Introduction

With increasing pressure on marine ecosystems from local and global stressors (Johnson et al. 2011; Krumhansl et al. 2016; Friedlander et al. 2018; Wernberg et al. 2019; IPCC 2019), methods to maintain resilience, buffer surrounding environments, and in some cases completely restore these systems are needed (Wood et al. 2019; Mcleod et al. 2018; Mearns et al. 1997; Fredriksen et al. 2020; Westermeier et al. 2014). An understanding of population genetics can provide insights into population connectivity, gene flow, dispersal patterns and diversity within and among populations (Durrant et al. 2018). Therefore, acquiring knowledge about population genetic structures can enable better monitoring and the development of conservation frameworks (Drechsler et al. 2003) in order to manage, preserve and rebuild systems under stress.

Macrocystis pyrifera is a globally significant kelp that holds high value not only for the ecosystem services it provides, but also as an economic resource in the aquaculture industry. Macrocystis pyrifera forms abundant and productive ecosystems within nearshore temperate waters, supporting a diversity of organisms through habitat and food provisioning (Graham et al. 2007). These habitats form the backbone of significant customary, commercial, and recreational fisheries. Macrocystis pyrifera is highly productive (Reed et al. 2008), has high affinity for nutrient uptake (Correa et al. 2016; Purcell-Meyerink et al. 2021) and high polysaccharide content (Ortiz et al. 2009), making it an extremely attractive species for single species and multitrophic aquaculture applications. However, a substantial decline of wild M. pyrifera beds has been reported in some regions of the world (Hay 1990; Johnson et al. 2011; Tait et al. 2021). Currently, there are concerted efforts to safeguard and restore the habitats that this species creates (Hawkins et al. 1999; Buschmann et al. 2014; Ferrari et al. 2018; Layton et al. 2020; Layton and Johnson 2021). Although M. pyrifera is a globally important and iconic species, there is limited research regarding coastal scale genotypic and phenotypic structure that is useful in management.

Some efforts have been made to quantify the genetic structure of M. pyrifera populations at global scales, as well as its interaction with environmental parameters, using microsatellite markers (Assis et al. 2023). These markers have been commonly applied in population genetic studies (Selkoe et al. 2016) due to their high polymorphism and the fact they are found across the genome (Phumichai et al. 2015). So far, 16 microsatellite loci have been isolated and characterized for M. pyrifera (Alberto et al. 2009). Seven to twelve microsatellite markers have been most commonly used to quantify genetic patterns of M. pyrifera in the northern Pacific (Macaya 2010; Alberto et al. 2011; Johansson et al. 2013, 2015; Klingbeil et al. 2022), in Chile (Macaya 2010; Camus et al. 2018; Buschmann et al. 2020) and New Zealand (Macaya 2010). Interestingly, from these studies, a strong genetic similarity was suggested at a global scale (Macaya 2010), whereas at a local scale, genetic breaks were found among populations (Johansson et al. 2015; Camus et al. 2018). Importantly, local environmental conditions are suggested to have a strong impact on species distributional ranges and likely create genetic barriers (Johansson et al. 2015; Camus et al. 2018; Assis et al. 2023). This said, greater resolution of population genetic structure at local scales is needed to better inform management.

New Zealand consists of over 700 islands, ranging from the subtropical north (Kermadec Islands at 29° S) to the subantarctic south (Campbell Island at 52° S) (Ross et al. 2009). A review of genetic structure between populations of different coastal marine species revealed that 33% of all studies (twenty out of fifty eight) found a north–south latitudinal split between populations, which included species such as the macroalga Gracilaria chilensis, a higher plant, many invertebrates, and several fish taxa (Gardner et al. 2010). The site-specific division of the north–south split is unclear but is estimated to be approximately at 37–39 °S in most studies. Wei et al. (2013) examined the genetic structure of the greenshell mussel, Perna canaliculus, and suggested that the exact location could be around Cape Campbell on the north-east coast of the south island. Only two of the 58 reviewed studies found an east–west differentiation in genetic structure.

In New Zealand, the genetic connectivity between coastal species is enabled due to a variety of factors including coastal currents, oceanic currents, discontinuities in environmental factors and previous geographical isolation (Ross et al. 2009; Kelly and Palumbi 2010; White et al. 2010; Stevens and Hogg 2004). Research on genetic variability of seagrass (Zostera muelleri) using random amplified polymorphic DNA found a clear north–south and east–west split and identified that coastal currents play a crucial role in this genetic break (Jones et al. 2008). Similar conclusions have been drawn for Adenocystis utricularis (Fraser et al. 2013), Bostrychia intricata (Ceramiales, Rhodophyta) (Fraser et al. 2013), Durvillaea antarctica (Collins et al. 2010), Carpophyllum maschalocarpum (Buchanan and Zuccarello 2012), and Ecklonia radiata (Nepper-Davision et al. 2021).

Macrocystis pyrifera beds can be found along the south coast of New Zealand’s North Island, the north, east and parts of the west coast of the South Island and on many of its smaller islands, such as Rakiura/Stewart Island, Chatham Islands and the Subantarctic Islands (Hay 1990; Nelson 2020). As in many other areas of the world, M. pyrifera beds in New Zealand have experienced significant decline (Hay 1990; Glover 2020; Tait et al. 2021). So far, only a coarse resolution of population genetic structure has been established (Macaya 2010), mainly focused on how New Zealand populations relate on a global scale, but little is known about the genetic structure of M. pyrifera beds along the coast of New Zealand.

This study utilized microsatellite techniques to quantify the genotypic diversity that currently exists in M. pyrifera populations across New Zealand. The aims were to identify population genetic structure and genetic diversity between and within M. pyrifera beds to gain a greater understanding at a national scale.

Materials and methods

Site selection and tissue collection

To identify population genetic structure of M. pyrifera, tissue samples were collected from multiple sites spanning eight geographical regions (Shears et al. 2008) across New Zealand (Fig. 1). At each site, the apical meristem was collected from mature individuals which were located >10 m apart under the University of Otago Special Permit 644-2. Generally 5–20 individuals were sampled per site and there were between 2 and 11 sites per region. From each individual, a 2 × 2 cm piece of tissue was excised and stored in silica gel until DNA extraction was performed.

Fig. 1
figure 1

Map of collection sites of Macrocystis pyrifera. Eight regions: Wellington, Marlborough, Canterbury, Otago, Southland, Fiordland, Stewart Island and Chatham Islands. Prevailing direction of major ocean currents (gray arrows): d’Urville Current (dUC), Westland Current (WC), Southland Current (SC), Wairarapa Coastal Current (WCC), Wairarapa Eddy (WE) and Rekohu Eddy (RE). (Further information and sites within each region in Supplementary Table S6)

Full size image

DNA extraction

DNA was extracted using a Chelex protocol (Zuccarello et al. 1999) with minor modifications. Briefly, tissue was introduced to 1.5 ml microfuge tubes containing 150 µl of 5% Chelex 100 (Bio-Rad, USA); the tissue was ground with a plastic pestle, then incubated for 10 min at 100 °C; following by 30-s vortex and another 10 min incubation at 80 °C; the sample was centrifuged for 5 min at 16,750 rcf. Finally, 30 µl of supernatant was transferred to a new tube.

DNA quantity was measured using a Nanodrop 200 (ND2000C, ThermoFisher, USA) at a wavelength of 340 nm. DNA quantity was adjusted to 100 ng/µl with sterile deionized water.

PCR amplification

Seven microsatellite primers, for dinucleotide repeats, were selected from Alberto et al. (2009), namely BC4, BC18, BC19, BC25, Mpy8, Mpy11 and Mpy14. Each DNA amplification reaction was carried out using BIOTAQ™ (Bioline). 25 µl of reaction contained 1 µl of DNA extraction (after dilution), 1X PCR buffer, 3 mM MgCl2 solution, 0.25 µM dNTPs, 0.5U of Biotaq polymerase, and 0.4 µM of each primer pair. The thermal profile was 35-cycle consisting of an initial denaturing step of 15 s at 95 °C, annealing 55 °C for primers BC4, Mpy8, Mpy11, and Mpy14 and 65 °C for primers BC18, BC19 and BC25 for 15 s; elongation for 40 s at 72 °C and a final 72 °C for 10 min.

The PCR products were first checked using 2% agarose gel electrophoresis. The PCR products that were clearly visible were used for DNA genotyping by capillary electrophoresis performed using a ABI3730xl DNA analyzer (Applied Biosystems, USA) run with GeneScan™ LIZ®600 standard at Department of Anatomy, University of Otago, New Zealand.

Allele calling

The resulting chromatograms were analyzed and scored using Geneious Prime® 2021.1.1 (https://www.geneious.com) with a semi-default function for microsatellite. Briefly, the genotyping files were imported to the software, sizing method was set as Local Southern, locus information was input with a range of fragment of markers in base pair (bp). To avoid scoring errors, each chromatogram was reviewed manually and compared with chromatograms of the same and different samples, looking for repeated profiles and identical fragment sizes to assign and correct the alleles based on the height of peaks. In difficult profiles showing stutters, the allele scoring followed the method of the manual routine from Pfeiffer et al. (2011) with slight modifications. For this study, the highest main peaks were coded as alleles and peaks directly after the main peak with no less than 70% of the height of the main peak were considered as different alleles. Specific alleles were assigned by round-up (Pfeiffer et al. 2011). In addition, MICROCHECKER software (van Oosterhout et al. 2004) was used to check for scoring errors and null alleles.

Data analysis

Both frequency-based and distance-based genetic indices were used to identify population genetic structure of M. pyrifera in this study.

Descriptive analyses were performed on GenAlEx software v.6.5 (Peakall and Smouse 2012) including number of alleles, observed heterozygosity (Ho), unbiased expected heterozygosity (He) and inbreeding coefficient (FIS). Allele richness was standardized at a sample size n = 30 for all regions using StandArich2 R package (https://github.com/UWMAlberto-Lab/StandArich2). To calculate confidence intervals for FIS, 1000 bootstrap replicates were run using heirfstat package (Goudet 2005) in R version 4.0.3 (R Core Team 2018). A probability test for Hardy–Weinberg equilibrium (HWE) was carried out using GenoDive v3.06, the heterozygosity-based Gis statistic (Nei 1987) was used with 999 permutations (Meirmans 2020). A pairwise FST was run using Arlequin v.3.5.2.2 (Excoffier and Lischer 2010) to calculate genetic differentiation among the eight regions. The significance of these values was tested with 10,000 permutations regions FST ranges between 0 and 1. The higher FST value, the greater genetic differentiation (Wright 1978).

Phylogenetic analyses of regions were conducted using POPTREE2 software (Takezaki et al. 2010). An Unweighted Pair-Group Method with Arithmetic mean (UPGMA) (Sneath and Sokal 1973) was constructed with 10,000 bootstrap replicates, the basis of Nei’s corrected value was chosen. The approach uses the most different region as the root of the tree and then it groups other regions into group structure based on the degree of similarity.

To discover biogeographical clusters, a Discriminant Analysis of Principal Components (DAPC) was conducted using adegenet package (Jombart et al. 2010) in R version 4.0.3 (R Core Team 2018) and STRUCTURE v.2.3.4 (Pritchard et al. 2000). DAPC transforms data using a principal components analysis, then genetic clusters are inferred using discriminant analysis and do not need external assumptions, whereas STRUCTURE uses K-means algorithms and external assumptions (e.g., HWE, Jombart et al. 2010). In DAPC, after conducting cross-validation as an optimisation procedure to identify the number of principal components with the lowest root mean square error, 27 principal components and 7 discriminant functions were retained for a discriminant analysis. A scatter plot was created using predefined biographical regions (eight in this case). In STRUCTURE, an admixture model with correlated allele frequencies was run. K value was set at 10 runs (K = 1–10) and the replication for each run was 10 with initial burn-in 100,000 iterations, followed by 300,000 Monte Carlo Marco chain repeats. Then simulation results from STRUCTURE were extracted using STRUCTURE HARVESTER (Earl and vonHoldt 2012) to identify the highest delta K value, which is also considered as the best fitting number of cluster (Evanno et al. 2005).

A Spatial Analysis of Molecular Variance (SAMOVA) was implemented using SAMOVA 2.0 (Dupanloup et al. 2002) to identify groups of geographical similarity. Different from DAPC and STRUCTURE, this method uses both geographical region and genetic variance to identify genetic structure of M. pyrifera. A priori K of groups of populations from 2 to 8 was chosen, and the initial annealing process was set to 100 times with 10,000 iterations. It is expected that the differences in genetic variance (FCT) between groups should increase with K (Dupanloup et al. 2002). To select the appropriate group, the optimal K should have a plateau in FCT value, and it should have no more than one or more single-region groups in the pattern (Magri et al. 2006).

Analyses of Molecular Variance (AMOVA) (Excoffier et al. 1992) were performed using Arlequin v.3.5.2.2. Regions were grouped into three categories: by genetic (STRUCTURE and DAPC, K = 3), geography and genetic (SAMOVA, K = 3) and by origin (8 regions). AMOVAs were used to identify the genetic variation at three hierarchical levels: among populations, among individuals within populations and among individuals.

Contemporary migration and self-recruitment were estimated using BAYESASS v3.0.4.2 (Wilson and Rannala 2003). Markov chain Monte Carlo analysis was run with 10,000,000 iterations, 1,000,000 burn-in length. To optimize the effectiveness of BAYEASS estimation, five runs were chosen each with a different number of random seeds (Faubet et al. 2007) and the best run was selected based on Bayesian deviance (Meirmans 2014).

To test the null hypothesis that genetic differentiation among regions is driven by stepping-stone gene flow, isolation by distance (IBD) was performed using Mantel test with 5000 permutations in GenAlEx software v.6.5 (Peakall and Smouse 2012).

The program BOTTLENECK (Piry et al. 1999) was used to identify recently bottlenecked genetic groups, which can limit genetic diversity, affect survival and adaptation to climate change (Li and Roossinck 2004), especially in areas that have been exposed to extreme weather conditions, and experienced a major decline in size such as has occurred in the Marlborough region (Hay 1990; Tait et al. 2021). A two-tailed Wilcoxon sign-rank test for heterozygosity excess (Luikart and Cornuet 1998) was used as the number of polymorphic microsatellite markers was less than 20 (Piry et al. 1999). In bottleneck populations, “mode-shift” test is not in an L-shaped distribution.

Results

102 polymorphic alleles were identified from 389 individuals across eight geographical regions using seven microsatellite loci. The number of alleles per locus ranged from 3 (BC4) to 33 (Mpy14). Results from MicroChecker indicated that there was a shortage of heterozygotes for most tested loci resulting in the possibility of null alleles, but no evidence of large dropout for any locus was found. Another possibility for missing heterozygotes is a Wahlund effect (structure between sampled sites within regions). Locus BC4, BC18, and BC19 showed stuttering.

The highest mean number of alleles per region was found in Marlborough, whereas Fiordland had the lowest with 8.571 and 4.000, respectively (Table 1). Allele richness after standardization (n = 30) revealed a slightly different pattern, the lowest allele richness was Fiordland with 3.614, but the highest one was Stewart Island with 6.918. Observed heterozygosity ranged between 0.272 and 0.474, while unbiased expected heterozygosity varied from 0.339 to 0.606. Inbreeding coefficient (FIS) was the greatest in the Marlborough region, while the Fiordland region showed the lowest value, 0.392 and 0.189, respectively (Table 1). All populations departed from Hardy–Weinberg equilibrium in all loci (P < 0.001, Table S2).

Table 1 Genetic diversity indexes of Macrocystis pyrifera regions, based on seven microsatellite loci
Full size table

The least genetically distinct regions were Otago and Southland with a pairwise FST value of 0.021, while the Fiordland and Wellington regions had the greatest genetic differentiation with a pairwise FST value of 0.376 (Table 2). Surprisingly, Fiordland and Stewart Island, which have the closest geographic distance, had the third highest genetic diversity (FST value of 0.318). The global FST was 0.146 (95% CI: 0.06–0.212). The difference in pairwise FST between regions was statistically significant in all comparisons (P < 0.05; Table 2).

Table 2 A matrix of pairwise FST value for eight Macrocystis pyrifera regions
Full size table

Genetic clustering was revealed from the UPGMA tree (Fig. 2). A clear genetic grouping, which was supported with a bootstrap value of 100%, was seen between the Fiordland region and the rest of the M. pyrifera populations in New Zealand (Fig. 2). Several other groups were formed but with lower bootstrap support (<80%): one containing the regions of Marlborough and Chatham Islands, and another containing Canterbury, Otago and Southland. Interestingly, Wellington and Stewart Island, which are ~900 km apart, formed the last group.

Fig. 2
figure 2

A UPGMA tree for Macrocystis pyrifera. Eight regions: Wel Wellington, MBr Marlborough, Can Canterbury, Ota Otago, Sou Southland, Ste Stewart Island, Cis Chatham Islands, Fio Fiordland. Numbers at the nodes represent the percentages from the bootstrap analysis

Full size image

STRUCTURE analysis highlighted the highest delta K was at 2 (maximum likelihood statistic = 346.6). Fiordland was genetically distinct from other regions (Fig. 3). When K = 3 and 4, most individuals from the Wellington, Stewart Island and Marlborough regions were joined in one cluster while in all other regions the cluster assignment was mixed. Interestingly, samples from the Marlborough, Canterbury, Otago and Southland regions revealed a transitional pattern, where individuals shared assignments equally with other clusters, with the exception of Fiordland. In terms of the islands, a very small proportion of ancestry was shared between Chatham Islands (North) and Stewart Island (South).

Fig. 3
figure 3

Structure analyses (K = 2, 3, and 4) for Macrocystis pyrifera sampled from eight regions across New Zealand (ordered from north to south): Wel Wellington, MBr Marlborough, Can Canterbury, CIs Chatham Islands, Ota Otago, Sou Southland, Ste Stewart Island, Fio Fiordland. Each bar represents an individual, with the proportion of color on each bar indicating the proportion of ancestries assignment of the individual to each of the clusters

Full size image

A similar pattern of genetic differentiation to STRUCTURE and the UPGMA tree was seen using DAPC approach (Fig. 4 and Fig. S1). Fiordland was isolated from others, while the remaining groups revealed very little spatial distance: Canterbury, Otago and Southland were grouped in one cluster, whereas the rest were classified as one single genetic cluster.

Fig. 4
figure 4

Discriminant analysis of principal components (DAPC). Scatterplot from DAPC with regions are shown as different colors and inertia ellipses (54% of variance), while dots indicates individuals. The top right is PCA eigenvalues (27), bottom right is the amount of variance explained by two discriminant eigenvalues for plotting. Wel Wellington, MBr Marlborough, Can Canterbury, Cis Chatham Islands, Ota Otago, Sou Southland, Ste Stewart Island, Fio Fiordland

Full size image

In the SAMOVA, FCT values reached a plateau at K = 6, and the second highest FCT was at K = 3 (Table S2). Five single groups were formed at K = 6, which is a signal of sub-structure (Magri et al. 2006). Therefore, K = 3 could be the optimal number of groups for this data set. However, when K = 3, M. pyrifera individuals grouped into three geographical groups: Wellington in group 1; Marlborough, Canterbury, Otago, Southland, Stewart Island and Chatham Islands in group 2; and Fiordland in group 3 (Table S2), which slightly differs from the UPGMA, STRUCTURE and DAPC analyses.

Overall, using different methods (UPGMA tree, STRUCTURE, DAPC and SAMOVA) the genetic structure of the Fiordland population was discontinuous from the rest of the other regions in New Zealand.

There was slight difference in the molecular variance when the regions were partitioned into different groups based on the previous analyses. However, the general pattern remained the same (Table 3). The majority of the molecular variance, over 60% occurred among individuals within regions (P < 0.001). Less than 30% of total variance was found among different groups (P < 0.001).

Table 3 AMOVA test for different grouped regions for Macrocystis pyrifera
Full size table

A wide range of self-recruitment rates were seen among the eight regions, with Southland having the lowest (67.4%) and Fiordland the highest (95.7%) (Table S4). The migration rate from other regions into Fiordland was greater than the migration rate out of Fiordland into other regions, with the greatest migration coming from Canterbury (0.0298 in versus 0.0074 out) (Table S4).

There was no correlation between genetic distance and geographical distance as the r value was close to zero and P = 0.54 when performing a Mantel test, which also means the null hypothesis of no IBD was inferred. The Mode-shift test was found to be L-shaped in all regions (Table S5), which indicates no population bottleneck.

Discussion

Macrocystis pyrifera beds are in a state of decline in many regions of the world (Johnson et al. 2011; Krumhansl et al. 2016; Filbee-Dexter and Wernberg 2018) and there is emerging evidence and a strong anecdotal narrative of decline along New Zealand coastlines (Hay 1990; Glover 2020; Tait et al. 2021). Efforts are being made to develop suitable approaches to restore kelp ecosystems (Eger et al. 2020; Layton et al. 2020; Layton and Johnson 2021). Understanding population genetic structure is an essential step for establishing restoration approaches and enhanced ongoing management. Using highly polymorphic microsatellite markers, this study indicates that a strong genetic discontinuity was found between the Fiordland region and the rest of M. pyrifera in New Zealand. Genetic structure of the East coast of the South Island was mixed, with no clear phylogeographic patterns emerging. Nevertheless, genetic connectivity among these eight regions was limited as evident from high and statistically significant pairwise FST values (Table 2) and strong self-recruitment (Table S4).

Genetic diversity is generally used as an indicator to identify the adaptive potential of a species or population (Reusch et al. 2005; Johnson et al. 2005; Hughes et al. 2018). Wernberg et al. (2018), for example, suggested that low genetic diversity in Ecklonia radiata may result in lower resilience to rising temperatures (Wernberg et al. 2018). Our results show similar genetic diversity among east coast regions, Chatham Islands and Stewart regions but a significantly lower genetic diversity in the Fiordland region. The latter has recently been exposed to a high frequency of marine heatwaves (Bell et al. 2022). While it is difficult to make an assumption on the adaptive potential of a population from neutral markers, such as microsatellites, the low genetic diversity of the Fiordland region may warrant further investigation and monitoring.

The Fiordland region also showed the greatest genetic differentiation from all other sampled regions, suggesting a biogeographic discontinuity. This same pattern has been reported in other species, such as greenshell mussel Perna canaliculatus (Wei et al. 2013), sea perch Helicolenus percoides (Lawton et al. 2010), and the sea star Coscinasterias muricata (Sköld et al. 2003; Perrin et al. 2004). It may indicate that stock populations in this region originate from a different glacial refugium compared to the east coast regions, a pattern also suggested for Macrocystis populations in Tasmania, Australia (Coyer et al. 2001). Another explanation may be isolation due to oceanographic circulation patterns around the South Island of New Zealand, although it remains unclear how the Fiordland currents (FC) connect to the Southland Current (SC) (Chandler et al. 2021). It is clear that the FC flows southwestward, controlled by a poleward pressure system (Stanton 1976; Ridgway and Dunn 2003), whereas the SC flows northeastwards extending from the upper continental slope southeast of Stewart Island (Chiswell et al. 2015; Stevens et al. 2021). Modeling suggests that it takes 59 and 118 days for a single particle which is released from Fiordland (Milford and Doubful Sounds) to reach Dunedin and Wellington, respectively (Chiswell and Rickard 2011). Isolation from neighboring kelp beds could then prevent genetic exchange resulting in low genetic diversity and isolation, which was consistent with what we observed in Fiordland. These findings highlight that Fiordland may be more susceptible to perturbation as recruits from neighboring regions is reduced (Westemeier et al. 1979; Madsen et al. 1999). Therefore, it is recommended that a greater understanding of the health of Fiordland M. pyrifera beds is gained in order to assess the risk of localized extinctions. In early 2022, Fiordland, and more generally the west coast of New Zealand’s South Island, experienced the strongest marine heatwave with respect to cumulative intensity (duration × intensity), a measure of accumulated heat stress, in the >40 years satellite time series that started in 1981. Temperatures were in some places >4 °C above average and breached current known thermal thresholds for M. pyrifera with maximum coastal temperatures of 20 °C observed (Bell et al. 2022). These types of events have increased in frequency and magnitude over the past 40 years (Oliver et al. 2018) a pattern predicted to continue in New Zealand (Behrens et al. 2022) prompting the need for a greater understanding of genetic structure, health and vulnerability.

Genetic connectivity on the east coast could be explained by the migration of floating adult M. pyrifera being transported by currents and by a relatively strong connectivity in the past. The former hypothesis is supported by other work, both at a global scale (Assis et al. 2023; Coyer et al. 2001; Macaya 2010; Macaya and Zuccarello 2010b, a) and regionally (Johansson et al. 2015; Camus et al. 2018). In the wild, it is unlikely that genetic exchange among regions can occur through spore dispersal as spores are short-lived and have estimated dispersal distances of <1 km (Gaylord et al. 2006). However, floating adult kelp thalli have been shown to be reproductive after 125 days (Hernández‐Carmona et al. 2006), and these rafts can allow genetic connectivity between giant kelp beds that are geographically distant (Reed et al. 2006). A modeled particle released on the east coast of the South Island (Bluff, Port Chalmers, Lyttleton, Dunedin, Timaru) can be found in the North within 10, and 72 days in Chatham Islands via transport by the SC (Chiswell and Rickard 2011). Research on other macroalgae also supports this hypothesis (Dayton 1973; Muhlin et al. 2008; Fraser et al. 2009). For example, Fraser et al. (2009) found that Durvillaea antarctica has a single widespread haplotype over 10,000 km due to high dispersal from rafting. However, a better understanding of nearshore and offshore currents around New Zealand is required to further understand population connectivity. In addition, the genetic similarity of the Chatham region and the mainland may be explained by past events such as a westward shift in the SC as a result of West Wind Drift at the end of Last Glacial Maxium (LGM).

The reason for the genetic relatedness between Wellington and Stewart Island (~900 km distant, FST = 0.032, P < 0.001) using indices, such as UPGMA, STRUCTURE and DAPC, is unclear. Chiswell and Rickard (2011) found that Kaikoura (northeast South Island) is the longest distance that a particle released in Wellington can reach along the coast of the South Island. The very high self-recruitment rate (88.7%) of the Wellington region also raises a question for this unusual pattern. In contrast, spatial analyses using SAMOVA, which requires geographical data, rejected this unusual pattern. Other studies also found that genetic variability of other seaweed and invertebrate species between Wellington and South Island is closer than from northern populations (Ross et al. 2009; Buchanan and Zuccarello 2012; Nepper-Davision et al. 2021). Therefore, the need for further research into the factors shaping this unique genetic pattern is crucial for a comprehensive understanding of marine population dynamics.

Ocean currents are not the only variable that may explain genetic connectivity among geographical regions, and a more complex approach could be applied to further investigate the patterns seen in this study. Seascape genetics, for example, considers both spatially variable structural and environmental conditions to explain the genetic patterns of marine species. Johansson et al. (2015) concluded that oceanographic transport models do not provide a consistently adequate explanation for genetic variability. Incorporating environmental conditions in the analysis may help uncover previously unnoticed patterns of genetic structure. Wei et al. (2013) found that SST contributes more to genetic variability of the endemic New Zealand greenshell mussel Perna canaliculus than other variables, including environmental variables (SST, current, water velocity) and three geospatial variables. Similar results have been reported in other organisms, such as bivalves (Belanger et al. 2012) and brown algae (Assis et al. 2014; Neiva et al. 2014) using environmental data. Therefore, the inclusion of other seascape components such as environmental predictor variables may improve the power of model prediction for population genetic structure of this study. Other possibilities exist for the patterns of regional discontinuity, self-recruitment and lack of IBD seen in New Zealand regions of M. pyrifera. Metapopulation factors, such as extinctions and recolonizations, possible after glacial cycles (founder events), variable reproductive success (sweepstake genetics) and patterns of dispersal (current and sea surface temperatures) that may have been different during the LGM, similar scenarios have been proposed for other marine species and M. pyrifera (Asiss et al. 2023). A more detailed sampling design plus more widespread genetic markers (SNPs) may be able to determine the evolutionary history of New Zealand M. pyrifera in future.

Conclusions

This study provides data on the genetic structure and genetic diversity among and within M. pyrifera regions in New Zealand. Although genetic variation was low between regions, significant boundaries were found between regions on the east and west Fiordland coasts. This finding adds more evidence of the isolation of organisms in Fiordland from the rest of New Zealand. It remains unclear why low genetic differentiation exists between Stewart Island and Wellington on the North Island and more information regarding nearshore and offshore current flow is required to clarify this. In addition, lower genetic diversity was found in the northern regions, at the northern distributional limit of the species in New Zealand, and it is proposed that environmental factors (e.g., high temperature) may contribute to this. Further analysis is needed to understand the interaction between genetic structure and environmental data as this may help interpret the evolution and future persistence of M. pyrifera in New Zealand.