Genome‐wide SNPs resolve spatiotemporal patterns of connectivity within striped marlin (Kajikia audax), a broadly distributed and highly migratory pelagic species

Abstract Genomic methodologies offer unprecedented opportunities for statistically robust studies of species broadly distributed in environments conducive to high gene flow, providing valuable information for wildlife conservation and management. Here, we sequence restriction site‐associated DNA to characterize genome‐wide single nucleotide polymorphisms (SNPs) in a broadly distributed and highly migratory large pelagic fish, striped marlin (Kajikia audax). Assessment of over 4,000 SNPs resolved spatiotemporal patterns of genetic connectivity throughout the species range in the Pacific and, for the first time, Indian oceans. Individual‐based cluster analyses identified six genetically distinct populations corresponding with the western Indian, eastern Indian, western South Pacific, and eastern central Pacific oceans, as well as two populations in the North Pacific Ocean (F ST = 0.0137–0.0819). F ST outlier analyses identified a subset of SNPs (n = 59) putatively under the influence of natural selection and capable of resolving populations separated by comparatively high degrees of genetic differentiation. Temporal collections available for some regions demonstrated the stability of allele frequencies over three to five generations of striped marlin. Relative migration rates reflected lower levels of genetic connectivity between Indian Ocean populations (m R ≤ 0.37) compared with most populations in the Pacific Ocean (m R ≥ 0.57) and highlight the importance of the western South Pacific in facilitating gene flow between ocean basins. Collectively, our results provide novel insights into rangewide population structure for striped marlin and highlight substantial inconsistencies between genetically distinct populations and stocks currently recognized for fisheries management. More broadly, we demonstrate that species capable of long‐distance dispersal in environments lacking obvious physical barriers to movement can display substantial population subdivision that persists over multiple generations and that may be facilitated by both neutral and adaptive processes. Importantly, surveys of genome‐wide markers enable inference of population‐level relationships using sample sizes practical for large pelagic fishes of conservation concern.


| INTRODUC TI ON
Genetic studies of wild populations have assisted the development of scientifically informed management and conservation efforts over the past five decades (Frankel, 1974;Franklin, 1980;Simberloff, 1988;Soulé, 1985). However, for many species of conservation concern, information on spatiotemporal patterns of genetic connectivity remains limited, challenging the ability of resource managers to develop recovery plans that consider genetic attributes of distinct populations (Frankham, 2010;Funk, McKay, Hohenlohe, & Allendorf, 2012;Palsbøll, Bérubé, & Allendorf, 2006). This is frequently the case for large pelagic fishes, which display broad spatial distributions and highly migratory life histories, and seasonally occupy multiple domestic and international management jurisdictions (Meltzer, 1994). The advent of next-generation sequencing (NGS; e.g., Mardis, 2008) represents a particularly important advancement in genetic insights possible with sampling designs practicable for large pelagic fishes, especially those that also occur in low frequency. Surveys of genome-wide variation are capable of providing new information on population structure and movement patterns to improve conservation and management efforts for these species.
The now widespread availability of NGS enables rapid and cost-effective surveys of thousands to hundreds of thousands of molecular markers across entire genomes, facilitating statistically robust assessments of neutral and adaptive genomic variation in nonmodel systems. Though such assessments have progressed toward unraveling complex relationships between fitness-related traits and underlying genomic architectures in more easily accessible systems (e.g., Pacific salmonids [Oncorhynchus spp., Salmo spp.]; Ayllon et al., 2015;Barson et al., 2015;Prince et al., 2017;Thompson et al., 2019), applications of NGS to large pelagic fishes are just getting started, and recent studies illustrate the utility of genomic methods for resolving spatial patterns of connectivity in pelagic systems.
Considerable potential exists for genomic surveys to provide valuable information for additional large pelagic fishes of conservation concern, including those found at comparatively lower frequencies or of lesser commercial value.
Striped marlin (Kajikia audax) is a large pelagic fish broadly distributed in temperate, subtropical, and tropical waters of the Pacific and Indian oceans (Nakamura, 1985), and supports valuable recreational and commercial fisheries throughout the species range. Tagging studies demonstrate that striped marlin is capable of long-distance movements spanning hundreds to thousands of kilometers over periods less than one year (Domeier, 2006;Holdsworth, Sippel, & Block, 2009;Ortiz et al., 2003;Sippel, Davie, Holdsworth, & Block, 2007), presumably to exploit seasonal spawning and feeding grounds. Despite such high dispersal capabilities, striped marlin display some degree of site fidelity to regions largely corresponding with known spawning grounds (Domeier, 2006;Ortiz et al., 2003).
Genetic studies based on microsatellites and mtDNA provide evidence for at least four genetically distinct populations of striped marlin in the Pacific Ocean (McDowell & Graves, 2008;Purcell & Edmands, 2011), and these results are generally consistent with available information on seasonal movements and spawning behavior. However, incongruities in results between genetic studies reflect uncertain population-level relationships for striped marlin in the central North Pacific and eastern central Pacific oceans, complicating the delineation of biologically relevant management units in these regions. Additionally, population structure for striped marlin in the Indian Ocean remains unexplored, and the degree of genetic connectivity between ocean basins is unknown.
Practical challenges to implementing biologically representative sampling designs have impeded rangewide studies of biological and genetic relationships in striped marlin. Catches of striped marlin outside of seasonal feeding or spawning assemblages are typically low, and known assemblages are often difficult to access. The location and timing of striped marlin spawning are also poorly understood, and few efforts exist to sample larvae or reproductively active adults.
These challenges have resulted in a lack of information on spatial genetic variation across the full range of striped marlin, leading to mismatches between populations characterized by distinct biological processes and stocks recognized by regional fisheries management organizations (RFMOs). For example, a single ocean-wide stock is presently recognized in RFMO assessment and management efforts in the Indian Ocean due to insufficient information on population structure. Additionally, uncertain population structure in some regions of the Pacific Ocean is further complicated by continued use of stock boundaries inconsistent with populations evident from available genetic and biological information. Such mismatches may be

K E Y W O R D S
DArT, genetic connectivity, highly migratory, large pelagic fish, population genomics, SNPs, striped marlin, temporal stability especially problematic for striped marlin because this species is estimated to be overfished or experiencing unsustainable levels of fishing effort in regions across the Indo-Pacific (IATTC, 2018;WCPFC, 2012WCPFC, , 2018. Studies that provide information to improve rangewide management and conservation efforts for striped marlin are timely not only because of the unsustainable status of most stocks (Collette et al., 2011;Cullis-Suzuki & Pauly, 2010), but also because habitat utilization and seasonal movements of striped marlin and other large pelagic fishes rely on environmental cues progressively influenced by a changing global climate (Carlisle et al., 2017;Dell'Apa, Carney, Davenport, & Vernon, 2018;Duery, Bopp, & Maury, 2014;Hazen et al., 2012;Mislan, Deutsch, Brill, Dunne, & Sarmiento, 2017;Muhling, Lee, Lamkin, & Liu, 2011;Pentz, Klenk, Ogle, & Fisher, 2018).
Here, we employ NGS of restriction site-associated DNA (Andrews, Good, Miller, Luikart, & Hohenlohe, 2016;Baird et al., 2008) to assess spatiotemporal patterns of genetic variation in striped marlin across the Pacific and Indian oceans. Relative to previous genetic studies of striped marlin, NGS-based methodology facilitates statistically robust assessment of genome-wide variation (Helyar et al., 2011;Luikart, England, Tallmon, Jordan, & Taberlet, 2003;Nielsen, Hemmer-Hansen, Larsen, & Bekkevold, 2009) despite pragmatic constraints on sampling efforts for this species. The primary objectives of this study were to: (a) determine the number and geographic extent of striped marlin populations in the Pacific and Indian oceans, (b) evaluate whether genetically distinct populations correspond with genetic variation potentially influenced by natural selection, and (c) assess the multigenerational stability of observed genomic variation. This work represents the first genomic assessment of striped marlin, and results provide novel insights into genetic connectivity within and between ocean basins, the role of putatively neutral and adaptive processes in facilitating population subdivision, and the stability of allele frequencies over decadal time periods. Samples consisted of fin tissue from striped marlin released alive following capture by recreational anglers or from striped marlin caught incidentally by commercial pelagic longline vessels targeting tunas and swordfish. Additional samples consisting of muscle tissue were obtained from local markets. All samples were preserved in 95% ethanol or a 10% dimethyl sulfoxide solution (Seutin, White, & Boag, 1991) and maintained at room temperature until DNA isolation.

| DArTseq™ 1.0 genotyping
DArTseq™ genotyping (Sansaloni et al., 2011) involves genomic complexity reduction followed by NGS and is similar to other commonly utilized approaches for NGS of reduced genomic representations

| SNP quality filtering
Additional quality filtering of SNP data received from DArT PL was performed in R version 3.3.1 (R Core Team, 2017) using the dartR v0.93 package (Gruber, Unmack, Berry, & Georges, 2018). Loci missing ≥10% of genotype calls were excluded from the dataset. Samples missing ≥20% of genotype calls were also excluded. To retain only high-quality SNPs with reliable genotype calls, loci with average reproducibility <95% were removed. All monomorphic loci were also removed. In instances where more than one SNP originated from a read alignment, a single SNP was randomly retained to reduce the probability of linked loci in the final dataset. Finally, any locus with a minor allele frequency <0.05 across all samples was removed to reduce the probability of PCR error or ascertainment bias resulting from nonrandom sampling of a gene pool (Bradbury et al., 2011;Roesti, Salzburger, & Berner, 2012).

| Identification of genetically distinct populations
We employed multivariate analyses and Bayesian-based simulations to infer the number and geographic extent of genetically distinct populations of striped marlin represented in our dataset.
Multivariate methods were selected for exploring population structure because these methods are computationally efficient and unconstrained by assumptions of Hardy-Weinberg equilibrium (HWE; Jombart, Pontier, & Dufour, 2009 sampling location, and statistical significance of HWE comparisons was determined using a critical value corrected by a modified false discovery rate (Benjamini & Yekutieli, 2001;Narum, 2006). Loci that did not conform to the expectations of HWE in more than one sample collection were removed. All STRUCTURE analyses were performed using an admixture model of ancestry (Falush et al., 2003), a burn-in of 50,000 followed by 500,000 Markov chain Monte Carlo simulations, and three iterations of each K. Default values were used for all other STRUCTURE settings, including the lack of a location prior. Previous evaluations of STRUCTURE performance demonstrate that the presence of strongly differentiated genetic clusters may obfuscate resolution of weakly differentiated clusters (Janes et al., 2017;Vähä & Primmer, 2006;Waples & Gaggiotti, 2006). In preliminary analyses of our data, the highest levels of genetic differentiation were observed between striped marlin sampled from the western Indian Ocean and the northern and eastern Pacific Ocean.  (Jakobsson & Rosenberg, 2007) and visualized in DISTRUCT v1.1 (Rosenberg, 2004). The most likely K for each dataset was identified using Structure Harvester v0.6.94 (Earl & vonHoldt, 2012;Evanno, Regnaut, & Goudet, 2005). Results from multivariate analyses and STRUCTURE simulations were collectively evaluated to determine the most likely scenario of spatial population structure for the striped marlin represented in our dataset. Based on this information, sample collections were combined into groups representing genetically distinct populations.

| SNPs putatively influenced by natural selection
To reduce the probability of committing type I or type II statistical errors (Lotterhos & Whitlock, 2014;Narum & Hess, 2011), we employed two approaches for identifying SNPs putatively under the influence of natural selection using a dataset in which loci not conforming to the expectations of HWE were removed. BayeScan v2.1 (Foll & Gaggiotti, 2008) implements a Bayesian-based algorithm that compares allele frequencies among populations to directly estimate the probability that each locus is exposed to natural selection (Beaumont & Balding, 2004;Foll & Gaggiotti, 2008). We performed BayeScan analyses using 10,000 iterations each for the burn-in, pilot runs, and final runs. We also used conservative prior odds for the neutral model (100:1) to reduce the probability of false positives in BayeScan results (Lotterhos & Whitlock, 2014). F ST outlier loci were identified from BayeScan output using a false discovery rate of 0.10. Loci putatively under the influence of natural selection were also identified using the FDIST2 outlier detection method (Beaumont & Nichols, 1996;Excoffier, Hofer, & Foll, 2009) implemented in Arlequin v3.5 (Excoffier & Lischer, 2010 To assess the relative contribution of SNPs putatively influenced by natural selection to observed population structure, we performed additional multivariate analyses using a dataset limited to F ST outlier loci. We also performed multivariate analyses using a subset of putatively neutral loci that contributed the most information to DAPC clustering. These loci were identified by performing DAPC using the full dataset, then scaling locus loadings, and calculating locus rank percentiles for discriminant functions one and two. Loci with rank percentiles ≥98.7% for each discriminant function were selected to produce a set of putatively neutral loci that corresponded with a similar number of markers as those identified in F ST outlier analyses. Any F ST outlier loci occurring in this putatively neutral set of markers were removed. For both datasets, PCoA and K-means clustering followed by DAPC were performed as described above.

| Genetic attributes of striped marlin populations
Populations of striped marlin resolved in multivariate and STRUCTURE analyses were characterized by assessing genetic diversity and the presence of SNPs exhibiting fixed differences among populations (i.e., private alleles) using a dataset in which loci not conforming to the expectations of HWE were removed. Observed and expected heterozygosities were calculated in the R packages poppR v2.5.0 (Kamvar, Tabima, & Grünwald, 2014) and dartR, respectively.
We used the R package PopGenReport v3.0.0 (Adamack & Gruber, 2014) to calculate rarefaction allelic richness. dartR was used to evaluate populations for the presence of private alleles.
Because inferences of demographic relationships may be biased by loci that deviate from a neutral model of evolution (Beaumont & Nichols, 1996;Luikart et al., 2003), SNPs previously identified as F ST outliers were removed prior to calculating pairwise levels of genetic differentiation and population-level inbreeding coefficients. Levels of genetic differentiation among populations were determined by calculating pairwise measures of F ST in Arlequin. Statistical significance of F ST values was assessed based on 10,000 permutations and a critical value corrected by a modified false discovery rate (Benjamini & Yekutieli, 2001;Narum, 2006). Inbreeding within populations was evaluated by calculating F IS in the R package diveRsity v1.9 (Keenan, McGinnity, Cross, Crozier, & Prodöhl, 2013). Confidence intervals (95%) for estimates of F IS were calculated based on 10,000 bootstrap iterations.
A putatively neutral dataset was also used to infer the degree of genetic connectivity among populations and to identify populations that serve as sources or sinks (Crowder & Norse, 2008;Howe, Davis, & Mosca, 1991), by calculating directional relative migration rates using the divMigrate function (Sundqvist, Keenan, Zackrisson, Prodohl, & Kleinhans, 2016) in the R package diveRsity. This approach provides relative bidirectional estimates of gene flow based on measures of genetic differentiation between populations, and is considerably less computationally intensive than maximum-likelihood or Bayesian methods (e.g., Beerli & Palczewski, 2010;Hey, 2010) for estimating migration rates from large genomic datasets.
We performed divMigrate calculations using all available measures of genetic differentiation so that the consistency of estimates among metrics could be assessed. Confidence intervals (95%) of relative migration estimates were calculated based on 10,000 bootstrap iterations.

| Temporal stability of population structure
To evaluate the temporal stability of allele frequencies within geographically distant regions, we used collections with sample sizes ≥15 individuals per sampling period, and for which temporally

| Population assignment
We assessed the ability of SNPs resolved in this study to accurately assign individuals to populations and to identify a minimum subset of loci for population assignment, using the R package assigner (Gosselin, Benestan, & Bernatchez, 2015). Assignment analyses were performed using a dataset in which loci not conforming to the expectations of HWE were removed. The methodological approach described by Anderson (2010) was implemented by randomly selecting 80% and 20% of samples from each population for training and holdout datasets, respectively. Training samples were used to rank loci based on F ST , and holdout samples were assigned to populations using DAPC as implemented in adegenet. Within DAPC assignment analyses, the function predict.dapc was used to predict posterior membership probabilities to each population for each individual.
We evaluated scenarios where 100-1,000 markers (in increments of 100) were used for population assignment. Analyses were repeated five times for each marker subset, and final results were produced by averaging across iterations.

| SNP quality filtering
The original DArT PL dataset consisted of 61,908 SNP loci (  (Table 2). This dataset is hereafter referred to as the "full dataset."

| Identification of genetically distinct populations
Multivariate analyses and Bayesian simulations were used to delineate populations of striped marlin using the full dataset and a dataset in which loci not conforming to the expectations of HWE were removed, respectively. At least five distinct clusters were resolved on PCoA axes one and two, which collectively explained 6.61% of total genetic variation ( Population-level calculations were performed using these six populations of striped marlin. The eight fish identified as putative migrants were retained with their original sample collections so that biologically realistic assemblages of striped marlin could be characterized; however, some calculations were performed a second time with these samples excluded (described below).   (Table S1).

| Genetic attributes of striped marlin populations
Loci previously found to deviate from a neutral model of evolution (n = 59; described above) were excluded to produce a putatively neutral dataset prior to calculating pairwise measures of genetic differentiation, inbreeding coefficients, and relative migration rates. Pairwise F ST values ranged from 0.0137 between EIO and WSPO, to 0.0819 between WIO and NPO2 (  (Table 3).
Relationships of genetic connectivity among populations inferred by calculating bidirectional relative migration rates (m R ) were similar across all three metrics of genetic differentiation (Jost's D, G ST , and N M ; Alcala, Goudet, & Vuilleumier, 2014;Jost, 2008;Nei, 1973), except the magnitude of these relationships was lower for calculations based on Jost's D. We therefore describe results for only one of these metrics (N M ; Figure 5). The largest relative migration rates F I G U R E 3 Results from discriminant analysis of principal components (DAPC) using the full dataset (n = 4,206 SNPs). (a) Bar plots colored to show posterior probabilities of assignment to a cluster. Scenarios for K equal to four through seven are shown. Horizontal bar at bottom delineates sample collections labeled as in Table 1. Horizontal bar at top delineates clusters corresponding with regional populations. (b) Scatter plot of discriminant functions one and two for scenario with K equal to six from (a). Samples are colored according to the legend. Inertia ellipses for each group are also shown were lower than migration rates between most Pacific Ocean populations. Additionally, relative migration rates between WIO and WSPO (m R ≥ 0.58) were higher than those between EIO and WSPO (m R ≤ 0.54), despite the closer geographic proximity of EIO to WSPO.
Accurate inference of relative migration rates is difficult under scenarios of high gene flow (Sundqvist et al., 2016); this may be particularly so for EIO and WSPO, especially given the small sample size for EIO (n = 8). Migration rates calculated between Indian Ocean and Pacific Ocean populations of striped marlin indicate that genetic connectivity between ocean basins is primarily facilitated by WSPO.
Relative migration rates calculated with putative migrants excluded from analyses produced relationships similar to those described here ( Figure S10).

| Temporal stability of population structure
We assessed the multigenerational stability of allele frequencies

| Population assignment
Overall assignment success was ≥90% (SE ± 2.29-4.55) when subsets of ≥200 SNPs were used to assign individuals to populations.

| D ISCUSS I ON
The primary goal of this study was to resolve spatiotemporal patterns of genomic variation across the full range of a broadly distrib-

| Biological context of genetically distinct populations
This study represents the first assessment of spatial genetic vari- during seasons that at least partially overlap between regions (Bromhead, Pepperell, Wise, & Findlay, 2003;Jones & Kumaran, 1964;Nakamura, 1983;Nishikawa, Kikawa, Honma, & Ueyanagi, 1978;Pillai & Ueyanagi, 1978;Ueyanagi, 1974). Our results demonstrate low degrees of shared ancestry and high levels of genetic differentiation between the population of striped marlin in the western Indian Ocean and populations in the Pacific Ocean.
In comparison, eastern Indian Ocean striped marlin exhibited a close genetic relationship with striped marlin in the western South Pacific Ocean; genetic differentiation between these regions was F I G U R E 5 Bidirectional relative migration rates among striped marlin (Kajikia audax) populations calculated using a dataset where loci not conforming to Hardy-Weinberg equilibrium and selective neutrality were removed (n = 4,106 SNPs). Open circles represent populations and lines connecting circles are weighted according to relative migration rate. Relative migration rates with 95% confidence intervals larger than 0.00 are denoted with an asterisk. Values shown here were calculated with putative migrants included  & Ueyanagi, 1965;Sippel et al., 2007) and occurrence of seasonal assemblages in waters as far south as Tasmania (Bromhead et al., 2003) suggest interoceanic movements of striped marlin around the Australian continent may be possible in at least some years.
Our results clarify previously ambiguous population-level relationships for striped marlin in the Pacific Ocean and are generally consistent with biological information available for this region. We Regional movements of striped marlin generally correspond with the genetically distinct populations described here (Domeier, 2006;Holdsworth et al., 2009;Ortiz et al., 2003;Sippel et al., 2007). The and a third population in the South China Sea (Williams, 2018).
Identifying mechanisms underlying differences in ocean-wide patterns of genetic connectivity among large pelagic fishes requires improved knowledge of species' biological characteristics (e.g., thermal preferences, dispersal capabilities, degree of fidelity to natal spawning grounds) and sensitivity to obvious or cryptic barriers to movement.

| Biological significance of statistically significant comparisons
The populations of striped marlin resolved in this study were separated by levels of genetic differentiation that were also highly statistically significant. However, in some instances, comparatively low levels of genetic differentiation were also sta- Genetic differentiation between populations of marine fishes is expected to be lower than in freshwater and anadromous fishes (Ward, Woodwark, & Skibinski, 1994), presumably due to greater opportunities for dispersal in marine environments. Distinguishing biologically meaningful levels of genetic differentiation from stochastic noise is therefore challenging in marine fishes (Waples, 1998) and may be particularly so for species with enhanced dispersal capabilities, such as large pelagic fishes. Under these circumstances, determining whether a statistically significant test result is also of biological significance may be context-dependent (Waples, 1998).
Further exploration of the levels of genetic differentiation expected for varying experimental designs, species life histories, and genome-wide markers (e.g., Alcala & Rosenberg, 2017;Jost et al., 2018) is necessary for assisting the interpretation of results from genomic studies of broadly distributed and highly migratory large pelagic fishes.

| Relative contributions of neutral and adaptive processes to population structure
Population census sizes (N c ) for many marine fishes are very large relative to terrestrial or freshwater species and may also correspond with comparatively large effective population sizes (N e ; Laconcha et al., 2015;Waples, Grewe, Bravington, Hillary, & Feutry, 2018; but see Hauser & Carvalho, 2008;Palstra & Ruzzante, 2008). For marine populations with large Ne, accumulation of allele frequency differences between populations due to neutral processes may be slow and counteracted by individuals that stray and reproduce in non-natal populations. This may be a greater possibility in marine fishes that are broadly distributed in pelagic environments and have high dispersal capabilities. It is therefore likely that additional mechanisms, such as those corresponding with regional selective pressures, contribute to the accumulation of genetic differences among populations of pelagic marine fishes (Hauser & Carvalho, 2008). In this study, we fishes. Such studies will ultimately inform our presently limited understanding of ecological and environmental variables, phenotypes, and loci involved in regional adaptive processes for large pelagic species.

| Sampling designs for studies of large pelagic fishes
For many highly migratory large pelagic fishes of conservation concern, implementing biologically informed sampling designs  in the central North Pacific and western North Pacific (at least in waters off Hawaii and Japan) may seasonally interact with more than one stock, and approaches that account for mixed-stock fisheries (Crozier et al., 2004)

| Future directions and concluding remarks
This and other recent surveys of genome-wide variation in highly migratory large pelagic fishes present some of the first genomic resources for these species, particularly for those found at comparatively lower frequencies or of lesser commercial value. Results from this study provide evidence for six genetically distinct populations of striped marlin in the Indo-Pacific and are valuable for improving rangewide conservation and management efforts for this species.
Additional work that employs fine-scale spatiotemporal sampling is necessary for identifying regional stock boundaries and whether these boundaries are seasonally dynamic. Such efforts will also help determine the contributions of distinct stocks to mixed-stock fisheries. Studies that implement tagging technology capable of monitoring detailed movements over periods longer than 1 year are presently lacking, but necessary for elucidating seasonal movement patterns within and between ocean basins, and for determining the degree to which populations of striped marlin display spawning site fidelity.
Equivalent genomic and tagging efforts across large pelagic fishes will provide valuable insights into broad-scale patterns of connectivity within and between ocean basins, and ecological and evolutionary factors influencing genetic connectivity in pelagic communities (Hand, Lowe, Kovach, Muhlfeld, & Luikart, 2015;Raeymaekers et al., 2017). The genomics era represents an important opportunity to provide novel information that improves management and conservation initiatives for wild populations (Meek & Larson, 2019), including those of large pelagic fishes, promoting the sustainable use of these ecologically and economically valuable resources.

ACK N OWLED G EM ENTS
We thank several recreational anglers across the Indo-Pacific for

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data available from the Dryad Digital Repository: https ://doi. org/10.5061/dryad.3j9kd 51cp (Mamoozadeh, Graves, & McDowell, 2019).