Testing alternate ecological approaches to seagrass rehabilitation: links to life-history traits

Authors


Correspondence author. Southern Seas Ecology Laboratories, DX 650 418, School of Earth and Environmental Sciences, University of Adelaide, Adelaide, SA 5005, Australia. E-mail: andrew.irving@adelaide.edu.au

Summary

1. Natural resources and ecosystem services provided by the world’s major biomes are increasingly threatened by anthropogenic impacts. Rehabilitation is a common approach to recreating and maintaining habitats, but limitations to the success of traditional techniques necessitate new approaches.

2. Almost one-third of the world’s productive seagrass meadows have been lost in the past 130 years. Using a combined total of three seagrass species at seven sites over 8 years, we experimentally assessed the performance of multiple rehabilitation methods that utilize fundamentally different ecological approaches.

3. First, traditional methods of transplantation were tested and produced varied survival (0–80%) that was site dependent. Secondly, seedling culture and outplanting produced poor survival (2–9%) but reasonable growth. Finally, a novel method that used sand-filled bags of hessian to overcome limitations of traditional techniques by facilitating recruitment and establishment of seedlings in situ produced recruit densities of 150–350 seedlings m−2, with long-term survival (up to 38 months) ranging from 0 to 72 individuals m−2.

4. Results indicate that facilitating seagrass recruitment in situ using hessian bags can provide a new tool to alleviate current limitations to successful rehabilitation (e.g. mobile sediments, investment of time and resources), leading to more successful management and mitigation of contemporary losses. Hessian bags have distinct environmental and economic advantages over other methods tested in that they do not damage existing meadows, are biodegradable, quick to deploy, and cost less per hectare (US$16 737) than the estimated ecosystem value of seagrass meadows (US$27 039 year−1).

5.Synthesis and applications. This research demonstrates how exploring alternate ecological approaches to habitat rehabilitation can expand our collective toolbox for successfully re-creating complex and productive ecosystems, and alleviate the destructive side-effects and low success rates of more traditional techniques. Moreover, new methods can offer economic and environmental solutions to the restrictions placed upon managers of natural resources.

Introduction

Optimally functioning habitats and ecosystems maintain the Earth’s biosphere and human welfare, but many face impacts from an increasing number of anthropogenic threats (Vitousek et al. 1997; Bellwood et al. 2004). Such impacts jeopardize the global value of environmental, economic, and social benefits provided by the world’s major biomes, which have collectively been estimated at ∼ US$46 trillion year−1, or nearly twice the global gross national product (Costanza et al. 1997). Our reliance on ecosystem goods (e.g. food) and services (e.g. nutrient cycling) has prompted many countries to invest substantially in habitat conservation and monitoring, natural resource management, and environmental planning (e.g. more than US$155 billion between 2004 and 2008 in the US alone: USGPO 2009). Whilst most agree on the need for healthy ecosystems, the merits and drawbacks of numerous approaches to best achieve and/or maintain them require careful consideration (Wilson et al. 2007).

Where habitats have been degraded or lost, management options have often adopted a compensatory approach by focussing on techniques of rehabilitation that ultimately aim to re-establish functional and self-sustaining habitat through human intervention (Marzluff & Ewing 2001; Elliott et al. 2007). Numerous approaches to rehabilitation have been studied across terrestrial and aquatic ecosystems, ranging from relatively simple transplants of individuals to understanding the importance of spatial arrangement and heterogeneity of habitats (Oertli et al. 2002). Notably, rehabilitation programmes show varied levels of success, including outright failure, which underscores the limitations to current techniques as well as the inherent challenges involved in re-creating complex ecosystems (Fonseca, Kenworthy & Thayer 1998). Nevertheless, repeated testing has taught us that limits to successful rehabilitation may be based upon a combination of biological realities (e.g. tolerance of species to transplantation), spatial considerations (e.g. variability of environmental stress amongst sites), and temporal restrictions (e.g. timing of reproductive activity) (Paling et al. 2003; Elliott et al. 2007). Such insights have led to a growing awareness that limitations may be alleviated by tailoring techniques to the biology of the system under study, particularly the life-history strategies and dispersal or colonization potentials of target species (e.g. Rehounkova & Prach 2010).

Coastal habitats provide enormous resources on a global scale (e.g. fisheries) but have proven sensitive to anthropogenic disturbances such as overfishing, pollutants that reduce water quality, and species invasions (Hughes 1994; Meinesz 1999; Connell 2007). In particular, seagrass meadows are frequently affected because their requirement for sandy and well-illuminated environments, such as estuaries and protected embayments, often places them near centres of human habitation (Larkum & West 1990; Ralph et al. 2006). Since 1879, coastal eutrophication, increased sedimentation and turbidity, disease, and meadow fragmentation through dredging and boating activities, have caused the loss of over 51 000 km2 of seagrass, ∼ 29% of the world’s total seagrass habitat (Waycott et al. 2009). Such losses threaten the significant ecological services that seagrasses provide (Duarte 2002; Orth et al. 2006), including the trapping and stabilization of sediments to reduce turbidity and coastal erosion (Orth 1977), high rates of primary productivity with associated nutrient cycling and carbon sequestration (Mateo et al. 2006), and the provision of food and habitat for numerous fish and invertebrates (Heck et al. 1995), including many of commercial value (Connolly 1994).

Rehabilitation is often considered a worthwhile option for degraded seagrass habitats because many seagrasses are notoriously slow-growing such that natural recovery, if it occurs (Kendrick et al. 2002), may take anywhere from tens to hundreds of years (e.g. Posidonia spp.: Kirkman & Kuo 1990; González-Correa et al. 2005). Transplantation is by far the most common method trialled around the world, followed by the planting of collected seeds and cultured seedlings (Fonseca, Kenworthy & Thayer 1998; Calumpong & Fonseca 2001). The overall success of such techniques is questionable (e.g. van Keulen, Paling & Walker 2003; Bell et al. 2008), with meta-analysis of seagrass rehabilitation projects in the USA identifying success rates of 35–50% (Fonseca, Kenworthy & Thayer 1998). Limitations to success are often underpinned by the severity of the physical environment, such as wave exposure and associated sediment mobility causing erosion of transplants (van Keulen et al. 2003), as well as problems of seed supply and seedling culture (Holbrook, Reed & Bull 2002). There may be additional ethical issues to consider, such as the very real potential for transplantation to damage and fragment donor meadows, exacerbating the original problem by causing a local net loss of seagrass (Bull, Reed & Holbrook 2004). In essence, there is a clear need for fundamentally different approaches to seagrass rehabilitation in order to overcome current limitations, and a more careful consideration of life-history traits of target species may provide novel and achievable solutions.

This paper presents an 8-year study that experimentally tests the performance of multiple seagrass rehabilitation methods that utilize fundamentally different ecological approaches. In the temperate waters adjacent to the coastal city of Adelaide in South Australia (population ∼ 1·2 million), at least ∼ 5200 ha of seagrass have been lost since the 1930s as a consequence of coastal eutrophication from wastewater inputs promoting the growth of epiphytes that smothered seagrasses and limited their capacity to photosynthesize (Neverauskas 1987; Shepherd et al. 1989). In recent years, the management of Adelaide’s wastewater has improved and a small amount of natural seagrass recovery has occurred in some areas (e.g. ∼ 4% near a major sewage outlet decommissioned in 1993: Bryars & Neverauskas 2004), suggesting successful rehabilitation may be possible. Therefore, we first tested the traditional and generic rehabilitation methods of (i) transplanting seagrass from healthy donor meadows into damaged sites, and (ii) seedling culture and outplanting, to asses their potential for large-scale rehabilitation.

In general, the southern Australian coastline represents some of the most wave-exposed conditions in which seagrasses are found anywhere in the world (Walker & McComb 1992), and sediment instability following seagrass loss appears to be a critical factor limiting natural recovery by preventing the establishment of new recruits (Clarke & Kirkman 1989; Shepherd et al. 1989), as well as the retention of transplants (Paling et al. 2003). Consequently, traditional techniques may not always be practical, and so we developed and tested a novel approach that alleviates limits imposed by sediment mobility and which also exploits a critical seagrass life-history trait to maximize recruitment potential. This technique essentially involves deploying sand-filled bags of hessian (made from jute fibres, also known as burlap) on the sea floor to provide a stable sediment environment and to also facilitate the recruitment of seagrass seedlings possessing a distinctive ‘grappling hook’ at their base (see Materials and methods for further details). This novel technique was compared to the performance of the more generic and traditional rehabilitation methods to assess the benefits of a tailored approach that exploits a useful life-history trait and specifically alleviates a major limitation to successful rehabilitation.

Materials and methods

Study sites

All research was done in the waters of Gulf St Vincent, South Australia, from 2003 to 2010. Gulf St Vincent is a large (∼ 6800 km2) and relatively shallow embayment (<50 m depth), with a strong spring-neap tidal cycle and moderate wave exposure (mean annual significant wave height ∼ 1 m) (Bye & Kämpf 2008). Much of the coastline is sandy and supports seagrass, mangrove, and salt marsh communities, with the city of Adelaide located centrally on the eastern shore. Up to 14 species of seagrass occur in the waters off Adelaide, with the most abundant meadows comprising Amphibolis antarctica (Labill.) Sonder ex Ascherson, Posidonia angustifolia Cambridge & Kuo and Posidonia sinuosa Cambridge & Kuo, as well as Heterozostera nigricaulis J. Kuo in more sheltered locales (Bryars, Wear & Collings 2008). The historical loss of seagrass in the region can be generally described as the fragmentation and seaward regression of meadows, including the expansion and coalescence of blowouts, with peak periods of loss coinciding with major coastal developments (e.g. wastewater treatment plants: Neverauskas 1987; Shepherd et al. 1989). All experiments described herein were done at multiple sites within this zone of seagrass loss to indicate the spatial generality of observed responses. Sites differed somewhat in environmental conditions (e.g. some exhibited greater wave exposure or were shallower than others; see Materials and methods), which was considered an important part of the study since an overarching aim was to develop a rehabilitation technique that is useful in a wide range of environmental conditions.

Seagrass transplants

The potential for rehabilitating seagrass using transplants of A. antarctica and H. nigricaulis was assessed using two techniques: (i) plugs, essentially a core of seagrass extracted from donor meadows with roots, rhizomes, and sediment intact, and (ii) sprigs, comprising a section of rhizome with roots and shoots but no sediment (see Fonseca, Kenworthy & Thayer 1998 for more details). Plugs were extracted using PVC piping of 10 cm diameter inserted 25 cm into the sediment (i.e. 25 cm depth of sediment was retained around roots and rhizomes of extracted seagrass), whilst sprigs were collected by gently excavating sediments at the edge of donor meadows until 30-cm long sections of rhizome were exposed and removed for transplantation.

Recipient sites were Henley Beach and West Beach, based partly on the historical existence of seagrass meadows at both sites, but also because they were thought to exhibit relatively little sediment movement making erosion of transplants less likely. Plugs and sprigs of A. antarctica were sourced from a donor meadow at Henley Beach at a similar depth as the recipient sites (6 m). Donor meadows of H. nigricaulis, however, were sought from shallower waters (3 m) 20 km northward at Section Bank because of its sparse distribution near recipient sites.

Plugs and sprigs were harvested and transplanted within 24 h in late February to early March 2003, with each recipient site receiving 20 plugs and 18 sprigs of each species. Prior to planting, sprigs were woven into a coarse-weave mat of hessian to anchor them to the substratum and help stabilize sediments. Hessian matting was also placed around the base of all plugs to stabilize surrounding sediments, with all transplants located within 1–2 m of remnant patches of seagrass. The survival of transplants was monitored for eight months, with the change in shoot density amongst surviving transplants calculated as an indicator of transplant expansion or contraction. Differences between techniques (plugs vs. sprigs) were tested using anova.

Seedling culture and outplanting

Rehabilitation by culturing and outplanting seagrass seedlings was assessed from 2003 to 2005. Fruits of P. angustifolia with no sign of discoloration or herbivory were collected from amongst those washed up on local beaches, as well as in situ from nearby meadows, in December 2003. All fruits were kept in sea water for ∼ one month until the majority had dehisced and seeds had sprouted leaves and roots.

In February 2004, seedlings were planted into biodegradable Jiffy pots (Jiffy International, Norway) filled with beach sand and placed into four fibreglass holding tanks connected to an open flow-through sea water system (∼ 11 L min−1). Eighty beach-collected and 80 meadow-collected seedlings were placed in each tank and were covered with 90% shadecloth to condition seedlings to average light intensities at the outplanting sites, which were measured at random intervals throughout 2003 using a Li-Cor quantum sensor. Seafloor light intensities were found to average 15–18% of surface irradiance (86·83 ± 22·71 μmol m−2 s−1), which was best replicated in the shallower culturing tanks by using shadecloth designed to reduce light intensity by 90%. Survival and growth (change in the maximum height) of cultured seedlings were sampled ∼ bi-monthly until December.

In February 2005, seedlings were outplanted at Grange and Semaphore. Seedlings were planted within hessian bags filled with sand to provide a stable sediment environment for establishment (see description under Recruitment facilitation below). On each bag, four clumps of ∼ five seedlings were planted (i.e. ∼ 20 seedlings per bag), with bags either placed on sand adjacent to natural seagrass or within meadows (n = 5 per site). Seedling survival and growth was sampled ∼ bi-monthly until September. Survival in culture and outplant experiments was analysed with repeated-measures anova (rm-anova), whilst growth was analysed with anova.

Recruitment facilitation

During transplant experiments, substantial numbers of A. antarctica seedlings recruited to the hessian matting. A. antarctica seedlings possess a distinctive ‘grappling hook’ structure at their base (Kuo & den Hartog 2006), which ordinarily anchors them to stems of adult A. antarctica, or leaf sheaths and exposed rhizome of Posidonia, after which they grow roots and establish themselves (Clarke & Kirkman 1989). The fibrous hessian provides an excellent surrogate, facilitating recruit densities up to ∼ 660 individuals m−2 (Wear, Tanner & Hoare, in press). Thus, a novel donor-independent rehabilitation method was developed whereby hessian bags were filled with sand and placed on the sea floor to facilitate in situ recruitment of A. antarctica seedlings.

The performance of hessian bags was tested in three experiments established at multiple sites in 2004, 2005 and 2007. All recruitment units consisted of a fine-weave hessian bag encased in a coarse outer weave (see Appendix S1), measuring 0·76 × 0·46 m, and filled with ∼ 25 kg of playpit sand. Bags were placed on sand either in blowouts or near the edge of receding meadows, spaced ∼ 0·5 m apart, and similarly oriented to waves and tide. Densities of A. antarctica recruits on each bag were sampled ∼ bi-monthly for 12 months, and about every six months thereafter. For the experiment established in 2007, the heights of three haphazardly selected recruits on each bag were concurrently sampled to estimate growth (i.e. change in height).

Bag performance was formally assessed as the density of A. antarctica recruits after 12 months, which gave adequate time for recruitment and substantial above- and below-ground growth. This approach achieved a standardized comparison amongst experiments deployed in different years, though we recognize that recruits are not fully established by 12 months and so sampling continued to examine longer-term dynamics. Here, we expected to initially observe high densities concurrent with the recruitment of juveniles, followed by a decline to lower but sustained densities as sources of juvenile mortality take their toll and remaining individuals grow into adults (i.e. as a new patch is created; see Discussion for further information). We were hopeful of observing subsequent increases in density as new patches independently grow and expand, yet such outcomes may take >10 years to manifest due to naturally slow expansion rates of most seagrasses (Kirkman & Kuo 1990; González-Correa et al. 2005). Analyses proceeded by first comparing amongst-site variation in recruit densities over time for each experiment (rm-anova), and then comparing across experiments by analyzing the density and growth of recruits at 12 months (anova).

Results

Seagrass transplants

Survival of transplanted seagrass after eight months varied from 0 to 80% depending on site, species, and technique. Plugs of both species survived relatively well at Henley Beach (75–80%), but survival at West Beach was lower, with fewer H. nigricaulis surviving relative to A. antarctica (15 vs. 55%; Fig. 1a). Survival of sprigs was also greater at Henley Beach (44–61%), with no sprigs surviving past June at West Beach due to erosion of the hessian matting (Fig. 1b). Across sites and species, 56% of transplanted plugs survived, compared to only 26% of sprigs.

Figure 1.

 Survival of transplanted (a) plugs and (b) sprigs of A. antarctica and H. nigricaulis at Henley Beach and West Beach during 2003.

Average shoot density of surviving transplants declined over time in all treatments. For A. antarctica, the decline was relatively small for both plugs (mean ± SE = 17·25 ± 7·15%) and sprigs (13·08 ± 13·52%), and did not differ between techniques (anova: F1,24 = 0·23, P = 0·637). Greater declines were seen in H. nigricaulis where plugs lost almost twice as many shoots as sprigs, on average (35·05 ± 5·58% vs. 21·17 ± 17·15%, respectively), though such differences were not statistically significant (anova: F1,22 = 3·55, P = 0·073).

Seedling culture and outplanting

Survival of cultured P. angustifolia seedlings declined over time, falling to 6–9% after 11 months, and did not differ between beach- and meadow-collected fruits (Fig. 2a; rm-anova: F1,6 = 1·57, P = 0·256). Seedlings that did survive to 11 months generally appeared healthy, with beach-collected fruits exhibiting, on average, nearly twice as much growth as their meadow-collected counterparts (6·03 ± 1·12 cm vs. 3·67 ± 0·61 cm, respectively). Such differences, however, were not statistically significant (anova: F1,49 = 3·94, P = 0·053).

Figure 2.

 Mean (± SE) survival of P. angustifolia seedlings (a) grown in culturing tanks during 2004, and (b) outplanted into the field during 2005. Cultured seedlings were grown from fruits collected at local beaches and from meadows in situ, whilst outplanted seedlings were placed on sand or within natural seagrass at Grange and Semaphore.

Outplanted seedlings experienced substantial mortality, particularly on hessian bags placed within natural seagrass where no seedlings survived more than three months (Fig. 2b). Up to ∼ 10% of seedlings placed on sand survived for eight months, but low overall survival meant no differences between sites or habitats were formally detected (rm-anova: F1,6 = 2·46, P = 0·100). Considerable growth was observed amongst the few survivors (4·28 ± 0·78 cm, all on sand at Grange), which all appeared healthy after eight months.

Recruitment facilitation

Recruitment of A. antarctica seedlings to hessian bags occurred in all experiments, though it was spatially variable. In 2004, recruit densities at Grange were greater than at Semaphore over much of the first year, peaking at 345 ± 35 individuals m−2, but became similar between sites by 12 months (Fig. 3a, Table 1). In 2005, maximum densities of 238–349 individuals m−2 occurred at all but the two southern-most sites (Seacliff and Brighton), with densities declining to near zero for all sites except Grange over the following 12 months (Fig. 3b, Table 1). For bags deployed in 2007, recruitment was generally poor at all sites until after May 2008, when bags at Brighton and Grange supported a moderate number of recruits after 12 months (∼ 45 ± 11 individuals m−2 and 134 ± 23 individuals m−2, respectively), whilst those at Semaphore and Largs Bay supported few recruits (Fig. 3c, Table 1).

Figure 3.

 Mean (± SE) density of A. antarctica recruits on hessian bags deployed in (a) 2004, (b) 2005, and (c) 2007. Broken lines indicate the samples taken after bags had been in situ for ∼ 12 months.

Table 1.   Results of repeated measures-anovas testing for spatio-temporal differences in the density of A. antarctica seedlings on hessian bags during the first 12 months from deployment in 2004, 2005, and 2007
YearSourcedfMSFP
  1. For each test, P-values have been conservatively adjusted by the Greenhouse-Geisser ε to compensate for inflated Type I error rates associated with any departures from sphericity (Myers & Well 2002).

2004Site1487630·4365·02<0·001
Error187499·50  
Time549383·518·06<0·001
Time × Site520234·043·300·033
Error906126·75  
2005Site497560·6939·96<0·001
Error452441·70  
Time2486111·78315·45<0·001
Time × Site866495·7343·15<0·001
Error901541·03  
2007Site311654·0731·93<0·001
Error29365·00  
Time511414·6148·33<0·001
Time × Site152885·3812·22<0·001
Error145236·21  

Comparing recruit densities at 12 months across all experiments emphasized the spatial and temporal variation. Amongst years, 2005 produced the least recruits, whilst spatial variation was most obvious in 2007 (Fig. 4, anova comparison amongst sites: F10,98 = 14·85, < 0·001). Interestingly, bags at Grange consistently retained the most recruits. For the experiment established in 2007, growth over the initial 12 months varied amongst sites from 0·50 ± 1·15 cm at Semaphore to 3·20 ± 1·07 cm at Brighton, though no differences amongst sites were detected (anova: F3,19 = 1·14, P = 0·360).

Figure 4.

 Comparison amongst sites and years of the mean (± SE) density of A. antarctica recruits on hessian bags after 12 months in situ. Letters above each column indicate the outcome of post-hoc pairwise comparisons.

Densities of A. antarctica were sampled for an additional 1–2 years beyond the first 12 months of each experiment, with a commonly observed pattern being a decline to lower but sustained densities over the life of the experiment (Fig. 3a–c). The magnitude of the decline was site dependent, with long-term densities reaching zero in some places but up to 72 individuals m−2 in others (best seen at Grange in Fig. 3a–c). Additionally, close observation of hessian bags at each sampling time showed that most had either become buried in the sandy sea floor and/or undergone severe or total degradation after ∼ 18–24 months in situ, such that their capacity to facilitate further recruitment was probably exhausted.

Discussion

The cumulative loss of almost one-third of the world’s seagrass meadows over the past 130 years (Waycott et al. 2009) underscores the need to conserve and rehabilitate such that the valuable resources and ecosystem services that seagrasses provide are maintained (Duarte 2002; Orth et al. 2006). Seagrass rehabilitation has a reasonably long history (since Addy 1947) that emphasizes donor-dependent techniques, particularly transplants (see table 1·6 in Fonseca, Kenworthy & Thayer 1998). Whilst often suitable for small areas (≤1 ha; Orth et al. 2006) in sheltered locales, transplant success is limited by numerous physical and biological features of the environment, as frequently demonstrated in the more exposed waters of southern Australia (Paling et al. 2003; van Keulen et al. 2003). Additionally, techniques such as transplants and seedling culture usually require a large commitment of time and resources that may render them impractical over large-scales. This paper shows how tailoring rehabilitation techniques to the biology of the target system or species may provide solutions that alleviate current critical limitations to rehabilitation success.

The performance of traditional and generic transplant methods of rehabilitation was mixed off the coast of Adelaide. Plugs generally out-performed sprigs in terms of survival (∼ 56 vs. 26%, respectively), consistent with studies in Western Australia using a related species (A. griffithii: van Keulen et al. 2003) and also with species in other regions (e.g. Phyllospadix torreyi in California: Bull et al. 2004). In Western Australia, strong water motion during storms excavated sprigs of A. griffithii before they could fully establish. In our study, sprigs failed completely at West Beach, which exhibited greater wave energy that excavated and dislodged the hessian matting used to anchor sprigs. Fewer plugs also survived at West Beach, but survival did not reach zero, suggesting plugs are a better choice than sprigs in places of stronger wave energy and sediment movement (van Keulen et al. 2003). Importantly, shoot density declined in almost all transplants, suggesting constraints on other aspects of transplant establishment and growth.

Tests of seedling culture and outplanting as a donor-independent method have been successful elsewhere (e.g. Italy: Balestri, Piazzi & Cinelli 1998), but proved challenging in our study. An earlier pilot study produced 95% survival of seedlings after six weeks (S. Seddon, unpublished data), in contrast to the poor survival (6–9%) after 11 months in the current experiment. A combination of excessive epiphyte growth on leaves and the 90% shadecloth used to condition seedlings to light intensities at outplanting sites may have reduced survival by limiting light availability. In a previous study, Kirkman (1978) controlled epiphytes on adult Zostera capricorni by blocking all light for three days, but since seedlings are unlikely to have sufficient carbon stores in their rhizome to survive such periods of darkness, it was questionable whether such complete shading would have also severely impacted seedlings in our study.

Outplanted P. angustifolia seedlings also performed poorly, with only 10 seedlings, ∼ 2% of those planted, surviving for eight months. Similar trials using P. australis in Western Australia, however, had greater success (30% survival after 11 months, J. Statton pers. comm.). Poor survival of outplanted seagrass seedlings is common (Holbrook et al. 2002; Bull et al. 2004, but see success in Balestri et al. 1998), yet juveniles of most species naturally experience high mortality (e.g. Holbrook et al. 2002). On balance, natural bottlenecks in juvenile survival may be difficult to overcome even if seedlings are grown to a seemingly more robust stage before exposure to field conditions.

Hessian bags represent a novel technique for alleviating some major limitations to successful seagrass rehabilitation, providing a stable sediment environment to overcome problems of erosion, as well as exploiting the grappling hook as a key life-history trait of juveniles to facilitate recruitment and establishment of habitat entirely in situ. In three experiments, A. antarctica seedlings consistently recruited to hessian bags, albeit with considerable spatial variability in density (Fig. 3). Initial recruitment typically averaged 150–350 seedlings m−2, and whilst natural recruitment levels within meadows are unknown, the production of recruits (i.e. number of seedling on adults) is much lower, averaging ∼ 37 seedlings m−2 (A.D. Irving, unpublished data). Longer-term abundance on bags (beyond 12 months) declined to lower levels but sustained densities of 0–72 individuals m−2, depending on location. As noted above, such declines following initial recruitment are representative of the natural situation as various causes of juvenile mortality take their toll. We also observed that this period of decline typically coincided with considerable growth of new shoots and leaves on the remaining individuals as they developed into adults. Additionally, the partial to complete degradation of hessian after ∼ 18–24 months in situ suggests that densities sampled around this time may be considered the direct end-product of hessian bags as a rehabilitation technique. Where seagrass remains, the question then becomes whether these new patches can expand and become self-sustaining. The answer is currently beyond the scope of the present study (though we continue to monitor our experiments), largely because of the generally slow expansion rate of seagrasses (i.e. in the order of decades: Kirkman & Kuo 1990; González-Correa et al. 2005) and thus the considerable amount of time that would be needed. Nevertheless, we recently observed (February 2010) a substantial increase in seagrass density between 21 and 30 months in situ for bags deployed at Grange in 2007 (Fig. 3c), suggesting the patches of seagrass remaining at this site after hessian bags have degraded are now expanding in their own right (i.e. the goal of creating new patches of expanding seagrass appears achievable).

Collectively, our results suggest that efforts to rehabilitate large areas of Adelaide’s seagrass could benefit by considering the inclusion of the hessian bag technique into project designs. The traditional techniques of transplanting and seedling culture are certainly possible, but appear unreliable for long-term success because of limitations imparted by wave energies and associated sediment mobility. Hessian bags alleviate this particular hindrance by providing a stable sediment environment, but also avoid potential stress and mortality due to handling plants by facilitating recruitment of seagrass in situ. In essence, hessian bags have three distinct environmental advantages over traditional methods tested here. First, they do not damage existing meadows like transplants, and also retain recruits likely to be lost from degraded systems. Secondly, they provide a stable sediment environment for seedlings to establish root systems. Thirdly, they are completely biodegradable, leaving no long-term negative impact. However, spatial variability in the performance of hessian bags was obvious, and probably relates to a combination of premature hessian degradation and stronger water movement at some sites that may dislodge seedlings and hessian (i.e. limitations to recruitment success are not completely overcome). In such places, transplanting plugs may be a worthwhile supplement, since up to 50% survival was observed in stronger wave environments (Fig. 1a: West Beach). However, to avoid the destructive side-effects of transplants, research aimed at improving the performance of hessian bags may be worthwhile. To this end, various options for improving hessian longevity (e.g. coating fibres with non-toxic biodegradable polymers) are currently being tested to give seagrass recruits more time to establish before the hessian degrades.

Economically, hessian bags appear to be a sound option for large-scale rehabilitation. Materials are inexpensive, and sites can be established by simply throwing bags off a boat. In contrast, transplants are slow and costly to establish, requiring expensive divers and specialist expertise to cover a much smaller area per unit time. Cost-benefit analyses for large-scale rehabilitation indicate that a sowing density of 1000 hessian bags ha−1 would cost ∼ US$16 737 (includes cost of materials, construction, and deployment; Wear et al. in press), whilst covering the same area using transplants would cost at least US$27 593 (Seddon et al. 2004). Investment in hessian bags thus compares more favourably with the global value of resources and ecosystem services provided by seagrasses, estimated at US$27 039 ha−1 year−1 (Costanza et al. 1997).

A potential limitation of hessian bags is their taxonomic breadth of application. Amphibolis is unique in dispersing viviparous seedlings that possess the grappling hook that anchors so well amongst hessian fibres. Most other seagrasses disperse via smooth and rounded fruits and/or seeds that are unlikely to become entangled so easily. There are some exceptions, such as fruits of the surfgrass Phyllospadix torreyi, which possess bristle-covered arms that could conceivably entangle with hessian (Bull et al. 2004 used this approach to successfully anchor fruits into braided nylon line). Besides entanglement, establishing Amphibolis may indirectly facilitate other species by creating more favourable conditions (e.g. stabilizing sediments; Clarke & Kirkman 1989). It may also be possible to exploit the stable sediment environment within bags whereby fruits and seeds are added when bags are filled with sand, essentially creating a concentrated ‘ready-mix’ seed pack somewhat similar to the successful technique of Harwell & Orth (1999).

In all our experiments, low long-term survival was common, especially as seedlings transitioned to adults. As noted above, significant juvenile mortality is representative of the natural world (Holbrook et al. 2002), and highlights a potentially unavoidable obstacle to successful rehabilitation. Calumpong & Fonseca (2001) observed that the success of terrestrial crops still cannot be guaranteed despite our collective millennia of experience in their cultivation, so perhaps we should not expect too much too soon given the relative infancy of ‘farming’ seagrasses. Even the best-understood seagrass systems on the planet may require transplanting of 300 units to ensure 100 survive (Fonseca, Kenworthy & Thayer 1998). This disproportionate expenditure of effort evokes the sentiment that preventing seagrass loss (conservation) is better than cure (rehabilitation); an ethos reinforced by the rapid pace at which large areas of seagrass are lost (weeks to years) relative to their protracted natural recovery (decades to centuries: Kirkman & Kuo 1990; González-Correa et al. 2005).

Anthropogenic pressures on habitats around the world have prompted many countries to implement better conservation and environmental management strategies to preserve their natural resources and associated ecosystem services (Vitousek et al. 1997; Bellwood et al. 2004). In systems that are beyond conservation, habitat rehabilitation can play a central role in reversing historical losses (Elliott et al. 2007). The great challenge of re-creating complex and productive ecosystems that are naturally dynamic over temporal scales of years to decades means that we must prepare for slow progress and as many failures as successes, if not more. However, perceived failures are often instructive to subsequent efforts by demonstrating the critical limits to successful rehabilitation, which may be alleviated or even overcome by tailoring techniques to the target system or species rather than adopting more generic approaches. Exploring such alternatives is likely to result in the development of novel techniques that expand our ecological toolbox for rehabilitation and ultimately bring us closer to the central goal of reversing historical impacts and being prepared for challenges in the future.

Acknowledgements

We thank those who collectively provided thousands of hours of field and laboratory assistance over the years, especially S. Drabsch, A. Eaton, Y. Eglinton, J. Nichols, L. Mantilla, B. Miller-Smith and K. Wiltshire. We also thank our colleagues for insightful discussion and encouragement, particularly D. Fotheringham, H. Kirkman, S. Murray-Jones, E. Paling, P. Ralph, S. Shepherd, J. Statton, M. van Keulen and D. Walker. Funding and continued support from the South Australian Department of Environment and Natural Resources, and the Adelaide and Mount Lofty Ranges Natural Resources Management Board, is gratefully acknowledged. This manuscript benefited from comments provided by the editors and two anonymous reviewers.

Ancillary