Strong population structure deduced from genetics, otolith chemistry and parasite abundances explains vulnerability to localized fishery collapse in a large Sciaenid fish, Protonibea diacanthus

Abstract As pressure on coastal marine resources is increasing globally, the need to quantitatively assess vulnerable fish stocks is crucial in order to avoid the ecological consequences of stock depletions. Species of Sciaenidae (croakers, drums) are important components of tropical and temperate fisheries and are especially vulnerable to exploitation. The black‐spotted croaker, Protonibea diacanthus, is the only large sciaenid in coastal waters of northern Australia where it is targeted by commercial, recreational and indigenous fishers due to its food value and predictable aggregating behaviour. Localized declines in the abundance of this species have been observed, highlighting the urgent requirement by managers for information on fine‐ and broad‐scale population connectivity. This study examined the population structure of P. diacanthus across north‐western Australia using three complementary methods: genetic variation in microsatellite markers, otolith elemental composition and parasite assemblage composition. The genetic analyses demonstrated that there were at least five genetically distinct populations across the study region, with gene flow most likely restricted by inshore biogeographic barriers such as the Dampier Peninsula. The otolith chemistry and parasite analyses also revealed strong spatial variation among locations within broad‐scale regions, suggesting fine‐scale location fidelity within the lifetimes of individual fish. The complementarity of the three techniques elucidated patterns of connectivity over a range of spatial and temporal scales. We conclude that fisheries stock assessments and management are required at fine scales (100 s of km) to account for the restricted exchange among populations (stocks) and to prevent localized extirpations of this species. Realistic management arrangements may involve the successive closure and opening of fishing areas to reduce fishing pressure.


| INTRODUCTION
Nearshore coastal ecosystems provide an important source of food for human populations, supporting 90% of the global wild fish harvest whilst accounting for only 7% of the ocean worldwide (Pauly et al., 2002). As global demand for seafood increases, many exploited species in these regions have declined and large numbers of fisheries are currently fished at unsustainable levels (Jackson et al., 2001;Smith et al., 2010). Sustainable management of exploited fish populations requires detailed knowledge of population structure, natural abundance and the degree of ecologically relevant exchange among harvested stocks (Pauly et al., 2002). The concept of "stocks" as population units has long been anchored in fisheries science and can be defined as groups of fish within a species that are self-recruiting; share similar growth rates and rates of natural and non-natural mortality; and may react more or less independently to harvesting (Cadrin, Kerr, & Mariani, 2013).
Methods for delineating stocks have advanced considerably in recent years and include genetic techniques, acoustic telemetry, tagging, otolith chemistry, demographic analysis, otolith shape and meristic data (Hawkins et al., 2016). Genetic approaches speak to both inter-and intragenerational timescales as they track the lifehistory stages of the fish (i.e., from fertilized eggs to adults) that are interchanged between locations and which subsequently participate in successful spawning. Otolith multi-elemental signatures and parasite assemblages are ecological markers recording the environment the juvenile and adult fish inhabit and reflect processes occurring within the individuals' lifetime. The integration of multiple techniques that operate over different temporal and spatial scales makes it possible to overcome many of the limitations of single technique approaches and greatly strengthens the inference available from stock structure studies (Abaunza et al., 2008;Begg & Waldman, 1999;Lleonart & Maynou, 2003;Waldman, 1999;Welch et al., 2009Welch et al., , 2015. As an example, Izzo et al. (2017) used an integrated approach to reveal four stocks of the sardine Sardinops sagax in Australian waters when it was previously considered a semicontinuous meta-population.
Fishes of the family Sciaenidae, commonly known as croakers, are important components of commercial, recreational and indigenous fisheries in tropical and temperate regions worldwide (Lenanton & Potter, 1987). They are targeted for their flesh and, increasingly, for their swim bladders, which are sold fresh or dried in South-East Asia (Ghosh et al., 2009;Sadovy & Cheung, 2003;Tuuli, 2010). Many species of Sciaenidae have declined in recent decades due to overexploitation, and several are now considered threatened. The blackspotted croaker Protonibea diacanthus (Lacapede, 1802) is a large species (>1.5 m max. total length; up to 42 kg mass) that is widely distributed throughout coastal waters and estuaries of the tropical Indo-West Pacific (Sasaki, 2001). In Australia, it is distributed along the northern coast from Shark Bay, Western Australia to Hervey Bay, Queensland (Bray, 2011). The species grows rapidly and has a maximum recorded age of 13 years (Phelan, 2008). Spawning of P. diacanthus in Australia occurs from August to December, and they appear to produce pelagic eggs and have a pelagic larval phase (Froese & Pauly, 2016;Leis & Carson-Ewart, 2000). Whilst inshore seasonal breeding aggregations of P. diacanthus are thought to occur (Welch et al., 2014), direct evidence of the behaviour of the species is limited (Phelan, 2008) and little is known about the size and locations of the aggregations. Commercial and recreational sectors both contribute to total landings (667 tonnes in 2005), but catches have increased recently due primarily to growth in the commercial sector (from 43 to 250 tons per annum between 1995 and 2005; Coleman, 2004;Phelan & Elphick, 2006). Intensive recreational fishing on the east coast of Queensland (Bowtell, 1995(Bowtell, , 1998 has also been attributed with reducing fish abundance and size to the point where the catches are almost exclusively limited to immature fish. There is also evidence of over-exploitation of P. diacanthus outside of Australian waters: overfishing in India in the 1980s led to local extirpation (James, 1994) and a fishery that once thrived in Hong Kong no longer exists (Sadovy & Cheung, 2003).
Management practices to support the sustainable harvest of P. diacanthus in Western Australia, Queensland and the Northern Territory are currently hindered by a lack of knowledge of the species' population structure. In this study, we examined the population structure broad-scale regions, suggesting fine-scale location fidelity within the lifetimes of individual fish. The complementarity of the three techniques elucidated patterns of connectivity over a range of spatial and temporal scales. We conclude that fisheries stock assessments and management are required at fine scales (100 s of km) to account for the restricted exchange among populations (stocks) and to prevent localized extirpations of this species. Realistic management arrangements may involve the successive closure and opening of fishing areas to reduce fishing pressure.

K E Y W O R D S
croaker, fisheries management, otolith chemistry, parasites, population genetics, stock discrimination of the species across north-western Australia with the expectation of spatially distinct stocks, which would be consistent with the observation of localized depletion. Moreover, we aimed to test the expectation that genetics provides information on a broader spatial scale than otolith chemistry and parasite data in this coastally distributed species. To do so and to demonstrate the power of integrating across techniques, we contrasted population structure results from the three methods and focused on one example where the assignment of individuals to adjacent populations was method-specific. Finally, we use our results to address the appropriate spatial scale for stock assessment and suggest fisheries management arrangements for P. diacanthus across north-western Australia.

| Samples
A total of 298 fish were collected from 11 sampling locations across north-western Australia from Roebuck Bay (Western Australia) to the Vanderlin Islands in the Gulf of Carpentaria (Northern Territory) ( Figure 1, Table 1). The decline of the species (Phelan, 2002) meant that samples could not be obtained from north-eastern Australia.
The lack of knowledge on the location and size of the aggregation grounds in P. diacanthus did not allow the exclusive collection of mature fish on spawning grounds, and thus, fish of different size were included in the analysis. Samples were collected with the assistance of Indigenous marine rangers, fishing tour operators and recreational and commercial fishers. The total length (TL), standard length (SL) and sex of each specimen were recorded. For genetic analyses, muscle tissue or fin clips were sampled and immediately placed into vials containing molecular grade 95% ethanol or 20% dimethyl sulfoxide (DMSO) solution in 5 mol/L NaCl and stored at 4°C in the field and −20°C in the laboratory. The sagittal otoliths were dissected from each fish, cleaned and rinsed thoroughly, dried, and stored in paper envelopes prior to preparation for trace element analysis. Gills and internal body organs were removed, placed into individual bags and frozen until examination for parasites. The same individuals were used for otolith microchemistry, genetic and parasite analyses.

| Data analysis
Population structure was tested in an hierarchical fashion, with genetics being deployed at the broad scale to guide population structure and ecological markers used at the finer scale. The nature of otolith microchemistry reflects fine-scale spatial variation between environments and is not useful for broad-scale variation. Indeed, lack of significance (i.e., otolith similarity) on a broad spatial scale does not indicate ecological population homogeneity. Fine-scale genetic analyses were used to integrate across the three methods for a pair of locations in NW Australia for which one presented a striking genetic heterogeneity.
Genomic DNA from all samples for genotyping was extracted using ISOLATE II Genomic DNA Kit (Bioline) following the manufacturer's instructions to produce 100 μl of eluted DNA from each sample. All the F I G U R E 1 Location of the 11 sampling sites across northern Australia showing the two jurisdictions (Western Australia and Northern Territory). The details for each of the sampling locations are provided in Table 1 DNA extracts were quantified using the Qubit v3 (ThermoFisher) fluorometric machine. Eleven species-specific microsatellite loci (Prd012,Prd023,Prd044,Prd042,Prd018,Prd045,Prd046,Prd020,Prd036,Prd049,Prd024) were genotyped across all the samples as part of multiplexes as described in Taillebois et al. (2016).
The potential for null alleles, large allele dropout and stuttering to interfere with scoring accuracy was checked for each microsatellite locus in each sample using MICROCHECKER version 2.2.3 (Van Oosterhout, Hutchinson, Wills, & Shipley, 2004). Summary statistics for microsatellite loci, including the number of alleles, allelic richness, expected and observed heterozygosity and fixation indexes, were obtained for each sampling location using GENALEX version 6.502 (Peakall & Smouse, 2006). Tests of conformance of genotypic proportions to Hardy-Weinberg equilibrium expectations were carried out for each locus at each sample location, and tests of genotypic equilibrium between pairs of microsatellites (linkage disequilibrium) were carried out for each sample location, using an exact probability test as implemented in GENEPOP version 4.5 (Rousset, 2008). The exact test was estimated using a Markov Chain that employed 1,000 dememorizations, 500 batches and 1,000 iterations per batch. We accounted for the linkage disequilibrium multiple testing by applying a Bonferroni correction to the p-value. A test with a p-value <.0045 (0.05/number of loci considered in each multiple test) was required to be considered significant.
In order to assess whether locations could be treated as independent genetically cohesive groups on a broad spatial scale, and to test for the extent of admixture within the groups, two contrasting clustering approaches were used. Population assignment and clustering was performed using the Bayesian model-based clustering program STRUCTURE version 2.3.4 (Pritchard, Wen, & Falush, 2003). For comparative purposes on the broad spatial scale, a discriminant analysis of principal components (DAPC, Jombart, Devillard, & Balloux, 2010) available in the adegenet package (Jombart, 2008) for R version 3.3.1 (R Core Team, 2017) was also used. An hierarchical approach was subsequently taken to explore potential substructure within the groups revealed by the STRUCTURE broad-scale analysis. The hierarchical analyses were also performed using STRUCTURE. Methodology for STRUCTURE and DAPC is detailed in Appendix A1 in Appendix S1.
To test alternate post hoc observations of genetic structure revealed by our different approaches, we assessed the per cent of genetic variance explained by the groupings deduced from STRUCTURE and DAPC results using an analysis of molecular variance (AMOVA) as implemented in ARLEQUIN. The significance of any differentiation was determined by permutation of 22,000 replicates.
Pairwise F ST values were calculated (i) to further investigate the population structure within the groups revealed with the clustering approaches and (ii) to test whether P. diacanthus is constrained by a pattern of isolation by distance (IBD). Fixation indices (F ST ) between pairs of sample localities were estimated as implemented in ARLEQUIN version 3.5.2.2 (Excoffier & Lischer, 2010). We performed a Mantel test To investigate the fine-scale population structure and provide information on the dispersal of fish, the microchemistry of the otolith was examined in two sections of the otoliths: the near core and the margin. The near core area is located just outside the first opaque zone ~500 μm from the core of the otolith and represents the juvenile phase; and the otolith margin (edge) was sampled adjacent to the sulcus acusticus to represent growth in the period prior to capture. A total of 11 trace elements ( 7 Li, 25 Mg, 23 Al, 49 Ti, 53 Cr, 55 Mn, 60 Ni, 63 Cu, 66 Zn, 88 Sr, 138 Ba) and the internal standard ( 43 Ca) were analysed from the two ablation zones in each otolith. Details on the general methodology for the otolith preparation, the laser ablation and data transformation are presented in Appendix A2 in Appendix S1. For all elements, the ratio of element isotope intensity to Ca intensity was used to estimate the element: 43 Ca ratio. These ratios were converted to molar T A B L E 1 Details of the 11 sampled locations of Protonibea diacanthus across northern Australia  All data analysis was performed using R (R Core Team, 2017).
Pairwise distributions of all variables used in each analysis for the parasite and otolith data sets were inspected for extreme outliers, unsatisfactory covariate distributions, and the presence of collinearity. Where necessary abundance data were log(Y + 1) transformed. Homogeneity of the covariance matrices for the grouping factor was assessed using Box's M-test based on the chi-square approximation with the biotools package in R (da Silva, 2016).
Fish are known to accumulate longer-lived parasite species with age (Rohde, 1982), which will affect differences in parasite fauna if samples contain individuals of different ages. All parasite species that were selected on prevalence were examined for their correlation with host length (as an indicator of age) using linear models.
Where length was significantly correlated with abundance, numbers greater than zero were adjusted to the mean host size (total length) using the methods described in Moore, Buckworth, Moss, and Lester (2003).
Spatial variation in parasite assemblages and the otolith near core and margin chemistry among regions and locations within regions were investigated separately using single-factor multivariate analysis of variance (MANOVA). Linear discriminant function analysis (LDFA) was conducted to provide a statistical and visual indication of the similarities of either the parasite communities or the multi-elemental otolith microchemical signatures among samples at the regional spatial scale using the MASS package in R (Venables & Ripley, 2002). LDFA classification success rates and an associated proportional chance criterion (the expected proportion of correct classification by chance alone) were calculated (Poulin & Kamiya, 2015) for each location within the regions. To show the separation achieved for the groups in the analysis, the first two discriminant functions for each individual were plotted and the 95% confidence ellipses around the centroid means of the first two discriminant functions for each group in the sample using the ellipse package in R (Murdoch & Chow, 2013).
To test the effect of distance between locations on the parasite assemblages, and the otolith elemental concentrations in the near core and the margin, we performed a Mantel test of the Jaccard dissimilarity using the mean differences in number and concentrations, respectively, between each pair of locations versus geographic distances among locations using the vegan package in R (Oksanen et al., 2017).
The individual assignment of fish was tested through a Bayesian approach across all fish for otoliths and parasites. Classes and posterior probabilities were calculated by jackknife cross-validation to assign individuals to most probable locations within each region. In a couple of locations where the genetic structure gave evidence of genetic heterogeneity within locations, STRUCTURE was used to calculate posterior probabilities and assign individuals to most probable locations. Individual mis-assignments were compared across techniques to test whether each technique provided different information or not (i.e., see whether the same fish were mis-assigned to the same locations between techniques).

| RESULTS
Genotypes from 11 microsatellite loci were obtained for 284 individuals of P. diacanthus. The level of missing data was low (2.15%) and as such was retained and identified as missing data in the subsequent analyses (missing data were calculated as the number of genotypes that could not be scored over the total number of genotypes present in the ideal data set, 11 × 284).  AMOVA was used to compare the different scenarios supported by STRUCTURE and DAPC. The 4-and 5-group scenario highlighted by STRUCTURE and DAPC, respectively, explained 2.15% (p-value = .000) and 2.28% (p-value = .000) of the total genetic variation. As both scenarios explained very similar among group genetic variation, they are both likely to represent biological population structure. However, the 4-group scenario revealed by STRUCTURE will be used in the discussion, as it is the result of unsupervised analyses, which was not the case for DAPC.   Figure 4). Average classification success in the LDFA among all locations was 54% for the otolith margin, 31% for the otolith near core and 67% for the parasites (Table 4) and was much higher than expected by chance (9% for the otolith and 10% for the parasites), suggesting a nonrandom distribution of the otolith microchemistry and parasite abundance variability across the sampled range. Classification success for both margin (71% for the Western region, 57% for the Darwin region and 74% for the Arnhem/Gulf region) and near core (57% for the Western region, 44% for the Darwin region and 38% for the Arnhem/ Gulf region) was higher when the data for each region were analysed separately, but the % classification success relative to the % expected by chance was higher when all sites were analysed simultaneously (Table 4, Figure 4). Similar to the otolith chemistry analysis, separation of the analyses into regions increased reclassification success in parasite abundances, with 81% for the Western region, 56% for the Darwin region and 89% for the Arnhem/Gulf region (Table 4, Figure 4).
For both otoliths and parasites, the classification success was greater for the Arnhem Gulf and Western regions than for the Darwin region.

BIC
Classification success was higher for the margin than the near core across all locations (except Wa and OD in the Darwin region), reflecting the fact that the margin sample incorporated growth at the sampling location immediately prior to collection (i.e., samples collected from the same location in the same year). In contrast, the near core (juvenile phase) samples incorporated growth from different years because fish in each location were of variable ages and sources may vary by age and cohort. Individual mis-assignments were compared between near core and margin, and there was no evidence of matching pattern of mis-assignment (Appendix B5 in Appendix S2). Fish that were mis-assigned using the near core were not necessarily mis-assigned using the margin; and if they were, the assigned location was not necessarily the same. Classification success was generally higher for the parasites than the otolith except BI and Wy where both margins and near cores had higher classification rates than parasites; and PI and AS where only the margins had higher classification rates than the parasites. Individual mis-assignments were compared between parasites and margin, and there was no evidence of matching pattern of mis-assignment (Appendix B5 in Appendix S2). Fish that were misassigned using the parasites were not necessarily mis-assigned using the margin, and if they were, the assigned location was not necessarily the same. locations with individuals presenting matching genotype proportions to individuals from CS location (Figure 2a). Further full Bayesian assignment tests between RB and CS revealed that individuals having outlier genetic signature did not necessarily have distinct otolith or parasite characteristics. Of the three individuals that were mis-assigned with the genetics, fish 3167 and 3540 were also mis-assigned using the near core and the parasites, respectively (Table 5). Assuming parasites and otolith margins provide information about postlarval development; and near core and genetic provide information about the location of early stage development, we tested the hypotheses that otolith near core would correspond to the genetics; and that parasites and otolith margin would provide similar assignments too. Off the eight fish that were mis-assigned using the near core, only one was similarly mis-assigned for the genetics (fish 3167). Of the four fish that were mis-assigned for the margin and the two for the parasites, only one was mis-assigned for both parasites and margin (fish 3518).

| DISCUSSION
In this study, the combined application of genetic markers, otolith chemistry and parasites across the same specimens provided three F I G U R E 4 Plots of the first two discriminant function scores showing spatial variation in the parasite assemblage and the multi-elemental otolith near core and margin signatures of Protonibea diacanthus collected from 11 locations within three regions: Western region, Darwin region and Arnhem/Gulf region. Ellipses are 95% confidence intervals around the group centroid for each location within each region, and data points represent individual fish independent lines of evidence and considerable power to detect the highly complex population structure of P. diacanthus across northern Australia. Relatively few studies to date have used this integrated approach to investigate population structure and connectivity or for stock identification (Abuanza et al., 2008;Welch et al., 2015;Izzo et al., 2017). These studies also concluded that the integrated use of multiple techniques considerably improved confidence in stock identification using index of stock differences (ISD) thresholds definition (Izzo et al., 2017;Welch et al., 2015) and were particularly effective for species with broadly dispersing larvae or adults that exhibit little or no genetic structure across their range (Waldman, 1999). As a case in point, our study revealed the existence of major genetic breaks along the north-western Australian coastline and thus the presence of several discrete breeding populations of P. diacanthus. Prior to settlement, genetic results suggest that currents and geographic barriers to dispersal may limit exchange among collection locations. The ecological markers (otoliths and parasites) strongly suggest that P. diacanthus is generally sedentary following larval settlement.

| Site fidelity and dispersal
Whilst the results of this study suggest high levels of site fidelity after  , 1998). Under these three scenarios, dispersal of fish among locations would not be necessary to explain the relationship between parasite assemblage similarity and geographic distance.
In the case of otolith chemistry, overlap in multi-elemental signatures among locations is likely to be driven by local geochemical characteristics, including the underlying seabed and nearby river catchment geology, as well as physicochemical factors such as water temperature and salinity and their effects on fish physiology and local ecological processes (Campana & Thorrold, 2001;Elsdon & Gillanders, 2002, 2004. Whilst such factors could potentially vary systematically over geographic gradients, we found no evidence of a relationship be- As previously mentioned, the generally lower classification success in the near core compared to the margin can, at least in part, be attributed to the fact that the near core samples incorporated growth from different years (i.e., different age cohorts) within each location (see also Gillanders, 2002). It is also possible that movement by fish T A B L E 4 Jackknife reclassification success of the linear discriminant function analysis (DFA) for the overall otolith near core and margin chemistry and parasite assemblage of Protonibea diacanthus Data are presented as the % of individuals that reclassify to their collection location. Poulin and Kamiya's (2015) proportional chance criterion is shown in bracket after the total classification success within each region.
T A B L E 5 Assignments of Protonibea diacanthus individuals between Roebuck Bay (RB) and Camden Sound (CS) locations based on a after the juvenile phase accounts for some the reduced classification success in the near core analyses. Yet despite the potential for variability among fish of different ages within a location, our analyses consistently revealed highly significant spatial differences in otolith chemistry signatures among locations. This finding agrees with our conclusion of location-level population structure based on parasite assemblages and, further, suggests that at least some spatial structuring is maintained from the early stages of the life history of P. diacanthus.
Even though ages were available for our samples, the low number of samples per location combined with the wide range of ages resulted in low numbers per cohort. Hence, cohorts could not be compared in this study, and further research is required to determine the extent to which mis-classification of fish can be attributed to temporal variation in otolith chemical signatures versus other potential causes (e.g., immigration from other locations).

| Population structure and connectivity
At broader spatial scales, the genetic analyses revealed the existence of major discontinuities along the north-western Australian coastline, resulting in several genetically discrete populations of P. diacanthus.
This finding is based on mixed age samples, and we would predict stronger spatial structuring if mature fish were collected on spawn-  Figure 1) (Semeniuk, 1996;Thackway & Cresswell, 1998) differ significantly in mangrove characteristics. The tip of the Dampier Peninsula has been identified as an important biogeographic break for marine species (Hutchins, 2001;Travers, Potter, Clarke, Newman, & Hutchins, 2010;Wilson, 2013) and coincide with a pronounced shift in the underlying geology (e.g., from sedimentary sandstone in the north to unconsolidated sand and silt in the south) and associated dominant benthic habitats (e.g., from coral reefs in the north to soft substrate communities in the south). This barrier also coincides with a prominent increase in tidal currents and associated water turbidity to the north.
The Kimberley-Canning border has the largest tropical tidal range (~12 m) and some of the fastest tidal currents in the world (2.5 m/s), and experiences massive volumes of freshwater in a highly turbid plume from the Fitzroy River during the monsoonal wet season (Wolanski & Spagnol, 2003). The discontinuity between the Kimberley and Canning geomorphology (macrotidal, jagged "ria" coastline with extensive mangrove forests vs smoother coastline with no rivers) creates contrasted marine coastal habitats, inhibiting larval dispersal and reinforcing spatial population structure. Likewise, quantitative analyses of fish assemblages have revealed two main biogeographic regions, the Kimberley and Canning coasts (Hutchins, 2001;Travers, Newman, & Potter, 2006;Travers, Potter, Clarke, & Newman, 2012;Travers et al., 2010;Wilson, 2013) (Broderick et al., 2011).
The offshore population of the AS off the coast of Arnhem Land and close to the Wessel Islands was a genetically isolated population.
The distance from the shore and the greater depths may prevent this population from exchanging individuals with coastal locations, which would explain the distinctiveness of that population. However, there is a significant information gap about this population and its degree of dependence upon other inshore populations of Australia or Papua, which requires further investigation.

| Integration across methods: example of RB and CS locations
Despite the overall evidence of a strong genetic break, Bayesian analysis at the individual level showed that RB consisted of an admixed population, with some fish having genotypes similar to CS individuals ( Alternatively, the genetically distinct immigrants could be the descendants of true immigrants-hence, they would be genetically distinct, but have the local otolith (near core and margin) and parasite signatures. The absence of a matching pattern between genetics and ecological methods in individual mis-assignment reveals that the genetically mis-assigned individuals from RB may (i) not be contemporary immigrants but rather a genetic signal of ancestral immigrants or (ii) be contemporary immigrants from sites not sampled in the study. Fine-scale sampling and temporal series collection would be required to further investigate the processes that drive the genetic composition of fish in the RB population.

| Management implications
The highly complex population structure of P. diacanthus shown by this study across northern Australia emphasizes the likely vulnerability of the species to overfishing, and importantly, guides the appropriate spatial scale for stock assessment and fisheries management. In particular, the population genetics and otolith and parasites data suggest that stock assessments and fisheries management on a spatial scale of as little as 100 s of km may be justified. Given genetic breaks among broad regions, it is unlikely that immigration of recruits from other regions would contribute to population recovery in the short term should one or more regions suffer severe population declines. At a finer scale, the parasite and otolith chemistry analyses suggest that movement of recruits among sites (e.g., reef complexes) is generally limited and that heavily depleted sites may be highly reliant on self-recruitment for population maintenance and, therefore, may be slow to recover once depleted. Within the Western Australia fisheries management jurisdiction, RB, CS and Wy were genetically distinct from each other, indicating that locationspecific stock assessments and location-specific management arrangements for P. diacanthus need to be considered by fisheries managers, or alternatively that the location-specific "stocks" identified are considered within assessment, monitoring and management frameworks.
Protonibea diacanthus is targeted by commercial, recreational and indigenous fisheries, which makes population modelling challenging over limited spatial scales of 100 s of kms. The difficulty with modelling over this spatial scale is the logistical challenges associated with collecting matching stock-specific data such as reproduction and spawning dynamics, growth rates, exploitation rates and natural mortality. This logistical challenge comes with a significant cost burden. These factors indicate that fine-scale management is problematical once the additional cost of compliance, regulation and enforcement is added. Within that framework, consideration needs to be given to a range of other options that limit fishing effort, including limited entry, limiting fishing effort, consideration of targeted fisheries spatial closures, rotational harvests or a combination of these approaches with other traditional fisheries management arrangements (e.g., co-operative management of local fisheries with Indigenous people). This approach has been used in the Northern Territory where a series of reef-scale closure areas have been implemented which has subsequently shown to aid the recovery of P. diacanthus and Lutjanus johnii (NTG 2016). Consideration is also being given to developing fine-scale areal catch limits for P. diacanthus based on information from this study (Thor Saunders pers. com) demonstrating that management at the scale of populations is achievable.

| CONCLUSIONS
Several lines of evidence support a model of low dispersal in P. diacanthus.
Significant spatial genetic population structure was supported by finer spatial-scale parasite assemblage and otolith chemistry data. There is evidence of a small amount of migration-most likely in the pelagic larval phase-from adjacent populations corresponding to prevailing currents, which was inferred from integration across data from three sources.
Our finding that P. diacanthus is highly structured across northern Australia from RB to the VI has demonstrated that the combination of ecological and genetic markers provides a more holistic view of population structure when deployed together on the same set of individuals. Genetics provides the overall framework for applying ecological methods and provides also considerable information about exchange between populations. Sensitive ecological methods, such as otolith chemistry and parasites, provide resolution of the fine-scale spatial separation within and between collection locations, but are unable to provide broader scale information over intergenerational timescales. Our results collectively demonstrate location-specific fidelity in the juvenile and adult phases and suggest that dispersal is likely to occur primarily during the very early life history as eggs and/or pelagic larvae. Further research (e.g., acoustic telemetry) would be required to directly quantify the extent of site attachment across the life history of P. diacanthus.
The finding that P. diacanthus across north-western Australia consists of a set of spatially structured populations with low connectivity sets a framework to develop approaches for the assessment and spatial management of the species. The collapse of the North Queensland stock (Phelan, Gribble, & Garrett, 2008) and stocks in India and China (James, 1994;Sadovy & Cheung, 2003) demonstrates that P. diacanthus is prone to local depletion and our findings suggest that this vulnerability is likely due to its low dispersal and high location-specific fidelity. Stock assessment models for this species cannot assume free and frequent movement of individuals among the genetic populations stock boundaries. Because population recovery will mainly depend on limited interchange of early life-stages such as eggs or larvae between neighbouring populations (stocks), it may take multiple generations to replenish depleted stocks and there is potential that some heavily depleted stocks may not recover at all.

This work was conducted under Charles Darwin University Animal
Ethics permit A13014. Voucher specimens of parasites will be deposited