Abstract
In recent years, it has become clear that frugivory and seed dispersal (FSD) by turtles and tortoises is much more common than previously thought. Yet, a review and synthesis is lacking. We here review published and unpublished records of chelonian FSD, and assess the role of chelonians as seed dispersers, from individual species to the community level. We first discuss the distribution of chelonian FSD and the characteristics of the fruit and/or seed species eaten and dispersed by chelonians. We then use the seed dispersal efficiency framework to explore the quantitative and qualitative components of seed dispersal by tortoises and turtles, embarking on a journey from when the fruits and/or seeds are consumed, to when and where they are deposited, and assess how efficient chelonians are as seed dispersers. We finally discuss chelonian FSD in the context of communities and chelonians as megafauna. We found that a substantial proportion of the world’s aquatic and terrestrial turtles and a major part of testudinid tortoises (70 species in 12 families) include fruits and/or seeds in their diet, and that furits of at least 588 plant species in 120 families are ingested and/or dispersed by chelonians. For some chelonians, overall or in certain seasons, fruit may even form the largest part of their diet. Contrary to seed dispersal by lizards, the other major reptilian frugivores, chelonian FSD is not an island phenomenon in terms of geographic distribution. Nevertheless, on islands especially tortoises are often among the largest native terrestrial vertebrates—or were, until humans got there. We synthesize our knowledge of chelonian FSD, and discuss the relevance of our findings for conservation and restoration, especially in relation to rewilding with large and giant tortoises.
Resumen En años recientes, se ha hecho claro que la frugivoría y dispersión de semillas (FDS) llevada a cabo por tortugas (quelónidos) es más común de lo antes pensado. No obstante, todavía carecíamos de una revisión y síntesis sobre este tema. En este artículo, revisamos récords (publicados y no publicados) sobre FDS por quelónidos, y evaluamos su rol como dispersores de semillas, desde el nivel de individuos, al nivel de comunidades. Primero, discutimos la distribución de FDS por quelónidos, y las características de las especies de frutos y/o semillas consumidas y dispersadas por tortugas. Luego hacemos uso del concepto de la eficiencia de dispersión de semillas como marco de referencia para explorar los componentes cualitativos y cuantitativos de la FDS por quelónidos, embarcándonos en un viaje desde cuando los frutos y/o semillas son consumidas, hasta cuando son depositadas. También evaluamos cuán eficientes son los quelónidos como dispersores de semillas. Finalmente procedemos a discutir la FDS por quelónidos en el contexto de comunidades, y como ‘megafauna’. Encontramos que una proporción substancial de las tortugas acuáticas del mundo y la mayor parte de las tortugas testudínidas (70 especies en 12 familias) incluyen frutos y/o semillas en su dieta que abarcan al menos 588 especies de plantas en 120 familias. En algunas especies, en general o en algunas estaciones, la mayor parte de su dieta está conformada por frutas y/o semillas. Más importante aún, y contrario a las lagartijas, que son otro grupo importante de reptiles que incurre en FDS, la frugivoría y dispersión de semillas por quelónidos no es un fenómeno de islas solamente, en términos de distribución geográfica. Empero, en islas, especialmente las tortugas terrestres, están entre los vertebrados nativos de mayor tamaño–o lo estuvieron, hasta que los humanos llegaron a ellas. En este artículo, hacemos una síntesis de las lecciones aprendidas hasta ahora sobre la FDS por quelónidos, y discutimos la relevancia de nuestros hallazgos para la conservación y restauración, especialmente en relación a proyectos de resilvestrar (‘rewilding’) con tortugas gigantes o de gran tamaño.
I. Introduction
Animal-mediated seed dispersal is the process by which animals disperse the seeds away from the mother plant (Fig. 1), and is an important ecological function that has profound ecological and evolutionary implications in ecosystems (Howe & Smallwood, 1982; Rezende et al., 2007; Stoner & Henry, 2008). The distribution and ecology of frugivory and seed dispersal (FSD) in most major vertebrate taxa has been thoroughly investigated and results synthesised (Estrada & Fleming, 1986; Levey, Silva, & Galetti, 2002), most recently for lizards (Iverson, 1985; Olesen & Valido, 2003; Valido & Olesen, 2007; Whitaker, 2011), and a start has even been made for crocodilians (Platt et al., 2013). However, a thorough overview and synthesis is still missing for chelonians.
Reviewing the origin and rise of frugivory and seed dispersal through deep time, Tiffney (2004) established that plants had the necessary morphological features for vertebrate dispersal by the Late Carboniferous (323.2–298.9 Ma), and that by the middle of the Mesozoic (252–66 Ma), several reptile lineages could have established specific FSD associations with plants. Given the long evolutionary history of chelonians, and the generally broad diet of many extant chelonians could have been among the early dispersers and ‘first movers’ in the evolutionary ecology of fruits (Ridley, 1930; van der Pijl, 1969; Tiffney, 1986; 2004). Perhaps the earliest example of frugivory by chelonians comes from a Campanian (83.6–72.1 Ma) coprolite that likely originated from a turtle, and that contained ca. 200 achenes of a Ranunculaceae sp. (Rodriguez-de la Rosa, Cevallos-Ferriz, & Silva-Pineda, 1998). Two fossilised specimens of Stylemys tortoises from the Oligocene (33.9–23.03 Ma) in South Dakota contained hackberry (Celtis) seeds (Marron & Moore, 2013). On the Bahamas, two out of three extremely well-preserved individual carapaces of the recently extinct (4,200–1,200 BP) giant tortoise (Chelonoidis alhuryorum) contained many seeds of two large-fruited species (wild mastic, Mastichodendron foetidissimum, and satinleaf Chrysophyllum oliviforme; both Sapotaceae) (Steadman et al., 2007; Franz & Franz, 2009). In historical times, now-extinct giant tortoises (Cylindraspis spp.) of the Mascarene Islands were observed by early settlers to include fruit in their diet. In Mauritius in the late 1600s the tortoises were reported to eat ‘apples’ (= endemic ebony Diospyros, Ebenaceae, and Sapotaceae fruits) (Hume & Winters, 2016). On nearby Rodrigues Island, the exiled French Huguenot François Leguat and his men ate many fruits from the forest, but “left the dates [= palm fruits, Arecaceae] for the turtles [= giant tortoises, Cylindraspis spp.]” (Leguat, 1708).
One of the first modern, experimental FSD studies was Rick & Bowman’s (1961) classic paper on how the germination rate of an endemic Galápagos tomato was dramatically improved by passing through the gut of the endemic giant tortoises. Additionally, Hnatiuk’s (1978) study of germinating seeds from the feces of Aldabra giant tortoises, and Iverson’s (1987) discussionof the likely frugivore mutualistsof the highly specialized Tambalacoque tree (Sideroxylon grandis; Sapotaceae), are two early examples of seminal thinking about the potential for seed dispersal interactions and germination enhancement of tortoises and island plants. It is thus ironic that, despite several calls for studies of turtles as seed dispersers (e.g., Moll & Jansen, 1995; Pérez-Emán & Paolillo, 1997), our understanding of chelonian FSD has progressed very little since then. In this review, we aim to summarise published and unpublished information about chelonian FSD in the wild, and synthesise and discuss the role of chelonians as frugivores and seed dispersers. We first present an overview of the taxonomical distribution of chelonian FSD, as well as of the taxonomical distribution of plants consumed by chelonians. We then use the concept of seed dispersal effectiveness (SDE) (Schupp, 1993; Schupp, Jordano, & Gómez, 2010) to discuss the quantitative and qualitative aspects of chelonian seed dispersal. We progress to discuss the function of chelonians as megafaunal seed dispersers in the FSD community, and their role in conservation and restoration efforts.
II. Methods and data
To synthesise data on FSD by chelonians, we performed a comprehensive literature search that included scientific articles, books, monographs, and theses. We used Google Scholar (http://scholar.google.com/; Google Inc.), as it has been found to include and exceed the results of other commonly used literature databases (specifically WoS and Scopus; see Svenning et al., 2016). We used the following search terms: ‘diet’, ‘frugivory’, ‘seed dispersal’, in combination with the Latin genera of chelonians (from van Dijk et al., 2014), or the keywords ‘chelonian’, ‘tortoise’, or ‘turtle’. No constraints on the year of publication or language were imposed (i.e., we found some articles in other languages, e.g., Spanish). We filtered the search results by reading the abstracts, and also went through the references of each text found to identify other potentially suitable articles. We added literature known by the authors to include diet information, but which did not appear in our search (mostly books). In addition, we added unpublished data based on our own observations and those shared by various researchers. For diet data, we only included information based on wild chelonians. For germination and gut passage experiments we included studies using captive chelonians conducted with fruits found in the natural habitat of the species. To give a more complete overview of some of the main variables that determine the outcome of seed dispersal, we reviewed information on gut retention time (GRT), and on movement ecology and habitat range of chelonians. We used the same approach as above, using each of the search terms, ‘gut retention time’, ‘movement’, ‘activity’ and ‘home range’ together with ‘tortoise’, ‘turtle’ or ‘chelonian’. See Supplementary Materials S1 for the resulting reference lists.
We followed van Dijk et al. (2014) for chelonian taxonomy, and the iPlant Collaborative for plant taxonomy (Boyle et al., 2013; http://tnrs.iplantcollaborative.org/). We used the amniote life-history database for data on chelonian body mass (Myhrvold et al., 2015). When studies only showed results graphically, we extracted the data from figures using WebPlotDigitizer ver. 4.0 (Rohatgi, 2017; https://automeris.io/WebPlotDigitizer/). We analysed and visualised the data using R v. 3.3.3 (R Core Team, 2017) and the package ‘ggplot2’ (Wickham, 2016).
We were able to extract data from a total of 167 studies on chelonian FSD, germination, GRT, and movement. We found a total of 106 studies containing data on FSD by wild chelonians. These arose from either focused FSD studies (i.e., studies focusing directly on the role of chelonians as frugivores and/or seed dispersers; n = 24), partial FSD studies (i.e., studies that examine diet in relation to/in a framing of FSD or examine gut passage, but not germination; n = 70), or diet studies (not framed in an FSD context; n = 12). The studies used several methods to obtain data on FSD by chelonians, including direct observation, camera traps (e.g., Wang et al.,2011), stomach flushing (e.g., Legler, 1977), or analysis of collected faeces (e.g., Nogales et al., 2017). Faecal collection methods ranged from simple picking up, to more creative approaches, such as collection with a miniature wheeled barrow mounted behind the animal (Josseaume, 2002), and, for marine turtles, collecting in cloaca-mounted bags (Amorocho & Reina, 2008). Determination of the seed content in the faeces was done with either direct counts of seeds, or counting any seeds that germinated from the dung (e.g., Hnatiuk, 1978). For chelonian GRT, we found 37 studies, which were conducted by feeding fruits and/or artificial particles. Finally, we found 24 studies on chelonian movement and home ranges.
There are inherent biases associated with the different methodologies when estimating chelonian FSD. In dietary studies, seeds might often be overlooked, or underreported/not specifically mentioned as plant diet components. For example, Mouden et al. (2006) have a long list of plants recorded in scat of the spur-thighed tortoise (Testudo graeca), many, but not all of which, overlap with those of Cobo & Andreu (1988), who specifically studied seeds dispersed by T. graeca. Also, Kabigumila (2001) and Hansen, Johnson, & Van Devender (1976) provide a long list of food plants found in the scat of the leopard tortoise (Stigmochelys pardalis) and the Mojave desert tortoise (Gopherus agassizii), respectively, but they did not specify whether these were fruits, seeds or other plant parts. Faecal analysis alone may provide a biased account of a species’ diet. For example, de Lima Magnusson, & da Costa (1997) describe the red side-necked turtle (Rhinemys rufipes) as a major frugivore “palm specialist” based on faecal analysis, but a subsequent study by Caputo & Vogt (2008), using stomach flushes, found relatively larger amounts of animal food items. Thus, faecal analysis tends to record more plant matter, while it can grossly underrepresent the importance of animal matter in the diet (Caputo & Vogt, 2008). However, using only stomach flushing may underestimate frugivory, as large seeds are hard to dislodge (Kennett & Tory, 1996; de Lima et al., 1997). A combination of both approaches, where possible, would seem to be ideal. Another aspect which can bias the available data is the seasonality in the diet of some chelonians, where fruit may only be a major part of the diet in some season(s). Short-term studies that do not span different seasons may underestimate fruit consumption and thus the potential for seed dispersal. This is important to into consideration, because only 9% of the studies we found considered seasonality in the diet of chelonians. All these factors underscore the need for a more comprehensive dietary sampling when considering the feeding type of chelonians and their role as frugivores and seed dispersers.
III. Distribution of chelonian FSD
(1) Taxonomical distribution
Chelonians comprise about 335 species, of which 275 are turtles and 60 are tortoises, spanning 94 genera in 14 families (van Dijk et al., 2014). We documented FSD in a total of 72 species with, distributed across all major chelonian phylogenetic groups (Table 2; except for Dermochelyidae, with the marine leatherback turtle, Dermochelys coriacea, as the only extant species).
There was a notable gap in FSD in the branches containing Platemys platychephala to Acanthochelys spp. (Chelidae), Pelochelys spp. and Chitra spp. (Trionichydae), and containing from Orlitia borneensis to Pangshura smithii(Geomydidae). However, FSD was recorded in other species within these three families. This pattern is likely due to the lack of focused dietary or FSD studies on these species. Moreover, as we will see below, habitat and seasonal influences on the diet of these groups may influence the levels of FSD in different locations and times of the year, therefore affecting sampling results. The few other chelonian species without any reported FSD have been described as purely carnivorous. Thus, frugivory is widespread in Testudines, with most taxa having at least one frugivorous representative at the genus level.
(2) Geographical distribution
Chelonians are widely distributed across the world, inhabiting habitats from tropical to temperate, from continents to islands and oceans, and they include terrestrial, aquatic and semi-aquatic, as well as marine species (see van Dijk et al., 2014 for individual species distributions). Chelonian species richness peaks in the southeastern USA, the Ganges Delta, Southeast Asia, and northern South America (Fig. 2a; data from Roll et al., 2017, provided by Y Itescu). Thus, unlike FSD by lizards (Olesen & Valido, 2003), FSD by chelonians is not restricted to islands, and they can thus potentially play a major role in continental and island ecosystems alike. The geographic distribution of species richness of chelonian species that engage in FSD is concentrated in the south-eastern USA and northern South America, highlighting the underrepresentation of studies for especially south-east Asia (Fig. 2b).
(3) Patterns of generalisation and specialisation
Specialisation or generalisation on fruits varies depending on the chelonian species, as expected by the different main feeding types (herbivores, carnivores or omnivores). Frugivorous tortoises can vary from generalist, specialist to opportunistic frugivores. For example, Chelonoidis tortoises in South America are generalist frugivores, consuming fruits having a variety of traits (Moskovits, 1985; Guzman & Stevenson, 2008). At the opposite end of the spectrum, we have the highly specialised Gibba turtle (Mesoclemys gibba [Phrynops gibbus]) that has been found to feed almost exclusively on palm fruits (Mauritia flexuosa, Aracaceae) during part of the year in the Rio Negro Basin in Brazil (Caputo and Vogt, 2008). Other species such as the common snapping turtle (Chelydra serpentina) are omnivorous, incorporating roughly the same amount of plant material (including fruits) and animal material in their diet (Ernst & Lovich, 2009). Lastly, there are species that are mostly carnivorous, which will eat fruits opportunistically, such as Blanding’s turtle (Emydoidea blandingii) (Rowe, 1992) and hinge-back tortoises (Kinixys spp.) (Luiselli, 2003). Overall, most frugivorous chelonian species are generalist frugivores that also include other plant material in their diet; this is especially true for tortoises (Testudinidae).
(4) Functional traits in relation to FSD
Frugivore species have inter- and intraspecific differences in functional traits, such as habits, size and age, sex, cognition and preferences, which may result in large differences in the seed dispersal services they provide (Jordano et al., 2007; Zwolak, 2017). Knowledge about these traits will help us understand the role of specific characteristics of frugivores in their effectiveness as seed dispersers.
(a) Habitat
Chelonians are a diverse group of vertebrates whose different habits, such as terrestrial, semi-aquatic, aquatic and marine, have allowed them to exploit many habitats and resources. Terrestrial plants and tortoises are perhaps the first that come to mind when thinking about seed dispersal in this group. With their preponderance of fleshy fruits, terrestrial plants could be considered more zoochorous than their aquatic counterparts, and because of their habitat, tortoises and terrestrial –or semi-aquatic–turtles are more likely to encounter fruits and disperse their seeds within terrestrial habitats. However, as we found, seed dispersal is also carried out by mainly aquatic species, both on land and in water, for terrestrial and aquatic plants, and even for coastal and marine plants in marine ecosystems. Yet, most studies of chelonian FSD have focused on terrestrial chelonians andlargely ignored the role of aquatic and marine species in seed dispersal (Moll and Jansen 1995). Ultimately, the habitats of both chelonians and of the plants they encounter will determine which fruits are available to each species, and whereto the seeds can be dispersed.
(b) Size and age
Tortoises and turtles exhibit great inter- and intraspecific size variation. Size generally increases with age in chelonians (Carr, 1952; Ernst and Lovich, 2009). From the perspective of FSD, the size of chelonians limits the size and the number of fruits and/or seeds they can swallow and pass through their guts. Furthermore, size may affect gut passage time (see section on mouth and gut passage treatment) and volume of the scat. Thus, size is expected to substantially affect the ability and effectiveness of chelonians as seed dispersers (see also section on chelonians as megafaunal seed dispersers).
Ontogenetic changes in diet may also occur in chelonians, with vegetation becoming more important as chelonians age and become larger (Moll, 1976); this seems to be common in omnivorous turtles (Clark & Gibbons, 1969; Georges, 1982; Hart, 1983; Sung, Hau, & Karraker, 2016). In the case of the omnivorous red sidenecked turtle (Rhinemys (Phrynops) rufipes), de Lima et al. (1997) found that most of the scat volume was palm seeds, and that the frequency of palm seeds increased with turtle size. These ontogenetic changes in diet may be accompanied by changes in gut morphology, as found in the green sea turtle (Chelonia mydas), with the ratio of long to short intestines increasing from 0.45 in post-hatchlings to 2.5 in adults, possibly reflecting a higher proportion of animal matter in the diets of young individuals (Davenport, Antipas, & Blake, 1989).
(c) Sex
Sexual dimorphism is common in chelonians, but the direction of sexual dimorphism depends on the species and even on habitat. For example, males of angulate tortoises (Chersina angulata) are larger than females, whereas females of leopard tortoises (Stigmochelys pardalis) are larger than males within the same habitat (Mason et al., 2000). In the case of Aldabra giant tortoises (Aldabrachelys gigantea), the population exhibits no sexual dimorphism on the east of Aldabra Atoll, but males gradually attain larger sizes compared to females towards the western side of the atoll (Turnbull et al., 2015). As described above, size is expected to have a differential effect on seed dispersal, and there may thus be differences in the seed dispersal provided by males and females, respectively. For example, where sexual dimorphism is present, the larger males of Aldabra giant tortoises are able to extend their necks to reach higher vegetation and fruits than the smaller females can (WF & DMH, pers. obs). Males and females may also exhibit different behaviours, e.g., habitat selection, which can affect the outcome of FSD. For example, most of the stomach contents of the omnivorous female smooth softshell turtle (Apalone mutica) were aquatic items, whereas stomach content of males was mostly terrestrial items and included more fruits (Plummer & Farrar, 1981). These sexual differences in terms of diet were attributed to the different microhabitat preferences (females forage in deep water, whereas males forage in the interface between aquatic and terrestrial habitats). Furthermore, males and females of some species may show differences in home range size and displacement distances (see below). Difference in habitat selection, home range size and displacement distances are not only expected to affect the ability of chelonians to exploit fruits, but also their effectiveness as seed dispersers.
(d) Cognition and behaviour
Chelonians, as other animals, rely on cognitive processes to acquire knowledge about their environment through their senses, leading to learning and memory creation. The sensory features of fruits play an important role in attracting frugivorous birds and aid in their selection (Schaefer, Spitzer, & Bairlein, 2008b), and this is expected to be the case for chelonians as well. Sight and olfaction are the sensory faculties that aid turtles and tortoises in the recognition of food sources. Chelonians can perceive images and distinguish colours in the human-visible spectrum (Granda & Stirling, 1965; Baylor & Fettitplace, 1975; Schwartz, 1975; Neumeyer & Jäger, 1985; Arnold & Neumeyer, 1987; Ammermüller, Muller, & Kolb, 1995; Ventura et al., 2001; Twig & Perlman, 2004; Mathger, Litherland, & Fritsches, 2007; Pellitteri-Rosa et al., 2010), and some have been shown to also have sensitivity to the ultraviolet spectrum (Ammermüller et al., 1998; Ventura et al., 1999; Zana et al., 2001). Additionally, chelonians have a highly developed olfactory (vomeronasal) system (Manteifel, Goncharova, & Boyko, 1992; Murphy, Tucker, & Fadool, 2001; Fadool, Wachiowiak, & Brann, 2001), which they can use to detect volatile chemicals excreted by plants from long distances (King, 1996), and also to smell fruits at close range, possibly to evaluate ripeness (WF, DMH, DM, pers. obs.).
Learning and memory of frugivores has an important impact on seed dispersal, because decision-making based on previous experiences can determine which plants and which fruits are selected and consumed, and ultimately where seeds are dispersed (reviewed in John et al., 2016). A model by John et al. (2016) testing frugivores with different spatial memory skills suggested that those with longer spatial memory are able to relocate food sources more efficiently, survive longer and disperse larger amounts of seeds. The also moved less at random around the landscape, which led to differences in terms of the spatial distribution of seeds dispersed compared to animals with shorter memory. Captive red-footed tortoises (Chelonoidis carbonarius) can navigate efficiently in their environment, and they can remember spatial location of food for at least two months (Soldati, 2015). Moreover, they were able to anticipate food availability over periods of 24 h, discriminating between the quality and quantity of food, and remembering these attributes for at least 18 months. In relation to large-scale movement patterns, individuals of both Galápagos (Blake et al., 2013) and Aldabra giant tortoises (Baxter 2015) have been shown to follow the same movement patterns in different years, implying that they have a persistent spatial memory.
Chelonians may use landmarks and different stimuli to orient themselves and find suitable food sources. For example, when tested in a T-maze, sulcata tortoises (Geochelone sulcata) and leopard tortoises (Stigmochelys pardalis) could discriminate between colours and shapes, and associate these features with navigation to food sources (Janisch, 2013). Red-footed tortoises (Chelonoidis carbonarius) can navigate between known localities where fallen fruits are available at certain seasons (Josseaume, 2002). Also, fallen fruits encountered are often from foraging activity of arboreal/aerial frugivores (Moll & Jansen, 1995), and it is thus possible that chelonians can use cues from other species to find food. This seems to be the case in in Malaysia, where painted terrapins (Batagur borneoensis) have been observed clustering in the water under a troop of leaf monkeys in trees above to eat berembang fruits (Sonneratia caseolaris, Lythraceae) that the monkeys were throwing into the water (Moll, 1980b).
By navigating the landscape based on previous experiences, chelonians can identify and exploit fruits. For example, (Legler, 1976) noted that the gulf snapping turtle (Elseya lavarackorum) in Australia exploits windfall fruits of fig trees, with large congregations of these turtles found around this resource. In addition, other aquatic species such as the black river turtle (Rhinoclemmys funerea) (Moll & Jansen, 1995) and the Central American river turtle (Dermatemys mawii) (Moll, 1989) have been observed clustering and waiting in water beneath fruiting Ficus glabrata(Moraceae) trees, and the painted terrapin (Batagur (Callagur) borneoensis) displays similar clustering in the water under falling berembang fruits in Malaysian rivers (Moll, 1980b). Similarly, the Travancore tortoise in India (Indotestudo travancoria)(Bonin, Devaux, & Dupré, 2006; Kanagavel & Raghavan, 2012), and in Brazil, the redfooted tortoise (Chelonoidis carbonarius) (Moskovits & Bjorndal, 1990) congregate beneath favoured fruiting trees to exploit these food resources. Notably, the tree Spondias testudinis (Anacardiaceae) was named for the yellow-footed tortoises (Chelonoidis denticulatus) (Mitchell & Daly, 1998) that flock beneath fruiting trees to feed on the large, yellow-brown fruits (D Daly, pers. comm.). Furthermore, aquatic Antillean sliders (Trachemys decussata) in Cuba will emerge onto the land in great numbers after rains to feed on fallen jobo (Spondias lutea) and Bagá(Anona palustris, Annonaceae) fruits that have fallen from riparian trees (Barbour & Carr, 1940). Thus, chelonians possess a landscape-scale spatial awareness of plants providing fruits.
(e) Fruit preferences
Animals rely on their ability to detect differences in food quality by using sensory adaptations, which allows them to circumvent some of the costs associated with foraging (Borges et al., 2011). Frugivores can establish and maintain preferences based on colour, odour and taste (Sorensen, 1983; Levey, 1987; Willson, Graff, & Whelan, 1990; Willson & Comet, 1993). As discussed above, chelonians have highly developed visual and olfactory systems, and are known to be attracted by smell and colour (see Harless & Morlock, 1979 for a review), which may lead to the establishment of preferences. Indeed, studies focusing on colour preferences in tortoises have found preferences for distinct visual stimuli. Probably the first study that explored colour preferences in chelonians was done by Grant (1960) on Texas tortoises (Gopherus berlandieri), which exhibited a strong preference for red, selecting food items dyed red after having initially rejected them (i.e., when the same food items had other colours). Subsequent studies using spur-thighed tortoises (Testudo graeca) (Pellitteri-Rosa et al., 2010), yellow-footed tortoises (Chelonoidis denticulatus) (Passos, Santo Mello, & Young, 2014), and Aldabra giant tortoises (Aldabrachelys gigantea) (Spiezio, Leonardi, & Regaiolli, 2017; DMH, unpublished) have shown a prevalent preference for yellow, and/or red colours. Furthermore, chelonians have been shown to discriminate between odours to identify potential mates and conspecifics (e.g., Auffenberg, 1965; Galeotti et al., 2009; Polo-Cavia, López, & Martín, 2009), and they also use scent to find food items (Germano et al., 2014). Although chelonians are also known to discriminate shapes (Janisch, 2013), we did not find any studies examining food or fruit shapes as visual stimuli, nor did we find any studies on taste discrimination.
Plants are known to employ visual and scent cues to signal ripeness in fruits to attract seed dispersers, which use these cues to assess their nutritional value (Brady, 1987; Kalko, Herre, & Handley, 1996; Schlumpberger, Clery, & Barthlott, 2006; Schaefer, McGraw, & Catoni, 2008a). Unripe fruits often have chemical compounds that make them unpalatable to seed dispersers (Sherburne, 1972; Schaefer, Schmidt, & Winkler, 2003), who may learn to associate visual and scent cues with unpalatability. Therefore, different colour and smell preferences may ultimately lead to distinct preferences for certain fruit traits. For example, many fruits are green when unripe, and yellow or red when ripe, and the ripening process is usually accompanied by the release of scents. Consequently, we can expect chelonians to have different preferences for different fruit species, be able to discern between ripe and unripe fruits, and show a preference for ripe ones, especially those that become yellow and red.
The degree to which chelonians act as valid seed dispersers rather than only as frugivores depends on the selection of fruits with viable seeds (usually ripe). Moskovits & Bjorndal (1990) showed that the red- (C. carbonarius) and yellowfooted tortoises (C. denticulatus) preferred fruits over other food items, and preferred fruits that were predominantly red or yellow and were fragrant while rejecting unripe fruits. Chelonians have been observed smelling ripe and unripe fruits at close proximity before eating or apparently rejecting them. For example, this behaviour has often been observed in Aldabra giant tortoises (WF and DMH, pers. obs.; Fig. 1c). Similarly, the eastern box turtle (Terrapene carolina carolina)seems to be able to distinguish between ripe and unripe fruits, preferring the ripe ones (Allard, 1948). However, it should be noted that Hermann’s tortoises (Testudo hermanni) consumes unripe green fruits of Ruscus aculeatus (Asparagaceae) when seasonally available (Del Vecchio et al., 2011), probably limiting their effectiveness as seed dispersers.
The only experimental study that we are aware of that simultaneously evaluated the perception of colour, olfaction and taste was by Grant (1960), studying the Texas tortoise. He proposed, based on feeding trials, that vision, olfaction, and taste, in that order, were used to by the tortoises to select food items. Thus, rather than just relying on one or the other, chelonians use sight and olfaction and taste to discriminate between possible food sources (Grant, 1960; Fitch, 1965; Pellitteri-Rosa et al., 2010), and when fruits and seeds become available in their habitat, they are probably effective at finding them (Moll & Jansen, 1995).
It is worth mentioning that, although it is often assumed that tortoises benefit from seed dispersal interactions with plants by obtaining food resources, it appears that some chelonians do not derive nutritive benefits from these interactions. For example, alligator snapping turtles (Macrochelys temminckii) eat many acorns (such as those of the willow oak, Quercos phellos), and significantly enhance their germination (Elbers et al. 2011). However, the acorns pass through the gut seemingly unscathed and unchanged, so why are they ingested by the turtles in large numbers? Are they covered by nutritious microorganisms after they have soaked in water and that benefit the turtles, or perhaps act as roughage to help grind up other ingested foods? More research is needed to determine the mechanisms by which chelonians are attracted to seemingly unnutritious fruits and/or seeds, and whether they truly present nutritional or other benefits to these reptiles.
IV Plants eaten and dispersed by chelonians
(1) Taxonomical distribution
Chelonians consume the fruits and/or seeds of a great number of plants, including at least 588 species belonging to 368 genera in 121 families. These plant species are distributed across the phylogenetic tree of angiosperms. These plant species occur in many different habitats, with a variety of growth habits, and possess fruits and seeds with a myriad of traits (see Supplementary Materials S2 for the list of plant fruit and/or seed species consumed and/or dispersed by chelonians). Only 18% of all plant families, however, had more than 10 species whose fruits and/or seeds are consumed and/or dispersed by chelonians (Table 1), with 27% of families represented by only a single plant species.
(2) Modes of dispersal
There are two modes of chelonian seed dispersal: endozoochory (dispersal of seeds through the ingestion of fruits and/or seeds), and epizoochory (dispersal of seeds on external body parts). Of these, endozoochory is by far the most common mode, forming the majority of cases reviewed in our study. It occurs in terrestrial, aquatic, and even in marine ecosystems. During the process of endozoochory, the handling behaviour, gut treatment and location of defecation all affect the ultimate quality of seed dispersal (see below). Epizoochory is a passive way of dispersal where seeds are stuck on the external parts of the animals until they are subsequently dropped, and other than movement away from the mother plant, the fruits or seeds are not affected further. Epizoochory has only been observed in two species of chelonians. The terrestrial Aldabra giant tortoises (Aldabrachelys gigantea), which disperse the sticky seeds of Plumbago aphylla (Plumbaginaceae) that adhere to their carapaces, and secondarily disperse seeds of various plant species on their carapaces after birds defecate on them (e.g., Ficus spp., Moraceae; WF and DH, pers. obs.). In Australia, the aquatic eastern long-necked turtle (Chelodina longicollis), disperses several wetland-associated plants whose seeds lodge on its carapacial algal mats (Burgin & Renshaw, 2008).
(3) Diversity of seeds
The species diversity of seeds potentially dispersed by chelonians varies by chelonian species and/or studies. Overall, frugivorous chelonians covered in our review each potentially disperse a high diversity of seeds, with a mean of 13.0 plant species per chelonian species (± 23.6; range: 1–123; see Supplementary Materials S2 for species dispersed), and for some chelonians fruits and seeds were major parts of their diets. For example, the Gopher tortoise (Gopherus polyphemus) disperses more than 50 species of seeds in pine savannah in the southestern USA (Birkhead, Guyer, & Hermann, 2005). For the big-headed Amazon river turtle (Peltocephalus dumerilianus) fruits and seeds were the most diverse components in the diet, with a total of 19 species found in the stomachs, and with Aracaceae (palm) seeds as the most common ones (Pérez-Emán & Paolillo, 1997). In the northern giant musk turtle (Staurotypus triporcatus), the large seeds of Diospyros digyna (Ebenaceae) comprised 63% of the volume of their stomach contents (Vogt & Guzman, 1988). It should be noted that although careful studies have documented many dry-seeded species dispersed –and potentially dispersed–by chelonians (e.g., Hnatiuk, 1978; Cobo & Andreu, 1988; Milton, 1992; Birkhead et al., 2005), there is likely an underestimation in the amount and diversity of such seed species when compared to fleshy-fruited species due to the difficulty of detection and/or identification.
(4) Plants only/mostly dispersed by chelonians
van der Pijl (1969) suggested that fruits dispersed by reptiles (saurochory) should be coloured, smelly, and borne near the ground or drop at maturity. Although strong FSD relationships have been documented between plants and some chelonians, there is a lack of evidence of any form of coevolution that has resulted in a chelonian seed dispersal syndrome (Herrera, 1985). As mentioned above (see section on preferences), although they may show preferences, chelonians potentially disperse fruits with a wide variety of s sizes, colours, and scents. For example, although they show preferences for certain fruits, Chelonoidis tortoises consume fruits with a variety of colours, including both fragrant and odourless ones (Moskovits, 1985; Guzmán & Stevenson, 2008).
However, certain plants may rely disproportionally on chelonians for seed dispersal. For example, while rodents and birds destroy the seeds of Pandanus aquaticus (Pandanaceae), northern Australian snapping turtles (Elseya dentata)defecate the seeds intact (Kennett & Russell-Smith, 1993). Similarly, European pond turtles (Emys orbicularis) disperse most of the seeds of Nymphaea alba(Nymphaceae) intact, while ducks, coots and fish destroy the seeds after gut passage (Calviño-Cancela, Ayres Fernández, & Cordero Rivera, 2007, and references therein). Wang et al. (2011) found that red-footed tortoises (Chelonoidis carbonarius) may be an important seed disperser of Syagrus flexuosa (Arecaceae), because the seeds were often defecated undamaged but are rarely found at all in the scat of other animals. Furthermore, Moll & Jansen (1995) suggested the black river turtle (Rhinoclemmys funerea) as an important seed disperser of Ficus glabrata (Moraceae) and Dieffenbachia longispatha (Araceae). This turtle is very abundant, practices “windfall” feeding in water under riparian fig trees, and emerges on riverbanks and defecates seeds while on land along shorelines in optimal growing locations for these plants. Tortoises may also be especially important for the dispersal of large-seeded plant species on islands (Heleno et al. 2011; Blake et al. 2012; Falcón 2018), which has important implications at the ecosystem level (see section on chelonians as megafaunal seed dispersers).
Grasslands (composed of grasses and sedges) are an important food source in the diet of different terrestrial chelonians (e.g., eastern Hermann’s tortoise, Testudo hermanni boettgeri) (Rozylowicz & Popescu, 2013). In the case of Aldabra giant tortoises on Aldabra, grasslands are the most preferred habitat (Walton et al.,in review), where the high grazing pressure led to the evolution of a specialised ‘tortoise turf’ plant community, whose seeds they disperse (Merton, Bourn, & Hnatiuk, 1976; Hnatiuk, 1978). For the green sea turtle (Chelonia mydas) in the Great Barrier Reef (Australia), seagrass is an important dietary component and it disperses its seeds (Tol et al., 2017). The only other known seed disperser in the Great Barrier Reef is the dugong (Dugong dugon) (Tol et al., 2017), which is considered vulnerable and occurs in low numbers, so turtles may be more important in terms of quantity. Additionally, the diamondback terrapin is also known to be a seed disperser for the eelgrass (Zosteraceae), a type of seagrass, in the Lower Chesapeake Bay (Tulipani & Lipcius, 2014).
Although chelonians do not necessarily seek for grass seeds per se (but see Kimmons and Moll, 2010, turtles may eat floating grass seeds from water surface), and rather act mainly as herbivores, grasses, sedges and seagrasses in general have traits that facilitate chelonian seed dispersal, and it could be important for the maintenance of such communities (Merton et al., 1976; Hnatiuk, 1978; Tol et al.,2017). As Janzen (1984) puts it, “the foliage is the fruit”, and the role of chelonians as seed dispersers in grass communities is likely to be of great importance in places where they reach high densities and levels of biomass, like in island ecosystems or in some places in Africa, especially when compared to other seed dispersers (e.g., Coe, Bourn, & Swingland, 1979; Branch, 2008). It should be noted, however, that the six-tubercled Amazon river turtle (Podocnemis sextuberculata) seems to be a predator of Poaceae and Cyperaceae seeds in the Amazonas, which constituted 92% of their stomach volume contents (Fachín-Terán & Vogt, 2014). In all cases, proper viability, germination and recruitment studies are necessary to determine whether effective seed dispersal occurs.
V Chelonian seed dispersal efficiency
The ultimate definition of efficient animal-mediated seed dispersal is that a dispersal event results in the successful establishment of new reproducing plant individuals. This, however, is far from always the case, as different frugivore species do not provide the same dispersal services to plants. The seed dispersal effectiveness (SDE) framework (Schupp, 1993; Schupp et al., 2010) provides a way to estimate the contributions of individual dispersal agents to the overall dynamic of plant populations Essentially, it quantifies the number of seeds dispersed by a frugivore multiplied by the probability that a dispersed seed produces a new adult plant. As such, the SDE framework has two components: a quantitative and a qualitative one, which, in turn, have many variables, demographic parameters and subcomponents.
The SDE framework can thus be used as a valuable organising tool to study the ecological and evolutionary implications of seed dispersal. Below we discuss chelonian FSD in the context of the SDE framework.
(1) Quantitative component
The quantitative component of SDE can be reduced to the number of foraging visits a chelonian makes to a fruiting plant multiplied by the number of seeds dispersed for each visit (Schupp et al., 2010). The former can be affected, for example, by the local abundance of both plants and chelonians, and the chelonian’s degree of frugivory, while the latter is influenced by the numbers of fruits and/or seeds handled per visit, handling behaviour, and body size (for body size, see section on chelonian functional traits).
(a) Local biomass and density
Perhaps the most comprehensive work to date on chelonian biomass and density is that of Iverson (1982), who argued that despite the important role that reptiles play in terms of the energetics at the ecosystem level, the study of chelonian abundance and biomass was a neglected subject. He calculated biomass of chelonians based on population density estimates, and analysed those data in terms of habit, habitat, and trophic position. He found that typical values of chelonian biomass are at least one order of magnitude higher than those of other ectotherm species. He also found indications that herbivorous chelonians, which often include fruits as part of their diet, appear to have higher biomass than omnivorous or carnivorous species. Finally, he found that annual production estimates in chelonians (with a maximum of 528 kg ha-1 yr-1) are similar to most other vertebrate groups, except for fishes; and that the maximum biomass for individual tortoise species could be as high as 586 kg ha-1. In terms of density, studies have provided estimates for several species; for example, 0.15–0.31 individuals ha-1, for the highly frugivorous yellow-footed tortoise in the Amazon (Guzmán & Stevenson, 2008), 0.85 tortoises ha-1 for leopard tortoises and 0.12 individuals ha-1 for angulate tortoises in South Africa (Mason et al., 2000).
In some species, chelonian biomass may be higher than that of many classes of larger mammals. For example, Branch (2008) indicated that the leopard and angulate tortoise biomass is about 13% that of all mammalian herbivores in South Africa’s Eastern Cape province, where tortoises can reach high densities (Mason et al., 2000). He posited that this meant that the total biomass of tortoises there almost equalled the combined biomass of kudu, buffalo, eland, and bushbuck, only being exceeded by that of elephants! Moreover, Coe et al. (1979) estimated the biomass of Aldabra giant tortoises to range between 253.42–353.87 kg ha-1 on Aldabra Atoll, which is much higher than that exhibited by large mammalian herbivores on Africa. However, it should be noted that chelonian biomass is limited by different factors, such as habitat type (e.g., in mesic vs. xeric habitats) (McMaster & Downs, 2006), and can differ between co-occurring species (Mason et al., 2000). Nevertheless, in general, we can expect the total numbers of seeds dispersed per hectare per year to be large for chelonians (see section on quantity of seeds dispersed), especially when considering the number of large seeds dispersed (Jerozolimski, Ribeiro, & Martins, 2009).
(b) Degree of frugivory
The degree of frugivory in chelonians varies between species, and within species it can vary at the population and at the individual level. For example, in Mexican giant mud turtles (Stauratypus triporcatus), fruits and seeds were the most important dietary component across two sites in Los Tuxtlas (Mexico), but the occurrence of frugivory ranged from 38–100% between populations, and fruits and seeds represented values between 55–82% of the stomach content volume examined (Vogt & Guzman, 1988). The degree of frugivory can also vary depending on the size of chelonians. For example, Sung et al. (2016) found a positive relationship between the size of big-headed turtles (Platysternon megacephalum) and the occurrence of fruits in their diet. Moreover, diet can vary much over short distances. Another aspect to take into consideration is the changes in diet depending on which habitat chelonians inhabit, and depending on seasons. Geoffroy’s sidenecked turtle (Phrynops (Rhinemys) geoffroanus) may have different diets depending on whether it inhabits clean or polluted rivers (Medem, 1960, cited in Fachín-Terán, Vogt, & Gomez, 1995); Souza & Abe, 2000), and depending on season (e.g., fruits of Myrtaceae and Sapotaceae were only found in its stomach during the season of rising water levels) (Fachín-Terán et al., 1995). Likewise, the Gibba turtle (R. gibbus)has been found to feed almost exclusively on palm fruits (buriti) only during part of the year in the Rio Negro Basin in Brazil (RC Vogt, pers. comm.). Similarly, inclusion of fruits in the diet can shift seasonally in the smooth softshell turtle (Apalone mutica) (Plummer & Farrar, 1981) and the Mexican mud turtle (Kinosternon integrum) (Macip-Rios et al., 2010). In addition, changes in diet can occur at the same location (e.g., a river) over time, as the habitat and food resources change over time (e.g., river changes from clean to polluted) (Moll, 1980a).
(c) Quantity of seeds dispersed
Propagule pressure influences the establishment of plants, and the number of seeds dispersed can thus determine the dynamics of plant recruitment. Studies on chelonians indicate that tortoises and turtles are capable of dispersing a high number and diversity of seeds. For example, in the red-footed tortoise (Chelonoidis carbonarius), Wang et al. (2011) reported that a single scat sample contained high numbers of seeds, ranging from 22 to 765 seeds. Moreover, Lagler (1943) found 11,065 seeds of Nymphaea alba in the digestive tract of one individual of the common snapping turtle (Chelydra serpentina). Combining information on density estimates and information on their diet and seed dispersal ecology, Guzmán & Stevenson (2008) estimated that yellow-footed tortoises disperse 160.70 seeds ha-1 per year.
(2) Qualitative component
The qualitative component of SDE can be reduced to the probability that a dispersed seed survives handling by chelonians in a viable condition (quality of treatment in the mouth and gut) multiplied by the probability that a viable dispersed seed will survive, germinate, and produce a new adult (quality of deposition) (Schupp et al., 2010).
(a) Mouth and gut passage treatment
Lacking teeth, most chelonians tend to swallow fruits and seeds whole (“gulpers”), rather than chewing them as other vertebrate groups do (Moll & Jansen, 1995). They use ‘lingual prehension’, which is the behaviour of using the tongue to touch food items to insert them into their mouths, and this is obligatory for tortoises (Wocheslander, Hilgers, & Weisgram, 1999; Bells et al., 2008). Amphibious emydids and geoemydids use their jaws to grasp food items in terrestrial habitats, a behaviour known as ‘jaw prehension’ (Heiss, Plenk, & Weisgram, 2008; Natchev et al., 2009; 2015). Moreover, and different from birds and monkeys, tortoises do not regurgitate/spit seeds. Thus, damage to seeds by the mouthparts of chelonians was minimal in the studies evaluated. For example, most of the large numbers of seeds of Nymphaea alba (Nymphaceae) found in the digestive tract of the common snapping turtle were mature, and very few of the coats were ruptured (Lagler, 1943). However, some chelonian species can damage seeds with their mouths before gut passage. For example Caputo & Vogt (2008) reported that seeds of several plant species were never recovered whole from stomach flushing in the red side-necked turtle (Rhinemys (Phrynops) rufipes). Similarly, seeds of two species of plants were found crushed inside the stomachs of the giant South American river turtle (Podocemis expansa) (Goulding, 1980).
After consuming the fruits or seeds, they pass to the stomach and through the gut before being defecated. The overall effect on seeds can vary, depending on digestion efficiency and gut retention time (GRT; the time seeds take to pass through the guts until being defecated). Food intake rates may differ among food types in herbivorous chelonians, which have a flexible dietary response, with the ability of switching between cell wall fermentation and extraction of cell contents depending on the diet (Bjorndal, 1989). Moreover, digestive efficiency is inversely related to food intake in tortoises (Meienberger, Wallis, & Nagy, 1993). In some instances, digestive efficiency can depend on the degree of herbivory the species considered, and upon the types of fruits consumed (e.g., in the box turtles Terrapene carolina and T. ornata) (Stone & Moll, 2009), while in others, such as yellow- and red-footed tortoises (Chelonoidis denticulatus and C. carbonarius, respectively), for a given diet, neither digestibility nor mass-specific intake varied between species, and neither did they vary by sex or body mass within each species (Bjorndal, 1989).
Chelonians seem to submit digesta to a similar degree of ‘gut washing’ as mammalian herbivores do (Franz et al., 2011). However, although herbivorous reptiles have similar digestibilities as mammalian herbivores (Bjorndal, 2012), overall chelonians are said to be inefficient feeders because their performance at digesting cellulose is lower when compared to mammalian herbivores, and they need to eat large quantities of food to satisfy their energy demands (Branch, 2008). As a result, plant items in their scat are often recognisable, and seeds often pass undamaged.
Compared to the other vertebrate groups, chelonians have relatively longer GRTs, with a mean of 7.65 days (± 5.89; for all species examined combined, Fig. 3), due to their low metabolic rates and food intake (Stevens & Hume, 2004; Franz et al., 2011). Gut retention times in chelonians may be affected by a myriad of factors. For example, GRT tends to vary across seasons, especially in habitats where there are wet and dry periods (e.g., Aldabrachelys gigantea) (Coe et al., 1979). Temperature also plays a role in regulating GRT, with increasing temperature leading to faster passage (Sadeghayobi et al., 2011). Moreover, GRT depends strongly on fruit species consumed and on overall diet composition (Bjorndal, 1989; Stone & Moll, 2006). For birds, secondary metabolites in fruits are known to affect GRTs (Murray et al., 1994; Wahaj et al., 1998), which is likely the case in chelonians as well. Furthermore, tortoises show variation in their intestinal morphology according to their feeding habits, and the length ratio of large to small intestines is positively related with GRT (Hailey, 1997). Also, chelonians may exhibit selective food retention based on particle size (Hatt et al., 2002), with coarser food being retained for longer (Hailey, 1997). Lastly, chelonians may exhibit antiperistalsis in the large intestine (i.e., contents are carried upwards) (Naitoh, Hukuhara, & Kameyama, 1975), which also likely affects GRT.
Overall, mean GRT seems to increase with species size (Fig. 3), likely due to the increasing length of digestive tracts (Hatt et al., 2002). However, although mean GRT scales with body mass across different tortoise taxa, Franz et al. (2011) reported that this relationship was not significant when looking only at tortoises with body mass > 1 kg. The reported effects of chelonian size on GRT varied by species in the studies reviewed. Body size did not influence GRT in the red- and yellow-footed tortoises (Bjorndal, 1989). When comparing GRT of hatchlings with that of adults of the aquatic Florida red-bellied turtle (Pseudemys nelsoni), Bjorndal & Bolten (1992) reported that although adults were, on average, 250 times larger, GRT was only 1.4 longer when compared to that of hatchlings.
Potentially muddying the waters, studies on the effect of tortoise size on GRT in Galápagos and Aldabra giant tortoises that used different methods yielded different results. Sadeghayobi et al. (2011) found no effect of size on GRT of Galápagos giant tortoises (carapace width range: 0.84–1.53 m) when fed artificial seeds. However, Hatt et al. (2002), using n-alkanes particles as GRT markers, reported that mean GRT was shorter for smaller Galápagos giant tortoises (mass range: 7–38 kg vs. 100–210 kg in adults). Similarly, in Aldabra giant tortoises, Falcón et al. (in revision) reported no effect of tortoise size (mass range: 0.6–104 kg) on GRT (mean GRT 15 days ± 4) when fed artificial seeds, whereas Waibel et al. (2013) reported that sub-adults (20–30 kg) had shorter mean GRT (13 days ± 1) when compared to adult individuals (75–80 kg; 18 days ± 2) when fed fruits of different plants. Thus, other factors such as differences in diet, hydration, food intake and temperature may be more relevant in determining chelonian GRTs within species.
Although seed size can also affect GRT in frugivores (e.g., Fukui, 2003; Figuerola et al., 2010), this does not seem to be the case for chelonians. Braun & Brooks (1987) found that seed size did not influence the GRT of the small, box turtle (Terrapene carolina) when fed fruits of different wild plants found in their habitat. Also, in larger chelonians such as the Chaco tortoise (Chelonoidis chilensis) (Varela & Bucher, 2002), the Galápagos giant tortoise (Sadeghayobi et al., 2011) and the Aldabra giant tortoise (Falcón et al.in revision), seed size does not affect GRT.
Overall, the GRT data suggests that within chelonian species, seeds of different sizes can be dispersed to similar distances.
(b) Seed deposition
After being consumed, fruits and seeds are processed in the gut and transported until they are eventually defecated. The state in which seeds are deposited by frugivores is affected by the combination of the mouth and gut treatments. In general, after handling and passage through chelonian guts, seeds are defecated without pulp, but this can be plant-species dependent as some seeds can pass with little physical change and still be covered with pulp (Rick & Bowman, 1961; Varela & Bucher, 2002; Hansen, Kaiser, & Müller, 2008; Waibel et al., 2013). Within the same species of plants, there may be differences in terms of seed damage depending on the species of chelonian that consumes them (Kimmons & Moll, 2010).
Damage to seeds tends to be minimal after defecation. For example, Rick & Bowman (1961) found that less than 1% of recovered seeds of Solanum cheesmaniae(Solanaceae) showed any signs of damage after gut passage. Similarly, virtually all the seeds of Solanum aldabrense were recovered intact from a single Aldabra giant tortoise scat (WF, pers. obs.). Also, painted turtles (Chrysemys picta) pass 99% of seeds intact (Padgett, Carboni, & Schepis, 2010). In addition, 90% of gut-passed seeds were intact for Chelonoidis carbonarius in the Pantanal (Wang et al., 2011). Moreover, most seeds were intact for after gut passage in C. denticulatus in the Brazilian Amazonia (Jerozolimski et al., 2009). Even for soft seeds without endocarp, like Syzygium mammilatum (Myrtaceae), substantial amounts of seeds survive gut passage undamaged (Hansen et al., 2008). As a result of the minimal damage experienced by seeds after chelonian gut passage, many of them remain viable. For example, studies reported between 90–100% of viability of seeds in the faeces of red-footed tortoises (Strong & Fragoso, 2006; Wang et al., 2011).
The location of seed deposition, and perhaps especially the distance from the source, are two key factors for determining what happens to seeds after defecation. This is largely affected by the frugivores’ movement ecology in combination with the GRTs. Only very rarely have chelonian FSD studies specifically included movement ecology (Moll & Jansen, 1995; Strong & Fragoso, 2006; Guzman & Stevenson, 2008; Jerozolimski et al., 2009). We therefore here include information on the movement ecology of chelonians as it affects seed deposition, germination success and ultimately plant recruitment.
Turtles and tortoises have varied home range sizes and movement distances, and these may vary depending on species and individuals within species. There is high variation of home range size between species, with the mean home range size generally increasing with species size (Fig. 4a). Overall, chelonians have a mean home range size of 14.8 ha (± 24.2; n = 41). There is a high within-species variation in home range size (Fig. 4a). Furthermore, chelonians show overall mean daily displacements of 103.9 m day-1 (± 114.3; n = 22), but displacement distances do not seem to be related to chelonian size (Fig. 4b). As for home ranges, there is a high variation within species.
In contrast to many other frugivores, turtles and tortoises are mostly solitary and thus disperse seeds scattered across the landscape (Varela & Bucher, 2002). Additionally, they often frequent areas expected to be of high recruitment probability for seeds growing into plants. For example, tortoises frequent tree gaps in forested areas to bask in the sun, and such gaps are very suitable recruitment areas for many plant species. A model parameterised with red-footed tortoise cognitive data suggested that the active use of gaps by tortoises enhances the probability of seed deposition in gaps and deforested areas (Soldati, 2015). Indeed, the congeneric yellow-footed tortoise (C. denticulatus), which is a major seed disperser, often deposits seed-rich dung in open habitats and treefall gaps (Josseaume 2002, cited in Bonin et al., 2006). In the wild, yellow- and red-footed tortoises favour microsites in open areas that are important for seed germination for resting, such as areas of debris piles, with fallen branches, vines or trees, where they presumably defecate more often than other sites (Moskovits & Bjorndal, 1990; Strong & Fragoso, 2006). Brown wood turtles (Rhinoclemmys annulata), are also known to frequent tree gaps (Moll & Jansen, 1995). Open areas are also often used by the gopher tortoise (Gopherus polyphemus), which are important areas of plant recruitment in pine savannah in the southestern USA (Birkhead et al., 2005). The European pond turtles, which disperses the seeds of the aquatic waterlily (Nymphaea alba, Nymphaceae), effectively disperse seeds between ponds, aiding in maintaining population connectivity and meta-population dynamics of the waterlily (Calviño-Cancela et al., 2007). Moreover, even aquatic species often spend time out of the water, increasing the probability of dispersing plants to suitable habitats (rather than in the water). For example, the black river turtle (Rhinoclemmys funerea) in Costa Rica regularly defecates on land (Jansen 1993, cited in Moll & Jansen, 1995).
(c) Seed and seedling fate
Seed deposition after zoochory has both spatial and temporal aspects, both of which affect the ultimate fate of seeds. Spatially, the Janzen–Connell model proposed that seeds that are dispersed away from maternal plants have a higher probability of survival as they can escape distance- and density-dependent seed- and seedling predation (Janzen, 1970; Connel, 1971). Both of these are ubiquitous interactions that result in strong establishment limitations for plants (Crawley, 2000; Wright, 2002; Paine & Harms, 2009). Temporally, Guzmán & Stevenson (2011) proposed that escape in time via endozoochory by animals with low metabolic rates and long GRTs, such as chelonians, may aid seeds by basically allowing them to ‘time travel’ into the future to escape from periods with high-intensity seed predation.
After being deposited in suitable habitats, viable seeds that escape predation and pathogens may eventually germinate, and a proportion of these survive and are recruited as adult plants. One of the factors that can affect germination percentage and rates of seeds consumed by chelonians is the gut treatment. For example, gut washing by the digestive fluids of frugivores may be an important mechanism which aids in increasing seed endocarp permeability, and thus enhance germination (Traveset, 1998). Germination percentage and rates can vary within plant genera and between plant species and on the frugivore species after gut passage (reviewed in Traveset, 1998). Effects on seed germination after gut passage can go from positive (enhanced germination), neutral (no effect), to negative (decreased germination). In the studies reviewed here, chelonian gut passage had a mixed effect, depending on the species of chelonian and of fruits/seeds consumed (Table 3). Compared to controls (depulped seeds), 29% of the cases, gut passage had a negative effect on germination, the effect was neutral for 39% of the cases, and in 32% of the cases, seed germination was enhanced.
In addition to depending on the species of chelonians and plants, factors such as chelonian ontogeny, seed size, within-species variation in seed dormancy, and external stimuli may affect seed germination. For example, tortoise age, which correlates with size, can affect the likelihood of seed germination after passage through the guts of Aldabra giant tortoises (Aldabrachelys gigantea), with smaller sub-adults increasing the probability of germination of some plant species when compared to larger adult tortoises, and this was attributed to the shorter GRTs of sub-adults (Waibel et al., 2013). Braun & Brooks (1987) found that after gut passage through the box turtle (Terrapene carolina), seed germination increased with increasing seed size. Plant species may also have different degrees of seed dormancy that may affect seed germination after gut passage (Rick & Bowman, 1961). External stimuli, such as the availability of light has been shown to have a differential effect on aquatic seed germination, with delayed germination after gut passage in light conditions (but with equal total germination to controls), and delayed germination during the first year, with subsequent increased germination speed and percentage in the long term in dark conditions (Calviño-Cancela et al., 2007). The authors suggest that in their natural habitat, the differential effect of gut passage in combination with light stimuli is expected to affect seed germination in turbid vs. clear bodies of water. Similarly, we can expect that seeds inside the dung of terrestrial species, with no direct light, to have a delayed germination, as dung disintegrates, and thus escape predators in time (assuming that the dung does not attract predators).
In terms of seedling growth and vigour, the few studies we found reported a positive effect of chelonian gut passage. For example, in the case of Syzygium mammilatum (Myrtaceae), gut passage through A. gigantea had negative effects on seed germination rate, but positive effects on seedling growth and health when grown ‘in shitu’ (i.e., grown in scat) (Hansen et al., 2008). In addition, Elbers & Moll (2011) found that common persimmon (Diospyros virginiana, Ebenaceae) and water tupelo (Nyssa aquatica, Nyssaceae) seeds had lower proportions of germinating seeds (compared to controls) after passage through the guts of alligator snapping turtles (Macrochelys temminckii), while the acorns of the willow oak (Quercus phellos, Fagaceae) had a higher proportion of germination after gut passage compared to controls. Passage of seeds of the grass Briza maxima (Poaceae) through the gut of the spur-thighed tortoise (Testudo graeca) led to seedlings growing larger and faster, although this may have been due to filtering of seed size, as only larger seeds were recorded passing through the gut (Cobo & Andreu, 1988).
(d) Secondary seed dispersal
Secondary seed dispersal is the process by which seeds that have been initially dispersed by a frugivore via endozoochory are consumed by a second disperser, for example, through coprophagy. Some chelonian species have been observed acting as potential secondary seed dispersers. For example, giant tortoises frequently eat each other’s scat on Aldabra Atoll (WF & DMH, pers. obs.), and red- and yellow-footed tortoises (Chelonoidis carbonarius and C. denticulatus, respectively) have been observed eating tortoise scat in Brazil (Moskovits and Bjorndal 1990). Also, Young (2003), states that tortoises (without specifying which species) are partial to eating dung from camels, sheep, and goats, who themselves are potential seed dispersers (Kuiters & Huiskes, 2010; Mancilla-Leytón, Fernández-Alés, & Vicente, 2011; Root-Bernstein & Svenning, 2016). Juvenile Central American river turtles (Dermatemys mawii) eat the scat of adults, presumably to obtain cellulolytic bacteria to aid in digestion of plant foods (Legler & Vogt, 2013). Forsten’s tortoise (Indotestudo forsteni) has been observed eating monkey scat, which contained fruit pulp (Ives et al., 2008), and thus likely also contained seeds. In addition, deer faecal pellets were found in the scat of the box turtle (Terrapene carolina bauri) (Platt et al., 2009). Although it is possible that they secondarily ingested some seeds from the deer scat, the authors stated that the contribution to the overall number of seeds found in the turtle’s dung is likely to be minimal. Also, North American box turtles (Terrapene carolina and T. ornata) regularly eat cow dung which often contains seeds (DM, pers. obs.)
Seeds in tortoise scat can also be potentially secondarily dispersed by non-chelonian species. For example, turtle doves (Streptopelia picturata) have been observed eating the contents of giant tortoise scat on Aldabra Atoll (WF, pers. obs). Moreover, dung beetles, which feed on scat and usually bury it, have been recorded amassing and dispersing scat of red- and yellow-footed tortoises in Brazil (Strong & Fragoso, 2006). In addition, land crabs (Cardisoma carnifex) and coconut crabs (Birgus lastro) have been observed eating giant tortoise scat containing grass and Ficus sp. seeds on Aldabra Atoll (WF, pers. obs.). We are unaware of any studies addressing the effects of secondary seed dispersal by chelonians, or other species consuming chelonian scat, on plant germination and/or recruitment and it thus remains to be seen whether effective secondary seed dispersal occurs in, or is promoted by, chelonians.
VI Chelonian FSD in a community context
The interactions between plants and frugivores do not occur in a vacuum, but are embedded in the ecological network of seed dispersal interactions between all plant species and all frugivores in the community (Bascompte & Jordano, 2007).
Therefore, if we truly want to know the role of chelonians as seed dispersers, we must look at their role in a community context. Studies on the role of chelonians as frugivores and seed dispersers at the community level are scarce (we found only four studies, described below) yet they provide valuable insights about their role in relation to other frugivores.
Donatti et al. (2011) studied seed dispersal interactions at the community level in the Brazilian Pantanal using bipartite interaction network analysis. For mutualistic plant–animal interactions, a bipartite network consists of nodes (vertices) and links (edges), which are represented by trophic levels (i.e., frugivores and plants in this case) and the interactions between them (interactions within trophic levels are not possible). The Pantanal seed dispersal network was hyper-diverse, with 46 species of frugivores interacting with 46 species of plants. In the network, the red-footed tortoise (Chelonoidis carbonarius), was the sixth most important frugivore in terms of the number of interactions (Fig. 5a). Given the diversity and the complexity of the network, based on the number of interactions in comparison to other frugivores and the fact that they are capable of dispersing large-seeded plants, red-footed tortoises are probably one of the most important dispersers in the Pantanal community.
Falcón (2018) studied seed dispersal interactions in the smaller plant–frugivore community of Aldabra Atoll (with ten frugivores and 37 plant species), home to Aldabra giant tortoises (Aldabrachelys gigantea). The network was highly generalised, and tortoises were the second most important seed dispersers in terms of the number of interactions. In total, A. gigantea dispersed the seeds of at least 20 fleshy-fruited plant species (grasses and sedges were not included; Fig. 5b), including large-seeded ones such as Cordia subcordata (Boraginaceae) and Guettarda speciosa(Rubiaceae). Moreover, he found that the network was most vulnerable to the loss of three particular frugivores, one of them being the giant tortoises. This study highlighted the importance of tortoises as megafaunal seed dispersers and suggests that the many recently extinct giant tortoises in the Indian Ocean (see Hansen et al. 2010) had a similarly pivotal role in their communities before being exterminated.
In Galápagos, Heleno et al. (2013) used network analysis to investigate the impact of alien plants on the seed dispersal networks in two islands, one of which harboured giant tortoises (Chelonoidis nigrus). They looked at the seed dispersal of both native and introduced plants by the different island frugivores. Giant tortoises here were the third most important seed disperser in terms of the number of interactions; Fig. 5c), and were especially important for fleshy-fruited plants. They also performed an analysis of the quantitative seed dispersal network, and stated that tortoises played an important role as seed dispersers based on the strength of interactions, and that the extirpation of tortoises on other islands in the Galápagos must have resulted in a negative impact on seed dispersal function at the community level.
Also in the Galápagos, Nogales et al. (2017) took a step further and studied the direct contributions delivered by different groups of frugivores, including giant tortoises, lizards, and three groups of birds, to the number of seeds dispersed, and the effect on germination. Frequency of occurrence of seeds was the highest in the scats of giant tortoises and medium-sized passerine birds, but the number of seed deposited per unit area was lowest for tortoises and lizards. In terms of seed emergence after gut passage, only a small proportion of seeds from all scat samples germinated (19%) within the study period, but those that originated from tortoise scat showed the highest emergence frequency compared to seeds dispersed by all the other disperser guilds. Based on the large frequency of occurrence and number of seeds found in the scat, as well as seed germination after gut passage, they concluded that Galápagos giant tortoises play a key role as seed dispersers in the Galápagos Islands.
VII Chelonians as megafaunal seed dispersers
On many islands worldwide, large and giant tortoises were present until recently, and were often the largest vertebrates in their respective faunas (Hansen et al., 2010). Giant tortoises on islands function as megafauna, capable of dispersing even very large seeds (Hansen & Galetti, 2009). Surprisingly, there is evidence that medium-sized tortoises in continental ecosystems can disperse unexpectedly large seeds. In Amazonia, (Jerozolimski et al., 2009) found that yellow-footed tortoises (Chelonoidis denticulatus), a tortoise with a mean length of 40 cm, dispersed seeds of the palm Attalea maripa (Arecaceae) of up to 40 × 17 mm, and (Mitchell & Daly, 1998) described how C. denticulatus tortoises easily swallowed the 50–60 mm large fruits of Spondias testudinis (Anacardiaceae), thus presumably capable of dispersing the ca. 40 × 30 mm large seeds. The two Brazilian Chelonoidis species may thus act as some of the last surviving heirs to several of the many large-seeded fruits left orphaned by late Pleistocene megafauna extinctions (Guimarães, Galetti, & Jordano, 2008), and Spondias mombin is thus perhaps not yet entirely “culturally deprived in [mammalian] megafauna-free forest”(sensu Janzen, 1985).
VIII Chelonian FSD and conservation/restoration
Chelonians are the most endangered of the major groups of vertebrates, exceeding birds, mammals, fishes and amphibians (van Dijk et al., 2014). Factors that affect the conservation of chelonians include habitat destruction, exploitation, and climate change. On a more positive note, chelonians have shown themselves to be key players in habitat restoration projects.
(1) Chelonian conservation
Roll et al. (2017) found that the distributional overlap of the range of chelonians with protected areas is only ca. 10%, which puts them at great risk, especially if they are habitat specialists. For example, the Northern Australian snapping turtle (Elseya dentata)resides in riverine habitats, and their diet consist mainly of fruits of riparian rainforest trees, so they are particularly vulnerable to changes in land management that may have negative effects on riparian forest habitats (Kennett & Tory, 1996). Thus, habitat modification and destruction not only affect chelonian populations, but can also affect the availability of fruit resources, which can lead to the loss of seed dispersal mutualisms.
Exploitation is another factor threatening the conservation of chelonian species, and the main causes are consumption as food resources, traditional medicine, and the pet trade. Known frugivorous chelonians are not exempt from suffering from exploitation, and for example, species of the turtle genera Trachemys and Pseudemys are the most exported turtles in the USA, with individuals being taken directly from the wild, or taken from the wild and subsequently bred in captivity (Moll & Moll, 2004; Mali et al., 2014). Similarly, species such as the radiated tortoise (Astrochelys radiata) (Leuteritz, 2003) and the spur-thighed tortoise (Testudo graeca), are prone to exploitation from their native habitat (Walker & Rafeliarisoa, 2012). Exploitation of chelonian populations may have important implications for seed dispersal as the reduction of frugivore populations can result in the functional extinction of seed dispersal mutualisms, even before the species of frugivore itself goes extinct (e.g., McConkey & Drake, 2006).
In addition, changes in temperature and precipitation due to anthropogenically-induced climate change are poised to affect many ectothermic species, including chelonians, harder than endothermic ones (Walther et al., 2002; Deutsch et al., 2008; Clusella-Trullas, Blackburn, & Chown, 2011; Ihlow et al., 2012). For example, turtle and tortoise species may respond strongly to precipitation, and their activity and movements decrease with increasingly dry periods (Luiselli, 2005; Baxter, 2015; Falcón et al., 2018). In addition, increasing droughts can affect the habitats of chelonians (Haverkamp et al., 2017), and potentially reduce shade availability, which is an important resource for thermoregulation (Merton et al.,1976; Moulherat et al., 2014). Moreover, increasing temperatures have been shown to decrease the activity of chelonians, and they may be particularly vulnerable to increases in air temperature in terms of thermoregulation (Lambert, 1981; McMaster & Downs, 2013; Falcón et al., 2018). Thus, the magnitude and outcome of chelonian FSD is very likely to be negatively affected by climate change.
(2) Rewilding and restoration
Overall, because frugivorous chelonians in general are efficient seed dispersers they are ideal candidates for rewilding and restoration efforts that have the resurrection of extinct seed dispersal interactions as a major focus. This is especially the case in island ecosystems, where many of the large-bodied frugivores have gone extinct (Heinen et al., 2017), and where giant tortoises are in general considered to be ecosystem engineers (Hansen et al., 2010). The best-studied example of this is the introduction of Aldabra giant tortoises to islands in the Mascarenes to restore the function left behind by the extinction of the endemic Cylindrapsis giant tortoises (Griffiths et al., 2010; Hansen et al., 2010). Here, they effectively disperse the seeds of several endemic and endangered plant species (Hansen et al., 2008; Griffiths et al., 2010), including the large-seeded Diospyros egrettarum (Ebenaceae) (Griffiths et al., 2011). Moreover, these tortoises have also been shown to have potential as seed dispersers of the huge fruits of Baobab trees (Adansonia rubrostipa, Malvaceae) in Madagascar, where giant tortoises also used to occur (Andriantsaralaza et al., 2013), and may soon find themselves being deployed as ecological restoration agents in Madagascar, too (Pedrono et al., 2017).
The perhaps most ‘extreme’ functional substitution can be found in Hawai’i, where recently extinct herbivorous and frugivorous giant flightless ducks and geese have been replaced by the large African spurred tortoise (Centrochelys sulcata) in the Makauwahi Cave Reserve on the island of Kauai (Burney et al., 2012). Although neither terrestrial nor fresh water chelonians ever reached Hawai’i by natural means, based on their ecology, the authors posited that the spurred tortoises could act as ecological substitutes for the extinct endemic frugivore-herbivores.
Rewilding with tortoises does not have to be necessarily limited to islands, and according to Sobral-Souza et al. (2017), the continental northern Atlantic Forest of Brazil, which is heavily defaunated and fragmented, and whose fragments are too small to reintroduce large mammalian frugivores, is another potential tortoise rewilding region. Based on studies highlighting the role of yellow- and red-footed tortoises as seed dispersers, especially for large-seeded plants (Chelonoidis denticulatus and C. carbonarius, respectively), and on the success of rewilding efforts with Aldabra giant tortoises, the authors argued that introducing these Chelonoidis spp. in fragments of the northern Atlantic Forest would be a way to mitigate the negative cascading effects of defaunation. To support their argument, they employed niche modelling based on known occurrence of tortoises, and assessed food availability and conservation co-benefits, and found that fragments in the northern Atlantic Forest are suitable for these tortoises.
X Conclusions
Chelonian FSD is geographically and taxonomically widespread. In contrast to other major classes of frugivorous reptiles, chelonian FSD is not mainly restricted to islands. However, and different to patterns of chelonian species richness, most FSD studies in turtles and tortoises come from the south-eastern USA and northern South America. Studies on chelonian FSD in south-east Asia, where chelonian species richness peaks, are notably scarce.
Likewise, chelonian FSD occurs widely across the angiosperm phylogeny, with at least one family represented in the major grades and clades. There is, however, an asymmetry of interactions, in which few plant families amass most of the unique pairwise interactions with chelonians.
Based on the studies reviewed here, we expect frugivorous chelonians to be, in most cases, efficient seed dispersers. Not only they can consume large quantities and a high diversity of fruit and/or seed species, but also damage by the mouth parts or after passage is minimal, resulting in many viable seeds. Moreover, compared to controls, passage of seeds through chelonian guts seldom causes negative impacts on seed germination, often resulting in neutral to positive effects, and can result in high seedling vigour.
Seed dispersal interactions do not occur in a vacuum, and the few studies that have investigated the role of chelonians from a community perspective have highlighted their importance in terms of not only the number and strength of interactions, but also the importance of their role as central species amongst frugivores in seed dispersal networks.
Large and giant tortoises (Testudinidae) were present on many islands worldwide, and were often amongst the largest vertebrates. It is in islands, especially, where they are/were prime dispersers of large-seeded plants. Nonetheless, the capacity of large testudinid species in continental ecosystems as megafaunal seed dispersers has also been demonstrated. Therefore, chelonians can act as megafaunal seed dispersers.
Finally, on the one hand, chelonians are amongst the most threatened taxa in the world. Not only they suffer from habitat loss and lack of protection, but they are also heavily exploited, and face an uncertain future due to pressures imposed by climate change. On the other hand, chelonians have a great potential to aid in the conservation of plant–frugivore mutualisms, which have vital implications for ecosystem functioning, and to be used as analogue species to restore lost interactions and functions.
XI Acknowledgements
We thank Dr. Yuval Itescu (Tel Aviv University) for providing us with the data on the global distribution of chelonians. Funding was provided to DMH and WF by the Swiss National Science Foundation (grant number 31003A_143940), and the Zoological Museum and the Department of Evolutionary Biology and Environmental Studies (both at the University of Zurich), for which we are thankful. DM would like to thank Missouri State University, and especially the Biology Department, for financial and other support of research he has participated in related to frugivory and seed dispersal in chelonians.