Investigating the impact of black rats (Rattus rattus) on the endemic and threatened avifauna of Christmas Island

Study site and bird conservation issues

Christmas Island (135 km²) is a basaltic sea mount capped with limestone, located within the equatorial tropics (10.448°S – 105.690°E) with a wet season driven by the north-western monsoon between approximately November and April, and a dry season driven by south-east trade winds from May to October (Director of National Parks 2014). The island geography is comprised of coastal cliffs and terraces and inland cliffs, and a central plateau of approximately 200 m elevation, with the highest point at 361 m. The central plateau is capped by basalt and largely covered by tall evergreen rainforest, coastal terraces are vegetated with deciduous forests, whereas inaccessible coastal cliffs are largely covered in scrub (Geoscience Australia 2014). Historic mining areas, where the closed forest vegetation and soil profile had been completely cleared leaving exposed limestone pinnacles, were later colonised by a dense layer of ground storey weeds, or scrub thickets of woody weeds (Walker et al. 2014). The Christmas Island settlement (population ~1800–3000) is in the north-eastern point of the island. Sixty-three percent of the island is designated as Christmas Island National Park, a Commonwealth reserve. Park management focuses on managing threats to the ecosystems (Director of National Parks 2014). Cat control has been underway since 2010, with that effort enhanced, and aiming for eradication, since 2020.

Christmas Island has high levels of endemism and high rates of extinction. At least 253 species are endemic including 21 vertebrates, of which six are now extinct or extinct in the wild – three reptile and three mammal species, including two endemic rats (Rattus macleari and Rattus nativitatis; Wyatt et al. 2008), while most of the remainder are listed as threatened (James et al. 2019). Christmas Island thrush (Endangered, Environment Protection and Biodiversity Conservation Act 1999; EPBC Act 1999), white-eye (Near Threatened; IUCN Red list), emerald dove (Endangered: EPBC Act 1999), imperial pigeon (Near Threatened; IUCN Red list) and the Indian Ocean red-tailed tropicbird (recommended Vulnerable, EPBC Act 1999; DCCEEW 2022) are believed to be threatened by black rats, and also by cats.

Transect surveys of bird abundance

To investigate relationships between forest bird abundance and invasive species, forest birds were surveyed along 22 fixed transects located across the island (Fig. 1). Transects followed forest tracks and were 800 m long. Each transect was surveyed three times per survey session. Three survey sessions were completed at 4-monthly intervals between April 2018 and February 2019 (see Fig. 1 for timetable). These sessions covered both dry and wet season conditions. The number of birds counted (seen and heard) for each species during each transect was the variable of interest, and was attributed with the transect identity, season, habitat type, year and season covariates (Table 1).

Fig. 1.

The location and timeframes of bird survey transects, thrush nest study sites, red-tailed tropicbird (RTTB) nest study sites, rat activity index (RAI) ink card monitoring transects and rat density monitoring sites in relation to Christmas Island habitat.

Nest monitoring

To estimate the impacts of rats (and other invasive species, including cats and yellow crazy ants) on bird nesting success, nest outcomes were monitored for two focal bird species, a forest exemplar (Christmas Island thrush) and a ground-nesting seabird exemplar (Indian Ocean red-tailed tropicbird).

For thrush, active nests were searched for in or near eight of the 22 bird transects (Fig. 1) between October and March in 2017–2018 and 2018–2019, by following behavioural cues of birds and by visually searching for nests within the branches and canopies of shrub and trees. For tropicbirds, sitting adults were found between May and September in 2017 and 2018 (see Fig. 1 for timetable) at three known breeding sites. Two of the sites were high density cliff-terrace breeding colonies, whereas the third site consisted of a lower density of scattered nests within cliff-top cavities located along a 500 m stretch of coastline (Fig. 1).

Once found, nests were monitored using infra-red motion-sensor cameras (Ltl Acorn 3510As, see Supplementary Table S1 for settings). Nests found at earlier stages i.e. nest building, laying and egg stages were given monitoring priority. Where possible, nests were selected at least 50 m apart to increase the independence of nest observations. Each nest was attributed with species, year and stage found. For thrush, the number of eggs and/or chicks present, the height of the nest from the ground and the tree species the nest was in were also recorded. Red-tailed tropicbirds only lay a single egg on the ground.

Camera data was analysed manually, and nest outcome (1: success, 0: failed) was determined for each nest, and for thrush nests, each individual egg or chick in the nest. Nests were classified as successful when fledglings left the nest. The nest was deemed to have failed when eggs were abandoned, chicks were abandoned, or nests were preyed upon as observed using camera data before the expected fledgling date (thrush: 28 days; red-tailed tropicbirds: 126 days). Where it could be determined, the stage and cause of failure were recorded; if unable to be determined, cause of failure was recorded as unknown. If the outcome of the nest could not be determined (e.g. the camera failed or moved) but sufficient time had passed so that nests could have successfully fledged, nest outcomes were recorded as unknown.

Environmental covariates

Each bird survey transect was attributed with a range of environmental covariates (Table 1).

The influence of rats on bird abundance

We examined whether bird abundance was related to rat density and activity, as well other environmental variables including yellow crazy ant presence and density (Table 1). The abundance of each bird species was modelled separately as a function of the candidate variables (Table 1) using generalised linear mixed models and the package ‘glmmTMB’ (Brooks et al. 2017; RStudio Team 2021). The abundance of the bird species at the survey site was the response variable in all species models except for emerald dove for which presence/absence was used, as the counts were low for this species. Gaussian error distributions were assumed for all species except emerald dove for which a binomial error distribution with a logit link function was used. Survey transect was specified as a random effect for all species. Alternative candidate model sets were built to avoid collinear terms (Pearson’s r > 0.6) using the GGally package (Schloerke 2020). Foliage projective cover (FPC) was collinear with canopy height (r = 0.67) so model subsets contained either one of these variables but not both. The ratios of residual deviance to residual degrees of freedom were assessed to check each model for over-dispersion. Model assumptions were verified by plotting residuals versus fitted values.

We used an information-theoretic approach for the model selection process (Burnham and Anderson 2002). First, the importance of model terms was determined using backwards selection. All possible combinations of the highest-ranking terms were then explored (ΔAICc value <2 and excluding collinear variables) using the dredge function in the MuMIn package v1.43.17 (Barton 2020). Models with an ΔAICc value <2 were considered to have similar support (Grueber et al. 2011). To illustrate the relative significance of terms, continuous variables were standardised to lie between 0 and 1 and the selected models were run using standardised variables. Model averaging was then performed across the 95% confidence set of competitive models based on the Akaike Information Criterion value (AICc) corrected for small sample size using the MuMIn package v1.43.17 (Barton 2020). Variables in the full averaged model set with a 95% confidence interval that did not overlap zero were considered to have an influence on the response variable.

The influence of rats on bird nest success

We examined thrush nest success (success/failure) in relation to rat density and activity, as well as yellow crazy ant abundance, year and habitat variables (Table 1), using generalised linear mixed models (binomial family with a logit link function) and the ‘lme4’ package (Bates et al. 2015). Transect was included as a random effect and year and nest height were included as fixed effects. We examined nest success in red-tailed tropicbirds with two separate generalised linear models (binomial family with a logit link function). The first model (a) included site or year as fixed effects and evaluated the effect of rat activity measured at the site, yellow crazy ant abundance and yellow crazy ant super colony presence, on nest outcome. The second model (b) included site as a fixed effect and evaluated the effect of rat activity measured at the nest level by nest cameras on nesting success using the 2018 subset of red-tailed tropicbird nest monitoring data only (Table 1). The information-theoretic approach, as described above, was used to quantify model support. Model derived estimates of nest survival are presented with 95% confidence intervals.

Patterns of bird abundance

Two hundred and fifteen bird surveys were completed at 22 transects in three survey sessions over 3 years. Thrush (in 214 surveys), imperial pigeon (in 211 surveys) and white-eye (in 215 surveys) were recorded on all or almost all surveys, whereas emerald dove were less commonly observed (in 156 surveys). Goshawk (in 13 surveys), swifts (in 88 surveys) and water hen (in 1 survey), occurred infrequently and were not included in analyses.

Rat activity was included in the top models explaining white-eye, imperial pigeon and emerald dove abundance, but not thrush abundance (Fig. 2; Table S2). There were fewer white-eye, and slightly more imperial pigeon and emerald dove where rat activity was higher, but these relationships were not significant (Fig. 2; Table S3). Yellow crazy ant abundance was included in the top models for all species except for the emerald dove, though was only significant for the white-eye (Fig. 2; Tables S2, S3). Where yellow crazy ant super colonies were present there were significantly more white-eye (Table S3; Fig. 2). There were also more thrush counted where yellow crazy ant super colonies were present, while imperial pigeon were less abundant when yellow crazy ant abundance was higher and if yellow crazy ant super colonies were present (Fig. 2; Table S3). None of these relationships were significant.

Fig. 2.

The 95% confidence set of model averaged coefficient estimates of influential predictor variables for Christmas Island bird abundance based on the generalised linear mixed models. Significant terms are those for which confidence intervals do not cross zero. Reference levels for categorical predictor variables are yellow crazy ant super colony absence, dry season and coastal habitat.

Season, habitat and microhabitat covariates were consistently included in the most parsimonious models of bird abundance, and often as significant predictors (Fig. 2; Tables S2, S3). Thrush were significantly more abundant with increasing tree density (Figs 2, S3). White-eye were significantly less abundant during the wet season, and significantly more abundant with increasing leaf litter cover (Figs 2, S3). White-eyes were also more common with increasing tree density and leaf litter cover, and were significantly less abundant with increasing canopy cover and during the wet season (Figs 2, S3). Finally, imperial pigeon were significantly more abundant with increasing canopy height and during the wet season, but were significantly less abundant with increasing tree density (Figs 2, S3).

Nest monitoring

Thrush

We found 91 thrush nests in a range of tree species (Table S4). Seven nests were not active following camera installation, and the outcome was undetermined for a further 10 nests (these 17 nests were removed from analyses). Of the remaining 74 nests, 29 (39%) fledged 56 young (Fig. 3). Failure was due to nest damage, broken eggs, abandonment, and predation. Goshawks (Accipiter fasciatus natalis) were the most common predator of both chicks (n = 22) and eggs (n = 3), followed by rats (n = 2 chicks; six eggs), while a giant centipede (Scolopendra subspinipes) preyed on one chick. Causes of failure were undetermined for 20 individuals from 14 nests (Fig. 3). Predation of chicks occurred on average within 9.3 days of hatching.

Fig. 3.

Outcomes for individuals and nests over two breeding seasons for thrush (2017–2018 and 2018–2019) and red-tailed tropicbird (2017 and 2018) on Christmas Island, Indian Ocean. Totals include the number of individuals for which the outcome was unknown (‘unknown’), the number of successes (‘successfully fledged’) as well as failed nests by causes of failure.

Rat impacts on nesting success

Thrush nest success was positively and significantly related to rat density (Fig. 4; Tables 2, S5). The relationship between rat density and nest survivorship was positive. Rat activity was not related to nest survivorship (Table 2).

Fig. 4.

The predicted relationships between probability of successful nest outcome and rat activity for the thrush and red-tailed tropicbirds at the site (a) and nest (b) level. Rat activity did not rank highly in models describing nest survivorship for red-tailed tropicbirds at the site level and is shown here to illustrate there was no trend.

Table 2.The most parsimonious generalised linear models of nest success (binary response variable 1, 0) for thrush and red-tailed tropicbirds candidate model sets.

Models	K	AICc	AICc Wt	LL
Thrush
Rat density	3	97.2	0.36	−45.4
Rat density + Yellow crazy ant abundance	4	98.6	0.19	−45.0
Rat density + Nest height	4	98.9	0.16	−45.1
Rat density + Yellow crazy ant super colony	4	99.0	0.15	−45.2
Rat density + Year	4	99.2	0.14	−45.3
Red-tailed tropicbirds a) rat activity measured at the site level
Site	3	116	0.6	−54.9
Site + Year	4	118	0.2	−54.7
Year	1	120	0.1	−59.0
Null model	2	121	0.0	−58.5
Yellow crazy ant abundance	2	122	0.0	−58.7
Yellow crazy ant super colony	2	122	0.0	−58.9
Rat activity	2	122	0.0	−59.0
Red-tailed tropicbirds b) rat activity measured at the nest level
Rat activity nest	2	50.5	0.70	−23.1
Rat activity nest + Site	4	54.5	0.10	−22.7
Site	3	55.8	0.10	−24.5
Null model	1	55.9	0.00	−26.9

Site was included as a random effect for thrush and a fixed effect for red-tailed tropicbirds.

K is the number of estimated parameters including the intercept; AICc, corrected Akaike information criterion; AICc Wt, weights for AICc; LL, log likelihood.

Red-tailed tropicbird nest success was not related to rat activity measured at the site level (Fig. 4; Table 2). Models containing site and year were the highest ranked (Table 2), although no individual sites or years had significantly different nest survivorship (Table S5). However, when rat activity measured at the nest level (2018 data only; model b) was analysed, models containing this variable ranked highest (Table 2). Where rat activity at the nest was lower, nest success was higher; although this relationship was marginal (P = 0.07; Fig. 4; Table S5).

Impacts of other invasive species on Christmas Island birds

Cat predation was not observed at any thrush nests and was observed at only one red-tailed tropicbird nest. However, our study did not capture an important thrush life-history stage extremely vulnerable to cat predation. Juvenile thrushes do not have well developed feathers and remain on the ground where they rely on parental feeding for approximately 30 days post-fledging (Davis et al. 2008). Diet studies show that cats prey upon thrushes on Christmas Island, and juveniles are likely to make up a significant proportion of this. Cats were significant predators of red-tailed tropicbirds before the cat eradication program began on Christmas Island (prior to 2010; Willacy et al. 2021), as has been observed elsewhere (Ratcliffe et al. 2010; Debrot et al. 2014; Vanderwerf and Young 2014; Boeken 2016; Luna et al. 2018).

White-eyes were significantly more abundant where yellow crazy ant super colonies were present. Davis et al. (2008) also found that white-eyes were more abundant in yellow crazy ant invaded forest on Christmas Island. Contact with yellow crazy ants is less likely for white-eye than for other birds as they rarely spend time on the ground where ant densities are highest, and their rapid flying behaviour while foraging may afford few opportunities for yellow crazy ants to swarm onto the birds. White-eyes may also benefit from the increased availability of scale insects caused by yellow crazy ants (Davis et al. 2008). The effect of yellow crazy ants, or centipedes, on nesting success could not be quantified because invertebrates do not typically trigger infra-red motion sensors. However, nest abandonment was observed in several thrush and red-tailed tropicbird nests with high ant activity when returning to collect cameras. Nest abandonment affected 18% of thrush nests and 14% of red-tailed tropicbird nests at the nest building (for thrush), egg laying and incubating stages, and may have been partly due to yellow crazy ant impacts (Davis et al. 2008; Varela et al. 2018; Willacy et al. 2021). Nest cameras inadvertently recorded a giant centipede preying upon a thrush chick; an impact previously undescribed on Christmas Island. On Phillip Island, New South Wales, native centipedes were the principal source of black-winged petrel (Pterodroma nigripennis) nestling mortality, with annual rates of predation varying between 11.1% and 19.6%, equating to 3724 nestlings (Halpin et al. 2021). Thus, further investigation of centipede impacts on declining Christmas Island birds is warranted.

There has been some loss of coastal terrace red-tailed tropicbird nesting habitat caused by the spread of the invasive vine Antigonon leptopus (Director of National Parks 2014), which smothers and destroys the native vegetation that the red-tailed tropicbirds nest beneath (Willacy et al. 2021). Invasive vines may contribute to nest failure if adults are selecting lesser quality nesting sites in more exposed locations. Defoliation of coastal vegetation at nesting colonies was observed as a result of late dry season sea spray events, and abandonment of nests and chick mortality occurred where nests or chicks were exposed. Intraspecific competition resulting in chick death was also observed when chicks moved into adjacent nest sites in response to vegetation defoliation caused by spray events and sun exposure.

Impacts of rats on Christmas Island birds

Spatial variation in black rat density and activity does not relate to spatial variation in the abundance of forest bird species on Christmas Island. In other words, bird abundance is not reduced in areas where rats are most abundant. Furthermore, while rats have been confirmed as a nest predator of thrush, the predation rate was minor compared to other sources of mortality. Rates of rat predation were lower (5% of nests) than recorded for species in the same family studied elsewhere. For example, black rats preyed upon 36% of critically endangered Hawaiian Puaiohi (M. palmeri) nests and some adult females (Vanderwerf et al. 2014), and predation by black rats is listed as a contributing factor in the extinction of several Turdus subspecies, including the Norfolk Island thrush (T. poliocephalus poliocephalus), Lord Howe Island thrush (T. poliocephalus vinitinctus) and New Caledonia’s Mare Island thrush (T. poliocephalus mareensis (IUCN 2017). Rat predation rates of red-tailed tropicbird nests on Christmas Island are also low (3%). While red-tailed tropicbird nesting success was slightly lower with increasing rat activity at nests, this effect was not significant. Again, this result contrasts with reports of rat predation on tropicbird nests elsewhere, which are extremely varied (Fleet 1972, 1974; Wirtz 1972; Ringler et al. 2015; Flores et al. 2017; Luna et al. 2018; Saunier et al. 2022).

There are four potential, non-exclusive reasons for why black rats have little impact on birds of Christmas Island. First, black rats occur at lower densities (4.2–32 rats ha⁻¹) on Christmas Island than reported on other tropical islands. On Europa, where nesting success of red-tailed tropicbirds was <10%, black rat densities ranged from 8 to 65 ha⁻¹ (Ringler et al. 2015) while on Kure Atoll, nest success of red-tailed tropicbirds was 0–11% and Pacific rat densities (R. exulans) ranged from 50 to 185 ha⁻¹ (Wirtz 1972). Rats frequently visited tropicbird nests on Europa. On average, there were 0.34 visits h⁻¹ nest⁻¹ (Ringler et al. 2015). On Christmas Island, rat activity at nests was lower, averaging 0.04 visits h⁻¹ nest⁻¹ (range 0–0.40 h⁻¹ nest⁻¹) which could help to explain lower rat predation rates. However, note that on Europa there was no correlation between nest fate and the number of rat visits (Saunier et al. 2022).

Second, the availability of other plant and invertebrate food resources on Christmas Island may be higher than on other islands (especially low lying and structurally simple coral atoll ecosystems), reducing the need for rats to target bird nests for food (Harper and Bunbury 2015). Black rats are omnivorous (Harper and Bunbury 2015). Diet analyses have shown that rats on Christmas Island largely eat vegetative material and invertebrates including crustaceans (C. Pink 2017, pers. comm.). Ochrosia sp., Pandanus sp., Scaveola taccada and Terminalia catappa occur on Christmas Island, and rats are known to consume these plants on islands elsewhere (Abe 2007; Fleet 1972). Many endemic macroinvertebrates (27 mollusc, 15 crustacean and 150 insect species) are also known to occur on Christmas Island (James et al. 2019). If there are abundant alternative food resources that are less costly and dangerous to find and forage upon, rat predation on birds may be low (Atkinson 1985; Towns et al. 2006; Jones et al. 2008). Fleet (1972) observed rats attacking red-tailed tropicbirds from behind and recorded most rat kills (85%) on chicks <8 days old, before they had become aware of their surroundings. This suggests that predation becomes more difficult for rats as red-tailed tropicbirds become increasingly more active and aggressive, unless they are starved or otherwise stressed. Saunier et al., observed rats repeatedly visiting nests until nestlings were accessible to attempt predation (2022). Seasonal changes in food availability may therefore influence rat predation rates. For example, if rats are resource stressed, they may be more likely to attempt predation of chicks (Zarzoso-Lacoste et al. 2011). On Kure Atoll, it was suggested that a seasonal failure of the preferred food (S. taccada fruit and foliage) due to storm damage caused increased red-tailed tropicbird nest predation by rats (Fleet 1972).

Third, birds on Christmas Island have co-evolved with nest predators and may have adaptations to reduce or offset rat predation rates (Atkinson 1985). Maclear’s rat (R. macleari) and the bulldog rat (Rattus nativitatus) were native to Christmas Island, becoming extinct in 1903 and 1910 after black rats were introduced, most likely from the co-introduction of a novel pathogen (Wyatt et al. 2008). The ecology of these rat species is poorly known, but it is possible that the Maclear’s rat, with its large size and powerful teeth (Tidemann 1989), was a predator of bird nests. Christmas Island also has a diverse and abundant land crab fauna, and crabs are known to prey on bird eggs and nestlings. Only one crab predation event was observed, but crabs may not trigger infra-red cameras as they are ectothermic, and it is possible that some unattributed nest failures were caused by crab predation. Atkinson (1985) suggested that bird species occurring between about 15 degrees north and 20 degrees south of the equator co-exist with native rats and abundant crab fauna and therefore may have adaptations that make them less vulnerable or less naïve to invasive mammalian predators (Atkinson 1985). Thrush nests were commonly observed on the mid-ribs and leaves of arenga palms (Arenga listeri) which appear difficult to access for both rats and crabs.

The native predator (goshawk; A. fasciatus natalis) also had an impact on the nesting success of both the thrush and to a lesser degree, red-tailed tropicbirds. Saunier et al. (2022) noted that brood neglect contributed to nest failure rates on Europa Island caused by black rats and birds of prey (2022). Failure due to brood neglect on Christmas Island was comparatively low. It is possible that on Christmas Island, where the combined natural predation risk to nests is high, adults may defend nestlings for longer, which also makes them less vulnerable to black rat impacts. For red-tailed tropicbirds, the influence of marine productivity on the length of foraging trips and therefore the incidence of unattended nests available for predation could affect this to some extent (Sommerfeld 2008).

Finally, like most tropical passerines, the Christmas Island thrush has a life history characterised by long lifespans (Bird et al. 2020), long breeding seasons with multiple breeding attempts per season, and small clutches (average 2; range 1–4; Ricklefs 1997; Davis et al. 2008; James and McAllen 2014). Nest predation is generally higher in tropical bird species, which has likely driven these life history traits (Pianka 2008) and may mean that Christmas Island thrush are able to offset nest losses with more breeding attempts over their lifetime, compared with passerines in temperate regions (Roper et al. 2010). Similarly, seabirds have ‘slow’ life histories with high survival and low fecundity (typically 1 clutch per year) which makes their populations extremely vulnerable to predation of adults but not young, unless predation of offspring is sustained (Jones et al. 2008; Sandvik and Einar Erikstad 2008). Therefore, if rat densities are held low, such as by competition with land crabs, and if abundant vegetative and invertebrate resources are available, then rat predation may never reach levels sufficient to cause population decline in long-lived birds (Atkinson 1985).

These results suggest that the risk posed by invasive species to endemic bird species can be location specific. Population size, conservation status and life-history traits influenced by evolutionary history (egg size, clutch size, fecundity, nesting behaviour, timing of breeding, defence by adults influenced by multi-invader context and the presence of native predators) affect the vulnerability of bird species to rat impacts (Jones et al. 2008). Rat population density, the availability of alternative resources and the seasonality of those resources influence whether rats will co-exist or impact the avifauna of an island ecosystem (Wirtz 1972; Jones et al. 2008; Woodward 2017). Therefore, understanding island-, site- as well as species-specific invasive predator impacts and responses to invasive predator management are critical to evaluations of management planning in invaded ecosystems.

Implications for future management on Christmas Island

Black rats currently have little impact on the bird species studied on Christmas Island, suggesting rat threat mitigation is not a high management priority. However, rats are known to be a key predator of vertebrate groups other than birds, such as reptiles and invertebrates, and can also affect vegetation. Therefore, further study of the relationships between black rats and the full suite of native species would inform a more thorough ecosystem-wide management strategy. The baseline data from this study on forest bird abundance, forest and seabird nesting success, rat population metrics and rat impacts provide a critical platform for detecting the potential ecosystem changes that could result from cat eradication on Christmas Island. Understanding if and how rat and bird populations respond over time as eradication progresses would be gained through repeated monitoring. Understanding how changes in rat impacts, nest failure or background rates of mortality might affect bird populations, would help to set thresholds for management triggers and improve the cost-effectiveness of monitoring (Bal et al. 2021).