Extreme seascape drives local recruitment and genetic divergence in brooding and spawning corals in remote north‐west Australia

Abstract Management strategies designed to conserve coral reefs threatened by climate change need to incorporate knowledge of the spatial distribution of inter‐ and intra‐specific genetic diversity. We characterized patterns of genetic diversity and connectivity using single nucleotide polymorphisms (SNPs) in two reef‐building corals to explore the eco‐evolutionary processes that sustain populations in north‐west Australia. Our sampling focused on the unique reefs of the Kimberley; we collected the broadcast spawning coral Acropora aspera (n = 534) and the brooding coral Isopora brueggemanni (n = 612) across inter‐archipelago (tens to hundreds of kilometres), inter‐reef (kilometres to tens of kilometres) and within‐reef (tens of metres to a few kilometres) scales. Initial analysis of A. aspera identified four highly divergent lineages that were co‐occurring but morphologically similar. Subsequent population analyses focused on the most abundant and widespread lineage, Acropora asp‐c. Although the overall level of geographic subdivision was greater in the brooder than in the spawner, fundamental similarities in patterns of genetic structure were evident. Most notably, limits to gene flow were observed at scales <35 kilometres. Further, we observed four discrete clusters and a semi‐permeable barrier to dispersal that were geographically consistent between species. Finally, sites experiencing bigger tides were more connected to the metapopulation and had greater gene diversity than those experiencing smaller tides. Our data indicate that the inshore reefs of the Kimberley are genetically isolated from neighbouring oceanic bioregions, but occasional dispersal between inshore archipelagos is important for the redistribution of evolutionarily important genetic diversity. Additionally, these results suggest that networks of marine reserves that effectively protect reefs from local pressures should be spaced within a few tens of kilometres to conserve the existing patterns of demographic and genetic connectivity.

The coral reef systems of the Kimberley in north-west Australia are a biophysically unique centre of coral biodiversity at the southern margin of the East Indies Coral Triangle (Wilson, 2013) and are among the world's most remote and least degraded ecosystems (Halpern et al., 2008). This region may also play an important role as a tropical refuge for photosymbiotic benthic fauna (Richards et al., 2019). However, some inshore Kimberley reefs bleached for the first time in 2016 (Gilmour et al., 2019;Hughes et al., 2018), highlighting that even these reefs that are far from urban centres and agricultural influences are susceptible to global warming. Macrotides (up to 12 m) combine with complex geomorphology to create powerful currents (>1m s −1 ; Ivey et al., 2016), which could be either strong conduits or barriers to dispersal of larvae among local populations.
These reefs also experience large variations in temperature, turbidity, nutrient concentrations and aerial exposure (Jones, Patten, et al., 2014;Richards, Garcia, Wallace, Rosser, & Muir, 2015;Schoepf, Stat, Falter, & McCulloch, 2015;Wilson, 2013). Limited cross-shelf and long-shore circulation (D'Adamo, Fandry, & Domingues, 2009;Treml & Halpin, 2012) suggest inshore populations are isolated from others in the region. Theory predicts that physical isolation coupled with strong selection pressures from extreme environmental heterogeneity will produce unique patterns of inter-and intra-specific genetic diversity and structure in populations (Felsenstein, 1976). This prediction has not been well tested in the Kimberley for reef-building corals, but records of new species (D. Jones, Patten, Bryce, Fromont, & Moore, 2014; and unique species/habitat associations (Richards, Bryce, Bryce, & Bryce, 2013) are beginning to substantiate this expectation.
Knowledge of larval connectivity is fundamental to spatial planning for coral reef conservation because it is a key ecological driver of population replenishment and recovery after disturbance (Cowen & Sponaugle, 2009). There is currently limited knowledge of metapopulation dynamics of most reefs and species, and even less understanding of how to integrate connectivity information into ecosystem management (Magris et al., 2014;Underwood, Wilson, Ludgerus, & Evans, 2013). Because genetic divergence among individuals and populations accumulates over multiple generations through genetic drift and differential selection when inter-breeding is restricted, spatial analysis of genetic structure is a pivotal method for measuring connectivity (Hedgecock, Barber, & Edmands, 2007).
This study characterized the genetic diversity and connectivity within and among populations of Acropora aspera (Dana, 1846) and Isopora brueggemanni (Brook, 1893) from Kimberley reefs of northwest Australia. Both these species are widespread branching corals that provide the three-dimensional habitat for many coral reef organisms throughout the Indo-Pacific. Although they both belong to the family Acroporidae, these two species differ in modes of reproduction. Acropora aspera is a broadcast spawner, releasing eggs and sperm into the water column where fertilization and larval development occur. The larvae spend a few days in the plankton before they are competent to settle (Appendix A). In contrast, I. brueggemanni is a brooder. Fertilization and larval development occur within the polyp before larvae are released at an advanced developmental K E Y W O R D S Acropora aspera, conservation genomics, Isopora brueggemanni, marine reserve networks, population connectivity, single nucleotide polymorphism stage capable of settling within a few hours (Appendix A). Both species are listed as vulnerable on the IUCN Red List of Threatened Species based on their geographic range and their susceptibility to bleaching and disease (Aeby et al., 2014;Richards et al., 2008).
Here, we investigated the eco-evolutionary processes that sustain the metapopulations of A. aspera and I. brueggemanni in north-west Australia by genotyping thousands of single nucleotide polymorphisms (SNPs) isolated from across their genomes. We first tested for cryptic diversity within samples identified as A. aspera or I. brueggemanni. We then measured the spatial distribution of genetic diversity at inter-archipelago (tens to hundreds of kilometres), inter-reef (kilometres to tens of kilometres) and within-reef (hundreds of metres to kilometres) scales to determine the relative strength of genetic connections. Finally, we explored key seascape drivers of metapopulation structure by testing whether heterogeneity in environmental factors such as temperature, turbidity and tide was associated with genetic differentiation and diversity of local coral populations.

| Sampling design
We sampled a range of spatial scales (Figure 1) We collected samples by walking on exposed platforms at spring low tides and removing one-centimetre fragments from coral colonies. Fragments were preserved in 100% ethanol. We photographed colonies and collected representative voucher specimens for taxonomic verification. We collected 534 Acropora aspera samples from 15 sites (between 24 and 83 colonies per site; Table 1) and 612 Isopora brueggemanni samples from 18 sites (between 20 and 60 samples per site; Table 2).

| SNP development, QC and diversity
We extracted genomic DNA from coral specimens using a saltingout protocol modified from Cawthorn, Steinman, and Witthuhn (2011) and purified with Zymo-Spin I-96 Filter plates. Genome-wide SNP data were generated using the next-generation sequencing platform and the DArT-seq protocol. DArT-seq is similar to other site-associated restriction enzyme-based library preparation methods (e.g. RAD-seq) and is a widely applied approach for exploring population genetic structure in species that lack genome assemblies (DiBattista et al., 2017;Pazmino, Maes, Simpfendorfer, Salinas-de-Leon, & van Herwerden, 2017;Thomas et al., 2020). Sequencing was carried out on an Illumina HiSeq 2,500 using 75-cycle singleend reads. Raw reads were processed using DArT's proprietary variant calling pipeline, DArTsoft-14. The call quality of the initial SNP data set was further assured by setting a cut-off of read depth per locus (coverage) <7, call rate >0.35 and minimum allele frequency >0.00075 for Isopora and >0.0017 for Acropora (further details of DArT-seq protocol in Appendix B). This development phase indicated the presence of highly divergent genetic lineages within A. aspera. We subsequently applied a stringent filter to the data to isolate loci suitable for inter-specific analysis. From the primary data set of 34,304 SNPs, we used adegenet (Jombart, 2008) and the dartR package (Gruber, Georges, Unmack, & Berry, 2017) to filter using call rate >0.95, coverage >20, minimum allele frequency >0.05 and max heterozygosity <0.75. In addition, we used the reproducibility statistic to filter out all loci with <0.999 correct calls across individuals.
These filters were chosen to minimize genotyping noise such as null alleles brought about by differences in the target sequences among divergent genetic groups. The final filtered A. aspera data set comprised of 585 SNPs. However, to make sure this stringent set of loci did not bias differentiation estimates, we also conducted our interspecific analysis with relaxed filters (call rate >0.80, coverage >20, a minimum allele frequency >0.01 and reproducibility >0.999). We did not filter for Hardy-Weinberg or gametic-phase disequilibrium at this stage of the analysis because large (potentially inter-specific) divergence would be associated with such disequilibrium, and removal of such markers would likely limit power of the analyses. Seven individuals with more than 15% missing data were removed.
We identified four distinct lineages in the A. aspera samples that often occurred in sympatry (see results). Due to low sample sizes in three of the four lineages, we only conducted population-level analyses on the most common and widespread lineage (Acropora asp-c). To  (Gonzalez et al., 2007), adegenet and pegas (Gonzalez et al., 2007;Paradis, 2010 Four loci were identified as outliers, and these were removed from subsequent analyses resulting in final data set of 2,894 loci. There was no indication of cryptic diversity in I. brueggemanni, and we filtered the primary data set (n = 23, 165 loci) using the same criteria as for Acropora asp-c. This resulted in 2,946 loci. We then filtered out loci that exhibited significant Hardy-Weinberg and linkage disequilibrium at each sampling site (n = 21). For Hardy-Weinberg disequilibrium, we removed 133 loci that showed departures from expectations at p < .05 in five or more of the 21 sites. For linkage disequilibrium, we removed 681 loci with r values >0.8 among five or more sites. These filters resulted in 2,132 SNPs. Six I. brueggemanni individuals with more than 15% missing data were removed.
We identified putative loci affected by selection as for the Acropora asp-c analysis. Initial analysis using the entire data set did not detect any outliers, but when OutFLANK was applied to the inshore data only, seven loci were identified as outliers and were removed from subsequent analyses resulting in final data set of 2,125 loci.

| Cryptic diversity
We tested for the presence of cryptic diversity within our collections with a cluster analysis that identified the optimal number of genetic clusters (K) and membership coefficients (q) of each colony to a range of clusters with the Bayesian software STRUCTURE v2.3 (Pritchard, Stephens, & Donnelly, 2000). Mean and variance of loglikelihoods and posterior probabilities of the number of clusters from K = 1 to 8 were inferred using correlated allele frequency with admixture model and burn-in of 10,000 and then 100,000 MCMC repetitions. We checked convergence of algorithms by assessing the stability of runtime α and Ln likelihood after burn-in, the variability in individual assignment proportions and the similarity score calculated with the online program CLUMPAK (Kopelman, Mayzel, Jakobsson, Rosenberg, & Mayrose, 2015) from ten replicate runs. As recommended by Wang (2016), we used a separate α for each population and applied an initial value of α = 0.25 (1/K ascertained from exploratory runs), and all other parameters were set as default values. CLUMPAK was used to summarize and graphically present the STRUCTURE results as well as to calculate optimal K using the ΔK method of Evanno, Regnaut, and Goudet (2005). We also considered alternative K values in addition to ΔK including Ln(Pr(X|K) values to identify the k for which Pr(K = k) is highest (Pritchard & Wen, 2004) and chose the K that best described the data and addressed a priori questions and expectations (see Meirmans, 2015;Pritchard & Wen, 2004). When divergent samples were detected (e.g. cryptic diversity or strong geographic divergence), we performed subsequent runs excluding these divergent samples to increase clustering accuracy among the genetically coherent samples (see Janes et al., 2017).
The initial analysis in STRUCTURE identified four divergent and sympatric lineages of A. aspera (see Results). We gauged the relative divergence among versus within these lineages by estimating TA B L E 2 Numbers of samples and unique colonies (genets) of Isopora brueggemanni collected from sites from the Kimberley coast and Ashmore Reef in north-west Australia. N is the total numbers of samples, Ng is the number of genets, and Ng:N is the genotypic richness We used the standardized distance option for the PCoA. We also calculated pairwise FST between lineages and the number of private alleles (P A ) in each lineage in GenAlEx to further estimate the magnitude of differentiation among lineages.

| Inter-archipelago to inter-reef population structure
We examined the population genetic structure at broad scales with STRUCTURE, PCoA and AMOVA using samples from the en-

| Inter-reef to within-reef population structure
We investigated population genetic structure at the inter-reef to within-reef scale in the Acropora asp-c lineage and I. brueggemanni using spatial autocorrelation analysis on the two archipelagos that were sampled in most detail in the inshore Kimberley (Dampier Peninsula and Buccaneer Archipelago). Spatial autocorrelation uses the spatial position and genetic identity of each individual. This analysis is therefore well-suited to establishing the finest scale of genetic structure, is sensitive to recent dispersal processes and is robust to most natural characteristics of plant or animal populations (Double, Peakall, Beck, & Cockburn, 2005;Epperson, 2005). We calculated the autocorrelation between the genetic distance (codominant genotypic) and geographic (Euclidean) distance of all pairs of individuals that fell within a given distance class and plotted each autocorrelation coefficient, r, against its distance class in GenAlEx. Under conditions of restricted gene flow, r is expected to be positive and stable at short-distance classes; then, a subsequent decline in r indicates the "genetic patch," and the y-intercept indicates a balance between genetic drift and gene flow before r becomes negative (Epperson & Li, 1996;Smouse & Peakall, 1999;Sokal & Wartenberg, 1983). Initial analysis of I. brueggemanni showed that the site Kooljaman (see Figure 1) was clearly separate from the general patterns of spatial genetic structure and so was excluded from this analysis. This decision also provided geographic consistency with the study of A. aspera. We tested for statistical significance of r at each distance class, by generating a 95% confidence interval about r via 1,000 bootstrap trials and drawing (with replacement) from within the set of pairwise comparisons for a specific distance class. We inferred significant spatial genetic structure when the confidence interval did not straddle r = 0.
We also estimated fine-scale genetic structure with AMOVA among Bathurst, Pope and Mermaid Islands reefs (F RT_REEFS ); between sites within these reefs (F SR_SITES ); and among all these sites (F ST_SITES ) for I. brueggemanni. This analysis was only possible in this species because we sampled replicate sites at these three reefs. We calculated pairwise F ST between all sites and tested for statistical significance with 999 random permutations.

| Environment, genetic structure and diversity
We quantified the effect of the environment on the population genetic structure and diversity of the A. asp-c lineage and I. brueggemanni using the Bayesian method implemented in GESTE (Foll & Gaggiotti, 2006). Specifically, we tested whether environmental heterogeneity was associated with variation in levels of genetic dif- for the nonindependence inherent in multiple pairwise comparisons (Foll & Gaggiotti, 2006;Riginos, Crandall, Liggins, Bongaerts, & Treml, 2016). We analysed only those sites where n ≥ 9 to account for small sample sizes at some sites in the Acropora asp-c lineage.
This meant that the genetic and geographic outlying site of Ashmore Reef was excluded. For both corals, we used a sample size of 10,000 and a thinning interval of 50 (total of 5 x 10 5 iterations), 10 pilot runs with a length of 5,000 and an additional burn-in of 50,000.
We included six environmental factors: latitude, longitude, the 90th We also investigated the seascape influences on the diversification of lineages within A. aspera by testing whether sites with greater environmental heterogeneity were associated with greater genetic diversity in these corals. To this end, we used a simple linear regression to correlate gene diversity (unbiased expected heterozygosity) with the same environmental factors used in the GESTE analysis at each site in the Acropora asp-c lineage.

| Cryptic diversity in Acropora aspera
The cluster analysis of 329 unique genotypes (genets) of A. aspera revealed four sympatric genetic lineages (hereafter referred to as Further, the PCoA supported the ΔK results, distinguishing four discrete lineages within the entire collection of A. aspera, two of which (asp-c and asp-d) were relatively closely related (Figure 2). There were major differences among these lineages across the genome, with private alleles in one of the four lineages at 247 loci (P A asp-a = 98, P A asp-b = 64, P A asp-c = 69 and P A asp-d = 26). Pairwise F ST between lineages was very large, ranging from 0.480 to 0.704 (Table 3). These estimates of divergence were highly congruent with the relaxed data set of 3,698 loci (Table D1) (Table 3), indicating strong genetic isolation even when living in sympatry. Morphological assessments in the field, along with preliminary assessments of skeletal material, showed no clear macro-morphological differences among the lineages ( Figure D3). Gene diversity within each lineage also varied greatly and was highest in Acropora asp-a (0.108) and lowest in Acropora asp-d (0.054; Figure D4).

| Inter-archipelago to inter-reef population structure
We focused on the Acropora asp-c lineage using 2,894 loci for subsequent population-level analysis in the Acropora data set. After removal of clones (final n = 169; Table 1 & Gilmour, 2009;Whitaker, 2004), and indicate Wahlund effects brought about by nonrandom mating within sites due to spatial and/ or temporal admixture. Gene diversity (unbiased expected heterozygosity at each site) was higher in the centre of the sampling area at the Buccaneer Archipelago sites than at the Dampier Peninsula or the central Kimberley and was very low at Ashmore Reef (Figure 3a).
After removal of clones (n = 561; Table 2  The largest pairwise differences were between the Ashmore site and all the other sites, with average F ST of 0.380 (±SE 0.011; Table   D2). Levels of subdivision were therefore weaker when Ashmore and other sites with sample sizes ≤ 8 were excluded in the AMOVA, but overall patterns and statistical significance were the same (Table   D3).
The geographic structuring observed across a range of spatial scales in Acropora asp-c was far more pronounced in I. brueggemanni. Utilizing 2,125 loci, STRUCTURE analysis revealed maximum ∆K was at K = 2, with very strong membership (q = 1) of all corals to either an offshore Ashmore cluster or an inshore cluster (except for West Montalivet that had q ̴ 50% to both clusters). However, at K > 2, clusters continued to segregate according to geography, and the Ln (Pr(X|K) method identified optimal K = 10. An additional cluster was formed by Kooljaman at K = 3 (Appendix E, Figure E1)    Table E1). Pairwise F ST was also notably high between Kooljaman and the other inshore sites (F ST = 0.241 ± SE 0.019).

| Inter-reef to within-reef population structure
There was significant genetic structure over fine scales within the

| Environment, genetic structure and diversity
There was a strong association between environment and genetic structure and diversity in both corals. The GESTE analysis revealed that tide formed the highest probability model for Acropora asp-c (p = .556) and for I. brueggemanni (p = .432; Table 4). All other models exhibited much lower probabilities (p ≤ .1). The slope of regression was negative for both corals (Table 4), showing that site-specific genetic differentiation (local F ST ) decreased with increasing tidal magnitude. However, the deviation from the regression was moderate for both corals (Table 4), suggesting other untested environmental factors also contributed to the genetic patterns.
Tidal magnitude also exhibited a strong association with gene diversity at each site within the Acropora asp-c lineage. Specifically, sites with bigger tides exhibited greater diversity (R 2 = 0.806, Figure 6). We observed a weaker positive relationship between F I G U R E 5 Spatial autocorrelation analyses of the genetic correlation coefficient (r) as a function of distance for the Acropora asp-c lineage (upper panel) and the I. brueggemanni (lower panel) corals sampled from the Dampier Peninsula and the Buccaneer Archipelago. The bootstrapped 95% confidence intervals were generated by 1,000 bootstrap trials. X-axes differ slightly because of more extensive spatial sampling of sites in I. brueggemanni

| D ISCUSS I ON
Strong genetic divergence and restricted population connectivity characterized the distribution of genetic diversity in two reef-building corals from north-west Australia. These characteristics were evident across a wide range of spatial scales in both the spawning coral, Acropora aspera, and the brooding coral, Isopora brueggemanni. This consistency between species with different life histories indicated that the heterogeneous seascape and powerful oceanographic currents of this region have important influences on their metapopulation dynamics.
Environmental influences not only promote strong genetic differentiation between bioregions and archipelagos (tens to hundreds of kilometres) and regular local recruitment within reefs (tens of metres to a few kilometres), but also rare longer-distance connectivity between reefs within archipelagos (kilometres to tens of kilometres). Underlying this spatial genetic structure, we discovered several highly divergent and cryptic lineages in A. aspera that co-occur on the same reef patch.

| Cryptic diversity and Kimberley corals
We detected four distinct genetic lineages in A. aspera that were not distinguished by macro-morphological characteristics. Pairwise F ST between the four lineages was large (F ST ≥ 0.469), and private alleles were observed at more than half the loci analysed. These results indicate inter-breeding between lineages is rare (sensu Moritz, 2002), despite often co-occurring on the same reef patch. This result is consistent with many genetic studies in other regions that have detected cryptic diversity in Acropora (e.g. Ladner & Palumbi, 2012;Ohki, Kowalski, Kitanobo, & Morita, 2015;Sheets et al., 2018;Wallace & Willis, 1994) as well as other scleractinian (e.g. Forsman, Barshis, Hunter, & Toonen, 2009;Miller & Babcock, 1997;Pinzon et al., 2013).
The detection of cryptic lineages is also consistent with extensive evidence throughout the Kimberley and north-west Australia of high inter-specific-level genetic diversity within reef-building coral species (Richards et al., 2016;Richards, Miller, Miller, & Wallace, 2013;Rosser, 2015Rosser, , 2016Thomas et al., 2014;Underwood et al., 2018). Although many mechanisms are likely involved in evolution of the distinct lineages detected here in A. aspera, timing of spawning (prezygotic barrier) is the best explanation for the maintenance of reproductive isolation among sympatric lineages. Indeed, direct evidence of species-level genetic differences has been observed in Acropora lineages that appear identical but spawn in either spring or autumn (Gilmour et al., 2016;Rosser, 2016;Rosser et al., 2020) or in different months of the same season (Dai, Fan, & Yu, 2000;Ohki et al., 2015;Wolstenholme, 2004). A revision of the taxonomic status of A. aspera that integrates genetic, micro-morphological and reproductive data is warranted.

| Population structure and connectivity
The overall level of genetic subdivision among reefs within the  different life histories at the offshore reefs of north-west Australia (Thomas et al., 2020). However, genetic differentiation was consistently correlated with geographic distance at all scales studied, despite differences in magnitude between species.
At the broadest scale, the largest genetic divergence occurred between offshore and inshore bioregions in both species. This result supports the absence of cross-shelf connectivity in other genetic (Underwood, 2009;Underwood et al., 2018), oceanographic (D'Adamo et al., 2009) and biodiversity studies (Richards, Bryce, & Bryce, 2018;Wilson, 2013). Also at broad scales, three distinct ge-  (Thomas et al., 2020;Underwood et al., 2009Underwood et al., , 2018Underwood, Smith, van Oppen, & Gilmour, 2007) but contrasts to recent evidence of panmixia over these local spatial scales (<100 km) in a different broadcast spawning coral (Acropora digitifera; Thomas et al., 2020). Therefore, the evidence gathered to date suggests many coral populations that are separated by more than a few tens of kilometres are demographically independent, but the environmental heterogeneity of the inshore Kimberley may further restrict connectivity in spawners.
At a local scale, colonies of both species less than 500 m apart were more closely related than more distance colonies. This distance indicates the genetic patch of complete mixing. The size and distinctness of the genetic patch are likely due to fine-scale environmental heterogeneity that influences survival after settlement (e.g. Johnson & Black, 1982).
However, we also suspect life histories play an important role. Larvae of brooders can settle soon after release and recruit very close to their parents. Here, significant differentiation was observed in I. brueggemanni between colonies and sites on the same reef, and the positive autocorrelation was much higher than for the spawner. In contrast, larvae of broadcast spawners spend at least a few days in the plankton. This means the fine-scale genetic patchiness of A. aspera also likely reflects the influence of sticky water and tidally driven eddies that concentrate larvae together and limit mixing of a wider larval pool (Andutta, Kingsford, & Wolanski, 2012;Selkoe et al., 2010;Wolanski & Spagnol, 2000).

| Management implications
Kimberley corals thrive in extreme conditions with especially wide ranges in temperature, irradiance and water quality (Wilson, 2013).
However, even in the Kimberley, bleaching occurs when anomalous heatwaves exceed those tolerances (Gilmour et al., 2019;Hughes et al., 2017;Schoepf et al., 2015). Recovery after such disturbances requires the continued production of demographically important numbers of recruits from local populations over small spatiotemporal scales. In addition, persistence of the metapopulation as a whole requires connectivity networks that enable rarer but evolutionarily important dispersal over broader scales (Gaggiotti, 2017 We found no evidence of contemporary cross-shelf connectivity, so inshore reefs rely on their own stocks not only to supply recruits every generation, but also for genetic diversity to adapt to climate change over multiple generations. These inshore populations are maintained by locally produced recruits at the scale of reef or reef patch, with very few brooded or spawned larvae dispersing and surviving more than 35 km from place of origin. However, our seascape analysis also revealed that genetic structure and diversity were strongly associated with tidal magnitude; sites with bigger tides were more connected to the entire metapopulation (in both species) and were more diverse (in Acropora asp-c). Therefore, strong tidally driven currents appear to have increased the likelihood of occasional larval dispersal between local populations. Conversely, the deep-water tidal current at Sunday Strait appears to be a semi-permeable barrier to dispersal of larvae between the genetically distinct Dampier Peninsula and Buccaneer Archipelago. This result is consistent with other studies that have shown strong oceanic currents often act as "leaky" barriers to larval dispersal that override the influence of biological factors on genetic structure such as planktonic period (Baums, Paris, & Cherubin, 2006;Hohenlohe, 2004;Suzuki et al., 2016).
The congruent patterns among two species with very different modes of reproduction suggest that a single spatial marine management strategy may be used to aid resilience of all coral populations in this region. We recommend that multiple sanctuary networks be spaced at distances no greater than a few tens of kilometres. More specifically, the Dampier Peninsula and Buccaneer Archipelago should be managed as demographically independent systems that sustain their populations through production of local recruits. Further, the population at Kooljaman was the most genetically divergent and depauperate of the I. brueggemanni sites and is probably small, isolated and close to the limits of its south-western range. Therefore, Kooljaman may be more vulnerable to local extinction compared with other reefs studied here. Lastly, the genetic signatures at Tide Rip and Mermaid islands were admixed between the Dampier Peninsula and Buccaneer Archipelago, indicating these reefs provide stepping stones for occasional genetic exchange between the archipelagos important for the adaptive capacity of the metapopulation. We suggest these islands should be considered conservation priorities.
The discovery of cryptic lineages within A. aspera also has implications for management. Such unrecognized diversity is probably common in these systems (Richards et al., 2016), and biodiversity estimates need to account for this (Fišer, Robinson, & Malard, 2017).
Recent evidence indicates ecosystem productivity increases with species richness in many wild populations (Duffy et al., 2017), and coral biodiversity enhances reef ecosystem function (Clements & Hay, 2019). Therefore, the discovery of unrecognized inter-specific diversity may well confer greater resilience to changing environment. Alternatively, if inter-breeding among lineages is rare, their effective population sizes will be smaller than expected, increasing their susceptibility to Allee effects and reproductive failure following reductions in density after disturbance (Knowlton, 2001). In addition, the relatively low genetic diversity of less abundant lineages such as Acropora asp-d may reflect a limited adaptive capacity. Such lineages are likely vulnerable to silent extinction.
Coral reefs worldwide are threatened by the increased frequency and severity of marine heatwaves (Hughes et al., 2018;Van Hooidonk, Maynard, & Planes, 2013). The impacts of such temperature anomalies appear to override the well-documented ecological benefits of no-take reserves (Graham et al., 2020). Therefore, the ecosystem trajectory of most coral reefs will primarily depend on the rate at which carbon emissions are reduced (Hughes et al., 2017).
Nevertheless, conservation strategies that sustain existing connectivity networks will be important (van Oppen & Gates, 2006). Such strategies that protect reefs from local pressures will promote demographic recovery in the short term by capitalizing on the natural variation in resilience to heatwaves of local populations and also the adaptive capacity of coral metapopulations in the longer term. This study illuminates the hidden genetic structuring of two key species of habitat-forming corals to support such local management actions. ZR, JU and JG acknowledge the support of ARC Linkage Project LP160101508 to explore coral resilience.

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