Determinants of geographical range sizes: a test using introduced New Zealand birds
Summary
1. The 34 species of birds that have been successfully introduced to New Zealand offer a unique opportunity to study patterns of variation in geographical range sizes, and to test mechanisms that may be responsible for that variation, because the New Zealand range sizes have established within the last 160 years and because there are data on the starting conditions, including the year of first recorded release and the subsequent effort put into the introduction of each species.
2. We collated data on geographical range sizes, life history traits, dates of introduction and initial introduction effort for the birds successfully introduced to New Zealand. To test whether range size–life history correlations show a consistent pattern between regions, we collated further data on geographical range sizes and life history traits for British breeding birds.
3. The geographical range sizes of birds introduced to New Zealand did not depend on the length of time since they were introduced. Instead, large geographical ranges were exhibited by species whose preferred habitat is widespread in New Zealand, species with life history traits associated with higher rates of population growth (high fecundity, fast development and small body size), species that are partial migrants in part of their natural range, and species that were initially introduced more often and in greater numbers to New Zealand.
4. The strength and direction of geographical range size–life history correlations in introduced New Zealand and British breeding birds were very similar. There was also a strong positive correlation between the geographical range sizes of the species introduced to New Zealand from Britain and their geographical range sizes in Britain. However, the similarity of the correlations between life history traits and geographical range sizes in both regions was not a simple consequence of this; the similarity persisted when species introduced from Britain were excluded from the New Zealand data.
5. We discuss the implications of these results for understanding variation in geographical range sizes in the introduced New Zealand avifauna.
Introduction
There has recently been much interest in exploring relationships between life history traits (especially body size), and abundance and range size in animal assemblages (Damuth 1981; Nee et al. 1991; Cotgreave 1993; Taylor & Gotelli 1994; Brown 1995; Silva & Downing 1995; Blackburn, Lawton & Gregory 1996; Gaston & Blackburn 1996; Blackburn & Gaston 1997; Blackburn, Gaston & Gregory 1997). A focus of these studies is a search for consistent patterns in relationships between these variables that might uncover the underlying mechanisms that structure abundance and range size distributions.
There has been a proliferation of studies in this area, but unfortunately there is little indication that any consistent patterns are emerging; for example, despite initial optimism of a simple negative relationship between body size and abundance in animal taxa (Damuth 1981), published studies show relationships that range from linear negative to linear positive, and include nonlinear and complex ‘polygonal’ relationships (Blackburn & Gaston 1997). Nevertheless, it is still possible that the different forms of these relationships are manifestations of the same ‘true’ underlying abundance–body size relationship. Polygonal relationships are often suggested to be truncated portions of larger linear negative relationships that arise from sampling a restricted range of body sizes in smaller sample areas (Lawton 1989; Currie 1993). Indeed, there is growing evidence that the form of both abundance- and geographical range size–body size relationships depend crucially on the spatial scale of the investigation (Gaston & Blackburn 1996; Blackburn & Gaston 1997). Given these uncertainties, a search for consistent, underlying patterns in relationships between life history traits and abundance or range size may be most productive if we compare data that have been collected at the same spatial scale or that are similar in other ways that might confound the relationships of interest.
The interaction of body size with abundance and geographical range size has received much attention, probably because body size is an easily measured species attribute. However, the mechanisms suggested to explain observed relationships of body size with abundance and range size almost always assume that body size acts as a surrogate for other life history variables; for example, of the five mechanisms suggested to explain observed range size–body size relationships in Gaston & Blackburn (1996), four mechanisms assume that body size correlates with range size indirectly through its correlation with other life history variables. Thus, a more productive way of identifying what determines the abundances or range sizes of species within a geographical region may be directly to relate abundance or range size with the variables thought directly to affect them, rather than indirectly looking for those effects through correlations with body size (Blackburn et al. 1996).
Our aims in this study are twofold. First, we test for consistent relationships between species life history traits and range size by comparing interspecific correlations between these variables in two avian assemblages that occupy regions of similar geographical extent at similar latitudes. Geographic extent and latitude are factors that could confound such comparisons because there is evidence that range sizes vary with both (Stevens 1989; Anderson & Marcus 1992; Ricklefs & Latham 1992; Letcher & Harvey 1994; Smith, May & Harvey 1994; Gaston & Blackburn 1996; Blackburn, Gaston & Lawton 1998; Gaston, Blackburn & Spicer 1998). Identifying consistent patterns in range size–life history relationships in two geographically similar but independent regions would give us some confidence that we were observing a true underlying pattern. Second, we use the results of these range size–life history correlations to try to distinguish between the most likely mechanisms structuring range size distributions at the geographical scale of our study regions.
We chose for comparison the assemblages of introduced birds in New Zealand and breeding birds in Britain. Britain and the two main islands of New Zealand are of similar area, allowing us to discount the possibility that a comparison of the observed relationships will be confounded by differences in the spatial scale of the investigation. We can also discount a second factor, variation in latitude, that has been shown to complicate interpretation of range size–body size relationships in New World birds (Blackburn & Gaston 1996), because both Britain and New Zealand occupy similar temperate latitudes. Given this correspondence in the area and latitude of the study regions, we predict that if there are consistent, underlying relationships between range size and species life histories, then we should find similar patterns of variation in both regions.
Our comparison has the additional feature that the two regions share 19 species in common. Of the 34 species successfully introduced to New Zealand, 17 are native to Britain, which was the source of introductions for these species, while a further two species have been successfully introduced to both New Zealand and Britain from elsewhere. Hence, if range size is a species-level attribute, being a function of particular life history traits, then in addition to observing similar range size–life history correlations among the full sets of species, geographical range size in New Zealand should correlate with geographical range size in Britain for the 19 species common to both regions.
Introduced New Zealand birds offer additional advantages for distinguishing between the mechanisms that might be responsible for variation in geographical range size. The introduction of birds to New Zealand can be viewed as a large-scale experiment, beginning in 1842, in which 34 species of birds were successfully released onto two large islands and have since spread by varying amounts to occupy their present geographical range. The acclimatization societies, responsible for introducing the vast majority of birds, kept records of the timing of introductions and the numbers of individuals released (Thomson 1922; Lamb 1964). We use these historical records, along with data on species life histories, to test four hypotheses about the factors that might underlie range size variation in introduced New Zealand birds:
1. Some species may have had a longer period in which to disperse, establish and attain a larger geographical range size than others (Gaston 1994), which predicts a negative correlation between the year of species introduction and present range size.
2. Species distributions may be determined by habitat availability such that species able to exploit habitats that are widespread in a region will have a large geographical range size in that region, predicting a positive correlation between suitable habitat area and geographical range size (Gaston 1994). This hypothesis encompasses Brown's niche breadth hypothesis (Brown 1984), in that those species that are able to exploit a greater variety of habitats, particularly if those habitats are widespread, should have a larger geographical range size.
3. Some species may establish more successfully because after colonizing a site they have a faster population growth, are therefore less vulnerable to local extinction, and consequently are able to occupy more sites than other species (Gaston 1988; Hanski 1991; Hanski, Kouki & Halkka 1993; Holt et al. 1997). We do not have data on rates of population growth to test this hypothesis directly. Instead, we test for significant correlations between geographical range size and life history traits that are probable correlates of population growth rate, including clutch size, number of broods per season, and development times.
4. Individuals of some species may have a greater propensity to move about the landscape, be recorded at more locations and so attain a greater geographical range size. We test this prediction in two ways: first, by searching for a positive correlation between range size and migratory tendency and second, by testing whether range size is positively correlated with breeding and natal dispersal distances.
Methods
Data sources
The 34 species of birds successfully introduced to New Zealand were identified from Turbott (1990), and data on geographical range size (the number of 9·14 × 9·14-km grid squares occupied) were taken from Bull, Gaze & Robertson (1985). Data on the geographical range size of breeding birds in Britain (the number of 10 × 10-km grid squares occupied) were taken from Gibbons, Reid & Chapman (1993). We excluded from the British dataset seabirds, introduced species and species that did not breed in Britain every year from 1981 to 1990.
Data on life history traits were taken from the following sources. For the birds introduced to New Zealand, Veltman, Nee & Crawley (1996) was the source for data on body mass, clutch size and the number of broods per season. For British birds, data on body mass came from the sources cited in Blackburn et al. (1996) and Bennett (1986) was the source for data on clutch size and the number of broods per season. For all species, Bennett (1986) was the source for data on egg mass, clutch size, clutch mass, hatchling mass, incubation time, fledging time, age at offspring independence from parent, age at first breeding and species maximum lifespan, except for five species introduced to New Zealand where data for incubation time, fledging time and age at first breeding were taken from Heather & Robertson (1996). We derived two additional measures from these data: maximum lifetime reproduction of a species (the product of egg mass, clutch size, number of broods per season and maximum lifespan) and annual clutch mass (clutch mass multiplied by number of broods per season). For all species, estimates of juvenile and adult survivorship were taken from Dobson (1990) and Sæther (1989), and data on resting metabolic rate (RMR) were taken from Bennett & Harvey (1987), ignoring the values these authors excluded from their analyses. Note that many of the above attributes are traits that correlate with population growth rate. Species with short development times (short intervals for incubation, fledging, age at independence and age at first breeding) and high fecundity (large clutch sizes, many broods per season) should have high population growth rates. Also note that for the species introduced to New Zealand, the vast majority of life history traits used in our analyses were measured outside New Zealand in the species’ natural ranges.
We calculated the area (as a percentage of total land area) of seven major habitat types in New Zealand (Table 1; these habitats comprise more than 90% of New Zealand's total land area) using the data in Newsome (1987). We quantified the propensity of the species introduced to New Zealand to utilize these seven habitats by scoring them as 1 for frequent use and 0 for infrequent use based on descriptions of habitat use in their natural ranges (from Witherby et al. 1941; Whistler 1963; Macdonald 1973; Cramp & Simmons 1977–93). For each species, we estimated the area (as a percentage of total land area) of suitable habitat in New Zealand by summing the area of habitat scored as frequent use.
| % of total New Zealand land area |
|
|---|---|
| Lowland—montane habitats | |
| Forest and closed shrubland | 32 |
| Unimproved grassland mixed with shrubland or forest | 16 |
| Unimproved grassland | 10 |
| Cultivated or developed pasture mixed with forest | 3 |
| Cultivated or developed pasture and arable land | 24 |
| Wetland and inland aquatic | 2 |
| Montane—alpine habitats | |
| Unimproved grassland and shrubland | 5 |
For the birds introduced to New Zealand, we used two measures for their propensity to move about the landscape: (i) the measure of migratory tendency in Veltman et al. (1996); species were scored as either residents in their natural range or as partial migrants if they migrate in only a part of their natural range (no obligatory migrants were successfully introduced to New Zealand, see Veltman et al. 1996); and (ii) the arithmetic means of breeding and natal dispersal distances in Britain from Paradis et al. (1998). The year of first recorded introduction of a species to New Zealand was taken from Thomson (1922) and additional sources cited in Duncan (1997). Veltman et al. (1996) was the source for the total number of introduction attempts in New Zealand and the total number of individuals of each species released.
Analysis
It is becoming routine in comparative studies to control for phylogenetic relatedness among species. This is because closely related species may share a similar suite of traits through common ancestry, not because those traits evolved independently in each species. Treating species as independent data points in a comparative study may therefore overestimate the number of times a relationship between traits has evolved, and could inflate the statistical significance of the relationship (see Harvey & Pagel 1991). Phylogenetic comparative methods overcome this problem of nonindependence by controlling for phylogenetic relatedness among species.
Geographical range size of the introduced New Zealand birds, however, is entirely the outcome of processes operating within the 160 years since introductions began. Species in New Zealand could therefore be treated as independent data points in range size–life history correlations, because these correlations are not confounded by shared ancestry. Nevertheless, phylogenetically related species might have similar range sizes because, through common ancestry, they share traits that influence range size, but these traits were not evaluated in the present study. Showing that a particular trait that correlates strongly with range size across species, also correlates strongly with range size independent of phylogeny would suggest that the trait in question was a robust correlate of range size, independent of other shared attributes. Therefore, we performed two types of analyses on the data. First, we compared correlations between geographical range size and life history traits, treating species as independent data points (across-species analyses). Second, we compared range size–life history correlations using the method of independent contrasts (Felsenstein 1985) to control for phylogenetic relatedness among species (within-taxon analyses).
Phylogenetic comparative methods require a hypothesis about the evolutionary relationships among taxa. We classified species using the phylogeny of Sibley & Ahlquist (1990), supplemented by the classification of Sibley & Monroe (1990) where Sibley & Ahlquist (1990) did not resolve groups to species (Fig. 1). We used this phylogeny while aware of the criticisms (e.g. Houde 1987; Sarich, Schmid & Marks 1989; Harsham 1994; but see Mooers & Cotgreave 1994). Despite the potential biases and subjectivity, it is currently probably the best avian phylogeny available.
Hypothesized phylogenetic relationships among the species of introduced New Zealand birds. Nomenclature follows Sibley & Monroe (1990).
The method of independent contrasts calculates the differences between trait values at each node on the phylogenetic tree. Each contrast then represents a phylogenetically independent event, because the taxa used in the calculation of each contrast share an immediate common ancestor. Thus, if two variables are correlated independent of phylogeny, then the independent contrasts should be correlated. Trait values at ancestral nodes are estimated as the mean of the trait values in each branch of the node, weighted by branch length (Felsenstein 1985). Because we lacked data for all branch lengths, we set branch lengths as equal. The independent contrasts for continuous variables (everything except migratory tendency) were calculated using the crunch option in caic (Purvis & Rambaut 1994). Relationships between these sets of contrasts were then analysed using ordinary least squares regression, with the regression line forced through the origin (Garland, Harvey & Ives 1992).
For migratory tendency, which was a dichotomous rather than a continuous variable, the independent contrasts were calculated using the brunch option in caic (Purvis & Rambaut 1994). If there is no difference in range size between partial migrants and resident species, then the independent contrasts should show no significant tendency to be either positive or negative. We tested for such a tendency using a one-sample t-test against an expected value of zero (Purvis & Rambaut 1994).
Results
Across-species analysis
Many life history traits correlated significantly with geographical range size in both the introduced New Zealand and the British birds (Table 2). Except for juvenile survivorship, the strength and direction of the range size–life history correlations in both regions were very similar (Fig. 2a; note that data on juvenile survivorship were available for only eight of the 34 species introduced to New Zealand). The general pattern was that species with traits associated with high population growth rates (small, rapidly developing species with high mortality and fecundity) had the largest geographical range sizes in both regions.
| New Zealand r |
n | Britain r |
n | |
|---|---|---|---|---|
| Log body mass | –0·621*** | 34 | –0·209** | 155 |
| Log RMR | –0·735*** | 17 | –0·319** | 80 |
| Log adult survivorship | –0·726** | 16 | –0·311** | 87 |
| Log juvenile survivorship | 0·467 | 8 | –0·349* | 41 |
| Log egg mass | –0·603*** | 33 | –0·278*** | 151 |
| Log clutch size | –0·063 | 34 | 0·058 | 151 |
| Log clutch mass | –0·524** | 33 | –0·264*** | 151 |
| Log annual clutch mass | –0·564** | 27 | –0·220** | 140 |
| Log hatchling mass | –0·729*** | 18 | –0·399*** | 72 |
| Log incubation time | –0·657*** | 34 | –0·312*** | 142 |
| Log fledging time | –0·561** | 27 | –0·342*** | 132 |
| Log age at independence | –0·391 | 11 | –0·137 | 77 |
| Log age at first breeding | –0·493** | 27 | –0·244** | 123 |
| Log lifespan | 0·055 | 26 | –0·034 | 124 |
| Log lifetime reproduction | –0·546** | 24 | –0·239** | 116 |
| Log broods per season | 0·632*** | 28 | 0·228** | 140 |
| Log breeding dispersal distance | –0·501* | 16 | – | – |
| Log natal dispersal distance | –0·530* | 16 | – | – |
| Log suitable habitat area in New Zealand | 0·338* | 34 | – | – |
| Log number of introductions | 0·466* | 25 | – | – |
| Log total number of birds released | 0·574** | 25 | – | – |
| Year of first recorded introduction | –0·184 | 26 | – | – |
- *P < 0·05; **P < 0·01; ***P < 0·001.
- RMR, resting metabolic rate.
Comparison of (a) across-species, and (b) within-taxon range size–life history correlation coefficients in New Zealand and Britain, for the 16 life history traits in Tables 2 and 3 with correlations for both regions. In both graphs ○ represents juvenile survivorship. The across-species, but not the within-taxon, New Zealand and British correlation coefficients were significantly correlated (across-species: r = 0·67, P < 0·01; within-taxon: r = 0·16, P = 0·55) when all 16 traits were included, but both were significantly correlated when the outlier juvenile survivorship was excluded (across-species: r = 0·95, P < 0·001; within-taxon: r = 0·59, P = 0·02).
In New Zealand, species that were partial migrants in their natural range had a significantly greater range size than species that were resident in their natural range (one-way anova; F1,32 = 40·5, P < 0·001). In addition, the area of suitable habitat, both measures of introduction effort, and breeding and natal dispersal distance were significant correlates of range size, but the year of first recorded introduction was not (Table 2).
Within-taxon analysis
At first glance, controlling for phylogeny appeared to weaken the relationships between range size and life history traits; most traits that were significant correlates of range size across-species were not significant correlates within-taxon in both New Zealand and Britain (Table 3). However, the changes in range size–life history correlations resulting from the different methods of analysis were not simply the result of weaker within-taxon relationships.
| New Zealand r |
n | Britain r |
n | |
|---|---|---|---|---|
| Log body mass | –0·298 | 33 | 0·194* | 101 |
| Log RMR | –0·478 | 16 | –0·011 | 60 |
| Log adult survivorship | –0·545* | 15 | –0·114 | 66 |
| Log juvenile survivorship | 0·521 | 7 | –0·308* | 33 |
| Log egg mass | –0·325 | 32 | 0·172 | 99 |
| Log clutch size | 0·355* | 33 | 0·137 | 99 |
| Log clutch mass | –0·167 | 32 | 0·253* | 99 |
| Log annual clutch mass | 0·059 | 26 | 0·241* | 96 |
| Log hatchling mass | –0·581* | 17 | 0·020 | 56 |
| Log incubation time | –0·144 | 33 | 0·014 | 96 |
| Log fledgling time | –0·210 | 26 | –0·105 | 90 |
| Log age at independence | –0·471 | 10 | 0·084 | 58 |
| Log age at first breeding | –0·210 | 26 | –0·015 | 87 |
| Log lifespan | 0·310 | 25 | 0·371*** | 84 |
| Log lifetime reproduction | 0·125 | 23 | 0·397*** | 80 |
| Log broods per season | 0·418* | 27 | 0·145 | 96 |
| Log breeding dispersal distance | –0·232 | 15 | – | – |
| Log natal dispersal distance | –0·137 | 15 | – | – |
| Log suitable habitat area in New Zealand | 0·051 | 33 | – | – |
| Log number of introductions | 0·860*** | 24 | – | – |
| Log total number of birds released | 0·768*** | 24 | – | – |
| Year of first recorded introduction | –0·361 | 25 | – | – |
- *P < 0·05; **P < 0·01; ***P < 0·001.
- RMR, resting metabolic rate.
If controlling for phylogeny had no effect on range size–life history correlations, then a plot of the across-species and within-taxon correlation coefficients should form a line that passed through the origin with slope equal to one. If the only effect of phylogenetic correction was to weaken range size–life history correlations (i.e. to shift both positive and negative values towards zero, as phylogenetic correction might be expected to do; Ricklefs & Starck 1996), then the resulting plot should form a scatter of points along a line that still passed through the origin but with a slope less than one. We tested whether this was the case by fitting a line through the values of the within-taxon and the across-species correlation coefficients in each region (Table 4). Because both sets of correlation coefficients are subject to measurement error, we fitted a line using the reduced major axis method (McArdle 1988). The slopes of the fitted lines in both regions did not differ significantly from one. However, in both regions the intercept terms were significantly positive, implying that rather than weakening correlations, controlling for phylogeny had made the across-species correlations consistently more positive within-taxa. This was most apparent in the British data where all 13 significant across-species correlations were negative, while five of the six significant within-taxon correlations were positive (compare Tables 2 and 3). Of the five significant positive within-taxon correlations, all were negative across-species, four of them (body mass, clutch mass, annual clutch mass and lifetime reproduction) significantly so. Hence, these four traits showed a significant pattern reversal after controlling for phylogeny.
| d.f. | Parameter estimate |
Standard error |
t-value | P | |
|---|---|---|---|---|---|
| New Zealand | |||||
| Intercept | 1 | 0·217 | 0·069 | 3·15 | < 0·01 |
| Slope | 1 | 0·845 | 0·123 | 1·26 | 0·23 |
| Britain | |||||
| Intercept | 1 | 0·327 | 0·071 | 4·58 | < 0·001 |
| Slope | 1 | 1·115 | 0·269 | 0·43 | 0·67 |
The changes were not as readily apparent in the New Zealand data, largely because most of the across-species correlations were more strongly negative to begin with, so that many of these correlations remained negative within-taxa. Nevertheless, the changes in both regions were very similar. For the same set of variables, phylogenetic correction resulted in the across-species correlation coefficients increasing by an average of 0·30 in Britain and 0·28 in New Zealand. The change in range size–life history correlations as a consequence of phylogenetic correction were themselves correlated between regions (r = 0·68, P < 0·001, n = 16) and of similar magnitude. Lifetime reproduction, for example, showed the greatest change in both regions with phylogenetic correction, increasing by 0·636 (from –0·239 to +0·397) in Britain and 0·671 (from –0·546 to +0·125) in New Zealand. With the exception of juvenile survivorship, there remained a significant within-taxon relationship between the range size–life history correlations in New Zealand and Britain (Fig. 2b).
In New Zealand, the two measures of introduction effort were robust predictors of range size, being significant correlates in both the across-species and within-taxon analyses. Indeed, both measures of introduction effort were stronger correlates after controlling for phylogeny than when species were treated as independent data points. Migratory tendency was also a significant correlate across-species and after controlling for phylogeny; in the within-taxon analysis, species that were partial migrants in their natural range had larger range sizes in New Zealand compared to species resident in their natural range (t = 2·27, d.f. = 5, P = 0·02).
Range size correlation between birds resident in both regions
Range size in New Zealand was significantly correlated with range size in Britain for the 19 species reported breeding in both regions (Fig. 3; r = 0·72, P < 0·001, n = 19), suggesting that range size may be a persistent species-level characteristic. The correlation between range size in New Zealand and Britain was even stronger after controlling for phylogeny (r = 0·88, P < 0·001, n = 18). We have also shown, however, that both the number of introductions and the total number of birds released were robust predictors of range size in New Zealand (Tables 2 and 3). Furthermore, Duncan (1997) has shown that, for the passeriform birds introduced to New Zealand from Britain, these two measures of introduction effort were positively correlated with total population size, and consequently with geographical range size, in Britain. Greater effort was made to introduce to New Zealand birds that were abundant and widespread in Britain, perhaps because these birds were readily captured for export or because they were desirable species (Duncan 1997). Hence, a positive correlation between range size in New Zealand and range size in Britain could arise indirectly in the following way: if species with large ranges in Britain were introduced to New Zealand more often and in greater numbers, then they would have had larger founding populations than species with small British ranges. If a large initial population size gave species an advantage that subsequently enabled them to attain a greater range size in New Zealand, perhaps by allowing them to pre-empt resources, then we would obtain a positive correlation between range size in Britain and range size in New Zealand that was not the direct result of range size being a characteristic linked to species life history traits.
Comparison of geographical range size in New Zealand and Britain, for the species confirmed breeding in both regions (r = 0·72, P < 0·001, n = 19). ○ represents species recorded in both New Zealand and Britain that were introduced from elsewhere (Branta canadensis and Phasianus colchius; correlation when these two species were excluded: r = 0·72, P < 0·001, n = 17).
We attempted to unravel the potentially confounded relationships between life history traits, introduction effort, and range size in both Britain and New Zealand. First, using multiple regression we identified a reduced set of attributes that could explain most of the variation in New Zealand range size (Table 5; note that many life history traits were highly cross-correlated, thus, while the number of broods per season was included in the multiple regression model, other traits associated with high fecundity or fast development could have substituted for this variable with little loss of explanatory power. A model including body mass, migratory tendency and number of individuals released, for example, had r2 = 0·77.) We then tested a set of causal hypotheses describing the relationships between this reduced set of attributes and range size in Britain and New Zealand, using path analysis (Fig. 4). Path analysis is an appropriate technique in this situation because we are interested in teasing apart correlations that result from indirect pathways. Of the two life history traits in the reduced set, only migratory tendency was a good predictor of range size in both Britain and New Zealand in the path model, although both traits indirectly affected range size through their correlation with the other trait. Nevertheless, Fig. 4 suggests that the significant correlation between range size in New Zealand and range size in Britain was in part a direct consequence of range size being a function of at least one life history trait (migratory tendency), and in part an indirect consequence of variation in historical introduction effort, that was, in turn, correlated with range size in Britain.
| Variable | r 2 | F |
|---|---|---|
| Log broods per season | 0·59 | 31·2*** |
| Migratory tendency | 0·14 | 10·8** |
| Log total number of birds released | 0·07 | 7·2* |
| Total | 0·80 |
- * P < 0·05;
- ** P < 0·01;
- *** P < 0·001.
Path diagram of hypothesized causal relationships among migratory tendency, log number of broods per season, log number of individuals released in New Zealand, and geographical range size in both Britain and New Zealand, for 18 of the 19 species confirmed breeding in both regions (Columba livia was excluded from this analysis because there are no complete records of the number of individuals released in New Zealand). U indicates the proportion of total variation in the dependent variable that is not explained by the combination of independent variables used in the analysis.
Discussion
To date, no consensus has emerged on the form of the relationship between species life history traits (notably body size) and range size in animal assemblages (Gaston & Blackburn 1996). We asked if there were consistent patterns in range size–life history correlations in avian assemblages in two regions of similar geographical extent and latitudinal position, using comparable measures of range size. With the exception of juvenile survivorship, range size and life history traits showed the same pattern of covariation across species in both our study regions, and a consistent, although different pattern of covariation after controlling for phylogeny. That correlations between juvenile survivorship and range size were not consistent between regions is not surprising; it was the trait for which we had by far the fewest data so that we would expect relationships with range size to be the least accurately replicated between regions.
The repeatability of range size–life history correlations between regions gives us some confidence that the relationships we observed reflect true, underlying patterns. However, the consistency in our results may have been inflated through the overlap in species composition, coupled with our use of the same life history data for those species common to both regions. Consequently, once we found a significant correlation between range sizes in New Zealand and Britain, similar correlations of range size with life history traits might have been expected. This does not detract from the interesting result that range sizes are correlated in the first place, but it does suggest that the similarities in range size–life history correlations may simply arise as a consequence of this first outcome (or vice versa). A closer inspection of our data shows this is not entirely the case. The introduced New Zealand birds are not a complete subset of the British birds; there were 15 species successfully introduced to New Zealand that do not breed in Britain. Hence, it is possible to compare range size–life history correlations in two independent groups by excluding from our New Zealand data those species also found in Britain. Across-species range size–life history correlation coefficients were still significantly correlated between New Zealand and Britain using this subset of the species introduced to New Zealand (r = 0·84, P < 0·001, n = 12; this analysis included only those 12 traits for which there were five or more data points in New Zealand). Thus, while similar range size–life history correlations are expected as a consequence of a significant correlation between range sizes in New Zealand and Britain for the 19 shared species, this is not the sole cause of the similarities; the 15 species introduced to New Zealand that are not found in Britain contribute independently to the same pattern.
Nevertheless, the consistency in our range size–life history correlations may have been inflated by the use of the same life history data for those species common to both regions. Ideally, we would have used life history data gathered independently from populations in each of the regions where we compared range size–life history correlations. Unfortunately, there are probably insufficient data on enough life history traits for introduced birds in New Zealand to permit a robust comparison. However, we do know that for introduced passerine birds, clutch sizes in New Zealand are consistently lower than in Britain, but that clutch sizes in the two regions are highly correlated (r = 0·91, n = 9, data from Niethammer 1970). Hence, the different environment in New Zealand has resulted in only a slight rearrangement of clutch sizes relative to Britain. If this holds true for other life history traits, then the New Zealand range size–life history correlations reported in this study (primarily using life history data from species’ natural ranges) should be similar to the correlations we would obtain if we used life history data collected in New Zealand.
A final confounding influence is that the similarity in range size–life history correlations between New Zealand and Britain could reflect variation in the accuracy of life history estimates. Consistent differences in the strength of correlations could arise because some traits were measured using more or less accurate data than others, the assumption being that all correlations would be equally strong given perfect data. We consider this situation unlikely and note that contrary to expectation, clutch size (a trait that is likely to be accurately measured) was one of the weakest correlates of range size across species (Table 2). In summary, despite the potential for the overlap in species composition and for our use of the same life history data artificially to inflate the consistency of range size–life history correlations, the available evidence suggests that such effects were minor and that the consistent outcomes we observed reflect a robust underlying pattern in range size variation.
Given that we find patterns in across-species range size–life history correlations that are consistent between New Zealand and Britain, what do these patterns tell us about the factors that might underlie range size variation? For the birds introduced to New Zealand, we can reject the first of the four hypotheses outlined in the Introduction. Range size was not significantly correlated with date of introduction, so there was no evidence that some species were more widely distributed because they had more time to disperse, establish and attain a larger range size. Indeed, following introduction the spread of most introduced birds in New Zealand appears to have been rapid, such that by the early 1900s most species were occupying close to their present range sizes (Thomson 1926; Heather & Robertson 1996).
Our results support two of the three remaining hypotheses and provide conflicting answers to the third. First, species whose preferred habitat was abundant in New Zealand tended to have a large geographical range size. More specifically, the species with the largest range sizes were those able to exploit the variety of human-modified, predominantly farmland habitats (unimproved grassland, developed pasture and cultivated land) that comprise 53% of the total land area (Table 1). Second, species with large geographical range sizes were those with high fecundity and fast development, traits associated with fast population growth rate. Fast developing species with high fecundity should have a greater chance of establishing new populations, and consequently have a larger geographical range size, because fast population growth rate will ensure that they are less vulnerable to local extinction when colonizing unoccupied sites (Gaston 1988; Hanski 1991; Hanski et al. 1993; Holt et al. 1997). This should especially apply in extensive farmland habitats where local patches are frequently disturbed by grazing, burning and cultivation and are thus repeatedly available for recolonization. Hence, the finding that suitable habitat area (largely measuring the degree to which species exploit human-modified habitats) and traits associated with fast population growth rate are both correlated with geographical range size, and are themselves correlated (suitable habitat area did not explain additional variation in range size after controlling for the number of broods per season and migratory tendency; Table 5), is consistent with a single mechanism: in predominantly human-managed, regularly disturbed landscapes, high population growth rate can generate high site occupancy and, consequently, a large geographical range size. In addition to similar geographical area and latitude, the repeatability of range size–life history correlations between New Zealand and Britain may reflect the similarity and extent of highly disturbed, artificial habitats in both regions.
Our results were equivocal with regard to the hypothesis that species showing a greater propensity to move about the landscape should have larger geographical range sizes. Somewhat counter-intuitively, those species with greater breeding and natal dispersal distances in Britain tended to have smaller range sizes in New Zealand. In Britain, Paradis et al. (1998) found a similar negative correlation, but suggested that breeding and natal dispersal distances were a consequence rather than a cause of range size variation; species with large geographical ranges were those able to exploit the most abundant habitats and, consequently, they did not need to disperse far to find suitable sites for reproduction. More predictably, range size was strongly correlated with migratory tendency, and this correlation persisted after controlling for other life history traits (Table 5). Superficially, the most likely explanation for this relationship is that species that migrate in part of their natural range may winter and breed in different parts of New Zealand and therefore tend to have larger range sizes than resident species. However, banding studies have found no evidence of regular long-distance movement in the partial migrants introduced to New Zealand (Heather & Robertson 1996). In the absence of this mechanism, it is not clear why range size is so strongly correlated with migratory tendency. Intriguingly, of the 13 partial migrants introduced to New Zealand, 11 have dispersed to at least one of the nine outlying island groups greater than 95 km from mainland New Zealand, compared with only three of the 21 resident species (from distributional data in Williams 1953; Heather & Robertson 1996). It is possible that migratory tendency reflects a propensity for long-distance dispersal, as opposed to the relatively short mean distances (less than 35 km) associated with breeding and natal dispersal. Species with high rates of long-distance dispersal may have wider distributions because they are capable of bridging large patches of unsuitable habitat and because they can repeatedly reoccupy widely scattered localities at which they would otherwise go locally extinct (the ‘rescue effect’; Brown & Kodric-Brown 1977).
Unexpectedly, we found a significant correlation between range size and the effort put into the initial introduction of birds into New Zealand; a relationship that persisted after controlling for the correlation between range size and life history traits (Table 5, Fig. 4). We are unsure how to interpret this result, but have already mentioned a possible explanation: species with large founding populations may have initially captured a greater proportion of any shared resources from species with smaller founding populations. This initial advantage could have compounded itself: those species initially able to capture a greater share of resources would have had a faster population growth and rate of spread, allowing them further to pre-empt resources at newly colonized sites as their ranges expanded. In addition, the species with larger founding populations were released at more sites throughout the country (Veltman et al. 1996; Duncan 1997), giving them a further headstart in range expansion and resource pre-emption. Interestingly, we would expect this effect to be most pronounced among closely related species that would compete for similar resources, precisely the result we found; the within-taxon correlations between geographical range size and both measures of introduction effort were stronger than the across-species correlations. This explanation implies that competitive interactions play a role in limiting range sizes, and that the outcome of those interactions depends on the initial starting conditions. Indeed, whatever the mechanism generating the correlation between range size and introduction effort, this relationship suggests that, in addition to being predictable from aspects of species biology, present range size variation also includes a component resulting from historical circumstances.
Finally, we found that range size–life history relationships varied with phylogenetic relatedness in a similar way in both regions. Our finding that in Britain range size correlated negatively with body size across-species but positively within-taxa mirrors a result found for the abundance–body size relationship for British birds (Nee et al. 1991). An explanation for these positive within-taxon correlations is that among closely related, ecologically similar species, large body size could be an advantage in interspecific competition, leading to lowered abundances and reduced range sizes in smaller-bodied species (Nee et al. 1991). It is plausible that the absence of a similar within-taxon pattern in New Zealand was a result of the overwhelming influence of initial population size on the outcome of competitive interactions between closely related species (see above). Nevertheless, controlling for phylogeny caused only a moderate shift in range size–body size correlations compared with other life history traits; lifetime reproduction, for example, showed a much stronger pattern reversal in both regions. Quite how, and why, abundance– and range size–life history relationships depend on phylogenetic relatedness remains unclear.
Acknowledgements
R. Duncan and C. Veltman were supported by a grant from the New Zealand Marsden Fund. T. Blackburn is a Leverhulme Special Research Fellow, and was supported in part by NERC grant GST/03/1211. We thank Frances Schmechel and two anonymous referees for insightful comments.
