Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Cassowary dispersal of the invasive pond apple in a tropical rainforest: the contribution of subordinate dispersal modes in invasion

Cassowary dispersal of the invasive pond apple in a tropical rainforest: the contribution of... INTRODUCTION Seed dispersal is influential as a determinant of plant population and community structure and dynamics ( Nathan & Muller‐Landau, 2000 ; Levey ., 2002 ; Schupp ., 2002 ). As many as 75% to 90% of tropical rainforest tree species ( Webb & Tracey, 1981 ; Howe & Smallwood, 1982 ), accounting for possibly 95% of all seeds ( Terborgh ., 2002 ), are dispersed by a suite of animals that may represent the majority of vertebrate biomass or species at a location ( Terborgh, 1986 ; Gautier‐Hion & Michaloud, 1989 ). These plant–disperser interactions are generally loose mutualisms ( Wheelwright & Orians, 1982 ; Herrera, 1985 ), with plants receiving dispersal from multiple disperser species which in turn utilize multiple plant species. Seed dispersal is also an important component of plant invasion processes ( Rejmánek, 1996 ; Richardson ., 2000 ) with the potential to influence an invasive species’ introduction, establishment, and subsequent spread ( Kot ., 1996 ; With, 2002 ). The diffuse nature of dispersal processes in rainforests should make them relatively prone to infiltration by invasive species. Because seed dispersal processes determine the proportions of seeds moved any given distance from the source plant, understanding the shape and extent of dispersal kernels is necessary for predicting potential rates and pattern of invasive spread ( Kot ., 1996 ). This information can be fundamental in the design and implementation of management ( Trakhtenbrot ., 2005 ). However, vertebrate‐produced dispersal kernels, i.e. the frequency distribution of dispersal distances away from parent plants ( Janzen ., 1976 ), are poorly described even for native species in tropical rainforests ( Levey ., 2002 ; Nathan ., 2002 ) and almost invariably consider only the dominant dispersal mode. For rainforest invaders, descriptions of vertebrate‐produced dispersal kernels are unavailable ( Buckley ., 2006 ). Almost invariably, descriptions of dispersal kernels have focused on the kernel produced by just a single dispersal vector (but see Redbo‐Torstensson & Telenius, 1995 ; Vander Wall & Longland, 2004 ; Jordano ., 2007 ; Dennis & Westcott, in press ), an approach that provides just a fragment of the total dispersal picture. In contrast, all species are dispersed by multiple biotic and abiotic vectors and it is the combined contributions of these multiple vectors that determine the final dispersal kernel ( Nathan, in press ; Dennis & Westcott, in press ). While dominant dispersal modes may shape overall population dynamics, subordinate dispersal modes may contribute key dispersal components that have an overwhelming effect on overall dispersal outcomes ( Higgins ., 2003 ). In this paper we describe the mutualism that exists between an invasive plant and its primary vertebrate disperser in Australia's tropical rainforests. In doing so we provide what is, to our knowledge, the first description of the dispersal kernels generated by a native disperser in a tropical rainforest for an invasive species. This dispersal is provided by an endangered frugivore, the southern cassowary ( Casuarius casuarius , L.) to pond apple ( Annona glabra , L.), a primarily water‐dispersed species that invades intact and mildly disturbed rainforests. Because handling effects and dispersal distance are important components of a disperser's dispersal service, we describe both the effects of gut passage and deposition on germination and the dispersal kernels generated by cassowaries for this invasive species. We discuss these results in the context of whether cassowary dispersal, as an example of a subordinate dispersal mode in terms of the quantity of seed dispersed, can significantly alter pond apple's pattern of invasion and dispersal outcomes over and above those provided by its primary dispersal mode, water. METHODS Study species and study sites Pond apple originates from the Neotropics and was introduced to Australia in 1912 as a wetland rootstock for commercial cultivars of the custard apple, Annona squamosa L. Currently limited in distribution to Queensland, major infestations are found in the wet, tropical north‐east ( Agriculture and Resource Management Council of Australia and New Zealand, 2000 ; Sugars ., 2006 ). Habitats susceptible to pond apple invasion include rainforest, wet land, and riparian plant communities. In these habitats, pond apple can displace native vegetation and form monospecific stands. In Australia, fruiting generally occurs in the summer, peaking for 6–8 weeks between December and March and coinciding with monsoonal rains. Fruits are 7–8 cm in diameter and contain between 100 and 200 seeds, each 12–15 × 8–10 mm. Though a large, fleshy fruit apparently adapted for vertebrate dispersal, in Australia water is considered to be the dominant dispersal agent by far since the buoyant seed and fruit are encountered in great numbers deposited at high water marks and the regional scale pattern of spread is strongly associated with water ( Swarbrick, 1993 ; Sugars ., 2006 ). The large fleshy fruit is also attractive to a variety of vertebrates including cassowaries, pigs ( Sus scofa ), flying foxes ( Pteropus spp.), and a variety of birds and small mammals. While there are many potential dispersers, the size of fruits and seeds means that only cassowaries, pigs, and flying foxes are likely to provide quantities of dispersal with definite records only for the former two. Observations of ripe fruits suggest that cassowaries provide as much as 84% of vertebrate dispersal, pigs provide c . 2%, and predatory rodents move 14%, almost all less than 1 m ( O'Malley, 2004 ). Dung surveys suggest that pond apple is included in pig diets when available but at much lower levels than is the case for cassowaries ( Setter ., 2002 ). The southern cassowary is a large, flightless bird, resident in Australia in the rainforests of the north‐eastern humid tropics ( Davies, 2002 ). The cassowary is listed as endangered in Australia ( Environment Australia, 1999 ). Cassowaries are generally solitary birds occurring in densities of up to 0.08/Ha in lowland forests ( Bentrupperbäumer, 1998 ). The diet consists overwhelmingly of fruits which are generally swallowed whole and the seeds passed intact and viable ( Stocker & Irvine, 1983 ; Bentrupperbäumer, 1998 ; Westcott ., 2005a ). Field observations and collections were performed in February and March 2000 at three sites within 10 km of Innisfail, North Queensland (17°32′ S, 146°02′ E). Dung searches were conducted at each site and the number of pond apple seeds in each dung was counted. Gut passage trials Ripe pond apple fruit was collected in the field and fed to two captive cassowaries at the Johnstone River Crocodile Farm, Innisfail. The trial was limited to these two birds as other captive birds in the region were all housed in pond apple free areas and we sought to eliminate any risk of accidental introductions. Trials were performed in February and March 2001. Cassowaries were hand‐fed pond apple fruit, in feeding episodes that lasted until they lost interest in eating or up to a maximum of 15 min. The birds were fed their usual diet before and after being fed pond apple. The time elapsed from feeding to each defecation was recorded, and the dungs were immediately collected and seeds were counted. The birds were monitored until all consumed seeds were voided. Dispersal kernel estimation Dispersal kernels produced by cassowaries for pond apple were estimated using a combination of movement from Westcott . (2005a ) and gut passage rate data from this study. Though the movement data were not collected from the same locations in which the pond apple infestations occurred, it was collected from a variety of similar habitats and landscape contexts, and we are satisfied that an adequate representation of the dispersal of pond apple is likely to receive. To estimate a seed's dispersal distance, we chose a point in time during a telemetry session and determined its location. This was considered to be equivalent to the time and location of a feeding session. Dispersal distance for a seed consumed in that feeding session was estimated by taking a seed's gut passage rate from the feeding trials and determining the cassowary's location at that time interval. The straight‐line distance between the feeding location and the new position is the animal's displacement and the dispersal distance. Cassowary foraging is not distributed randomly throughout the day and so we distributed starting points, the equivalent of feeding sessions, over the day according to the daily distribution of foraging ( Westcott ., 2005 , Fig. 2 ). We assume that in each feeding session, multiple seeds are consumed and so we randomly selected 200 gut passage rates from the sample of rates obtained in the gut passage trials and determined the dispersal distance for each. The number of starting points used in a telemetry session was determined by the proportion of a day's foraging that would have been included in the timeframe of that session. For a full day this was 20 start times. This process is repeated across all movement pathways, and the dispersal kernel was derived as the frequency distribution of all dispersal distances. This process was repeated 10 times, and the mean percentage and 95% CI of seeds dispersed in each 25 m distance category were determined. 2 Mean cassowary‐produced dispersal kernel for pond apple. Bars indicate the dispersal kernel (left y ‐axis); the dashes represent the 95%CI around this kernel based on 10 kernels (left y ‐axis); the thick solid line represents the cumulative frequency of pond apple seeds dispersed as a function of distance (right y ‐axis). Germination trials Germination trials were conducted to document the effects of both gut passage and deposition pattern on seeds. The seeds used were from dung collected during the gut passage trials and from fresh dung and fruit collected in the wild. To examine gut passage effects we used two single seed treatments: (1) Single seeds with dung material attached (hereafter dung fruit) and (2) single seeds from fresh fruit with arils still attached (fallen fruit). To examine the effect of deposition pattern we used two additional treatments. Because pond apple in cassowary dung is deposited in a clump with other pond apple seeds and because unconsumed fruits fall to the ground whole, we also used two ‘clumped’ treatments: (3) Entire ripe fruit (whole fruit), because unconsumed pond apples fall to the ground whole, and (4) 20 seeds clumped with dung material attached (clumped dung fruit), because cassowaries deposit seeds in clumps. Treatments were placed on the surface of a 1 : 10 mixture of potting mix and a medium–heavy organic clay. Single seed treatments were placed in pots of 73 × 97 mm, and clumped seed treatments and entire fruit in pots of 125 × 140 mm. Pots were arranged randomly and were watered to saturation with tap water each day. Pots were checked twice weekly for germination, defined here as the emergence of the radicle at least 5 mm from the seed. The experiment was ended when all seeds were either germinated or deemed non‐viable upon inspection. We used Fisher exact tests to compare proportions of seeds germinating under the different treatments. The simple distributions of time to germination were analysed using survival analysis to enable seeds that were un‐germinated at the end of the trials to be taken into account. For comparisons of single seed treatments, log‐rank tests were used. Because multiple seeds were monitored in each pot for comparisons involving whole fruit and clumped treatments, these comparisons were conducted using Cox's F ‐tests with the degrees of freedom of the significance test set to the number of pots rather than to the number of seeds to avoid pseudo‐replication. To minimize the risk of type 1 errors when more than one test was performed, false discovery rate control was applied following Verhoeven . (2005 ). In all cases, tests remained significant after control and are reported at their original significance values. RESULTS Field observations and collections Cassowary dungs containing pond apple seeds were found at all three sites, and at one, Coquette Point, on each monitoring day. The presence of adult plants in the vicinity indicates that these dungs were deposited in locations suitable for germination. All 13 dungs collected contained pond apple seed, and 11 contained only pond apple. The average dung contained 199, ±66, SD seeds, ranging from three to 842 ( n = 13). Over 2500 seeds in total were collected with just one of these showing signs of damage. Casual observations elsewhere indicate that germination in cassowary dung is common in the wild (pers. obs.). Two cassowaries (one adult and one sub‐adult) were observed at Coquette Point. On one occasion, the adult bird was observed swallowing a whole pond apple fruit. Though dung indicated the presence of cassowaries at the other sites, they were not observed. Gut passage trial In total, 13 pond apples were fed to the two captive cassowaries with a total of 1512 seeds recovered from their dung. The frequency distribution for pond apple seed gut passage rates through cassowaries was strongly leptokurtic ( Fig. 1 ). Median gut passage time was 180 min (range 65–1675 min, lower and upper quartiles 150–250 min, respectively). Approximately 65% of seeds were passed between 90 and 240 min post‐consumption. A further 30% were passed between 240 and 600 min post‐consumption. 1 Gut passage rates of pond apple seeds through cassowaries. The bars indicate the frequency distribution of gut passage (left y ‐axis). The dashed line indicates the cumulative proportion of seeds recovered (right y ‐axis). Dispersal kernel estimation Dispersal was estimated to occur between 12 and 5212 m with a median dispersal distance of 387 m (quartile range 112–787 m; Fig. 2 ) and a mean distance of 605 m. Twenty‐four per cent of seeds were dispersed less than 100 m, and the modal dispersal distance was 38 m. Seventeen per cent of seeds were dispersed over 1000 m, with 5% and 1% of seeds being transported more than 2062 and 4037 m, respectively. Germination trials In total, 359 seeds were used in the single seed dung treatment and 107 seeds were used in the single seed fallen fruit treatment. A total of 1569 seeds in 10 pots and 1640 seeds in 12 pots were used in the clumped and whole fruit treatments, respectively. With the exception of whole fruits, all seed treatments showed two bouts of germination ( Fig. 3 ). The first began 40–60 days post‐sowing and continued for c . 40 days. Germination ceased with the beginning of cool weather in June. Dormancy continued until the weather warmed again in October and germination continued while warm weather prevailed ( Fig. 3 ). In contrast, there was little further germination in the whole fruit treatment after the first germination bout, most remaining seeds having rotted by day 200. 3 Cumulative germination curves for the four seed treatments: single seed with aril attached (fallen fruit, n = 107 seeds), single seed passed through a cassowary with dung attached (dung fruit, n = 359 seeds), seeds passed through a cassowary and placed in clumps with dung attached (clumped dung fruit, n = 1569 seeds), and whole fruit ( n = 1640 seeds). The proportion of seeds germinating across all treatments varied from 88% to 90% for the single seed treatments and from 79% to 82% for the clumped treatments. Though gut passage increased the proportion of seeds germinating in the single seed trials at the end of the first season ( Fig. 3 ), by the end of the second season there was no difference between the two treatments (dung fruit vs. fallen fruit, Fisher exact P = 0.70). The dung fruit treatment resulted in an increased proportion of germination when compared with the clumped dung fruit treatment (Fisher exact P = 0.01). In contrast, there was no difference between the proportion of seeds germinating in the clumped dung fruit and the whole fruit treatments (Fisher exact P = 0.57) though a greater proportion of whole fruits had germinated by the end of the first season ( Fig. 3 ). In the single seed germination trials, the time to germination of the dung fruit treatment was significantly shorter than that of the fallen fruit treatment (log‐rank test = –2.17, P = 0.03) due to a higher germination during the first season in the former ( Fig. 3 ). In the clumped seed trials, nearly all germination occurred during the first season in the whole fruit treatment resulting in a lower time to germination (median = 84 days) than observed in the clumped dung fruit treatment (median = 288) (Cox's F ‐test, F 10,12 = 4.98, P 0.05). In contrast, the dung fruit treatment showed shorter time to germination (median = 242 days) than the clumped dung fruit treatment (Cox's F ‐test 12,359 = 2.05, P < 0.05). DISCUSSION Cassowaries as dispersers of pond apple Our field observations confirmed that cassowaries are high‐quality consumers and dispersers of pond apple and that they are capable of dispersing large quantities of pond apple seed and commonly do so in infested areas. Cassowaries prefer fruit species that produce predictable, spatially concentrated, and high biomass crops, i.e. easily harvested ( Wright, 2005 ). Pond apple fits this description as individual fruits are large (mean mass = 226 g ± 49.3 SD), individual trees produce large crops and frequently occur in dense stands with relatively synchronous ripening. Consistent with this, the majority of dungs encountered contained only pond apple despite the fact that the species fruits during the period of high fruit availability and diversity in the region ( Crome, 1975 ; Westcott ., 2005b ). Gut passage through animals can be variable in its effect on seed germination. Passage through frugivore guts modifies germination probability or rate in about 50% of cases with germination enhancement, i.e. more rapid germination or increased germination probability, occurring c . twice as frequently as inhibition ( Travaset, 1998 ). In our experiments, a high proportion of ingested seeds were viable after gut passage, irrespective of deposition pattern ( Fig. 3 ). There was no difference in the final proportion of seeds germinating from the different single seed treatments, however, seeds that had both passed through a cassowary and were planted with faecal material still adhering showed higher germination frequencies at the end of the first season and shorter times to germination overall. Deposition pattern influenced germination over and above gut passage. Clumping of seeds from dung, the deposition pattern provided in undisturbed cassowary dung, resulted in a lower proportion of seeds germinating and a longer time to germination than did the single seed treatments. This latter treatment represents the deposition pattern for seeds that are removed from dung uneaten or from a fallen fruit. In contrast, whole fruit germinated far more rapidly than did single seeds and the two cassowary gut passage treatments. The driver of this effect is unknown. It is possibly due to a better germination environment existing within the husk of the fruit. Alternatively, because seeds are effectively trapped inside the fruit husk in this treatment, this rapid germination may be a competitive response by seeds to high density situations. If density‐dependent mortality is higher in this situation, then a growth advantage derived through earlier germination may enhance an individual's probability of survival. These results indicate that cassowary consumption has a significant influence on pond apple germination though the nature of this influence is dependent on the final context in which the seed germinates. Whether these differences in timing of germination associated with gut passage and deposition pattern translate into enhanced survival and establishment probabilities at the individual level or into differences in the rate of the invasion process is not immediately clear. If seed mortality during dormancy exceeds that of seedlings or if seedling mortality after a longer dormancy period exceeds that of earlier germinating seedlings, then earlier germination will be advantageous. In contrast, under the opposite conditions slower germination rates will be advantaged. Both these circumstances could potentially be encountered if seed and seedling mortality factors vary in space and time or with deposition pattern. If this is the case, then gut passage and the resulting more even spread of germination across two growing seasons may provide an overall advantage to pond apple by spreading risk. Thus, the results of these experiments suggest that the benefit that accrues from cassowary gut passage and deposition for seed germination and survival will vary depending on current and local conditions but that overall, cassowary gut passage provides a broader range of germination outcomes than those experienced by seeds that fall within whole fruit at least. This, combined with the absence of a clear negative effect, suggests that cassowaries provide high quality seed handling to pond apple. Cassowary‐produced pond apple dispersal kernels Estimated dispersal distances for pond apple seeds ranged up to 5212 m, and while the modal distance was just 38 m, the tail of the kernel is long and fat ( Fig. 2 ). These distances are sufficient to remove seeds well beyond a current infestation and to start fresh infestations. Significantly, this long‐distance dispersal will not be restricted to dispersal within rainforest. Cassowaries are known to move across the landscape, travelling between drainages and vegetation types as distinct as mangroves and rainforest, crossing open ground, and sometimes moving kilometres from rainforest (D.A.W., unpubl. data; Bentrupperbäumer, 1998 ). Such cross‐boundary and long‐distance movement would allow pond apple access to all parts of the landscape including locations where infestations might not be predicted if water dispersal alone is considered, e.g. upstream, uphill, and across catchment boundaries. While long‐distance dispersal has been the focus of much recent attention, it needs to be remembered that short‐distance dispersal is also important, particularly in driving local population dynamics and invasion pattern ( Bolker & Pacala, 999 ; Nathan, 2005 ). The large proportion of seeds dispersed < 50 m, the modal dispersal distance of 38 m, and our observation of dung containing pond apple seeds at infestations indicate that cassowaries will also provide short‐distance dispersal or dispersal directed to existing infestations as birds move from one fruiting tree to the next. In this manner cassowaries will contribute to the consolidation and expansion of these existing infestations ( Bolker & Pacala, 1999 ). Implications of subordinate dispersal modes for population spread Most reported dispersal kernels represent the contributions of one disperser or dispersal mode (but see Redbo‐Torstensson & Telenius, 1995 ). When such kernels are considered alone, the implicit assumption is that they adequately represent the dispersal received by the plant species, despite most plants being dispersed by multiple dispersers or through multiple modes ( Vander Wall & Longland, 2004 ; Dennis & Westcott, in press ; Nathan, in press ). Our results highlight the dangers of failing to consider and incorporate the contributions of multiple dispersers and modes, particularly in the context of plant invasions and their management. Under its dominant dispersal mode, pond apple dispersal will be largely limited to movement with water currents, i.e. downstream and to high‐water lines. This results in a pattern of spread that is largely defined by hydrology, water bodies, and associated vegetation. The addition of cassowary dispersal dramatically alters this picture with supplementary dispersal, both short and long distance, that is unconstrained by hydrology. This dispersal can take seeds to well above high water lines, along or up rather than down altitude contours, or even between drainages. Just as rare, long‐distance dispersal events are influential in determining a population's rate of spread ( Kot ., 1996 ), even small quantities of cassowary dispersal across drainage boundaries or up‐hill will dramatically change the rate and direction of invasive spread by providing access to parts of the landscape that would otherwise be inaccessible, increasing the proportion of habitat susceptible to invasion. This will occur irrespective of whether the dispersal provided is long or short distance, so long as it crosses barriers to the dominant dispersal mode. Such unexpected patterns of spread will greatly complicate management with areas assumed to be weed‐free, potentially harbouring weed populations and continuing to seed areas in which weeds have been eradicated. In complex and closed environments such as rainforests, such infestations may go undetected. Thus, despite providing a subordinate dispersal service, the novel transport method provided by cassowaries and other vertebrate dispersers will act to increase the rate, directional spread, and geographical extent of pond apple invasion. Once having gained access and established beyond the reach of the dominant dispersal mode, cassowary dispersal will also act to maintain populations at these locations, even if these populations are not self‐sustaining and in the face of strongly directional (downstream) dispersal ( Levine, 2003 ). This highlights the need for management plans to consider all likely dispersal. We might expect such effects from novel but subordinate ‘natural’ dispersal modes to not be uncommon ( Vander Wall & Longland, 2004 ). An extreme natural example of the surprises that can arise from subordinate dispersal modes is Symphonia globulifera , a species normally dispersed by forest birds, dispersed on oceanic currents from Africa to the Neotropics ( Dick ., 2003 ). Management Implications Overall, our results show that cassowaries are capable of providing quality short‐ and long‐distance dispersal to pond apple and are capable of acting as effective vectors in the establishment of new infestations. Cassowary dispersal differs in important respects from pond apple's dominant dispersal mode, hydrochory, in that infestations can be established upstream and across drainage boundaries from the source. Though we have only documented cassowary dispersal here, other vertebrate dispersers would provide qualitatively, if not quantitatively, similar outcomes. This is a significant finding as management strategies emphasize water dispersal ( ARMCANZ, 2000 ). Modification of eradication efforts to include searches for seedlings in a radius scaled to the dispersal kernels described here would incorporate additional vertebrate dispersal effects into eradication efforts. Our results indicate that to maximize the probability that all recruits are found, these search zones should extend over a radius of c . 5 km. Shorter, and more realistic, radii could be searched with an approximate certainty of encountering all seedlings that correspond to the cumulative frequency given in Fig. 2 . In summary, both the documented gut passage effects on germination and the estimated dispersal distances indicate that the interaction with cassowaries is a positive one for pond apple. Pond apple appears to have benefited from the pre‐existing, diffuse mutualisms that are common between rainforest plants and their dispersers in that it has attracted a high quality disperser, the cassowary. In doing so it has obtained a subordinate dispersal mode that provides markedly distinct dispersal outcomes to those of its primary mode, water dispersal, and does so in terms of both short‐ and long‐distance dispersal. Importantly, cassowaries and other vertebrate dispersers facilitate access to sites and parts of the landscape that would be unattainable by water transport alone. Provision of such subordinate dispersal services will significantly alter the pattern of invasion and has major implications for managers and researchers alike. ACKNOWLEDGEMENTS We thank Mick and Margaret Tabone at the Johnstone River Crocodile Farm in Innisfail for allowing continued access to and use of their much loved birds, and Shane Campbell for advice. Shane Campbell, Dane Panetta, Tony Grice, Andrew Ford, Ran Nathan, Richard Corlett, Yvonne Buckley, and anonymous reviewers all provided comments on the manuscript at different stages and their contributions have greatly improved it. This work was conducted under the Animal Ethics Approval Number OB15/12. The Australian Government's Marine and Tropical Sciences Research Facility contributed to part of this work. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Diversity and Distributions Wiley

Cassowary dispersal of the invasive pond apple in a tropical rainforest: the contribution of subordinate dispersal modes in invasion

Loading next page...
 
/lp/wiley/cassowary-dispersal-of-the-invasive-pond-apple-in-a-tropical-xpSDyjNA24

References (47)

Publisher
Wiley
Copyright
© 2007 The Authors
ISSN
1366-9516
eISSN
1472-4642
DOI
10.1111/j.1472-4642.2007.00416.x
Publisher site
See Article on Publisher Site

Abstract

INTRODUCTION Seed dispersal is influential as a determinant of plant population and community structure and dynamics ( Nathan & Muller‐Landau, 2000 ; Levey ., 2002 ; Schupp ., 2002 ). As many as 75% to 90% of tropical rainforest tree species ( Webb & Tracey, 1981 ; Howe & Smallwood, 1982 ), accounting for possibly 95% of all seeds ( Terborgh ., 2002 ), are dispersed by a suite of animals that may represent the majority of vertebrate biomass or species at a location ( Terborgh, 1986 ; Gautier‐Hion & Michaloud, 1989 ). These plant–disperser interactions are generally loose mutualisms ( Wheelwright & Orians, 1982 ; Herrera, 1985 ), with plants receiving dispersal from multiple disperser species which in turn utilize multiple plant species. Seed dispersal is also an important component of plant invasion processes ( Rejmánek, 1996 ; Richardson ., 2000 ) with the potential to influence an invasive species’ introduction, establishment, and subsequent spread ( Kot ., 1996 ; With, 2002 ). The diffuse nature of dispersal processes in rainforests should make them relatively prone to infiltration by invasive species. Because seed dispersal processes determine the proportions of seeds moved any given distance from the source plant, understanding the shape and extent of dispersal kernels is necessary for predicting potential rates and pattern of invasive spread ( Kot ., 1996 ). This information can be fundamental in the design and implementation of management ( Trakhtenbrot ., 2005 ). However, vertebrate‐produced dispersal kernels, i.e. the frequency distribution of dispersal distances away from parent plants ( Janzen ., 1976 ), are poorly described even for native species in tropical rainforests ( Levey ., 2002 ; Nathan ., 2002 ) and almost invariably consider only the dominant dispersal mode. For rainforest invaders, descriptions of vertebrate‐produced dispersal kernels are unavailable ( Buckley ., 2006 ). Almost invariably, descriptions of dispersal kernels have focused on the kernel produced by just a single dispersal vector (but see Redbo‐Torstensson & Telenius, 1995 ; Vander Wall & Longland, 2004 ; Jordano ., 2007 ; Dennis & Westcott, in press ), an approach that provides just a fragment of the total dispersal picture. In contrast, all species are dispersed by multiple biotic and abiotic vectors and it is the combined contributions of these multiple vectors that determine the final dispersal kernel ( Nathan, in press ; Dennis & Westcott, in press ). While dominant dispersal modes may shape overall population dynamics, subordinate dispersal modes may contribute key dispersal components that have an overwhelming effect on overall dispersal outcomes ( Higgins ., 2003 ). In this paper we describe the mutualism that exists between an invasive plant and its primary vertebrate disperser in Australia's tropical rainforests. In doing so we provide what is, to our knowledge, the first description of the dispersal kernels generated by a native disperser in a tropical rainforest for an invasive species. This dispersal is provided by an endangered frugivore, the southern cassowary ( Casuarius casuarius , L.) to pond apple ( Annona glabra , L.), a primarily water‐dispersed species that invades intact and mildly disturbed rainforests. Because handling effects and dispersal distance are important components of a disperser's dispersal service, we describe both the effects of gut passage and deposition on germination and the dispersal kernels generated by cassowaries for this invasive species. We discuss these results in the context of whether cassowary dispersal, as an example of a subordinate dispersal mode in terms of the quantity of seed dispersed, can significantly alter pond apple's pattern of invasion and dispersal outcomes over and above those provided by its primary dispersal mode, water. METHODS Study species and study sites Pond apple originates from the Neotropics and was introduced to Australia in 1912 as a wetland rootstock for commercial cultivars of the custard apple, Annona squamosa L. Currently limited in distribution to Queensland, major infestations are found in the wet, tropical north‐east ( Agriculture and Resource Management Council of Australia and New Zealand, 2000 ; Sugars ., 2006 ). Habitats susceptible to pond apple invasion include rainforest, wet land, and riparian plant communities. In these habitats, pond apple can displace native vegetation and form monospecific stands. In Australia, fruiting generally occurs in the summer, peaking for 6–8 weeks between December and March and coinciding with monsoonal rains. Fruits are 7–8 cm in diameter and contain between 100 and 200 seeds, each 12–15 × 8–10 mm. Though a large, fleshy fruit apparently adapted for vertebrate dispersal, in Australia water is considered to be the dominant dispersal agent by far since the buoyant seed and fruit are encountered in great numbers deposited at high water marks and the regional scale pattern of spread is strongly associated with water ( Swarbrick, 1993 ; Sugars ., 2006 ). The large fleshy fruit is also attractive to a variety of vertebrates including cassowaries, pigs ( Sus scofa ), flying foxes ( Pteropus spp.), and a variety of birds and small mammals. While there are many potential dispersers, the size of fruits and seeds means that only cassowaries, pigs, and flying foxes are likely to provide quantities of dispersal with definite records only for the former two. Observations of ripe fruits suggest that cassowaries provide as much as 84% of vertebrate dispersal, pigs provide c . 2%, and predatory rodents move 14%, almost all less than 1 m ( O'Malley, 2004 ). Dung surveys suggest that pond apple is included in pig diets when available but at much lower levels than is the case for cassowaries ( Setter ., 2002 ). The southern cassowary is a large, flightless bird, resident in Australia in the rainforests of the north‐eastern humid tropics ( Davies, 2002 ). The cassowary is listed as endangered in Australia ( Environment Australia, 1999 ). Cassowaries are generally solitary birds occurring in densities of up to 0.08/Ha in lowland forests ( Bentrupperbäumer, 1998 ). The diet consists overwhelmingly of fruits which are generally swallowed whole and the seeds passed intact and viable ( Stocker & Irvine, 1983 ; Bentrupperbäumer, 1998 ; Westcott ., 2005a ). Field observations and collections were performed in February and March 2000 at three sites within 10 km of Innisfail, North Queensland (17°32′ S, 146°02′ E). Dung searches were conducted at each site and the number of pond apple seeds in each dung was counted. Gut passage trials Ripe pond apple fruit was collected in the field and fed to two captive cassowaries at the Johnstone River Crocodile Farm, Innisfail. The trial was limited to these two birds as other captive birds in the region were all housed in pond apple free areas and we sought to eliminate any risk of accidental introductions. Trials were performed in February and March 2001. Cassowaries were hand‐fed pond apple fruit, in feeding episodes that lasted until they lost interest in eating or up to a maximum of 15 min. The birds were fed their usual diet before and after being fed pond apple. The time elapsed from feeding to each defecation was recorded, and the dungs were immediately collected and seeds were counted. The birds were monitored until all consumed seeds were voided. Dispersal kernel estimation Dispersal kernels produced by cassowaries for pond apple were estimated using a combination of movement from Westcott . (2005a ) and gut passage rate data from this study. Though the movement data were not collected from the same locations in which the pond apple infestations occurred, it was collected from a variety of similar habitats and landscape contexts, and we are satisfied that an adequate representation of the dispersal of pond apple is likely to receive. To estimate a seed's dispersal distance, we chose a point in time during a telemetry session and determined its location. This was considered to be equivalent to the time and location of a feeding session. Dispersal distance for a seed consumed in that feeding session was estimated by taking a seed's gut passage rate from the feeding trials and determining the cassowary's location at that time interval. The straight‐line distance between the feeding location and the new position is the animal's displacement and the dispersal distance. Cassowary foraging is not distributed randomly throughout the day and so we distributed starting points, the equivalent of feeding sessions, over the day according to the daily distribution of foraging ( Westcott ., 2005 , Fig. 2 ). We assume that in each feeding session, multiple seeds are consumed and so we randomly selected 200 gut passage rates from the sample of rates obtained in the gut passage trials and determined the dispersal distance for each. The number of starting points used in a telemetry session was determined by the proportion of a day's foraging that would have been included in the timeframe of that session. For a full day this was 20 start times. This process is repeated across all movement pathways, and the dispersal kernel was derived as the frequency distribution of all dispersal distances. This process was repeated 10 times, and the mean percentage and 95% CI of seeds dispersed in each 25 m distance category were determined. 2 Mean cassowary‐produced dispersal kernel for pond apple. Bars indicate the dispersal kernel (left y ‐axis); the dashes represent the 95%CI around this kernel based on 10 kernels (left y ‐axis); the thick solid line represents the cumulative frequency of pond apple seeds dispersed as a function of distance (right y ‐axis). Germination trials Germination trials were conducted to document the effects of both gut passage and deposition pattern on seeds. The seeds used were from dung collected during the gut passage trials and from fresh dung and fruit collected in the wild. To examine gut passage effects we used two single seed treatments: (1) Single seeds with dung material attached (hereafter dung fruit) and (2) single seeds from fresh fruit with arils still attached (fallen fruit). To examine the effect of deposition pattern we used two additional treatments. Because pond apple in cassowary dung is deposited in a clump with other pond apple seeds and because unconsumed fruits fall to the ground whole, we also used two ‘clumped’ treatments: (3) Entire ripe fruit (whole fruit), because unconsumed pond apples fall to the ground whole, and (4) 20 seeds clumped with dung material attached (clumped dung fruit), because cassowaries deposit seeds in clumps. Treatments were placed on the surface of a 1 : 10 mixture of potting mix and a medium–heavy organic clay. Single seed treatments were placed in pots of 73 × 97 mm, and clumped seed treatments and entire fruit in pots of 125 × 140 mm. Pots were arranged randomly and were watered to saturation with tap water each day. Pots were checked twice weekly for germination, defined here as the emergence of the radicle at least 5 mm from the seed. The experiment was ended when all seeds were either germinated or deemed non‐viable upon inspection. We used Fisher exact tests to compare proportions of seeds germinating under the different treatments. The simple distributions of time to germination were analysed using survival analysis to enable seeds that were un‐germinated at the end of the trials to be taken into account. For comparisons of single seed treatments, log‐rank tests were used. Because multiple seeds were monitored in each pot for comparisons involving whole fruit and clumped treatments, these comparisons were conducted using Cox's F ‐tests with the degrees of freedom of the significance test set to the number of pots rather than to the number of seeds to avoid pseudo‐replication. To minimize the risk of type 1 errors when more than one test was performed, false discovery rate control was applied following Verhoeven . (2005 ). In all cases, tests remained significant after control and are reported at their original significance values. RESULTS Field observations and collections Cassowary dungs containing pond apple seeds were found at all three sites, and at one, Coquette Point, on each monitoring day. The presence of adult plants in the vicinity indicates that these dungs were deposited in locations suitable for germination. All 13 dungs collected contained pond apple seed, and 11 contained only pond apple. The average dung contained 199, ±66, SD seeds, ranging from three to 842 ( n = 13). Over 2500 seeds in total were collected with just one of these showing signs of damage. Casual observations elsewhere indicate that germination in cassowary dung is common in the wild (pers. obs.). Two cassowaries (one adult and one sub‐adult) were observed at Coquette Point. On one occasion, the adult bird was observed swallowing a whole pond apple fruit. Though dung indicated the presence of cassowaries at the other sites, they were not observed. Gut passage trial In total, 13 pond apples were fed to the two captive cassowaries with a total of 1512 seeds recovered from their dung. The frequency distribution for pond apple seed gut passage rates through cassowaries was strongly leptokurtic ( Fig. 1 ). Median gut passage time was 180 min (range 65–1675 min, lower and upper quartiles 150–250 min, respectively). Approximately 65% of seeds were passed between 90 and 240 min post‐consumption. A further 30% were passed between 240 and 600 min post‐consumption. 1 Gut passage rates of pond apple seeds through cassowaries. The bars indicate the frequency distribution of gut passage (left y ‐axis). The dashed line indicates the cumulative proportion of seeds recovered (right y ‐axis). Dispersal kernel estimation Dispersal was estimated to occur between 12 and 5212 m with a median dispersal distance of 387 m (quartile range 112–787 m; Fig. 2 ) and a mean distance of 605 m. Twenty‐four per cent of seeds were dispersed less than 100 m, and the modal dispersal distance was 38 m. Seventeen per cent of seeds were dispersed over 1000 m, with 5% and 1% of seeds being transported more than 2062 and 4037 m, respectively. Germination trials In total, 359 seeds were used in the single seed dung treatment and 107 seeds were used in the single seed fallen fruit treatment. A total of 1569 seeds in 10 pots and 1640 seeds in 12 pots were used in the clumped and whole fruit treatments, respectively. With the exception of whole fruits, all seed treatments showed two bouts of germination ( Fig. 3 ). The first began 40–60 days post‐sowing and continued for c . 40 days. Germination ceased with the beginning of cool weather in June. Dormancy continued until the weather warmed again in October and germination continued while warm weather prevailed ( Fig. 3 ). In contrast, there was little further germination in the whole fruit treatment after the first germination bout, most remaining seeds having rotted by day 200. 3 Cumulative germination curves for the four seed treatments: single seed with aril attached (fallen fruit, n = 107 seeds), single seed passed through a cassowary with dung attached (dung fruit, n = 359 seeds), seeds passed through a cassowary and placed in clumps with dung attached (clumped dung fruit, n = 1569 seeds), and whole fruit ( n = 1640 seeds). The proportion of seeds germinating across all treatments varied from 88% to 90% for the single seed treatments and from 79% to 82% for the clumped treatments. Though gut passage increased the proportion of seeds germinating in the single seed trials at the end of the first season ( Fig. 3 ), by the end of the second season there was no difference between the two treatments (dung fruit vs. fallen fruit, Fisher exact P = 0.70). The dung fruit treatment resulted in an increased proportion of germination when compared with the clumped dung fruit treatment (Fisher exact P = 0.01). In contrast, there was no difference between the proportion of seeds germinating in the clumped dung fruit and the whole fruit treatments (Fisher exact P = 0.57) though a greater proportion of whole fruits had germinated by the end of the first season ( Fig. 3 ). In the single seed germination trials, the time to germination of the dung fruit treatment was significantly shorter than that of the fallen fruit treatment (log‐rank test = –2.17, P = 0.03) due to a higher germination during the first season in the former ( Fig. 3 ). In the clumped seed trials, nearly all germination occurred during the first season in the whole fruit treatment resulting in a lower time to germination (median = 84 days) than observed in the clumped dung fruit treatment (median = 288) (Cox's F ‐test, F 10,12 = 4.98, P 0.05). In contrast, the dung fruit treatment showed shorter time to germination (median = 242 days) than the clumped dung fruit treatment (Cox's F ‐test 12,359 = 2.05, P < 0.05). DISCUSSION Cassowaries as dispersers of pond apple Our field observations confirmed that cassowaries are high‐quality consumers and dispersers of pond apple and that they are capable of dispersing large quantities of pond apple seed and commonly do so in infested areas. Cassowaries prefer fruit species that produce predictable, spatially concentrated, and high biomass crops, i.e. easily harvested ( Wright, 2005 ). Pond apple fits this description as individual fruits are large (mean mass = 226 g ± 49.3 SD), individual trees produce large crops and frequently occur in dense stands with relatively synchronous ripening. Consistent with this, the majority of dungs encountered contained only pond apple despite the fact that the species fruits during the period of high fruit availability and diversity in the region ( Crome, 1975 ; Westcott ., 2005b ). Gut passage through animals can be variable in its effect on seed germination. Passage through frugivore guts modifies germination probability or rate in about 50% of cases with germination enhancement, i.e. more rapid germination or increased germination probability, occurring c . twice as frequently as inhibition ( Travaset, 1998 ). In our experiments, a high proportion of ingested seeds were viable after gut passage, irrespective of deposition pattern ( Fig. 3 ). There was no difference in the final proportion of seeds germinating from the different single seed treatments, however, seeds that had both passed through a cassowary and were planted with faecal material still adhering showed higher germination frequencies at the end of the first season and shorter times to germination overall. Deposition pattern influenced germination over and above gut passage. Clumping of seeds from dung, the deposition pattern provided in undisturbed cassowary dung, resulted in a lower proportion of seeds germinating and a longer time to germination than did the single seed treatments. This latter treatment represents the deposition pattern for seeds that are removed from dung uneaten or from a fallen fruit. In contrast, whole fruit germinated far more rapidly than did single seeds and the two cassowary gut passage treatments. The driver of this effect is unknown. It is possibly due to a better germination environment existing within the husk of the fruit. Alternatively, because seeds are effectively trapped inside the fruit husk in this treatment, this rapid germination may be a competitive response by seeds to high density situations. If density‐dependent mortality is higher in this situation, then a growth advantage derived through earlier germination may enhance an individual's probability of survival. These results indicate that cassowary consumption has a significant influence on pond apple germination though the nature of this influence is dependent on the final context in which the seed germinates. Whether these differences in timing of germination associated with gut passage and deposition pattern translate into enhanced survival and establishment probabilities at the individual level or into differences in the rate of the invasion process is not immediately clear. If seed mortality during dormancy exceeds that of seedlings or if seedling mortality after a longer dormancy period exceeds that of earlier germinating seedlings, then earlier germination will be advantageous. In contrast, under the opposite conditions slower germination rates will be advantaged. Both these circumstances could potentially be encountered if seed and seedling mortality factors vary in space and time or with deposition pattern. If this is the case, then gut passage and the resulting more even spread of germination across two growing seasons may provide an overall advantage to pond apple by spreading risk. Thus, the results of these experiments suggest that the benefit that accrues from cassowary gut passage and deposition for seed germination and survival will vary depending on current and local conditions but that overall, cassowary gut passage provides a broader range of germination outcomes than those experienced by seeds that fall within whole fruit at least. This, combined with the absence of a clear negative effect, suggests that cassowaries provide high quality seed handling to pond apple. Cassowary‐produced pond apple dispersal kernels Estimated dispersal distances for pond apple seeds ranged up to 5212 m, and while the modal distance was just 38 m, the tail of the kernel is long and fat ( Fig. 2 ). These distances are sufficient to remove seeds well beyond a current infestation and to start fresh infestations. Significantly, this long‐distance dispersal will not be restricted to dispersal within rainforest. Cassowaries are known to move across the landscape, travelling between drainages and vegetation types as distinct as mangroves and rainforest, crossing open ground, and sometimes moving kilometres from rainforest (D.A.W., unpubl. data; Bentrupperbäumer, 1998 ). Such cross‐boundary and long‐distance movement would allow pond apple access to all parts of the landscape including locations where infestations might not be predicted if water dispersal alone is considered, e.g. upstream, uphill, and across catchment boundaries. While long‐distance dispersal has been the focus of much recent attention, it needs to be remembered that short‐distance dispersal is also important, particularly in driving local population dynamics and invasion pattern ( Bolker & Pacala, 999 ; Nathan, 2005 ). The large proportion of seeds dispersed < 50 m, the modal dispersal distance of 38 m, and our observation of dung containing pond apple seeds at infestations indicate that cassowaries will also provide short‐distance dispersal or dispersal directed to existing infestations as birds move from one fruiting tree to the next. In this manner cassowaries will contribute to the consolidation and expansion of these existing infestations ( Bolker & Pacala, 1999 ). Implications of subordinate dispersal modes for population spread Most reported dispersal kernels represent the contributions of one disperser or dispersal mode (but see Redbo‐Torstensson & Telenius, 1995 ). When such kernels are considered alone, the implicit assumption is that they adequately represent the dispersal received by the plant species, despite most plants being dispersed by multiple dispersers or through multiple modes ( Vander Wall & Longland, 2004 ; Dennis & Westcott, in press ; Nathan, in press ). Our results highlight the dangers of failing to consider and incorporate the contributions of multiple dispersers and modes, particularly in the context of plant invasions and their management. Under its dominant dispersal mode, pond apple dispersal will be largely limited to movement with water currents, i.e. downstream and to high‐water lines. This results in a pattern of spread that is largely defined by hydrology, water bodies, and associated vegetation. The addition of cassowary dispersal dramatically alters this picture with supplementary dispersal, both short and long distance, that is unconstrained by hydrology. This dispersal can take seeds to well above high water lines, along or up rather than down altitude contours, or even between drainages. Just as rare, long‐distance dispersal events are influential in determining a population's rate of spread ( Kot ., 1996 ), even small quantities of cassowary dispersal across drainage boundaries or up‐hill will dramatically change the rate and direction of invasive spread by providing access to parts of the landscape that would otherwise be inaccessible, increasing the proportion of habitat susceptible to invasion. This will occur irrespective of whether the dispersal provided is long or short distance, so long as it crosses barriers to the dominant dispersal mode. Such unexpected patterns of spread will greatly complicate management with areas assumed to be weed‐free, potentially harbouring weed populations and continuing to seed areas in which weeds have been eradicated. In complex and closed environments such as rainforests, such infestations may go undetected. Thus, despite providing a subordinate dispersal service, the novel transport method provided by cassowaries and other vertebrate dispersers will act to increase the rate, directional spread, and geographical extent of pond apple invasion. Once having gained access and established beyond the reach of the dominant dispersal mode, cassowary dispersal will also act to maintain populations at these locations, even if these populations are not self‐sustaining and in the face of strongly directional (downstream) dispersal ( Levine, 2003 ). This highlights the need for management plans to consider all likely dispersal. We might expect such effects from novel but subordinate ‘natural’ dispersal modes to not be uncommon ( Vander Wall & Longland, 2004 ). An extreme natural example of the surprises that can arise from subordinate dispersal modes is Symphonia globulifera , a species normally dispersed by forest birds, dispersed on oceanic currents from Africa to the Neotropics ( Dick ., 2003 ). Management Implications Overall, our results show that cassowaries are capable of providing quality short‐ and long‐distance dispersal to pond apple and are capable of acting as effective vectors in the establishment of new infestations. Cassowary dispersal differs in important respects from pond apple's dominant dispersal mode, hydrochory, in that infestations can be established upstream and across drainage boundaries from the source. Though we have only documented cassowary dispersal here, other vertebrate dispersers would provide qualitatively, if not quantitatively, similar outcomes. This is a significant finding as management strategies emphasize water dispersal ( ARMCANZ, 2000 ). Modification of eradication efforts to include searches for seedlings in a radius scaled to the dispersal kernels described here would incorporate additional vertebrate dispersal effects into eradication efforts. Our results indicate that to maximize the probability that all recruits are found, these search zones should extend over a radius of c . 5 km. Shorter, and more realistic, radii could be searched with an approximate certainty of encountering all seedlings that correspond to the cumulative frequency given in Fig. 2 . In summary, both the documented gut passage effects on germination and the estimated dispersal distances indicate that the interaction with cassowaries is a positive one for pond apple. Pond apple appears to have benefited from the pre‐existing, diffuse mutualisms that are common between rainforest plants and their dispersers in that it has attracted a high quality disperser, the cassowary. In doing so it has obtained a subordinate dispersal mode that provides markedly distinct dispersal outcomes to those of its primary mode, water dispersal, and does so in terms of both short‐ and long‐distance dispersal. Importantly, cassowaries and other vertebrate dispersers facilitate access to sites and parts of the landscape that would be unattainable by water transport alone. Provision of such subordinate dispersal services will significantly alter the pattern of invasion and has major implications for managers and researchers alike. ACKNOWLEDGEMENTS We thank Mick and Margaret Tabone at the Johnstone River Crocodile Farm in Innisfail for allowing continued access to and use of their much loved birds, and Shane Campbell for advice. Shane Campbell, Dane Panetta, Tony Grice, Andrew Ford, Ran Nathan, Richard Corlett, Yvonne Buckley, and anonymous reviewers all provided comments on the manuscript at different stages and their contributions have greatly improved it. This work was conducted under the Animal Ethics Approval Number OB15/12. The Australian Government's Marine and Tropical Sciences Research Facility contributed to part of this work.

Journal

Diversity and DistributionsWiley

Published: Mar 1, 2008

There are no references for this article.