Dispersal distances predict subspecies richness in birds

Authors

  • Belliure,

    1. Laboratoire d’Ecologie, CNRS UMR 7625, Université Pierre et Marie Curie, Bât. A, 7ème étage, 7 quai St. Bernard, Case 237, F-75252 Paris Cedex 05, France
    Search for more papers by this author
  • Sorci,

    1. Laboratoire d’Ecologie, CNRS UMR 7625, Université Pierre et Marie Curie, Bât. A, 7ème étage, 7 quai St. Bernard, Case 237, F-75252 Paris Cedex 05, France
    Search for more papers by this author
  • Møller,

    1. Laboratoire d’Ecologie, CNRS UMR 7625, Université Pierre et Marie Curie, Bât. A, 7ème étage, 7 quai St. Bernard, Case 237, F-75252 Paris Cedex 05, France
    Search for more papers by this author
  • Clobert

    1. Laboratoire d’Ecologie, CNRS UMR 7625, Université Pierre et Marie Curie, Bât. A, 7ème étage, 7 quai St. Bernard, Case 237, F-75252 Paris Cedex 05, France
    Search for more papers by this author

Belliure Dr Laboratoire d’Ecologie, CNRS UMR 7625, Université Pierre et Marie Curie, Bât. A, 7ème étage, 7 quai St. Bernard, Case 237, F-75252 Paris Cedex 05, France. Tel.: +33 1 44 27 38 23; fax: +33 1 44 27 35 16; e-mail: jbelliur@snv.jussieu.fr

Abstract

Dispersal ability has been hypothesized to reduce intraspecific differentiation by homogenizing populations. On the other hand, long-distance dispersers may have better opportunities to colonize novel habitats, which could result in population divergence. Using direct estimates of natal and breeding dispersal distances, we investigated the relationship between dispersal distances and: (i) population differentiation, assessed as subspecies richness; (ii) ecological plasticity, assessed as the number of habitats used for breeding; and (iii) wing size, assessed as wing length. The number of subspecies was negatively correlated with dispersal distances. This was the case also after correcting for potential confounding factors such as migration and similarity due to common ancestry. Dispersal was not a good predictor of ecological plasticity, suggesting that long-distance dispersers do not have more opportunities to colonize novel habitats. Residual wing length was related to natal dispersal, but only for sedentary species. Overall, these results suggest that dispersal can have a homogenizing effect on populations and that low dispersal ability might promote speciation.

Introduction

A central challenge for organismal biologists is to establish links between ecological and behavioural traits of individuals and the evolution of species. One such link is expected between dispersal ability and the magnitude over which populations differ genetically or ecologically. It is well known that dispersal is a life history trait that has profound effects on populations, both from an ecological and from an evolutionary perspective ( Johnson & Gaines, 1990). As a demographic process, dispersal plays an important role in the distribution of organisms; with its potential for gene flow, dispersal exerts a crucial effect on the genetic structure of populations. Recent theoretical work has shown that dispersal may be associated with population divergence and with ecological plasticity, but results are controversial. Tendency to diversification appears linked to dispersal propensity in some cases ( Schluter, 1996; Doebeli & Ruxton, 1997; Smith et al., 1997 ; Losos et al., 1998 ; Owens et al., 1999 ), but to low dispersal propensity in other circumstances ( Zera, 1981; Liebherr, 1986; Baker et al., 1995 ; Bohonak, 1999). Ecological plasticity in its turn has been theoretically associated with dispersal and finding new environmental conditions ( Rosenzweig, 1995; Owens et al., 1999 ), but also to restricted dispersal ( Murrenet al., in press).

Long-distance dispersal has been shown to be important in invasion processes ( Shaw, 1995), and in the genetic structuring of populations ( Ibrahim et al., 1996 ). It is known that populations colonizing novel environments can evolve extremely rapidly ( Carrol et al., 1997 ; Losos et al., 1997 ; Reznick et al., 1997 ), as normal genetic variants may accumulate more rapidly in small than in large populations ( Kimura, 1979). Therefore, invasion processes may play a prominent role in speciation ( Schluter, 1996; Doebeli & Ruxton, 1997; Smith et al., 1997 ; Losos et al., 1998 ; Owens et al., 1999 ). In their study, Owens et al. (1999 ) showed dispersal ability to be associated with indices of ecological generalism, supporting Rosenzweig’s (1995) ‘geographical’ model of diversification whereby the probabilities of a lineage becoming species rich is closely associated with its chances of finding and successfully colonizing new areas.

On the other hand, low dispersal ability can facilitate geographical isolation and divergence. In a recent study on different animal groups including marine, freshwater and terrestrial invertebrates, increased differentiation among populations was associated with decreased dispersal ability ( Bohonak, 1999). Studies on flightless insects ( Zera, 1981; Liebherr, 1986) and birds ( Baker et al., 1995 ) reveal high levels of genetic differentiation among populations and low genetic variability. Restricted dispersal ability has been associated with increased ecological plasticity, although theoretical results are controversial ( Murrenet al., in press).

Life history, morphological, behavioural and habitat-associated traits all contribute to dispersal ability, and can provide the basis for studies investigating long-term consequences of dispersal. For example, wing morphology in birds is associated with migratory abilities ( Pennycuick, 1975). Over the past two decades, many reviews have attempted to relate genetic variation in natural systems to a variety of individual, population and species-level traits ( Selander, 1976; Nevo, 1978; Sole Cava & Thorpe, 1991; Palumbi, 1994). However, the relationships among population structure, indirect estimates of gene flow and dispersal remain ambiguous and resistant to generalization (e.g. Larson et al., 1984 ; Liebherr, 1988; Boileau et al., 1992 ; Pogson et al., 1995 ).

Dispersal can occur at two different moments in the life of individuals. Natal dispersal is the movement an individual makes from its birth site to the place where it reproduces; breeding dispersal is the movement from one home range to another between attempts at reproduction (following the definitions by Greenwood & Harvey, 1982). The circumstances and the decisions involved in each of these dispersal events are presumably not the same, and the proximate mechanisms inducing individuals to leave may also be different. For example, natal dispersal in birds appears to be related to body condition ( Nilsson & Smith, 1985; Ferrer, 1992) and plasma levels of corticosterone ( Belthoff & Dufty, 1995, 1998; Silverin, 1997). Any attempt to describe the possible consequences of dispersal should take these considerations into account, as observations related to one or the other type of dispersal may have different interpretations.

Using direct estimates of natal and breeding dispersal distances ( Paradis et al., 1998 ), we attempt here to identify consequences of dispersal for bird species with variable ecological characteristics and migratory habits. We investigated the relationship between dispersal distances and (i) population differentiation (assessed as the relative number of subspecies), (ii) ecological plasticity (assessed as the number of habitats used for breeding) and (iii) wing size (assessed as wing length).

Materials and methods

Data collection

We used dispersal distances of 75 bird species breeding in Britain, as reported in Paradis et al. (1998 ). Data on dispersal distances, to the nearest kilometre, were collected by the British Trust for Ornithology during the period 1909–94. Natal dispersal was estimated using birds ringed in their year of birth and recovered at breeding age; breeding dispersal was estimated using birds ringed at breeding age (for more details on compilation of dispersal distances, see Paradis et al., 1998 ).

Likelihood for population diversification was computed as the number of subspecies described for each species. Data were extracted from Howard & Moore (1991).

Ecological plasticity was estimated as the number of different habitats in which each species has been known to breed. To attain greater precision and completeness, we recorded the habitat preferences that Cramp & Simmons (1977–1994) provide for each species. As the authors define the principal terms used in the glossary, to minimize ambiguity and unconscious overlap of terms in our data, we have considered a habitat only when it appeared in their glossary.

Paradis et al. (1998 ) showed an influence of body size, population size, breeding range and migratory status on dispersal distances. We took these factors into account in the regression models. Body size, as well as wing length, were extracted from Cramp & Simmons (1977–1994) as the averages for both sexes. Population size was estimated as the number of pairs breeding in Britain and Ireland, as reported in Paradis et al. (1998 ) from Gibbons et al. (1993 ). Breeding range was estimated according to the distribution maps in Cramp & Simmons (1977–1994). The total number of latitudinal degrees and the product of the total number of latitudinal and longitudinal degrees at which the species have been identified to breed covary to a large extent for the species considered (r=0.86, n=75, P < 0.0001). Thus, the latitudinal range of breeding area was considered as a measure of the breeding range. Species were categorized as migrant (i.e. trans-Saharan migrants) or sedentary, as reported in Paradis et al. (1998 ).

Statistical analyses

We considered the geometric mean of the natal and breeding dispersal distances, as reported in Paradis et al. (1998 ). Distances were log10-transformed before analysis.

Stepwise multiple regression analyses (Proc. Reg. SAS 1990) were conducted to investigate the effect of dispersal distances and potential covariates on the number of subspecies and the number of breeding habitats. Through this exploratory statistical technique ( Tabachnick & Fidell, 1989), we explored if dispersal distances constitute a good predictor of subspeciess richness and diversity of breeding habitats. Although natal and breeding dispersal covary to a large extent (r=0.76, n=67, P < 0.0001; Fig. 1), they were treated separately, because they reflect behavioural decisions taken at quite different moments in the avian life cycle.

Figure 1.

 Mean breeding dispersal distance (km) in relation to mean natal dispersal distance (km). N=67.

When the number of subspecies was regressed on dispersal distance, the covariates considered were breeding range (hereafter range), body mass, number of breeding habitats and population size. When the number of breeding habitats was regressed on dispersal distance, the covariates considered were range, body mass and population size. As migration has a significant effect on dispersal, with migrant species dispersing further than sedentary species ( Paradis et al., 1998 ), a second set of analyses was performed using migration as a covariate.

To cope with the problem of dependence among species due to common ancestry, we analysed the same multiple regression models using standardized contrasts for each variable according to the method described by Felsenstein (1985). Information on phylogenetic relationships was obtained from Martin & Clobert (1996). They constructed their phylogenetic hypothesis based on Sibley & Ahlquist (1990) and on the most recent information available for North America and Europe (see Martin & Clobert, 1996, for more details on this information). The only polytomy considered in the study was for three species of the genus Turdus, and it was resolved following Pagel (1992). We adopted the software CAIC ( Purvis & Rambaut, 1995) to calculate standardized contrasts. All the branches were assigned the same length – equivalent to assuming that change occurs only at speciation events ( Purvis & Rambaut, 1995). As discrete variables do not meet the assumptions for calculation of contrasts, separate analyses on migrant and sedentary species were performed. Contrasts were analysed by forcing the regression through the origin ( Purvis & Rambaut, 1995).

All analyses were performed using SAS (SAS Release 6.4, SAS Institute, 1990).

Results

Dispersal and subspecies richness

Stepwise multiple regression analysis showed that range and breeding dispersal were significant predictors of number of subspecies: species with large distribution ranges had more subspecies and species with long dispersal distances had fewer subspecies (Table 1; Fig. 2). Replacing breeding with natal dispersal gave similar results: number of subspecies was negatively correlated with natal dispersal, positively correlated with range and negatively correlated with population size (Table 2; Fig. 3). The number of habitats was not a good predictor of the number of subspecies under both regression models (P > 0.1 in both cases).

Table 1.   Stepwise multiple regression analysis of the number of subspecies in birds considering breeding dispersal distances. Only predictor variables with P < 0.1 are reported. Sample size is 67 species. Thumbnail image of
Figure 2.

 Number of subspecies in relation to mean breeding dispersal distance (km). N=67.

Table 2.   Stepwise multiple regression analysis of the number of subspecies in birds considering natal dispersal distances. Only predictor variables with P < 0.1 are reported. Sample size is 75 species. Thumbnail image of
Figure 3.

 Residual number of subspecies in relation to mean natal dispersal distance (km). N=75.

To remove the possible influence of migration, we included it as a covariate in a second set of analyses. An ANCOVA showed that both migration and breeding dispersal were independently significantly correlated with the number of subspecies (Table 3). Natal dispersal was no longer significantly correlated with the number of subspecies when adding migration to the model (P=0.086).

Table 3.   Analysis of covariance of the number of subspecies in birds with migration as factor, breeding dispersal, range, body mass, population size, and number of habitats as covariates. Sample size is67 individuals. R2 of the model is 0.42. Thumbnail image of

We used the same multiple regression model to test whether the number of subspecies was correlated with dispersal after controlling for similarity due to common ancestry. As for the cross-species analysis, natal dispersal was negatively correlated with the number of subspecies (Table 4). However, breeding dispersal was not a significant predictor of number of subspecies (P=0.12). To take the effect of migration into account we repeated the analyses separately for migratory and sedentary species. The stepwise multiple regression showed that the number of subspecies was negatively correlated with natal dispersal in sedentary (Table 5), but not in migratory species (P > 0.05). The number of subspecies was not significantly correlated with breeding dispersal when considering migratory and sedentary species separately (all P values > 0.05).

Table 4.   Stepwise multiple regression analysis of independent contrasts of the number of subspecies in birds. Only predictor variables with P < 0.1 are reported. Sample size is 71 contrasts. Thumbnail image of
Table 5.   Stepwise multiple regression analysis of independent contrasts of the number of subspecies in sedentary birds. Only predictor variables with P < 0.1 are reported. Sample size is 56 contrasts. Thumbnail image of

Dispersal and ecological plasticity

To investigate whether dispersal propensity was correlated with ecological plasticity we regressed the number of habitats occupied by a species on dispersal and potential covariates. Only population size was significantly correlated with number of habitats, species exploiting a large number of breeding habitats having larger population sizes (slope (SE)=0.057 (0.026), partial R2=0.06, n=75, P=0.032). An ANCOVA including migration as a covariant did not change the conclusions.

After controlling for similarity due to common ancestor, again only population size was positively correlated with the number of breeding habitats (slope (SE)=0.064 (0.032), partial R2=0.06, n=72, P=0.047). When considering migratory and sedentary species separately, the number of breeding habitats was not significantly correlated with natal or breeding dispersal (all P values > 0.05).

Dispersal and wing length

We calculated the residual wing length from a regression of log10-transformed wing length on log10-transformed body mass and related it to the dispersal distances. Both natal and breeding distances were positively related to residual wing length (natal dispersal: F1,73=19.34, P < 0.001, n=75, R2=0.210; breeding dispersal: F1,65=6.83, P=0.011, n=67, R2=0.095).

An ANCOVA including migration as a covariate showed that residual wing length and migratory status were correlated to both natal and breeding dispersal distances, with the interaction terms being marginally nonsignificant for both natal (residual wing length × migration: F1,73=3.94, P=0.051, n=75) and breeding dispersal (residual wing length × migration: F1,65=3.66, P= 0.060, n=67). We repeated the analyses for migratory and sedentary species separately. Migrant species had larger residual wing length than sedentary species (F1,73=6.63, P=0.012, n=75). Residual wing length was positively related to natal and breeding dispersal for sedentary species (natal dispersal: F1,57=17.89, P < 0.001, n=59, R2=0.239, Fig. 4A; breeding dispersal: F1,50=7.00, P=0.011, n=52, R2=0.12), but not for migrant species (natal dispersal: F1,14=0.042, P=0.84, n=16, R2=0.003, Fig. 4B; breeding dispersal: F1,13= 0.46, P=0.51, n=15, R2=0.034).

Figure 4.

 Mean natal dispersal distance (km) in relation to residual wing length for sedentary (A; N=59) and migrant (B; N=16) species.

After controlling for similarity due to common ancestry, residual wing length was positively related to natal dispersal (F1,70=5.06, P=0.028, n=71, R2=0.067), but not to breeding dispersal (F1,64=0.0008, P=0.98, n=65, R2 < 0.001). When considering migratory and sedentary species separately, residual wing length was positively related to natal dispersal for sedentary species (F1,55=4.37, P=0.041, n=56, R2=0.074), but not for migrant species (F1,14=0.96, P=0.34, n=15, R2=0.064). Breeding dispersal was not significantly related to residual wing length in either migrant (F1,13=0.007, P=0.94, n=14, R2=0.0005) or sedentary (F1,49=0.06, P=0.82, n=50, R2=0.0011) species.

Discussion

The main results of this study are that: (i) natal dispersal is a good predictor of subspecies diversity; (ii) natal dispersal is a good predictor of subspecies diversity until we consider migration status, after which this relationship only persists for sedentary species; (iii) dispersal propensity is not related to diversity of breeding habitats, suggesting that dispersal ability alone is a poor predictor of such ecological plasticity; and (iv) residual wing length is related to natal dispersal only for sedentary species. We conducted correlational analyses to provide some links between ecological and behavioural traits of individuals and the evolution of species. Therefore, we do not interpret causation in the relationships suggested and discussed here.

Although a number of factors are likely to determine the level of divergence among populations, this study shows that poor dispersal ability might generate higher rates of subspeciation. Once the effect of potential confounding variables such as breeding range and migration status was controlled, there was a negative relationship between the number of subspecies and dispersal, species with the shortest dispersal distances having the highest subspeciess richness. The existence of subspecies as biological entities is disputed ( Mayr, 1942, 1963). However, there is clear evidence that populations considered to represent subspecies differ in appearance. Such incipient morphological differentiation within species can be considered to be an initial step in the phenotypic divergence among populations ( Houde & Endler, 1990). This negative relationship between dispersal and subspecies richness was already found by von Haartman (1949) for a reduced number of bird species. Other studies using different indicators of population divergence obtain the same relationship – see Pulido (1991) for birds, and Bohonak (1999) for a diverse group of invertebrates – and it is the tendency observed in flightless insects and birds ( Zera, 1981; Liebherr, 1986; Baker et al., 1995 ), which have reduced dispersal abilities.

Considering both measures of dispersal, results showed that natal dispersal persisted as a significant predictor of subspecies richness after controlling for the lack of independence between species. In birds, natal dispersal usually accounts for most dispersal in the life of an individual (see Fig. 1); in general, little relocation occurs after the first breeding attempt (reviewed in Greenwood & Harvey, 1982). This could explain the different results and lack of significance obtained when considering breeding dispersal in the study.

We found a positive association between subspecies richness and geographical breeding range. The same relationship was reported by Møller & Cuervo (1998) for a different set of bird species. An extensive geographical range may cause populations to diverge due to different local selective pressures. Paradis et al. (1998 ) found a negative correlation between geographical range and dispersal power, which corroborates our results. Therefore, populations far removed from others have higher rates of divergence when dispersal ability is poor. A limited gene flow reduces the homogenization of the genetical architecture of populations potentially adapted to local selective pressures.

Subspecies richness was higher in sedentary species, and the interaction between dispersal and migration showed that the number of subspecies was negatively correlated with dispersal in sedentary, but not in migratory species. Although this interaction between dispersal and migration should be interpreted cautiously, due to the differences in the sample sizes between migratory and sedentary species, the study provides the opportunity to compare the homogenizing effects of dispersal and migration on populations, and suggests the importance of distinguishing between these two behavioural events, often used synonymously.

Wing length was a good predictor of the distance moved by individuals, as it was positively related to dispersal and migration. Migration could lead to the evolution of a morphology adapted to long-distance movements, hence facilitating the evolution of relatively uncostly dispersal. Although migrant species had greater residual wing length than sedentary species, the relationship between residual wing length and dispersal distance was not significant for these species, while it was significant for sedentary species. Migratory species should have larger wing span for enduring flight ( Pennycuick, 1975). This and other such factors that facilitate migration should also be those involved in dispersal ability for these species.

We considered the number of breeding habitats as a potential confounding factor when analysing the effect of dispersal on population differentiation. A large number of breeding habitats could reflect colonizing abilities, and the chances of a lineage becoming diversified have been associated with ecological generalism ( Owens et al., 1999 ), and with chances of finding and successfully colonizing new areas ( Rosenzweig, 1995). However, in our study the number of breeding habitats was not a significant predictor of the number of subspecies. Theoretical work has described an interplay between local adaptation and genetic homogenization in populations ( Nuismer et al., 1999 ). The use of a high number of breeding habitat types could generate population divergence if local adaptation is more important than genetic drift among populations. Thus, the lack of association between subspecies and ecological plasticity could be the result of genetic drift counteracting local adaptation in the populations occupying different habitat types. Moreover, our data lack information on the diversity of breeding habitats within subspecies. This information would be necessary to indicate the extent to which local adaptation is determining divergence.

In summary, this work attempts to provide some links between ecological and behavioural traits of individuals and the evolution of species. The study failed in finding ecological plasticity in breeding habitats as a consequence of dispersal, and leaves open the possibility of genetic drift counteracting the role of ecological plasticity in population diversification through local adaptation. Overall, the study suggests that dispersal can have a homogenizing effect on populations and that low dispersal ability might promote speciation.

Acknowledgments

The study was given financial support by ‘Training and Mobility of Researchers’ contract from the European Union to J.B. (FMBICT972583).

Ancillary