Long distance biotic dispersal of tropical seagrass seeds by marine mega-herbivores

Terrestrial plants use an array of animals as vectors for dispersal, however little is known of biotic dispersal of marine angiosperms such as seagrasses. Our study in the Great Barrier Reef confirms for the first time that dugongs (Dugong dugon) and green sea turtles (Chelonia mydas) assist seagrass dispersal. We demonstrate that these marine mega-herbivores consume and pass in faecal matter viable seeds for at least three seagrass species (Zostera muelleri, Halodule uninervis and Halophila decipiens). One to two seagrass seeds per g DW of faecal matter were found during the peak of the seagrass reproductive season (September to December), with viability on excretion of 9.13% ± 4.61% (SE). Using population estimates for these mega-herbivores, and data on digestion time (hrs), average daily movement (km h) and numbers of viable seagrass seeds excreted (per g DW), we calculated potential seagrass seed dispersal distances. Dugongs and green sea turtle populations within this region can disperse >500,000 viable seagrass seeds daily, with a maximum dispersal distance of approximately 650 km. Biotic dispersal of tropical seagrass seeds by dugongs and green sea turtles provides a large-scale mechanism that enhances connectivity among seagrass meadows, and aids in resilience and recovery of these coastal habitats.

annually 16 . Many factors that influence seagrass meadow maintenance and recovery have been examined in both temperate and tropical habitats, but little is known concerning the modes of dispersal for these populations 1 .
The resilience of seagrass meadows is dependent on interactions between physical (e.g., location, climate, water currents and tidal flow) and biological factors (e.g. species and genetic diversity, life history strategy, population connectivity) [27][28][29] . Following large-scale losses of seagrass, seed germination from the sediment seed bank and subsequent seedling growth, is thought to be one of the main pathways for natural revegetation of disturbed habitats 10,13 . Seed banks are replenished via seed production within the meadow and from the input of seeds, or propagules, from other more distant meadows. Deposition of seeds from outside sources results in more abundant seed banks and an increase in genetic diversity, culminating in an increased resilience to disturbance 28,29 . However, most seagrass seeds are negatively buoyant. This inhibits dispersal by abiotic means unless seeds are attached to floating plant fragments 2,14 . Significant knowledge gaps remain on mechanisms of seed dispersal in seagrass, levels of connectivity between meadows from seed dispersal 1,2 , and particularly on the importance of biological dispersal vectors for maintaining the resilience and recovery capacity of seagrass meadows.
Research has primarily focused on abiotic dispersal of seagrass seeds and propagules 1,29 . However, herbivores and omnivores in the marine environment, including crustaceans, echinoderms, fish, reptiles, birds, sea turtles and Sirenians (dugongs and manatees), consume seagrass either directly or indirectly while feeding 6,[10][11][12]30 . As a result, fruiting bodies and seeds can be consumed, providing a potential for biotic dispersal 1 . Recent laboratory studies have shown that seeds of Zostera marina and Halophila ovalis can survive consumption by herbivorous fish and maintain or enhance their ability to germinate 31 . However, the ability of seeds to survive digestion by herbivores is species specific, as Thalassia hemprichii seeds did not survive consumption by waterfowl and fish under controlled laboratory conditions 9 . While these results indicate biotic dispersal of some seagrass by specific vectors is possible, there still remains a dearth of information on the role of animals, particularly mega-herbivores, in seagrass dispersal.
In tropical habitats marine mega-herbivores, dugongs (Dugong dugon) and green sea turtles (Chelonia mydas), consume seagrasses in shallow coastal waters [32][33][34] . Dugongs conusme almost exclusively seagrass and include in their diet 24 out of the 26 seagrass species that co-occur in Australian waters 33 . Green sea turtles consume either a seagrass or algae dominated diet depending on their home range, however they will consume both food types when present 35 . Dugongs can consume up to 40 kg wet weight of seagrass daily 33,36 and green sea turtles consume up to 2.5% of their body weight in algae and/or seagrass daily 37 . Both marine mega-herbivores have digestion times ranging from several days to weeks; 6-8 days for dugongs and 7-14 days for green sea turtles (time variations are dependent on the digesta/food consumed, with a herbivorous diet taking longer to digest) [37][38][39] . Dugongs and green sea turtles actively move across their home ranges (green sea turtle median home range of 75.7 km 2 ; dugong median home range of 453.2 km 2 ) 40 and can travel large distances when actively migrating (tagged dugongs have been observed to move at 25.9 ± 2.23 km per day when undertaking macro-scale movements, while green sea turtles can travel up to 40 km per day) [41][42][43] . The traits of high consumption rates, slow digestion and long distance movement combine to make dugongs and green sea turtles suitable vectors for long distance dispersal of seagrasses (Fig. 1). For mega-herbivores in marine systems to disperse seagrasses, seeds must remain viable after passage through the gut; a factor not explored in previous studies. Our objectives in the present study were to (1) determine whether seagrass seeds pass through the digestive systems of marine mega-herbivores and remain viable and (2) determine the species, number, and potential dispersal distances for those seagrass species consumed by dugongs and green sea turtles in the Great Barrier Reef (GBR) Region. Our results quantify the potential for marine mega-herbivores to disperse seagrass seeds at different spatial scales and provide direct evidence of biotic dispersal by marine mega-herbivores.

Results
Faecal Samples. Marine mega-herbivore faecal samples collected floating on the water's surface varied in size and mass; ranging from 0.002 g DW to 7.696 g DW, with an average of 1.37 ± 0.18 (SE) g DW ( Supplementary  Fig. S1).
Except for H. decipiens, all species of seagrass seeds found in faecal samples included at least one seed with a split coat (Z. muelleri split seeds = 39.73%; Halodule uninervis split seeds = 100%; Halophila spp. split seeds = 50%). Time and/or site of collection had no significant effect on the proportion of seeds which had their seed coat split (site p = 0.661; time p = 0.291) (Figs 2b and 3).
Viability. Germination of seeds collected from mega-herbivore faeces occurred prior to viability testing for Z. muelleri, Halophila decipiens and Halophila spp., but not for H. uninervis. As germination is an indication of  viability, these seeds were considered viable. Of the remaining non-germinated seeds, only Z. muelleri seeds tested as viable. The mean percentage of viable Z. muelleri seeds collected was 9.13% ± 4.61% (SE). Although the percentage of viable seeds was lower in samples collected in the earlier months (September and October) and peaked towards the end of the year (November and December), time of collection and/or site did not have a significant effect on the proportion of viable seeds (site p = 0.630; time p = 0.597) (Fig. 2c).

Discussion
This is the first study to confirm the role of marine mega-herbivores, dugongs (Dugong dugon) and green sea turtles (Chelonia mydas), in the biotic dispersal of seagrass seeds. As the origin of the faecal samples (dugong or turtle) could not be accurately determined in the field, samples were combined into a single mega-herbivore category. We found 1-2 seeds per g DW in marine mega-herbivore faecal matter, with 9.13% ± 4.61% (SE) of these seeds maintaining viability after digestion. We found over half of the seeds excreted had a split seed coat, potentially increasing the chance of germination 44 . Time and site had only a minor effect on the number of seeds found per gram of marine mega-herbivore faeces. This is most likely due to different seagrass locations peaking in reproduction in the flowering season for these species.
For most seagrass species, long distance movement of reproductive structures (flowers, fruits and seeds) is possible only when these structures are attached to floating fragments 2,29,45,46 . The potential movement of viable seeds by biotic dispersal, confirmed in our study, is an alternative and likely important pathway by which seagrass can colonise new or recovering locations. By identifying a potential long-distance dispersal vector for seagrasses, these results have direct implications for an increased understanding of the connectivity between seagrass meadows and existing natural resilience and recovery processes in seagrass meadows.
The dispersal capability of mega-herbivores depends on the availability of plant reproductive material, the number of mega-herbivores in the region, the amount of time the seagrass material stays within the digestive systems of herbivores and the potential distance travelled between consumption and excretion of the plant material 1, 2 . In the GBR the availability of viable seagrass reproductive material varies temporally, spatially and between species 47,48 . In this study, maximum seed abundances in mega-herbivore faecal samples occurred in September and October while the greatest proportion of viable seeds were found in November and December. This reflects the dominance of Z. muelleri seeds in the samples and follows the observed periods of maximum flowering/reproductive density for Z. muelleri in this region 49 . In addition to its greater availably to mega-herbivores compared to other seagrass species across the sites sampled (Supplementary Table S2), the abundance of Z. muelleri seeds in the collected mega-herbivore faeces can be attributed to its flowering strategy. Zostera muelleri seeds are produced on branching flowering shoots that occur at the top of the meadow canopy 49 . The location and concentration of seeds above the sediment surface may make it easier for herbivores, green sea turtles in particular, to consume the seeds compared to species which flower at the base of the plant (e.g. H. uninervis) 50 . Therefore, the impacts of the timing and flowering shoot morphology on the biotic dispersal capability of seagrasses by mega-herbivores is likely to be species specific.
The GBR region supports a dugong population of at least 4,000-6,000 individuals 51, 52 and an estimated population of 855,000 (95% CI: 55,000-1,200,000) green sea turtles (green sea turtle estimates based on best available data collected in the southern GBR) 53 . If these marine mega-herbivores conservatively pass one average size faeces per day, they have the potential to disperse as many as 500,000 viable seeds daily during the peak seagrass reproductive season (September to November). Based on current mega-herbivore population estimates, that would be an average of up to 2,500 viable seeds per day for dugongs and up to 500,000 viable seeds per day for green sea turtles. However, caution should be given when using these estimated seed dispersal numbers due to limitations in mega-herbivore abundance and migration data. Refined estimates of the number of individuals in the resident and migratory dugong and green sea turtle populations in the GBR, and the variability in the proportion of the turtle population that migrate into and out of the region, will be necessary to improve the accuracy of these dispersal estimates. However, the numbers presented here are likely to be conservative.
Due to the importance of connectivity in maintaining resilient seagrass meadows, it is important to quantify how many and how far seeds are dispersed 2, 54 . For biotic dispersal of seeds by mega-herbivores, dispersal distance is dependent on how far individual herbivores move while the seagrass material moves through their digestive system. Green sea turtles have a digesta retention time of 156-325 hours 38 , with an average travelling speed of 1.89 ± 0.12 km h 41, 42 , creating a potential dispersal distance of 277-652 km. However, most green sea turtles have a median home range of 75.7 km 2 which may result in local rather than long-distance dispersal when turtles are not migrating 40 . Dugong gut passage times range from 146-166 hours 39 . They can travel on average at a speed of 1.3 ± 0.11 km h when swimming long distances 43 , potentially leading to a dispersal distance of 173-234 km; a distance less than the median home range of dugongs (453.2 km 2 ) 40 . However, pedigree analysis in south-east Queensland found that only 1-3% of dugongs undertake largescale movements 55 , while recapture and telemetric data found long distance movement across foraging grounds is possible, with travel distances ranging up to 560 km 43,55 . This suggests that dugongs have the potential to be long distance dispersers, however are more likely to disperse locally, similar to green sea turtles.
The actual distance travelled by seeds collected in this study cannot be measured. However, we found seeds in marine mega-herbivore faeces at one of our sites (Pioneer Bay) 100 s of kilometres from where seagrass flowering is common 48 , supporting the conclusion that these herbivores are an effective disperser of seagrass seeds. Distances shown in this study for seed dispersal, and previous studies on abiotic dispersal, indicate that there is a strong potential for the dispersal of seagrass among isolated meadows on reefs and islands and for the dispersal among countries by these mechanisms 29,33,56 .
Increases in the severity and occurrence of storms that have the ability to decimate seagrass meadows 20,27,57,58 have led to questions about the ability of seagrasses to re-establish or re-colonise after large losses 27,29,59 . Our present understanding of seagrass recovery mechanisms highlight the importance of the scale of the disturbance. Seagrass biomass loss on the scale of meters to tens of meters can recover primarily through rhizome extension 1,2,29,45 . Seagrass expansion by clonal growth can occur quickly, allowing for rapid recovery from small scale impacts 2, 45 . However, clonal growth alone would be slow to re-colonise a large area if most seagrass biomass was lost 2 . To re-colonise a large area devoid of seagrass, input from either viable seagrass fragments or seeds would be required 2,45 . Some coastal seagrass meadows have low connectivity to other meadows based on abiotic factors alone (such as wind and currents) 29 . Marine mega-herbivores may be vital for connectivity and gene flow among these meadows, as these animals are more likely to deposit seeds in habitats viable for seagrass growth compared to abiotic processes. The process of digestion aids in scarification, which through splitting the seed coat, removes physical dormancy and provides a cue for germination for some plant species 12 . Ultimately this may enhance germination rates and contribute to the recovery of impacted meadows. Deposition of genetically distinct seeds via mega-herbivore dispersal would increase the genetic diversity of meadows, thereby increasing their resiliency to disturbance events 60 .
Our study has confirmed biotic dispersal of viable seagrass seeds by marine mega-herbivores. As a result, tropical seagrass seeds have the potential to be dispersed far greater distances (in the hundreds of kilometres) than most previous reports suggest. The potential importance of mega herbivores in biotic dispersal is further enhanced by the large geographic distribution of these animals (green sea turtles from the tropical and sub-tropical ocean basins 56 and dugongs from the east coast of Africa to the Indo-Pacific between latitudes 27 degrees north and south) 33 , which includes a large proportion of the world's tropical and sub-tropical seagrass meadows. Biotic dispersal is also more likely to carry seeds to areas that are viable habitats for seagrass to grow, due to mega-herbivores actively searching for seagrass as a primary food source. This dispersal mode has the ability to contribute to the resilience of seagrass meadows and aid in recovery after loss. Our findings suggest that the conservation of green sea turtles and dugongs is likely to be far more important for maintaining the delicate balance between seagrass meadow recovery and loss than previously realised. A detailed understanding of the interconnection between marine mega-herbivores and their seagrass food is necessary to ensure a sustainable future for both these animals and their seagrass habitats.

Methods
Dugong and green sea turtle faecal samples were collected monthly between September and December 2014 from three coastal seagrass meadows in central GBR, Queensland; Pioneer Bay near the Whitsunday Island Group, Upstart Bay near the city of Bowen and Cleveland Bay near the city of Townsville (Fig. 4). Seagrass meadows at all sites consisted of mixed species assemblages and extended from the intertidal zone to the subtidal zone (Supplementary Table S2). Samples were collected during the period of seagrass maximum sexual reproduction, when 9 out of the 14 species present in the region flower 47 . During sample collections, green sea turtles were observed foraging at all sites, while dugongs were seen foraging at Pioneer Bay and Cleveland Bay. As it is not possible under field conditions to confidently determine from which of the two species the faeces had originated, and genetic differentiation was at the time outside the scope of this research, faecal samples from dugongs and green sea turtles were combined under a single 'marine mega-herbivore' category.
A total of 60 faecal samples were collected over four months; 4 samples in September, 15 in October, 19 in November and 22 in December. Faecal samples were transported on ice to the laboratory where they were stored at 3-5 °C until processing. All samples were processed within 30 days of collection. Samples were sieved across three size fractions to aid in seed species identification (1.4mm, 750 µm and 250 µm) and then placed in 34 ± 1 PSU seawater to prevent osmotic stress 44 . Seagrass seeds found in faecal samples were identified to species and counted prior to the removal of the seed coat. Once the seed coat was removed, all non-germinated seeds were stained with a 0.5% Tetrazolium solution for a total of 48 hours to determine viability 44 . A positive stain (tissue turns brown to red in colour; Fig. 5a) indicated normal cellular metabolism within cells, signifying that the seed was still active and capable of germination 44,61 . Any seeds removed from the faeces that had already germinated were counted as viable (Fig. 5b). Seeds with a split seed coat were also counted per sample. Faecal matter was dried at 220 °C in an oven until a consistent dry weight was reached. Samples were then weighed and reported as g DW.
Due to the low number of seeds collected per sample, seed data was combined across seagrass species and reported as the total number of seagrass seeds per g DW of faecal matter, the proportion of seeds with a split seed coat per g DW of faecal matter and the proportion of viable seagrass seeds per g DW of faecal matter (Fig. 2). Statistical analyses were conducted in the statistical software environment R (R version 3.2.5) 62 . Differences in the abundance of seagrass seeds per g DW of facecal matter over time (month) and site were analysed using generalized linear mixed effects models with a quasi-Poisson distribution (GLM) 62 ; a Tukey post hoc test was applied. Quasi-Poisson regression is a generalized form of Poisson regression which corrects for overdispersion in count data 63 . Changes over time (month) and site in the proportion of seeds with a split seed coat and the proportions of seeds that stained as viable (positive stain + germinated seeds) per g DW of faecal matter were analyzed using logistic regression mixed effects models with a binomial distribution (GLM) 62 ; a Tukey post hoc test was applied. Logistic regression was selected due to the binary response variable and the small number of seeds found per sample 64 .  (a) A Zostera muelleri seed (with seed coat removed) taken from marine mega-herbivore faeces returned a positive stain, turning the cotyledon pink; (b) A germinated Zostera muelleri seed removed from marine mega-herbivore faeces.