The Wadden Sea is the world’s most extensive intertidal flat system and stretches along the coasts of the Netherlands, Germany, and Denmark (Figures 1A, B). This shallow coastal sea has been awarded UNESCO World Heritage status, reflecting its high ecological and economic importance. We performed the experiment on two separate study sites in the Western Dutch Wadden Sea: Zachte Bed Oost (53°6’25.664’’N, 5°9’11.538’’E) and Vlakte van Kerken (53°6’21.168’’N, 4°55’11.982’’E) (Figure 1C), with annual mean salinities of 26.4 PSU (± 3.9 SD), and 28.6 PSU (± 2.5 SD), respectively (van Weerdenburg and Vroom, 2021). According to the procedure for reintroduction measures defined by the Standards of the Society for Ecological Restoration, suggesting reintroducing an organism in its original range from which it has disappeared (Gann et al., 2019), we chose sites located in or close to former eelgrass beds. Zachte Bed Oost is located in an area that had been populated by subtidal eelgrass before it vanished in the 1930s, and Vlakte van Kerken is located close to a former eelgrass bed (Figure 1C, green polygons). Both locations fall under the Natura 2000 protection, as the entire Dutch Wadden Sea has this protection status. In addition, Zachte Bed Oost is within an area closed off to mussel seed fishery since 2015. Vlakte van Kerke is closed off for hand-cockle fishery and in close proximity (~200 m) to an area closed off to shrimp and mussel seed fishery, which are the most prominent bottom-disturbing fisheries in the Dutch Wadden Sea. Considering legal restrictions, like marine protection status, improves efficient habitat selection (Pogoda et al., 2020). Furthermore, the sites’ shallow bathymetries make it physically impossible for mussel seed- and shrimp fishers to fish on both sites. Potential study sites were pre-selected based on suitable bathymetry, sediment grain size, and position within areas partially closed to fisheries. In the scope of a system analysis combined with field measurements in 2020 and 2021, we searched for locally low to moderate flow velocities, relatively high underwater light availability (measured as photosynthetically active radiation (PAR)), and minor wave height. Based on their promising sets of abiotic conditions, we chose Zachte Bed Oost and Vlakte van Kerken as our two study sites. Their shallow bathymetric positions (-135 cm and -130 cm below mean water level (MWL), respectively) ensure that seagrass remains permanently submerged and lays well within the depth range of former seagrass beds of the 1930s, which were growing at depths between -50 cm and -230 cm below MWL (van der Heide et al., 2007). A shallow water column during low tide might enable more light to reach the plants, leading to less light limitation than in deeper areas. Light limitation is currently one of the most pressing bottlenecks for subtidal eelgrass restoration in the Dutch Wadden Sea (van der Heide et al., 2007, 2006). For light-saturated growth, subtidal eelgrass requires 7 mol photons m^-2 day^-1 (Thom et al., 2008). In June 2021, we measured a mean photosynthetic photon flux density (PPFD) of 7.0 mol photons m^-2 d^-1 at Zachte Bed Oost and 9.0 mol photons m^-2 d^-1 at Vlakte van Kerken. These values are at or above this saturation limit, indicating that the sites might meet the light requirements.

Eelgrass grows in two morphologically distinct varieties: flexible intertidal and robust subtidal (van Katwijk and van Tussenbroek, 2023 and sources within). In 2022, 650 hectares of intertidal eelgrass grew in the Dutch Wadden Sea due to a long-term research project through seed-based restoration (Gräfnings et al., 2023b). While remarkable restoration success has been accomplished for the intertidal variety, the subtidal variety, which grows under permanently submerged conditions, is still absent. For the subtidal eelgrass variety, one unsuccessful restoration attempt was carried out in 1950 (Korringa, unpublished data). Since then, this study has been the first to investigate restoration possibilities further within the scope of a field experiment. Due to COVID-19 border restrictions, we could not access donor plants from the nearest preferred donor site in Limfjord, Denmark. Instead, the collection of approximately 900 seagrass shoots took place in June 2021 in the bay of Eckernförde (54°27’49.9’’N, 9°56’42.0’’E), Germany, at 2 m depth (~14 PSU). Multiple shoot counts were performed by placing a 0.5 × 0.5 m metal frame on randomly selected patches within the meadow, resulting in a conservative estimation of 235 shoots m^-2. Based on recent satellite images, the donor meadow covered approximately 12,000 m². Hence, our harvest impact was estimated to have affected less than 0.05% of the meadow. To minimize the impact on the donor meadow, apical shoots, including rhizomes and roots, were carefully collected by hand at the edge of multiple patches. To fulfill permit requirements to limit the risk of unwanted transport of seagrass-associated species, sediment was removed entirely from the belowground structures, followed by a thorough rinse of aboveground plant structures, removing all visibly attached species. This step was a requirement of the Directorate General for Public Works and Water Management (Rijkswaterstaat) (registration nr. RWSZ2021–00011036). During transportation, seagrasses were kept moist and cool. Overnight storage of the plants took place in aerated artificial seawater with higher salinity (Tropic Marine Classic, 20 PSU) to aid the acclimatization of the plants from the Baltic Sea to the higher salinity levels of the Wadden Sea. We constructed transplant units (TPUs) out of three seagrass shoots with 10 cm long rhizome fragments (mean wet weight = 15.2 gr, SE = 5.6), which we anchored with a 20-gram metal nail with thin metal wire (0.5 mm), according to the method established in Denmark by Lange et al. (2022). Transplant units were constructed one day after harvest and kept in shaded Wadden Sea surface water for 1.5 days until planted.

We performed the experiment at two sites: Vlakte van Kerken and Zachte Bed Oost (Figure 1C). We subjected the seagrass transplants to two treatments: altered hydrodynamic forcing (barrier vs. control) and altered sediment stability (BESE vs. control stability). The treatments were applied in a full-factorial design, resulting in four treatments replicated 12 times at each study site (n = 12).

For the first treatment, to manipulate hydrodynamic forcing, we constructed two barriers per site with jute bags filled with ~33 kg sand. We used mortar sand (overall grain size ≤ 3mm), which has sharp edges and a more complex structure, increasing the stability of the bags. The bags were piled up (L:H:W: ~7:0.5:0.6 m) in a slightly bent curve with an arc length of 7.2 m at a radius of 7.0 m and 59°central angle (Figure 2A). The direction of the barriers was perpendicular to the predominant directions of current flow and waves to shelter the seagrasses from the incoming tidal current. Predominant wind direction data from the Royal Netherlands Meteorological Institute (KNMI) was used as a proxy for predominant wave direction. We measured the predominant heading of the current flow with tilt current meters (Lowell Instruments LLC, East Falmouth, United States) three months before the experiment. Per site, the two barriers and bare controls were arranged alternating in a line next to each other (Figure 2C). We stacked three layers of 1 × 10 m coconut fiber mesh (Ø 10 mm; mesh size = 2 cm) under each of the four barriers and secured them with four L-shaped rebars (l = 50 cm; Ø 10 mm). One meter behind each barrier or bare control, we semi-randomly arranged twelve plots in two rows containing equal numbers of plants treated with increased or control sediment stability (Figure 3).

For the second treatment, to increase sediment stability, we buried three-dimensional artificial structures, mimicking seagrass root mats, completely sub-surface. The structures consisted of Biodegradable Elements for Starting Ecosystems (BESE-elements, http://www.bese-products.com/, Figure 2B) (BESE BV, Culemborg, Netherlands), composed of a carbon neutral biodegradable polymer made from potato-waste Solanyl C1104M (Rodenburg Biopolymers, Oosterhout, the Netherlands) (Temmink et al., 2020; Marin-Diaz et al., 2021). We stacked two layers of BESE (45 cm × 45 cm), resulting in a 4 cm high 3-D honeycomb-shaped matrix with a 12 cm diameter circular planting hole cut in the center. Each structure was secured with three L-shaped rebars (l = 50 cm; Ø 10 mm). The control sediment stability plots (45 cm × 45 cm) consisted of bare sediment. The experiment was set up (i.e., barrier construction and sediment-stabilizing structures burial) one month before planting to allow sediment to settle around the restoration tools. To start the experiment, we manually planted three seagrass transplant units just below the sediment surface in the center of each plot within a diameter of 12 cm. For this, we lifted the sediment with an elongated trowel and gently pushed the units into the gap at an angle. The experiment was terminated thirteen weeks after planting.

We monitored the seagrass transplants while snorkeling during daylight low tides, which occurred approximately monthly. At Zachte Bed Oost, we monitored 1, 5, 8, and 13 weeks after planting. Similar monitoring timing occurred at Vlakte van Kerken, except that the 8-week sample was impossible due to high turbidity that impaired underwater visibility. Plants did not grow sufficiently to determine growth parameters like expansion or elongation rate, so we assessed whether the transplant units still contained alive shoots. We checked whether the three transplant units held live shoots at each plot. We recorded the death of a transplant unit when all leaves of the same transplant unit were absent or completely discolored or when the entire transplant unit was missing. To monitor transplant dislodgement, we recorded the presence or absence of transplant anchors (i.e., nails) in week five at both locations. We took pictures of all plants during monitoring events. Additionally, we visually estimated the epiphyte cover when visibility and duration of the low tide during monitoring allowed it. We estimated the percentage of epiphyte cover of the seagrass leaves twice at Zachte Bed Oost (weeks five and eight) and once at Vlakte van Kerken (week five). The experiment was terminated at the last monitoring event, after thirteen weeks, when no surviving shoots were present anymore.

We monitored the abiotic conditions of current flow velocity and heading direction with tilt current meters (Lowell Instruments LLC, East Falmouth, United States). The loggers recorded current data at a sampling rate of 8 Hz and bursting duration of 10 seconds per ten minutes. To prevent the heading and current velocity measurements from being expressed relative to the magnetic north, we corrected for the magnetic declination, which is the angle between the magnetic north and true north. We determined the declination of 1.48° east with the “National Centers for Environmental Information (NOAA) Declination Calculator” (https://www.ngdc.noaa.gov/geomag/calculators/magcalc.shtml), according to the experimental site’s locations in the western Wadden Sea. The current reduction occurred during incoming tides through the barriers’ position regarding the seagrass plots. Therefore, we categorized the records’ heading degrees into incoming (Vlakte van Kerken: ~20°; Zachte Bed Oost: ~40°) and outgoing tides (Vlakte van Kerken: ~180°; Zachte Bed Oost: ~230°), for the velocity reduction calculations. For both sites, data of barriers and controls was aligned based on time stamps and expressed as percentual current velocity reduction or current attenuation. The maximum wave height was measured with wave gauge sensors (Ocean Sensor Systems Inc., Coral Springs, United States) at a sampling rate of 10 Hz and bursting length of 7 minutes per 15 minutes. We derived the minimum and maximum pressure values per measurement moment. From that, we calculated the maximum wave height per burst. Light intensity was monitored as photosynthetically active radiation (PAR) throughout the experiment, approximately 50 m next to the experimental setup with Odyssey Integrating PAR Sensors (Dataflow Systems PTY Limited, Christchurch, New Zealand). Loggers were cleaned off barnacle and algae biofouling during monitoring events, and only PAR data collected within three weeks after cleaning events were used in analyses to ensure the reliability of the data, as the PAR sensors were especially susceptible to biofouling. In addition, eight weeks after setting up the barriers, hence four weeks after planting, sediment surface samples were taken by extracting the top 5 cm layer with a 50 ml syringe (∅ 3 cm). We took three samples between seagrass plots behind each barrier and control, totaling 24 samples per site (Figure 2C, brown circles). Samples were analyzed for median sediment grain size and silt percentage (% with sediment grain size < 63 μm). Freeze-dried sediment samples were analyzed by the Royal Netherlands Institute for Sea Research (NIOZ, Texel) with a particle size analyzer (Beckman Coulter LS 13 320, Aqueous Liquid Module) using Polarization Intensity Differential Scattering (PIDS) technology. We took drone pictures of the experimental setup eight weeks after planting. Bathymetric information of the exact transplanting position and large-scale changes in bathymetry within the experimental setup were collected using the Trimble® R8 GNSS System (Global Navigation Satellite System) rtk-dGPS with a vertical resolution of 15 mm, one week and four weeks after planting by either measuring at the plots directly or in an approximately 1.5 × 1.5 m resolution around the setup. Due to the unexpected nature of the large bathymetry change next to the barriers, we did not take repeated bathymetry measurements in the scour holes. Bathymetry differences over time are, therefore, likely an underestimation.

Statistical analyses were performed with R version 4.3.2 (R Core Team 2014). Throughout this study, we refer to the standard error with SE. We predicted the odds of shoot survival with a Generalized Linear Mixed-Effect Model (GLME) using the lme4 package (Bates et al., 2015). We converted the survival data into a binomial distribution. We considered a plot as having survived (1) if at least one of the three seagrass transplant units still had at least one alive shoot. On the other hand, we noted death (0) if none of the three seagrass transplants contained living shoots. We set hydrodynamic treatment, sediment stability treatment, experimental site, time (in days after transplantation), and interaction between time after transplantation and study site as fixed effects. Transplant unit ID and plot ID were applied as nested random effects. Model estimates on a logit scale were back-transformed by exponentiation to generate survival odds. We computed 95% Confidence Intervals (CIs) and calculated p-values using a Wald z-distribution approximation. For the abiotics, we fitted linear mixed-effect (LME) models with the nlme package (Pinheiro et al., 2023) to predict the effect of the hydrodynamic barriers on current flow and wave height, with the hydrodynamic treatment as a fixed effect and time as a random effect. Models were run separately for both locations. Additionally, we fitted a linear model with the median grain size as the dependent variable and hydrodynamic treatment and experimental site as independent variables. The difference in silt content between the study sites was analyzed non-parametrically for barrier and control treatments using a Kruskal-Wallis test, as assumptions for normality were not met. We validated model assumptions by plotting residuals versus fitted values to verify homogeneity and assessing Q–Q plots of the residuals to check for normality and residuals versus each explanatory variable to check for independence. Additionally, the Shapiro-Wilks test (α >.05) was used to test for normality of variance, and Levene’s test (α >.05) was used for testing homogeneity of variance. Changes in bathymetry depending on the presence of the barriers were tested with a Welch two-sample t-test due to unequal variances of the groups.

We investigated if the reintroduction of self-facilitating feedbacks can facilitate the restoration of subtidal eelgrass in a dynamic ecosystem. For this, we tested the effects of increased sediment stability and the manipulation of hydrodynamic forcing on seagrass survival. The reported survival probabilities are not absolute but relative survival chances. They are expressed in relation to the survival probabilities of the other level of the same treatment. In line with our hypothesis, we found that short-term survival chances of eelgrass transplants treated with sediment stabilizing root mimics were +67% higher than the bare controls (GLME, beta = -1.11, 95% CI [-1.96, -0.26], p <.01) (Supplementary Table 1). The mean survival of the four different treatment combinations was similar at the same monitoring event, with mean survival peaking at 15%. At Zachte Bed Oost, treatment effects were visible after five weeks, with increased sediment stability resulting in increased survival. We found a highly significant interaction effect between transplantation time and study site, as shoots survived longer at Zachte Bed Oost (beta = 0.18, 95% CI [0.11, 0.25], p <.001) (Figure 4). Interestingly, the overall seagrass survival at Zachte Bed Oost was still higher after eight weeks than after five weeks at Vlakte van Kerken, so seagrass transplants survived for a more extended period at Zachte Bed Oost. After five weeks, we discovered that the survival rate of plants treated with the barriers at Vlakte van Kerken was merely 8% of the survival rate observed at Zachte Bed Oost. This suggests a stark difference in survival rates between the two locations, with Vlakte van Kerken showing lower plant survival relative to Zachte Bed Oost. Contrasting to our expectations, barriers decreased transplant survival probabilities by -74% (beta = -1.33, 95% CI [-2.29, -0.36], p <.01).

All seagrass shoots died in the course of three months after transplantation, so there was no long-term survival of the transplant units. We noticed that many shoots and leaves broke off at a straight edge, approximately 3 cm close to the ground, where leaves split from the leaf sheath (Supplementary Figures 1A, B). The remnants were often overgrown with epiphytes (see Supplementary Figure 1A for a representative photo) and/or discolored entirely. We found that the proportion of mortality caused by dislodgement was relatively low in most treatment groups (Supplementary Figure 2). At Vlakte van Kerken, the median percentage of mortality caused by dislodgement was 16.7%, while at Zachte Bed Oost, it was 4.2%. Interestingly, at Zachte Bed Oost, we recorded both the highest and lowest proportion of mortality due to uprooting. The BESE treatment without a barrier resulted in 0% mortality due to dislodgement, while the barrier treatment without BESE resulted in more than 30%.

The abiotic conditions at both sites did not differ substantially. We found a tidal range of ~1.9 m at Zachte Bed Oost and ~1.7 m at Vlakte van Kerken. The median sediment grain sizes at the two study sites were 176 μm (SE = 2.4) and 174 μm (SE = 2.4), respectively. From June to August 2021, the average current flow velocities without barriers (16.15 cm s^-1, SE = 0.09) were significantly higher at Zachte Bed Oost compared to the average at Vlakte van Kerken (14.72 cm s^-1, SE = 0.09) (beta = 0.76, SE = 0.17, t_(18426) = 4.51, p <.001). We recorded maximum current flow velocities of 42.63 cm s^-1 and 42.21 cm s^-1 and mean wave heights of 11.26 cm (SE = 0.11) and 8.55 cm (SE = 0.08) (Supplementary Table 2) at Zachte Bed Oost (Figure 5A) and Vlakte van Kerken (Figure 5B), respectively.

We measured the light irradiance on land and underwater as Photosynthetic Photon Flux Density (PPFD), the sum of all recorded photosynthetically active radiation on a day. The land light irradiance when we conducted the experiment was relatively low, as the 2021 mean (55.32 mol photons m^-2 day^-1; SE = 2.75) was 28% lower than the 2022 mean (76.95 mol photons m^-2 day^-1; SE = 2.96) (Figure 6B). Our underwater measurements (Figure 6A) revealed that the monthly mean PPFD in June, July, and August 2021 at Zachte Bed Oost (7.0; 12.3; 5.9 mol photons m^-2 d^-1, respectively) and Vlakte van Kerken (9.0; 12.4; 9.2 mol photons m^-2 d^-1, respectively) laid well above the threshold for long-term survival of established meadows of 3 mol photons m^-2 d^-1. 75% of the recordings at Zachte Bed Oost and 86% of the records at Vlakte van Kerken exceeded the long-term survival threshold. Furthermore, measurements drew close to or exceeded the light requirement for light-saturated growth of 7 mol photons m^-2 d^-1 (Thom et al., 2008), as 48% and 58% exceeded the threshold for light-saturated growth at Zachte Bed Oost and Vlakte van Kerken, respectively (Figure 6A). It is important to address the difference between the measured and the actual light availability, as light-competing species (epiphytes and macroalgae) likely reduced the light availability (Sand-Jensen, 1977; Hauxwell et al., 2003).

At Vlakte van Kerken, very high loads of macroalgae, predominantly sea lettuce (Ulva lactuca), grew and floated throughout the water column. Sea lettuce attached in very high quantities to the barriers, sensor poles (Supplementary Figure 3), and base of the seagrass transplant units. The high weight of the floating macroalgae suffocated the seagrass. At Zachte Bed Oost, only a small amount of floating macroalgae was present. In our drone picture (Figure 7), the quantity of macroalgae can be seen. It did not seem to affect the seagrass negatively.

We quantified the epiphyte cover on seagrasses once at Vlakte van Kerken after five weeks and twice at Zachte Bed Oost after five and eight weeks. We found that after five weeks, at Vlakte van Kerken, the seagrass epiphyte cover was significantly higher than at Zachte Bed Oost (F_1,1 = 25.96, p <.01). Here, only approximately half of the epiphyte cover was detected. Comparing the cover at Zachte Bed Oost at 5 and 8 weeks reveals no statistically detectable difference between these monitoring events (F_1,1 = 2.78, p = .17) (Figure 8). Epiphyte volume (Supplementary Figure 1A) and weight added considerable drag to the plants leading to breakage of the leaves.

When growing in low densities, sparse seagrass rhizome networks may not be able to stabilize the sediment (Suykerbuyk et al., 2016). When reintroducing this self-facilitating feedback with transplants in a bare system, very high quantities would be required to initiate the sediment-stabilizing feedback. Artificially inducing the same feedback can minimize the need for donor material (Temmink et al., 2020). Thus, the root-mimicking structures that we used for the sediment treatment were designed to induce this stabilizing feedback. Seagrass restoration can be aided by overcoming the settlement threshold of low sediment stability with BESE structures (Gagnon et al., 2021). Studies in Sweden and the US found increased restoration success for subtidal eelgrass transplants by applying the buried root mimics (Temmink et al., 2020; van der Heide et al., 2021). However, the restoration success can be conditional (van der Heide et al., 2021), as too little hydrodynamics rendered the BESE unnecessary, while too strong hydrodynamics overpowered the effect of the root mimics. Apparently, the hydrodynamic conditions in the Wadden Sea, where our study was carried out, lay within the windows where BESE is useful. It also means that if our barriers had worked in the anticipated way, the relative effect of the root mimics would likely have been smaller, and these measures have the potential to interact. Here, we found that BESE increased short-term survival probabilities, meaning that seagrass restoration success in the Dutch Wadden Sea may benefit from artificially induced sediment-stabilizing feedbacks.

Contrary to our expectations, the 7-m hydrodynamic sandbag barriers did not have the intended effect of creating a calmer growth environment for seagrass transplants. Furthermore, many BESE-elements were washed clear, reversing their effect. The barriers were ineffective in reducing maximum wave height and reduced current velocities by less than 10%. Here, one should keep in mind that these values might be underestimated, as we only measured 3.4 m behind the barriers. Regardless, the barriers did not have their desired effect on seagrass survival, and their ameliorating effect was seemingly too low for ecological relevance. In a mussel reef restoration experiment in the Dutch Oosterschelde estuary, breakwaters did not increase the mussel coverage, probably because the barriers were too shallow or narrow to sufficiently reduce hydrodynamic stress (Schotanus et al., 2020). In our study, an increased barrier height might have asserted ecologically relevant hydrodynamic stress amelioration. Because the barriers deflected water, scour holes formed along the edges, likely due to locally increased current flow and resulting sediment transport. The pronounced edge effect would probably be relatively smaller if barriers were longer. Furthermore, an alternative barrier design might result in a lower edge effect, as, for instance, a semi-permeable barrier would not deflect as high volumes of water.

Installing functional underwater barriers is challenging in practice because of the high hydrodynamic forces they are supposed to ameliorate. Even though several attempts in close-by systems were made to establish structures targeted specifically at seagrass restoration, only a few results have been published. In the Danish estuary Vejle Fjord, barrier boulder reefs were installed to facilitate seagrass restoration within the ongoing “Sund Vejle Fjord” project. Similar to our observations, the barriers considerably affected sediment dynamics after installation. The effects on eelgrass restoration attempts cannot be quantified before a new steady state of the sediment dynamics is established (T. Banke, pers. comm., Feb. 7, 2024). In the Dutch Lake Grevelingen (R. Cronau, pers. comm., Feb. 9, 2024), the Dutch intertidal Wadden Sea (L. Govers, pers. comm., Jan. 24, 2024), and the German Baltic Sea (L. Kamperdicks & M. Paul, pers. comm., Feb. 1, 2024), underwater barriers were dislodged due to the hydrodynamic forcing. In the Dutch Lake Markermeer, a series of breakwaters were successfully used to relieve hydrodynamical stress and stimulate the growth of submerged macrophytes. However, this restoration effect occurred ten years after the installation, attributed to underwater light limitations (van Zuidam et al., 2022). These examples highlight the challenges and the necessity of applying restoration tools to reduce hydrodynamic stress below its limiting threshold.

The current velocity maxima for established eelgrass meadows range for instance between 0.5 – 1.8 m s^-1 (Koch, 2001 and sources within) and 1.2 – 1.5 m s^-1 (Fonseca et al., 1983). We conducted field measurements in the summer of 2022 and found mean current velocities of 0.27 m s^-1 at Zachte Bed Oost and 0.26 m s^-1 at Vlakte van Kerken, values well below the stated threshold. However, dislodgement of unanchored transplants may start at lower hydrodynamic threshold values of 0.16 – 0.23 m s^-1 (Carus et al., 2022), as transplants may be more vulnerable to dislodgement and the adverse effects of high current speeds than established meadows. This was why our experimental design included, firstly, shoot anchorage and, secondly, the barrier treatment to mimic the hydrodynamic forcing reducing feedback of an established meadow. Even though the monitoring values during our experimental phase were lower than in 2022, the hydrodynamic stress, a combination of waves and currents, still seemed too high.

Conducting transplantation efforts across different sites leads to spreading the risk, which is important in dynamic coastal systems like the Wadden Sea (van Katwijk et al., 2009). Our study further evidences this, as we found significant differences in survival between the two study sites. We showed that the survival probabilities when planted at Vlakte van Kerken were 77% of the survival probabilities at Zachte Bed Oost. Depending on the treatment, seagrass at Vlakte van Kerken declined from 100% to 3 – 11% in the course of only five weeks. The second site, Zachte Bed Oost, showed higher and longer short-term survival, but survival percentages dropped to similar lows three weeks later. While the abiotic conditions of both sites seem quite similar (Supplementary Table 2), the biotic conditions deviated strongly.

The presence of antagonistic species like fast-growing pleustophytic macroalgae can compromise the success of transplantation activities (Sfriso et al., 2023). Vlakte van Kerken functions as a basin where high macroalgae loads accumulate. The border to the next tidal basin is 2 km north of the site (Figure 1C), and 1.4 km to the west lies the barrier island Texel. Combined with the incoming tides from the south from the tidal basin “Marsdiep”, this leads to macroalgae accumulation. The algae biomass (predominantly Ulva lactuca) attached to the barriers and transplants (Supplementary Figure 3) seemed to have suffocated the seagrass. This observation aligns with previous studies in the Wadden Sea (Bos et al., 2004, 2005; van Katwijk et al., 2009). Other adverse effects of floating algae are competition for light (Hauxwell et al., 2001, 2003) and unfavorable biogeochemical conditions like lowered redox conditions and potentially toxic concentrations of ammonium (NH₄⁺) (Hauxwell et al., 2001).

For our site selection, a system analysis was performed based on abiotic factors like bathymetry, sediment grain size, position within areas partially closed to fisheries, flow velocity, underwater light availability, and wave height. To effectively choose an optimal site, it is essential to consider not only abiotic factors but also to incorporate biotic data into the decision-making process. Next to floating and epiphytic macroalgae, examples of species that could affect seagrass habitat suitability in the Wadden Sea are bioturbators like blow lugworms (Arenicola marina) (Philippart, 1994), grazers like mud snails (Peringia ulvae) (Gräfnings et al., 2023a), and herbivores of seedlings like the European green crab (Carcinus maenas) (Infantes et al., 2016). Taking the occurrence and the species’ antagonistic or mutualistic effect into account could further improve effective site selection.

Even though we found treatment effects on short-term survival, no long-term survival occurred. When seagrasses are exposed to single or multiple stressors, the plants’ carbon budget undergoes alterations: the carbon acquisition can be lowered, and the reserves can be drained. Energy can be allocated to defense or repair processes instead of growth, reducing growth and increasing mortality (Moreno-Marín et al., 2018). Abiotic differences between donor and transplantation sites led to sudden changes in water composition (salinity mean and fluctuation) and light availability. The roots of the single shoots were physically disconnected from the surrounding nutrient-providing sediment. Using rhizome fragments, shoots were disconnected from their extensive rhizome network and access to energy through carbon storage. Furthermore, as the transplantation occurred relatively late in the growing season (i.e., mid-June), the timing probably also adversely affected the survival. Lastly, light was limited due to fairly low irradiance during the experimental period, with epiphytic growth on the seagrass leaves further limiting light availability. Especially at Vlakte van Kerken, the high floating and epiphytic macroalgae loads overshadowed the treatment effects of the barriers and root mimics. We hypothesize that the high exposure to cumulative stress caused seagrass die-off in the scope of this experiment.

Our study is an important first step towards restoring the once vast (~150 km²) subtidal eelgrass meadows in the Dutch Wadden Sea. This study identified several current bottlenecks and insights for knowledge-based decision-making for future restoration activities. We show that artificially reintroducing positive feedback, i.e., sediment stabilization, positively affected short-term survival and suggest incorporating measures that induce this feedback in future restoration activities. Even though the second treatment was unsuccessful in inducing the positive feedback of creating shelter from hydrodynamic forces, we remain convinced that reducing hydrodynamic forces could be an important measure to bridge establishment thresholds for seagrass (Temmink et al., 2020). The installation of inter- and subtidal barriers is challenging due to the high hydrodynamic forces they are designed to ameliorate (R. Cronau, pers. comm., Feb. 9, 2024; L. Govers, pers. comm., Jan. 24, 2024; L. Kamperdicks & M. Paul, pers. comm., Feb.1, 2024) and altered sediment dynamics (T. Banke, pers. comm., Feb. 7, 2024). The barriers need sufficient height and length to ameliorate the hydrodynamic forces to an ecologically relevant degree (Schotanus et al., 2020). A positive effect can be overshadowed by other bottlenecks like low light availability (van Zuidam et al., 2022) or the presence of antagonistic species (Sand-Jensen, 1977; Hauxwell et al., 2001). It should be further investigated if barriers with an adjusted design (e.g., altered height, length, permeability) can reduce hydrodynamic forcing to an ecologically relevant degree. Another aspect that requires further research, as the sudden translocation to a system with twice the salinity levels (14 vs. 28 PSU) likely posed additional stress, is whether donor plants can tolerate the Wadden Sea salinity fluctuations and if an acclimatization phase with gradual salinity increases promotes field survival.

We recommend lowering high cumulative stress levels by carefully considering the following aspects when planning future restoration activities. Firstly, an eelgrass population from a donor meadow with more similar abiotic Wadden Sea conditions should be used (van Katwijk et al., 2009). Secondly, the annual variations of carbon reserves (Vichkovitten et al., 2007) should be considered when planning transplantation activities. Lastly, we found that the site selection played a major role in the survival of seagrass transplants, as macroalgae (Supplementary Figure 3) and epiphytes (Figure 8; Supplementary Figure 1A) competed for light (Sand-Jensen, 1977; Hauxwell et al., 2001). They added drag and physical stress to the plants on one of our experimental sites (Vlakte van Kerken). This added drag resulted in higher dislodgement of the transplant units. To effectively choose an optimal site, it is essential to consider a habitat suitability map based on human impact and abiotic factors (Pogoda et al., 2023, 2020) and incorporate biotic data into the decision-making process.

To conclude, for seagrass restoration in the Wadden Sea, one should carefully consider 1) the reintroduction of positive feedbacks through restoration tools, 2) lowering high cumulative stress levels regarding donor population choice and timing, and 3) site selection, taking both local biotic (e.g., floating macroalgae) and abiotic conditions (e.g., light conditions) into account. Optimizing these restoration facets might lower the additive stress that eelgrass transplants face to the degree that allows long-term survival in the Dutch Wadden Sea.