There is increasing evidence that restoration ecologists should be most concerned with restoring species interactions rather than species diversity per se [1]. Rewilding with taxon substitutes, the intentional introduction of exotic species to replace the ecosystem functions of recently extinct species, is one way to reverse ecosystem dysfunction following the loss of species interactions [2]. This is highly controversial [3], in part because of a lack of rigorous scientific studies [4]. Here we present the first empirical evidence of an in situ rewilding project undertaken as a hypothesis-driven ecosystem management option. On Ile aux Aigrettes, a 25-hectare island off Mauritius, the critically endangered large-fruited endemic ebony, Diospyros egrettarum (Ebenaceae), was seed-dispersal limited after the extinction of all native large-bodied frugivores, including giant tortoises. We introduced exotic Aldabra giant tortoises, Aldabrachelys gigantea, to disperse the ebony seeds. Not only did the tortoises ingest the large fruits and disperse substantial numbers of ebony seeds, but tortoise gut passage also improved seed germination, leading to the widespread, successful establishment of new ebony seedlings. Our results demonstrate that the introduction of these exotic frugivores is aiding the recovery of ebonies. We argue for more reversible rewilding experiments to investigate whether extinct species interactions can be restored.

The extinction of large herbivores, often keystone species, can dramatically modify plant communities and impose key biotic thresholds that may prevent an ecosystem returning to its previous state and threaten native biodiversity. A potentially innovative, yet controversial, landscape-based long-term restoration approach is to replace missing plant-herbivore interactions with non-native herbivores. Aldabran giant (Aldabrachelys gigantea) and Madagascan radiated (Astrochelys radiata) tortoises, taxonomically and functionally similar to the extinct Mauritian giant tortoises (Cylindraspis spp.), were introduced to Round Island, Mauritius, in 2007 to control the non-native plants that were threatening persistence of native species. We monitored the response of the plant community to tortoise grazing for 11 months in enclosures before the tortoises were released and, compared the cost of using tortoises as weeders with the cost of using manual labor. At the end of this period, plant biomass; vegetation height and cover; and adult, seedling, flower, and seed abundance were 3-136 times greater in adjacent control plots than in the tortoise enclosures. After their release, the free-roaming tortoises grazed on most non-native plants and significantly reduced vegetation cover, height, and seed production, reflecting findings from the enclosure study. The tortoises generally did not eat native species, although they consumed those native species that increased in abundance following the eradication of mammalian herbivores. Our results suggest that introduced non-native tortoises are a more cost-effective approach to control non-native vegetation than manual weeding. Numerous long-term outcomes (e.g., change in species composition and soil seed bank) are possible following tortoise releases. Monitoring and adaptive management are needed to ensure that the replacement herbivores promote the recovery of native plants.

No abstract