Island rewilding with giant tortoises in an era of climate change

Abstract

Replacing recently extinct endemic giant tortoises with extant, functional analogues provide the perhaps best examples of island rewilding to date. Yet, an efficient future application of this conservation action is challenging in an era of climate change. We here present and discuss a conceptual framework that can serve as a roadmap for the study and application of tortoise rewilding in an uncertain future. We focus on three main ecological functions mediated by giant tortoises, namely herbivory, seed dispersal and nutrient cycling, and discuss how climate change is likely to impact these. We then propose and discuss mitigation strategies such as artificial constructed shade sites and water holes that can help drive and maintain the ecosystem functions provided by the tortoises on a landscape scale. The application of the framework and the mitigation strategies are illustrated with examples from both wild and rewilded populations of the Aldabra giant tortoise, Aldabrachelys gigantea, in the Western Indian Ocean.

This article is part of the theme issue ‘Trophic rewilding: consequences for ecosystems under global change’.

1. Island rewilding with tortoises

Rewilding aims to restore missing ecological functions by replacing recently extinct or extirpated species [1,2]. In the case of globally extinct species, this involves the introduction of a functional analogue non-native species—a highly controversial conservation action, especially when aiming at introducing large carnivores or megaherbivores in continental ecosystems [3]. Island rewilding can be thought of as less controversial, mainly because most island species went extinct much more recently than continental ones, because extinct megafaunal species were much smaller, and because island ecosystems typically did not harbour large carnivores [4–6].

The use of extant giant tortoises to replace recently extinct (i.e. within the past several hundred years) ones is perhaps the best example of island rewilding to date. Until human arrival, many species of giant tortoise were once widespread across islands worldwide. The two sole remaining extant taxa (the Galapagos giant tortoise, Chelonoidis niger/Chelonoidis spp., and the Aldabra giant tortoise, Aldabrachelys gigantea) are increasingly used to rewild these dysfunctional island ecosystems (see Hansen et al. [6] for details on islands and projects). Giant tortoises are central species in island rewilding efforts because the extinct species were ecosystem engineers that drove many important dynamics and processes [6]. However, in contrast to mammalian megafauna, tortoises are ectotherms and need to be able to thermoregulate efficiently, and they are often rewilded to habitats that are at least seasonally dry, where the ability to successfully maintain their water balance can depend on very small margins. A consideration of how ongoing and predicted climate change can affect the ecological outcome of tortoise rewilding is thus imperative, but so far missing in the literature.

We here examine the effects of two main components of climate change, increasing temperature and decreasing precipitation, and how we expect them to affect the function of giant tortoises in rewilding projects. Specifically, we concern ourselves with examining the likely consequences of climate change for the following three ecosystem functions: herbivory, seed dispersal and nutrient cycling. While another climate change component, sea-level change, will also increasingly affect many islands, we do not include this in our study. Once land has disappeared below the sea, it is no longer available to terrestrial tortoises as a habitat, and no (cost-effective) mitigation measures would be able to change this in a meaningful way. With this in mind, though, any ongoing and future tortoise rewilding project that includes areas of low-lying habitat should make it a priority to construct a digital terrain model (e.g. [7]) of the project site to enable an evaluation of the effects of future sea-level rise.

Our primary goal is to present and discuss a conceptual framework that can serve as a roadmap for the study and application of tortoise rewilding in an uncertain future. We then illustrate applications of the framework and the mitigation strategies with examples from A. gigantea tortoise rewilding projects in the Western Indian Ocean, especially the Mascarene Islands, one of the global ‘rewilding hotspots’ in terms of the number of rewilding projects [1,6]. However, many of our arguments are directly applicable to rewilding with Galapagos giant tortoises, too. While these tortoises have so far only been used for rewilding within the Galapagos Archipelago [8], they hold great promise for applications on, for example, many Caribbean islands that harboured endemic giant tortoises until humans arrived [6], including several species that were closely related to the Galapagos giants [9,10]. Lastly, the proposed use of large and giant tortoises as non-native conservation megaherbivores on islands without a history of native tortoises [11] could also be modified according to our arguments here.

2. Ongoing and predicted climate changes in the Western Indian Ocean

Ongoing climate change is poised to increase the mean environmental temperatures, the frequency of both extreme thermal and precipitation events, and, depending on the region, climate change may increase or decrease the mean precipitation [12]. Evaluating and planning for the effects of projected climate change in rewilding projects depends on having good local projections of future climate change trajectories, especially for precipitation patterns that can vary dramatically across spatial scales. Based on a global model with a coarse spatial resolution, simulations project precipitation declines during the twenty-first century for most of Madagascar and the Mascarenes, whereas further north in the Western Indian Ocean, from northern Madagascar and up, islands can expect to experience an increase in precipitation [13].

However, the changes experienced on individual islands may differ substantially from global predictions. For example, regional simulations predict a precipitation decline for Aldabra [14], which is borne out by actual precipitation data from Aldabra that show an annual 6 mm decline in rainfall in the last four decades [15], as well as an increase in drought frequency [16]. The situation is even further complicated by precipitation patterns often varying dramatically across single islands, be they large volcanic islands with significant altitudinal variation, such as Mauritius [17,18], or much smaller, flat coralline atolls, such as Aldabra [19]. Overall, given the small size of the current tortoise rewilding projects (tens to hundreds of hectares [6]), it is of paramount importance to gather site-specific and fine-scale, long-term data to enable the development of efficient predictive models and conservation management plans. However, predictions by even the best models can be thwarted by extreme weather events, which are projected to increase in the South-western Indian Ocean, both for wind and precipitation events [20,21]. This adds another layer of complexity when considering mitigation strategies.

3. Climate change and tortoises

Tortoises are ectotherms and rely on access to an appropriate range of environmental temperatures for efficient thermoregulation, and on sufficient precipitation to regulate their water budget. Ectotherms are thus particularly vulnerable to changing climatic conditions, especially to increased temperatures and decreased precipitation that will strain their thermoregulatory capabilities and water budgets [22–25]. The distribution ranges of tortoises and turtles strongly depend on environmental temperatures, and climate change, together with habitat fragmentation, is expected to result in range contractions for 86% of all chelonian species [26]. Because of their terrestrial habit, tortoises are particularly vulnerable. On warm days, many tortoises depend on vegetation to find suitable shade sites used for thermoregulation. Climate change is expected to alter plant phenology and reduce primary productivity and foliage area [27], which would in turn limit the shade resources used by tortoises. Although they can use waterholes to thermoregulate, declining precipitation will limit the availability of these often ephemeral resources, and the decrease of available water (free or in vegetation) will likely have negative impacts on their water budgets. The overall effects will be significant changes in tortoise behaviour (see below), and because tortoises in particular tend to have long generation times, they will have limited abilities to quickly respond to selective pressures.

4. Tortoise rewilding and climate change: a conceptual framework

We here focus on three ecological functions mediated by giant tortoises, namely herbivory, seed dispersal and nutrient cycling (figure 1a). In general, from an ecosystem function point of view, we argue that giant tortoises should be considered megafaunal herbivores and frugivores [5]; that is, they are responsible for major top-down regulation of functions that directly structure much of the vegetation community and the soil [28].

Figure 1.

An overview of ecosystem functions provided by Aldabra giant tortoises (a), and the expected effects of climate change on tortoise activity (b), movement (c) and spatial clustering of ecosystem functions ((d); see main text for details). (Online version in colour.)

(a). Herbivory

Herbivory is the consumption of plant material by animals and is an important ecosystem function that can regulate plant population and community dynamics, which in turn can influence processes at the ecosystem level [29]. Especially on oceanic islands, giant tortoises are, or were, often the largest herbivore species with the greatest cumulative biomass [6]. Giant tortoises engage in three distinct types of herbivory: grazing (feeding on ground-level vegetation; grasses, sedges, herbs), browsing (feeding on leaves of shrubs and trees) and detritivory (feeding on wilted leaves on the ground).

Giant Aldabra tortoises typically spend a lot of their time grazing and prefer habitats dominated by low grasses and sedges. The highest densities of giant tortoises in both the wild and rewilded populations are thus typically found in such habitats, where the intense grazing pressure maintains a relatively species-rich plant community, the so-called tortoise turf [30]. By browsing, giant tortoises can modify the vegetation structure and create habitat heterogeneity and can even structure the population biology or the trait evolution of plant species. For example, browsing giant tortoises in Galápagos regulate the density, distribution and mode of reproduction (inducing sexual over vegetative reproduction) of endemic Opuntia cacti, which are considered keystone plant species on the islands [31]. In the Mascarene Islands, many plant species show heterophylly, different leaf types close to the ground compared to leaves higher up. The shift between leaf types typically occurs around the maximum feeding height of the extinct endemic giant tortoises (Cylindraspis spp.), and heterophylly has thus been interpreted as an anti-tortoise-browsing adaptation in these islands. Notably, non-native Aldabra giant tortoises strongly prefer the upper over the lower leaf types, lending strong support for this hypothesis [32]. In addition, giant tortoises may eat leaf litter and thus act as detritivores. For example, although giant tortoises on Aldabra prefer to forage on living vegetation, during the dry season, when many plant species lose their foliage, giant tortoises readily incorporate fallen leaves into their diets [30].

(b). Frugivory and seed dispersal

Seed dispersal is an important ecosystem function that serves as a bridge between the end of the reproductive cycle and offspring establishment of plants, and many species of plants rely on animals to disperse their seeds, especially in the tropics and subtropics [33]. For example, through seed dispersal, animals can influence plant recruitment and distribution [34], genetic population connectivity [35] and the ability of plants to reach or migrate to more suitable areas for recruitment [36]. Large frugivores often constitute a disproportionally important part of community-level seed dispersal interactions. A lack of large frugivores as a result of extinction can affect plant recruitment and lead to reduced genetic variation in plants with fleshy fruits with large seeds [37] and may drive rapid evolution of smaller seeds that negatively affect seed survival, germination and seedling growth [38]. Moreover, changes in climate can influence seed dispersal indirectly by influencing frugivores and how they interact with plants [39].

A general understanding of chelonians as important seed dispersers in many ecosystems around the world is emerging; they are capable of ingesting and moving large quantities of seeds, with largely positive gut passage effects on germination and seedling growth (reviewed in [40]). Giant tortoises in particular are known to disperse the seeds of many plants on oceanic islands. For example, both on their native Aldabra Atoll, and on the islands to which they have been introduced for rewilding purposes, Aldabra giant tortoises effectively disperse the seeds of many species of plants [41], and they play a central role in the seed dispersal network of the atoll [40]. The same is true for Galápagos giant tortoises on their islands [42].

(c). Nutrient cycling

Nutrient cycling is the movement and biological, chemical and physical transformation of nutrients (organic and inorganic matter) within ecosystems, between living organisms and the earth and the atmosphere. The cycling of nutrients in terrestrial ecosystems is an important ecosystem function that links the above- and below-ground components and influences processes and properties at the community and ecosystem levels [43]. The sheer quantitative scale of feeding and defecation activities of large mammalian herbivores, especially if their annual cycle includes migratory or long-distance events, generally increases spatial heterogeneity and availability of plant nutrients [44,45]. A loss of large animals can greatly degrade both spatial and quantitative aspects of nutrient cycling [46].

Very little is known about tortoises and their effects on nutrient cycling or on soil nutrients in general; we know of only two studies that even mention this. The first, by Kaczor and Hartnett [47], focused on the burrowing gopher tortoise (Gopherus polyphemus) in North America and found that the burrowing activity of these tortoises created ‘islands’ of soil with relatively low nutrient levels, which increased local plant diversity. The second is a short comment in an update about the Makauwahi Cave Reserve rewilding project on Hawaii, where sulcata tortoises (Centrochelys sulcata) are used as functional replacements for extinct giant flightless ducks, which states that soil levels of nitrate, phosphate and potassium were ‘significantly elated in the tortoise-grazed paddocks' [48]. It should be a priority for tortoise rewilding projects to evaluate the impact of giant tortoises on soil nutrients and nutrient cycling. On Aldabra Atoll and rewilded islands, the vast majority of visible faecal matter is of chelonian origin (figure 2a), which makes it plausible that tortoise-mediated nutrient cycling and nutrient transporting functions are of a magnitude similar to that of large mammals in continental ecosystems. Additionally, the commonly observed detritivory of wilted leaves by tortoises may promote faster cycling of nutrients from leaf litter.

Figure 2.

Wild and rewilded Aldabra giant tortoises, Aldabrachelys gigantea, in their ecosystems. (a) A typical tortoise turf on Aldabra Atoll, with a very high density of not only giant tortoises, but also of their faeces. These form the basic currency of nutrient cycling and seed dispersal in the ecosystem. (b) On Aldabra Atoll, by providing shade during midday, a single tree on the otherwise open and exposed coastal turf directly structures the ecosystem engineering activities of giant tortoises for hundreds of metres around it. (c) Spatial knowledge of sites that can hold ephemeral water resources may be critical. Here, a giant tortoise has made it in time for a drink to an area on Aldabra Atoll, where rain water forms shallow puddles for a few hours. (d) Giant tortoises in the François Leguat Reserve on Rodrigues Island often move long distances along linear features in the landscape, such as on this man-made path. (Online version in colour.)

(d). Tortoise behaviour and the outcome of ecosystem functions

The spatial extent and outcome of the above-mentioned three tortoise-mediated ecosystem functions are largely determined by two main behavioural components, tortoise activity and tortoise movement, both of which are known to be strongly affected by temperature and precipitation (e.g. [49–51]). In the case of Aldabra giant tortoises, Falcón et al. [52] described the activity patterns and thermoregulation of tortoises on Aldabra Atoll. They found strong effects of air temperature on activity levels, with maximum activity levels recorded between 26 and 32°C. As air temperature increased above 32°C, tortoise activity levels decreased steeply, until reaching near-zero values for air temperatures above 35°C, probably because such environmental temperatures exceed the maximum core temperatures typically reached by the tortoises and thus affect their ability to thermoregulate efficiently. For the Aldabra giant tortoise, core temperatures above 36–38°C are lethal [53]. While these tortoises can survive a long time without water, they usually become much less active during drier periods [52]. Other studies have reported similar results in other tortoise species, e.g. Stigmochelys pardalis in Swaziland and in South Africa [54,55], and in North African populations of Testudo graeca [56].

Altering the daily timing and level of activity is likely to be the foremost mechanism of behavioural thermoregulation [57]. Based on the responses of tortoise activity and movement patterns to changes in temperature and precipitation reported in the above-mentioned studies, we expect that the magnitude of tortoise activity and movement will be negatively affected by both temperature increases and precipitation decreases due to climate change. Specifically, we expect that the daily timing of the turning points (going from low to high activity, and vice versa) and the magnitude of the levels of activity will be affected, leading to longer periods of inactivity and lower magnitudes of activity (figure 1b). A projection of observed patterns of activity of giant tortoises on Aldabra Atoll assuming a temperature increase of 3°C due to ongoing climate change seems to fit our expected pattern well (online electronic supplementary material, figure S1).

As a result of decreased activity, we expect that movement patterns will be affected, with a general shift towards shorter movement distances (figure 1c). In turn, we expect the resulting spatial patterns of tortoise-mediated ecosystem functioning to change significantly. As tortoise activity decreases and the need-based linking to shade and water sites increases, the spatial extent of tortoise-mediated ecosystem engineering will decrease, increasingly fragment, and clump around the major shade and water resources (figure 1d). However, we also suggest there might be a potential for an ‘unexpected’ increase of long-distance dispersal events (the hump at the right-hand side of the projected distribution of movement distances in figure 1c) as tortoises can move far to get to water or favoured feeding grounds in stressful conditions. Thus, even in a scenario with increased spatial clustering and fragmentation of ‘high-density’ tortoise ecosystem engineering, long-term genetic connectivity of plant populations could possibly be maintained, and in larger-scale rewilding projects even plant range shifts could be facilitated. This could be the case in future rewilding projects on large islands such as Madagascar, where climate change is expected to result in the shift of climatically suitable habitats for plants, and where plant dispersal is limited [58]. The Galapagos Islands, where giant tortoises disperse seeds over large distances and across altitudinal gradients via their migratory routes, provide an example of this. Here, while some seeds are currently ‘wasted’ (i.e. not dispersed into climatically suitable areas for establishment), it is expected that some of these areas will become favourable under climate change, and thus, that tortoises will aid plants in tracking these changes of climatic suitability [59].

5. Mitigation strategies

Rewilding projects on islands so far cover very small areas, from tens to a few hundred hectares, which typically do not allow for natural migration, be it latitudinal or altitudinal, to track optimal thermal regimes or seasonally abundant resources. Thus, active management, including provisioning of key resources, should be considered. Obviously, a first critical step is to consider which tortoise species is most likely to provide long-term, sustained ecosystem functions for a given rewilding site. This ideally requires detailed knowledge about all candidate species' thermal and hydric ecology (e.g. optimum body temperature, critical maxima for temperature regulation, drought resistance and behavioural adaptations for thermoregulation). Although studies exist for some tortoise species (e.g. [52,60,61]), we currently lack information about this for most species. There is thus a clear need for more information on the thermal and hydric ecology of many tortoise species.

(a). Providing shade and water sites

The distribution of shade sites is probably the single most important parameter for predicting and managing the presence and intensity of tortoise-mediated ecosystem functions at small scales. At larger landscape scales, the distribution of water sites will likely drive seasonal presence/absence of tortoises and serve as waystations for long-distance tortoise movement. Manipulating and optimizing the spatio-temporal availability of these two key resources should therefore be an obvious priority for the management of tortoise-mediated ecosystem functions.

For shade sites, a first step should be to identify likely future thermal refugia for tortoises at the landscape scale for projected climate change scenarios; for example, by following the methods developed by Barrows et al. [62]. Armed with such information, it is then possible to target areas without sufficient refugia and provide artificially constructed shade sites. A single tree can provide a thermal refuge for tens of giant tortoises, and directly structure the presence and intensity or tortoise-mediated ecosystem functions in the nearby landscape (figure 2b). An optimal placement of shade sites thus requires a detailed knowledge of tortoise movement and thermoregulation behaviour, especially in relation to how far the tortoises can be expected to roam from the nearest available shade site within their daily activity cycle. For the medium-sized T. graeca tortoises in Morocco, modelling suggested that they should be able to move up to 1 km between shade sites, but empirical movement data showed that in reality tortoises never moved farther than 500 m between shade sites. For giant tortoises, Swingland and Frazier [53] studied the risk of A. gigantea overheating while grazing on the coastal turf on Aldabra, a narrow strip of seasonally highly productive grassland. They found that a majority of tortoises would graze up to an average of 200–250 m away from the nearest shade-providing plant. Given the ongoing and projected temperature increases, there is likely a continuing contraction of daily maximum tortoise displacements away from shade sites. There is an urgent need for empirical behavioural data to be able to efficiently plan shade mitigation strategies.

Compared to artificially constructed shade sites, providing water sites is likely to present greater logistical challenges; e.g. in terms of sourcing and transporting water. Luckily, tortoises do not need to drink daily; they have very large urinary bladders that they use as water reservoirs and as a store for nitrogenous waste [63]. On Aldabra, before the dry season Aldabra giant tortoises drink copiously and can hold water in their bladder for several months; they will only void their bladders once water becomes available again. If rain falls unexpectedly during the dry season, tortoises will move as fast as they can to reach known sites where water will collect on the surface, as the harsh sun can cause such water to evaporate within hours (figure 2c; W. Falcón and D. Hansen, personal observation). Of relevance to understanding this behaviour in relation to tortoise rewilding, Jørgensen [63] mentions a translocation experiment, where captive desert tortoises that had recently been released into the wild were unable to locate and make use of ephemeral water pools after rains, and thus, unlike resident wild tortoises with long-term experience, they could not replenish their water resources.

Thus, one important aspect for any kind of project management that supplies shade or water sites is that the tortoises need to be able to find and use them. Tortoises have an excellent spatial memory and an ability to navigate across landscapes, often following the same migratory routes (e.g. [64,65]), and they are known to congregate around favoured trees during the fruiting season [66,67]. Such abilities are likely the result of years or decades of experience, although the potential for social learning [68] should not be discounted. Especially in early stage rewilding projects, it could thus be beneficial to purposefully release tortoises near resource-rich sites. Additionally, it may be possible to facilitate navigation by newly released tortoises. In the François Leguat reserve on Rodrigues Island, both soft-released (released from a known habitat with resources and allowed to freely enter/exit) and hard-released (released into unknown habitat without prior acclimatization) Aldabra giant tortoises often followed linear features, such as paths or edges of vegetation, in the habitat (figure 2d) [69]. This suggests that the creation of such features could both facilitate the navigation to water stations and encourage long-distance movements to maintain plant genetic structure at the landscape scale via seed dispersal.

(b). Integration with plant restoration

Many island restoration projects that include tortoise rewilding also feature substantial planting efforts. For example, in the François Leguat Reserve on Rodrigues, the first restoration phase from 2006 to 2011 completely transformed degraded secondary shrub by the planting of more than 120 000 seedlings of native trees. Once the forest was developing, the tortoises were released into the developing forest in the second restoration phase, from 2011 onward [70]. Continuing planting at the project largely aims to fill gaps where saplings have died, using whatever seedlings are available. However, a more targeted, long-term supplementary planting strategy that integrates tortoise rewilding and plant restoration with the specific aim to mitigate the likely negative effects of climate change on tortoise-mediated ecosystem engineering would be ideal. In general, the available plant species, and their capacity of responding to climate change, will determine the quantity and quality of resources that tortoises can use for feeding and shade, which, in turn, will determine the density of tortoises that an area can support and their activity patterns. In the François Leguat Reserve on Rodrigues, tortoises prefer to seek shade under dense thickets of Pandanus plants, whose seeds are dispersed by tortoises, but which provides very little nutrition (D. Hansen, personal observation). On the other hand, many palm species (e.g. Latania and Dictyosperma) produce highly nutritious fruits, yet they provide comparatively little shade. In this case, planting small areas with both kinds of plants in close proximity seems advisable.

(c). Tortoise husbandry

It may be necessary to manage tortoise densities from becoming too high to avoid unwanted negative effects, such as exposure of tree roots with subsequent plant mortality caused by shade-seeking tortoises [30,71]. As argued in Hansen et al. [6], the relative ease of moving and penning giant tortoises means that they lend themselves well to partial husbandry. This could be done on a seasonal basis, or in the case of extreme weather events. Especially in the case of smaller rewilding sites, where rewilding with megafauna is already a very hands-on approach to restoration, considering some level of active population management should be deemed preferable to losing the eco-engineering function of the tortoises.

6. Conclusion and outlook

We believe that the use of giant tortoises in rewilding projects on islands represents an excellent choice for maximizing the quantity and quality of positive restoration outcomes. Yet, the use of these megafaunal ectotherms also presents special challenges in an era of ongoing and rapid climate change. The framework we presented here provides a starting point for practitioners and researchers involved in tortoise rewilding projects to consider how climate change will likely impact tortoise-mediated ecosystem functions at their project site, and plan for corresponding mitigation strategies. We largely covered the interplay between tortoise ecology and ecosystem function, but it is evident that there will be longer-term climate change mediated impacts on tortoise demography and evolution that need to be considered. For example, the sex determination of Aldabra giant tortoises is temperature dependent, with males likely being produced at lower temperatures than females, as is the case for the few tortoises where this has been studied [72]. However, we could find no published values for pivotal temperatures of the transitional temperature range. It is clear that gaining maximum benefits from the proposed mitigation strategies rely on a detailed understanding of the underlying ecological and evolutionary processes and dynamics. We have pointed out important knowledge gaps that should become research priorities.

Data accessibility

This article has no additional data.

Authors' contributions

W.F. and D.H. conceived of the ideas and wrote the manuscript.

Competing interests

We have no competing interests.

Funding

Our work was supported by the Zoological Museum of the University of Zurich.