Journal list menu

Volume 57, Issue 10 p. 1988-1998
Free Access

Using genomics to design and evaluate the performance of underwater forest restoration

Georgina Wood

Corresponding Author

Georgina Wood

Centre for Marine Science and Innovation, School of Biological, Earth and Environmental Sciences, Sydney, NSW, Australia


Georgina Wood

Email: [email protected]

Search for more papers by this author
Ezequiel M. Marzinelli

Ezequiel M. Marzinelli

School of Life and Environmental Sciences, The University of Sydney, Sydney, NSW, Australia

Singapore Centre for Environmental Life Sciences Engineering, Nanyang Technological University, Singapore, Singapore

Sydney Institute of Marine Science, Sydney, NSW, Australia

Search for more papers by this author
Adriana Vergés

Adriana Vergés

Centre for Marine Science and Innovation, School of Biological, Earth and Environmental Sciences, Sydney, NSW, Australia

Sydney Institute of Marine Science, Sydney, NSW, Australia

Search for more papers by this author
Alexandra H. Campbell

Alexandra H. Campbell

USC Seaweed Research Group, University of the Sunshine Coast, Sunshine Coast, QLD, Australia

Search for more papers by this author
Peter D. Steinberg

Peter D. Steinberg

Centre for Marine Science and Innovation, School of Biological, Earth and Environmental Sciences, Sydney, NSW, Australia

Singapore Centre for Environmental Life Sciences Engineering, Nanyang Technological University, Singapore, Singapore

Sydney Institute of Marine Science, Sydney, NSW, Australia

Search for more papers by this author
Melinda A. Coleman

Melinda A. Coleman

Department of Primary Industries, National Marine Science Centre, Coffs Harbour, NSW, Australia

Search for more papers by this author
First published: 27 June 2020
Citations: 18


  1. Restoration is an emerging intervention to reverse the degradation and loss of marine habitat-formers and the ecosystem services they underpin. Current best practice seeks to restore populations by transplanting donor individuals chosen to replicate genetic diversity and structure of extant, nearby populations. However, genetic characteristics are rarely empirically examined across generations, despite their potential role in influencing restoration success.
  2. We used genomics to design a restoration program for lost underwater forests of Phyllospora comosa, a dominant forest-forming macroalga that went locally extinct from reefs off Sydney, Australia. Population genetic diversity and structure of nearby extant populations informed choice of donor sites. We tested whether donor provenance influenced adult transplant survival, condition (via metrics of epibiosis) and the genetic characteristics of recruits at restoration sites.
  3. Extant populations of Phyllospora within a 100-km radius of Sydney comprised three distinct genetic clusters with similar levels of genetic diversity. We transplanted reproductive adults from two of these sites, with the aim of restoring five Phyllospora forests with levels of genetic structure and diversity similar to donor populations.
  4. Donor provenance influenced survival and condition of transplanted adults and recruitment levels varied significantly among restoration sites. Yet, recruitment was rapid and genetic diversity and structure of the F1 generation resembled extant populations. This likely occurred because transplanted individuals reproduced synchronously and rapidly post-transplantation, prior to mortality of adult donor transplants.
  5. Synthesis and applications. As restoration and the need to ‘future-proof’ marine ecosystems increase globally, it will be critical to understand and harness the role of donor provenance, genetic diversity and structure in restoration success. By incorporating ecological and genomic data into restoration design and assessment, this study demonstrates that evidence-based selection of macroalgal donors can result in F1 generation recruits with similar levels of genetic diversity and structure as extant populations, despite an effect of provenance on transplant survival and condition. This study also highlights the need for ongoing refinement of transplantation techniques to ensure future recruitment success.


Global habitat degradation affects essential ecosystem functions and services and is one of the most serious threats faced by humanity (IPBES, 2018). Recovering degraded or lost habitats and maintaining ecosystem functions through active interventions, such as restoration, are now a major focus for management and conservation (UN, 2019). Understanding the factors that enhance restoration success is therefore crucial. In particular, decisions regarding the geographic or genetic origin (‘provenance’) of donor biological material can profoundly influence restoration success (Miller et al., 2017) but are only beginning to receive attention in many systems (Breed et al., 2018; Mijangos, Carlo, Spencer, & Craig, 2015).

Provenance generally defines the genetic diversity, identity and structure of a restored population and thus much of the population's ability to persist, respond and adapt to subsequent change (Bischoff, Steinger, & Müller-Schärer, 2010). Many guidelines for restoration practitioners recommend mimicking genetic characteristics of natural populations by sourcing donors from multiple extant, ‘local’ populations (Bischoff et al., 2010; Bucharova et al., 2019). Alternative strategies involving the use of distant populations and/or intentionally admixing to increase genetic diversity to enhance ecosystem function or stress tolerance are increasingly being proposed in response to human impacts such as landscape fragmentation and climate change (Broadhurst et al., 2008; Kettenring, Mercer, Reinhardt Adams, & Hines, 2014; Wood et al., 2019).

Most restoration programs, however, are still undertaken without any a priori knowledge of background population genetics (Mijangos et al., 2015). Even when some knowledge of provenance is utilized in restoration design, the maintenance of genetic diversity or structure and its relationship to restoration success is rarely quantified, particularly over more than one generation (Mijangos et al., 2015). This can lead to restored populations with poor genetic characteristics and subsequent failure (Granado, Neta, Nunes-Freitas, Voloch, & Lira, 2018; Williams, 2001). To rectify these gaps, the combined use of genetic tools and provenance trials embedded into the design, implementation and assessment of restoration projects is needed (Breed et al., 2018).

Underwater macroalgal forests underpin ecosystem goods and services along temperate rocky coastlines (Steneck & Johnson, 2013) but they are declining in many places around the world (Krumhansl et al., 2016). Restoration of macroalgal forests is still in its early stages (e.g. Layton et al., 2020) and long-term success is difficult to predict. The youth of such programs creates an opportunity to include provenance and other demographic considerations in their design.

Here, we used genomics to delineate appropriate provenance and assess effects on restoration of Phyllospora comosa (hereafter, Phyllospora), a dominant macroalga that forms extensive underwater forests along the south-east coast of Australia (Coleman & Wernberg, 2017). Phyllospora underpins coastal biodiversity and valuable ecosystem functions and services such as secondary production and nutrient cycling (Coleman & Wernberg, 2017), but disappeared from 70 km of Sydney's coastline in the 1970–1980s (Coleman, Kelaher, Steinberg, & Millar, 2008). While this decline was likely due to poor water quality, which has since improved (Coleman et al., 2008; Scanes & Philip, 1995), Phyllospora has not returned naturally—likely due to recruitment limitation. However, transplanted Phyllospora can survive and reproduce in Sydney (Campbell, Marzinelli, Vergés, Coleman, & Steinberg, 2014) demonstrating that active interventions are likely to be effective in re-establishing Phyllospora forests where they have been lost.

To identify appropriate provenance for Phyllospora restoration, we used single nucleotide polymorphisms (SNPs) to characterize genetic diversity, structure and the effect of geographic distance on extant populations surrounding the gap in distribution. We used this genetic information to design a restoration program spanning five sites within the distributional gap, with the aim to mimic genetic diversity and structure of nearby extant populations. We quantified differences in morphology, survival and condition of transplanted individuals of different provenance and assessed subsequent effects on the establishment, genetic diversity and likely origin of F1 (first generation of crossed) recruits. If provenance influenced restoration success via the survival and condition of transplanted adults, we predicted that this would result in differences in the genetic diversity and structure of the F1 generation compared to extant donor populations. We also used FST (genetic distance) outlier tests to determine the presence of genetic selection.


2.1 Study species

Phyllospora is a dioecious perennial macroalga found from Port Macquarie in northern New South Wales (NSW) to Southern Tasmania in Australia (Underwood, Kingsford, & Andrew, 1991; Wormersley, 1987). In NSW, Phyllospora forms dense beds on exposed rocky reefs from the low tide mark to c. 5-m depth (Underwood et al., 1991). It has a life span of ~2–6 years and is reproductive year-round (Coleman & Kelaher, 2009). Reproduction occurs only via sexual mechanisms, via spawning of motile sperm into the water column which fertilize stalked eggs attached to the fronds of female adults and drop to the seafloor (Burridge, 1990).

2.2 Determination of appropriate provenance for restoration

2.2.1 Sample collection

To characterize the patterns of genetic diversity and structure of extant Phyllospora populations, we sampled three sites north and three sites south of the distributional gap in Sydney [Bateau Bay (BB), Terrigal (TE), Palm Beach (PB), Cronulla (CR), Shark Park (SP) and Shellharbour (SH); Figure 1], over the Austral summer of 2016 (November–December). Sites were characterized by a canopy of predominantly Phyllospora, although some also had small numbers of the kelp Ecklonia radiata present. At each site, 30 individuals were haphazardly sampled (>1-m apart, see Coleman & Kelaher, 2009) at 1- to 5-m depth from 500 m2 of reef. Only reproductive individuals were sampled as this is the life stage that is used in restoration efforts. Lateral branches were removed from thalli and kept cool until processing (within 48 hr). From each individual, 30 unfouled apical tips were removed, rinsed in fresh water and dried to remove external salt, epiphytes and water (Coleman & Brawley, 2005). Samples were snap-frozen in liquid nitrogen and stored at −80°C.

Details are in the caption following the image
(a) Genetic structure of Phyllospora populations. Left: STRUCTURE plot showing individuals from extant populations assigned to three inferred clusters (A, B, and C). Each column represents an individual; different colours within columns indicate maximum likelihood probability of belonging to different clusters. Right: Map of extant and restored sites coloured according to average probability of belonging to each genetic cluster. Sites from top to bottom are as follows: BB, Bateau Bay; TE, Terrigal; PB, Palm Beach; WH, Whale Beach; FW, Freshwater; SO, South Head; CO, Coogee; MA, Maroubra; CR, Cronulla; SP, Shark Park; SH, Shell Harbour. The three pie charts on either side of the black semicircle represent extant populations and four smaller pie charts on the line represent recruits at restored sites. (b) Photograph showing transplanted algae from BB (shorter and lighter) and SP (taller and darker)

Frozen frond tips (~25 mg) were ground to a powder in a Qiagen Tissuelyser 2000 using stainless steel beads without thawing. DNA was then extracted using the Qiagen DNeasy Plant Mini kit with some modifications (see Appendix S1). DNA was then cleaned using a Qiagen PowerClean Pro Cleanup kit.

2.2.2 Genotyping

One hundred and seventy-seven samples from across the six sites were genotyped using an Agena Bioscience MassARRAY with iPlex GOLD technology on a custom panel of 354 SNP loci. These had been previously established by genotyping by sequencing (GBS) runs of seven samples at the Australian Genome Research Facility ( that were subjected to preliminary assay design (MassARRAY software) to select the top SNPs with a high minor allele frequency (MAF) and reasonable flanking sequence. The quality of genotyping was then assessed for each locus in the sample collection using the r poppr package (Kamvar, Tabima, & Grünwald, 2014). SNPs and samples with a call rate below 90% of the total, or with a MAF below 0.05 were excluded from further analysis.

As restoration success could be impacted by local adaptation, we tested whether extant populations were under any selection pressure prior to further analysis by searching for loci with high FST values using BayeScan 2.1 (Foll & Gaggiotti, 2008). A total of 10 separate runs were performed, from 50,000 to 500,000 iterations with a 10% burn-in period. After the runs, an FDR correction of q-values of 0.05 was applied to avoid the occurrence of false positives. No FST outliers were detected and the entire filtered data matrix was used for subsequent analysis.

2.2.3 Genetic diversity and structure in extant populations

Exact tests for Hardy–Weinberg equilibrium (HWE) deviations were calculated across all loci using hierfstat (Goudet, 2005). We then estimated linkage disequilibrium (LD) for each locus and across all loci using Fisher's exact tests in the r package genepop 1.1.2 (Rousset, 2008) with 10,000 dememorization and in 100 batches with 999 iterations per batch. An FDR correction was applied for multiple testing. None of the pairwise comparisons of loci were found to be significant in LD. Five loci deviated from HWE across all sites and were removed, leaving 113 loci across 177 samples.

Genetic differentiation and diversity were then evaluated by generating the estimates of observed heterozygosity (HO), expected heterozygosity (HE) and heterozygote excess (FIS) for each locus and for each sampling group using the r package diveRsity (Keenan, McGinnity, Cross, Crozier, & Prodöhl, 2013). FIS estimates were assessed for significance using 1,000 permutations with 95% confidence intervals. We calculated allelic richness (with allele counts rarefied by the minimum number of individuals genotyped) and the Shannon–Weiner diversity index using the hierfstat and poppr packages respectively.

Genetic structure was assessed by estimating Wier and Cockham's FST. Pairwise comparisons of FST between sampling locations and their significance were assessed using bootstrapping (999) to construct 95% confidence intervals in hierfstat. To identify the number of genetic groups in the dataset, we used STRUCTURE version 2.3.4 (Pritchard, Stephens, & Donnelly, 2000). The number of possible genetic clusters (K) varied from one to seven and was assessed using 20 independent runs with a 10,000 burn-in time and a Markov Chain Monte Carlo iteration of 200,000. This was performed with a model that allowed admixture and assumed correlated allele frequencies with no prior information. To determine the most probable value of real clusters (K), we used the ad hoc criterion (Evanno, Regnaut, & Goudet, 2005). The software CLUMPAK (Kopelman, Mayzel, Jakobsson, Rosenberg, & Mayrose, 2015) was then used to find the optimal alignment of multiple replicate analyses of each K and to display them. To calculate the percentage of genetic variation attributed among and within sites, an analysis of molecular variance (AMOVA) was performed using poppr, also using the broader genetic clusters identified with the STRUCTURE analysis (‘A’, ‘AB mixed’, ‘B’ and ‘C’ clusters; see Figure 1) as a priori groupings.

We also performed a Mantel test using the ade4 package in r (Dray & Dufour, 2007) to examine the relationship between geographic distance and genetic distance. For the geographic distance matrix, we used the r dist function to calculate the Euclidean distances in geographic space between collection sites based on their coordinates. For the genetic distance matrix, we used the collection site-based pairwise FST values generated from hierfstat.

2.3 Restoration experiment

Donor male and female Phyllospora thalli were haphazardly collected from SP (predominately characterized by the genetic cluster B, mixed with some of cluster A) and BB (predominately genetic cluster A; see Sections 3 and 4). Ninety ‘healthy’ (i.e. <5% of the thallus had visible marks of grazing, epibiosis and bleaching) individuals from each population were transplanted to each of five ‘restoration’ sites [Whale Beach (WH), Freshwater (SW), South Head (SO), Coogee (CO) and Maroubra (MA); N = 180 per site] that had previously been identified as suitable habitat (moderately exposed and having large, flat boulders at 4- to 5-m depth) in the austral spring of 2017 (October–November). Restoration sites were varying distances from the two donor sites (30–90 km; see Figure 1) but were all located within the Sydney metropolitan coastline and had no Phyllospora present. These habitats were composed of a mixture of ‘fringe’ (Underwood et al., 1991) which included the laminarian Ecklonia radiata and fucoid Sargassum sp., turfing coralline algae and barren. Differences in morphology between algae of each provenance were tested by comparing thallus length and number of branches using t tests in r. Algae were transplanted over 1–3 days by cable-tying individuals in natural densities (15 algae/m2) to 6 × 2 m2 plastic mats per site that had been attached to the rocky reef (as per Campbell et al., 2014). Mats were placed on top of bare rock or turfing corallines, 0.5 to 5-m apart depending on substrata availability. All transplanted thalli were treated in the same way (e.g. similar collection approach, time out of the water; Campbell et al., 2014). Individuals from the two donor sites were evenly distributed (mixed) across each of the mats at each site, and they were identified using cable tie attachment units of different colours (Figure 1).

2.3.1 Survival and condition of transplants

The number of adult algae from each donor site (BB or SP) with holdfast, stipe and fronds present was quantified in each mat within each site, 6 and 9 months after transplantation. To assess transplant condition, we also quantified the percentage of epibiosis, which can negatively affect algae and indicate stress (D'Antonio, 1985; Sogn Anderson, Moy, & Christie, 2019). To compare the effects of donor provenance (BB vs. SP, as a fixed factor), restoration site (random factor) and their interaction on (a) the percentage of survivors per mat and (b) percentage of epibiosis per individual, we fitted linear mixed models in the lme4 package followed by F tests with Kenward–Roger approximation using the KRmodcomp function in the pbkrtest package in r (Halekoh & Højsgaard, 2014). ‘Mat’ was fitted as a random effect to account for non-independence between individuals of different provenance interspersed on the same mat. This was done for the data collected at 6 months only, as the numbers of adults remaining 9 months following transplantation were very low (see Section 3). Normality and homogeneity of variance assumptions were checked visually with histograms of the residuals and scatterplots of residuals versus fitted values respectively.

2.3.2 Recruitment and origin of the F1 generation

We quantified total numbers of Phyllospora F1 recruits at each site 10 months after transplantation, when recruitment had occurred at most sites. We tested whether numbers of recruits found in or around each mat was dependent on total donor survival (at 6 months) using a linear model (data were pooled across sites).

We then quantified genetic diversity and the relative contribution of BB versus SP donors to reproduction by sampling recruits at each restoration site as described above, although only one apical tip was taken from each recruit to avoid mortality. Overall, genomic data from a total of 30 recruit samples [20 for population genetic analysis at the most successful site (CO) and two to five from the remaining sites for comparison of genetic assignment] were genotyped and refiltered as described above (extant populations). Only loci that were common to both donor and recruit populations following filtering were used. To test for potential selection, we pooled the samples from all sites and ran BayeScan with the run details described above. No FST outliers were detected, leaving 102 loci across 30 samples for subsequent comparative analysis.

As we had low sample sizes of recruits at most sites (see Section 3), measures of genetic diversity could only be meaningfully interpreted for the recruit samples collected from CO, where n = 20. To check the number of genetic clusters among donor and recruit groups, multivariate analyses were carried out using discriminate analyses of principal components (DAPC) using adegenet (Jombart, 2008). The optimal number of clusters was selected based on the lowest Bayesian information criterion (BIC). Genetic assignment of recruits to parent populations was then performed using Monte Carlo and K-fold cross-validation coupled with machine learning classification algorithms in the assignPOP package in r (Chen et al., 2018). We used the existing data from the BB and SP donor populations as reference groups and used the Bayesian model approach to estimate membership probabilities to these. Data were cross-validated with Monte Carlo simulations to assess self-assignment rates of individuals from donor sites to donor populations. We did not test for differences in genetic assignment of recruits among restored populations as recruit sample sizes were very uneven between sites.


3.1 Characterizing genetic diversity and structure of extant populations

Mean allelic richness and genetic diversity estimates varied only slightly among the six extant populations sampled, with no difference between observed and expected heterozygosity (Bartlett's test, p = 0.63; Table 1). Northern sites were characterized by low, positive FIS values (indicative of inbreeding trend), and southern sites were characterized by low, negative FIS values (indicative of outbreeding trend); however, none of the sites differed significantly from zero or each other (Table 1).

Table 1. Genetic diversity of Phyllospora from the six extant sites surrounding Sydney
Site n Total alleles ARa H o b H e c SD d FISe all loci Genetic clusterf
BB 26 234 1.98 0.34 0.34 3.26 0.01 0.81 0.16 0.02
TE 28 232 1.95 0.35 0.35 3.40 0.01 0.94 0.04 0.02
PB 29 231 1.94 0.31 0.32 3.37 0.03 0.47 0.40 0.12
CR 30 229 1.93 0.34 0.33 3.40 −0.03 0.7 0.25 0.04
SP 31 232 1.96 0.33 0.32 3.43 −0.03 0.03 0.93 0.03
SH 30 233 1.97 0.32 0.31 3.43 −0.01 0.03 0.035 0.93
  • Abbreviations: BB, Bateau Bay; TE, Terrigal; PB, Palm Beach; CR, Cronulla; SP, Shark Park; SH, Shell Harbour.
  • a Rarefied allelic richness (AR).
  • b Observed heterozygosity (HO).
  • c Expected heterozygosity (HE).
  • d Shannon–Weiner diversity (SD).
  • e Inbreeding coefficient (FIS). None significant.
  • f Proportion of genetic clusters represented at each site.

All pairwise FST tests between pairs of sites confirmed that populations were genetically different (global FST: 0.05; Table S1). STRUCTURE analysis revealed three genetic clusters (Table S2): BB and TE formed a ‘northern’ cluster, SP and SH each formed their own cluster with little admixture, while PB and CR (on either side of the gap in distribution) comprised mixes of the ‘northern’ and SP clusters (Figure 1). A small but significant amount of genetic variation was explained by genetic clusters (AMOVA, 3.15%, p = 0.01, Table S3) and sites (AMOVA, 2.99%, p = 0.01), while the majority of genetic variation occurred among individuals (AMOVA, 94.8%, p = 0.07). Geographic distance and pairwise FST values between each pair of sites were positively correlated (Mantel test, p = 0.01, r = 0.6, Figure 2).

Details are in the caption following the image
Relationship between geographic distance (km) and genetic distance (pairwise FST) between Phyllospora collected at six extant sites, fitted with linear regression. Comparison between the two donor sites Bateau Bay (BB) and Shark Park (SP) is shown in red. 95% confidence intervals shaded in grey. The minor peak around 60–70 km corresponds to the gap in distribution around Sydney

3.2 Differences between transplants from each provenance

At the time of transplantation, there were significant differences in the total thallus length (t86 = −3.48, p < 0.001) and number of branches (t66 = 7.58, p < 0.001) between algae from the two donor sites, with algae from BB significantly shorter and more branched (88.12, SE 3.81 cm and 24.88, SE 1.72 respectively) than algae from the south (107.02, SE 3.94 cm and 10.88, SE 0.66 respectively).

Between 13% and 40% of transplanted adults remained in restored sites after 6 months (Figure 3) and c. 8% (SE 2.73) remained across all sites after 9 months. At 6 months, survival of adults sourced from BB (31.3%, SE 3.56%) was significantly higher than those sourced from SP (19.1%, SE 3.56%; F1,4 = 11.002, p = 0.03). Most of the remaining algae had higher levels of epibiosis than when transplanted (<5%), with significantly higher epibiosis on adults sourced from SP (30.24%, SE 3.30%) than from BB (22.92%, SE 3.30%; F1,3.7 = 8.5970, p = 0.047). Both of these patterns were consistent across all sites.

Details are in the caption following the image
Transplant survival and condition. (a) Survival and (b) epibiosis of adult Phyllospora transplants at restoration sites after 6 months. Coloured bars depict the provenance. *p < 0.05

3.3 Recruitment and gene flow

Phyllospora recruits were observed at four transplant sites (Figure 4a), with most recruits found on mats (92%) or within 30 cm of these (8%) 10 months after adult transplantation, except for two recruits found 10-m away from the mats at Coogee. Total numbers of recruits differed widely between restoration sites (WH: 6, FW: 11, SH: 0, CO: 96, MA: 5) and were not significantly related to total adult survival at 6 months (F1,28 = 1.349, p = 0.26).

Details are in the caption following the image
Recruitment at restored sites. (a) Photograph of restored Phyllospora recruits at Freshwater. (b) Genetic structure of donor and restored (boxed in legend) populations, as inferred from discriminant analysis of principal components. (c) Membership probability plot showing probability of restored F1 recruits genetic assignment to Bateau Bay (BB) and Shark Park (SP) donors at four restored sites. Horizontal-dotted lines represent 95% confidence intervals for assignment to each site (=likely a pure bred cross between donor from the same site) (d) Summary table showing number of recruits found (total), genotyped (n) and assigned to donor populations, 10 months after transplantation. Percentages of recruits significantly assigned under 0.95 thresholds are shown alongside those assigned under threshold values of 0.8 (in brackets), as the accuracy of assignment was ~80%

Allelic richness and genetic diversity estimates varied slightly between the F1 generation and donor populations (Table 2). Observed heterozygosity was larger than or equal to expected heterozygosity in all sites, although this trend was non-significant (Bartlett's test, p = 0.41). All of the F1 populations were characterized by low, negative FIS values with WH, FW and MA having values that differed significantly from zero, which is indicative of outbreeding, although the number of replicates were low (Table 2).

Table 2. Genetic diversity of Phyllospora from donor populations and restored F1 recruits
Site n Total alleles ARa H o b H e c SD d FISe all loci
BB 26 204 1.37 0.36 0.36 3.26 −0.01
WH 2 173 1.39 0.38 0.29 0.69 −0.32
FW 5 188 1.36 0.39 0.32 1.61 −0.23
CO 20 202 1.36 0.36 0.35 3.00 −0.04
MA 3 182 1.37 0.38 0.30 1.10 −0.24
SP 31 202 1.34 0.35 0.33 3.43 −0.05


  • Significant values in bold.
  • Abbreviations: Donor sites in grey: BB, Bateau Bay; SP, Shark Park. F1 of restored populations in white: WH, Whale Beach; FW, Freshwater; CO, Coogee; MA, Maroubra.
  • a Rarefied allelic richness (AR).
  • b Observed heterozygosity (HO).
  • c Expected heterozygosity (HE).
  • d Shannon–Weiner diversity (SD).
  • e Inbreeding coefficient (FIS).

Discriminate analyses of principal components partitioned the donor and F1 recruit populations into two genetic clusters (Figure 4b). Across all transplant sites, 17 recruits were assigned to SP and 13 were assigned to BB (Figure 4c). Many individuals appeared admixed, likely indicating breeding between male and female donor algae from each of these sites respectively. Probability of assignment varied from 0.503 to 1, with lower probability of recruits assigned to BB (0.80, SE 0.07) than SP (0.89, SE 0.07; Figure 4d). Monte Carlo cross-validation using assignPOP showed the mean self-assignment rates for the BB and SP populations following training were 77% and 83%, respectively, indicating there was some uncertainty around delineation between genetic clusters.


Provenance is widely considered as vital for restoration success; however, it is rarely empirically examined. Here, we characterized three genetic groups of the dominant, forest-forming seaweed Phyllospora surrounding Sydney and, for the first time, used this genetic information to design a restoration program. Our empirical tests showed that, despite differences in survival and condition of donors of different provenance, the subsequent F1 generation had levels of genetic diversity and structure that resembled a mix of the extant donor populations, likely due to the rapid reproduction of transplanted individuals. The use of genomics thus provides critical information for restoring genetically diverse Phyllospora forests. Given that population mixing was achieved, similar mixing techniques could be extended to ‘future-proof’ populations via purposely selecting donors with desirable genetic traits in the context of climate change and other stressors. However, total recruitment varied significantly between sites, and recruitment levels were generally low. This suggests that while it is possible to use population mixing to restore desired levels of genetic diversity and structure, other factors such as donor provenance, transplant numbers and environmental factors at restoration sites must also be considered during future restoration efforts in order to achieve long-term success.

4.1 Extant population genetics and provenance choices

We found a small amount of genetic variation between extant populations, although they had similar levels of heterozygosity and no evidence of inbreeding overall. There was a trend for allelic richness to be slightly lower closer to the gap in its distribution near Sydney, possibly reflecting the impact of fragmentation of Phyllospora forests on dispersal and gene flow. Overall, our results are largely consistent with prior research indicating that Phyllospora has relatively low genetic diversity across its central range and disperses widely, facilitated by the presence of large, gas-filled vesicles that aid rafting on the ocean surface (Coleman et al., 2008, 2011). This previous work has suggested that dispersal is largely nonlinear and likely influenced by local eddies and/or northward-flowing currents that prevail inland off the coast of NSW. However, by using SNPs we were able to detect a weak but significant pattern of isolation by distance and previously uncharacterized genetic structure. This suggests that Phyllospora's dispersal patterns are also highly influenced by the East Australian Current, which flows in a north-southerly direction.

The lack of strong differentiation or evidence of selection in Phyllospora populations around Sydney indicated that restoring with the aim to resemble the genetic structure of surrounding extant populations could be achieved by sourcing donor algae from BB to SP, that is sites within a ~60-km radius of the centre of Sydney's coastline. Moreover, the genetic structure detected between the extant populations surrounding Sydney suggested that populations restored into Sydney should be comprised of a mixture, primarily of genetic clusters A and B. Sourcing from two, rather than one, area would lead to a more realistic genetic mix of these clusters while minimizing potential harvesting impacts on more genetically mixed sites bordering of the distributional gap (i.e. PB and CR).

4.2 Provenance effects on adult survival and condition

Provenance influenced Phyllospora transplant survival and condition, with higher survival and better condition of transplants sourced from BB. This was unlikely due to our transplantation method because algae from both sites were handled similarly. Differences in susceptibility to herbivory may explain some of the observed patterns, as restoration sites with greater fouling and differences in survival also had the highest levels of grazers (G. Wood, unpubl. data). Donors from SP had fewer branches and were longer than the ‘bushier’ algae from BB, potentially leaving them susceptible to higher stress and biomass loss caused by grazing. Generally, traits that influence herbivory in Phyllospora and other species, for example tissue chemistry and morphology, are highly plastic to local abiotic conditions and unlikely to be passed onto the next generation (Peters, 2015; Weigner, 2016). In this instance, high wave exposure and strong currents generally experienced at SP compared to BB likely contributed to Phyllospora's elongated morphology. Future restoration efforts intending to improve transplant survival and condition may however benefit from further work to determine the underlying mechanism(s) driving provenance effects and whether they manifest into adult stages of the restored F1 generation and select donors based on this.

4.3 Provenance effects on recruits

Restoration resulted in rapid reproduction; as a result, 80% of sites contained an F1 generation (i.e. the first generation to recruit in 3–4 decades) within 6 months. Where the F1 generation had similar recruitment density to natural populations (CO), observed heterozygosity and allelic richness were similar to that of donor populations. Interestingly, despite transplants from the north having higher survival, slightly more recruits were attributed to reproduction between donors from the south or were putative crosses between the north and south. This could be due to a slight bias in the validity of assignment test estimates because donor populations were very genetically similar to begin with, or a comparatively lower reproductive output from adults from the north. However, we contend that the F1 populations are likely composed of a genetic mix of contributions of each provenance due to the life history of Phyllospora and fucoids in general (Pearson & Serrao, 2006). Synchronous reproduction and gamete release are likely induced by osmotic and/or hydrostatic stress occurring during transplantation. This not only results in restoration achieving its aims of mimicking patterns of genetic diversity and structure of extant populations, but also overcomes a major bottleneck of marine restoration—the need to maintain adult donor plants on engineered structures in situ, that can be rapidly removed due to waves and during storms (Campbell et al., 2014).

While sites with low recruit density (Whale, Freshwater, South Head and Maroubra) exhibited some evidence of outbreeding (negative FIS values) in the F1 generation, we believe that outbreeding depression or maladaptation was unlikely to be the cause of lower recruitment as the level of genetic differentiation observed in this study is generally considered low. Instead, we contend that the low densities of recruits were likely due to the high levels of grazing at these sites (G. Wood, pers. obs.). Nevertheless, crossing over comparatively similar genetic distances has reportedly caused fitness problems in F2 generations of other marine species, for example pink salmon (Gharrett, Smoker, Reisenbichler, & Taylor, 1999). The effects of low recruitment at some sites are also likely to cause genetic bottlenecks as effective population size diminishes due to mortality. Restoration efforts are continuing at these sites and in some sites, recruitment continued to increase the winter following this study. Subsequent efforts to enhance recruitment at some sites have included supplemental transplantations to ‘boost’ recruitment numbers and in some cases, herbivore exclusion or removal. Future work monitoring gene flow and diversity in restored populations beyond the F1 generation will also be critical in determining such effects.

4.4 Implications and future directions for seaweed restoration

As underwater forests are predicted to continue to be affected by environmental stressors into the future, there is an increasing need to apply genetic and genomic techniques to underwater restoration projects (Wood et al., 2019). This study demonstrates that Phyllospora populations can be mixed to achieve desired genetic structure and diversity, at least in the F1 generation. Although results will vary depending on the algal species being restored, mixing donor populations of other dioecious seaweeds is also likely to conserve diversity and heterozygosity, perhaps more so than for other species with more complex life histories, such as the with true kelps, for example Ecklonia radiata or Macrocystis spp.

Additional positive effects of genetic diversity documented in other systems, for example, increased establishment rates, fitness and ecosystem functioning may also be conferred. However, further investigation of whether these apply to underwater forests is still needed. Future work in this field will also determine if mixing algae containing desirable genetic characteristics may be used to ‘future-proof’ populations. Determining ideal provenance mixtures and available genetic variation for selection would necessitate sampling populations at larger scales and would be facilitated by a multi-modelling approach which incorporates differences in regional and heritable differentiation (Breed et al., 2019; Rossetto et al., 2019). Experimental trials to determine progeny fitness and expression of desirable traits such as growth and productivity across scales and generations will facilitate this outcome (Weeks et al., 2011). Overall, the ability to manage genetic diversity, as has been demonstrated here, will likely play a crucial role in maintaining underwater forests into the future.


The authors thank the volunteers who contributed to this project, particularly Damon Bolton, Gary Truong, Sofie Voerman, Sam Baxter and John Turnbull. They are also grateful to Antoinne Minne for his assistance in composing Figure 1. This work was funded by the Australian Research Council through a Linkage Project to P.D.S., E.M.M., A.V., A.H.C. and M.A.C. (LP160100836), Discovery Projects to A.V. and P.D.S. (DP170100023) and P.D.S. and E.M.M. (DP180104041) and the Ecological Society of Australia through a Holsworth Wildlife Research Endowment to G.W. and E.M.M.


    All the authors conceived and designed the study; G.W. and E.M.M. collected the data; G.W., E.M.M. and M.A.C. analysed the data and led the writing of the manuscript. All the authors contributed critically to the drafts and gave final approval for publication.


    Data available from the Digital Dryad Repository (Wood et al., 2020).