Invasive rats alter woody seedling composition on seabird-dominated islands in New Zealand
Invasive rats (Rattus rattus, R. norvegicus, R. exulans) have large impacts on island habitats through both direct and indirect effects on plants. Rats affect vegetation by extirpating burrowing seabirds through consumption of eggs, chicks, and adults. These seabirds serve as ecosystem engineers, affecting plant communities by burying and trampling seeds and seedlings, and by altering microclimate. Rats also directly affect plant communities by consuming seeds and seedlings. We studied the direct and indirect impacts of rats on the seedlings of woody plants on 21 islands in northern New Zealand. We compared seedling densities and richness on islands which differed in status with respect to rats: nine islands where rats never invaded, seven islands where rats were present at the time of our study, and five islands where rats were either eradicated or where populations were likely to be small as a result of repeated eradications and re-invasions. In addition, we compared plots from a subset of the 21 islands with different burrow densities to examine the effects of burrowing seabirds on plants while controlling for other factors that differ between islands. We categorized plant communities by species composition and seedling density in a cluster analysis. We found that burrow densities explained more variation in seedling communities than rat status. In areas with high seabird burrow density seedling densities were low, especially for the smallest seedlings. Species richness and diversity of seedlings, but not seedling density, were most influenced by changes in microclimate induced by seabirds. Islands where rats had been eradicated or that had low rat populations had the lowest diversity and richness of seedlings (and adults), but the highest seedling density. Seedling communities on these islands were dominated by Pseudopanax lessonii and Coprosma macrocarpa. This indicates lasting effects of rats that may prevent islands from returning to pre-invasion states.
KeywordsRattus rattus Rattus norvegicus Ecosystem engineer Seedling community Predator eradication
The ability of predators to shape ecosystems has been recognized for decades (Hairston et al. 1960; Paine 1966; Schmitz et al. 2000; Terborgh et al. 2001). When predators are introduced, they can have cascading effects on ecosystems both above and below ground (Schmitz et al. 2000; Hairston et al. 1960; Terborgh et al. 2001; Croll et al. 2005; Fukami et al. 2006). Island systems can be particularly vulnerable to invasion and some of the most devastating invasive predators on islands are rats (Rodentia: Muridae; Atkinson 1985; Courchamp et al. 2003; Jones et al.2008), which consume not only fauna, but flora as well (Allen et al. 1994; Towns and Daugherty 1994; Campbell 2002). Many seabirds nest on islands, finding respite from predators more common in mainland areas. Thus, when rats invade islands they find easy prey and can severely reduce or eliminate seabird populations (Drever and Harestad 1998; Major et al. 2007; Jones et al. 2008). This may be especially devastating to ecosystems if eliminated seabirds nest in burrows (order Procellariiformes–prions, petrels, and shearwaters; also order Charadriiformes–auks). Burrowing seabirds have been termed “ecosystem engineers” because of their strong impacts through soil perturbation and nutrient addition (Mulder and Keall 2001; Campbell and Atkinson 2002; Bancroft et al. 2005; Fukami et al. 2006; Wardle et al. 2007).
Seabirds feed at sea and deposit guano at nesting sites on land (e.g., Burger et al. 1978; Mizutani and Wada 1988; Wainright et al. 1998; Anderson and Polis 1999; Hawke and Newman 2004). When rats consume seabird eggs, chicks, and adults, they interrupt these marine-derived nutrient additions to soils. While guano deposition lowers soil pH (Ward 1961; Okazaki et al. 1993; Mulder and Keall 2001) and potentially reduces nutrient availability to plants (Blakemore and Gibbs 1968; McLaren and Cameron 1990), dead seabird chicks, egg shells, and occasional dead adults increase rates of N and P deposition in the soil (Furness 1991). In addition, seabird trampling and burrowing activity can change litter levels and soil structure, decreasing seed germination and seedling survival (Gillham 1961; Campbell 1978; Maesako 1999; Mulder and Keall 2001; Ellis 2005). These changes together influence plant communities in areas where seabirds are found (Ellis 2005).
Three Rattus species are highly invasive on islands: Rattus exulans, Rattus rattus, and Rattus norvegicus (Atkinson 1985; Drake and Hunt 2009). Since the mid-1800s approximately 45 island groups worldwide have been invaded by rats (R. rattus, R. norvegicus), often replacing populations of the smaller R. exulans (Atkinson 1985; Thorsen et al. 2000). In addition to fauna, rats consume seeds, seedlings, and adult plant parts (Campbell 1978; Campbell et al. 1984; Wilson et al. 2003; Grant-Hoffman and Barboza 2010). During the past decade there have been numerous projects focused on eradicating rats (Towns and Broome 2003) and a recent review reported 332 successful rodent eradications from 284 islands covering a total of 47,628 ha (Howald et al. 2007). However, the removal of rats does not necessarily result in the rapid return of seabird colonies (Gaze 2000; Miskelly and Taylor 2004), resulting in islands that lack both rats and seabirds.
The flora and fauna of New Zealand are especially vulnerable to rats because they have evolved for the last 16 million years in the absence of terrestrial mammals except bats (Atkinson 2006; Worthy et al. 2006). In addition, due to many accessible islands with different rat statuses, New Zealand is a particularly appropriate system for studying the short-term effects of invasive rats. We studied 21 islands off the coast of the North Island of New Zealand. Some of these islands have never been invaded by rats (uninvaded islands), some had rat infestations of either R. rattus or R. norvegicus at the time of our study (invaded islands), and on some islands rats had been eradicated entirely or repeatedly eradicated after subsequent re-invasions, keeping populations low (managed islands). It should be noted that on some of the invaded islands, R. rattus or R. norvegicus may have replaced R. exulans; however, at the time of our study there was no evidence of remaining populations of R. exulans. Here we examine both direct and indirect effects of invasive rats on woody seedling communities on islands.
Island characteristics are more complex than either seabird burrow density or rat status. For example, some islands have been burned or farmed, and factors known to be correlated with plant diversity, such as distance to mainland and island size, also varied. However, we sought to evaluate seabird burrow density and rat status as potential drivers for woody seedling communities. In addition, the effects of R. rattus and R. norvegicus are undoubtedly different. However, these two species of rats do overlap in their effects (Towns et al. 2006). Similarly, there were at least ten different species of Procellariiformes across the study islands. These burrowing seabirds differ in breeding times and duration of breeding season (Warham 1990), but we expected some overlap in their effects. Due to limited islands with histories of only R. rattus and R. norvegicus and overlapping seabird species, coupled with other logistical constraints, we did not partition effects of different rat or seabird species. Despite these limitations to our study, we expected woody seedling communities to be similar in density, species richness, and diversity on islands with similar rat status and seabird burrow densities, due to similar pressures of consumption by rats and nutrient addition and disturbance by seabirds. Other studies on these islands have shown differences in soil and leaf chemical characteristics, C sequestration, decomposition rates, and invertebrate populations in the presence or absence of seabird populations (Fukami et al. 2006; Wardle et al. 2007, 2009; Towns et al. 2009).
We did not expect seedling communities on managed islands, where eradications or eradication attempts have taken place within the last 25 years, to revert to pre-invasion states. Some impacts of rats may take decades to reverse; for example, shifts in the relative abundance or even the complete absence of seeds in the seed bank following selective herbivory. In addition, many of the impacts of rats are mediated by seabirds. Seabirds are philopatric and once removed from islands may take many years to return, or need careful translocations to re-establish populations (Warham 1990; Miskelly and Taylor 2004; Priddel et al. 2006).
Our study is a short-term study and focuses on seedlings rather than adult plants. However, our results are likely to be relevant for the management of islands. We considered woody seedlings that may or may not become an ecologically significant component in the ecosystem. While the seedlings we have considered contribute little to the overall plant biomass present on these islands and may have little influence over ecosystem function, they will determine future community composition and thus future ecosystem function (Grime 1998). Vegetation on islands that have been invaded by rats may continue to diverge from that on islands with large seabird colonies, as the mature vegetation may have developed prior to rat invasion and their full impact may not yet be visible. If the seedling community is not a desirable mix of species (desirable for specific management or restoration goals), then active restoration of the plant community may be necessary to establish a desirable mix of vegetation needed to support native animal populations or other restoration goals.
Materials and methods
We established study plots on 21 islands in warm temperate northern New Zealand (3–350 ha, Electronic Supplementary Material 1). All islands are within 48 km of the North Island and most are of volcanic origin (Electronic Supplementary Material 1). Islands were classified according to rat status: seven islands with rats present (invaded), nine islands where rats were never present (uninvaded), and five islands where rats have been successfully eradicated or repeatedly eradicated after subsequent re-invasions (managed). It is likely that all islands had colonies of burrowing seabirds (petrels, shearwaters, or prions) before the introduction of R. rattus and R. norvegicus (Holdaway 1999; Worthy and Holdaway 2002). R. exulans may also have caused extinctions of smaller-bodied seabirds on these islands, but did not likely reduce overall seabird burrow densities as dramatically as the larger rat species R. rattus and R. norvegicus (Worthy and Holdaway 2002). Definitive evidence of pre-rat seabird densities or time of extirpation is often lacking. Further, on islands where seabirds are no longer present the time since extirpation is often unknown, but is <170 years ago (the time of European colonization). Due to some similarities in behavior of these seabird species, our lack of knowledge of exact species composition of present and historical seabird populations and inability to determine densities of different species due to short visits to islands, and overlap in forest use by some seabird species, we did not distinguish between seabird species. However, it should be noted that differences in seabird species may account for some differences in vegetation and that we could not distinguish between these impacts.
Woody seedling communities and environmental variables
We sampled plots (10 m × 10 m) in secondary coastal forest (Court et al. 1973; Atkinson 2004) in the most mature stands on each island between late January and mid-April in 2004 and 2005. All islands have been disturbed, for example by burning or cultivation, and “mature stands” have regenerated after these modifications and are likely to differ in age. On islands with seabirds present, two plots were placed on seabird colonies (areas with relatively high densities of seabird burrows) and two were placed in areas with few or no seabird burrows. On islands with no seabirds all four plots were placed in secondary coastal forest. Within all plots we positioned 21 regularly spaced 1-m2 subplots (nine subplots in 2004 using a stratified random sampling design). Within each subplot, we identified all seedlings of woody trees to species and counted seedlings in five height classes (0–15 cm, 16–45 cm, 46–75 cm, 76–105 cm, 106–135 cm). We measured and identified all adult trees and stems ≥2.5 cm diameter at breast height (1.5 m) in all plots. Other vascular plants in plots included ferns, grasses, sedges, and mostly non-native herbaceous species. Analyses used the more extensive dataset from 2005, but to maximize the number of species captured we combined data from 2004 and 2005 for the cluster analysis.
Soil temperature (using a HANNA Instruments HI 145 digital thermometer inserted approximately 10 cm into the soil), soil moisture (using a Delta-T thetaprobe inserted approximately 6 cm into the soil), canopy cover (using a spherical densiometer; Forest Densiometers, Bartlesville, Okla), and a litter sample from a 0.1-m2 area were taken in every third subplot (seven samples per plot). We also measured mean air temperature and relative and absolute humidity using dataloggers (Hobo H8 ProTemp/RH dataloggers; Onset Computer) on two plots per island (2004–2005). Soil pH, N concentration, and Olsen’s P concentration were measured from samples obtained from two randomly located samples (one on each of two plots) on each island (Fukami et al. 2006). We counted all burrow entrances within the 100-m2 plot, but the ratio of entrances to actual burrows is probably not 1:1 (as there may be multiple burrow entrances or unused burrows) and may differ slightly for different bird species (Warham 1990). Therefore, our results are correlated with seabird burrow density, but the relationship with seabird density may be weaker. Island area and distance to mainland were obtained from the Rodent Invasion Project sponsored by the Auckland University Department of Statistics (http://www.stat.auckland.ac.nz/research/rodent-invasion/map/, downloaded January 2008). Rat status was based on the knowledge of New Zealand Department of Conservation staff who monitor and maintain bait stations placed on the islands, as well as observations such as partially chewed fruits.
Relationship between woody seedling community and environmental variables
To determine which environmental variables explained the most variation in seedling density, species richness, and diversity we used an information-theoretic approach (Akaike 1973; Burnham and Anderson 2002; Stephens et al. 2005). Candidate models included four variables known to be affected by seabird burrow density that have the potential to affect seedling growth and survival (canopy density, soil pH, total soil N concentration, soil Olsen’s P concentration; Fukami et al. 2006; Mulder et al.2009); and three variables potentially affected by seabird burrow density (soil moisture, soil temperature, litter weight). In addition, since seedling communities may be driven by biogeographical variables, such as climate or island location, that are unlikely to be driven by seabird burrow density or rat status, we included six additional variables (January and June mean temperature, January and June relative humidity, island area, and distance to the mainland). We used Akaike’s information criteria (AIC; Akaike 1973) adjusted for small sample size (AICc; Burnham and Anderson 2002) to select the models that best explained the variation in the data. We report results from these tests and parameter estimates for all models not statistically distinguishable from the best model (AICc < 2).
Composition of woody seedling communities
Differences among and within islands
We performed analyses at two scales (among islands and within islands), to determine if rat status or seabird burrow density was correlated with differences in seedling density, species richness, and Shannon-Weiner diversity (H′).
At the among-island scale we asked if rat status was correlated with variation in seedling density, richness, or diversity. We used analysis of covariance (ANCOVA; sequential SS) and included island size and distance to mainland (to account for established relationships on islands between area and distance to mainland and species richness), burrow density and burrow density squared (to test for non-linearity), and rat status (a fixed variable with three levels: uninvaded, managed, and invaded). Islands were the experimental unit (n = 21) since plots within islands share many characteristics (e.g., past disturbance, island size, distance to mainland, and impacts of chance events).
Species richness at the whole-island scale is likely affected by variables such as island size and distance to seed source as well as past disturbance. To account for differences between islands driven by these factors we re-analyzed plot-level species richness using a model that included whole-island woody species richness, burrow density, and rat status. Values for whole-island species richness were available for 18 of the 21 islands (Cameron et al. 2007; Bellingham et al. 2010; P. J. Bellingham unpublished data; see Electronic Supplemetary Material) and were positively correlated with island size (t15 = 4.76, P = 0.0003) but not distance to the mainland (t15 = −1.45, P = 0.17).
Both rat history and vegetation characteristics may be related to island location (if an island was invaded the neighboring island also had a high probability of being invaded and vice versa), and islands may not be independent units. Therefore, we repeated our analyses using island groups as experimental units (n = 12); since the results from these analyses were not qualitatively different from those using the islands as independent units, we report only the results using all 21 islands as experimental units.
Burrowing seabirds are colonial nesters, congregating in dense populations in some areas of islands and sparsely nesting in other areas of the same island, and using mean burrow densities per island may obscure some of the more subtle effects of seabird densities. Therefore, we performed analyses at the within-island scale to ask whether, for a given island, burrow density had a consistent impact. We limited these analyses to the 12 islands that had a minimum difference of 0.05 burrows per m−2 between high- and low-density areas. For this analysis plots were the experimental unit (three to four plots per island, n = 42). We used an ANCOVA (PROC GLM in SAS; SAS Institute 2003) to examine the relationships between burrow density (as continuous variable) and woody seedling density per square meter, species richness per plot, and species diversity (H′). We included islands as blocks to account for variation between islands.
In addition, to account for inherent differences in seedling densities (generally areas with many seabirds have fewer seedlings than areas with few seabirds and this may affect the number of plant species encountered) we performed rarefaction analysis for species richness at both among- and within-island scales (Gotelli and Entsminger 2006). We compared rat status and high and low burrow densities. Cluster analyses indicated strong differences in composition between plots with fewer than 0.05 burrows m−2 and those with more than 0.05 burrows m−2. Therefore, we categorized high and low burrow density as more than or less than 0.05 burrows m−2.
To determine if smaller seedlings were generally more susceptible to damage by invasive rats or seabird burrow density, we examined the relationship between seedling density per square meter in five height classes, and burrow density and rat status at both the among- and within-island scale. We performed a MANOVA including all height classes followed by ANOVAs for each class if the MANOVA was significant [protected ANOVAs (Scheiner 2001)].
Responses of individual woody plant species
To determine if differences in densities of individual plant species were significantly correlated to either invasive rat status or seabird burrow density, we examined differences in seedling densities for those plant species that were present in at least four plots (within-island scale) or four islands (among-island scale). We used the same within-island and among-island ANCOVA models as for the community-level data. Four species (Coprosma repens, Melicytus ramiflorus, Pittosporum crassifolium, Streblus banksii; scientific names follow http://www.nzflora.landcareresearch.co.nz, downloaded January 2008) showed strongly binomial responses (either they were present and abundant or absent). For these species we performed G-tests at the among-island scale (considering three levels rat status) and logistic regressions at the within-island scale (considering seabird burrow density as a continuous variable).
Relationship between woody seedling community and environmental variables
Variables included in the best models based on Akaike’s information criteria adjusted for small sample size (AICc) for three response variables [seedling density per square meter, plant species richness, and plant species diversity (H′)]
Seedling densitya (four models)
Species richnessa (six models)
Diversity, H′a (seven models)
Range of R2 values for top models
January absolute humidity
January average air temperature
June absolute humidity
June average air temperature
Distance to mainland
Soil total N
Soil Olsen P
The best four models for seedling density included two out of six large-scale and biogeographical variables (June air temperature, island area) and four microsite variables, although for only one of these (soil temperature) was there strong support (95% confidence intervals did not overlap with zero).
The best models for species richness included strong support for a negative relationship with summer absolute humidity and distance to the mainland, as well as for negative relationships with soil pH and total soil N concentration. Diversity was driven by summer temperature (positively), humidity (negatively), and island area (positively), but also by litter weight, soil pH, soil total N and Olsen’s P concentrations (all negatively). Results from biogeographical variables, such as distance to mainland, agreed with predictions from island biogeography theory (MacArthur and Wilson 1967).
Composition of woody seedling communities
Islands grouped by burrow density (Fig. 1). The first delineation, explaining most of the variation (R2 = 0.65), was between two highly burrowed, uninvaded islands (Middle Island, 0.52 burrows m−2; Green Island, 1.01 burrow m−2) and all other islands. The next clear delineation was between the remaining islands with a burrow density > 0.05 burrows m−2, and those islands with a burrow density < 0.05 burrows m−2 (an additional 10% of variation explained). Invaded islands and most uninvaded islands were clustered together, while managed islands were interspersed throughout these two categories (Fig. 1). Overall, the location of the managed islands in the diagram was better explained by their seabird densities than by their rat status: the two managed islands that clearly fell within the uninvaded group, Otata (0.07 burrows m−2) and Whenuakura (0.15 burrows m−2), also had the highest burrow densities of the managed islands. The three islands with very low burrow densities (mean <0.04 burrows m−2) were adjacent to invaded islands in the diagram.
Differences among islands
Low species diversity on managed islands suggests that one or a few plant species perform exceptionally well when rats are eradicated and seabird burrow densities are low (indicating that seabird colonies have not fully recovered). This was supported by a closer examination of relative abundance of individual species. On managed islands one or two species accounted for between 78 and 94% of the seedlings found. Pseudopanax lessonii was the dominant species (38–94% of seedlings) on four of the five managed islands, while the fifth island, Te Haupa, was dominated by Coprosma macrocarpa (78%). Other abundant species on managed islands included Pittosporum crassifolium, Dysoxylum spectabile, and Macropiper excelsum s.l.
Burrow density did not explain species richness or H′ for mature trees (P > 0.1 for both). However, adult vegetation showed the same trends as the seedling community with respect to species diversity and rat status: invaded and uninvaded islands were generally similar (H′ = 1.30 and 1.43; SD = 0.33, 0.44; species richness = 8.33, 8.12; SD = 2.45, 3.27, respectively) while managed islands had lower values (H′ = 0.92, SD = 0.46; species richness = 1.52, SD = 3.27), although these differences were not significant (P = 0.07 for both).
Differences within islands
Responses of individual woody plant species
Fifteen of 31 woody species were present as seedlings in at least four plots at the among-island scale and 16 of 31 species were present on at least four islands at the within-island scale. At the among-island scale, densities of individual species (as tested by MANOVA) were not explained by burrow density. However, differences between rat statuses were significant (Roy’s greatest root, F15,3 = 14.88, P = 0.02) and were analyzed further. Different trends were seen for different species. P. lessonii seedlings were significantly more abundant on managed islands than on invaded or uninvaded islands (F2,14 = 9.91, P = 0.002). There was a mean of 21 P. lessonii adults 100 m−2 on managed islands, versus five and four per uninvaded and invaded islands, respectively. Seedlings of Melicytus novae-zealandiae (χ22 = 13.28, P = 0.001) and S. banksii (χ22 = 10.03, P = 0.007) were significantly more abundant on uninvaded islands and absent from plots on managed and invaded islands. There was a mean of 8 adult S. banksii trees 100 m−2 on uninvaded islands, no adults on managed islands, and only one adult on all eight invaded islands combined. One other species showed marginal relationships with rat status: P. crassifolium (χ12 = 4.97, P = 0.08) was more abundant on uninvaded and managed islands and less abundant on invaded islands.
At the within-island scale a MANOVA looking at all species simultaneously was not significant (P > 0.2) so no further analyses were performed. However, results for logistic regression showed that C. repens increased with increasing burrow density (χ12 = 6.29, P = 0.01).
Overall, we found evidence for direct impacts of rats on the woody seedling community of these New Zealand islands, as well as indirect effects mediated through changes in seabird densities.
Effects of seabirds
Soil N and P concentrations, which are increased by seabirds (Ellis 2005; Fukami et al. 2006), were correlated with comparatively low seedling species richness and diversity. Low species diversity in areas of high seabird density is commonly reported (Ellis 2005) and very high nutrient inputs and low soil pH can inhibit seed germination and seedling growth (Hilhorst and Karssen 2000). There appear to be few plants in our system that can withstand these very high nutrient loads coupled with the physical disturbance of burrowing activities. Contrary to what we expected, we found no direct significant correlations between burrow density and species richness or diversity at the within-island or among-island scales. However, differences in island diversity driven by biogeography or past disturbance (beyond seabirds and rats) may be partially responsible: when whole-island diversity was included in the model, there was a negative relationship with burrow density, as expected if high burrowing activity results in reduced survival. Furthermore, we may have lacked power to detect differences between low and intermediate burrow densities. Our cluster analysis suggested that large changes occurred in the seedling community around a threshold of 0.05 burrows m−2.
Physical disturbance by seabirds was closely correlated with decreases in seedling density. High levels of burrowing and trampling can increase damage to plants, limiting plant growth (Maesako 1991). At very high seabird burrow densities (>50 burrows 100 m−2) woody seedling density is low and woody seedling survival may be more a function of chance and less dependent on competition based on seed or seedling characteristics. We found that smaller seedlings (those < 75 cm in height) were more strongly affected than larger seedlings. This suggests that smaller seedlings are more susceptible to death by trampling and uprooting of seedlings and burial of seeds during activity on the ground (Furness 1991; Campbell and Atkinson 2002). By the time seedlings reach 75 cm in height, the negative impacts of high burrow density are no longer discernible.
Our study took a correlational approach that did not allow us to determine cause and effect. Seabirds actively choose where they nest based on soil and vegetation (Carter 2002), and it is possible that some of the vegetation differences we saw within islands were driven by seabird choice rather than vegetation responses to seabird impacts. While we did not find significant differences in slope, aspect, or elevation between plots within islands (C. P. H. Mulder, unpublished data), experimental study with careful tracking of seabird nesting choice on a new island would be required to truly determine the extent to which seabird activities determine vegetation characteristics and vice versa.
Responses of individual plant species
Overall, there was more evidence for direct impacts of rats on seedlings of individual species than for indirect impacts mediated through seabird densities, and only one of 16 species showed a relationship with seabird burrow density: C. repens increased with increasing burrow density and is an early successional species that is often prostrate (Poole and Adams 1964). In heavily burrowed areas trees will often topple over and those that can re-grow from this position have an advantage (Cameron 1990; Bellingham and Sparrow 2000). The reported sensitivity of C. repens to rat effects (Campbell and Atkinson 2002) may be the result of positive impacts of high seabird densities (as described above), which occur only where rats are absent.
Effects of rats
When considering general measures of the seedling communities we found differences in managed islands compared with invaded and uninvaded islands. Overall, we had fewer managed islands (five managed islands compared to nine uninvaded and seven invaded), and control measures as well as time since control or eradication of invasive rats differed on these islands. Therefore, it is difficult to discern if the differences we found were truly due to management of rats or to idiosyncrasies of individual islands. However, we did find more small seedlings on managed islands than on invaded or uninvaded islands. Other studies on New Zealand Islands have found increases in seedling numbers after eradication of R.norvegicus (Allen et al. 1994). In addition, the related R. exulans reduces recruitment and establishment of many New Zealand tree and shrub species through consumption of seeds and plant parts (Campbell and Atkinson 1999, 2002; Campbell 2002). A release from herbivorous pressures of rats may favor quicker growing competitive dominants (see review in Olff and Ritchie 1998) and our results may reflect the recovery of C. macrocarpa, P. lessonii, and other early successional plant species that are vulnerable to rat consumption. Although seedlings of these species may be vulnerable to rats, mature individuals of these species were common on invaded islands, setting the stage for a rapid increase in seedling populations in the absence of both rat consumption and disturbance by seabirds. This is also supported by results for species richness and diversity, which were lower on managed islands than on invaded and uninvaded islands. In addition to rapid seedling emergence by a few plant species, selective consumption by rats may have lead to species being eliminated from the seed bank entirely, reducing species richness.
In the short term, plant communities on managed islands are not reverting to communities similar to those on uninvaded islands. However, whether these changes in the structure of the seedling community will persist as the forest matures is less certain, but consistent patterns of diversity up to sapling sizes and low species richness and diversity on managed islands in adult vegetation suggest that they may.
Responses of individual plant species
Consistent with impacts of rats on community structure, we found evidence for negative impacts of rats on a number of individual species. Previous studies in New Zealand have noted sensitivity of seedling densities of C. macrocarpa,D. spectabile, M. novae-zealandiae,Pittosporum crassifolium,Pseudopanax lessonii, and S. banksii to rats (Atkinson 1985; Campbell and Atkinson 1999, 2002). We found four of 16 woody plant species tested showed at least marginal associations with rat status (M. novae-zealandiae, Pittosporum crassifolium,Pseudopanax lessonii, and S. banksii), and all have been previously identified as species that are sensitive to R. exulans (Atkinson 1985; Campbell and Atkinson 1999, 2002). We found strong support for direct rat predation for two species: S. banksii and M. novae-zealandiae, which were entirely restricted to uninvaded islands, except for adult trees found on Motueka, an island with an active burrowing seabird population. Time since rat invasion is uncertain on this island, but the presence of seabirds indicates that rats may have invaded the island recently. Neither S. banksii nor M. novae-zealandiae were found on managed islands as adults or seedlings and both of these plants have fleshy seeds which may be vulnerable to rat consumption.
Several species recovered well once rats were removed from an island. C. macrocarpa, D. spectabile, Macropiper excelsum,Pseudopanax lessonii, Pittosporum crassifolium all had numerous seedlings on managed islands. These species may be consumed by rats, but this consumption may not be as devastating as for other species. For example, R. exulans (a smaller rat species than those on our study islands) consumes leaves and bark of P. lessonii but it does not appear to eat the fruit or seeds (Campbell and Atkinson 1999) and suppression of P. lessonii may take place at the seedling rather than the seed stage. R. exulans also suppresses the recruitment of C. macrocarpa and D. spectabile (Campbell and Atkinson 2002) and in the case of C. macrocarpa consumes the fruit, bark, twigs, and seedlings (Campbell 1978; Campbell et al. 1984). Maintenance of viable seeds in the seedbank and rapid regeneration of these species following the removal of rats may also play a role in their recovery. For example, P. lessonii regenerates frequently after disturbance (burning, herbivory; Campbell and Atkinson 1999). It has also been postulated that C. macrocarpa and D. spectabile seeds are merely quiescent (not dormant) and can germinate readily after seed fall (Fountain and Outred 1991), potentially out-competing slower germinating species if the pressure of rat consumption is removed. Evidence from the seedling community suggests that rat removal will result in a community with a very different composition than on islands with current rat invasions or on islands where no rats have invaded.
We found strong correlations between measures of seedling communities and both rat status and seabird burrow density, which suggest that these factors are influencing and may be driving these seedling communities. However, we recognize that these islands are small and prone to disturbances such as salt spray and wind, which can topple larger trees and impose severe mortality (Gillham 1960; Cameron 1990; de Lange et al. 2006) and these disturbances may compound those imposed by seabirds or rats. In addition, the time scales which we considered, about 170 years for extirpation of seabirds and < 25 years for invasive rat control, are short in ecological terms and our results may, therefore, have little predictive power for longer term changes to these systems (Yodzis 1988). Also, seabirds likely actively choose nesting sites based on vegetation and soils (Carter 2002) and our observed correlations between seabird burrow density and vegetation characteristics may reflect these choices and not a response to seabird presence. Finally, the islands selected for rat control programs may have shared characteristics of which we are not aware, or that have resulted from management programs (e.g., by increasing the probability of arrival of some propagules). However, incorporating a better understanding of how invasive rats and burrowing seabirds are affecting island vegetation, both synergistically and singly, will improve management of these islands.
For permission to work on the islands they own or for which they are kaitiaki (guardians), we thank the following iwi: Ngāti Hako, Ngāti Hei, Ngāti Manuhiri, Ngāti Paoa, NgātiPuu, Ngāti Rehua, and Ngātiwai, as well as the Ngamotuaroha Trust, the Ruamāhua Islands Trust, John McCallum, Oho Nicholls, Bryce Rope, and the Neureuter family. We thank the New Zealand Department of Conservation for facilitating our visits to the island they administer. We also thank Karen Boot, Tadashi Fukami, Larry Burrows, Ewen Cameron, Aaron Hoffman, Richard Parrrish, Rob Chapelle, Gaye Rattray, Brian Karl, David Wardle, Holly Jones, Nora Leipner, Walter Hoffman and Anders Hoffman for all of their help. We are grateful to Ewen Cameron for providing the list of whole-island flora needed to calculate whole-island species richness. Comments by Henrik Moller and an anonymous reviewer improved the manuscript. This study was supported by the US National Science Foundation (DEB, 0317196), Marsden Fund of the Royal Society of New Zealand, the New Zealand Department of Conservation, the New Zealand Foundation for Research, Science, and Technology (Sustaining and restoring biodiversity OBI), and the Teaching Alaskans, Sharing Knowledge (TASK)/NSF Graduate Teaching Fellows in K-12 Education Program. The experiments performed for this research abide by all current laws within the countries in which they were performed.
- Akaike H (1973) Information theory and an extension of the maximum likelihood principle. In: Petrov BN, Csaki F (eds) Second International Symposium on Information Theory. Akademiai Kiado, Budapest, pp 267–281Google Scholar
- Allen RB, Lee WG, Rance BD (1994) Regeneration in indigenous forest after eradication of Norway rats, Breaksea Island, New Zealand. NZ J Bot 32:429–439Google Scholar
- Anderberg MR (1973) Cluster analysis for applications. Academic Press, New YorkGoogle Scholar
- Atkinson IAE (1985) The spread of commensal species of Rattus to oceanic islands and their effects on island avifaunas. ICBP Tech Publ 3Google Scholar
- Atkinson IAE (2004) Successional processes induced by fires on the northern offshore islands of New Zealand. NZ J Ecol 28:181–193Google Scholar
- Bellingham PJ, Towns DR, Cameron EK, Davis JJ, Wardle DA, Wilmshurst JM, Mulder CPH (2010) New Zealand island restoration: seabirds, predators, and the importance of history. NZ J Ecol 34:115–135Google Scholar
- Blakemore LC, Gibbs HS (1968) Effects of gannets on soil at Cape Kidnappers, Hawke’s Bay. NZ J Sci 11:54–62Google Scholar
- Burger AE, Lindeboom HJ, Williams AJ (1978) The mineral and energy contributions of guano of selected species of birds to the Marion Island terrestrial ecosystem. S Afr J Antarct 8:59–70Google Scholar
- Burnham KP, Anderson DR (2002) Model selection and multimodel inference: a practical information-theoretic approach, 2nd edn. Springer, New YorkGoogle Scholar
- Cameron EK (1990) Flora and vegetation of Middle Island, Mercury Islands Group, eastern Coromandel, northern New Zealand. J R Soc NZ 20:273–285Google Scholar
- Cameron EK, de Lange PJ, McCallum J, Taylor GA, Bellingham PJ (2007) Vascular flora and some fauna for a chain of six Hauraki Gulf islands east and southeast of Waiheke Island. Auckl Bot Soc J 62:136–156Google Scholar
- Campbell DJ (1978) The effects of rats on vegetation. In: Dingwall PR, Atkinson IAE, Hay C (eds) The ecology and control of rodents in New Zealand nature reserves. Department of Lands and Survey information series no. 4, pp 9–120Google Scholar
- Campbell DJ (2002) Changes in numbers of woody plant seedlings on Kapiti Island after rat eradication. Sci Conserv 193:5–26Google Scholar
- Campbell DJ, Atkinson IAE (1999) Effects of kiore (Rattus exulans Peale) on recruitment of indigenous coastal trees on northern offshore islands of New Zealand. J R Soc NZ 29:265–291Google Scholar
- Campbell DJ, Moller H, Ramsay GW, Watt JC (1984) Observations on foods of kiore (Rattus exulans) found in husking stations on northern offshore islands of New Zealand. NZ J Ecol 7:131–138Google Scholar
- Court DJ, Hardacre AK, Lynch PA (1973) The vegetation of the Aldermen Islands: a reappraisal. Tane 19:41–67Google Scholar
- Drever MC, Harestad AA (1998) Diets of Norway rats, Rattus norvegicus, on Langara Island, Queen Charlotte Islands, British Columbia: implications for conservation of breeding seabirds. Can Field Nat 112:676–683Google Scholar
- Fountain DW, Outred HA (1991) Germination requirements of New Zealand native plants: a review. NZ J Bot 29:311–316Google Scholar
- Furness RW (1991) The occurrence of burrow-nesting among birds and its influence on soil fertility and stability. In: Meadows PS, Meadows A (eds) The environmental impact of burrowing animals and animal burrows. Oxford University Press, Oxford, pp 53–65Google Scholar
- Gaze P (2000) The response of a colony of sooty shearwater (Puffinus griseus) and flesh-footed shearwater (P. carneipes) to the cessation of harvesting and the eradication of Norway rats (Rattus norvegicus). NZ J Zool 27:381–393Google Scholar
- Gillham ME (1961) Some interactions of plants, rabbits and seabirds on South African islands. J Ecol 49:275–294Google Scholar
- Goldberg DE, Scheiner SM (2001) ANOVA and ANCOVA: field competition experiments. In: Scheiner SM, Gurevitch J (eds) Design and analysis of ecological experiments, 2nd edn. Oxford University Press, New York, pp 77–98Google Scholar
- Gotelli NJ, Entsminger GL (2006) EcoSim: Null models software for ecology. Version 7. Acquired Intelligence, Kesey-Bear, Jericho, Vt. http://garyentsminger.com/ecosim.htm
- Holdaway RN (1999) A spatio-temporal model for the invasion of the New Zealand archipelago by the Pacific rat Rattus exulans. J R Soc NZ 25:91–105Google Scholar
- MacArthur RH, Wilson EO (1967) The theory of island biogeography. Princeton University Press, PrincetonGoogle Scholar
- McLaren RG, Cameron KC (1990) Soil science. Oxford University Press, OxfordGoogle Scholar
- Okazaki M, Oshida Y, Malony R, Warham J (1993) Effects of sooty shearwaters Puffinus griseus on surface soils on Motuara Island, New Zealand. J Yamashina Inst Ornithol 25:137–143Google Scholar
- Poole AL, Adams NM (1964) Trees and shrubs of New Zealand. Owen. Government Printer, WellingtonGoogle Scholar
- SAS Institute (2003) The SAS system for Windows V9.1. SAS Institute, CaryGoogle Scholar
- Scheiner SM (2001) MANOVA: multiple response variables and multispecies interactions. In: Scheiner SM, Gurevitch J (eds) Design and analysis of ecological experiments. Oxford University Press, Oxford, pp 99–115Google Scholar
- Towns DR, Broome KG (2003) From small Maria to massive Campbell: forty years of rat eradications from New Zealand islands. NZ J Zool 30:377–398Google Scholar
- Towns DR, Daugherty CH (1994) Patterns of range contractions and extinctions in the New Zealand herpatofauna following human colonization. NZ J Zool 21:325–339Google Scholar
- Ward WT (1961) Soils of Stephens Island. NZ J Sci 4:493–505Google Scholar
- Warham J (1990) The petrels: their ecology and breeding systems. Academic Press, LondonGoogle Scholar
- Wilson DJ, Lee WG, Webster RA, Allen RB (2003) Effects of possums and rats on seedling establishment at two forest sites in New Zealand. NZ J Ecol 27:147–155Google Scholar
- Worthy TH, Holdaway RN (2002) Oceanic birds in the terrestrial environment. In: The lost world of the Moa. Canterbury University Press, Christchurch, pp 440–501Google Scholar