(di,j,#d)

log(Fp,i,j)∼N(fi,j,#f)

log (Vp,i,j)∼N(vi,j,#v)

log(Kp,i,j)∼N(ki,j,#k)

860  |    Journal of Applied Ecology HORSWILL et aL.

tropical species also matured earlier compared to the temperate species and returned predominantly larger credible intervals for annual fecundity (Figure3B and F). The relationshipwithmatura-tion was further demonstrated by the coefficient of the relation-ship with biome (97.5 credible interval for ψa,Equation2:−2.000to−0.003:Figure3B).Basedonthemedianvalues,spawningduration

was longer in tropical species (Figure3D). In contrast, spawningfrequency and batch fecundity did not clearly differ between the twohabitatgroupings (Figure3CandE),althoughbatchfecundityincreased and was more variable for the three slow- growing bluefin species(Figure3E).

Large intra- population variation in the raw data for spawning durationwasnot reduced throughmodel fitting (seeSupportingInformationFigureS1).Furthermore,thelargedisparitybetweenthe mean published values of maturation for the two populations of Atlantic bluefin tuna was maintained (Figure3B, SupportingInformation Figure S1). Intra-species variability appeared as dif-ferent median estimates of traits within species, for example, in survival(Figure3A)andsomaticgrowth(Figure3G),aswellasdif-ferences in the credible interval for each trait, for example, in age ofmaturity (Figure3B),spawningduration (Figure3D),spawningfrequency(Figure3C)andannualfecundity(Figure3F;SupportingInformationTableS3).Withinthethreebluefintunaspecies,thebreeding strategy of Pacific bluefin tuna appeared more distinct, comparedtotheothertwospecies(Figure3).

3.2 | Validation testing

Life- history traits were recreated convincingly for a popula-tion that was represented by multiple other populations of the same species (south Pacific albacore tuna), with the exception of spawning duration (Figure4A). For all other traits, median pos-terior values were similar to the original data as well as the val-ues reconstructed by the model that included all available data. Furthermore, traits were returned within tight credible inter-vals. For a species that was represented by a single population (Southernbluefintuna),themodelconvincinglyrecreatedtheme-dian values for survival, spawning frequency and annual fecundity (Figure4B).However,maturitywasimputedtobelowerthantheoriginal data and the median value reconstructed by a model that incorporatedallavailabledata.Furthermore,thecredibleintervalsof reconstructed traits increased considerably, compared to the model including all available data (Supporting Information TableS4).Inferencewasonlyimprovedbyrefittingthemodelwiththerespectivelife-historydataincorporated(SupportingInformationFigureS2).

Examination of the coefficient parameters for the effect of habitat preference demonstrated that age of maturity was the only trait with a step difference between biomes, as opposed to a more continual change from the fastest growing tropical species to the slowest growing temperate species. Removing the remaining habitat terms from the demographic functions generated marginal changes to the posterior credible intervals for annual fecundity and the median values for spawningduration. Furthermore, the remainingtraitswere largely unchanged (Supporting Information Figure S3).However, the overall fit of a model without a complete set of habitat terms was reduced compared to a model including these terms (min-imumeffectivesampleformodelwithouthabitatterms=2,604,allr̂≤1.13;minimumeffectivesamplesizeformodelincludinghabitat

F IGURE  2 The strength of trade- offs between a seven trait life-historystrategyforthesevenspeciesofprincipalmarkettunas.(A) Original data (based on mean values) and (B) reconstructed data (median posterior values), both on the scale of the linear predictor. Speciesaredemarkedusingcolour.Colourrampcorrespondstothe observed gradient in the observed data for somatic growth rate (Juan-Jordáetal.,2016),fromfasttoslow,withinthetropicalandtemperate groupings. Pearson correlation coefficients between traits shown in the top left hand corner of each panel, not shown for the original dataset of annual fecundity due to sample size. Units: age of maturity (years), spawning frequency (days), spawning duration (proportion of year), batch fecundity (number of oocytes per individual per batch), annual fecundity (number of ooctyes)

     |  861Journal of Applied EcologyHORSWILL et aL.

terms=3,764,all r̂≤1.01).Removingthefecunditytraitsfromtheinference framework resulted inamarginal change in thecredibleintervals of the reconstructed life- history parameters (less than 5% for themajority of populations,max. 7%; Supporting InformationTable S5). The imputedmedian life-history traitswere also highlycorrelated with those estimated by the model that included these additionalfourmetricsoffecundity(SupportingInformationFigureS4).

4  | DISCUSSION

In this study, we applied a formal hierarchical Bayesian approach that connects different life- history traits in order to quantify trade- offs and synergies (i.e. negative and positive correlations) and pre-dict missing values. This approach provides potentially more precise estimates of life- history traits for use in population assessments, compared to surrogate data assigned from similar species or other populations of the same species. We compiled our starting dataset

from an extensive repository of life- history studies on scombrids (Juan-Jordá etal., 2016) that offers a more exhaustive coverageof available data sources than other global sources of life- history parameters, such as FishBase.We apply the approach to the 23populationsofprincipalmarket tuna recognizedglobally.As theseare closely related species that inhabit large geographic areas and migrate over broad latitudinal gradients, a straightforward assign-ment of missing life- history information using phylogenetic or envi-ronmental determinants (e.g. Thorson et al., 2017) is insufficient to resolve life- history differences among populations. The seven- trait life- history strategies reconstructed here provide insights into the variationwithin- and between-species. By taking some degree ofobservation error and spatio- temporal variation into account during model fitting, we also quantify the uncertainty associated with each reconstructed value. Propagating this uncertainty into projections of population change allows realistic estimates of extinction and over- exploitationrisktobequantified.

Our framework performedwell at reconstructingmissing life-history values when information was available for other populations

F IGURE  3 Reconstructed life- history traits by habitat, species and population. Median values are shown as points and 95% credible intervals are shown as bars. Blue indicates parameters with data available (all observed data points fell within the credible intervals), orange indicatesparameterswithdatamissingintheoriginaldataset.Speciesandpopulationslistedintheobservedorderofsomaticgrowthrate(Juan-Jordáetal.,2016),fromfasttoslow,withinthetropicalandtemperategroupings.Somaticgrowthisadjustedtoincludepopulation-specific residuals. Units: age of maturity (years), spawning frequency and spawning duration (days), batch fecundity (number of oocytes per individual per batch, ×107), annual fecundity (number of ooctyes, ×108)

862  |    Journal of Applied Ecology HORSWILL et aL.

of the same species. Furthermore, inference of survival and ageof maturity were not degraded by running the model without the four fecundity metrics that are often more difficult to measure in batch spawning incomebreeders, like tunas. Themarginal changein model fit created by adding (or removing) the fecundity metrics is likelytoreflectthelargegapsinthesedatasets,suchthattheother

life- history traits have a stronger influence on resolving fecundity than vice versa. Inference was weaker when predicting the life-historytraitsofasingle-populationspecies(Southernbluefintuna).Forthisspecies, themedianposteriorvalueswererelativelyaccu-rate for traits where the original data appeared close to the inter- species linear regression with growth rate (e.g. survival and annual fecundity). However, the observed values of maturation and batch fecundity are both higher than expected for its growth rate, that is, these values have large residuals from the linear regression, thus re-ducingpredictivecapacity(Figure3BandE).Furthermore,Southernbluefin tuna exists at the edge of the parameter space where the likelihoodsurfacemaybelesswelldefined.Thisislikelytohavecon-tributed to the wider credible intervals resulting from this aspect of the model validation exercise.

The direction and strength of the evolutionary correlations that constrain the life- history strategies emerged during model fitting. In agreement with theory, several key life-history traits (somaticgrowth, survival and maturity) appeared along a gradient. Here, a slow life- history strategy was characterized by slower rates of body growth, higher rates of adult survival and higher ages of matu-rity, while a fast life- history strategy included faster rates of body growth, lower rates of survival and earlier maturation (Figure2;Stearns,1992).Previouslyundefinedlife-historytrade-offsbetweenthe different aspects of reproduction for batch- spawning income breeders were also inferred through model fitting. In general, a slow life- history strategy was associated with shorter spawning dura-tions and higher batch fecundities, while a fast life- history strategy had longer spawning durations and lower batch fecundities. Colder habitats select for slower life histories in fishes, including shorter spawning seasons, becauseoffspring survival is strongly linked toseasonal food availability, for example spring algal blooms (Lowerre- Barbieri,Ganias,Saborido-Rey,Murua,&Hunter,2011).However,the described trade- offs between reproductive traits may also imply some directional selection between spawning duration and batch fe-cundities, whereby temperate species can produce large batches of oocytes because the energetic cost is temporally restricted. Variable strategies within the slow- growing bluefin tunas also indicate pop-ulation- or species-specific trade-offs. For example, Pacific blue-fin tuna returned elevated median values for batch fecundity and spawning frequency and lower values for survival and age of matu-rity, relative to the other bluefin species. This result suggests that it is not possible for these large, slow- growing tunas to increase their reproductive output without trading- off their survival rates.

Modelling the full set of life- history traits across a species’ range allows differences between populations to be examined. Clear dif-ferences among populations, as well as species, were observed in so-maticgrowth,spawningfrequencyandspawningduration(Figure3C,D, G) that may reflect diverging strategies associated with habitat or lifestyle.InagreementwithJuan-Jordáetal.(2013,2015),broadsim-ilarities in survival, maturation and spawning duration could be drawn based on whether the species predominantly inhabits tropical or tem-perate waters during its annual cycle. The slower life- history strate-giesof temperatespecies,especiallybluefin tunas, is likely tobea

F IGURE  4 Predicting the life- history information for (A) the SouthernPacificpopulationofalbacoretuna,(B)Southernbluefintuna.Two-dimensionalkerneldensityplotscorrelatethefullposterior distribution for each life- history trait combination. Colour ramp corresponds to the probability associated with values: highest probability in yellow and lowest probability in blue. Original data points are shown as red circles and red dashed lines, and the mean imputed values from the model including all available data are shownasblacktrianglesandblackdashedlines.Dataavailabilityforeachtraitareshownasstacksofblocks(seeFigure1);shadedblocksindicatethatlife-historydatawereavailableforthatpopulation, those shaded grey represent data removed and imputed inthevalidationtest.TunapicturesfromDianeRomePeebles.SeeFigure3fortraitunits

A

B

Albacore

Southernbluefin

     |  863Journal of Applied EcologyHORSWILL et aL.

contributingfactorintheiroverexploitation(Juan-Jordáetal.,2015).Tropical species also returned predominantly larger credible inter-vals associated with annual fecundity, compared to the temperate species. However, the median values appeared to be similar across the two habitat groupings, despite temperate species spawning for a shorter period compared to tropical species. The smallest credible intervals for annual fecundity were obtained for the two temperate populations with observations for this trait. Therefore, this apparent trend is likely to reflect data availability, such thatmore empiricalestimates will improve parameter estimation of annual fecundity, as well as the resolution of associated traits and trade- offs. Complex breeding strategies mean that estimates of annual fecundity are par-ticularlydifficulttoobtainforthesespecies(Farley,Williams,Hoyle,Davies,&Nicol,2013).Theimputedcredibleintervalsthusprovideaunique insight into breeding strategies.

By reconstructing missing trait values based on the life- history cor-relations that connect different characteristics, we imputed values that are potentially more precise single estimates than the measured data. Thisislikelytobethecasefortraitsthataredifficulttoresolveorthatdemonstrate large intra- population variability, such as batch fecundity and spawning duration, respectively. However, for traits that are reliably resolved, such as growth rate, there is an argument for not using recon-structed values. The life- history trait values reconstructed in this study largely reflect the population-level mean value for the available data, indi-catingthatshrinkagetowardsthespecies-levelmeanduringmodelfittingis limited(SupportingInformationFigureS1).Bylimitingthenumberofcovariates in the demographic functions (Equation 2), model fitting is pre-dominantly driven by the original data and the variance–covariance matrix connecting traits. Thus, divergent traits within species were maintained, for example age of maturity for the two populations of Atlantic bluefin tuna. There is a current debate about this reported divergence (ICCAT, 2017). The previously published age of maturity for the eastern population may be an underestimate because samples outside the Mediterranean are not included in the original analyses, increasing the proportion of mature fish from the resident population (ICCAT, 2014). Thus, the inclusion of fu-ture data as model input may change the inference of this trait.

The described approach for imputing missing life- history data could potentially be applied to any data- limited group of species. Within a fisheries context, the reconstructed life- history data can beemployed indata-limitedassessments,suchasecological riskassessments and length-based models (Hordyk, Ono, Prince, &Walters, 2016; Rudd & Thorson, 2018). In addition, the approach has immediate application to the world- wide management of fish-eries for tuna inwhich the steepness of the Stock RecruitmentRelationship(SRR)isused.WeillustratetheideasfortheBeverton-Holt SRR (BH-SRR), butour results apply tootherSRRs aswell.WiththeBH-SRR,wepredictthenumberofrecruits,thatis,indi-viduals added to the population in the youngest age class (R), as a function of the spawning biomass (Bs), which we write as R(Bs)=

!Bs1+"Bs

where λ and β are parameters. The former is a measure

of the maximum per capita productivity, that is, related to annual fecundity, and the latter a measure of the strength of density

dependence.Steepness(h) is defined to be the reproduction when spawning biomass is 20% of its unfished level (B0) relative to re-

production at the unfished level; h= R(0.2B0)

R(B0) (see Mangel et al.,

2010, 2013 for reviews). For the BH-SRR, steepness is (Mangelet al., 2010, 2013):

where W̄ is the average spawning biomass of a female that has sur-vived to age a (Φ(a)), and the probability of being mature at age a (pm(a)) and mass at age a (W(a)) according to W̄=

∑

a

"(a)pm (a)W(a). Our

methodology provides values for survival and maturation and can easily be adapted to include mass as an additional life- history trait.

ApplicationofEquation6todeterminesteepnessrequiresknow-ing λ, which is the product of mass- specific fecundity δ (eggs, or larvae for those species with live birth, per unit spawning biomass) and sur-vival from the egg or larval stage until recruitment to the population ϕEL, thus, λ=δϕEL.Survivaltorecruitmentintothepopulationcanbemodelled using well- established allometries between size and mortal-ity(e.g.Brodziak,Mangel,&Sun,2015;Kai&Fujinami,2018;Mangel,Brodziak,&DiNardo,2010;Simonetal.,2012),andthedescribedhi-erarchical model of life- history traits allows the computation of total fecundity.IntheSupportingInformationAppendixS4,weshowhowto obtain mass- specific fecundity from total fecundity under appro-priateassumptions.Thus,oncemass-specificfecundityisknown,theparameter λcanbedeterminedforadata-poorstock.

Our analyses reveal that by considering multiple life- history traits, populations and species simultaneously in a Bayesian framework, it is possible to statistically maximize the utility ofsparse datasets, quantify the trade- offs that connect different aspects of an organism's life- history and access traits that are difficult to measure empirically. The percentage of missing life- history data will inevitably be higher for other, commercially less important groups of fishes, such that a Bayesian hierarchical ap-proach informed by the principles of life- history evolution, includ-ing metabolic theory, could further assist in population- specific trait reconstruction.

ACKNOWLEDG MENTS

ThisworkwaspartiallyfundedbyUKNaturalEnvironmentalResearchCouncil grantNE/P004180/1 to CH and JM,MarineAlliance forScienceandTechnologyScotlandgrantSG411toCH;USNationalScienceFoundation grantsDEB-1556779 toHKK,DEB-1555729toMM.NKDwassupportedbytheNaturalScienceandEngineeringResearchCouncilandCanadaResearchChairsProgram,andMJJJbyaMarieCurieInternationalOutgoingFellowshipwithinthe7thEuropeanCommunityFrameworkProgrammeFP7-PEOPLE-2013-IOF. Thanks are also extended to Haritz Arrizabalaga for discus-sions regarding the maturation of Atlantic bluefin tuna, and to Olaf Jensen,JonBrodziak,JamesThorsonandtwoanonymousreviewsfor providing comments on an earlier version of this manuscript.

(6)h=!W̄

4+!W̄

864  |    Journal of Applied Ecology HORSWILL et aL.

AUTHORS’ CONTRIBUTIONS

All authors contributed to the design and planning of this study; M.J.J.J. collated the scombrid life-historydataset;C.H. conductedall analyses and wrote the first draft of the manuscript, and all au-thors contributed substantially to revisions and gave final approval for publication.

DATA ACCE SSIBILIT Y

The tuna dataset was compiled from https://doi.org/10.1890/15-1301.1(Juan-Jordáetal.,2016),supplementedwithindividualvaluesfrom https://doi.org/10.1016/j.seares.2012.08.005 (Aranda, Medina, Santos, Abascal, & Galaz, 2013); https://doi.org/10.1371/journal.pone.0060577(Farleyetal.,2013);https://www.iccat.int/Documents/CVSP/CV068_2012/n_2/CV068020387.pdf (ICCAT 2012); https://doi.org/10.1111/j.1095-8649.2002.tb02398.x (Medina, Abascal, Megina, & Garcia, 2002) and https://doi.org/10.7755/fb.111.3.4 (Zudaire, Murua, Grande, & Bodin, 2013). The data read- in file for the life-historymodel are available in Supporting InformationAppendixS2.

ORCID

Cat Horswill https://orcid.org/0000-0002-1795-0753

Holly K. Kindsvater https://orcid.org/0000-0001-7580-4095

Maria José Juan-Jordá https://orcid.org/0000-0002-4586-2400

Nicholas K. Dulvy https://orcid.org/0000-0002-4295-9725

Jason Matthiopoulos https://orcid.org/0000-0003-3639-8172

R E FE R E N CE S

Aranda, G., Medina, A., Santos, A., Abascal, F. J., & Galaz, T. (2013).Evaluation of Atlantic bluefin tuna reproductive potential in the westernMediterranean Sea. Journal of Sea Research, 76, 154–160. https://doi.org/10.1016/j.seares.2012.08.005

Beverton,R.J.H.(1992).Patternsofreproductivestrategyparametersin some marine teleost fishes. Journal of Fish Biology, 41, 137–160. https://doi.org/10.1111/j.1095-8649.1992.tb03875.x

Brodziak, J., Mangel, M., & Sun, C.-L. (2015). Stock-recruitment resil-ience of North Pacific striped marlin based on reproductive ecol-ogy. Fisheries Research, 166, 140–150. https://doi.org/10.1016/j.fishres.2014.08.008

Burgess,M.G.,Polasky,S.,&Tilman,D.(2013).Predictingoverfishingand extinction threats in multispecies fisheries. Proceedings of the National Academy of Sciences of the United States of America, 110(40), 15943–15948. https://doi.org/10.1073/pnas.1314472110

Charnov, E. L. (1993). Life history invariants: Some explorations of symmetry in evolutionary ecology. Oxford, UK: Oxford University Press.

Congdon, P. (2003). Applied Bayesian modelling. New York, NY: JohnWileyandSons.https://doi.org/10.1002/0470867159

Costello,C.(2012).Statusandsolutionsfortheworld'sunassessedfisher-ies. Science, 338, 517–520. https://doi.org/10.1126/science.1223389

Denney,N.H., Jennings, S.,&Reynolds, J.D. (2002). Life-history cor-relates of maximum population growth rates in marine fishes. Proceedings of the Royal Society B: Biological Sciences, 269(1506), 2229–2237. https://doi.org/10.1098/rspb.2002.2138

FAO(2014).The state of world fisheries and aquaculture: Opportunities and challenges. Rome, Italy: Food and Agriculture Organization of theUnitedNations.

FAO (2016).The state of world fisheries and aquaculture: Contributing to food security and nutrition for all.Rome, Italy:FoodandAgricultureOrganizationoftheUnitedNations.

Farley,J.H.,Williams,A.J.,Hoyle,S.D.,Davies,C.R.,&Nicol,S.J.(2013).Reproductive dynamics and potential annual fecundity of SouthPacific albacore tuna (Thunnus alalunga). PLoS ONE, 8(4), e60577. https://doi.org/10.1371/journal.pone.0060577

Freckleton,R.P.,&Jetz,W.(2009).Spaceversusphylogeny:Disentanglingphylogenetic and spatial signals in comparative data. Proceedings of the Royal Society B: Biological Sciences, 276(1654), 21–30. https://doi.org/10.1098/rspb.2008.0905

Gelman,A.,&Hill,J.(2006).Data analysis using regression and multilevel/hierarchical models. Cambridge, UK: Cambridge University Press. https://doi.org/10.1017/CBO9780511790942

Hordyk,A.R.,Ono,K., Prince, J.D.,&Walters,C. J. (2016).A simplelength- structured model based on life history ratios and incor-porating size- dependent selectivity: Application to spawning po-tential ratios for data-poor stocks. Canadian Journal of Fisheries and Aquatic Sciences, 73(12), 1787–1799. https://doi.org/10.1139/cjfas-2015-0422

Hordyk, A.,Ono, K., Sainsbury, K., Loneragan,N., & Prince, J. (2014).Someexplorationsofthelifehistoryratiostodescribelengthcom-position, spawning- per- recruit, and the spawning potential ratio. ICES Journal of Marine Science, 72, 204–216.

Hutchings,J.A.,Myers,R.A.,Garcia,V.B.,Lucifora,L.O.,&Kuparinen,A. (2012). Life-history correlates of extinction risk and recoverypotential. Ecological Applications, 22(4), 1061–1067. https://doi.org/10.1890/11-1313.1

ICCAT. (2012). Report of the 2011 ICCAT South Atlantic andMediterranean albacore stock assessment sessions. Col. Vol. Sci.Pap. ICCAT 68:387–497. Retrieved from https://www.iccat.int/Documents/CVSP/CV068_2012/n_2/CV068020387.pdf https://doi.org/10.1073/pnas.1506215112

ICCAT. (2014). Report of the 2013 bluefin meeting on biological param-etersreview.Col.Vol.Sci.Pap.ICCAT,1–75.Retrievedfromhttps://www.iccat.int/Documents/Meetings/Docs/2013-BFT_BIO_ENG.pdf

ICCAT. (2017). Report of the 2017 ICCAT bluefin data preparatory meeting. Retrieved from https://iccat.int/Documents/Meetings/Docs/2017_BFT_DATA_PREP_ENG.pdf

ISC.(2016).Pacific bluefin tuna Stock Assessment 2016 (Vol. 6).Sapporo,Japan: International Scientific Committee for Tuna and Tuna-LikeSpeciesintheNorthPacificOcean.

ISSF.(2018).Statusoftheworldfisheriesfortuna.ISSFTechnicalReport2018-02. Washington, DC: International Seafood SustainabilityFoundation.

IUCN.(2017).TheIUCNredlistofthreatenedspecies.Version2017-2.RetrievedSeptember1,2017,fromhttp://www.iucnredlist.org

Jetz,W.,& Freckleton, R. P. (2015). Towards a general framework forpredicting threat status of data- deficient species from phylogenetic, spatial and environmental information. Philosophical Transactions of the Royal Society B: Biological Sciences, 370(1662), 20140016. https://doi.org/10.1098/rstb.2014.0016

Juan-Jordá,M. J., Mosqueira, I., Cooper, A. B., Freire, J., & Dulvy, N.K. (2011). Global population trajectories of tunas and their rela-tives. Proceedings of the National Academy of Sciences of the United States of America, 108(51), 20650–20655. https://doi.org/10.1073/pnas.1107743108

Juan-Jordá,M. J.,Mosqueira, I., Freire, J., & Dulvy, N. K. (2013). Theconservationandmanagementoftunasandtheirrelatives:Settinglife history research priorities. PLoS ONE, 8(8), e70405. https://doi.org/10.1371/journal.pone.0070405

     |  865Journal of Applied EcologyHORSWILL et aL.

Juan-Jordá, M. J., Mosqueira, I., Freire, J., & Dulvy, N. K. (2015).Population declines of tuna and relatives depend on their speed of life. Proceedings of the Royal Society B: Biological Sciences, 282, 20150322. https://doi.org/10.1098/rspb.2015.0322

Juan-Jordá,M.J.,Mosqueira, I.,Freire, J.,Ferrer-Jordá,E.,&Dulvy,N.K. (2016). Global scombrid life history data set. Ecology, 97(3), 809. https://doi.org/10.1890/15-1301.1

Kai,M.,&Fujinami,Y.(2018).Stock-recruitmentrelationshipsinelasmo-branchs:ApplicationtotheNorthPacificblueshark.Fisheries Research, 200, 104–115. https://doi.org/10.1016/j.fishres.2017.10.025

Kenchington,T.J.(2014).Naturalmortalityestimatorsforinformation-limited fisheries. Fish and Fisheries, 15(4), 533–562. https://doi.org/10.1111/faf.12027

Kindsvater,H.K.,Dulvy,N.K.,Horswill,C.,Juan-Jordá,M.J.,Mangel,M.,&Matthiopoulos,J.(2018).Overcomingthedatacrisisinbiodiversityconservation. Trends in Ecology and Evolution, 33, 676–688. https://doi.org/10.1016/j.tree.2018.06.004 https://doi.org/10.1016/j.tree. 2018.06.004

Lowerre-Barbieri,S.K.,Ganias,K.,Saborido-Rey,F.,Murua,H.,&Hunter,J.R. (2011).Reproductive timing inmarine fishes:Variability, tem-poral scales, and methods. Marine and Coastal Fisheries: Dynamics, Management, and Ecosystem Science, 3, 71–91. https://doi.org/10.1080/19425120.2011.556932

MacFadyen,G.(2016).Estimates of global sales values from tuna fisheries: Study for phase 3 report. Lymington, UK: Poseidon Aquatic Resource Management Ltd.

Mangel,M.,Brodziak,J.,&DiNardo,G.(2010).Reproductiveecologyandscientific inference of steepness: A fundamental metric of popula-tion dynamics and strategic fisheries management. Fish and Fisheries, 11(1), 89–104. https://doi.org/10.1111/j.1467-2979.2009.00345.x

Mangel,M.,MacCall,A.D.,Brodziak,J.,Dick,E.,Forrest,R.E.,Pourzard,R.,&Ralston,S.(2013).Aperspectiveonsteepness,referencepoints,and stock assessment. Canadian Journal of Fisheries and Aquatic Sciences, 940, 930–940. https://doi.org/10.1139/cjfas-2012-0372

Medina,A.,Abascal,F.J.,Megina,C.,&Garcia,A.(2002).Stereologicalassessment of the reproductive status of female Atlantic northern bluefin tuna during migration to Mediterranean spawning grounds throughtheStraitofGibraltar.Journal of Fish Biology, 60, 203–217. https://doi.org/10.1111/j.1095-8649.2002.tb02398.x

Myers, N., Mittermeier, R. A., Mittermeier, C. G., da Fonseca,G. A. B., & Kent, J. (2000). Biodiversity hotspots for conser-vation priorities. Nature, 403(6772), 853–858. https://doi.org/10.1038/35002501

Nadon,M.O.,&Ault,J.S.(2016).Astepwisestochasticsimulationap-proach to estimate life history parameters for data- poor fisheries. Canadian Journal of Fisheries and Aquatic Sciences, 73(12), 1874–1884. https://doi.org/10.1139/cjfas-2015-0303

Pearson,R.G.,Stanton,J.C.,Shoemaker,K.T.,Aiello-Lammens,M.E.,Ersts,P. J.,Horning,N.,…Akçakaya,H.R. (2014).Lifehistoryandspatial traits predict extinction risk due to climate change.Nature Climate Change, 4(3), 217–221.

Rice,J.,Harley,S.,Davies,N.,&Hampton,J.(2014).Stock assessment of skipjack tuna in the western and central Pacific Ocean. Majuro, Marshall Islands:WesternandCentralPacificFisheriesCommission.

Rudd,M.B.,&Thorson,J.T.(2018).Accountingforvariablerecruitmentand fishingmortality in length-based stock assessments for data-limited fisheries. Canadian Journal of Fisheries and Aquatic Sciences, 75, 1019–1035. https://doi.org/10.1139/cjfas-2017-0143

Salguero-Gómez,R.,Jones,O.R.,Jongejans,E.,Blomberg,S.P.,Hodgson,D.J.,Mbeau-Ache,C.,…Buckley,Y.M.(2016).Fast–slowcontinuumand reproductive strategies structure plant life- history variation worldwide. Proceedings of the National Academy of Sciences of the United States of America, 113(1), 230–235.

Simon,M., Fromentin, J.-M., Bonhommeau, S., Gaertner,D., Brodziak,J.,&Etienne,M.-P. (2012).Effectsofstochasticity inearly lifehis-tory on steepness and population growth rate estimates: An illus-tration on Atlantic bluefin tuna. PLoS ONE, 7, e48583. https://doi.org/10.1371/journal.pone.0048583

Stearns,S.C.(1992).The evolution of life histories.NewYork,NY:OxfordUniversity Press.

Sutherland,W.J.,Armstrong-Brown,S.,Armsworth,P.R.,Brereton,T.,Brickland,J.,Campbell,C.D.,…Watkinson,A.R.(2006).Theiden-tification of 100 ecological questions of high policy relevance in the UK. Journal of Applied Ecology, 43(4), 617–627.

Then,A.Y.,Hoenig,J.M.,Hall,N.G.,Hewitt,D.A.,Then,A.Y.,Hoenig,J.M.,…Hewitt,D.A. (2015).Evaluatingthepredictiveperformanceofempirical estimators of natural mortality rate using information on over 200 fish species. ICES Journal of Marine Science, 72(1), 82–92.

Thorson,J.T.,Munch,S.B.,Cope,J.M.,&Gao,J.(2017).Predictinglifehistory parameters for all fishes worldwide. Ecological Applications, 27, 2262–2276. https://doi.org/10.1002/eap.1606

Thorson,J.T.,Taylor,I.G.,Stewart,I.J.,&Punt,A.E.(2014).Rigorousmeta- analysis of life history correlations by simultaneously analyzing multiple population dynamics models. Ecological Applications, 24(2), 315–326. https://doi.org/10.1890/12-1803.1

Winemiller, K. O. (2005). Life history strategies, population regula-tion, and implications for fisheries management. Canadian Journal of Fisheries and Aquatic Sciences, 62(4), 872–885. https://doi.org/10.1139/f05-040

Zudaire,I.,Murua,H.,Grande,M.,&Bodin,N.(2013).Reproductivepo-tentialofYellowfinTuna (Thunnus albacares) in the western Indian ocean. Fishery Bulletin, 111, 252–264.

SUPPORTING INFORMATION

Additional supporting information may be found online in the SupportingInformationsectionattheendofthearticle.

How to cite this article: Horswill C, Kindsvater HK, Juan-JordáMJ,DulvyNK,MangelM,Matthiopoulos J.Globalreconstructionoflife-historystrategies:Acasestudyusing tunas. J Appl Ecol. 2019;56:855–865. https://doi.org/10.1111/1365-2664.13327

Horswill et al. Overcoming missing life-history data (supporting information)

1

Supporting information for Horswill et al. Global reconstruction of life-history strategies

# Appendix S1: Code for model of life-history traits # Fitted in JAGS: package ”jagsUI” v1.4.9 writeLines(" writeLines(" model{ #Priors and constraints beta2[1]~dnorm(0,0.1) beta2[2]~dnorm(0,0.1) beta2[3]~dnorm(0,0.1) beta2[4]~dnorm(0,0.1) beta2[5]~dnorm(0,0.1) beta2[6]~dnorm(0,0.1) beta2[7]~dnorm(0,0.1) Sigma[1:(ntraits),1:(ntraits)] <- inverse(Tau_mat[,]) Tau_mat[1:(ntraits),1:(ntraits)] ~ dwish(Omega[,],(ntraits)+1) for(i in 1:(ntraits)) { xi[i]~dunif(0,5) xi1[i]~dunif(0,5) } for (i in 1:ntraits){ mu_raw[i]~dnorm(0,15) mu_raw1[i]~dnorm(0,50) } Sigma1[1:(ntraits),1:(ntraits)] <- inverse(Tau_mat1[,]) Tau_mat1[1:(ntraits),1:(ntraits)] ~ dwish(Omega1[,],(ntraits)+1) species_prec[1] ~ dnorm(20,0.1)T(0.0,) species_prec[2] ~ dnorm(20,0.1)T(0.0,) species_prec[3] ~ dnorm(20,0.1)T(0.0,) species_prec[4] ~ dnorm(20,0.1)T(0.0,) species_prec[5] ~ dnorm(20,0.1)T(0.0,) species_prec[6] ~ dnorm(20,0.1)T(0.0,) species_prec[7] ~ dnorm(20,0.1)T(0.0,) # __________ Observation component ______________________ Spawn_d[1]~dnorm(mu[1,1],species_prec[1]) Spawn_d[2]~dnorm(mu[2,1],species_prec[1]) Spawn_d[3]~dnorm(mu[3,1],species_prec[1]) Spawn_d[4]~dnorm(mu[3,1],species_prec[1]) Spawn_d[5]~dnorm(mu[3,1],species_prec[1]) Spawn_d[6]~dnorm(mu[3,1],species_prec[1]) Spawn_d[7]~dnorm(mu[3,1],species_prec[1]) Spawn_d[8]~dnorm(mu[4,1],species_prec[1]) Spawn_d[9]~dnorm(mu[5,1],species_prec[1]) Spawn_d[10]~dnorm(mu[5,1],species_prec[1]) Spawn_d[11]~dnorm(mu[8,1],species_prec[1]) Spawn_d[12]~dnorm(mu[9,1],species_prec[1]) Spawn_d[13]~dnorm(mu[9,1],species_prec[1]) Spawn_d[14]~dnorm(mu[11,1],species_prec[1]) Spawn_d[15]~dnorm(mu[11,1],species_prec[1]) Spawn_d[16]~dnorm(mu[11,1],species_prec[1]) Spawn_d[17]~dnorm(mu[11,1],species_prec[1])

Horswill et al. Overcoming missing life-history data (supporting information)

2

Spawn_d[18]~dnorm(mu[11,1],species_prec[1]) Spawn_d[19]~dnorm(mu[12,1],species_prec[1]) Spawn_d[20]~dnorm(mu[12,1],species_prec[1]) Spawn_d[21]~dnorm(mu[12,1],species_prec[1]) Spawn_d[22]~dnorm(mu[13,1],species_prec[1]) Spawn_d[23]~dnorm(mu[14,1],species_prec[1]) Spawn_d[24]~dnorm(mu[15,1],species_prec[1]) Spawn_d[25]~dnorm(mu[15,1],species_prec[1]) Spawn_d[26]~dnorm(mu[15,1],species_prec[1]) Spawn_d[27]~dnorm(mu[15,1],species_prec[1]) Spawn_d[28]~dnorm(mu[15,1],species_prec[1]) Spawn_d[29]~dnorm(mu[16,1],species_prec[1]) Spawn_d[30]~dnorm(mu[16,1],species_prec[1]) Spawn_d[31]~dnorm(mu[18,1],species_prec[1]) Spawn_d[32]~dnorm(mu[18,1],species_prec[1]) Spawn_d[33]~dnorm(mu[19,1],species_prec[1]) Spawn_d[34]~dnorm(mu[20,1],species_prec[1]) Spawn_d[35]~dnorm(mu[20,1],species_prec[1]) Spawn_d[36]~dnorm(mu[21,1],species_prec[1]) Spawn_d[37]~dnorm(mu[22,1],species_prec[1]) Spawn_d[38]~dnorm(mu[22,1],species_prec[1]) Spawn_d[39]~dnorm(mu[23,1],species_prec[1]) Phi[1]~dnorm(mu[1,2], species_prec[2]) for (i in 2:4){ Phi[i]~dnorm(mu[2,2], species_prec[2]) } Phi[5]~dnorm(mu[3,2], species_prec[2]) for (i in 6:8){ Phi[i]~dnorm(mu[4,2], species_prec[2]) } for (i in 9:11){ Phi[i]~dnorm(mu[5,2], species_prec[2]) } Phi[12]~dnorm(mu[6,2], species_prec[2]) for (i in 13:17){ Phi[i]~dnorm(mu[7,2], species_prec[2]) } for (i in 18:22){ Phi[i]~dnorm(mu[8,2], species_prec[2]) } Phi[23]~dnorm(mu[9,2], species_prec[2]) for (i in 24:25){ Phi[i]~dnorm(mu[10,2], species_prec[2]) } for (i in 26:29){ Phi[i]~dnorm(mu[11,2], species_prec[2]) } for (i in 30:32){ Phi[i]~dnorm(mu[12,2], species_prec[2]) } for (i in 33:35){ Phi[i]~dnorm(mu[13,2], species_prec[2]) } for (i in 36:37){ Phi[i]~dnorm(mu[14,2], species_prec[2]) } Phi[38]~dnorm(mu[15,2], species_prec[2]) for (i in 39:45){ Phi[i]~dnorm(mu[16,2], species_prec[2]) } for (i in 46:50){ Phi[i]~dnorm(mu[17,2], species_prec[2]) } for (i in 51:52){ Phi[i]~dnorm(mu[18,2], species_prec[2]) } Phi[53]~dnorm(mu[19,2], species_prec[2]) for (i in 54:57){ Phi[i]~dnorm(mu[20,2], species_prec[2]) } for (i in 58:60){ Phi[i]~dnorm(mu[21,2], species_prec[2]) } for (i in 61:63){ Phi[i]~dnorm(mu[22,2], species_prec[2]) }

Horswill et al. Overcoming missing life-history data (supporting information)

3

for (i in 64:71){ Phi[i]~dnorm(mu[23,2], species_prec[2]) } m[1]~dnorm(mu[3,3], species_prec[3]) m[2]~dnorm(mu[4,3], species_prec[3]) m[3]~dnorm(mu[11,3], species_prec[3]) m[4]~dnorm(mu[15,3], species_prec[3]) m[5]~dnorm(mu[16,3], species_prec[3]) m[6]~dnorm(mu[16,3], species_prec[3]) m[7]~dnorm(mu[19,3], species_prec[3]) m[8]~dnorm(mu[20,3], species_prec[3]) m[9]~dnorm(mu[21,3], species_prec[3]) m[10]~dnorm(mu[22,3], species_prec[3]) m[11]~dnorm(mu[22,3], species_prec[3]) m[12]~dnorm(mu[23,3], species_prec[3]) m[13]~dnorm(mu[23,3], species_prec[3]) m[14]~dnorm(mu[23,3], species_prec[3]) Fec[1]~dnorm(mu[3,4], species_prec[4]) Fec[2]~dnorm(mu[5,4], species_prec[4]) Fec[3]~dnorm(mu[5,4], species_prec[4]) Fec[4]~dnorm(mu[5,4], species_prec[4]) Fec[5]~dnorm(mu[9,4], species_prec[4]) Fec[6]~dnorm(mu[10,4], species_prec[4]) Fec[7]~dnorm(mu[11,4], species_prec[4]) Fec[8]~dnorm(mu[12,4], species_prec[4]) Fec[9]~dnorm(mu[13,4], species_prec[4]) Fec[10]~dnorm(mu[13,4], species_prec[4]) Fec[11]~dnorm(mu[14,4], species_prec[4]) Fec[12]~dnorm(mu[15,4], species_prec[4]) Fec[13]~dnorm(mu[15,4], species_prec[4]) Fec[14]~dnorm(mu[15,4], species_prec[4]) Fec[15]~dnorm(mu[16,4], species_prec[4]) Fec[16]~dnorm(mu[18,4], species_prec[4]) Fec[17]~dnorm(mu[20,4], species_prec[4]) Fec[18]~dnorm(mu[20,4], species_prec[4]) Fec[19]~dnorm(mu[21,4], species_prec[4]) Fec[20]~dnorm(mu[22,4], species_prec[4]) Spawn_f[1]~dnorm(mu[5,5], species_prec[5]) Spawn_f[2]~dnorm(mu[5,5], species_prec[5]) Spawn_f[3]~dnorm(mu[5,5], species_prec[5]) Spawn_f[4]~dnorm(mu[5,5], species_prec[5]) Spawn_f[5]~dnorm(mu[9,5], species_prec[5]) Spawn_f[6]~dnorm(mu[11,5], species_prec[5]) Spawn_f[7]~dnorm(mu[12,5], species_prec[5]) Spawn_f[8]~dnorm(mu[13,5], species_prec[5]) Spawn_f[9]~dnorm(mu[13,5], species_prec[5]) Spawn_f[10]~dnorm(mu[15,5], species_prec[5]) Spawn_f[11]~dnorm(mu[15,5], species_prec[5]) Spawn_f[12]~dnorm(mu[15,5], species_prec[5]) Spawn_f[13]~dnorm(mu[15,5], species_prec[5]) Spawn_f[14]~dnorm(mu[16,5], species_prec[5]) Spawn_f[15]~dnorm(mu[18,5], species_prec[5]) Spawn_f[16]~dnorm(mu[20,5], species_prec[5]) Spawn_f[17]~dnorm(mu[20,5], species_prec[5]) Spawn_f[18]~dnorm(mu[21,5], species_prec[5]) Spawn_f[19]~dnorm(mu[21,5], species_prec[5]) Spawn_f[20]~dnorm(mu[22,5], species_prec[5]) annual_fec[1]~dnorm(mu[11,6], species_prec[6])

Horswill et al. Overcoming missing life-history data (supporting information)

4

annual_fec[2]~dnorm(mu[22,6], species_prec[6]) for (i in 1:3){ k[i]~dnorm(mu[1,7], species_prec[7]) } for (i in 4:8){ k[i]~dnorm(mu[2,7], species_prec[7]) } for (i in 9:17){ k[i]~dnorm(mu[3,7], species_prec[7]) } for (i in 18:23){ k[i]~dnorm(mu[4,7], species_prec[7]) } for (i in 24:45){ k[i]~dnorm(mu[5,7], species_prec[7]) } for (i in 46:49){ k[i]~dnorm(mu[6,7], species_prec[7]) } for (i in 50:53){ k[i]~dnorm(mu[7,7], species_prec[7]) } for (i in 54:71){ k[i]~dnorm(mu[8,7], species_prec[7]) } for (i in 72:81){ k[i]~dnorm(mu[9,7], species_prec[7]) } for (i in 82:83){ k[i]~dnorm(mu[10,7], species_prec[7]) } for (i in 84:87){ k[i]~dnorm(mu[11,7], species_prec[7]) } for (i in 88:96){ k[i]~dnorm(mu[12,7], species_prec[7]) } for (i in 97:105){ k[i]~dnorm(mu[13,7], species_prec[7]) } for (i in 106:111){ k[i]~dnorm(mu[14,7], species_prec[7]) } for (i in 112:124){ k[i]~dnorm(mu[15,7], species_prec[7]) } for (i in 125:138){ k[i]~dnorm(mu[16,7], species_prec[7]) } for (i in 139:150){ k[i]~dnorm(mu[17,7], species_prec[7]) } for (i in 151:154){ k[i]~dnorm(mu[18,7], species_prec[7]) } for (i in 155:158){ k[i]~dnorm(mu[19,7], species_prec[7]) } for (i in 159:163){ k[i]~dnorm(mu[20,7], species_prec[7]) } for (i in 164:167){ k[i]~dnorm(mu[21,7], species_prec[7]) } for (i in 168:173){ k[i]~dnorm(mu[22,7], species_prec[7]) } for (i in 174:183){ k[i]~dnorm(mu[23,7], species_prec[7]) } # __________ Demographic trait functions ______________ for (l in 1:(ntraits)){ mu[1,l]<- gamma[1,l] + alpha[1,l] + beta2[l]*H[1] mu[2,l]<- gamma[1,l] + alpha[2,l] + beta2[l]*H[1] mu[3,l]<- gamma[1,l] + alpha[3,l] + beta2[l]*H[1] mu[4,l]<- gamma[1,l] + alpha[4,l] + beta2[l]*H[1] mu[5,l]<- gamma[1,l] + alpha[5,l] + beta2[l]*H[1] mu[6,l]<- gamma[2,l] + alpha[6,l] + beta2[l]*H[2] mu[7,l]<- gamma[2,l] + alpha[7,l] + beta2[l]*H[2] mu[8,l]<- gamma[2,l] + alpha[8,l] + beta2[l]*H[2] mu[9,l]<- gamma[2,l] + alpha[9,l] + beta2[l]*H[2] mu[10,l]<- gamma[2,l] + alpha[10,l] + beta2[l]*H[2]

Horswill et al. Overcoming missing life-history data (supporting information)

5

mu[11,l]<- gamma[2,l] + alpha[11,l] + beta2[l]*H[2] mu[12,l]<- gamma[3,l] + alpha[12,l] + beta2[l]*H[3] mu[13,l]<- gamma[3,l] + alpha[13,l] + beta2[l]*H[3] mu[14,l]<- gamma[3,l] + alpha[14,l] + beta2[l]*H[3] mu[15,l]<- gamma[3,l] + alpha[15,l] + beta2[l]*H[3] mu[16,l]<- gamma[4,l] + alpha[16,l] + beta2[l]*H[4] mu[17,l]<- gamma[5,l] + alpha[17,l] + beta2[l]*H[5] mu[18,l]<- gamma[5,l] + alpha[18,l] + beta2[l]*H[5] mu[19,l]<- gamma[5,l] + alpha[19,l] + beta2[l]*H[5] mu[20,l]<- gamma[5,l] + alpha[20,l] + beta2[l]*H[5] mu[21,l]<- gamma[6,l] + alpha[21,l] + beta2[l]*H[6] mu[22,l]<- gamma[7,l] + alpha[22,l] + beta2[l]*H[7] mu[23,l]<- gamma[7,l] + alpha[23,l] + beta2[l]*H[7] } # __________ Species loop ______________________ for (j in 1:nspecies){ E0_raw[j,1:(ntraits)] ~ dmnorm(mu_raw[],Tau_mat[,]) for (m in 1:(ntraits)){ gamma[j,m] <- xi[m]*E0_raw[j,m]}} # __________ Population loop ______________________ for (j in 1:(nstocks)){ E0_raw1[j,1:(ntraits)] ~ dmnorm(mu_raw1[],Tau_mat1[,]) for (m in 1:(ntraits)){ alpha[j,m] <- xi1[m]*E0_raw1[j,m]}} } ", con="model.txt")

Horswill et al. Overcoming missing life-history data (supporting information)

1

Supporting information for Horswill et al. Global reconstruction of life-history strategies

# Appendix S2: Life history data for the Principal market tunas. Compiled from Juan-Jorda et al. (2016), also see Table S1 k<-c(-1.38230234,-1.180907531,-1.133203733,-0.84397007,-0.446287103,-0.261364764,-0.187535124,-0.187535124,-1.514127733,-0.891598119,-0.798507696,-0.755022584,-0.713349888,-0.693147181,-0.509160344,-0.430782916,-0.020202707,-2.302585093,-2.207274913,-1.692819521,-1.63475572,-1.237874356,-1.078809661,-1.237874356,-1.197328262,-0.916290732,-0.839329691,-0.776528789,-0.693147181,-0.597837001,-0.597837001,-0.597837001,-0.510825624,-0.287682072,-0.261364764,-0.261364764,-0.248461359,-0.056570351,-0.051293294,-0.051293294,0.09531018,0.09531018,0.223143551,0.223143551,0.262364264,-2.216407397,-2.137070655,-1.838851077,-1.820158944,-1.973281346,-1.609437912,-0.900663476,-0.616186139,-2.120263536,-2.047942875,-1.958995389,-1.771956842,-1.672910335,-1.660731207,-1.660731207,-1.609437912,-1.565421027,-1.527857925,-1.294627173,-1.272965676,-1.272965676,-1.251763468,-1.184170177,-1.139434283,-1.108662625,-0.30788478,-2.207274913,-1.771956842,-1.714798428,-1.614450454,-1.491654877,-1.30933332,-1.272965676,-1.177655496,-1.078809661,-0.494296322,-2.071473372,-1.966112856,-2.563949857,-2.120263536,-1.456716825,-0.371063681,-1.973281346,-1.320506621,-1.26940061,-1.157089218,-0.960242619,-0.867500568,-0.841647189,-0.814185509,-0.585190039,-0.653926467,-0.585190039,-0.572701027,-0.572701027,-0.525939262,-0.517514612,-0.510825624,-0.415515444,-0.415515444,-1.737271284,-1.237874356,-1.203972804,-1.080281332,-0.94160854,-0.916290732,-1.725971729,-1.609437912,-1.386294361,-1.386294361,-1.237874356,-1.231001477,-1.108662625,-0.867500568,-0.84397007,-0.789658081,-0.671385689,-0.572701027,-0.410980289,-2.244316185,-2.225624052,-2.055725015,-1.966112856,-1.966112856,-1.924148657,-1.903808973,-1.897119985,-1.832581464,-1.771956842,-1.692819521,-1.687399454,-1.676646662,-1.660731207,-2.918771232,-2.465104022,-2.26336438,-2.183025858,-2.040220829,-1.903808973,-1.832581464,-1.790360448,-1.754463684,-1.714798428,-1.46967597,-1.46967597,-2.538307427,-1.576970722,-1.139434283,-0.967584026,-1.766091722,-1.210661792,-1.177655496,-1.139434283,-1.604450371,-1.38230234,-1.378326191,-1.052683357,-1.002393431,-2.268183666,-2.120263536,-1.754463684,-1.714798428,-2.688247574,-2.659260037,-2.525728644,-2.375155786,-2.353878387,-2.302585093,-2.538307427,-2.538307427,-2.419118909,-2.407945609,-2.154165088,-2.154165088,-1.771956842,-1.62964062,-1.609437912,-1.203972804) k2 <- (k-mean(k))/sd(k) Phi<-c(-0.309806043,-0.657516324,-0.657516324,-0.657516324,-0.309806043,-0.657516324,-0.309806043,-0.046498493,-1.270176095,-0.882256298,-0.046498493,0.62127774,-0.657516324,-0.657516324,-0.046498493,0.164478627,0.164478627,0.164478627,0.340067608,0.557839041,0.557839041,0.62127774,0.62127774,0.93639519,0.93639519,0.490227377,0.490227377,0.93639519,1.022943267,-0.223846411,-0.214703327,0.064249252,-0.882256298,-0.470075308,-0.046498493,-0.046498493,0.490227377,-0.470075308,-0.046498493,1.102680113,1.481780096,1.87005982,2.121858717,2.121858717,2.202059753,0.164478627,0.490227377,0.490227377,0.507550982,0.841780558,-0.657516324,0.93639519,0.164478627,0.255962358,0.490227377,0.62127774,1.102680113,-0.046498493,1.102680113,1.715649798,1.022943267,1.370532854,1.533108894,1.309935591,1.427691951,1.427691951,1.673028439,1.905215789,1.939181133,1.97203367,2.034675635)

Horswill et al. Overcoming missing life-history data (supporting information)

2

Phi2 <- (Phi-mean(Phi))/sd(Phi) Fec<- c(12.80511929,13.32937755,12.47221362,12.24200913,13.72667934,13.65299163,13.99783211,14.5574479,14.26658618,14.73262735,14.93718812,14.58561878,15.05533302,14.81245919,15.60727003,14.18808001,14.58702247,14.80323231,16.58809928,15.99026228) Fec2<-(Fec-mean(Fec))/sd(Fec) Spawn_f<-c(0.631271777,0.58221562,1.490654376,0.609765572,0.530628251,0.262364264,1.208960346,0.182321557,0.2390169,0.58221562,0.688134639,0.431782416,0.500775288,0.09531018,0.262364264,0.086177696,0.09531018,1.190887565,1.193922468,0.182321557) Spawn_f2<-(Spawn_f-mean(Spawn_f))/sd(Spawn_f) Spawn_d<-c(365.00,152.08,365.00,365.00,365.00,365.00,365.00,121.67,182.50,365.00,182.50,152.08,212.92,121.67,121.67,121.67,182.50,212.92,,91.25,152.08,273.75,365.00,152.08,152.08,182.50,212.92,365.00,365.00,91.25,121.67,273.75,365.00,91.25,365.00,365.00,91.25,45.63,60.83,91.25) Spawn_d[Spawn_d==365.00] <- 364.99 Spawn_d2<-Spawn_d/365 Spawn_d2.1<-log(Spawn_d2/(1-Spawn_d2)) #logit Spawn_d3<-(Spawn_d2.1-mean(Spawn_d2.1))/sd(Spawn_d2.1) m<- c(0.405465108,0.693147181,1.504077397,0.875468737,2.397895273,2.442347035,0.678033543,0.78845736,1.386294361,1.098612289,1.098612289,2.079441542,2.442347035,2.76000994) m2<-(m-mean(m))/sd(m) annual_fec<- c(18.6101,18.48889) annual_fec2<-(annual_fec-mean(annual_fec))/sd(annual_fec) jags.data <- list(nstocks = 23, nspecies = 7, ntraits = 7, Omega=diag(7),Omega1=diag(7), #Habitat: 1-tropical; 0-temperate H=c(1,0,1,0,1,0,0), #Somatic growth k=k2, #Adult Survival Phi=Phi2, #Fecundity Fec=Fec2, #Spawning freq Spawn_f=Spawn_f2, #Spawning duration Spawn_d=Spawn_d3, #Age maturity m=m2, #annual fecundity

Horswill et al. Overcoming missing life-history data (supporting information)

3

annual_fec=annual_fec2)

Horswill et al. Overcoming missing life-history data (supporting information)

1

Supporting information for Horswill et al. Global reconstruction of life-history strategies

Appendix S3. Initial model structure

During initial model development three different model structures were created. Each one

included a different body metric; somatic growth rate (K), asymptotic body size (L∞) or

maximum observed body size (Lmax), alongside the remaining six life-history traits

(Φ ,A,S,D,F,V). Data were collated from Juan-Jordá et al. (2016) supplemented by additional

studies for batch and annual fecundity (supp. info. T1.), limiting data points to those

representing females when sex information was available. The fit of these three models was

compared based on the Brooks-Gelman-Rubin diagnostic tool for each parameter. These

goodness of fit criteria indicated poorer mixing of the MCMC in the models that included

asymptotic body size and maximum observed body size (min. effective sample size for model

including L∞ = 2512, all ≤ 2.00; min. effective sample size for model including Lmax = 2186,

all ≤ 1.05). The model that best described the data incorporated somatic growth rate (min.

effective sample size for model including K = 3764, all ≤1.01). Consequently, the results

presented relate to the model that included growth rate. The models based on the other two

variables followed largely identical structures.

Body size is often viewed as the principal driver of life-histories (Blueweiss et al. 1978; Sibly

& Brown 2007). However, correlations constructed with maximum body size are likely to

include larger residual error introduced by single individuals. In contrast, the estimation of

asymptotic size and growth rate from the von Bertalanffy growth curve typically requires large

sample sizes. The von Bertalanffy growth rate is an emergent property of physiology (foraging

efficiency) and behaviour (anti-predator behaviour) (Mangel 2006), and appears, at least in this

context, to provide greater predictive power for estimating life-history traits, when compared

to asymptotic size. This finding is in agreement with Jensen (1996) and Juan-Jordá et al. (2015).

In particular, the latter of these studies reports that time-related traits, such as growth rate,

rather than size-related traits, such as maximum size, better explain rates of population decline

for tuna and their relatives after accounting for fishing mortality.

Blueweiss, L., Fox, H., Kudzma, V., Nakashima, D., Peters, R. & Sams, S. (1978).

Relationships between body size and some life history parameters. Oecologia, 37, 257–

272.

r̂

r̂

r̂

Horswill et al. Overcoming missing life-history data (supporting information)

2

Jensen, A.L. (1996). Beverton and Holt life history invariants result from optimal trade-off of

reproduction and survival. Can. J. Fish. Aquat. Sci., 53, 820–822.

Juan-Jordá, M.J., Mosqueira, I., Freire, J. & Dulvy, N.K. (2015). Population declines of tuna

and relatives depend on their speed of life. Proc. R. Soc. B Biol. Sci., 282, 20150322.

Juan-Jordá, M.J., Mosqueira, I., Freire, J., Ferrer-Jordá, E. & Dulvy, N.K. (2016). Global

scombrid life history data set. Ecology, 97, 809.

Mangel, M. (2006). The Theoretical Biologist’s Toolbox. Cambridge University Press,

Cambridge.

Sibly, R.M. & Brown, J.H. (2007). Effects of body size and lifestyle on evolution of mammal

life histories. Proc. Natl. Acad. Sci., 104, 17707–17712.

Horswill et al. Overcoming missing life-history data (supporting information)

1

Supporting information for Horswill et al. Global reconstruction of life-history strategies

Appendix S4. From total fecundity of the population to mass specific fecundity

As described in the main text, in order to compute steepness, and the related management

quantities, we require mass specific fecundity but the methods described in the main text

produce total annual fecundity. We now show how to determine from total fecundity.

To do so, we use a standard age structured population model with constant rate of natural

mortality (e.g. Mangel 2006, Kindsvater et al 2016). We let denote the steady state

number of individuals of age a, with a=0 denoting the recruited class. Then for a>0 we have

and for a=0

(A.1)

where, as in the main text, is the survival through the egg and larval stage to recruitment

and is the steady stage production of eggs by the population. It is given by

(A.2)

Since and is a constant with respect to a, we have

(A.3)

where, as in the main text, is the average biomass of a spawning female.

The total steady state population size is given by

(A.4)

where, as in the main text, is maximum age. Thus, if we know total fecundity and total

population size – both predicted by our methods, we are able to compute mass specific

fecundity from Eqns A.3 and A.4 since

(A.5)

d

d

M N(a)

N(a) = N(0)e−Ma

(0)1

ELENE

fb

=+

ELf

E

( ) ( ) ( )ma

E N a p a W ad=å

N(a) = N(0)e−Ma d

(0) ( ) ( ) (0)Mam

aE N e p a W a N Wd d-= =å

W

NT

NT = N(a) = N(0)e−Maa∑

a∑ = N(0) e−Ma

a∑ = N(0) 1− e

MAmax

1− e−M⎡

⎣⎢

⎤

⎦⎥

Amax

max1 11

MA

MT

e EW e N

d -

é ù-= ê ú-ë û

Horswill et al. Overcoming missing life-history data (supporting information)

2

Knowing steepness provides information about various quantities relevant to fisheries

management (Mangel et al., 2013). For example, for fast-growing species (which many tunas

are), the fishing mortality that gives maximum sustainable yield ( ) is well approximated

by . The Spawning Potential Ratio at maximum sustainable yield (

) is defined to be lifetime offspring production of an individual when the population is fished

at divided by lifetime offspring production of an individual in an unfished population (see

Kindsvater, et al. 2016 for review). For a fast growing species, it is well approximated by

(Mangel et al., 2013).

Mangel, M. The Theoretical Biologist’s Toolbox. (Cambridge University Press, 2006).

Mangel, M., MacCall, A.D., Brodziak, J., Dick, E., Forrest, R.E., Pourzard, R., & Ralston, S.

(2013). A perspective on steepness, reference points, and stock assessment. Canadian

Journal of Fisheries and Aquatic Sciences, 940, 930–940.

Kindsvater, H. K., Mangel, M., Reynolds, J. D. & Dulvy, N. K. Ten principles from

evolutionary ecology essential for effective marine conservation. Ecol. Evol. 6, 2125–2138

(2016).

FMSY

FMSYM

= 4h1− h

−1 SPRMSY

FMSY

SPRMSY =1− h4h

Horswill et al. Overcoming missing life-history data (supporting information)

1

Supporting information for Horswill et al. Global reconstruction of life-history strategies

Figure S1. Correlation of raw (observed) life-history data and reconstructed (modelled) median

values for each trait. Species are demarked using colour. Colour ramp corresponds to the

observed gradient in the observed data for somatic growth rate, from fast to slow, within the

habitat groupings: tropical and temperate. All values are on the scale of the linear predictor,

1:1 line shown as grey line for reference.

Horswill et al. Overcoming missing life-history data (supporting information)

1

Supporting information for Horswill et al. Global reconstruction of life-history strategies

Figure S2. Refitting the life-history information for the Southern bluefin tuna population. The

six simulations represent the life history information supplied as data: A. Somatic growth only,

B. Somatic growth and survival, C. Somatic growth and maturity, D. Somatic growth and

spawning frequency, E. Somatic growth and spawning duration, F. Somatic growth and batch

fecundity. Two-dimensional kernel density plots between the posterior values of each life-

A. Growth B. Growth + Survival

C. Growth + Maturity

F. Growth + Batch fecundityE. Growth + Spawning duration

D. Growth + Spawning freq.

Horswill et al. Overcoming missing life-history data (supporting information)

2

history trait. Original data points are shown as red circles and red dashed lines, and the mean

imputed values from the full model are shown as black triangles and black dashed lines.

Horswill et al. Overcoming missing life-history data (supporting information)

1

Supporting information for Horswill et al. Global reconstruction of life-history strategies

Figure S3. Reconstructed life-history traits by habitat, species and population for a model

without the habitat covariate included in all demographic functions, with the exception of

maturity. The 95% credible intervals are shown as bars. Blue indicates parameters that also had

data available (all observed data points fell the credible intervals), orange indicates parameters

with data missing from the original dataset. Species and populations listed in the observed order

of somatic growth rate, from fast to slow, within the tropical and temperate groupings. Somatic

growth is adjusted to include population specific residuals. Units: age of maturity (years),

spawning frequency and spawning duration (days), batch fecundity (number of oocytes per

individual per batch, x107), annual fecundity (number of ooctyes per individual per year, x108).

Horswill et al. Overcoming missing life-history data (supporting information)

1

Supporting information for Horswill et al. Global reconstruction of life-history strategies

Figure S4. Correlation of the median posterior values for survival, age of maturity and somatic

growth between a model including all seven life-history traits and a model including three traits

only (survival, age of maturity and growth).

Horswill et al. Overcoming missing life-history data (supporting information)

1

Supporting information for Horswill et al. Global reconstruction of life-history strategies

Table S1. Number of studies available for each life-history trait across the 23 populations for

principal market tuna.

Species Population

Som

atic

gro

wth

rate

Asy

mpt

otic

bod

y si

ze

Max

imum

bod

y si

ze

Max

imum

age

(use

d to

de

rive

rate

s of s

urvi

val)

Age

mat

urity

Spaw

ning

freq

uenc

y

Spaw

ning

dur

atio

n

Bat

ch fe

cund

ity

Ann

ual F

ecun

dity

Skipjack tuna Katsuwonus pelamis

Eastern Atlantic population

3 3 5 1 - - 1 - -

Eastern Pacific population

5 5 4 3 - - 1 - -

Indian population

9 9 18 1 1 - 5 1 -

Western Atlantic population

6 6 9 3 1 - 1 - -

Western Pacific population

22 22 25 3 - 4 2 3 -

Albacore tuna Thunnus alalunga

Indian population

4 4 7 1 - - - - -

Mediterranean population

4 4 5 5 - - - - -

Northern Atlantic population

18 18 17 5 - - 1 - -

Northern Pacific population

10 10 12 1 - 1 2 1 -

Southern Atlantic population

2 2 4 2 - - - 1* -

Southern Pacific population

4 4 4 4 1 1 5 1 1*

Atlantic population

9 9 8 3 - 1 3 1 -

Horswill et al. Overcoming missing life-history data (supporting information)

2

Yellowfin tuna Thunnus albacores

Eastern Pacific population

9 9 10 3 - 2 1 2 -

Indian population

6 6 11 2 - - 1 1* -

Western Pacific population

13 13 17 1 - 4 5 3 -

Southern bluefin tuna Thunnus maccoyii

Southern population

14 14 10 7 2 1 2 1 -

Bigeye tuna Thunnus obesus

Atlantic population

12 12 13 5 - - - - -

Eastern Pacific population

4 4 6 2 - 1 2 1 -

Indian population

4 4 5 1 1 - 1 - -

Western Pacific population

5 5 12 4 1 2 2 2 -

Pacific bluefin tuna Thunnus orientalis

Pacific population

4 4 7 3 1 1 1 1 -

Atlantic bluefin tuna Thunnus thynnus

Eastern Atlantic population

6 6 10 3 2 1 2 1* 1*

Western Atlantic population

10 10 14 8 3 - 1 - -

*Supplementary studies to Juan-Jordá et al. (2016):

Aranda, G., A. Medina, A. Santos, F. J. Abascal, and T. Galaz. 2013. Evaluation of Atlantic

bluefin tuna reproductive potential in the western Mediterranean Sea. Journal of Sea Research

76:154–160.

Farley, J. H., A. J. Williams, S. D. Hoyle, C. R. Davies, and S. J. Nicol. 2013. Reproductive

dynamics and potential annual fecundity of South Pacific albacore tuna (Thunnus alalunga).

PLoS ONE 8:e60577.

ICCAT. 2012. Report of the 2011 ICCAT South Atlantic and Mediterranean albacore stock

assessment sessions. Col. Vol. Sci. Pap. ICCAT 68:387–497.

Medina, A., F. J. Abascal, C. Megina, and A. Garcia. 2002. Stereological assessment of the

reproductive status of female Atlantic northern bluefin tuna during migration to Mediterranean

spawning grounds through the Strait of Gibraltar. Journal of Fish Biology 60:203–217.

Horswill et al. Overcoming missing life-history data (supporting information)

3

Zudaire, I., H. Murua, M. Grande, and N. Bodin. 2013. Reproductive potential of Yellowfin

Tuna (Thunnus albacares) in the western Indian ocean. Fishery Bulletin 111:252–264.

Horswill et al. Overcoming missing life-history data (supporting information)

1

Supporting information for Horswill et al. Global reconstruction of life-history strategies

Table S2. Key parameters in the model of life-history traits and method of estimation

Parameter Parameter notation

Equation Estimation Prior

Coefficient for habitat term 2 Estimated

Species-specific mean trait value

2 Estimated

Population-specific trait values

2 Estimated

Mean expectation for species-level MVN

3 Estimated

Mean expectation for population-level MVN

3 Estimated

Variance-covaraince matrix for species-level MVN

3 Estimated

Variance-covaraince matrix for population-level MVN

3 Estimated

Scale matrix for species-level Inverse Wishart

4 Fixed

Scale matrix for population-level Inverse Wishart

4 Fixed

Scaling parameter for species-level scale matrix

4 Estimated

Scaling parameter for population-level Inverse Wishart

4 Estimated

Precision of observation component

5 Estimated

y (0,5)N

g1( , )iMVN µ S

w 2( , )jMVN µ S

1µ (0,15)N

2µ (0,50)N

iS ( ,8)iInvWishart W

jS ( ,8)jInvWishart W

iW 8I

jW 8I

ix (0,5)U

jx (0,5)U

t (20,0.1)N

Horswill et al. Overcoming missing life-history data (supporting information)

1

Supporting information for Horswill et al. Global reconstruction of life-history strategies

ST3. Imputed life-history traits for each population of principal market tuna on the natural scale (median posterior values and 95% credible intervals

of the posterior distribution)

Species Population

Spec

ies-

mea

n so

mat

ic g

row

th

Surv

ival

Age

mat

urity

(yea

rs)

Spaw

ning

freq

uenc

y (d

ays)

Spaw

ning

dur

atio

n (d

ays)

Bat

ch fe

cund

ity (x

106 )

Ann

ual F

ecun

dity

(x10

8 )

Med. 95-CI Med. 95-CI Med. 95-CI Med. 95-CI Med. 95-CI Med. 95-CI Med. 95-CI

Skipjack tuna

Katsuwonus pelamis

Eastern

Atlantic

0.33 0.22, 0.49

0.40 0.29, 0.53

1.80 0.50, 6.74

1.80 0.58, 5.31

364.91 357, 365.00

0.35 0.21,

0.70

1.15 0.61,

2.22

Eastern Pacific 0.61 0.44,

0.84

0.35 0.27,

0.44

1.84 0.44,

7.81

2.25 0.81,

6.17

325.43 23.73,

364.68

0.34 0.21,

0.61

1.16 0.61,

2.24

Indian 0.49 0.38,

0.63

0.38 0.28,

0.50

1.53 1.08,

2.17

1.73 0.63,

4.48

364.98 364.79,

365.00

0.35 0.25,

0.51

1.16 0.62,

2.22

Western

Atlantic

0.21 0.15,

0.29

0.42 0.33,

0.52

1.97 1.38,

2.79

2.28 0.62,

7.79

256.60 7.28,

364.10

0.36 0.21,

0.81

0.16 0.60,

2.23

Horswill et al. Overcoming missing life-history data (supporting information)

2

Western Pacific 0.67 0.57,

0.78

0.34 0.26,

0.43

1.75 0.44,

6.91

2.27 1.96,

2.64

363.50 324.99,

364.95

0.33 0.25,

0.43

1.16 0.61,

2.23

Albacore tuna

Thunnus alalunga

Indian 0.15 0.10,

0.21

0.64 0.53,

0.75

4.57 1.24,

16.84

1.61 0.61,

4.35

205.78 0.06,

364.96

1.01 0.60,

1.77

1.18 0.86,

1.57

Mediterranean 0.28 0.20,

0.39

0.51 0.42,

0.60

3.57 0.81,

15.69

1.60 0.53,

5.19

329.16 0.15,

365.00

0.96 0.48,

1.69

1.18 0.80,

1.67

Northern

Atlantic

0.23 0.19,

0.27

0.61 0.54,

0.68

4.46 1.34,

14.83

1.60 0.67,

4.02

230.05 7.37,

362.44

0.98 0.59,

1.65

1.18 0.89,

1.52

Northern

Pacific

0.24 0.19,

0.30

0.62 0.52,

0.72

4.59 1.40,

15.15

1.67 1.26,

2.23

209.61 16.05,

356.12

0.97 0.67,

1.38

1.18 0.90,

1.50

Southern

Atlantic

0.15 0.10,

0.24

0.67 0.58,

0.76

4.89 1.33,

17.53

1.58 0.59,

4.27

206.91 0.07,

364.95

0.97 0.67,

1.40

1.18 0.86,

1.57

Southern

Pacific

0.19 0.14,

0.27

0.65 0.58,

0.72

4.54 3.24,

6.39

1.36 1.03,

1.81

174.19 30.93,

329.31

1.03 0.73,

1.52

1.20 1.16,

1.25

Yellowfin tuna

Thunnus albacores

Atlantic 0.34 0.26,

0.43

0.47 0.37,

0.56

2.10 0.55,

8.22

3.06 2.21,

4.11

219.23 28.60,

353.94

2.37 1.65,

3.21

1.14 0.60,

2.17

Eastern Pacific 0.56 0.44,

0.72

0.41 0.32,

0.51

1.93 0.49,

7.81

1.27 1.03,

1.58

364.95 358.23,

365.00

2.26 1.61,

2.93

1.14 0.60,

2.22

Indian 0.31 0.23,

0.41

0.51 0.4,

0.62

2.29 0.64,

8.52

2.01 0.74,

5.22

276.65 11.49,

363.81

2.55 1.90,

3.76

1.14 0.60,

2.16

Western Pacific 0.35 0.28,

0.43

0.45 0.34,

0.56

2.32 1.63,

3.3

1.74 1.50,

2.02

359.61 318.18,

364.43

2.51 2.00,

3.24

1.14 0.62,

2.12

Southern bluefin tuna

Thunnus maccoyii

Southern 0.15 0.12,

0.19

0.82 0.77,

0.86

10.73 8.29,

13.78

1.14 0.85,

1.54

140.50 6.50,

349.10

5.96 3.60,

9.87

1.13 0.77,

1.66

Horswill et al. Overcoming missing life-history data (supporting information)

3

Bigeye tuna Thunnus

obesus

Atlantic 0.14 0.12,

0.18

0.62 0.54,

0.69

2.15 0.56,

8.36

1.52 0.53,

4.62

351.86 0.80,

365.00

2.06 1.18

3.83

1.13 0.59,

2.14

Eastern Pacific 0.22 0.16,

0.32

0.57 0.46,

0.66

1.98 0.59,

6.77

1.32 1.00,

1.76

363.66 327.83,

364.95

1.90 1.23,

2.72

1.12 0.61,

2.10

Indian 0.25 0.18,

0.36

0.57 0.44,

0.68

1.98 1.4,

2.8

1.64 0.65,

4.63

279.95 10.75,

364.14

2.01 1.11,

3.51

1.13 0.60,

2.11

Western Pacific 0.27 0.19,

0.37

0.61 0.53,

0.69

2.20 1.55,

3.10

1.11 0.91,

1.37

364.96 363.26,

365.00

2.12 1.54

2.96

1.13 0.60,

2.13

Pacific bluefin tuna

Thunnus orientalis

Pacific 0.15 0.10,

0.21

0.72 0.62,

0.80

4.06 2.87,

5.74

3.18 2.55,

3.91

111.27 1.61,

356.78

14.81 8.76,

24.5

1.12 0.76,

1.63

Atlantic bluefin tuna

Thunnus thynnus

Eastern

Atlantic

0.09 0.07,

0.12

0.80 0.72,

0.86

3.13 2.44,

4.07

1.25 0.94,

1.67

84.61 3.14,

335.15

8.63 5.20,

14.24

1.07 1.03,

1.11

Western

Atlantic

0.13 0.10,

0.16

0.84 0.8,

0.87

10.89 8.82,

13.39

1.33 0.45,

3.90

122.99 1.66,

358.41

8.70 4.08,

18.62

1.09 0.78,

1.58

Horswill et al. Overcoming missing life-history data (supporting information)

1

Supporting information for Horswill et al. Global reconstruction of life-history strategies

Table S4. The change in credible intervals (CI) associated with the reconstructed life history

traits for Southern bluefin tuna, comparing a model that incorporates all available data and a

model that includes somatic growth rate only.

Southern bluefin Full model Reduced model % Change

2.5 CI Median 97.5 CI 2.5 CI Median 97.5 CI

Survival 0.80 1.14 1.47 -1.31 0.30 1.91 376.05

Maturity 0.81 1.13 1.45 -1.59 0.06 1.72 419.05

Spawning duration -1.38 -0.70 -0.02 -1.42 -0.69 0.04 7.71

Spawning frequency -1.69 -1.01 -0.29 -2.88 0.13 3.13 327.66

Batch fecundity 0.68 1.13 1.59 -2.80 0.45 3.69 612.43

Annual fecundity -4.60 -0.12 4.40 -4.46 -0.06 4.29 -2.78

Somatic growth -1.10 -0.81 -0.52 -1.09 -0.80 -0.51 0.00

Horswill et al. Overcoming missing life-history data (supporting information)

1

Supporting information for Horswill et al. Global reconstruction of life-history strategies

Table S5. Percentage change in the credible intervals (CI) of the life-history traits reconstructed using a model with seven life-history traits

(including the four fecundity metrics) compared with traits reconstructed using a model based on three life-history traits (somatic growth, survival

and age of maturity). Variables on the scale of the linear predictor.

Trait Species / Stock Model with 7 traits Model with 3 traits

% change in 95 CI 2.5 CI median 97.5 CI 2.5 CI median 97.5 CI Survival Katsuwonus pelamis Eastern Atlantic stock -1.60 -1.05 -0.44 -1.54 -1.00 -0.40 -0.44

Katsuwonus pelamis Eastern Pacific stock -1.71 -1.29 -0.86 -1.73 -1.30 -0.88 -0.61 Katsuwonus pelamis Indian stock -1.69 -1.16 -0.57 -1.65 -1.13 -0.58 -4.20 Katsuwonus pelamis Western Atlantic stock -1.39 -0.95 -0.49 -1.37 -0.93 -0.46 1.69 Katsuwonus pelamis Western Pacific stock -1.78 -1.35 -0.93 -1.79 -1.35 -0.93 1.71 Thunnus alalunga Indian stock -0.44 0.09 0.67 -0.42 0.11 0.69 0.14 Thunnus alalunga Mediterranean stock -0.97 -0.56 -0.13 -0.97 -0.56 -0.14 -1.34 Thunnus alalunga Northern Atlantic stock -0.40 -0.07 0.27 -0.41 -0.07 0.27 1.48 Thunnus alalunga Northern Pacific stock -0.51 -0.02 0.52 -0.54 -0.03 0.54 4.87 Thunnus alalunga Southern Atlantic stock -0.23 0.22 0.76 -0.22 0.25 0.79 2.65 Thunnus alalunga Southern Pacific stock -0.23 0.13 0.52 -0.24 0.13 0.52 1.81 Thunnus albacares Atlantic stock -1.19 -0.74 -0.32 -1.18 -0.75 -0.33 -2.43 Thunnus albacares Eastern Pacific stock -1.45 -1.00 -0.55 -1.48 -1.03 -0.60 -2.51 Thunnus albacares Indian stock -1.03 -0.56 -0.04 -1.05 -0.57 -0.05 0.69 Thunnus albacares Western Pacific stock -1.35 -0.81 -0.31 -1.41 -0.83 -0.31 5.67 Thunnus maccoyii Southern stock 0.80 1.14 1.47 0.80 1.14 1.48 0.79 Thunnus obesus Atlantic stock -0.39 -0.03 0.33 -0.38 -0.02 0.34 -1.10 Thunnus obesus Eastern Pacific stock -0.77 -0.27 0.19 -0.79 -0.27 0.18 1.82 Thunnus obesus Indian stock -0.86 -0.25 0.29 -0.86 -0.27 0.25 -3.20 Thunnus obesus Western Pacific stock -0.43 -0.05 0.34 -0.42 -0.04 0.35 0.70 Thunnus orientalis Pacific stock -0.01 0.48 0.98 -0.03 0.45 0.93 -2.98

Horswill et al. Overcoming missing life-history data (supporting information)

2

Thunnus thynnus Eastern Atlantic stock 0.52 1.00 1.46 0.49 0.98 1.44 1.14 Thunnus thynnus Western Atlantic stock 0.99 1.30 1.60 0.99 1.30 1.61 0.85

Maturity Katsuwonus pelamis Eastern Atlantic stock -2.70 -1.11 0.55 -2.60 -1.05 0.52 -4.08 Katsuwonus pelamis Eastern Pacific stock -2.88 -1.08 0.73 -2.86 -1.16 0.57 -5.14 Katsuwonus pelamis Indian stock -1.75 -1.31 -0.87 -1.73 -1.30 -0.86 -0.60 Katsuwonus pelamis Western Atlantic stock -1.44 -0.99 -0.56 -1.43 -1.00 -0.55 -0.24 Katsuwonus pelamis Western Pacific stock -2.87 -1.14 0.58 -2.96 -1.19 0.54 1.59 Thunnus alalunga Indian stock -1.58 0.06 1.70 -1.49 0.07 1.63 -4.68 Thunnus alalunga Mediterranean stock -2.12 -0.25 1.61 -2.11 -0.31 1.56 -1.53 Thunnus alalunga Northern Atlantic stock -1.48 0.03 1.54 -1.49 0.00 1.49 -1.26 Thunnus alalunga Northern Pacific stock -1.42 0.07 1.57 -1.48 0.05 1.56 1.72 Thunnus alalunga Southern Atlantic stock -1.48 0.15 1.75 -1.37 0.16 1.72 -4.41 Thunnus alalunga Southern Pacific stock -0.37 0.05 0.48 -0.38 0.05 0.48 1.00 Thunnus albacares Atlantic stock -2.59 -0.91 0.80 -2.50 -0.92 0.65 -7.02 Thunnus albacares Eastern Pacific stock -2.75 -1.02 0.73 -2.79 -1.05 0.67 -0.78 Thunnus albacares Indian stock -2.40 -0.80 0.84 -2.40 -0.81 0.81 -0.71 Thunnus albacares Western Pacific stock -1.23 -0.79 -0.35 -1.22 -0.79 -0.35 -1.91 Thunnus maccoyii Southern stock 0.81 1.13 1.45 0.81 1.13 1.45 -0.08 Thunnus obesus Atlantic stock -2.57 -0.88 0.82 -2.50 -0.87 0.74 -4.46 Thunnus obesus Eastern Pacific stock -2.50 -0.99 0.56 -2.54 -1.00 0.51 -0.03 Thunnus obesus Indian stock -1.43 -0.99 -0.55 -1.42 -1.00 -0.57 -2.52 Thunnus obesus Western Pacific stock -1.29 -0.86 -0.42 -1.28 -0.85 -0.41 0.40 Thunnus orientalis Pacific stock -0.52 -0.09 0.35 -0.52 -0.08 0.34 -0.86 Thunnus thynnus Eastern Atlantic stock -0.72 -0.41 -0.08 -0.74 -0.42 -0.09 0.46 Thunnus thynnus Western Atlantic stock 0.89 1.15 1.41 0.88 1.15 1.41 0.47

Growth Katsuwonus pelamis Eastern Atlantic stock -0.28 0.30 0.85 -0.33 0.24 0.80 0.14 Katsuwonus pelamis Eastern Pacific stock 0.70 1.16 1.62 0.74 1.19 1.64 -1.98 Katsuwonus pelamis Indian stock 0.51 0.86 1.20 0.49 0.83 1.18 -1.41 Katsuwonus pelamis Western Atlantic stock -0.78 -0.34 0.10 -0.77 -0.33 0.11 -0.82 Katsuwonus pelamis Western Pacific stock 1.06 1.29 1.52 1.07 1.30 1.53 -0.25 Thunnus alalunga Indian stock -1.34 -0.83 -0.33 -1.33 -0.84 -0.34 -1.88

Horswill et al. Overcoming missing life-history data (supporting information)

3

Thunnus alalunga Mediterranean stock -0.44 0.05 0.55 -0.45 0.06 0.56 2.32 Thunnus alalunga Northern Atlantic stock -0.47 -0.22 0.03 -0.48 -0.22 0.03 1.82 Thunnus alalunga Northern Pacific stock -0.49 -0.17 0.16 -0.49 -0.16 0.17 1.37 Thunnus alalunga Southern Atlantic stock -1.44 -0.79 -0.15 -1.47 -0.82 -0.18 0.14 Thunnus alalunga Southern Pacific stock -0.93 -0.45 0.03 -0.92 -0.45 0.04 -0.03 Thunnus albacares Atlantic stock -0.03 0.33 0.67 -0.01 0.34 0.69 0.14 Thunnus albacares Eastern Pacific stock 0.71 1.06 1.41 0.70 1.05 1.40 -0.24 Thunnus albacares Indian stock -0.23 0.19 0.61 -0.21 0.21 0.62 -0.54 Thunnus albacares Western Pacific stock 0.09 0.37 0.66 0.08 0.38 0.67 2.69 Thunnus maccoyii Southern stock -1.10 -0.81 -0.52 -1.10 -0.81 -0.52 -0.43 Thunnus obesus Atlantic stock -1.18 -0.88 -0.57 -1.18 -0.87 -0.57 0.21 Thunnus obesus Eastern Pacific stock -0.75 -0.26 0.24 -0.78 -0.28 0.22 1.10 Thunnus obesus Indian stock -0.57 -0.08 0.43 -0.53 -0.04 0.46 -0.88 Thunnus obesus Western Pacific stock -0.45 0.00 0.46 -0.49 -0.03 0.42 -0.18 Thunnus orientalis Pacific stock -1.38 -0.86 -0.34 -1.36 -0.84 -0.33 -0.74 Thunnus thynnus Eastern Atlantic stock -1.96 -1.52 -1.09 -1.96 -1.52 -1.09 0.14 Thunnus thynnus Western Atlantic stock -1.38 -1.04 -0.71 -1.38 -1.05 -0.70 1.49