Mixed effects of ingestion by the Aldabran giant tortoise (Aldabrachelys gigantea) on the germination of alien plant species on the Mascarene Islands

Master - Thesis Mixed effects of ingestion by the Aldabran giant tortoise (Aldabrachelys gigantea) on the germination of alien plant species on the Mascarene Islands Ellen Annika Waibel Zürich December 2009 University of Zürich Institute of Environmental Sciences Supervisor: Prof. Dr. Bernhard Schmid Dr. Matthias Albrecht

Abstract Abstract The use of non-native species as analogue species to restore lost ecological interactions is currently hotly debated. On the Mascarene Islands, the suitability of the Aldabran giant tortoise, Aldabrachelys gigantea as extant ecological replacement for the extinct giant tortoises is currently investigated; to see if they can restore lost grazing and seed dispersal functions. However, tortoises are likely to include not only native species in their diets, but also the numerous alien species, potentially facilitating the germination and dispersal of fleshy-fruited, animal-dispersed species. In a pot experiment that run over four months we investigated the effects of ingestion by A. gigantea on the germination success and the germination rate of five alien, partly invasive, fleshy-fruited species on Rodrigues. Feeding of distinctly coloured plastic pellets together with the fruits of these species allowed us to test for individual tortoise effects on seed germination patterns. Overall, ingestion by A. gigantea increased the probability of germination and germination rate, but reduced the number of germinated seeds in pots in which germination occurred. However, both the sign and the magnitude of these effects differed among the experimental plant species. Furthermore, individual tortoises influenced the percentage and rate of germination differently. Moreover, for germination percentage, these tortoise identity effects varied among plant species. Some, but not all of this variation was explained by tortoise age. Ingestion by sub-adult tortoises increased the germination whereas adult tortoises reduced seed germination compared to that of control seeds. We conclude that the consequences of Aldabran giant tortoise ingestion are contingent on the involved plant species. Our findings also emphasize the importance of considering intraspecific effects of frugivore ingestion on seed germination. 1

Table of contents Table of contents Abstract...1 1. Introduction...3 2. Methods...6 2.1 Study site and study species...6 2.2 Feeding of the tortoises and further handling of the seeds...8 2.3 Germination experiment...9 2.4 Statistical analysis...11 3. Results...13 3.1: Effects of tortoise ingestion...13 3.2: Effects of tortoise age and identity...15 4. Discussion...18 4.1: Effects of tortoise ingestion...18 4.2. Effects of tortoise age and identity...21 5. Conclusions and conservation implications...23 6. Acknowledgements...26 7. References...27 Index of figures Figure 1 Overview of the experimental plant species; before and after ingestion (1)...9 Figure 2 Experimental design...11 Figure 3 mean percentage of germination, comparison between the control and ingested seeds of the species. The mean of the ingested seeds was calculated out of all treatments concerning this species. The bars show the standard error of the mean...14 Figure 4 effect of time to germination on percentage of germination of the pooled experimental plant species, without S. cumini. Comparison between control and ingested seeds. Bars represent the standard error of the mean...15 Figure 5 Germination rate, referring to Table 4; comparison between adult, sub-adult and control seeds of the pooled experimental species, S. cumini was excluded (as was left out of the model for germination rate)...16 Index of tables Table 1 Table 2 Table 3 Table 4 Fruit and seed characteristics of the plant species used in the experiment. From all species, 30 fruits were taken to calculate the mean number of seeds per fruit, the seed size and fruit size....7 Mixed-model analyses of variance to test the probability of germination...16 Mixed-model analyses to test the percentage of germination of the germinated seeds..17 Mixed-model analyses to test the germination rate (S. cumini excluded)...17 2

Inroduction 1. Introduction Vertebrates play an important role in the reproduction and seed dispersal of many plant species (Van der Pijl 1982; Olesen and Valido 2003). Seed dispersal is important for plant population persistence and ecosystem health. Especially plants producing fleshy fruits are likely to be dispersed by vertebrates, and might benefit from this dispersion as new suitable habitats can be reached. Several studies have shown that dispersal away from the parent tree can enhance the germination success (Janzen 1970; Hansen, Kaiser et al. 2008). The quality of animal seed dispersal does not only depend on spatial distribution of the seeds but also on the effects of gut-passage (Traveset 1998). Seed germination success and rate can be altered by animal ingestion; seed germination is more often enhanced than inhibited (Traveset 1998). The removal of the pulp, mechanical or chemical abrasion of the seed coat during gut passage, or the provision of a nutrient rich, moist microsite in the animals faeces have been attributed to have significant effects on seed germination (McDiarmid, Ricklefs et al. 1977; Barnea, Yomtov et al. 1990; Malo and Suarez 1996). Most research concerning the effect of frugivory has been done with birds and mammals (Traveset 1998). The importance of reptiles as seed dispersers has often been underestimated, especially in island ecosystems (Olesen and Valido 2003). For example, in the Mascarene archipelago, reptiles and birds compose a large part of the fauna as few mammals were able to colonise these isolated islands (Rick and Bowman 1961; Olesen and Valido 2003). On these islands, reptiles and birds perform many of the key functions, such as grazing, pollination and seed dispersal (Olesen and Valido 2003; Cheke and Hume 2008). Like many island ecosystems, the arrival of humans was followed by the extinction of much of the endemic fauna and flora (Whittaker 2007). Moreover, humans introduced numerous alien species, of which some became invasive (Cheke and Hume 2008). The loss of key species and the introduction of alien species lead to dysfunctional ecosystems (Cheke and Hume 2008; Hansen, Kaiser et al. 2008). Alien plant species are a serious problem worldwide, suppressing native plants and developing new interactions between plants and animals (Bartuszevige and Gorchov 2006; Westcott, Setter et al. 2008). Vila and D'Antonio found that the fruit preference of frugivores and seedling survival can facilitate the spread of an aggressive introduced species (Vila and D'Antonio 1998). Dispersal modes play an important role in invasion patterns and influence the ultimate success of a plant (Westcott, Setter et al. 2008). Rodrigues is one of the three Mascarene Islands located 560 km east of Mauritius. Despite only being colonized in 1601, many plant and animal species went extinct and large 3

Inroduction proportion of the native fauna and flora became endangered through habitat destruction and degradation, overexploitation and the introduction of alien invasive species. 305 of the alien species on the island have already been naturalised and comprise 78% of the flora (Cheke and Hume 2008). Francois Leguat, who provided the first description of the island in 1708, noted two species of giant tortoises which he believed created the little under-story in the forests through grazing (Leguat 1708). It is assumed that they were the predominant large native herbivores(cheke and Hume 2008). The two giant tortoise species, Cylindraspis vosmaeri and Cylindraspis peltastes, are believed to have evolved from Cylindraspis that first arrived on Mauritius and then spread to Réunion and Rodrigues, where sympatric speciation took place (Austin, Arnold et al. 2003). The harvesting by European sailors resulted in their extinction. The loss of these giant tortoise species was maybe one the most severe ones for the Mascarene Islands as many co-evolved plant-animal interactions are likely to be suffering from a legacy of lost ecosystem services and functions (Cheke and Hume 2008). Overexploitation also almost caused the extinction of Aldabrachelys gigantea, the only surviving giant tortoise species in the Indian Ocean. As a conservation measure, A. gigantea was brought to Mauritius in the 1880s, for captive breeding (Stoddart, Peake et al. 1979). Today, Aldabran giant tortoises are captive-bred at La Vanille Nature Park (Weaver and Griffiths 2008), Mauritius, and in 2006, they were brought to the Francois Leguat Reserve in Rodrigues. Furthermore populations exist on two Mauritian offshore islands. In 2002, they were introduced to Ile aux Aigrettes and in 2007 to Round Island. Both Islands are National Parks and managed by Conservation Services. Research about the introduction of the tortoises will determine if they can be further used as ecological analogues that may help to restore grazing and seed dispersal functions once fulfilled by the extinct giant tortoises (Griffiths, Jones et al. 2009). Preliminary results from research on Round Island suggested that the tortoises have a positive effect on the native vegetation, by preferentially grazing the alien grasses and herbs (Griffiths, Jones et al. 2009), which are the greatest threat to most habitats on the Mascarene Islands (Strahm 1989). Yet little is known about dispersal modes of alien species (Vila and D'Antonio 1998; Bartuszevige and Gorchov 2006; Westcott, Setter et al. 2008; Linnebjerg, Hansen et al. 2009) and no research was found dealing with the effects of ingestion by reptiles on alien fleshyfruit species. In addition former research concentrated on the general effect of animal ingestion but not on the intraspecific differences among a frugivorous species. Variation in age, size or other physical morphologies might alter the germination success as well. Moreover only a few studies give attention to germination of seeds after ingestion by tortoises 4

Inroduction (but see: Braun and Brooks 1987; Cobo and Andreu 1988; Moll and Jansen 1995; Strong and Fragoso 2006; Guzman and Stevenson 2008; Moolna 2008). Considering conservation management the introduction of tortoises as an analogue species on the Mascarene Islands may lead to significant changes in the ecosystem. Ingestion can profoundly alter the fate of seeds and seedling survival, which will be crucial for the final composition of plant communities (Loiselle 1990; Traveset 1998). As it in known that Aldabran tortoises are generalists (Gibson and Hamilton 1983) they are likely to feed also on alien fleshy-fruited species and dispersal will alter the dispersal mode of both native and invasive plant species. The main aim of this study is to assess the potential impact of the introduction of Aldabran giant tortoises as analogues on the success and rate of seed germination of five fleshy-fruited plant species alien and partly invasive in the Mascarene Islands. The following questions were examined: (1) How does ingestion by the Aldabran giant tortoise alter the germination success and germination rate of these species (2) Do plant species, differing in seed size, respond differently to tortoise ingestion? (3) Do individual Aldabran giant tortoises influence the germination success and germination rate of ingested seeds differently? (4) How does tortoise age affect the germination of ingested seeds? 5

Methods 2. Methods 2.1 Study site and study species The experiment was carried out at Francois Leguat Reserve (63 22 11.24 E, 19 45 15.76 S), near Anse Quitor on the south-western tip of Rodrigues. Rodrigues is part of the Mascarene archipelago located 590 km east of Mauritius in the Indian Ocean. The island is characterized by a tropical maritime climate in summer (November to April) and trade winds from subtropical anticyclones in winter (May to October). The mean temperature ranges from 21 C in winter to 29 C in summer. The mean annual rainfall for Rodrigues (Pointe Canon) is 1105 mm (Mauritius-Meteorological-Services). The Francois Leguat Reserve possesses over 400 Aldabran giant tortoises in a 25 ha park area. Prior to the introduction of tortoises the area was weeded, removing most alien plants. The tortoises are kept in natural enclosures, resulting from the collapse of limestone caves. The aim of the park is to conserve the Aldabran giant tortoises, create a popular tourist attraction and carry out conservation research on using the Aldabran giant tortoises as an analogue species in the Mascarene Islands. Several theories about the evolution of the Aldabran tortoise are discussed in literature. Either that the genus evolved on Madagascar and then spread over to the Seychelles or that it evolved on the Seychelles and colonised Madagascar afterwards. Nevertheless both theories place Madagascar as the source of the tortoises that colonised Aldabra (discussed in Austin, Arnold et al. 2003). The Aldabran giant tortoise was brought to Mauritius and neighbouring islands in the 1880s for conservation purposes. Several prominent scientists of that time, including Hooker, Own, Darwin and Newton saw a serious decline of the population through harvesting and thus the demand for conservation (Stoddart, Peake et al. 1979). Aldabrachelys gigantea and Geochelone nigra in Galapagos belong to the same genus and are the only surviving giant tortoises, but about 13 smaller species belong to the genus (Arnold 1979). Today, roughly 100000 Aldabran giant tortoises live on the Aldabra atoll (Bourn, Gibson et al. 1999). They roam free on the islands and feed on plant material, mainly on leaves and fruits (Hnatiuk 1978). The largest male found by Grubb (1971) weighed around 300 kg with an carapace length of 106 cm. Females are generally smaller, the largest observed weighed about 65 kg and measured 79 cm in carapace length (Grubb 1971). Rodrigues has 132 native plant species, of which 44 (33%) are endemic. It also has 305 naturalised alien species which comprise 78% of the flora (Cheke and Hume 2008). For the experiment, five alien, fleshy-fruited species were used (Table 1, Fig. 1). From the plant 6

Methods species that had ripe fruits during the time of the experiment these five were chosen as they are attractive for tortoises and because of their status of invasion in Rodrigues. Due to their fruit colour they are likely to be eaten by tortoises (Dodd 2001) and most of the species are invasive throughout the Mascarene Islands. Table 1 Fruit and seed characteristics of the plant species used in the experiment. From all species, 30 fruits were taken to calculate the mean number of seeds per fruit, the seed size and fruit size. Scientific name Family Fruit colour Fruit size (mm) Number of seeds per fruit Seed size (mm) Wikstroemia indica Thymelaeceae red/orange 8.1 ± 0.1 1 ± 0.0 5.8 ± 0.1 Syzygium cumini Myrtaceae black 19.0 ± 0.4 1 ± 0.0 13.4 ± 0.5 Lantana camara Verbenaceae black 4.4 ± 0.9 1 ± 0.0 4.3 ± 0.7 Veitchia merrillii Arecaceae red/orange 31.4 ± 0.3 1 ± 0.0 18.8 ± 0.2 Mimusops coriacea Sapotaceae yellow 38.4 ± 0.8 4 ± 0.2 19.6 ± 0.2 Wikstroemia indica (L., C. A. Mey) This alien herb introduced to Rodrigues in 1828 is widespread and invasive in the higher areas of the island. It grows in open habitats and can reach a height of 1.70 m (A. Waibel, personal observation). Its natural range extends from south-east Asia to Australia (Guy Rouillard and Guého 1999). Syzygium cumini (L., Skeels) This alien tree is widely distributed in Indo-Malaysia and alien to Rodrigues and Mauritius, where it is naturalised and grows on the coast but also in lowland and upland forest. It was introduced to Rodrigues around 1800 for its fruits. The seeds of S. cumini show polyembryony. After fertilisation, the egg apparatus degenerates, and the upper half of the nucellus produces several embryos, of which several may survive, resulting in polyembryony (Narayanaswami 1960). Lantana camara (L.) This highly invasive shrub, listed among the 100 worst invasive organisms (ISSG-list, 2), grows in dense coppices, up to 1 m in height, and suppressing other vegetation. It was introduced to Rodrigues around 1920 (Cheke and Hume 2008) and is the dominant vegetation in the south of Rodrigues. L. camara originates from tropical America. 7

Methods Veitchia merrillii (Becc., H.E Moore) This species originates in the Philippines and was introduced to the Mascarene Islands for its ornamental beauty. On Rodrigues the species inhabits coastal and lowland forests. It is often cultivated in gardens, but to date not considered as invasive on the Mascarenes. Mimusops coriacea (DC, Miq.) This partly invasive tree originates from Madagascar and was introduced to Rodrigues in 1800 for its fruits. In Rodrigues, the tree grows mainly in upland forests (pers. obs.), but occurs also in coastal regions and lowland forests (Guy Rouillard and Guého 1999). 2.2 Feeding of the tortoises and further handling of the seeds From 24 April to 11 May 2009, fruits from the experimental plant species were fed to 5 adult (est. 60 120 years old, 75 180 kg) and 22 sub-adult Aldabran giant tortoises (10 15 years old, 20 30 kg). The adults and sub-adults were kept in two different enclosures. The subadults were fed with all 5 plant species. The adults were fed with L. camara, S. cumini and M. coriacea. Roughly the same amount of fruits from one species was fed to each tortoise by feeding them individually out of plastic bowels. To be able to identify the droppings of each individual tortoise, each tortoise was also fed distinctively coloured plastic pellets (between 1 and 5 mm in diameter) The pellets were obtained from a fabric producing the pellets for industrial use, 16 different coloured pellets with round edges to avoid injuries in the gut, were available (master batches, Albert GmbH & Co. KG, 32257 Bünde, Nordrhein-Westfalen). The pellets were presented at each meal, together with banana or papaya. Between 20 and 60 pellets were fed to each tortoise at each meal. The enclosures were checked daily for faeces, and all faeces were collected. Figure 1 shows the experimental plant species, before and after ingestion by the tortoises. The retention time in the gut of the tortoises varied between adults and sub-adults. First seeds were found after a week in the sub-adults faeces, whereas for the adults it took around two weeks. In some cases even one month after feeding, seeds were still found in the faeces. All seeds were taken out of the faeces and the seeds of the experimental species were kept in paper bags, in dry shady conditions, until the date of sowing. Because some droppings did not contain any pellets and trampling took place, it was not possible to unambiguously assign enough seeds of 8

Methods all plant species to some individual tortoises. For the sub-adults, we could assign enough seeds from each experimental species to a total of 8 tortoise individuals with the exception of M. coriacea, for which seeds could be assigned to 7 tortoise individuals. For the adults, enough seeds of all species could be assigned to each of the 5 individual tortoises. The seeds from droppings that could not be assigned to an individual were mixed for each plant species. The mixtures obtained from adult and sub-adult tortoises were kept separately. The seeds obtained from droppings were stored between 10 and 24 days, until enough seeds for the germination experiment were available. Also the faecal material, freed from any seeds, was kept separately for each individual tortoise. The faecal material that could not be assigned to an individual was stored as mixtures, separately for adults and sub-adults. V. merrillii V. merrillii M. coriacea M. coriacea L. camara L. camara W. indica W. indica S. cumini S. cumini Figure 1 Overview of the experimental plant species; before and after ingestion (1) 2.3 Germination experiment To investigate the effect of ingestion by A. gigantea on the germination success and time to germination of the five experimental plant species, the following experiment was conducted in a nursery at Francois Leguat Reserve from 19 May to 8 September 2009. The fruits for the 9

Methods experiment were collected around the island. The nursery (3 x 9 m) was covered by Sharlon cloth to protect the plants from strong solar radiation. For the control treatment, all fruits were collected on the day of planting or the day before. For each species, the control fruits and the ingested seeds were planted in monoculture on the same day. All species were planted from 19 May to 4 June 2009. In the ingestion treatment, the seeds were planted in a layer of 100 ml of seed-free faecal material obtained from the same tortoise individual as the seeds. Where the seeds could not be assigned to an individual tortoise, a mixture of faecal material from sub-adults or adults, respectively, was used, according to the age class to which the ingested seeds were assigned to. In the control treatment, seeds were planted in a layer of 100 ml of soil. All pots (8 cm diameter) were filled with 400 ml of soil before the seeds/ fruits were placed in the pot. In both treatments, the seeds were covered with a 5 mm layer of fine soil. The soil was collected in the dry river bank of Anse Quitor River, in the nature reserve near the Francois Leguat Reserve. For planting, eight seeds were taken for the gut-passage treatment and 8 or 2 fruits for the control. Except for L. camara, in both, gut-passage treatment and control, 6 seeds were taken for planting. M. coriacea was the only species which had a multi-seeded fruit (Table 1), the mean number of seeds was calculated out of 30 fruits and determined as 4. Thus the number of seeds was roughly constant among the plant species. Three replicates were planted for each individual tortoise. In addition, 9 replicates for the adult mixture and 21 for the sub-adult mixture were planted of each treatment. The experimental design is illustrated in Figure 2. The tortoises cut some of the seeds of S. cumini, resulting in seed-propagules. Both intact seeds and seed propagules were used for the experiment. Eight of them were randomly taken from the available seeds/ propagules from an individual or a seed mixture. The pots were put into three blocks, each block containing an equal number of pots of all treatment types. The replicates from the individual tortoises, the seed mixtures from adults and sub-adults and the control pots were divided at random but in equal numbers among the blocks. Within the blocks, the pots were placed at random. In total, 588 pots were planted. 10

Methods Tortoise gut-passage Fresh fruits Sub-adult tortoises Control Adult tortoises Seeds of individual sub-adults Seeds of individual adults Seed-mixture of all sub-adults Seed-mixture of all adults Figure 2 Experimental design The plants were watered to keep the soil moist. Slug poison was used to prevent herbivory by snails. The pots were checked every day for germination. Each time a plant germinated, a toothpick was placed next to it to avoid the double-counting of seedlings. Germination was defined as the first emergence of any part of the seedling (Moolna 2008). Also survival was checked every day. If a seedling was found to be dead, the toothpick and the dead seedling were removed from the pot. The mortality of the seedlings was very low during the experimental period and was therefore not analysed (10 plants died during the entire experiment). 2.4 Statistical analysis For the analysis of germination success (i.e. the probability of germination and the percentage of germinated seeds at the end of the experiment) we used generalized and general linear models. To analyse germination probability a logistic model with quasi-binomial errors and a logit-link function were fitted to the binary response variable germination (pots with 1 seedling at the end of the experiment) or no germination (pots with 0 seedlings at the end of the experiment). The deviance for the treatment factor, nested with block, was partitioned into 11

Methods a contrast ingestion (seeds ingested by tortoises vs. control seeds), a contrast tortoise age (adults vs. sub-adults), a contrast composition (ingested seeds from a mixture of droppings that could not be attributed to a specific individual tortoise vs. seeds from droppings that could be attributed to individual tortoises) and the remaining deviance of the treatment factor (individual tortoise identity). The deviance for plant species was decomposed into a linear contrast seed size and the remaining deviance tested before the treatment contrasts in the model (Table 2). Treatment contrasts were tested against the remaining deviance of the treatment factor, the species contrasts and their interactions against the according species contrasts x remaining treatment deviance interactions and the remaining treatment deviance against the pots as the residuals (Table 2). To account for overdispersion, the data quasi-f tests were performed by calculating ratios of mean deviance changes (Crawley 2007). To analyse the percentage of germinated seeds (including only pots in which germinations occurred) a linear model with the same terms fitted sequentially in the same order as described above for the logistic model was used (Table 3). Due to the polyembryony of Syzygium cumini, values of this species could be > 1. To achieve normality and homoscedasticity of the residuals the response variable was log-transformed. The time to germinate was analysed using a generalized linear model with quasi-binomial errors and a complementary log-log link function recommended for analyses of interval counts (Egli and Schmid 2001). Germination percentage was calculated for each time interval (each of four month) separately, using the total number of sown seeds minus the already germinated seeds for the calculations of the percentage germination during a time interval (Egli and Schmid 2001). The factor time (month), decomposed into a linear contrast and the deviation from linearity, was included into a model with the same terms sequentially fitted as described for the analyses of germination probability and germination percentage (Table 4). Syzygium. cumini was excluded from this analysis due to its polyembryony. Since in all analyses, plant species explained a large proportion of the variance, separate models for each plant species were fitted to further explore how the effects differed for the different plant species. The same models as described above were fitted (without the plant species contrasts and their interactions). All analyses were carried out with R 2.7.2 (R Development Core Team 2008-08-25). Arithmetic means ± 1 standard error are reported. 12

Results 3. Results 3.1: Effects of tortoise ingestion Tortoise ingestion increased the overall germination probability but reduced the percentage of germinated seeds (Table 2, 3). Since values of Syzygium cumini for the percentage germination exceeded in many pots 100% due to its polyembryony, particularly for the control seed treatment (Fig. 3), this species could possibly have masked the effects in the other plant species. Therefore, S. cumini was included as a factor in a linear model analysing percentage of germinated seeds to test how the effects change when accounting for the variation explained by S. cumini. When accounting for the variation explained by S. cumini in the model, the effect of ingestion was reversed, i.e. the percentage of germination was higher for ingested seeds (F 1,16 = 7.06, P = 0.017 ) (Fig. 3). In addition the species x ingestion interaction was no longer significant. When analyzing plant species separately, tortoise ingestion reduced the germination probability and percentage germination of S. cumini (F 1,12 = 8.06, P = 0.015; F 1,12 = 38.67, P < 0.001) and germination probability of W. indica (F 1,7 = 6.64, P = 0.037). However, only 1 ingested and 6 non-ingested seeds of this latter species germinated during the experiment (Fig. 3). Conversely, both germination probability and percentage germination of M. coriacea were increased by ingestion (F 1,11 = 21.36, P < 0.001; F 1,10 = 4.99, P = 0.049). The germination success of L. camera and V. merrillil was not significantly affected (neither germination probability nor germination percentage: all P > 0.1). In L. camera, however, only few seeds germinated during the experiment, thus the non-significant effect of ingestion may be due to the low statistical power in these analyses (Fig. 3). Ingestion of seeds led to an increased germination rate, i.e. a higher proportion of seeds germinated in the early compared to the late months of the experiment (Table 3, Fig. 4). However, the magnitude and sign of the effect of ingestion on germination rate differed among plant species. The ingested seeds of V. merrillil, M. coriacea and L. camera germinated faster then the control seeds (F 1, 8 = 25.73, P = 0.001; F 1,13 = 123.17, P < 0.001; F 1,14 = 5.95, P = 0.029). Due to the low number of seeds of W. indica that germinated, germination rate was not analysed for this species. A large part of the variation in germination success and time to germination among species was explained by seed size; large-seeded species were more likely to germinate than smaller seeded species (Table 2 and 3). More importantly, the probability of germination of 13

Results small-seeded species was relatively less increased by tortoise ingestion than that of largeseeded species (Table 2). Conversely, the germination percentage of small-seeded seeds was relatively less reduced by tortoise ingestion than that of large-seeded species. However, when accounting for the polyembronic S. cumini (see above), the ingestion x seed size interaction was no longer significant for the percentage of germinated seeds. 200 180 Control Tortoise Percentage of germination 160 140 120 100 80 60 40 20 0 V. merrillii W. indica S. cumini L. camera M. coriacea pooled species pooled species, without S. cumini Figure 3 mean percentage of germination, comparison between the control and ingested seeds of the species. The mean of the ingested seeds was calculated out of all treatments concerning this species. The bars show the standard error of the mean 14

Results 50 40 Control Tortoise Percentage of germination 30 20 10 0 1 2 3 4 Time [month] Figure 4 effect of time to germination on percentage of germination of the pooled experimental plant species, without S. cumini. Comparison between control and ingested seeds. Bars represent the standard error of the mean 3.2: Effects of tortoise age and identity Germination success was higher for seeds ingested by sub-adult tortoises than for those ingested by adults (Table 4, Fig. 5). However, when analysing the germination probability (Table 2) and percentage of germinated seeds (Table 3) separately, the effect of age was not significant. Furthermore, tortoise age had no significant influence on the germination rate (Table 4). However, how age classes affected the probability of germination varied among plant species (Table 2). This variation could not be attributed to seed size differences among plants (Table 2 4). Tortoise identity, after accounting for age class differences, explained a significant proportion of the variation in the percentage of ingested seeds that germinated and in the germination rate (Table 3, 4). Furthermore, individual tortoises affected the percentage germination of germinated seeds of different plant species differently (Table 3). Differences 15

Results in seed size among plant species partly accounted for these individual tortoise effects on percentage germination (Table 3). 50 45 Adult Sub-adult Control Percentage of germination 40 35 30 25 20 15 Adult Sub-adult Control Figure 5 Germination rate, referring to Table 4; comparison between adult, sub-adult and control seeds of the pooled experimental species, S. cumini was excluded (as was left out of the model for germination rate) Table 2 Mixed-model analyses of variance to test the probability of germination d.f. Error Deviance Deviance change (%) F-value P-value Block 2 R 0.18 0.01 0.13 0.881 Seed size 1 SSxT 282.61 21.03 269.15 <0.001 Species 3 SxT 157.37 11.71 103.34 <0.001 Ingestion 1 T 10.72 0.80 12.24 0.002 Tortoise age 1 T 1.45 0.11 1.66 0.213 Tortoise ID (T) 20 R 17.52 1.30 1.23 0.222 Seed size X Ingestion 1 SSxT 20.26 1.51 19.30 0.001 Seed size X Age 1 SSxT 0.4 0.03 0.38 0.546 Seed size X Tortoise ID (SSxT) 15 R 15.75 1.17 1.48 0.109 Species X Ingestion 3 SxT 23.15 1.72 15.20 <0.001 Species X Age 1 SxT 3.25 0.24 6.40 0.019 Species X Tortoise ID (SxT) 21 R 10.66 0.79 0.71 0.821 Residuals (pot) (R) 587 800.69 59.57 16

Results Table 3 Mixed-model analyses to test the percentage of germination of the germinated seeds d.f. Error SS %SS F-ratio P-value Block 2 R 0.13 0.08 0.44 0.646 Seed size 1 SSxT 7.58 4.39 8.44 0.014 Species 3 SxT 56.09 32.48 51.37 <0.001 Ingestion 1 T 6.60 3.82 7.06 0.017 Composition (mix or ID) 1 T 0.01 0.01 0.01 0.904 Tortoise age 1 T 2.50 1.45 2.67 0.122 Tortoise ID (T) 16 R 14.95 8.66 6.26 <0.001 Seed size X Ingestion 1 SSxT 4.90 2.84 5.46 0.039 Seed size X Composition 1 SSxT 0.07 0.04 0.08 0.785 Seed size X Age 1 SSxT 0.12 0.07 0.14 0.719 Seed size X Tortoise ID (SSxT) 11 R 9.88 5.72 6.02 <0.001 Species X Ingestion 2 SxT 23.17 13.42 31.82 <0.001 Species X Composition 2 SxT 0.86 0.50 1.18 0.361 Species X Age 1 SxT 0.40 0.23 1.10 0.329 Species X Tortoise ID (SxT) 7 R 2.55 1.48 2.44 0.019 Residuals (pot) (R) 287 42.86 24.82 Table 4 Mixed-model analyses to test the germination rate (S. cumini excluded) d.f. Error Deviance Deviance change (%) F-ratio P-value Block 2 R 3 0.03 1.48 0.228 Time (month, numeric) 1 TNxT 3.4 0.04 0.89 0.355 Time (month, factor) 2 TFxT 1319.8 14.16 400.52 <0.001 Seed size 1 SSxT 1259.2 13.51 913.41 <0.001 Species 2 SxT 413 4.43 32.27 <0.001 Ingestion 1 T 143.2 1.54 32.51 <0.001 Composition (mix or ID) 1 T 0.5 0.01 0.11 0.740 Tortoise age 1 T 19.8 0.21 4.50 0.046 Tortoise ID (T) 21 R 92.5 0.99 4.41 <0.001 Time (numeric) X Ingestion 1 TNxT 311.4 3.34 81.84 <0.001 Time (numeric) X Composition 1 TNxT 1 0.01 0.26 0.614 Time (numeric) X Age 1 TNxT 8.9 0.10 2.34 0.141 Time (numeric) X Tortoise ID (TNxT) 21 R 79.9 0.86 3.81 <0.001 Time (factor) X Ingestion 2 TFxT 140.6 1.51 42.67 <0.001 Time (factor) X Composition 2 TFxT 5.6 0.06 1.70 0.195 Time (factor) X Age 2 TFxT 5.4 0.06 1.64 0.206 Time (factor) X Tortoise ID (TFxT) 42 R 69.2 0.74 1.65 0.005 Seed size X Ingestion 1 SSxT 5.1 0.05 3.70 0.075 Seed size X Composition 1 SSxT 0.3 0.00 0.22 0.648 Seed size X Age 1 SSxT 7 0.08 5.08 0.041 Seed size X Tortoise ID (SSxT) 14 R 19.3 0.21 1.38 0.153 Species X Ingestion 2 SxT 42.7 0.46 3.34 0.088 Species X Composition 2 SxT 0.3 0.00 0.02 0.977 Species X Tortoise ID (SxT) 8 R 51.2 0.55 6.41 <0.001 Residual (pot) (R ) 1740 5318.6 57.06 17

Discussion 4. Discussion 4.1: Effects of tortoise ingestion Tortoise ingestion influenced the germination success significantly; it increased the overall germination probability, reduced the germination percentage of germinated seeds and influenced the germination rate of the pooled experimental plant species. But the magnitude and sign of these effect differed among the plant species. The direction of the ingestion effect of the percentage of germinated seeds turned around when S. cumini was fitted before the other factors. Then the percentage of germination was increased by tortoise ingestion. Several other researches have looked at the influence of ingestion of tortoises on germination and found that after gut passage, most of the seeds are still able to germinate and some species do even germinate better (Braun and Brooks 1987; Cobo and Andreu 1988; Moll and Jansen 1995; Strong and Fragoso 2006; Guzman and Stevenson 2008; Moolna 2008). Our results agree with the fact that seeds are still able to germinate after ingestion. The review of Traveset (1998) shows, that in 56 % of previous experiments the germination was unaffected. When there was an effect, it was more often positive. But the germination rate was more often accelerated (47%) then delayed (16%). However this research focused on germination effects on native plant species. Generally it has been assumed that the fruits of some plant species which have coevolved with their frugivores are adapted towards ingestion. In special cases it is even possible that ingestion by the frugivore is obligatory. This was found for the Galapagos tomatoes which seem to need the ingestion of the Galapagos tortoises for regeneration (Rick and Bowman 1961; Gibbs, Marquez et al. 2008) and the tambalacoque tree on Mauritius which is thought to require the gut-passage by the nowextinct dodo (Temple 1977). In contrast to that, alien species, which do not originate from habitats where they are used to ingestion, are likely to be influenced negatively by ingestion. In our study the germination success of alien plant species was tested. The germination success of 40% of the present case species was reduced. Contrary, 80% of the species obtained an accelerated germination rate and only 40% were unaffected concerning germination success after ingestion. The seeds of S. cumini were broken by the tortoise s mouth while chewing. Approximately half of the ingested seeds were cut into smaller propagules. A study with Syzygium mamillatum, endemic to Mauritius, also showed seed fragmentation through Aldabran tortoises: the seeds were broken during gut-passage and only 15.8% passed the gut 18

Discussion unharmed (Hansen, Kaiser et al. 2008). Hanson et al. discuss that this causes the lower germination success after gut-passage, only 18% of the ingested S. mamillatum seeds germinated. Compared to this S. cumini had a high percentage of germination after ingestion (64%), which might partially be caused by its polyembryony. To see whether the propagules were able to germinate as well three pots with only propagules were planted additionally to the main experiment. In one pot 38% of the propagules germinated. Still the control seeds germinated better and therefore the breaking of the seeds could be a major cause for the reduced germination success after ingestion. Seeds from the other experimental plant species had no visible damage, neither through tortoise feeding nor gut-passage. The removal of the fruit pulp has been attributed to be an important effect of fruit ingestion by animals (Traveset 1998). We can not determine whether de-pulping or another trait of ingestion influenced the germination success because manually de-pulped fruits were not included in the control treatment. We did not aim to test these components of ingestion separately but addressed our questions towards the general effect of ingestion as it would take place in nature. Therefore the defecated seeds were replanted into the faeces of the tortoises and the whole effect of the tortoise was taken into account; on the one side the removal of the fruit pulp and the exposure of the seeds to the chemicals in the gut and on the other side the influence of the nutrient rich faeces are given. In addition, the treatments aimed to concentrate on the effects of the individual tortoise (age and individual effects). The germination of the ingested seeds, as it might occur in nature, was compared with the germination of the seeds in the fresh fruit. Further the experimental approach tried to simulate natural conditions as close as it was possible. The pots were planted in a semi closed nursery and exposed to all weather conditions. Thus the results are likely to be more suitable for transferring them into nature (Rodriguez-Perez, Riera et al. 2005; Samuels and Levey 2005). However, our results indicate that three species might be influenced by the mechanical removal of the pulp. Although germination success of M. coriacea was reduced after ingestion, the ingested seeds germinated faster compared to the control seeds. The seeds of M. coriacea are covered by a thick layer of fruit pulp, which was completely removed after ingestion. The fruit pulp rotted very slowly in the control treatment (6-8 weeks) and had to break open before the seeds were able to germinate (A. Waibel, personal observation). A similar effect was found for V. merrillii. The germination success did not differ between ingested and control seeds but similarly to M. coriacea the time to germination was accelerated after ingestion. The fresh fruits of V. merrillii are coated with a thin layer of red fruit pulp which was removed during gut-passage. The missing of the fruit flesh could have 19

Discussion caused the faster germination after gut passage. In addition, the seed coat of V. merrillii and M. coriacea is very hard and might not have been abraded in the gut. Also for W. indica the whole fruit pulp was removed during gut-passage. The seeds are surrounded by a fine grey coat, which was also removed by gut-passage. This might be the reason for the low germination success after ingestion. The germination of W. indica was very generally very low and only one seed germinated after gut passage. The germination success of the most invasive species of the experiment, L. camera, was not influenced by the tortoise gut passage. But the germination rate for ingested seeds decreased. However the number of germinated seeds of L. camera was very low. It might be that through the low statistical power no significant influence by ingestion was found. It seems that a part of the differences in germination success between plant species can be explained by differences in seed size. An ecological explanation for this is suggested by Levey and Grajal (1991); small seeds may retain longer in a bird s digestive tract than larger seeds. Longer retention time may cause more abrasion and this might influence the germination ability of seeds. Liu, Platt et al. (2004) found that ingestion by Florida box turtles increased germination rate and germination percentage of large seeds more than that of small seeds. In Contrast to that Braun and Brooks (1987) found no correlation between seed size and seed retention time in box turtles guts. Also for other taxa, no conclusive evidence has been found about the role of seed size shaping the direction of the effects of gut passage on germination success. A meta-analysis for birds and non-flying mammals found that positive effects of gut passage increased with seed size. However, when taking the phylogenetic relatedness into account, gut passage effects were not contingent on seed size but were similar between birds and non-flying mammals (Verdu and Traveset 2004). As we could not measure the retention time of the seeds in the gut of Aldabran giant tortoises we can not reveal whether small seeds stay in the gut longer than large seeds. In addition, no visible signs of strong mechanical abrasion of the seeds were observed (A. Waibel, personal observation). In our study the germination probability (but not percentage germination of germinated seeds) of large-seeded species was relatively more increased by tortoise ingestion. However, this effect was mainly due to a strong increase in germination probability of M. coriacea, the experimental plant species with the largest seed size. M. coriacea and V. merrillii, having the largest seeds showed an increases germination rate after ingestion. Both species have hard seed coats and the ingestion might therefore not have damaged the seeds and thereby not have negatively influenced the germination ability. 20

Discussion 4.2. Effects of tortoise age and identity Several studies have compared the effects of ingestion of different fruigvorous animal taxa on germination patterns and found that the germination of a plant species can be different depending on the animal species that consumed it (Lieberman and Lieberman 1986; Traveset 1998). It was found that the same plant species can respond differently to the same frugivore, depending on environmental conditions (Barnea, Yomtov et al. 1990). However, few is known about intraspecific variability of ingestion effects of the same frugivore, for example whether such effects differ among life history stages. As we used distinguishable plastic pellets as markers we were able to test for individual tortoise effects on the germination of ingested seeds. The findings from this study emphasize the importance of considering intraspecific variability when investigating ingestion effects. The results not only showed that tortoise identity within an age class strongly affected the germination success of ingested seeds, but also the influence of the individual tortoises. These individual effects even varied among the experimental plant species. It was shown that the retention time of food differs among life stages of tortoises. Hatt, Gisler et al. (2002) found that the food phase of Galapagos tortoises differed in retention time for adult and juvenile tortoises (12 and 9 days). In a study with A. gigantea body mass affected food intake and transit time (Hamilton and Coe 1982) and Tracy, Zimmerman et al. (2006) showed longer food retention time in juvenile desert tortoises compared to hatchlings. Unfortunately, in our study it was not possible to measure the exact retention time of the seeds in the tortoises digestive tract. Nevertheless rough estimates during field work indicate that the retention time of food was longer for adult tortoises. Sub-adult tortoises started to excrete the experimental seeds after approximately eight days whereas it took at least 12 days for the seeds to be defecated by adult tortoises (A. Waibel, pers. observation). In contrast to our findings, it was found that increased retention time of a chilli species in avian guts was positively related to germination percentage (Tewksbury, Levey et al. 2008). Our findings suggest that germination success of seeds ingested by sub-adult tortoises was higher than that of seeds ingested by adult tortoises. It could be that the observed increased retention time damaged seeds and thus reduced the germination ability. Our analyses suggest that not only age but also intra-age-specific individual effects alter the germination success. The tortoises used in this experiment were classified into two age classes: adults and sub-adults. But within these two classes size and age varied among the individuals. Especially the age and size of the adults differed (est. between 60 and 120 years 21

Discussion old, 75 180 kg). The sub-adults were all of similar age and size (between 10 15 years old, 20 30 kg). The morphological differences indicate that the gut of the tortoises is different as well and therefore influences the passing seeds in different ways. However, it is unlikely that the variation among the sub-adult tortoises can be related to size and age as the differences concerning this were relatively small. The food eaten by the individual tortoise might have influenced the experimental seeds in the guts as well. Additionally to the experimental fruits, the tortoises were fed with plant material, in the same way they are fed throughout the year. The same additional food was offered to adult and sub-adult tortoises. But the individual tortoises were observed to have preferences and not feed equally on the additional food. To test the individual tortoise effects on the germination of ingested seeds, distinguishable plastic pellets were fed the tortoises as markers. To our knowledge, markers have so far, only been used to determine food retention time in animals guts. N-alkanes were used by Hatt, Gisler et al. (2002) to determine mean retention time in Galapagos tortoises. And transit time in desert tortoises was measured with the help of plastic markers (Meienberger, Wallis et al. 1993). The practice of using plastic pellets to assign the droppings towards individual tortoises could be realized relatively easy during field work, but also raised problems. The feeding of the pellets was handy as the tortoises swallowed them unsolicited together with fruits like banana or papaya. They were clearly visible and occurred together with the experimentally fed seeds in the faeces. Anyway it was not possible to unambiguously assign each dropping to a tortoise. Especially in the enclosure of the sub-adults trampling took place. The sub-adults were closer together as the adults and trampled the faeces over time (even checking the enclosure for faeces several times a day did not prevent the trampling). Thus some faeces were mixed and could not be unambiguously assigned to an individual tortoise. Trampling was no problem in the enclosure of the adult tortoises as it was a lot bigger and hosted only seven tortoises during the experiment. Another problem was that not all droppings contained plastic pellets. Faeces that could not be unambiguously assigned to an individual tortoise were used for the mixture treatment. It was possible to find droppings of each tortoise containing the experimental seeds but it was not always enough for planting the three replicates. Despite theses difficulties this method seems to be a promising approach to study intraspecific individual effect. Therefore it can be recommended for further research having the possibility to keep animals in big enclosures. 22

Conclusion 5. Conclusions and conservation implications Alien, invasive plant species are a mayor threat to native ecosystems worldwide. Research with birds has shown that ingestion can enhance germination and dispersal to suitable habitats can support the spreading of invasive species (Vila and D'Antonio 1998; Bartuszevige and Gorchov 2006). In Mauritius the introduced Red-Whiskered Bulbul spread out the seeds of invasive plant species. Lingustrum robustum and Clidemia hirta germinate better after ingestion and the Bulbul carried the seeds to non invaded habitats (Linnebjerg, Hansen et al. 2009). Also cattleianum, which is highly invasive to Mauritius, was spread by the Red- Whiskered Bulbuls. Psidium cattleianum is much less pernicious in Rodrigues because the Bulbul was not introduced there (Cheke and Hume 2008). To our knowledge, only two studies have investigated gut-passage effects of A. gigantea. Both focused on the effect of ingestion on the germination of plant species endemic to Mauritius (Hansen, Kaiser et al. 2008; Moolna 2008). The germination success differed after the ingestion of Aldabran giant tortoises. The results of Moolna indicate enhanced germination success for Diospyros egrettarum. Hansen et al suggests that Syzygium mamillatum could benefit from the tortoise ingestion as the seeds would be dispersed. The germination success of alien and native plant species on Mauritius after ingestion by skinks was compared by N. Zuël (2009). He found that native plant species had a better germination after gut passage than alien ones. As skinks, tortoises feed on various plant species (native and aliens) thus this study focused on the effect of gut passage of Aldabran Tortoises on alien plant species on Rodrigues. The results of this study can be of importance for further introduction of Aldabran giant tortoises as an analogue species on the Mascarene Islands. How Aldabran tortoises would disperse seeds over the island can not be deduced from this study, but observations throughout the experiment indicate that the chosen species are likely to be eaten by the tortoises in the wild. Dodd (2001) found that bright coloured fruits are preferred by box turtles and they selected red and orange fruits. The tortoises of this study obviously liked to feed on the fruits of V. merrillii (red), W. indica (orange) and M. coriacea (yellow). Also S. cumini was favoured; its smell might have been attractive for the tortoises. Considering the status of invasion, it seemed to be important to include L. camera in the experiment, to find out whether Aldabran giant tortoises would feed on the fruits and how ingestion would influence its germination. Even though the results show that tortoises enhance the germination of L. camera it is not very likely that they would feed on it in the wild. The fruits were not eaten voluntary and had to be mixed with other fruit for 23