Using transplantation to restore seagrass meadows in a protected South African lagoon

Site description

Langebaan Lagoon (33°08′ S, 18°03′ E) is a marine protected area forming part of the West Coast National Park, covering ~280 km² and stretches over a 38 km coastline (Fig. 1). It is the only estuarine tidal lagoon in South Africa, found within the cool-temperate bioregion (van Niekerk et al., 2019). The lagoon is marine dominated and permanently open to the sea, with no large freshwater inputs (Day, 1959) and characterised by intertidal mud and sandflats that support seagrass beds (Pillay et al., 2010). Seagrass extent within Langebaan Lagoon has been heavily impacted by human activities, resulting in ~38% loss of seagrass between 1960 and 2007 (Pillay et al., 2010), with some sites losing up to 99% of their seagrass cover (Pillay et al., 2010).

Transplantation site selection

Within Langebaan Lagoon, site selection was based on qualitative assessments using aerial photography over the last three decades. Areas showing repeated loss and recolonisation of seagrass, or presence of growth- limiting macroalgae, were avoided (Hauxwell et al., 2001) and areas of sandflats with remnants of small stable seagrass patches targeted (Lange et al., 2022). Pre-transplant field inspections assessed the environmental conditions of potential transplantation sites, including identifying potential localised human disturbance, any recently emerged seagrass patches, wave action and water depth range.

The donor site (Centre Banks, 33°07′18″ S, 18°02′42″ E) was selected for transplantation due to meadow extent and stability, and proximity to the restoration site (<1 km; Fig. 1). The restoration site (Oesterwal, 33°07′20″ S, 18°03′21″ E) was evaluated in 2020 as meeting the key site selection criteria for restoration (Fonseca, Kenworthy & Thayer, 1998; Calumpong & Fonseca, 2001) relative to the donor seagrass meadow since it (1) had comparable wave action, water depths and sediment type (<1.0 m mean sea level and sandy sediment; Compton, 2001), (2) had a history of supporting seagrass meadows, (3) had sufficient sandflat area lacking Z. capensis and macroalgae prior to restoration, (4) had limited human activity, that could lead to disturbance and (5) could support similar quality seagrass habitat (Fig. 1). All works were conducted under the following permits: Department of Forestry, Fisheries and the Environment (DFFE: RES2021-68, RES2022-17) and South African National Parks (SANParks: CRC/2023/017--2020/V1).

Transplantation method

Three transplant methods were carried out using donor seagrass from Centre Banks: (1) cores (diameter (ø) = 10 cm) with seagrass plants in original intact sediment; three seagrass shoots bundled and anchored with (2) metal pegs (ø 1.6 mm × ø 15 cm uncoated wire bent into a U-shaped stake), and (3) pre-soaked bamboo pegs (ø 1.2 cm × ø 15 cm bamboo cane with a vertical slit to create a V-shape; Fig. S1). Post collection, seagrass cores were placed in ø 12.5 cm plastic pots lined with cotton sheets to keep the sediment intact. Shoots were harvested using a spade and washed in seawater, thus ensuring a minimum of three rhizomal nodes with roots per shoot. During harvest, high density intertidal inner areas of the donor meadow were targeted over a ∼400 m² area to minimise localised disturbance. Donor material was harvested at low tide and transplanted in three intertidal sites (upper, upper-mid and mid, Fig. S2) at Oesterwal (Fig. 2) over a period of six hours per day with volunteer help. To prevent plant stress, cores and shoots were stored in large plastic containers and covered in pre-soaked cotton sheets to prevent desiccation before being planted by hand. For cores, sediment in the transplant plots was dug out using a corer, then seagrass cores were inserted. The planting depth of each core was carefully aligned to the sediment surface, with displaced sediment redistributed amongst the core and the adjacent sandflat to create a homogeneous elevation of the planting plot (Paulo et al., 2019). Shoots were bundled into groups of three per peg, with five pegs per equivalent core (15 shoots/core, from here on also referred to as cores; Fig. S1). The shoots were anchored with pegs and the rhizome was gently pushed 10 cm into the sediment (Lange et al., 2022).

Transplantation was conducted in a phased approach in 2021 during the austral winter. Translocation was staggered to spatio-temporally minimise risks during the transplantation phase (van Katwijk et al., 2009). Cores were transplanted between 26–28^th May, shoots anchored with metal pegs were transplanted between 27–30^th July, and shoots anchored with bamboo pegs were transplanted between 23–26^th August. Each type of transplant material was planted into spatially discrete transplant plots incorporating nine cores, using three different planting patterns: line, dense and bullseye (Fig. 2). Planting patterns were replicated three times per site, using fully crossed and repeated measures (Fig. 2). To account for differences in water depth, bioturbation, and physical stress (desiccation and wave stress) along the tidal gradient, treatments were replicated at three sites from the upper, upper-mid and mid-intertidal zone, separated by 50 m equidistant intervals (Figs. 2 and S2). A total of 81 transplant plots were planted using 243 cores and 7,290 shoots, with treatment replicates representing ~7.29 m² of transplanted Z. capensis.

Post-transplant monitoring

All transplant plots were left to acclimate overnight before photos were taken ~1.5 m above each plot to obtain baseline seagrass area data using an iPad Pro (1^st generation, Apple Inc., Cupertino, CA, USA) at low tide the following day. Following the transplants, a regular monitoring protocol was established; at first, monitoring of plots was conducted at low tide weekly for the first eight weeks post-transplant, then fortnightly for the following two months, then monthly, for a total of 18 months following transplantation for each planting material. Transplant plots were monitored over the following dates cores: 28^th May 2021–13^th October 2022, shoots anchored with metal pegs: 30^th July 2021–15^th December 2022, and shoots anchored with bamboo pegs: 26^th August 2021–16^th January 2023. Photos of each transplant plot were recorded to analyse seagrass area cover (cm²) and sandprawn bioturbator activity (determined by number of excavated K. kraussi holes per plot). Measures of epiphytic fouling (% relative cover), three seagrass blade lengths (mm) to determine average canopy height, and abundance of macro-epifaunal species were also taken for each transplant plot.

Data analysis

Within each transplant plot, area cover (cm²) and excavated K. kraussi hole density were analysed in ImageJ (version 1.54d; Schneider, Rasband & Eliceiri, 2012). Average canopy height (mm) was determined at each monitoring point by randomly measuring three shoots per plot, from the sediment to leaf tip, and taking the average. To calculate macrofaunal species diversity, the Shannon-Wiener diversity index was calculated with the vegan package. The transplant plot survival rate was obtained through the following equation:

$${\displaystyle \frac{(\mathrm{C}\mathrm{u}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{n}\mathrm{o}.\mathrm{o}\mathrm{f}\mathrm{s}\mathrm{e}\mathrm{a}\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{s}\mathrm{s}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{e}\mathrm{s})}{(\mathrm{I}\mathrm{n}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{a}\mathrm{l}\mathrm{n}\mathrm{o}.\mathrm{o}\mathrm{f}\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{n}\mathrm{s}\mathrm{p}\mathrm{l}\mathrm{a}\mathrm{n}\mathrm{t}\mathrm{e}\mathrm{d}\mathrm{s}\mathrm{e}\mathrm{a}\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{s}\mathrm{s}\mathrm{c}\mathrm{o}\mathrm{r}\mathrm{e}\mathrm{s})}\times 100}$$

All analyses were carried out in R (version 4.3.0; R Core Team, 2023). Generalized linear mixed models (GLMM) were used to examine the effect of the response variables on transplants survival and area cover over time. The response variables: planting material, planting pattern, site, epiphyte coverage (%) and number of sandprawn holes, were included as fixed factors. Transplant plot and monitoring timepoint (number of days since transplantation was used as monitoring was more frequent initially post-transplantation) were included as random effects, to account for taking repeated measurements of transplant plots over time. Normality and homoscedasticity of variance were assessed using the Shapiro-Wilk test and Levene’s tests respectively. Where these assumptions were not met, data was transformed (log x + 1) and models refitted. Main effects were assessed, followed by inclusion of interactive effects, then the optimal model structure was determined by selecting the model with the lowest Akaike’s Information Criterion (AIC) value. Diversity data were tested for normality, and where deviations found, data was transformed (log x + 1) and model refitted. Main effects were assessed as aforementioned, and model structure selected using the lowest AIC value. The ggplot2 package was used for graphical presentations.

Transplant survival (%)

A total of 243 cores and 7,290 shoots were transplanted across 81 transplant plots, with a mean survival rate of ~58% after 18 months across all treatments. Across all study treatments, all transplant plots survived for a minimum of 31 weeks (full dataset: https://github.com/vonderHeydenLab/Watson_et_al_2023_SeagrassRestoration), after which the loss of transplant plots differed across transplant source material, planting pattern and site (Fig. 3). There were multiple significant main effects and interactions that impacted transplant survival (Table S1).

Cores outperformed other transplant materials, with a survival rate of ~40% after 18 months across all treatments (Fig. 3). The GLMM showed that cores (p < 0.001), and shoots anchored with metal pegs (p < 0.001) showed positive, significant effects on survival (Table S1). Biotic interactions significantly affected survival, with lower numbers of sandprawn holes (p = 0.02), or lower epiphytic fouling (p < 0.001), leading to increased survival (Table S1). The interaction between a higher number of sandprawn holes and transplanted cores (p < 0.001), significantly reduced survival (Table S1).

The upper-intertidal site delivered the best outcomes for long-term survival (mean survival ~62% across treatments) for transplant plots, across all planting materials and planting patterns (Fig. 3). Transplant survival was highest in the upper-intertidal site, using cores planted in a bullseye pattern (mean survival ~93%; Fig. 3). The interaction between a lower number of sandprawn holes and transplant plots in the upper-mid (p < 0.001), and mid-intertidal sites (p = 0.01) significantly increased survival (Table S1).

The use of a bullseye planting pattern, in the upper-mid site and a lower number of sandprawn holes interacted with both cores (p = 0.05), and shoots anchored with metal pegs to positively affect survival (p < 0.001; Table S1). Similarly, survival significantly increased in interactions between cores (p = 0.02), or shoots anchored with metal pegs (p = 0.01) planted in a bullseye pattern and lower numbers of sandprawn holes (Table S1). The model also showed a significant increase in survival between the interaction of shoots anchored with metal pegs, planted in the bullseye pattern, in the mid-intertidal site, with lower numbers of sandprawn holes (p = 0.01; Table S1). In the GLMM intercept, comprising of shoots anchored with bamboo pegs planted in a line pattern in the upper-intertidal planting site, was significant (p < 0.001), with negative impacts on survival (Table S1).

Transplant area (cm²)

In surviving transplant plots, area cover increased on average by ~126% (mean across all treatments), with some plots expanding in area by over ~400% from across 18 months of monitoring (Fig. 4A). The GLMM showed using cores (p < 0.001), or shoots anchored with metal pegs (p < 0.001) had positive effects on area cover (Table S2). Conversely, area cover decreased when cores (p < 0.001) or shoots anchored with metal pegs interacted with higher numbers of sandprawn holes (p = 0.01; Table S2). Under the combination of using cores planted in a bullseye pattern in the upper-intertidal site, area cover increased by ~394% (mean across replicates; Fig. 4). The upper-intertidal site showed an average increase in area cover by ~191% (mean across all treatments and replicates; Fig. 4). Area cover was found to be negatively affected by interactions between cores transplanted into the upper-mid (p = 0.04), or mid-intertidal (p < 0.001) sites, or with higher numbers of sandprawn holes (p < 0.001; Table S2). Seagrass area cover increased when transplanted cores interacted with lower number of sandprawn holes in combination with either the upper-mid (p = 0.03) or mid-intertidal (p < 0.001) sites (Table S2). The GLMM intercept, made by shoots anchored with bamboo pegs planted in a line pattern in the upper planting site, was significant (p < 0.001), with negative impacts on area cover (Table S2).

Transplant area cover increased in transplant plots planted in a bullseye planting patterns, with ~147% (mean across replicates) average increase in seagrass cover after 18 months (Fig. 4B). However, cores planted in a bullseye pattern, with high numbers of sandprawn holes caused negative effects on area cover (p = 0.01; Table S2). With a lower number of sandprawn holes this combination of interacting factors (cores, planted in a bullseye pattern) positively affected area cover in both the upper-mid (p < 0.001), or mid-intertidal (p = 0.03) sites (Table S2). Higher numbers of sandprawn holes (p < 0.001), and higher epiphyte cover (%; p < 0.001) also had negative independent effects on area cover (Table S2).

All transplant materials were found to undergo a post-transplantation stress phase where seagrass cover initially decreased at 2 months, and then steadily recovered, as shown in Figs. 3 and S3. This pattern was observed across treatments, with more significant recovery in area cover found in transplanted cores, planted in denser planting arrangement on the upper shoreline (Fig. 3). Over time, transplant plots in the upper-mid and mid-intertidal sites steadily declined after signs of recovery, irrespective of transplant material, planting pattern or site (Fig. 3).

Across the monitoring period, 14 macro-epifaunal species were recorded within the transplant plots, including the endangered false-eelgrass limpet (Siphonaria compressa). The GLMM indicated that species diversity was positively affected by transplanted cores (p < 0.001), but negatively affected by shoots anchored with metal (p < 0.001) and bamboo pegs (p < 0.001; Table S3). Transplanted cores had the highest Shannon diversity index score, with diversity indices driven by the abundance of sandy anemones (Bunodactis reynaudi; full dataset: https://github.com/vonderHeydenLab/Watson_et_al_2023_SeagrassRestoration).

The authors acknowledge that although recovery of the donor meadow was not explicitly monitored from the study outset, a once-off monitoring visit 18 months post-transplantation (October 2022) found no evidence of visible holes or long-term meadow scarring. This suggests that the donor meadow had regenerated across the harvest area, and that sympathetic study design had facilitated donor meadow recovery without intervention.

Importance of site selection for transplant survival

The significance of site selection in restoration initiatives is well acknowledged by restoration ecologists. This study employed widely used restoration strategies to determine the applicability of current transplantation best practise in a South African context, showing that planting in the upper-intertidal shorelines increases seagrass survival (average ~62% across all treatments) and area cover (average ~191% across all treatments), compared to upper-mid and mid sites. Several studies have demonstrated that identifying sites with suitable growing conditions continues to require a disproportionately high number of transplantation trials (Leschen, Ford & Evans, 2010; Wendländer et al., 2019; Lange et al., 2022). By incorporating best practice guidelines to rigorously inspect field sites prior to transplantation, this likely improved restoration outcomes. For example, targeting areas with a history of seagrass growth, assessing local environmental conditions, and ensuring minimal anthropogenic disturbance all improved our understanding of local site dynamics (Fonseca, Kenworthy & Thayer, 1998; Calumpong & Fonseca, 2001). Our approach also allowed identification of potential biotic stress factors, including bioturbation and algal growth, with opportunities to integrate this work into habitat suitability modelling (Erftemeijer et al., 2023).

The effect of transplantation material and method on seagrass persistence and expansion

Implementing restoration projects using cores and anchored shoots is typically a trade-off between the desire for sufficiently resilient transplants (i.e., cores containing an engineered microbiome), and reducing donor meadow impacts. Post-transplantation stress is reduced in cores, and the risk of losing the transplanted area due to stochastic disturbance events is also lowered (van Katwijk et al., 2016; Paulo et al., 2019). In this study, transplanted cores outperformed other transplant materials, showing the highest survival (average ~40% across all treatments) and led to significantly increased area cover (average ~126% across all treatments). This outcome mirrors the results from the global analysis by van Katwijk et al. (2016), with cores showing higher integrated success than transplanted shoots anchored with pegs. Monitoring of the donor meadow after 18 months provided evidence that sensitive removal of donor material does not negatively impact donor meadows in the long-term.

Experimental seagrass transplant trials have suggested that using more compact planting patterns promotes long-term transplant persistence, particularly in isolated transplant plots, as clumping plants mimic dense root mats, promoting self-facilitation via sediment stabilisation and nutrient sharing (Morris & Doak, 2002; van Katwijk et al., 2016; Temmink et al., 2020). From our research, more compact patterns (dense and bullseye), showed higher survival, and the bullseye planting pattern significantly increasing area cover (average ~147%). This result may be linked to the more influential factor of site selection, which at a local scale may lead to interactions of spatially isolated Z. capensis transplant plots with other ecosystem engineering processes, such as bioturbators. In Langebaan Lagoon, Z. capensis has been shown to be spatially excluded by sandprawns (K. kraussi) in higher densities, that subsequently increase sediment suspension and destabilise sediments (Siebert & Branch, 2005; Hanekom & Russell, 2015). Qualitative field observations also showed bioturbator pressure from Greater flamingos (Phoenicopterus roseus) and Lesser flamingos (Phoeniconaias minor) which both form donut-shaped depressions due to trampling and their circular filter feeding technique, resulting in sediment resuspension, and increased nutrient flows, causing root destabilisation and enhanced biofilm production (Gihwala, Pillay & Varughese, 2017; El-Hacen et al., 2018). Bioturbator interactions in combination with (1) small, spatially isolated transplant plots, (2) planted at sites across a tidal gradient where sandprawn density is known to increase towards the lower intertidal zone (Siebert & Branch, 2006), likely led to negative density-dependent effects and account for significantly lower survival rates and area cover in the upper-mid and mid-intertidal zones, in comparison to the upper-intertidal zone, in our study.

Wider impacts of seagrass restoration in Langebaan Lagoon

In this study, re-establishing seagrass patches has a positive impact on seagrass-associated species, including for South Africa’s most endangered marine invertebrate, the false-eelgrass limpet (S. compressa), which only lives on Z. capensis and is endemic to Langebaan Lagoon (Angel et al., 2006). Species diversity was higher in transplanted cores (Table S3), likely due to the increased survival and area cover (Fig. 3). In total, 14 macro-epifaunal species were recorded, which is lower than recorded by Pillay et al. (2010) at the nearby Klein Oesterwal, likely explained by the increased survey effort across the tidal gradient and the inclusion of infaunal species in Pillay et al. (2010). Although species diversity was not investigated in bare sandflats, previous studies have highlighted that sandflats without seagrass (including through seagrass loss), have reduced richness and abundance (Orth et al., 2006; Pillay et al., 2010).

In addition to ecological service provision, this project also provided social benefits. The community impact of this work resulted in SANParks staff and 40 Project SeaStore volunteers contributing >1,600 working hours to assist with transplantation and monitoring, highlighting the project’s success in connecting communities with a typically lesser-known ecosystem. Longer-term, this is also likely to promote stewardship of seagrass meadows and aid future conservation efforts (Unsworth et al., 2019b).

Recommendations for future restoration trials in southern Africa

The principal challenge to upscaling Z. capensis restoration is the with vastly different environmental conditions found within estuaries across its range, therefore restoration efforts will need to be population and site specific (Adams, 2016). For example, attempts in two estuarine systems in South Africa, Klein Brak and the Knysna Estuary, both showed a 100% loss of Z. capensis transplants within 3-months (Mokumo, Adams & von der Heyden, 2023), despite following a similar approach to the work presented here. Both estuaries have strong freshwater inflows and fluctuations in physicochemical conditions, water levels and turbidity (Mokumo, Adams & von der Heyden, 2023), making their environments more variable than the conditions found in Langebaan, with changes in salinity a likely variable in the loss of transplanted cores in Klein Brak. Conversely, a restoration project in a relatively protected site characterised by sandy and muddy flats in Maputo Bay, Mozambique, also without significant freshwater flow, recorded survival rates of 75% after 12 months for transplanted Z. capensis cores (Amone-Mabuto et al., 2022). These studies highlight that site conditions and seagrass population dynamics are significant considerations before extending restoration attempts into other South African estuaries.

Another key challenge of working with Z. capensis is the lack of abundant seeds, making restoration using seeds unfeasible despite having shown considerable success in other Zostera species (Marion & Orth, 2010; Unsworth et al., 2019a). Work by Jackson (2022) discovered low intrapopulation clonality, which suggests sexual reproduction is more prevalent than once thought and a few inflorescences have been documented in several South African estuaries including Berg, Breede and Olifants in the west, and uMhlathuze, Mngazana and Swartkops estuaries on the south-east coast. However, without pursing seed storage (Yue et al., 2019) or controlled assisted evolution (Pazzaglia et al., 2021), it remains unlikely that the use of seeds for Z. capensis restoration will be an immediately viable option. Alternatively, pursuing techniques to reduce impacts on donor meadows, such as in vitro seagrass propagation methods (J Stephens, 2021, unpublished data) or growing donor material in seagrass aquariums to increase cover before transplanting back into the field (A Bossert, 2023, unpublished data) could also advance sustainable restoration efforts of dynamic meadows.

Retrospectively, bioturbator disturbance from flamingos was deemed to have more of an impact than anticipated, but was not explicitly monitored during this study. To overcome bioturbator pressure (sandprawns, flamingos and other wading birds), there is an opportunity for future restoration attempts to pursue ecologically sympathetic bioturbator exclusion methods. Novel approaches could also employ bird guano fertilisation to accelerate recovery (Kenworthy et al., 2018), or biodegradable establishment structures that mimic dense belowground root mats and suppress sediment mobility (Temmink et al., 2020). It is likely that supporting transplants with bioturbator exclusion measures would reduce post-transplantation stress and promote long-term persistence.