Cryptic Diversity and Patterns of Structure in Nine Species of North Pacific Rockfishes (Sebastes spp. and Sebastolobus spp.)

ABSTRACT

Understanding how habitat and life history traits contribute to population genetic structure is critical for evaluating conservation risks and identifying optimal management of wild populations. Species derived from adaptive radiations can provide a useful model for studying such patterns because the presence of some consistent traits allows for closer examination of attributes which differ. Here, we examine population structure in rockfishes (genera Sebastes and Sebastolobus), a group of speciose, recently diverged marine fishes that exhibit a continuum of genetic structure, from cryptic species to putative genetic panmixia. In this study, we analyse low‐coverage whole genome resequencing data from nine rockfish species and identify four major types of genetic differentiation: cryptic diversity, isolation‐by‐distance (IBD), no detectable structure (putative panmixia) and no detectable structure with a single genomic region putatively underlying local adaptation. Two species, Dark ( Sebastes ciliatus ) and Northern ( S. polyspinis ) rockfishes, each exhibited cryptic diversity consisting of two sympatric groups differentiated at every chromosome throughout the genome. Black rockfish in Southeast Alaska also had two genetic groups, despite an overall signal of IBD from the West Coast of North America through the Gulf of Alaska (GOA). Five species showed evidence of genetic panmixia throughout the sampled range, while Yelloweye Rockfish exhibited high gene flow across ~4000 km, from southern California to the GOA, but had strong differentiation between the Salish Sea and the outer coast. Of the species in which no structure was detected, only Blackspotted Rockfish had one outlier region exhibiting signatures of potential local adaptation. Our results highlight how multiple mechanisms likely promote genetic structure in rockfishes, hinting at tentative links to depth distributions and offshore habitats that might influence patterns of differentiation in rockfishes more broadly. Further, this study provides genetic data for some of the most abundant rockfishes in commercial fisheries in the Northeast Pacific, enhancing the scientific information available for informing fisheries management.

1. Introduction

Population structure in many marine fishes is shaped by a complex combination of life history traits and historical processes and is often low due to limited barriers to dispersal and large effective population sizes, or recent divergence that can lead to patterns approximating panmixia (Bekkevold et al. 2005; Gonzalez and Zardoya 2007; Timm et al. 2024). When population structure in marine fishes is detectable, it is frequently due to reduced gene flow caused by interactions between species traits, such as larval retention, self‐recruitment or behaviour and oceanography, including currents, fronts and eddies, which often play an outsized role during the larval dispersal phase (Baetscher et al. 2019; Cowen and Sponaugle 2009). Population structure and species diversification are also shaped by historical processes, particularly sea‐level height, which influences habitat availability and can lead to divergence in more nearshore and shallow‐dwelling taxa (Ludt and Rocha 2015). Differences in selection regimes across habitats can also influence the presence and type of population structure, with evidence for genetic differentiation at loci associated with local adaptation increasingly documented in genomic studies (Savolainen et al. 2013).

Genetic analyses are often used to help identify population structure that could be relevant for management and conservation of wild populations. Increasingly, high‐resolution genomewide data can reveal additional genetic diversity that may not have been evident from prior studies using a smaller number of genetic markers. Further, genomewide data can reveal novel patterns of differentiation, including outlier peaks and evidence for links between genome architecture and local adaptation (Tigano and Friesen 2016; Wellenreuther et al. 2019). Such datasets provide new information about population structure and genetic diversity that can be valuable for identifying geographic stock boundaries, or cryptic diversity and informing which populations should be managed as independent stocks (Benestan 2020).

In the Northeast (NE) Pacific, rockfishes (Sebastes spp. and Sebastolobus spp.) are ubiquitous from the Gulf of California northward through the Aleutian Islands and Bering Sea (Love et al. 2002) and represent an important component of commercial and recreational fisheries throughout their range. With > 100 species worldwide, some rockfishes inhabit nearshore environments and kelp forests, while others are found in deep, offshore waters. Rockfishes are live‐bearing, with pelagic larvae that settle into juvenile habitat after spending months in the water column (Love et al. 2002). Genetic analyses of rockfishes have revealed multiple types of population structure along the speciation continuum including sister species with very low genomewide differentiation, isolation‐by‐distance, strong divergence at biogeographic boundaries, sympatric ecotypes and cryptic species (e.g., Andrews et al. 2018; Benestan, Fietz, et al. 2021; Hess et al. 2023; Longo et al. 2022). An absence of genetic structure has been described in some species as well (Grass Rockfish; Martinez et al. 2017; Kelp Rockfish; Gilbert‐Horvath et al. 2006).

Consistent with the variety of forms of genetic structure, the mechanisms of differentiation are varied and include a combination of behavioural, life history and environmental traits, such as mate‐choice and reproduction via internal fertilization (Helvey 1982), depth segregation of adults (Behrens et al. 2021; Ingram 2011; Shum et al. 2015), specific habitat requirements, particularly for nearshore and shallow water species (< 100 m depth; Carr 1991), and varied and extended larval dispersal that can facilitate long‐distance gene flow (30–170+ days depending on species; Love et al. 2002; Miller and Shanks 2004). Introgression between species is yet another mechanism that seems to be involved in species diversification (Buonaccorsi et al. 2005; Saha et al. 2017; Schwenke et al. 2018; Sykes et al. 2023; Wray et al. 2024).

One of the more intriguing developments in studies on rockfishes is the discovery of cryptic genetic diversity within species previously identified as monotypic based on morphology. Examples of cryptic species pairs identified using various genetic methods include Rougheye/Blackspotted (Gharrett et al. 2005; Orr and Hawkins 2008), Blue/Deacon (Frable et al. 2015), Dark/Dusky (Orr and Blackburn 2004) and Vermilion/Sunset (Hyde et al. 2008). Additional cryptic diversity has been identified using high‐resolution and genome‐wide genetic data, including within Atlantic Redfish (Benestan, Fietz, et al. 2021), Golden Redfish (Saha et al. 2017) and Vermilion and Sunset rockfishes (Harms et al. 2025; Longo et al. 2022; Wray et al. 2026). In the Atlantic Redfish species, S. mentella , three ecotypes are distributed across depth (shallow vs. deep) and geographic location, highlighting how multiple mechanisms of differentiation could be at play, and that changes in sea level may be particularly important for genetic divergence (Benestan, Fietz, et al. 2021; Hyde and Vetter 2007). Depth similarly appears instrumental in structuring genetic groups of Sunset Rockfish, with one exclusively inshore and the other primarily slightly deeper offshore (Longo et al. 2022). In contrast, within genetic groups of Vermillion Rockfish, differentiation appears structured latitudinally, with a northern, central and southern group along the California coast (Longo et al. 2022). Collectively, these studies suggest that genetic studies of rockfishes may continue to uncover cryptic diversity, especially in relatively nearshore habitats.

With the recent availability of rockfish reference genomes (e.g., De‐Kayne et al. 2025; He et al. 2019) and decreasing cost of generating whole genome resequencing data, it is feasible to generate individual genotypes for hundreds of samples from multiple species, thus allowing us to examine patterns of genetic differentiation at unprecedented resolution in some of the most abundant rockfishes encountered in commercial, federally managed U.S. fisheries. Here, we use low‐coverage whole genome resequencing (lcWGS) to identify patterns of genetic structure in nine rockfish species: Dark Rockfish ( Sebastes ciliatus ), Dusky Rockfish ( Sebastes variabilis ), Northern Rockfish ( Sebastes polyspinis ), Shortraker Rockfish ( Sebastes borealis ), Blackspotted Rockfish (Sebastes melanostictus), Rougheye Rockfish ( Sebastes aleutianus ), Yelloweye Rockfish ( Sebastes ruberrimus ), Black Rockfish ( Sebastes melanops ) and Shortspine Thornyhead ( Sebastolobus alascanus ). The nine species analysed in this study represent multiple habitat types (demersal, pelagic and slope), and although they are economically valuable, most have few if any data describing genetic structure. Of the species subject to prior studies, Northern Rockfish and Black Rockfish exhibited isolation‐by‐distance (IBD; Gharrett et al. 2012; Hess et al. 2023), Yelloweye Rockfish had structure associated with geographic boundaries inside the Salish Sea versus the NE Pacific coast (Andrews et al. 2018), and Shortraker Rockfish showed signals of structure but did not have clear genetic breakpoints or patterns of IBD across sampling locations (Matala et al. 2004). A similar analysis with Blackspotted Rockfish and Rougheye Rockfish likewise identified variation across groups of collections, but did not identify population structure (Gharrett et al. 2007). Thus, for most of these species, our study represents the first individual‐based genetic analysis.

In this study, we aim to identify patterns of genetic structure, and then to determine whether life history traits may provide some generalizable hints to types of genetic structure. For example, do species with similar depth distributions or habitat use (benthic vs. the water column) exhibit similar patterns of population structure? Are widely‐distributed species more likely to exhibit IBD or no detectable genetic structure? Do we identify cryptic diversity, and if so, is there a depth or latitudinal gradient that resembles results found in other species? The nine species in this study do not represent the full‐breadth of rockfish life history diversity, but include some of the largest species' ranges (i.e., Yelloweye Rockfish and Shortspine Thornyhead) and some of the most abundant species in the NE Pacific. This study will provide substantial additional data for understanding high‐latitude and offshore rockfishes and explore the fisheries management implications associated with multiple patterns of genetic structure.

2. Methods

2.1. Samples

For the nine species, sample sizes ranged from 58 for Dark Rockfish to 184 for Black rockfish (Table 1). The geographic range of samples varied by species based on species distribution and sample availability (Figure 1). Samples for Northern, Dark and Dusky rockfishes were restricted to Alaska waters, while the five other Sebastes species included samples distributed farther south along the coast of North America. Shortspine Thornyhead had the greatest geographic spread with samples from Southern California to the Bering Sea.

TABLE 1.

Samples and sequencing efforts for each rockfish species. Regions outside of the species range are denoted with a dash. Regions within a species range without samples contain a zero. Only species for which samples from a second sequencing batch were analysed separately are indicated in parentheses. Sample sizes are post‐quality filtering. Polymorphic sites are based on alignment to the Korean rockfish genome ( S. schlegelii ; NCBI assembly GCA_014673565.1). Only sites from the first batch of Northern rockfish were used to alleviate the sequencing batch effect.

Species	Sampling by geographic region									Sequencing efforts
	Bering Sea	Aleutian Islands	Western GOA	Central GOA	SE Alaska	West Coast	Salish Sea	Southern CA	Total samples	Sequencing batches	Sequencing facility	Mean sequencing depth	Polymorphic sites (n)
Shortspine Thornyhead Sebastolobus alascanus	18	10	25	25	15	0	0	4	97	1	Novogene	2.1	254,711
Blackspotted Rockfish S. melanostictus	4	21	0	13	14	26	—	—	78	1	Azenta	3.5	3,725,521
Rougheye Rockfish S. aleutianus	6	2	10	0	14	26	—	—	58	1	Novogene	2.2	1,819,672
Northern Rockfish S. polyspinis	17	59	28	8	0	0	—	—	112	2 (batch artefacts)	Novogene and University of Oregon	2.5	1,070,699
Shortraker Rockfish S. borealis	6	20	62	0	0	18	—	—	106	2 (batch artefacts)	Novogene and Azenta	3.3	4,701,373
Dusky Rockfish S. variabilis	0	0	33	23	18	0	—	—	74	2 (no batch effect)	Novogene and University of Oregon	2.7	2,424,209
Dark Rockfish S. ciliatus	0	0	40	24	4	—	—	—	68	2 (no batch effect)	Novogene and University of Oregon	2.9	2,639,434
Yelloweye Rockfish S. ruberrimus	0	0	0 (40)	17 (32)	0 (24)	53	30	22	122 (96)	2 (analysed separately)	Novogene and University of Oregon	2.1	1,758,200
Black Rockfish S. melanops	0	0	39	59	64	18	0	4	184	2 (no batch effect)	Novogene and University of Oregon	2.9	2,418,550

FIGURE 1.

Species distribution maps, sampling locations and genomewide principal components analysis for nine rockfishes. (A) No genetic structure was identified in five species, consistent with genetic panmixia throughout the sampled range. (B) Two species have population structure associated with geography: Black Rockfish exhibits isolation‐by‐distance (IBD; see Figure S5) and Yelloweye Rockfish shows differentiation between the inside waters of the Salish Sea and the outer coast. (C) Genomewide data for Dark and Northern rockfishes reveal two genetically distinct groups in each species and two intermediate individuals in Dark Rockfish. Evidence of a batch effect from multiple sequencing lanes is present in Shortraker and Northern rockfishes (see also Figures S2 and S8). In (A) and (B), colour corresponds to sampling location, whereas for (C), colour corresponds to groups defined by genetic analysis. Species distributions are shown in light blue with observation data from the Ocean Biodiversity Information System (OBIS).

2.2. Library Preparation and Sequencing

We generated lcWGS data for 995 rockfish samples across nine species (Table 1). Samples were obtained from fishery surveys and research collections with the goal of covering the majority of each species' geographic range (Table S1). DNA was extracted from tissue samples with Nucleospin 96 plate extractions (Machery‐Nagel), and the presence of high‐molecular weight bands was verified using a 2% agarose gel. Extracts were quantified (Quant‐iT PicoGreen) and normalized to 10 ng for standardized input for library preparation. Libraries were prepared using the Illumina Tagmentation kit, following Baym et al. (2015) and Therkildsen and Palumbi (2017). Each sample was individually barcoded and normalized using a SequalPrep 96‐plate (Invitrogen) before pooling barcoded libraries and then size‐selecting for fragments > 300 bp with a ratio of 0.6× AMPure XP beads (Beckman Coulter). The final library was eluted in 18–20 μL of TLE buffer, quantified with the Qubit HS kit and visualized on an E‐gel (Life Technologies) to verify library size. Because rockfish genomes are ~900 Mb, we sequenced samples for this study in batches of two pooled libraries (192 samples) on each S4 lane of an Illumina NovaSeq 6000 system to obtain ~2.5× sequencing coverage per individual. We used 2 × 150 bp sequencing for each S4 lane. Libraries were sequenced at one of three sequencing facilities: Azenta, Novogene and the University of Oregon. Due to opportunistic sampling across the species' range, some species were sequenced across multiple batches, meaning some samples were included in one lane of sequencing, while other samples of that species may have been sequenced at a later point in time at a different sequencing facility (as indicated in Table 1).

2.3. Bioinformatics, Data Filtering and Genotype Likelihoods

Data processing for all species followed the same initial bioinformatics workflow for lcWGS data (Timm et al. 2024) based on scripts from Therkildsen and Palumbi (2017). Bioinformatic processes were run on the NOAA Fisheries bioinformatics high‐performance computing cluster. The bioinformatics processing steps include trimming adapter sequences, aligning fastq files to the reference genome, calculating sequencing depth by individual, and computing genotype likelihoods (GLs). For this dataset, each species was processed separately. First, raw sequencing data were sorted by unique barcode combinations, and adapters and poly‐G tails were trimmed with TRIMMOMATIC v0.39 (Bolger et al. 2014) and fastp (Chen et al. 2018), respectively. Subsequently, trimmed sequences were aligned to the Korean Rockfish genome ( Sebastes schlegelii ; scaffold N50 of 7.9 Mb; NCBI assembly GCA_014673565.1) using BWA v0.7.17 (Li and Durbin 2009). The Korean Rockfish was one of only two Sebastes species with chromosome‐level genome assemblies, and represented an outgroup for all Sebastes species in this study (Figure S1). The other species, Honeycomb Rockfish ( S. umbrosus ), is not an outgroup, and thus had the potential to introduce reference bias. Available data indicate that karyotype is highly conserved among rockfish species, and that there are few large structural rearrangements (De‐Kayne et al. 2025; N. Lou, unpublished), perhaps with the exception of sex‐determining regions (Sykes et al. 2023). Duplicate reads were identified and removed using picard v2.23.9 (http://broadinstitute.github.io/picard/), and overlapping read pairs were clipped with BamUtil v1.0.5 (Jun et al. 2015). The depth of coverage per sample was subsequently calculated with SAMtools v1.11 (Li et al. 2009), and samples with mean coverage less than 1× were removed from analyses. GLs for polymorphic sites were calculated in ANGSD v0.933 (Korneliussen et al. 2013) using the following filtering criteria: a minimum base quality and mapping quality score of 20, a minimum minor allele frequency of 0.05, a SNP p‐value of 10⁻¹⁰, a minimum depth across all samples equal to the sample size, and a maximum depth across all samples of 20× the sample size.

2.4. Exploration of Population Structure

Genotype likelihoods were used to generate a covariance matrix for whole genome principal component analyses (PCA) with PCAngsd v0.99 (Meisner and Albrechtsen 2018). For species with multiple sequencing batches, a small number (12) of duplicate samples were included in both batches to evaluate sequencing batch effects (Lou and Therkildsen 2022). After verifying that duplicate samples appeared in the same location on the PCA, one sample from each duplicate pair was removed and the data re‐analysed to generate a new covariance matrix. For species that displayed separation of batches in PC space, batches with samples from a smaller geographic range were analysed separately, when possible, to avoid introducing a sequencing batch effect. For Northern and Shortraker rockfishes, retaining two batches in a single analysis was necessary to provide representative samples across geographic regions, despite a batch effect (see Results). Iterative efforts to reduce the batch effects included increasing the minimum number of individuals required for calling sites so that samples from both batches would be included (minInd), increasing the minimum base quality (minQ) and mapping quality (minMapQ) and finally, in the case of Northern Rockfish, restricting the analysis to just sites from one batch.

In order to interpret patterns of neutral structure, we identified whether sex‐associated regions, regions associated with structural variation or outlier regions were potentially influencing patterns of differentiation. For example, when whole genome PCA plots of the first two principal components (PC1 and PC2) appeared to be influenced by sex, we used a genome scan of pairwise F _ST between males and females based on metadata sex‐information to identify which genomic regions appeared differentiated by sex, and then removed these genomic regions in order to evaluate the genomewide PCA without sex (following the approach in Benestan et al. 2017; Timm et al. 2024). A similar approach was used with putatively neutral chromosomal inversions (i.e., inversions that did not appear correlated with geography).

To further evaluate genetic structure, we performed an admixture analysis using GLs in NGSadmix v32 (Skotte et al. 2013). For species that showed no clustering in the genomewide PCA, we performed three iterations each of K = 2 and K = 3 to test for multiple ancestral populations even though no groups were evident from the PCA data. For species with clustering or a batch effect in the PCA, we added K = 4 and, depending on the PCA pattern, K = 5 to the analysis to account for potential additional groups in the data. We used default parameters with no minimum minor allele frequency (minMaf 0).

Species with any indication of a correlation between geographic and genetic distance were further tested with a correlation plot using the metric for genetic distance, F _ST/(1−F _ST). Since some collections were combined by proximate geographic locations, the sampling location coordinates were averaged for isolation‐by‐distance (IBD) analyses (e.g., the coordinates for all California samples averaged to Monterey Bay in central California). Minimum marine distances between geographic regions were calculated using the marineDistances R script (Assis et al. 2013).

2.5. Population Differentiation

Subsequent population analyses were conducted on assigned groups based on clustering patterns in the whole genome PCA. Species with no evident clustering were grouped by geographic sampling location, while species with PCA clustering, suggestive of population structure, were grouped by genetic cluster.

To examine patterns of differentiation across the genome of each species, we calculated pairwise F _ST. For species without obvious clustering in the whole genome PCA, we compared geographic regions that represented the farthest geographic extent of the sampled species range, selecting groups with 10 or more individuals to avoid upward biases in F _ST estimates (Lou et al. 2021). For species that showed genetic clustering in the whole genome PCA unrelated to geographic regions, we compared genetic groups. We calculated site allele frequency likelihoods for each relevant within‐species group (e.g., geographic region, genetic cluster and/or sex) using SNPs previously identified in the full species' dataset (sites flag). ANGSD filtering parameters were the same as for the full dataset with adjustments for each group sample size. We calculated pairwise F _ST from the folded site frequency spectra for each group pair using realSFS in ANGSD. There were approximately 40 pairwise comparisons across all species meeting the minimum group size criterion. To limit the plots to a manageable number, we only included plots for the sampling locations separated by the greatest geographic distance for each species.

2.6. Putatively Adaptive Outlier Regions

Despite a lack of obvious clustering by geography in the genomewide PCA, we wanted to further evaluate whether samples from different geographic regions might harbour adaptive loci distinct from the neutral variation displayed in the PCA. Outlier loci can be indicative of local adaptation and would be relevant for fisheries management and consideration of stock structure (Nielsen et al. 2009). Because of the evolutionary value of identifying such outliers, various methods have been used for outlier detection with whole genome sequencing data (Fariello et al. 2017; Luu et al. 2017; Whitlock and Lotterhos 2015). However, many methods require called genotypes from higher coverage sequencing data or permutation tests that are too computationally expensive for a study of this scale. Furthermore, parameter choice using these methods affects whether a given SNP appears as an outlier. Since the purpose of outlier detection in our study was to identify loci that may be relevant for stock structure, we used a method that characterized the general shape of differentiation across the genome coupled with a conservative threshold to identify the most notable peaks. To do this, we first calculated z‐scores from the per‐SNP F _ST estimates across the whole genome. Next, we incorporated genomewide, per‐SNP F _ST estimates into a chromosome‐level Gaussian kernel smoothing function using the smoothr R package (Strimas‐Mackey 2025) and a bandwidth of 4. Using the previously calculated z‐scores, we identified significant peaks with z‐scores > 10. The peak boundary was based on where the significant kernel smoothed peak dropped below a z‐score of 5. Previous studies have used similar z‐score thresholds when aiming to remove false positive signals and identify regions of strong differentiation (e.g., Auteri and Knowles 2020; Provost et al. 2022).

For the one species in which we identified a significant genomic peak (Blackspotted Rockfish, see Results), we analysed SNPs within the peak boundary to better characterize allele differences across sampling regions. We visualized linkage disequilibrium (LD), performed local PCA, estimated heterozygosity and calculated the number of copies an individual has of the major allele at each locus from GLs (allele dosage). Linkage disequilibrium was calculated in ngsLD v1.2.0 (Fox et al. 2019), and r ² values were plotted as a heatmap with LDheatmap (Shin et al. 2006). For allele dosage, the major allele at each locus was standardized to the major allele from the region with the largest sample size (West Coast). Allele dosages were calculated as the sum of the heterozygous GL plus two times the homozygous major GL. Calculations for allele dosages were rounded to 0 (homozygous minor), 1 (heterozygous) or 2 (homozygous major) to represent the genotype for an individual at a single SNP. Average heterozygosity within the peak boundary was calculated for each individual in ANGSD while retaining base and mapping quality filters. Individual heterozygosity was then compared to the corresponding clustering patterns in the local PCA and allele dosages.

3. Results

Overall, the average sequencing depth by species was 2.1× to 3.3×, and the number of polymorphic sites varied considerably, with between 1.07 and 4.7 million SNPs across the eight Sebastes species and ~250,000 SNPs in Shortspine Thornyhead (Table 1).

3.1. Rougheye Rockfish

Rougheye Rockfish had 1.8 million polymorphic sites and 2.2× mean sequencing depth (Table 1). Based on the PCA, Rougheye Rockfish showed no clustering by geography from Oregon through the Bering Sea (Figure 1A) with additional support for no population structure from the admixture analysis, where at K = 2 mixed ancestry was present in similar proportions across geographic regions and individual ancestry proportions changed across model iterations (Figure S3). The genome scan identified very low levels of differentiation and no notable peaks of divergence between geographically far sampling locations (Figure 2B). Average pairwise F _ST estimates were 0.011–0.018.

FIGURE 2.

Genome scans for species that appear panmictic based on genomewide PCA. The pairs of sampling locations shown represent the largest geographic distance in each species. Blue lines represent kernel smoothing (with a bandwidth of four) based on per‐SNP F _ST from genotype likelihoods. Red dashed lines represent an F _ST z‐score of 10. Only Blackspotted has a peak on chr 14 that is significant at this threshold.

3.2. Shortraker Rockfish

The dataset for Shortraker Rockfish included 4.7 million polymorphic sites and a mean sequencing depth of 3.3× across two sequencing batches (Table 1). The PCA revealed two clusters along PC2 that did not separate by geographic region, with 16 of 106 individuals found in one of the clusters (Figure S2A). A genome scan revealed a putative chromosomal inversion that was responsible for the variation on PC2 (Figure S2). When reanalyzed without the chromosome containing the putative inversion (chr 19), the PCA showed no clustering (Figure 1A), but did show evidence of a sequencing batch effect (Figure S2B). Admixture results appear primarily influenced by the batch, which shows up as different proportions of ancestry for most samples from the Western GOA region, even at K = 2 (Figure S3). A genome scan based on sampling population location identified no notable peaks of divergence and pairwise F _ST values ranged from 0.07 to 0.010 (Figure 2).

3.3. Dusky Rockfish

Dusky Rockfish had 2.4 million polymorphic sites and a mean sequencing depth of 2.7× (Table 1). The individual‐based PCA showed no clustering from SE Alaska through the Western GOA (Figure 1A) and the lack of structure was also apparent in the admixture results, where the majority of individuals had mixed ancestry at K = 2 and/or switched ancestry among model iterations (Figure S2). Based on the genome scan, there were no notable regions of elevated differentiation, and the mean pairwise F _ST ranged from 0.008 to 0.010, with little variation across comparisons (Figure 2).

3.4. Shortspine Thornyhead

Shortspine Thornyhead had ~250,000 polymorphic sites and 2.1× mean depth (Table 1). The PCA showed no clustering from Southern CA to the Bering Sea (Figure 1A), and the admixture analysis showed a combination of mixed ancestry and inconsistent ancestry across model runs at K = 2. Using just collections with sufficient sample sizes for pairwise F _ST calculations and genome scans (samples from SE Alaska to the Bering Sea), there were no regions of elevated F _ST (Figure 2). In fact, Shortspine Thornyhead had the lowest average pairwise F _ST of all species in the study (range = 0–0.005; Table 2).

TABLE 2.

Average F _ST estimated within each species between locations with ten or more samples. For species with distinct genetic groups that were not associated with geography, F _ST was calculated between individuals grouped by PCA clustering rather than geography.

Species	Comparison	Average F ST
Rougheye Rockfish ( S. aleutianus )	West Coast–SE Alaska	0.01115
	West Coast–Western GOA	0.01432
	SE Alaska–Western GOA	0.01771
Shortraker ( S. borealis )	West Coast–Western GOA	0.006868
	West Coast–Aleutian Islands	0.009566
	Western GOA–Aleutian Islands	0.007099
Dusky Rockfish ( S. variabilis )	SE Alaska–Yakutat	0.01004
	SE Alaska–Western GOA	0.009494
	Yakutat–Western GOA	0.007737
Blackspotted Rockfish ( S. melanostictus )	West Coast–SE Alaska	0.012194
	West Coast–GOA	0.012707
	West Coast–Aleutian Islands	0.009701
	SE Alaska–GOA	0.016698
	SE Alaska–Aleutian Islands	0.013394
	GOA–Aleutian Islands	0.013584
Shortspine Thornyhead ( Sebastolobus alascanus )	SE Alaska–Central GOA	0.003886
	SE Alaska–Western GOA	0.004441
	SE Alaska–Aleutian Islands	0.000625
	SE Alaska–Bering Sea	0.000956
	Central GOA–Western GOA	0.002929
	Central GOA–Aleutian Islands	0.001762
	Central GOA–Bering Sea	0.003758
	Western GOA–Bering Sea	0.005126
	Western GOA–Aleutian Islands	0.003046
	Aleutian Islands–Bering Sea	0
Yelloweye Rockfish ( S. ruberrimus )	Southern CA–West Coast	0.006799
	Southern CA–GOA	0.011309
	West Coast–GOA	0.008751
	Salish Sea–Southern CA	0.026417
	Salish Sea–West Coast	0.024055
	Salish Sea–GOA	0.028388
Black Rockfish ( S. melanops )	California–West Coast	0.022267
	California–SE Alaska 1	0.047437
	California–SE Alaska 2	0.067457
	California–Central GOA	0.075291
	California–Western GOA	0.079429
	West Coast–SE Alaska 1	0.03531
	West Coast–SE Alaska 2	0.05217
	West Coast–Central GOA	0.059859
	West Coast–Western GOA	0.065092
	SE Alaska 1–SE Alaska 2	0.01567
	SE Alaska 1–Central GOA	0.019852
	SE Alaska 1–Western GOA	0.022288
	SE Alaska 2–Central GOA	0.016895
	SE Alaska 2–Western GOA	0.01943
	Central GOA–Western GOA	0.004816
Northern Rockfish ( S. polyspinis )	Cluster 1–Cluster 2	0.019563
Dark Rockfish ( S. ciliatus )	Cluster 1–Cluster 2	0.012563

3.5. Yelloweye Rockfish

Yelloweye Rockfish included 1.76 million polymorphic sites and mean sequencing depth of 2.1× across two sequencing batches (Table 1). In the first sequencing batch, Yelloweye Rockfish appeared to have geographic structure based on the individual‐based PCA and admixture plots. The PCA showed two clusters, one with Salish Sea individuals tightly clustered apart from all other samples, which included individuals sampled from Southern California to the GOA (Figure 1B). The admixture plot similarly showed the Salish Sea distinct from all other regions at K = 2, while no additional geographic groups appeared, even at K = 4 (Figure S3). In a second sequencing batch that was restricted to samples from the GOA and analysed separately, Yelloweye Rockfish showed no geographic clustering (Figure S4). Excluding the Salish Sea samples, kernel smoothed F _ST showed minimal differentiation in Yelloweye Rockfish (sampled from Southern CA to the GOA; Figure 2) and pairwise F _ST ranged from 0.007 to 0.028 (Table 2).

3.6. Black Rockfish

Black Rockfish included 2.4 million polymorphic sites and 2.9× mean depth (Table 1). Multiple clusters in the PCA were associated with geography, with proximity among clusters matching the geographic distribution of sampling locations. This pattern suggested isolation‐by‐distance across the sampling range (Figure 1B), which was confirmed by testing the correlation between genetic distance (F _ST/[1−F _ST]) and marine distance (in km; R ² = 0.88; Figure S5). Black Rockfish in the Western and Central GOA were tightly clustered together in the PCA, whereas samples from Washington State to Southern California were more dispersed across both PC1 and PC2. In the admixture results, this gradient is most evident at K = 2, with alternative ancestry types in the GOA and West Coast/Southern California, and mixed ancestry in SE Alaska. Excluding the small number of samples from Southern California, average pairwise F _ST was greatest between the West Coast and Western GOA (F _ST = 0.065; Table 2). In addition to IBD, Black Rockfish samples from SE Alaska segregated into two PCA clusters, and chromosome‐level PCAs and a genome scan showed that divergence between these two SE Alaska groups was prevalent throughout the genome (Figure S6).

3.7. Dark Rockfish

Dark Rockfish had 2.6 million polymorphic sites and 2.9× mean depth (Table 1). Genetic differentiation evident in the genomewide PCA showed two previously unrecognized genetically distinct groups as well as two individuals that appear intermediate to the genetic groups (Figure 1C). Geographically, the genetic groups are present in sympatry in the Western GOA; however, only one of the genetic groups appears in SE Alaska, albeit with limited sampling from that region (Figure 1C). Admixture plots for Dark Rockfish showed two groups, although some individuals exhibited < 0.5 proportion of the alternate ancestry (Figure S3). The two intermediate individuals from the PCA show ~50/50 ancestry in the admixture plots (Figure S3). Chromosome‐level PCAs showed separation between genetic groups in all but one chromosome (Figure 3). The whole genome F _ST scans for Dark Rockfish displayed multiple genome regions of high F _ST, but typically not exceeding a pairwise F _ST of 0.2, while the genomewide average F _ST was 0.012 between the genetic groups (Table 2).

FIGURE 3.

Chromosome‐level analyses show differentiation across the genome for species with cryptic diversity. PCA for Dark (A) and Northern (B) rockfishes aligned to the Korean Rockfish genome assembly (Sebastes schlegleii). Points are coloured according to groups from the genomewide PCA (Figure 1). Chromosomes 20, 22 and 24 are associated with sex in Northern rockfish and highlighted in pink. PC1 separates males from females and PC2 separates genetic groups in these chromosomes. The genomic architecture of cryptic diversity differs between Dark Rockfish (C) and Northern Rockfish (D). Green lines represent kernel smoothing based on F _ST from genotype likelihoods using a bandwidth of four. The dashed red line represents an F _ST z‐score of 10. Northern Rockfish has higher overall F _ST values than Dark Rockfish, and thus, the z‐score of 10 is at 0.4 in Northern Rockfish and at 0.2 in Dark Rockfish. Comparisons for F _ST are based on genetic groups from the genomewide PCA.

3.8. Northern Rockfish

Northern Rockfish had ~1.1 million polymorphic sites with 2.5× mean sequencing depth across two batches (Table 1). The genomewide PCA initially showed strong clustering based on putative sex‐determining regions that appeared to comprise the majority of three chromosomes (chr 20, 22, 24; Figure S7), although the actual chromosome structure could be affected by reference genome bias. After removing the three sex‐associated chromosomes, PC1 separated two genetic clusters that were not associated with geographic sampling location (Figure 1C), while PC2 was dominated by a sequencing batch effect (Figure S8). Despite this batch effect, admixture results were consistent for two genetic groups (Figure S3) and chromosome‐level PCAs revealed clear separation between the genetic groups in every chromosome (Figure 3). The distribution of samples from each group extends from the Western GOA through the Aleutian Islands and Bering Sea, but one group appears to be much more common in the Aleutian Islands and Bering Sea (Figure 1). The genome scan showed many peaks of elevated F _ST, and an average genomewide F _ST = 0.019 (Table 2).

3.9. Blackspotted Rockfish

Blackspotted Rockfish had 3.7 million polymorphic sites and 3.5× mean depth from one sequencing batch (Table 1). The PCA for Blackspotted Rockfish showed no clustering by geography from Oregon through the Bering Sea (Figure 1A) and the admixture plot similarly showed split ancestry across most individuals at K = 2, indicating that K = 1 fits the data best (Figure S3). From the genome scans, average pairwise F _ST estimates ranged from 0.010 to 0.017 (Table 2). Blackspotted Rockfish was the only species with no evidence of population structure that had a significant F _ST peak in the genome scan. The peak on chr 14 in the West Coast–Aleutian Islands comparison was ~30 kilobases wide and had a maximum F _ST of 0.34 (Figure 4). Linkage disequilibrium was elevated between SNPs within the majority of the peak (r ² ~0.5), indicative of either reduced recombination or selection. The local PCA for SNPs within the peak boundary displayed three main clusters on PC1 that are likely indicative of homozygous major, heterozygous and homozygous minor haplotypes (Figure 4C), consistent with per‐individual heterozygosity calculations (Figure S9). The allele dosage heatmap also showed these three haplotypes (Figure 4D), suggesting that the homozygous major cluster mostly consists of samples from the West Coast, SE Alaska and GOA, whereas the heterozygous and homozygous minor groups are mainly made up of Aleutian Islands and Bering Sea samples. At least one individual from each geographic region was in the heterozygous cluster.

FIGURE 4.

The Blackspotted F _ST peak is associated with geographic groups. The significant peak on chromosome 14 is differentiated between West Coast and Aleutian Island individuals (A). The red dashed line is a z‐score of 10 and the vertical grey dashed lines are the start and end of the kernel smoothed peak boundary used for local analyses. The LD heatmap (r ²) shows elevated linkage in the peak region (B). Local PCA shows three main clusters, where the positive PC1 space mostly consists of individuals from the Bering Sea and Aleutian Islands (C). The allele dosage heatmap shows each SNP within the peak boundary (columns) for each individual (rows), grouped by geographic region, and arranged based on location in PC1 space (D). Allele dosage represents the number of copies an individual has of the major allele for each SNP, represented as a 0 (homozygous minor), 1 (heterozygous) or 2 (homozygous major) and missing data are shown in grey. Individuals from the Aleutian Islands and Bering Sea display a lower dosage of the major allele than the other regions.

4. Discussion

Across nine Pacific rockfish species representing a variety of habitat and depth ranges, we observed varied structure from putative panmixia to cryptic genetic divergence, suggesting that multiple interacting processes shape differentiation in this group. Five species in this study exhibited patterns approximating genetic panmixia throughout the sampling range, with one additional species (Yelloweye Rockfish) putatively panmictic across ~4000 km, but with strong genetic differentiation between the inside waters of the Salish Sea and the outer coast. Two genetic groups of Black Rockfish in SE Alaska may also be the product of limited dispersal based on geography, pointing to two species where mechanisms limiting dispersal impact genetic structure. Similar to genetic studies of other rockfish species, we identified cryptic genetic diversity in Dark and Northern Rockfishes, consisting of two groups with overlapping geographic distributions. The mechanisms promoting divergence likely differ between Dark Rockfish and Northern Rockfish, with either the depth range or longitudinal gradient of oceanographic conditions potentially implicated. This study highlights the value of genomewide data from a set of closely related species for examining patterns of population structure, with the potential to highlight possible mechanisms underpinning genetic differentiation and provide improved scientific evidence to inform stock structure for commercially harvested wild species.

4.1. Causes of Gene Flow and Geographic Structure

Long distance dispersal facilitated by extended pelagic larval duration (PLD) is one of the key species characteristics thought to contribute to gene flow in marine species (Cowen and Sponaugle 2009). Longer PLD theoretically provides greater opportunities to homogenize genetic structure, while a shorter larval duration may be associated with geographic structure. Although PLD is only available for two species in this study, Black Rockfish (60–80 days; Lotterhos et al. 2014) and Shortspine Thornyhead (14–15 months; Love et al. 2002) follow theoretical expectations, where Shortspine Thornyhead has a longer dispersal period and no evidence of population structure, while Black Rockfish has a shorter PLD and evidence of IBD. More generally, it is unclear whether a shorter PLD generally corresponds with population genetic structure, although it is commonly assumed that offshore and shelf species have a longer PLD than nearshore rockfishes (Lotterhos et al. 2014; Love et al. 2002). However, at least one species of every habitat type in this study (demersal, pelagic and slope) has no obvious signals of population structure, suggesting that PLD and habitat‐type may be disconnected in how they correlate with structure. While rockfish data are limited, a study of four Mediterranean species (three fishes and one lobster) found no IBD and no genetic structure in species with PLD of 4–6 months, but identified small‐scale IBD in a species with a PLD of 20–35 days (Benestan, Rougemont, et al. 2021). Those lines of evidence might point to no detectable genetic structure in species with a long PLD, and yet genetic structure has been identified despite the potential for longer dispersal distances (e.g., Brown Rockfish: Buonaccorsi et al. 2005; Black rockfish: Hess et al. 2023; Lotterhos et al. 2014; Miller and Shanks 2004; this study). Overall correlations between PLD and F _ST are surprisingly poor (Weersing and Toonen 2009), which suggests other mechanisms may facilitate gene flow or population differentiation and points to the importance of population size and divergence time as two additional factors that affect F _ST (Faurby and Barber 2012).

Physical or biogeographic barriers can restrict larval dispersal or promote self‐recruitment (Baetscher et al. 2019), in some cases leading to genetic structure (e.g., Hess et al. 2011). In this study, a primary example of genetic structure associated with physical geography is in Yelloweye Rockfish, where the only genetically distinct population occurs in the Salish Sea (Figures 1 and 2), which is connected to the Pacific Ocean through narrow outlets (i.e., ‘inside’ waters) and harbours unique oceanography (i.e., sills and freshwater input). Salish Sea Yelloweye Rockfish likely have a limited exchange of migrants or larvae with outer coast populations (Andrews et al. 2018; Siegle et al. 2013); however, Andrews et al. (2018) did identify 5% of individuals with Salish Sea ancestry on the outer coast as well as a few putative hybrids. Another example of physical barriers promoting differentiation may be found in the two genetically distinct populations of Black Rockfish in SE Alaska, one of which may represent dispersal more restricted to waters inside the fjordlands and effectively separated from the outer Pacific coast (Figures 1 and S6). Unfortunately, these samples were obtained from port‐sampling and lack precise capture locations, making it difficult to unequivocally associate the genetic structure in SE Alaska Black Rockfish with geography. Otherwise, the IBD structure we identified in Black Rockfish is consistent with previous studies, which suggested either limited dispersal or extinction and recolonization events (Hess et al. 2023; Lotterhos et al. 2014).

4.2. Cryptic Diversity Is Common in Rockfishes

Although cryptic genetic diversity has been repeatedly identified across rockfish species, our discovery of cryptic diversity in both Dark and Northern rockfishes was still surprising given the lack of obvious barriers to gene flow and the presence of genetic groups in sympatry. Both species have co‐occurring genetic groups in the Western GOA; however, the distribution of genetic groups hints at the possibility of some spatial segregation (Figure 1). In Dark Rockfish, only one of the genetic groups appears in SE Alaska, while in Northern Rockfish, one of the genetic groups is dominant in the Aleutian Islands and Bering Sea (Figure 1). No previous studies have analysed population structure in Dark Rockfish and the previous study of Northern Rockfish identified a pattern of IBD based on estimating allele frequencies from collection locations rather than individual genotypes (see below for details). Cryptic diversity with genetic groups inhabiting spatially segregated geographies with some geographic overlap has also been observed in Vermilion ( S. miniatus ) and Sunset (S. crocotulus) rockfishes in California (Longo et al. 2022) and in Blue ( S. mystinus ) and Deacon (S. diaconus) rockfishes along the west coast (Frable et al. 2015).

The presence of cryptic diversity in two of the nine species we analysed suggests that this type of genetic variation may be even more common across rockfishes than previously appreciated, and yet, the environmental or habitat characteristics that might increase the likelihood of cryptic diversity are uncertain. Using our collection of nine species, we can compare and contrast available habitat and distribution data between species with cryptic diversity and those without (i.e., Table 3). For example, oceanographic mechanisms and currents could be similar for Dark Rockfish and Dusky Rockfish due to their similar geographic range (Figure 1), but differences in depth distributions might promote or impede genetic differentiation. Dark Rockfish tend to be shallower (10–153 m; Love et al. 2002), whereas Dusky Rockfish, which appear putatively panmictic, are found deeper along the outer continental shelf (30–525 m; Love et al. 2002). Northern Rockfish have an intermediate depth distribution (100–200 m; Love et al. 2002). Shallower environments may cultivate additional genetic differentiation due to divergence caused by sea‐level changes (as suggested by Hyde and Vetter 2007 and Benestan, Fietz, et al. 2021). Species that span latitudinal or longitudinal gradients may also exhibit cryptic diversity (e.g., Longo et al. 2022), which could be the case in Northern Rockfish from the GOA through the Aleutian Islands. Such gradients may be driven by oceanographic currents and eddies, for example, the Alaska Gyre (East–West) or the California Current (North–South). Conversely, neither cryptic diversity nor population structure was identified in any of the demersal, offshore species with geographic distributions spanning the West Coast, from California through the Bering Sea (i.e., Rougheye, Shortraker and Blackspotted rockfishes and Shortspine Thornyhead; Figure 1; Table 3), perhaps hinting at species life history characteristics that are less likely to lend themselves to genetic differentiation.

TABLE 3.

Ecological information for each rockfish species. Species distributions are based on observation data from the Ocean Biodiversity Information System (OBIS). Habitat information is from the Alaska Department of Fish and Game. Inference of population structure is from the genomewide PCA.

Species	Species distribution	Habitat	Depth	Population structure
Shortspine Thornyhead ( Sebastolobus alascanus )	Southern CA—Bering Sea	Demersal	100–850 m	None detected
Blackspotted Rockfish ( S. melanostictus )	Northern CA—Bering Sea	Slope	300–500 m	None detected
Rougheye Rockfish ( S. aleutianus )	Central CA—Bering Sea	Slope	300–500 m	None detected
Northern Rockfish ( S. polyspinis )	British Columbia—Bering Sea	Slope	50–300 m	Cryptic diversity
Shortraker Rockfish ( S. borealis )	West Coast—Bering Sea	Slope	100–600 m	None detected
Dusky Rockfish ( S. variabilis )	West Coast—Bering Sea	Pelagic	30–525 m	None detected
Dark Rockfish ( S. ciliatus )	SE Alaska—Bering Sea	Pelagic	10–153 m	Cryptic diversity
Yelloweye Rockfish ( S. ruberrimus )	Southern CA—Bering Sea	Demersal shelf	50–400 m	Inside/Outside
Black Rockfish ( S. melanops )	Southern CA—Bering Sea	Pelagic	< 350 m	Isolation‐by‐distance

Efforts to understand the ecological role of cryptic diversity may be important for fisheries management. For example, studies that have associated ecotypes with observed genetic differentiation (e.g., Benestan, Fietz, et al. 2021) are well suited to identify potential habitat or life history characteristics that could be relevant considerations for managed rockfish species (Cadigan et al. 2022). Thus, future sampling efforts for Dark and Northern rockfishes in Alaska could prioritize collections of phenotypic and depth data to explore whether these genetic groups represent phenotypically diverged ecotypes.

4.3. Individual‐Based Analyses Reveal Variation Within Geographic Collections

For Northern, Shortraker, Blackspotted and Rougheye rockfishes, previous genetic studies estimated allele frequencies from microsatellite data by collection location, which were further used to estimate F _ST between collections (Gharrett et al. 2005; Gharrett et al. 2012; Matala et al. 2004). Our individual‐based analysis of Northern Rockfish from the Aleutian Islands through the Western GOA identified two discrete, but geographically overlapping genetic groups (Figures 1C and 3); yet these same samples produced a signal of IBD using the previous approach (Gharrett et al. 2012). In this case, the proportion of the two genetic groups changes from East to West across the sampling range such that allele frequency changes suggestive of IBD masked the underlying cryptic diversity in this species. Similarly, a microsatellite study of Shortraker Rockfish identified population structure, but not IBD, using allele frequencies estimated by sampling location (Matala et al. 2004). The comparison with our individual‐based results for these species highlights how a priori decisions about grouping samples (i.e., by collection location or morphology) can affect resulting allele frequency estimates and inference of population structure, whereas individual‐based analyses are less likely to be biassed (Allendorf et al. 2010).

4.4. Limitations of Low‐Coverage Data

Despite the cost‐efficiency and high‐resolution of lcWGS data, analysis can, at times, be limited by particulars of the low‐coverage data. Batch effects are a well‐known consequence of sequencing samples for lcWGS across multiple sequencing lanes (Lou and Therkildsen 2022). Although not all species we sequenced across multiple batches produced batch effects (see Table 1), there were three species for which the batch effect produced data artefacts. Despite attempts to mitigate them using stringent filtering thresholds, batch effects remained in the genomewide PCA for Northern and Shortraker rockfishes (Figure 1; Figures S1 and S5). For Yelloweye Rockfish, we analysed batches separately, providing rangewide data in one batch and data for the GOA in the other (Figure S4). Batch effects are problematic because they can obfuscate true biological patterns and because iterating analyses with increasing filter stringency is computationally intensive. Although the patterns of population structure for species in this study with batch effects can be identified from a thorough examination of multiple facets of the data (i.e., > 2 principal components in the PCA; genome scans based on pairwise F _ST, and Admixture plots), avoiding batch effects through careful choice of the geographic distribution of samples in a sequencing library provides an opportunity to avoid the considerable effort of trying to minimize batch effects after the fact. As we found in Northern, Yelloweye and Shortraker rockfishes, it may not be possible to completely eliminate a batch effect.

Another consideration for WGS more generally is the effect of aligning data to the reference genome of a different species (i.e., reference bias). We aligned all species in this study to the Korean Rockfish, the one chromosome‐level genome assembly available at the time of this study that represented an outgroup (Figure S1; Hyde and Vetter 2007). Korean Rockfish diverged from the eight Sebastes species in this study ~10 million years ago (MYA) and diverged from Shortspine Thornyhead over 15 MYA (Kolora et al. 2021). The greater phylogenetic distance between Korean Rockfish and Shortspine Thornyhead likely explains the small number of polymorphic sites mapped in the more divergent Sebastolobus species (< 25% the number of SNPs in the species with the second‐lowest number of SNPs; Akopyan et al. 2025). Finally, in the genomic peak in Blackspotted Rockfish, reference bias could be responsible for the portions of the outlier region in which Blackspotted Rockfish sequencing data do not map (Figure 4D). Mapping to conspecific reference genomes, when available, would more accurately characterize F _ST outliers, including SNPs unassociated with structural variants (Akopyan et al. 2025). Despite these limitations, alignment to the same chromosome‐level reference genome assembly may be preferable for comparisons across species.

4.5. Management Implications

Genetic structure is typically considered in fisheries management, since genetic variation provides evidence for reproductive isolation across generational time scales (Waples 1998); however, the absence of genetic structure does not mean the absence of ecologically relevant populations or fishery stocks (Cadrin 2020; Waples and Gaggiotti 2006). This is a particularly important consideration for large, recently diverged, highly connected populations, as is true for many marine fishes. Most of the rockfishes in this study exhibit structure that approximates genetic panmixia, which provides little information to inform spatial fisheries management. Instead, spatial management may rely on demographic or life history characteristics (i.e., growth, age‐at‐maturity), or political boundaries and/or fishing fleet dynamics (Cheng et al. 2025). Local depletion can still occur in panmictic species because of mismatches between the time required to accumulate genetic differences among groups versus the time scales of fisheries management (Waples and Gaggiotti 2006).

Species that exhibit cryptic diversity present a distinct challenge for management. The genetically distinct groups of Northern and Dark rockfishes in sympatry in the Western GOA (Figure 1) may be candidates for management as separate stocks, if they are indeed genetically and/or demographically isolated. A similar question has been applied to Atlantic Redfish, where two sympatric and morphologically similar species are present in the Northwest Atlantic, and multiple genetically identifiable ecotypes are present within one of these species (Benestan, Fietz, et al. 2021). Improved biological information for both species and ecotypes is needed to effectively manage this complex stock structure (Cadigan et al. 2022).

The increased availability of chromosome‐level genome assemblies coupled with whole‐genome resequencing data has cleared a path for considering small genomic regions that contain putatively adaptive loci when identifying stock structure. In Blackspotted Rockfish, the putatively adaptive genomic peak we identified on chr 14 (Figure 4) has different frequencies of alleles in Bering Sea and Aleutian Islands samples compared to the GOA and West Coast, but these regions are already managed separately (Spencer et al. 2022). Similarly, Lingcod ( Ophiodon elongatus ) harbours a putatively adaptive inversion with a geographic transition in frequency from north to south along the West Coast of North America (Longo et al. 2020). Based on this chromosomal inversion, the existing stock management boundary was moved from the Oregon‐California border farther south to Northern California, where an area of admixture with both inversion types is present (Field et al. 2021; Longo et al. 2020).

5. Conclusions

The increasing quality of genomic data for commercially harvested marine species provides an opportunity to characterize species diversity and refine population information that can then be used to help inform management boundaries. In the case of Pacific rockfishes, genomewide data and individual‐based analyses revealed cryptic diversity as well as structure approximating panmixia, with tentative links to habitat, depth, oceanographic currents and overall species geographic range that might influence the patterns of genetic differentiation in rockfishes more broadly. Although the nine species and sequencing data for ~1000 samples in this study represent a considerable dataset for exploring mechanisms that promote various forms of genetic structure, a future project would benefit from expanding the analysis to represent additional life history traits (e.g., subtidal and kelp forest species) a broader geographic range (e.g., Western Pacific and Atlantic species), and potentially explicitly test physical and environmental variables.

Author Contributions

D.S.B., W.L., N.H., L.E.T. and P.H.S. designed the research. D.S.B., N.H., L.E.T. and R.N.L. performed the research. D.S.B., N.H. and L.E.T. analysed data. D.S.B. and N.H. wrote the paper and W.L., P.H.S., J.R.H., K.M.N., R.N.L., G.L.O. and W.C. edited the paper.

Funding

Funding from the North Pacific Research Board (project 2112) supported this work. This work made use of the Sedna high performance computing cluster hosted at and administered within NOAA Fisheries, which has been funded by NOAA Fisheries advanced technologies strategic initiatives.

Ethics Statement

All samples used in this study were collected consistent with the National Oceanic and Atmospheric Administration (NOAA) guidelines for scientific research collections.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: The maximum clade credibility (MCC) chronogram from Ingram & Kai (2014) subset to all rockfish species in this study including the reference genome species (blue). The phylogenetic tree was constructed from seven mitochondrial and two nuclear genes (data from Hyde and Vetter 2007). The MCC chronogram was created using the posterior distributions of 1000 trees from a Markov Chain Monte Carlo analysis (see methods in Ingram & Kai (2014)). Tree data from Ingram & Kai (2014) were imported from TreeBase to R using the treebase package (Boettiger & Temple, 2012), subset to the species within this study, and plotted with the ape package (Paradis & Schliep, 2018).

Figure S2: (A) The Shortraker Rockfish ( S. borealis ) genomewide PCA is influenced by an inversion on chromosome 19 that is unrelated to geographic region. (B) The individual‐level PCA for Shortraker shows a sequencing batch effect on PC1 in addition to the inversion on PC2. (C) The genomewide PCA does not show a batch effect with PC2 versus PC3. (D) The genome scan comparing the West Coast and Western GOA to the Aleutian Islands does not show any notable peaks, consistent with the PC1 clustering being due to a batch effect. Batch 1 included the Western GOA samples and two West Coast samples, while Batch 2 included the three other sampling locations. Plot C also has the inversion on chr 19 (included in B) removed.

Figure S3: Admixture plots for each of nine species of rockfishes. Species are shown with K = 2 and K = 3 unless there were additional clusters evident in the genomewide PCA, and then additional values for K were tested. The Shortraker rockfish plot shows the influence of a sequencing batch effect (see also Figure S1).

Figure S4: Yelloweye rockfish from the Gulf of Alaska show no population structure (96 samples; A–B). Samples were sequenced in a separate batch from the 122 Yelloweye rockfish samples included in the main text (Figure 1). Genomewide PCA for Yelloweye samples from Alaska, the West Coast and Southern California, excluding the Salish Sea, show no clustering with the first three principal components (C–D).

Figure S5: Black rockfish ( S. melanops ) show a strong correlation between geographic distance and genetic distance (see Table 2 for F _ST estimates).

Figure S6: Genetic structure in Black rockfish in SE Alaska shows differentiation at every chromosome consistent with the two genetic groups from SE Alaska in the whole genome PCA (Figure 1) (A). Genetic differentiation shows peaks of elevated F _ST across the genome (B). Blue lines represent kernel smoothing based on F _ST using a bandwidth of four. The dashed red line represents a z‐score of 10.

Figure S7: Genetic differentiation (F _ST) between males and females for Northern Rockfish aligned to the Korean Rockfish genome ( S. schlegelii ; NCBI assembly GCA_014673565.1). Chromosomes highlighted with red bars are putatively sex‐associated and were removed for the whole genome PCA.

Figure S8: Northern rockfish genomewide PCA shows evidence of a batch effect from two sequencing lanes. The majority of samples in batch 2 are from the Western Gulf of Alaska. The primary axis of variation (PC1) divides two genetic groups that are not defined by geographic region (see Figure 1C).

Figure S9: Local PCA of the Blackspotted Rockfish outlier peak with points coloured by per‐individual average heterozygosity in the peak region. Individuals in the center cluster of PC1 (putative heterozygous haplotypes) had greater heterozygosity estimates than those of the two putative homozygous clusters at the far left and right of PC1.

Table S1: Individual location and collection years for each sample in this study. Sample sources include the NOAA Alaska Fisheries Science Center (AFSC), NOAA Northwest Fisheries Science Center (NWFSC), NOAA Southwest Fisheries Science Center (SWFSC), Alaska Department of Fish and Game (ADFG), Oregon Department of Fish and Wildlife (ODFW) and the Department of Fisheries and Oceans Canada (DFO).

Data Availability Statement

Raw sequence reads are deposited in the SRA (BioProject PRJNA1416903). Analyses are available on GitHub (https://github.com/DianaBaetscher‐NOAA/ak‐rockfish).