Long‐term monitoring highlights the positive responses of the seabird community to rat eradication at Tromelin Island, Western Indian Ocean

The eradication of rats (Rattus norvegicus, R. exulans and R. rattus) on islands is essential for the preservation of island ecosystems, including seabird populations, which are particularly vulnerable to rat predation. However, the long‐term positive effects of rat eradication on seabird colonies and populations are often understudied. Brown rats (R. norvegicus) were eradicated from Tromelin Island in 2005. No other significative active restoration actions (such as artificial social attraction, translocation or habitat manipulations) have been implemented, which provides a unique opportunity to investigate natural seabird recolonization and recovery processes after rat eradication. We used seabird annual nest counts conducted from 2005 to 2022 and georeferenced data of occupied nests, to precisely describe the rebuilding of the seabird community of Tromelin Island. Despite the lack of any active restoration actions after the eradication of rats, and the remoteness of the island, 17 years after rat eradication, the seabird community increased from two to seven breeding species, and from 353 to 4758 breeding pairs (total for all species). The recovery of masked and red‐footed booby populations was mostly due to the improved breeding success and in fine to auto‐recruitment. Inter‐ and intra‐specific social attraction accelerated the arrival of new species and boosted their population growth. On a finer spatial scale, recolonizing species showed different patterns of colony settlement linked to intra‐specific attraction. The dynamics of the community of Tromelin Island after rat eradication can be regarded as a natural experiment that informs the processes of colony settlement, population dynamics and connectivity in tropical seabirds. Our results demonstrate the huge benefits of eradicating rats from seabird islands, even when no additional active restoration actions are feasible.

Seabirds are K-selected species with reduced reproductive rates, high adult survival and low population growth rates.Furthermore, many species are philopatric (they breed at their place of birth), which reduces the probability of a young adult from recruiting at a site that is not its natal colony (Friesen et al., 2007).
Recolonization processes can be actively manipulated by human actions such as artificial social attraction with the use of decoys or acoustic signals (Parker et al., 2007) and by translocating pre-fledging chicks to a predator-free island (Gummer, 2003).These active conservation strategies have led to well-documented seabird recoveries and recolonizations worldwide (Jones & Kress, 2012;Kappes & Jones, 2014).In the absence of active post-eradication conservation actions, the pace of recovery and recolonization highly depends on the life histories of the species involved: survival, age at maturity, reproductive outputs, coloniality, social behaviors, philopatry, nest site fidelity (the tendency to breed each year at the same place), and dispersal behaviors (Borrelle et al., 2018).In particular, philopatric species are less prone to recolonize islands where rats have been eradicated because of a lack of dispersive individuals.
Previous studies have shown the beneficial effects of rat eradication on seabird populations (Brooke et al., 2018;Jones, 2010).However, monitoring is generally limited in duration (but see Dunlop et al., 2015;Barbraud et al., 2021).It is thus particularly important to conduct long-term monitoring of seabird populations and assemblages at places where rats or other introduced predators have been eradicated.This is particularly true when no active posteradication conservation actions are made, as it informs on natural population and colony settlement processes.These situations can be regarded as a "natural experiment" which informs the seabird population ecology and recovery potential.
Several theories have been proposed to explain the process of colony settlement by considering environmental and social factors (Danchin et al., 2008).Studying the recolonization patterns of seabirds following the eradication of a predator from an island provides a valuable opportunity to test these theories.First, in the absence of any form of social attraction, the pace of seabird settlement would remain constantly low, resulting in a linear and slow population growth (the "Neutral hypothesis"; Figure 1.1a).In most cases however, there is some level of social attraction.The "Inadvertent Island Scale Social Attraction Hypothesis" (Danchin et al., 2004) (Figure 1) suggests that birds are attracted on the island by conspecifics.They may settle anywhere on the island as soon as suitable habitats are available ("Island scale attraction only"; Figure 1.1b).Some finer social attraction processes may drive the pattern of settlement at a smaller scale.The Inadvertent Social Information theory (Danchin et al., 2004) predicts that the density of breeding birds would inadvertently influence the decision of conspecifics regarding their choice of breeding site.If many individuals breed successfully in an area, it means that this area is favorable, so new individuals will try to settle there.There are two distinct scenarios within this hypothesis.The first one predicts that new breeders would "copy" the habitat of experienced breeders by breeding in the same place despite the high density, until the environment is saturated ("Habitat copying hypothesis"; Wagner et al., 2000;Danchin et al., 2001).This is because this place is assumed to be favorable socially (for meeting partners or for obtaining information from other birds on foraging spot location-the information center hypothesis, Ward & Zahavi, 1973) or/and environmentally (favorable site for successful breeding) (Figure 1.1c).Conversely, according to the "Ideal Free Distribution model", competition can affect the choice of breeding sites of individuals before habitat saturation (Danchin et al., 2008;Kennedy & Gray, 1993) (Figure 1.1d).Individuals may evaluate the intensity of competition in an occupied area and would prefer to move to a new area (away from the initial area or in the outskirts).
According to Dunlop (2005Dunlop ( , 2009)), once started, the recolonization of an island by seabirds typically has three stages: (1) a slow settlement of a few breeding pairs coming from other population(s), (2) an exponential growth of the population, mostly due to immigrants socially attracted by the first settlers, (3) a flattening of the growth due to a decrease in the importance of immigration in population growth and an increase of the auto-recruitment rate until auto-recruitment becomes the main factor of growth.The growth stops when populations reach carrying capacity, for example, when all breeding habitats are completely used (Figure 1.2).This dynamic can occur if no other factors limit population growth (no other predator on the island, no major cause of mortality at sea and no limiting food resources at sea).To our knowledge, since Dunlop's (2005Dunlop's ( , 2009) ) publications no other studies have tested if this dynamic is observed at places where rats have been eradicated.
Here, we study the recovery and recolonization processes of the seabird community of Tromelin Island (Western Indian Ocean, Figure 2) after brown rats (Rattus norvegicus) eradication.At the end of 18th century, at least eight seabird species bred on Tromelin.The presence of rats was first described in 1954 (Brygoo, 1955) but they were probably introduced in 1761, as were mice (Mus musculus), when a ship wrecked on the island.Because of the impact of rats, the seabird species richness dropped to two breeding species in 1992 (Le Corre, 1996).In 2005, only 130 pairs of red-footed boobies (Sula sula) and 224 pairs of masked boobies (Sula dactylatra) were breeding on Tromelin (Le Corre et al., 2015).In view of the urgency of the situation, rats were successfully eradicated at the end of 2005 (Le Corre et al., 2015).Although the goal was to eradicate rats and mice at the same time, mouse eradication failed (Le Corre et al., 2015).Apart from this eradication, no other conservation action has been implemented to facilitate the return of seabirds.Eight years after rat eradication, in 2013, 855 pairs of red-footed boobies and 1090 pairs of masked boobies were breeding on the island, an apparent increase of 22 and 23% per year, respectively.Furthermore, in 2014, three white tern (Gygis alba) nests, one brown booby (Sula leucogaster) nest and 11 other non-breeding seabird species were recorded on Tromelin (Le Corre et al., 2015).The seabird community (breeding species and population sizes) has been monitored yearly since that time.
The aim of this study is to precisely describe the natural recovery and recolonization of seabirds on Tromelin Island after rat eradication, without implementing any post-eradication interventions to aid community rebuilding.Furthermore, this study also aims to discuss the observed patterns, in light of the theoretical framework that elucidates colony settlement among seabirds.
F I G U R E 1 Theoretical processes of seabird colony settlements.Inspired by the theories of Danchin (2008) and Dunlop (2005).

| Study area
Tromelin (15.52 S/54.25 E) is a small (1 km 2 ) coralline island on the north-east of Madagascar (Figure 2).The island is classified as a Nature Reserve since 1975 by prefectural decree (no.13/DG/IOI) and is managed by the French Southern and Antarctic Lands Administration (Terres australes et antarctiques françaises-TAAF) since 2007.Fishing is forbidden within 12 nautical miles of the coastline and the island has no human-modified habitats except the airstrip and a small scientific base occupied by three people who are relieved every 3 months.Mice are now the only introduced mammals present on the island.A recent study based on the deployment of camera traps revealed that at Tromelin Island mice very rarely interact with seabird chicks or adults (Manoury, 2017).Although mice have been reported to prey upon seabirds at some locations (Wanless et al., 2007), their impact on seabirds at Tromelin is thus probably limited.
The maximum elevation of the island is ca.6 m asl (Marriner et al., 2010).Its native vegetation is very simple and composed of only one species of native shrub, the beach heliotrope (Heliotropium foertherianum), six native herbs, one cryptogenic and five introduced plants (Boullet & Hivert, 2023).The soil is composed of coral sand and debris.At the end of the 18th century the seabird breeding community was composed of at least eight species: the great frigatebird (Fregata minor), lesser frigatebird (F.ariel), sooty tern (Onychoprion fuscatus), white tern, red-footed booby, masked booby, brown noddy (Anous stolidus) and lesser noddy (Brooke, 1981).Archeological excavation confirmed the ancient presence of these species plus that of one or two species of tropicbirds (white-tailed tropicbird (Phaethon lepturus) and/or red-tailed tropicbird (Phaethon rubricauda) (Laroulandie & Lefèvre, 2014).This archeological study estimated that in the 18th century the island probably hosted hundreds of thousands pairs of seabirds, including a massive sooty tern colony of at least 350,000 pairs (Laroulandie & Lefèvre, 2014).Among the species that bred or are currently breeding on Tromelin, sooty terns, masked boobies and brown noddies are ground-nesting species.White terns, lesser noddies, frigatebirds, and red-footed boobies breed or bred on beach heliotrope.

| Population size of breeding species and observation of non-breeding species
All pairs of masked boobies and red-footed boobies were counted annually from 2005 to 2020 and from 2005 to 2022, respectively (except in 2010, 2011, 2014 and 2015).Red-footed boobies breed all year round on Tromelin.Masked boobies breed seasonally from August to March.For logistical reasons, the first censuses (2005-2013) of these two species were done during different months over the years.From 2005 to 2008, the censuses took place in December and January.In 2009 and 2013, they were conducted in May, while in 2012, they occurred in August.Since 2014, censuses of red-footed boobies are done annually in June, whereas those of masked boobies are done between August and October, just after the laying peak.
Red-footed boobies are polymorphic, with plumage ranging from white to brown morphs with intermediate brown-dominated (considered brown morphs here) or white-dominated (considered white morphs here) individuals (Le Corre, 1999).Tromelin is the only island of the Indian Ocean where the two morphs are abundant with 1/3 brown and 2/3 white.In all other colonies of the Indian Ocean, the white morph is dominant, except at Europa Island where most birds are of the white-tailed brown morph (Le Corre, 1999).The morph ratio observed at Tromelin Island remained stable between 1954 and 2014 (Brygoo, 1955;Le Corre, 1999;Le Corre et al., 2015;Staub, 1970), suggesting that the population growth observed after rat eradication was mostly due to autorecruitment and not to immigration (Le Corre et al., 2015).In 2012, 2013 and from 2016 to 2020 the morph of all the adults captured for banding operations was noted.The difference between the ratios across the years was tested using a chi-square test to see if it remained stable over time, which would indicate no (or low) immigration.
For the newly established species, all breeding pairs were counted at the first detection and then annually just after their laying peak until 2020, 2021 or 2022 (depending on the species).Counts occurred in July for white terns, in June for sooty terns and in May or June for brown noddies.Camera traps were also deployed opportunistically at putative burrows of shearwaters to increase our observation effort.
Since 2017, to detect new colonization as early as possible, the warden in charge of environmental observations on the island noted opportunistic observations of non-breeding seabird species.For each observation, the observer noted the species, the date, the number of birds and the behaviors (flight with no social interaction, flight with vocalizations, prospecting flights, landing on the island or courtships).For all species, population size estimates were obtained through exhaustive nest counts by prospecting the entire island by walk, with one or two people and using GPS (Global Position System) in track mode, to avoid double counting and to ensure full coverage of the island.In very high nestdensity areas, temporary marks were also made on the ground (by adding line on the sand) to mark already counted nests, to avoid double counting.For all species, each adult or pair on an empty nest, incubating, rearing a chick or chick alone, was considered as a breeding pair.

| Annual population growth
We first estimated the mean annual population growth rate between 2005 (t 0 ) and the last year of counts (t) for boobies and between the year of settlement (t 0 ) and the last years of counts (t) for other species with Equation (1).We also estimated the annual population growth of all breeding species between each season t and the previous one (t À 1) with Equation (2).For each equation, N i is the number of breeding pair at season i.
Rats are known to prey upon eggs and chicks of seabirds including boobies (as suggested in Le Corre et al., 2015).We thus assumed that rat eradication, conducted in December 2005, has led to an increase of the breeding success of masked and red-footed boobies from 2006 onward.As a consequence of an increased breeding success, we expected an increase of the recruitment of new adults and thus an increase of the breeding population size 6 years after (6 years being the median age of first breeding for the two species of boobies on Tromelin, Saunier et al., in prep).In order to test this time lag process, we compared the population growth rate immediately after rat eradication (2006)(2007)(2008)(2009), with the growth rate observed from 2012 onward, 2012 being the first year of possible recruitment of the chicks hatched after rat eradication, as in Brooke et al. (2018).To study population growth before the return of the birds that hatched in 2006 we used Equation (1), where N t 0 is the number of breeding pairs in 2006 and N t is the number of pairs in 2009.To estimate the mean population growth rate after the return of these birds (hatched in 2006), we used Equation (1) with t 0 being the year 2012 and t the year 2020 for masked boobies and 2022 for red-footed boobies.

| Spatial distribution and nest density
Seabird nests were georeferenced during the censuses, each year between 2017 and 2022 (but not all species in each year, see more details in results).We constructed a nest density map by species and year to study the spatial distribution of colonies over time.We first built a grid of 1934 cells of 20 m Â 20 m across the island except on the beach where species present at Tromelin do not breed.The spatial analysis of nest distribution was performed with the software QGIS 3.24.1 (QGIS Development Team, 2021).We then estimated whether the number of occupied cells and the nest density in occupied cells varied across time using two Kendall correlation tests (Kendall, 1938).
All statistical analysis was performed with R (R Core Team, 2021), with a significance threshold set at 0.05.

| RESULTS
3.1 | Seabird recovery: population growth of masked and red-footed boobies before and after 2012 Between 2005 and 2006 the red-footed and masked booby populations increased with an annual population growth rate of 2.9 and 1.6 respectively.Then, they decreased with an annual population growth rate of 0.96 for red-footed boobies and 0.95 for masked boobies from 2006 to 2009 (Table 1, Figure 3).The two populations experienced a rapid increase between 2009 and 2012.There was no field work in 2010 and 2011, so the exact starting point of the increase is unknown.The masked boobies reached 1261 pairs in 2020, while the red-footed boobies reached 1850 pairs in 2022 (Table 1, Figure 3).The average annual population growth rate between 2012 and the last count (2020 or 2022 depending on the species) was 1.09 for masked boobies and 1.11 for red-footed boobies.The annual growth rate between 2006 and the last count varied between 0.7 and 2.01 for the masked booby and between 0.6 and 1.5 for red-footed booby with no clear pattern (Figure 4).

| Seabird recolonization
Seventeen years after the eradication of rats, five species of seabirds have successfully recolonized the island.White terns settled in 2014 (three pairs), followed by brown noddies in 2015 (four pairs), sooty terns in 2016 (seven pairs), wedge-tailed shearwaters in 2017 (one or two pairs, but no chick survived until 2020), and lesser noddies in 2020 (first fledgling observed in 2021) (Table 1, Figure 3).A sixth species, the brown booby, has made only one breeding attempt, which failed in 2014, and no other breeding attempts of this species have been observed since.Thus, this species is not considered as a breeder at Tromelin.
In 2022, 353 pairs of white terns, 377 pairs of sooty terns and 910 pairs of brown noddies were breeding on Tromelin.Two pairs of wedge-tailed shearwaters and Abbreviations: NA, means that no census was conducted this year for the concerned species.
seven pairs of lesser noddies bred on the island in 2020 and 2021 respectively (Table 1, Figure 3).The average annual population growth rates between the year of recolonization and the last counts were 1.98, 2.17 and 1.94 for the white tern, brown noddy and sooty tern respectively.For all newly established species the largest population growth occurred between the year of recolonization and the following year (Figure 4).Currently, all three populations continue to increase over time, but at a lower rate (Figures 3 and 4).

| Breeding habitats and nest densities
Masked boobies breed on bare ground, in the central part of the island (Appendix S2).The average density was between 2.2 (SD: ±1.5) and 3.3 (SD: ±2.8) nests/400 m 2 .In 2020, 29% of the cells of the island were occupied by this species.The density per cell and the number of occupied cells slightly decreased between 2018 and 2020 (Figure 5).
Red-footed boobies breed in high density (from 2.9 (SD: ±2.7) to 5.0 (SD: ±5.6) nests/400 m 2 ) on mature F I G U R E 3 Seabird population changes at Tromelin Island since 2005 for masked boobies (Sula dactylatra), red-footed boobies (Sula sula), white terns (Gygis alba), brown noddies (Anous stolidus), sooty terns (Onychoprion fuscatus), wedge-tailed shearwaters (Ardenna pacifica) and lesser noddies (Anous tenuirostris).The first curves on booby plots correspond to the mean annual population between 2006 and 2009, the second from 2012 to 2020.On the other plots the curves corresponds to the mean annual population growth between the year of recolonization and the last counts.Black arrows correspond to the date of rat eradication.
beach heliotropes located mostly in the western and northern parts of the island, and in lower density in the southern part of the island (Appendix S3).Nest density decreased slightly between 2017 (4.2 (SD: ±4.2) nests/400 m 2 ) and 2022 (3.5 (SD: ±3.1) nests/400 m 2 ) (Figure 5) while the number of occupied cells increased from 343 in 2017 to 529 in 2022 (Figure 5).In 2022 redfooted boobies occupied 27% of the cells of the island.
In 2017, sooty terns only bred in the south-west of the island (Appendix S5) at a relatively low density (4 (SD: ±5.8) nests/400 m 2 on average and a maximum of 16 nests/400 m 2 ).In 2018, a second area was colonized to the east (Appendix S5).Since 2020, sooty terns bred only on the east, but in two separate areas.These areas were larger and denser across time, with an average density of 18 nests/400 m 2 (SD: ±16) and a maximum of 64 nests/400 m 2 in 2022 and a maximum nest density of 153 nests/400 m 2 .Nest density increased during the period as well as the number of occupied cells (Figure 5).
White terns breed on beach heliotropes, in low densities (1.3 (SD: ±0.05) nests/400 m 2 ) mostly on the northwestern coasts and in the center of the island (Appendix S6).The number of occupied cells significantly increased from 37 in 2016 to 249 in 2021 (Figure 5), while nest density increased very slightly (Figure 5).
The two-last species (wedge-tailed shearwater and lesser noddy) arrived very recently and are still too few (a maximum of three and seven pairs respectively) for quantitative analysis of their distribution.Their current distribution and densities are given in Appendices S7 and S8 for information.

| Observations of non-breeding species
Eleven species of non-breeding seabirds were observed at Tromelin between 2017 and 2021 (roseate tern (Sterna dougallii), common tern (Sterna hirundo), Saunders's tern (Sternula saundersi), lesser crested tern (Thalasseus bengalensis), brown skua (Stercorarius antarcticus), brown booby, Mascarene petrel (Pseudobulweria aterrima), great frigatebird, lesser frigatebird, white-tailed tropicbird and red-tailed tropicbird) among which 2 were using the island as a regular roosting place (the great and lesser frigatebirds), and two occasionally landed or performed nuptial displays, suggesting a possible future settlement (the white-tailed and the red-tailed tropicbird) (Appendix S9).The other species were migrant that visited the island only very occasionally.A Mascarene petrel was photographed in front of a cavity by one of our camera traps deployed for detecting shearwaters, which is very surprising as it is an endemic species of Reunion Island (see D 'Orchymont et al., 2020 for more details on this observation).

| DISCUSSION
The seabird community of Tromelin Island has changed tremendously since rats were eradicated in 2005.The number of breeding species has increased from two in 2005 to seven in 2021-2022.The number of pairs has increased by a factor of 13 in 17 years, going from 353 to 4758.Two major processes are of interest in this case study, (1) recovery of the two populations which were still breeding when the rats were eradicated and (2) recolonization of locally extinct species.
Direct anthropogenic threats are currently negligible on Tromelin and no actions such as artificial social attraction (with dummies or vocalizations) or translocation of chicks has been implemented after the eradication of rats.The dynamics of this seabird community after rat eradication can thus be regarded as a natural experiment that inform on the processes of colony settlement, on population dynamics and connectivity and on natural resilience of tropical seabirds.

| Recovery process
4.1.1| The recovery of the booby populations: a three-phase process Overall, both populations have increased significantly since rat eradication.The masked booby population has increased 6-fold in 15 years and the red-footed booby population 14-fold in 17 years.The apparent increase in both booby populations observed between 2005 and 2006 is due to the fact that before rat eradication, many breeding pairs were not censused because they failed early in the breeding season, which may have led to an underestimation of the actual number of breeding pairs before rat eradication (Le Corre et al., 2015).Then, the increase in breeding success, due to rat eradication, has induced a population increase 6 years later.This time lag is due to the delayed age at maturity (Saunier et al., in prep;Cubaynes et al., 2011).This kind of time lag has also been suggested for the red-footed booby population of Surprise Island, New Caledonia after rat eradication (Philippe-Lesaffre et al., 2023).
Sulids are known to be highly philopatric and thus little immigration is supposed to occur between populations (Huyvaert & Anderson, 2004;Parker et al., 2007;Steeves et al., 2005).The ratio of the two morphs of red-footed boobies has remained remarkably stable at Tromelin (Appendix S1), which suggests that no or very few birds originating from other colonies immigrated to Tromelin.Indeed, if a significant proportion of birds originating from other nearby colonies (Aldabra, Cosmoledo, Saint Brandon, Chagos) had migrated to Tromelin, the proportion of white birds would have increased, as these populations are composed predominantly of this morph (Le Corre, 1999).Conversely if a significant number of migrant birds had come from Europa Island (which is the only population in the Indian Ocean to be composed almost entirely of brown morph birds), the proportion of brown birds would have increased.
Thus, from 2012 onwards, the growth of the redfooted booby colony is very likely due mostly to autorecruitment.
The observed recovery can be summarized as a threephase process: 1.An apparent quick population growth during the first year after rat eradication, attributed to an underestimation of the number of breeders before rat eradication, because of early breeding failure at incubating stage due to rat predation (here between 2005 and 2006).See also Le Corre et al. ( 2015) 2. A decreasing phase during which the population continued to decrease at a rate probably similar to what it was before rat eradication, until the individuals that hatched at the time of eradication recruited as new breeders.This phase was globally observed from 2006 to 2009 but probably occurred until 2011 knowing the time lag between hatching and age at first breeding.3.An observed increase of the population, since 2012, due to the recruitment of individuals that hatched at the time of eradication and after.It is suggested that this phase will continue until the population reach the carrying capacity, which may be related to the amount of breeding habitat available on the island, or to limited food resources at sea (see Saunier et al., in prep).
The breeding areas of red-footed boobies did not vary during our study.However, in 1995, red-footed boobies were breeding mostly in the central part of the island and not on the northern and western coastline, like now (Le Corre, 1996).This may be explained by the distribution of beach heliotropes, the only breeding habitat of the species on Tromelin, which was denser in the central part of the island in 1993-1995 (Author, pers. commun.) (Author, pers. commun.).
Masked boobies breed on fine sandy substrate in the center of the island.Although the breeding area has expanded as the population size has increased, it has remained broadly the same since 1968 (Le Corre, 1996;Staub, 1970).

| Time-space recolonization pattern
Five seabird species recolonized Tromelin between 2014 and 2022.It took 9 years before the first recolonization event was detected (pre-recolonization stage in Figure 1.2).This time lag is probably due to the geographical distance between Tromelin and the closest source populations (Saint Brandon and Round Island, Mauritius; Appendix S10; Borrelle et al., 2018).It is of interest that the three first species that recolonized the island are terns.Terns are known to occasionally change their breeding place and have a lower level of philopatry than other seabird species (Dunlop et al., 2015;Hughes, 2013;Appendix S10).They are also very mobile, giving them a great dispersal capacity (Appendix S10) and the three species that recolonized the island are globally abundant in the Western Indian Ocean (Appendix S10).As suggested by Borrelle et al. (2018) these three traits (low philopatry, high abundance and dispersal ability) probably favored the recolonization of Tromelin by terns.Within terns, the lesser noddy may be less prone to dispersion (compared to the brown noddies, sooty terns and fairy terns) as they took more time to resettle and they are still in very low number.This is not due to a lack of source populations in the region, as the lesser noddy is extremely abundant in the Western Indian Ocean (Appendix S10) but may be due to a higher level of philopatry, which would require further investigations.
The recolonization process observed for white terns, sooty terns and brown noddies follows the three steps suggested by Dunlop (2005Dunlop ( , 2009) ) (Figure 1.2).The year after the first settlement, the population growth was very strong due to immigration (stage 1 in Figure 1.2).Auto-recruitment is impossible the year after recolonization as brown noddies start breeding at three and sooty terns at 5 or 6 years old (Dunlop, 2005;Harrington, 1974).The population growth rate then slowly declined each year.This phase, is proposed in Dunlop (2005) (stage 2 in Figure 1.2).The white tern and sooty tern population growth rates seem to slowly reach a stable value (Stage 3 in Figure 1.2).This suggests that immigration plays a less significant role in population growth.In other words, if we assume that the number of immigrants remains constant over time, the proportion of immigrants in the populations decreases as the overall populations increase.The pattern is different for the brown noddy population which, surprisingly, increased greatly in 2022.

Interspecific social attraction
The very high population growth in the first year after settlement is due to a strong social attraction.The presence of parents with their chicks informs prospecting birds about the quality of the breeding habitat and the absence of predators of chicks or eggs (Inadvertent Social Information hypothesis; Stamps, 1991;Forbes & Kaiser, 1994;Danchin et al., 2004).Furthermore, spatial and temporal patterns also suggest that fine scale interspecific social attraction also occurred.Indeed, during 9 years, no species recolonized Tromelin.Then it is remarkable to observe the succession of settlements: white terns settled first, brown noddies the year after and then sooty terns and lesser noddies.Brown noddies, sooty terns, wedge-tailed shearwaters and lesser noddies settled exactly in the same place, suggesting that brown noddies may have attracted sooty terns and both species then attracted wedge-tailed shearwaters and lesser noddies.

Intraspecific social attraction
Although favorable breeding habitats for ground-nesting terns are available throughout the island, sooty terns and brown noddies breed in well-defined clusters, suggesting strong intraspecific inadvertent social attraction at habitat scale.Over time, as both populations increased, they spread out more than they increased their nest density.This pattern of colony settlement is in accordance with the "Ideal Free Distribution" hypothesis (Fretwell & Lucas, 1970, Figure 1.1d).
On the other hand, the white terns breed at low-density throughout the island, as soon as there are favorable breeding habitats (beach heliotropes).The nest density at each occupied cell remained relatively the same since they settled although the total number of pairs sharply increased.This resulted in a great increase in the number of occupied cells.White terns do not seem to be attracted by conspecifics at habitat scale as they still breed at low density, but their rapid increase suggests that social attraction may occur at the scale of the entire island.It suggests that their settlement follows inadvertent social attraction hypothesis at the island scale only (Figure 1.1b).

| The case of the wedge-tailed shearwater
Four of the five species that recolonized the island had not been observed breeding there since 1856.The fifth species, the wedge-tailed shearwater, had never been observed on the island before.It is all the more surprising that no evidence of its previous presence has been historically evoked on the island (Laroulandie & Lefèvre, 2014).This suggests that the seabird community, after restoration, gradually moves towards the original ecosystem, even though it does not reach an exact replica of that ecosystem.While Procellaridae are assumed to be highly philopatric (Borg, 2002;Danckwerts et al., 2021;Thibault, 1995;Thibault & Bretagnolle, 1998;Warham, 1990), a recent study on the population genetics and connectivity of the wedge-tailed shearwater at the Indo-Pacific scale (Herman et al., 2022) suggests that the populations are not structured at the scale of an oceanic basin, assuming some gene flow, and thus reduced philopatry for this species.This could explain the colonization of Tromelin by some wedge-tailed shearwaters 12 years after rat eradication.

| Implication for conservation and perspectives
The eradication of rats from Tromelin Island has resulted in a very positive response within the seabird community, with an increased in both species diversity and population size.Notably, four of the species that had previously disappeared from the island as a consequence of rat predation, have recolonized the island.Populations are still far from their carrying capacities on land and most of them are still increasing.Furthermore, some species such as tropicbirds and frigatebirds are commonly observed on Tromelin and may settle in the future.
Populations of seabirds that reproduce on islands invaded by rats typically experience a decline.For instance, in tropical environments, several seabird populations are currently diminishing due to predation by black rats (Appendix S11).However, it is worth noting that once these rodents are eradicated, seabird populations tend to recover significantly (Appendix S11).Our results demonstrate that natural recolonization by seabirds can occur without any active post-eradication measures.This emphasizes the importance of evaluating the need for implementing additional post-eradication management measures such as translocation or artificial social attraction.This result is crucial because post-eradication active measures are not always necessary to promote the recovery of seabird populations, and their implementation can result in significant expenses that could be avoided.
The response of the seabird community of Tromelin Island has been spectacular in terms of demographic recovery and recolonization.However, other local limiting factor (such as marine resources, interspecific competition, breeding habitat limitation, etc.) may hamper this recovery at other places.For instance, at Surprise Island (New-Caledonia) no new species settled after rat eradication, some already breeding species slightly increased while others remained stabled or decreased (Philippe-Lesaffre et al., 2023).
Rat eradication also had a positive effect on the island's vegetation (Le Corre et al., 2015) and indirectly on the nearby adjacent marine ecosystems by increasing the seabird biomass and thus the amount of nutrients discharged into the ocean via guano (Benkwitt et al., 2021).It will be of particular interest to continue documenting these changes and the interactions with terrestrial and marine ecosystems to improve our understanding of the potential cascading effects that occur on a tropical island after the eradication of an introduced predator.

T
A B L E 1 Number of breeding pairs of seabirds on Tromelin island since rat eradication in 2005.

F
I G U R E 4 Annual population growth by year since recolonization of white tern (Gygis alba), brown noddy (Anous stolidus), sooty tern (Onychoprion fuscatus) and wedge-tailed shearwater (Ardenna pacifica) and since rat eradication for redfooted boobies (Sula sula) and masked boobies (Sula dactylatra).The horizontal line represents a stable population (population growth of 1).F I G U R E 5 Temporal changes in nest density and number of occupied cells for masked boobies (Sula dactylatra) (a and f), red-footed boobies (Sula sula) (b and g), brown noddies (Anous stolidus) (c and h), sooty terns (Onychoprion fuscatus) (d and i) and white-terns (Gygis alba) (e and j) on Tromelin island.The p-values and the coefficients τ stem from Kendall correlation tests.