Testing habitat suitability for shellfish restoration with small‐scale pilot experiments

The global loss in ecosystem engineers has initiated calls for restoration, which includes the UN declaration of 2021–2030 as the decade of ecosystem restoration. As researchers dive into this decade it is important to consider the current state of an ecosystem to ensure restoration success. Pilot‐scale restoration has been recommended by global guidelines and standards as an effective starting point for restoration to provide valuable initial information to increase efficiency and success of subsequent larger‐scale restoration. To test habitat suitability, 4 t of green‐lipped mussels were placed onto three 2.25 m2 plots at five locations with differing benthic environments and assessed over 2 years. This study had a mean of 85% mussel survival at four of the five locations, while the restored shellfish in one location were completely extirpated, most likely by sea star predation. This pilot‐scale restoration revealed location‐specific differences that have implications for ensuring larger‐scale restoration success, including mussel survival, density, and health, and recommends a timeframe of at least 18 months to properly assess habitat suitability. Overall, the results validated the efficacy of using small‐scale pilot experiments to test habitat suitability and optimize location selection to maximize success and efficiency of larger‐scale restoration efforts.


| INTRODUCTION
Anthropogenic stressors can cause ecosystems to shift into an alternative stable state that can be difficult to reverse without a necessary large change or intervention (Beisner et al., 2003). This has been demonstrated with organisms that are ecosystem engineers, or habitatforming species, where a stressor causes a large-scale decline, resulting in a shift in the community to a lower, less biodiverse state. Some ecosystems can have many factors influencing decline, such as coral reefs, where climate change, disease, pollution, and overfishing are all compounding stressors on an ecosystem (Pendleton et al., 2016). In this case a single intervention is insufficient to return the degraded ecosystem back to a functional coral reef and multiple types of intervention are needed (Boström-Einarsson et al., 2020). Other ecosystems experience trophic disruptions from stressors, such as the loss of kelp forest ecosystems as a result of increased grazing pressure from sea urchins caused by overfishing of urchin predators (Catton, 2019). The anthropogenic stressors experienced by habitat-building species are often very localized, such as for reef-forming shellfish, where different impacts occur in different areas such as disease, pollution, and/or overharvesting which can diminish populations beyond recovery (Beck et al., 2009;Gillies et al., 2018;Newell, 2004;Paul, 2012). The natural recovery of these habitat-building species can occur in some instances, especially when causes of decline have been alleviated (e.g., Dankers et al., 2001;Jones & Schmitz, 2009). However, in other cases natural recovery does not occur, and intervention is required to initiate the recovery of these species and the ecosystem services they provide (e.g., Alder et al., 2021b;Schotanus et al., 2020;Schulte et al., 2009;Wilcox et al., 2018).
With increasing awareness of the decline of global marine ecosystems and the lack of natural recovery of key engineer species there are increasing efforts toward their conservation and restoration in many parts of the world Fitzsimons et al., 2020;Waltham et al., 2020). While the United Nations has declared 2021-2030 as a decade of international focus on ecosystem restoration, there have been concerns that there is a lack of understanding of ecological processes in some areas to underpin the success of restoration initiatives (Cooke et al., 2019;Waltham et al., 2020;Young & Schwartz, 2019). With a lack of understanding, marine restoration projects can be costly and inefficient, and even ineffective in some cases Toone et al., 2021). Inefficient ecosystem restoration initiatives can be compounded by a push to rapidly scale up restoration projects to achieve faster and greater outcomes (e.g., The Bonn Challenge; Papadopoulou et al., 2017). In contrast, the international standards produced by the Society of Ecological Restoration suggest testing small-scale recovery responses in areas where knowledge gaps are present (Gann et al., 2019).
Pilot restoration studies can be critical for determining the potential effectiveness of subsequent larger-scale restoration efforts by filling in knowledge gaps, validating methods, and improving overall restoration efficiency (e.g., Brumbaugh & Coen, 2009;Cook et al., 2022). Close monitoring of pilot restoration studies may also help to isolate specific causes of decline that can then be targeted for mitigation, along with providing a method for trialing new techniques (e.g., corestoration, Gagnon et al., 2021). Shellfish restoration best practice guidelines have recommended initial smaller scale pilot experiments to assess habitat suitability as a means to ultimately improve the effectiveness of wide-scale restoration initiatives (Fitzsimons et al., 2020). Small-scale shellfish restoration experiments have been undertaken with the intention to stay small-scale, for example when engaging in citizen science (e.g., Brumbaugh et al., 2000), and smaller restoration efforts over time have been shown to make a large impact (e.g., McClenachan et al., 2020). However, when restoration efforts are initially aiming to scale up to larger scales it may be beneficial to firstly perform a small-scale habitat suitability assessment experiment to determine factors that may impede restoration. Although smallscale pilot experiments may increase restoration time, cost, and resources, they have the potential to produce more efficient restoration efforts and therefore save time, money, and resources in the long run. In addition, pilotscale experimental methods may also be paired with other technical approaches, such as habitat suitability models, to increase success for large-scale location selection.
Green-lipped mussels (Perna canaliculus) are a species of reef-building shellfish endemic to New Zealand (Jeffs et al., 1999) and currently targeted for large-scale restoration due to the loss of the wild mussel beds and lack of recovery (Handley, 2017;Paul, 2012). Commercial dredging and handpicking from 1920 to 1990 extirpated extensive wild mussel beds (i.e., >1000 km 2 in total area) throughout New Zealand with little evidence for subsequent natural recovery since (Handley, 2015;McLeod, 2009;Paul, 2012;Wilcox et al., 2018). Although commercial harvesting pressure has been removed in most locations, the lack of natural recovery of former populations indicates that a restoration intervention may be necessary to jumpstart their recovery. The cause/s preventing the natural recovery of these once extensive mussels is unknown. Likely causes may include insufficient larval supply, a lack of established adult mussel habitat into which juveniles can recruit, and/or unsuitable environmental conditions due to environmental changes since wild harvesting ceased. For example, significant changes in the coastal marine environment during this time include increased sedimentation McLeod et al., 2012;Morrison et al., 2009), increased nutrient outputs , increased mussel aquaculture (Clough & Corong, 2015), and depletion of fish and macroalgal populations (Hauraki Gulf Forum, 2020;Urlich & Handley, 2020). It is likely that any or a combination of these stressors could have made the environmental conditions unsuitable for re-establishment of self-perpetuating mussel populations.
In this pilot restoration study, we evaluate small-scale experimental plots of mussels in shallow (5-7 m) subtidal areas where mussels were historically present to determine the current environmental suitability for supporting mussel populations. Specifically, this study aimed to test small experimental plots of restored mussels to inform decisions for future large-scale restoration efforts in the region. As well as its relevance locally, the study more broadly provides insight for using small-scale pilot experiments to assess habitat suitability to maximize efficiency and success of future restoration efforts in a new restoration area.

| Study area and location description
Pelorus Sound/Te Hoiere is located within the Marlborough Sounds at the top of Aotearoa-New Zealand's South Island (Figure 1). Pelorus Sound is a 50 km long drowned river valley system that is comprised of multiple arms and bays with a steep topography and high annual average rainfall leading to regular excessive sediment influxes (Urlich & Handley, 2020). Historically wild mussels were found in higher abundances in the inner Pelorus Sound, specifically Kenepuru Sound (Flaws, 1975;Stead, 1969), but these mussels had been extensively handpicked and/or dredged by the early 1980s . A lack of natural recovery of mussel populations in Kenepuru Sound has recently been evaluated and only 3% of the original wild mussel populations remain, almost entirely in the intertidal, which may have prevented their removal by dredging .
Five locations throughout the inner Pelorus Sound where mussels were present historically (Flaws, 1975;Stead, 1969) were selected inside a 30 km radius along a sediment mud and organic content gradient ( Figure 1; Grant Bay, M aori Bay, Skiddaw, Te Mara, Weka Point). In November 2019 the sediment grain size at each location was determined from samples collected using a Ponar grab (scoop area 0.05 m 2 , maximum depth 15 cm) at three points within each location, except for Grant Bay, which was too rocky for the grab and was instead sampled by SCUBA divers using sediment cores to collect interstitial sediment among cobbles or in rocky crevices. Sediment samples were pooled at each location and stored at À20 C until analysis where they were run on a Malvern mastersizer 3000 and classified using the Wentworth grain size classification (Wentworth, 1922).

| Mussel deployment
In January 2020 a total of 4 t of 2-year-old adult greenlipped mussels (mean shell length 91.3 ± SE 0.7 mm) from a mussel farm in Grant Bay were translocated to the five study locations. Donor mussels were grown from wild spat collected on settlement ropes in Pelorus Sound. The mussels grown on suspended ropes were collected using a commercial mussel harvesting vessel that had been modified so that the mussels remained largely in clumps held together by the mussel byssus threads. The clumped mussels were placed into 15 individual bulk bags, commonly used to transport mussel crop, with a target weight of 256 kg per bag. Six plots of 1.5 Â 1.5 m on the seafloor were randomly assigned as either a restored plot (mussels added) or a control plot (no mussels added), with each plot $2 m apart and at 5-7 m water depth. At each of the restored plots, one 256 kg bag of mussels was emptied into a 1.5 m Â 1.5 m quadrat, and the mussels spread out by divers. The location of each restoration and control plot were marked with the use of steel pegs or small weights with subsurface floats. All mussels were transferred into the plots within 24 h of harvest from the mussel farm.
Divers measured mussel survival and density by haphazardly placing a 0.25 Â 0.25 m quadrat three times in each restored plot at least 0.5 m from the edge, a method modified from Wilcox and Jeffs (2019). To estimate the survival and density of mussels the dead and alive mussels inside the sampling quadrat were counted, where either whole shells or two half shells were considered as one dead mussel. The length and width of each restored plot was measured at its greatest distance to estimate changes in areal extent. The percent survival of mussels for each plot was calculated by estimating the proportion of the live mussels out of the total of both live and dead mussels (empty shells) recorded for each restored plot.
Mussel condition was used as an indicator of mussel health status (Lucas & Beninger, 1985). A sample of 15 randomly selected mussels, five from each restored plot, were collected at deployment to determine initial condition and length, and at every subsequent sampling event, except for in February 2020 due to field logistics. Mussels were stored at À20 C until processing, where they were dried at 60 C for 48 h. The dry condition index was then determined according to a protocol from Lucas and Beninger (1985) using the following formula: dry weight of flesh Â 100/dry weight of shell.
The abundance and size of the common mussel predator, the eleven-armed sea star, Coscinasterias muricata, in the experimental plots was determined at every sampling point by divers gathering all sea stars from the area within the plot and 1 m beyond the perimeter for each of the restored and control plots. All sea star lengths were measured from the tip of the longest arm across the body to the tip of the opposite arm before being relocated approximately 2 km away from the collection locations to prevent them from returning to the study locations.

| Mussel settlement
In February 2021, 13 months after mussel deployment, artificial settlement collectors were placed inside the mussel plots for a 4-week period (February-March 2021) to assess settlement of mussels. The Marine Farming Association has been undertaking monthly spat sampling in the study area since 1975, so the sampling period was chosen from this data for the higher mussel collections in these months (Atalah & Forrest, 2019). Collectors comprised of a weight and a float attached to a square PVC pipe frame that holds three sections of 250 mm polypropylene rope. This rope is the same material that is used to catch mussel spat for aquaculture in New Zealand (Alfaro & Jeffs, 2003) as it mimics filamentous algae reported to be suitable for mussel settlement (Buchanan & Babcock, 1997). One collector was placed directly inside each mussel plot and three were placed at the same depth 50 m away from the mussel plots as controls at each location, a method modified from Wilcox et al. (2020). The collectors were recovered from the seafloor and the ropes were bagged and frozen until processing to aid in the removal of mussel settlers from the rope (Alfaro & Jeffs, 2003). Upon processing the ropes were rinsed with fresh water, the contents filtered through a 100 μm sieve, and the material retained on the sieve was examined and all P. canaliculus mussels were removed and counted.

| Statistical analyses
Assumptions of normality and equivalence of variance in the data were both visually assessed using a quantilequantile plot and then assessed with a Shapiro-Wilk test. A two-way analysis of variance (ANOVA) was used to compare the mean shell length, with location and sampling month as factors. To assess mussel condition index and plot area the data was log transformed to meet the normality assumptions and then a two-way ANOVA was performed. Post hoc Tukey tests were used to identify differences among locations and over time. Mussel survival and sea star counts were assessed using a Poisson general linear model (glm) due to the nature of count data. A linear regression was constructed to analyze the relationship of mussel survival and cumulative seastar abundances. Mussel density and mussel settlement numbers was assessed using a quasi-poisson glm due to overdispersion in the count data. Sea star length was assessed using a glm as it did not meet the assumptions of normality. Pairwise Wilcoxon rank sum tests were used to determine pairwise location differences in sea star counts, sea star lengths, mussel density and survival, mussel plot area, and mussel settlement. All tests were performed using R statistical software version 3.2.3 (R Core Team, 2021).

| Benthic assessment
Benthic sediment characteristics varied by location prior to the deployment of mussels along a mud and organic content gradient from Grant Bay to Weka Point (Table 1). Chlorophyll a did not follow the same gradient with the highest (8.1 mg/L) recorded at Te Mara and the lowest (0.7 mg/L) at Skiddaw (
Location and sampling date had a significant effect on sea star abundance, along with the interaction between the two variables (p < .05). Grant Bay recorded higher numbers of sea stars per plot than three of the five locations (Skiddaw, Te Mara, Weka Point; p < .05, Figure 3), while M aori Bay recorded higher sea star counts than Skiddaw and Te Mara (p < .05). Mean sea star counts across all locations were higher in June for both sampling years, although not significant, indicating a possible seasonality in the sea star predation. Sea stars collected at the most seaward location, Grant Bay and mid M aori Bay location were larger in size than the other three locations (25.3 ± 0.3 and 19.7 ± 0.4 cm, respectively, p < .001), which did not differ from each other (Skiddaw 15.5 ± 0.7, Te Mara 13.2 ± 0.9, Weka Point 16.1 ± 0.7 cm; p > .05). Sampling date had an effect on sea star length (p < .05) and there was a significant interaction between location and sampling date (p < .05), however, mean sea star length varied throughout the sampling period with no specific trends over time. Overall, the relationship between cumulative seastar abundances and mussel mortality indicated a significant relationship where higher abundances of seastars correlated with lower mussel survival (F (1,85) = 117; p < .001; R 2 = .579; Figure 4).

| Mussel density and area
The mussels spread out over time with mean mussel density decreasing from 915 ± 45 mussels/m 2 1 month postdeployment to 267 ± 15 mussels/m 2 from the four remaining locations after 2 years (Table 2). Location and sampling date had a significant effect on mussel density (p < .05), along with the interaction between the two T A B L E 1 Properties of the five locations in Pelorus Sound prior to mussel deployment along a mud gradient. (p < .05). All locations had significantly lower densities at their final sampling date (24 months) compared with the first sampling (1 month, all p < .05). The mussel density at Grant Bay differed from Skiddaw and Te Mara in an overall pairwise analysis (both p < .05), but no other location differences were found. At deployment all mussel plots were standardized at 2.25 m 2 , and after 1 month the mean area across the five locations was 5.6 ± 0.2 m 2 . The mussel plot area continued to increase significantly (p < .001) among the four surviving locations increasing to a mean of 17.7 ± 2.0 m 2 after 2 years (Table 2). Grant Bay did not continue to grow in area and by 13 months had a mean area of 5.9 ± 0.2 m 2 ( Table 2) and was significantly smaller than the other four locations (p < .001).

| Mussel growth and condition
Upon deployment in January 2020 the mussels all had high condition regardless of location (19.0 ± 0.4; p > .05; Figure 5), with no differences in mussel shell length (91.3 ± 0.7 mm; p > .05). After 5 months all locations exhibited a decrease in mussel condition, with Grant Bay having the lowest (6.8 ± 0.4) compared with Te Mara (11.6 ± 0.8; p < .001). Mussels at most locations remained at the lower condition for the 13-month sampling in February 2021 except for those at Weka Point, which reported an increase in condition almost comparable to that at deployment (14.7 ± 0.8; p > .05). After 2 years the mean shell length was 115.1 ± 1.2 mm and the mean condition was 11.0 ± 0.6 for the four remaining locations. All locations had significantly increased condition at 24 months compared to 13 months (p < .001) with Te Mara reporting the highest condition (15.2 T A B L E 2 Mussel condition, length, density, and plot area after 13 and 24 months. F I G U R E 4 The relationship between mussel survival and cumulative sea star abundance in each mussel plot recorded over 24 months across the five locations. Lines and shading represent 95% confidence intervals from the linear model. ± 1.0), that was comparable to deployment (p > .05; Table 2; Figure 5). Skiddaw had lower mussel condition (p < .05) and a lower increase in shell length than the other three remaining locations at 24 months (p < .05; Table 2). Overall analysis of mussel condition and mussel shell length showed a significant effect of sampling date (p < .001) and location (p < .001), and a significant interaction between sampling date and location (p < .05).

| Mussel settlement
The number of mussel settlers on the collectors varied by location (p < 0.05; Table 2), and only between control and mussel plots at Te Mara (Mussel 620 ± 112, Control 80 ± 27; p < .001). The three locations, M aori Bay, Skiddaw, and Weka Point had no difference in mussel settlement between each other (p > .05). Te Mara had the highest mussel settlement on the collectors than all other locations and Grant Bay had the lowest (p < .01; Table 2).

| DISCUSSION
One major factor that can impede shellfish restoration success is post-transplantation survival. This study tested a potential technique to mitigate the risk of high mortality by using small-scale pilot experiments to test habitat suitability, optimize location selection, and thereby improve prospects for successful outcomes from subsequent larger-scale restoration initiatives. This study demonstrates the productive use of small-scale pilot experiments for restoration in a new area, particularly due to the ability for multiple factors in habitat suitability to be tested in conjunction, and revealed location specific differences in predator abundances, environmental factors leading to decreased mussel condition and growth, and juvenile mussel settlement ( Figure 6).
This pilot experiment tested the suitability of five locations inside a 30 km radius across a sediment mud and organic content gradient, with location having little effect on initial mussel survival. However, 1 year post-restoration one location (Grant Bay) experienced a dramatic decline in survival, eventually resulting in the loss of all deployed mussels by 18 months. Overall, the mean survival of all mussels in this study was 73.1 ± 10.1% after 24 months, which is higher than recorded in one F I G U R E 6 Conceptual diagram summarizing location specific differences in habitat suitability that were determined using pilot-scale experimental mussel plots in Pelorus Sound, New Zealand.
F I G U R E 5 Mean mussel condition per restored mussel plot over 24 months at each of the five restoration locations in Pelorus Sound, New Zealand. Error bars show 95% confidence interval.
green-lipped mussel restoration study in the North Island of New Zealand that reported a 26.2 ± 4.6% survival after 2 years postdeployment across four similarly mud, sand and shell hash sites, most likely due to predation and/or smothering from sediment (Wilcox et al., 2018). Deploying mussels into dense clumps formed by byssus threads prior to harvest may have helped to improve their establishment in the present study, including their resistance to predators, as has been found for the blue mussel, Mytilus edulis (Bertolini et al., 2018) and for subadult green-lipped mussels elsewhere in New Zealand (Alder et al., 2021a).
One month postdeployment the dense mussel beds appeared to form strong byssal mats regardless of benthic environment. The density of mussels steadily decreased naturally over the 2-year period as the mussels selforganized by spreading out into clumps or as individuals died. The extent of this behavior varied among locations, but this did not appear to correspond with differences in mussel survival between locations. However, this behavior did appear to align with sediment grain size composition, with mussels spreading out more to reach lower densities at locations with higher mud content. Facilitating initially higher densities and allowing for mussels to self-organize over time may aid in overall mussel survival, as mussels have been shown to stay tightly clumped as a defense mechanism, where attaching to conspecifics provides anchoring security (Commito et al., 2014). Mussel density and plot area are not typical metrics reported or targeted for shellfish restoration . However, measuring density and area in small pilot restorations can help to show location differences that allow for targeted restoration plans, specifically the number of mussels needed to cover an area, and how that may need to be altered at locations with known predators or high wave disturbance to ensure high mussel survival.
Little is known about the ecology of the eleven-armed sea star, although this species is known to pose the greatest threat to the persistence of both restored and natural populations of mussels in other parts of the country (e.g., Ohiwa Harbor; Paul-Burke & Burke, 2013). Elevenarmed sea stars actively migrated into the restored mussel beds from the surrounding habitats throughout the 2 years of sampling as sea stars were fully removed after each sampling point and were always found in all of the plots again at the next sampling point. In addition, these sea stars preferentially targeted the restored beds, as there were very few sea stars found in the control plots. These results confirm behavior previously recorded in a mussel restoration effort in the Hauraki Gulf of the North Island of New Zealand, with starfish predation attributed to 30% of the total loss of the restored mussels (Wilcox & Jeffs, 2019). High densities of these sea stars can commonly be found underneath marine farms (Inglis & Gust, 2003), likely feeding on mussels dropped during the harvesting and cleaning process, which may explain their high abundance at Grant Bay as that location was ca.100 m from a marine farm. In comparison, Te Mara was over 2000 m from the nearest marine farm and experienced the lowest predation, while the other three locations ranged from 400 to 1000 m from a marine farm. Elevated winter predation activity of this sea star (Wilcox & Jeffs, 2019) may explain the higher numbers of sea stars sampled on the restored mussels in June (i.e., Austral mid-winter) for both years of sampling. Sea stars were less abundant at locations with increasing mud content, a finding that could help explain historical patterns of mussel abundance in the inner parts of Pelorus Sound (Flaws, 1975;Stead, 1969). Without assessing habitat suitability prior to location selection for largescale restoration, it can be difficult to gauge the type of environmental stressors that will impact transplantation survival. In this study, eleven-armed sea star predators appeared minimal upon location assessments using dropdown cameras and at mussel deployment by diver observations, but due to their mobile nature only a pilot experiment revealed the significant extent and variation in localized predation pressure from this predatory species.
Efforts to visually assess new juvenile mussel recruitment by divers to the restored mussel plots over the 2-year period were unsuccessful and may be due to new recruits being difficult to see in often marginal underwater visibility. In a mussel restoration experiment in the North Island of New Zealand only three new mussel recruits were visually recorded by divers over a 2-year period (Wilcox et al., 2018), as compared with the use of artificial settlement collectors that indicated adult mussel beds increased settlement success (Wilcox et al., 2020). In this study, mussel settlement onto settlement collectors was greatest at Te Mara, the location that experienced the least current flow and the highest mussel condition. Settlement was significantly greater on collectors within mussel plots compared with control sites at this location, supporting the findings of Wilcox et al. (2020). Although no juvenile recruits were found, larval supply does not appear to be limiting recruitment to restored mussel beds in the Marlborough Sounds as sampling in the area has revealed no substantial decline in mussel settlement over the last four decades (Toone et al., 2022). Understanding recruitment is vital for the long-term success of restored mussels and performing pilot restoration experiments can help identify if juvenile recruitment is already happening into restored areas and the possible barriers that may be occurring in that process. For example, in oyster restoration small-scale restoration efforts have been used to assess bottlenecks in recruitment including substrate limitations in an area (Brumbaugh & Coen, 2009). This initial pilot research may be particularly important in an area that is thought to have limited larval supply (Fitzsimons et al., 2020), as restoration methods may need to be adjusted to enable the restored mussels to selfsustain in the long term, increasing efficacy of the restoration methods.
When assessing habitat suitability for shellfish restoration an important consideration is food availability for mussel health and growth, particularly in early life stages (Phillips, 2002). Encouragingly, the location with the highest condition and growth (Te Mara), also had the highest juvenile mussel settlement. Although average mussel growth found in this study (11.9 mm/year) was comparable to a similar restoration study in North Island of New Zealand (11.7 mm/year; Wilcox et al., 2018), it was much lower than recorded for longline ($63 mm/ year) aquaculture green-lipped mussels of a similar size (Dawber,2004;Hickman,1979). This was expected as longline mussels are known to grow faster due to the less stressful environment in the water column (Dawber, 2004;Hickman, 1979). One of our study locations (Skiddaw) had a slower growth rate than the others, suggesting that growth in this location may have been constrained in some way (i.e., by food availability; Marsden & Weatherhead, 1999). Furthermore, the mussels at this location also had lower condition compared with the other four locations, and there were lower levels of chlorophyll a found in the sediment at this location. The mean overall mussel condition at deployment (19.0 ± 0.4) was similar to the mussel condition recorded from farmed mussels used for restoration in the Hauraki Gulf in the North Island (16-18;McLeod et al., 2012) and was higher than previously recorded from wild intertidal populations in the South Island (6.95-8.77; Marsden & Weatherhead, 1999). The decline in mussel condition index that occurred at all locations after the mussels were deployed to the seabed may be a result of the stress associated with the translocation process, as has been seen in other restoration studies (McLeod et al., 2012;van Kampen, 2017), or a seasonal change that occurs naturally with mussels due to processes such as recent breeding activity decreasing condition (Çelik et al., 2012). Overall, Weka Point and Te Mara had higher condition than the other locations. However, both locations had the highest mud content in the sediment, in contrast to other studies with lower mussel condition recorded in higher suspended sediment loadings (McLeod et al., 2013). Considering the mussel condition and growth results from pilot restorations can help identify location specific factors that may impact overall mussel health and survival. Condition can also be used as a proxy for fecundity which has been shown to decrease with lower food availability (Hickman et al., 1991;Oliveira et al., 2021). Thus, choosing locations that have ample food availability and using mussels with higher condition may increase future larval supply and therefore increase the chance of creating self-sustaining restored beds.
One consideration needed for using small-scale pilot experiments prior to larger-scale restoration is the length of time necessary to properly assess habitat suitability and reveal location differences in restoration success. The results of this study indicate that timeframe is particularly important, as it took 12 months for Grant Bay to begin to decline in survival and 18 months for the mussels at that location to be completely extirpated. These results are indicating that longer than 18 months may be required to assess habitat suitability in mussel restoration, particularly with regards to understanding the survival and growth of new recruitment. Long-term monitoring (>5 years) has been suggested for larger-scale restoration efforts in order to allow sufficient time for evaluation of full ecosystem recovery , however, 1-3 years is recommended for developing and monitoring pilot studies by global shellfish restoration guidelines (Fitzsimons et al., 2020). Restoration effort timelines can be limited by funding, in some cases, but this study indicates that inefficient location selection cost can be much higher as it could result in the whole restoration effort being unsuccessful. The results of this research, along with global shellfish guidelines (Fitzsimons et al., 2020) recommends the use of pilot experiments in a new area targeted for restoration. However, if this is not feasible, this study indicates that in initial location assessments more direct assessments of predator recruitment abundances, food availability, and larval settlement may aid in initial location selection. This study shows that small-scale pilot restoration experiments can reveal location-specific differences in habitat suitability. This knowledge can be used to improve location selection, produce habitat suitability models, and increase the success of large-scale restoration efforts. These location differences include the abundance of predators, and environmental factors such as limited food availability that affect mussel survival, health, growth, and density.
Although differences between locations were ascertained, the overall mussel survival, growth, and condition over 2 years at four of the five locations in this smallscale pilot experiment confirms that the benthic environment in this New Zealand system remains suitable for the survival of mussels, despite previous studies indicating significant historical changes in the benthic environment in the last 50 years (Handley, 2015). The results of this study suggest that timeframe is critical when assessing habitat suitability and recommend a length of at least 18 months. Although small-scale pilot experiments may increase restoration time and resources, these results indicate that without testing habitat suitability first any larger-scale restoration efforts would have been hindered by inefficient location selection that could have led to high levels of predation or low food availability. Using small-scale pilot studies first therefore saves costs in the long run by identifying location specific factors that will reduce restoration success, thus allowing for efficient location selection ensuring higher survival and success of the restoration effort. These implications are important for increasing global restoration success, particularly in new areas where large-scale declines of habitat-building species may have resulted in a shift to an unsuitable stable state.