SEARCH

SEARCH BY CITATION

Keywords:

  • barn swallow;
  • climate change;
  • condition-dependence;
  • evolution;
  • Hirundo rustica;
  • phenotypic plasticity;
  • remote sensing;
  • survival

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References

The ability of organisms to respond evolutionarily to rapid climatic change is poorly known. Secondary sexual characters show the potential for rapid evolutionary change, as evidenced by strong divergence among species and high evolvability. Here we show that the length of the outermost tail feathers of males of the socially monogamous barn swallow Hirundo rustica, feathers that provide a mating advantage to males, has increased by more than 1 standard deviation during the period from 1984 to 2003. Barn swallows from the Danish population studied here migrate through the Iberian Peninsula to South Africa in fall, and return along the same route in spring. Environmental conditions on the spring staging grounds in Algeria, as indexed by the Normalized Difference Vegetation Index, predicted tail length and change in tail length across generations. However, conditions in the winter quarters and at the breeding grounds did not predict change in tail length. Environmental conditions in Algeria in spring showed a temporal deterioration during the study period, associated with a reduction in annual survival rate of male barn swallows. Phenotypic plasticity in tail length of males, estimated as the increase in tail length from the age of 1 to 2 years, decreased during the course of the study. Estimates of directional selection differentials for male tail length with respect to mating success, breeding date, fecundity, survival and total selection showed temporal variation, with the intensity of breeding date selection, survival selection and total selection declining during the study. Response to selection as estimated from the product of heritability and total selection was very similar to the observed temporal change in tail length. These findings provide evidence of rapid micro-evolutionary change in a secondary sexual character during a very short time period, which is associated with a rapid change in environmental conditions.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References

Sexual selection is a strong evolutionary force that potentially can lead to rapid evolutionary divergence in secondary sexual characters (Darwin, 1871). Three different features of sexual selection are hypothesized to give rise to rapid evolutionary change. First, sexual selection is often directional and very intense (Andersson, 1994). Secondly, secondary sexual characters often have large heritabilities (review in Pomiankowski & Møller, 1995), which is the second component affecting response to selection. Thirdly, secondary sexual characters have a large potential for evolutionary change as reflected by high evolvabilities measured as the additive genetic coefficient of variation (review in Pomiankowski & Møller, 1995).

Given this assumed potential for rapid evolutionary change in secondary sexual characters, we should expect to see cases of rapid change when selection pressures change. The reason for this expectation is that the costs of secondary sexual characters would change concordant with environmental conditions. We are only aware of one example of a species showing rapid evolutionary response to altered sexual selection pressures: Badyaev & Martin (2000) and Badyaev et al. (2000, 2002) have shown for the housefinch Carpodacus mexicanus that sexual dimorphism can evolve rapidly when a species changes its distribution. Periods of dramatic environmental change are often characterized by dramatic evolutionary changes (Lynch & Lande, 1993; Hoffmann & Parsons, 1997). Many potential examples of rapid change in phenotype related to climate have been reported recently (e.g. Winkel & Hudde, 1997; Brown et al., 1999; Crick et al., 1999; Dunn & Winkler, 1999; Sæther et al., 2000; Moss et al., 2001). However, it remains unclear whether these changes are phenotypic or micro-evolutionary in nature. Current concern has been raised about the ability of organisms to adapt to long-term environmental change. Previous studies have suggested that genetic changes in response to climate change do occur (Bradshaw & Holzapfel, 2001; Etterson & Shaw, 2001; Réale et al., 2003). Here we demonstrate that rapid micro-evolutionary change in a secondary sexual character of a passerine bird is associated with climatic change, which has the potential to quickly change the selection regime for free-living organisms. This demonstration of evolutionary change was possible due to a long-term population study of the barn swallow Hirundo rustica in Denmark, during which detailed morphological and other phenotypic characters have been recorded annually for each new cohort of individuals. Thus, we could follow changes in the phenotype of cohorts in relation to changes in environmental conditions. This change was compared with the predicted response to selection based on information on selection and heritability of the character in question, using the breeder's equation (Lynch & Walsh, 1998).

Many secondary sexual characters are condition-dependent (Andersson, 1994), which implies that only individuals in prime condition are able to produce the most exaggerated secondary sexual characters. Individuals in prime condition may be better able to produce exaggerated traits than individuals in poor condition, suggesting that such individuals may be in a better condition both before and after the production of secondary sexual characters (e.g. Møller, 1991a; Jennions et al., 2001). Therefore, we can predict that a deterioration of environmental conditions will be particularly harmful for individuals with small secondary sexual characters, causing individuals in poor condition to be selected against. We used information on the Normalized Difference Vegetation Index (NDVI) in the winter quarters and during migration to index environmental conditions outside the breeding season of the barn swallow (Szép & Møller, 2004). The North Atlantic Oscillation (NAO) index was used as a measure of climatic conditions in large parts of Northern Europe in general and in the study site in Denmark in particular during the breeding season. The NAO is a major source of atmospheric mass balance between pressure centres over Ponta Delgada, Azores and Stykkisholmur, Reykjavik, Iceland (Hurrell, 1995).

The barn swallow is a small (ca. 20 g), socially monogamous, semi-colonial passerine feeding on insects caught on the wing. The outermost tail feathers are on average ca. 20% longer in males than in females in the Danish study population, whereas other morphological characters are approximately of similar size (Møller, 1994). Long-tailed males enjoy several kinds of mating advantages, as demonstrated by observations and experiments (Møller, 1988, 1994; Saino et al., 1997). This secondary sexual character is condition-dependent because poor conditions in the winter quarters, ectoparasites and radioactive contamination cause the production of smaller characters (Møller, 1990, 1992, 1993). Long tails of males may improve manoeuvring ability (e.g. Møller, 1991c; Norberg, 1994; Evans & Thomas, 1997), suggesting that long tails provide males with a natural selection advantage. However, no selective advantage in terms of reproductive success or survival has ever been demonstrated (Møller et al., 1998), questioning whether these behavioural differences provide any fitness advantages derived from natural selection.

Study area

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References

Data were collected in a study area of ca. 45 km2 at Kraghede (57°12′N, 10°00′E), Denmark, during 1984–2003 (see Møller, 1994 for details). Adult barn swallows were captured in mist nets, and phenotypic traits including the length of the two outermost and the central tail feathers measured with a ruler to the nearest mm by A.P.M. The length of the central tail feathers was first measured from 1989 onwards. Body mass was recorded with a Pesola spring balance to the nearest 0.1 g. We measured the morphological characters accurately as shown by high repeatability of the measurements of characters during subsequent capture events and to a lesser extent during following breeding seasons (Møller, 1994, unpublished data).

Individuals with broken or damaged feathers were excluded from the analyses. The tip of the outermost tail feathers of barn swallows is rounded, being composed of small barbs, and any broken barb leaves an irregular shape of the feather that is readily visible. The number of birds with damaged feathers was less than 3% in any given year, which is thus unlikely to have biased the analyses. Likewise, all individuals involved in experiments were excluded if belonging to treatment groups (but not untreated controls). This exclusion should not cause any bias as individuals were randomly assigned to treatments, and experimental birds accounted for less than 10% of all individuals. A total of 1642 males and 1678 females were used in the analyses.

North Atlantic Oscillation

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References

The NAO is a major source of atmospheric mass balance between pressure centres over Ponta Delgada, Azores and Stykkisholmur, Reykjavik, Iceland, and it has a major influence on the geographical distribution of passage of low pressures and hence precipitation and temperature in Europe (Hurrell, 1995). An index of NAO is estimated as the difference in normalized sea level pressures by division of each monthly pressure by the long-term standard deviation (1865–1984). High index values are associated with high winter temperatures and high levels of precipitation in Denmark and Scandinavia (Hurrell, 1995). In our analysis, we used winter NAO values (mean of December–March) from the Climatic Research Unit of the University of East Anglia (UK) (http://www.cru.uea.ac.uk/cru/data/nao.htm).

Normalized Difference Vegetation Index

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References

We estimated environmental conditions during migration and in the African winter quarters of the barn swallow from the NDVI, which is based on satellite images indicating the condition of rainfall-dependent vegetation in space and time (Prince & Justice, 1991). NDVI provides an estimate of the amount and vigour of vegetation at the land surface that is directly related to the level of photosynthetic activity (Prince & Justice, 1991). The NDVI is likely to reflect the abundance of insects, because they depend on plant productivity. Although many migratory birds are insectivores, the NDVI may still provide reliable information on food abundance for such species. This is particularly likely to be the case for species like the barn swallow that rely on swarming insects such as termites in Africa, as such insects become active in direct response to precipitation. NDVI is estimated from data collected by National Oceanic and Atmospheric Administration (NOAA) satellites, and processed by the Global Inventory Monitoring and Modelling Studies at the National Aeronautics and Space Administration (NASA) (Prince & Justice, 1991). Vegetation indices derived from the (NOAA) Advanced Very High Resolution Radiometer sensor such as the NDVI have been used in qualitative and quantitative studies of numerous phenomena related to the condition of vegetation (Tucker et al., 1991). NDVI has been used as an indicator of relative biomass and greenness of the vegetation (Paruelo et al., 1997; Chen & Brutsaert, 1998; Boone et al., 2000), precipitation (Schmidt & Karnieli, 2000) and quantity and quality of bird habitats (Wallin et al., 1992). Maurer (1994) indicated a correlation between abundance of birds and NDVI, and NDVI has the potential to be used for mapping bird distributions at large spatial scales (Osborne et al., 2001). NDVI data are available at NASA for the entire African continent and for the Mediterranean countries at http://edcintl.cr.usgs.gov/adds/adds.html. These images are calibrated for inter-sensor and intra-sensor degradation for the period 1982–2002.

The NDVI data are available for each pixel that covers 8 km × 8 km (64 km2). The values were subsequently averaged to obtain a mean estimate for each square of 0.25° by 0.25°. This was done for each period of the annual cycle of the barn swallow in Africa [fall (September–November), winter (December–February) and spring (March–May)] using the WinDisp v4.0 software (http://www.fao.org/WAICENT/faoinfo/economic/giews/english/windisp/windisp.htm). We calculated NDVI data for squares where at least 30% of the pixels had valid data. We only included squares in the analyses if there was NDVI data available for each year for which we had survival estimates (1984–2002). This criterion caused exclusion of 15% of the ca. 46 000 squares in Africa and around the Mediterranean.

Survival analysis

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References

Survival analysis was based on data of 1227 adult males and 1272 adult females captured and recaptured between 1984 and 2002. We estimated adult annual survival rates using capture–mark–recapture methods (Lebreton et al., 1992). The goodness-of-fit of the general Cormack–Jolly–Seber (CJS) model (Clobert & Lebreton, 1987) to the data was calculated using the program release (Burnham et al., 1987), and bootstrap function of the program mark (White & Burnham, 1999). Modelling and estimation of survival and recapture rates and individual covariates was made with MARK (White & Burnham, 1999). Model selection among the studied models was based on the Akaike's information criterion (AIC) by selecting the model with the lowest AICc value (Anderson et al., 1994), Akaike weight (Anderson & Burnham, 1999), and likelihood ratio tests were made for comparing a model with the nested ones (Lebreton et al., 1992). This procedure allows choice of the best model while using rigorous and objective criteria for choice of the simplest model that fits the data.

We tested for effects of environmental conditions in the breeding areas, on migration and during winter on annual survival with capture–mark–recapture modelling methods (Lebreton et al., 1992). We used analysis of deviance to test for the significance of group covariates (NDVI or NAO value) on survival rate (Skalski et al., 1993) [global model: St (survival rate dependent on time), P (recapture rate); constant model: S, P; covariate model: S(NDVI or NAO), P]. We identified areas of potential importance for survival of the barn swallow by investigating relationships between annual survival rate and NDVI (see Szép & Møller, 2004 for details). A brief summary of this approach follows.

Areas were identified using the Spearman rank correlation matrix between the time series of survival estimates and the time series of NDVI data for each square. For each of the three periods (fall, winter and spring) we identified squares that had a positive correlation between annual survival estimate and NDVI exceeding a critical threshold. We only investigated positive correlations as migratory birds have populations that are regulated outside the breeding season (see discussion in Szép & Møller, 2004). We then identified aggregations of squares with positive correlations between survival and NDVI values as potential areas of importance for survival. Briefly, the argument is that a large number of positive correlations between survival and NDVI are unlikely to occur in neighbouring squares by chance, and that a large number of continuous squares should occur with a probability of finding a square with a given threshold correlation between NDVI and survival raised to the power of the number of squares in that aggregation (Szép & Møller, 2004). Thus, the probability of finding even as few as five continuous squares together is exceedingly small. We used the mean NDVI values of these continuous areas as independent variables in stepwise linear regression models with annual survival estimate as the dependent variable. We chose the areas that explained the largest amount of variation across the three time periods of the annual cycle, provided they explained at least 10% of the variance in a partial correlation analysis.

We identified areas of importance for survival using six different thresholds of the rank correlation between NDVI and survival rate (rS ≥ 0.20, 0.30, 0.40, 0.50, 0.60 and 0.80) to determine the robustness of the conclusions to the choice of threshold. The results for the threshold of 0.60 provided the best fit, because a value of 0.80 yielded an insufficient number of areas, whereas values less than 0.60 yielded many more areas with an overall poorer fit (see also Szép & Møller, 2004). An area was defined as the aggregation of neighbouring squares that had a positive correlation exceeding the threshold rank correlation coefficient. In a second series of analyses we defined an area as the aggregation of squares that were maximally separated by one square without data or not exceeding the threshold. This procedure allowed us to investigate the robustness of our conclusions to the exclusion of squares that did not have NDVI data for the entire period. Use of this second criterion did not provide quantitatively different results than the use of the first criterion (see Szép & Møller, 2004). These procedures resulted in a very small number of continuous areas of potential importance for survival. As we simultaneously investigated NDVI in all these continuous areas of potential importance as predictors of survival in a single analysis of deviance (Skalski et al., 1993), this approach avoided any problems arising from inflated probabilities due to multiple comparisons.

This method for identification of areas of importance for survival may superficially appear to suffer from circular reasoning. However, we emphasize that we have recommended that identified areas of potential importance for survival should be independently validated using other sources of information (see details in Szép & Møller, 2004). We have proposed a number of such methods of cross-validation including coincidence of identified areas with areas of known distribution, coincidence of identified areas with areas where banded breeding birds have been recovered, comparison of isotope profiles of feathers from birds captured in identified areas and in breeding areas, and comparison of genetic profiles, for example, from microsatellite or RFLP typing, of birds captured in identified areas and in breeding areas (Szép & Møller, 2004). Here, we validated the areas identified as being of importance for survival with the location of banding recoveries of breeding birds from Denmark, and we tested the distribution of these recoveries against the null hypothesis of a random distribution of recoveries across Africa.

Selection analysis

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References

We calculated directional selection differentials as the change in standardized mean phenotype associated with four different selection episodes and total selection: (1) mating success [either mated (fitness = 1) or unmated (fitness = 0)], (2) breeding date (where fitness decreases with delay in breeding date) (see Møller, 1994), (3) fecundity selection (fitness equalling the number of offspring produced during the breeding season), and (4) survival selection (survival from 1 year to the next with fitness = 1 for survivors and fitness = 0 for nonsurvivors), using the procedures recommended by Arnold & Wade (1984). This was done by regressing relative fitness (fitness for each individual divided by mean fitness) on standardized tail length before selection for each cohort (standardized to a mean of zero and a variance of 1). For the analysis of survival we estimated individual survival from recapture or re-sighting of individuals. This procedure is justified by the very high and constant recapture rates of barn swallows in this study (see Møller & Szép, 2002; this study). We related directional selection differentials to year and NDVI. In addition, we compared the change in tail length during the study (calculated as change in standard deviation units) with the expected value from the breeder's equation (Falconer & Mackay, 1996; Lynch & Walsh, 1998): R = h2×S, where R is response to selection, h2 is heritability and S is selection differential. We calculated S from total selection as estimated above, and multiplied this with the number of generations, where generation time T = A + P/(1 − P), where A is age at first breeding (1 year; Møller, 1994) and P is annual survival rate (0.343; Møller & Szép, 2002). This gives a T of 1.522, resulting in the study period of 20 years equalling 13.141 generations.

Heritability estimates used for the estimation of response to selection were obtained from parent–offspring regression, using phenotypic values of sons and fathers when both were yearlings to avoid any bias due to age effects. Daughters were excluded because very few females recruit locally into the study population (Møller, 1991b, 1994). We investigated biases in these estimates using phenotypic traits of fathers (the abundance of three different species of ectoparasites differing in degree of virulence) and reproductive variables from the nest of origin of the offspring (laying date, clutch size, brood size) as independent variables in multiple regressions.

Genetic correlations were estimated for length of outermost and central tail feathers in males and for length of outermost tail feathers in males and females using correlation coefficients based on phenotypic values of offspring and parents, following the procedures described by Falconer & Mackay (1996). The overall genetic correlation coefficient was estimated as the mean value of the two estimates.

Statistical analysis

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References

Each yearling was only used once in the statistical analysis to avoid problems of pseudo-replication. We tested for temporal change in tail length and length of the central tail feathers using a full factorial two-way ancova with sex as a factor and year as a continuous variable. We tested for the importance of NDVI in identified areas of importance for survival in predicting tail length of male and female barn swallows using a forward stepwise regression procedure. Change in tail length across cohorts was estimated in units of standard deviations, as the mean difference between two subsequent cohorts, divided by the standard deviation of the first of these cohorts. We adjusted the degrees of freedom for tests of significance of correlation coefficients for temporal auto-correlation (Bartlett, 1952), ensuring that the adjusted degrees of freedom were equal to the actual sample size.

Survival estimates and environmental conditions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References

We estimated annual adult survival rates using capture–mark–recapture analysis to identify areas of importance for annual survival. Heterogeneity in the data was modelled using survival models, and this heterogeneity was accounted for in subsequent models. A goodness-of-fit test of the CJS model with RELEASE (Burnham et al., 1987), which only assumes time effects (t) on survival (S) and recapture rate (P), showed that this model (Ss*t, Ps*t) fitted the male data (χ2 = 24.83, d.f. = 17, P =0.099; TEST 3), but not the female data (χ2 = 43.40, d.f. = 17, P < 0.001; TEST 3).

In females, the main cause of the lack of fit was that the probability of recapture during the (i + 1)th year was lower for females banded in the ith year than for females banded before the ith year. In the years 1991, 1993, 1994 and 1995 this difference was significant [1991: χ2 =14.73, d.f. = 1, P < 0.001; 1993: χ2 = 7.40, d.f. = 1, P =0.007; 1994: χ2 = 5.54, d.f. = 1, P = 0.02; 1995: χ2 =4.37, d.f. = 1, P = 0.04; TEST 3.Sri (Burnham et al., 1987)].

Because of the lack of fit of the CJS model to the data of adult females, we considered the (Sm*fa1*fa2*t, Ps*t) model, according to which the survival of males (m) varies differently among years (t) from the survival of newly banded females (fa1) and from already banded females (fa2) and the recapture rate varies differently among years between the sexes (s*t). Bootstrap GOF of the starting general model showed fit to the data (P =0.25, deviance based; 1000 repetition). The variance inflation factor (c-hat) for the general model was 1.093 (deviance based), and this value was used for correcting the AIC value for overdispersion of data (Anderson et al., 1994).

Model selection among alternative, reduced models of the general model showed that the Sm+fa1+fa2+t, Pt3 model had the lowest QAICc value (QAICc = 3612.79) (Table 1). This model assumes that survival rate does not differ between age classes of males (m). In addition, it assumes that survival differs between the two age classes of females (fa1 and fa2), varying in parallel, with that of newly banded, 1-year-old females (fa1) and earlier banded, 2-year-old or older females (fa2), during the study period (m+fa1+fa2+t), and that the recapture rate differs for three 6-year periods (t3), being similar for all birds. The model fitted the data (P = 0.93; bootstrap GOF, deviance based; 1000 repetition).

Table 1.  Akaike information criterion (QAICc) value, Delta QAICc, QAICc weight, number of identifiable parameters (np), and Q-deviance values of the investigated models calculated from mark (White & Burnham, 1999)
ModelQAICcDelta QAICcQAICc weightNpQ-deviance
  1. S, survival value; P, recapture rate; m, male; f, female; a1, newly banded individuals; a2, already banded individuals (banded in the previous or earlier seasons); s, sex; t, time (year); area (1–6): NDVI value of areas 1 – Spain, spring; 2 – Tunisia, fall; 3 – Egypt, fall; 4 – South Africa, spring; 5 – SE Angola, fall; 6 – Algeria, spring; NAO, North Atlantic Oscillation Index for the period December–March; t3, three equally long time periods between 1984 and 2002; +, no interaction, or *interaction between sexes(s) or ages (a1, a2) over time. The models with the lowest QAICc are shown in bold. Variance inflation factor (ĉ), estimated for the general model (ĉ = 1.09) was used for correction of the AIC and deviance value for overdispersion of the data (Anderson et al., 1994).

Sm+fa1+fa2+area(1–3), Pt33593.3600.310249206.985
Sm+fa1+fa2+area(1–5), Pt33593.960.60.2299911203.557
Sm+fa1+fa2+area(1–4), Pt33594.771.40.1537510206.376
Sm+fa1+fa2+area(1–6), Pt33595.161.80.1264012202.740
Sm+fa1+fa2+area(1–2), Pt33596.403.040.067938212.034
Sm+fa1+fa2+area(1–6), Pt33597.103.730.0480213202.659
Sm+fa1+fa2+area(1–6), NAO, Pt33597.183.810.0461313202.740
Sm+fa1+fa2+t, Pt33612.7919.430.0000223198.126
Sm+fa1+fa2+t, Pt3618.4725.11021207.862
Sm*t,fa1+fa2+t, Pt33630.4937.12040181.141
Sm+fa1+fa2+NAO, Pt33635.2641.8907252.900
Sm*fa1*fa2*t, Pt33639.4346.06056157.106
Sm,fa1,fa2, Pt33642.3148.9406261.958
Sm*fa1*fa*t, P3645.7952.43054167.609
Sm*fa1*fa2*t, Ps*t3672.5079.14086127.449
Sg*t, Ps*t3684.9391.56070173.481

On the basis of the Sm+fa1+fa2+t, Pt3 model, the survival rate of newly ringed, 1-year-old females was on average 0.14 lower compared with earlier ringed females. On average, the survival rate of males was on average 0.10 higher than for 1-year-old, ringed females, and it was close to that of 2-year-old or older females. The values of recapture rate were constant and very high for the three periods (1985–1990: 0.971; 1991–1996: 0.989; 1997–2002: 0.882).

In the selected model we used 18 survival rates of adult males for 1984–2002 for calculating correlation coefficients between survival rate and environmental conditions during migration and in winter.

Selection of areas of importance for survival

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References

On the basis of the threshold values (rS ≥ 0.6; considering only the neighbouring squares; the size of the selected areas is equal to or larger than three squares), we selected 40 areas during fall, six areas during winter and nine areas during spring. During the stepwise regression analysis, we selected six areas (three areas during fall and three areas during spring) that had a large influence on estimated survival, based on the Sm+fa1+fa2+t, Pt3 model (Table 1). Thus, the potentially large number of continuous areas of importance for survival was reduced to a mere six areas, following the procedure that we adopted. We considered these six areas in the subsequent capture–recapture analysis as group covariates.

Survival and environmental conditions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References

Modelling survival rate with group covariates provided a model that explained annual changes in survival rate with NDVI in Tunisia, Egypt and Angola in fall and NDVI in South Africa, Algeria and Spain in spring (see Fig. 1 for location and extent of these selected areas). We considered all different combinations of the six identified areas in the models. When the model with NDVI from the six areas was fitted to the data, this model was among the best models based on the QAICc value (Table 1). Models that considered fewer areas did not provide a better fit than the model considering six group covariates [Sm + fa1 + fa2 + (area1–6), Pt3] as the delta QAICc value was lower than 2 and QAICc weight was 0.11 (Table 1). NDVI in areas in Africa and in the Mediterranean region were positively correlated with survival rate of adult barn swallows, whereas we found no significant influence of NAO on survival rate [analysis of deviance, anodev: F =2.65, d.f. = 1,16, P = 0.12; global model: S(m+fa1+fa2 + t) p(t3); constant model: S(m,fa1,fa2) P(t3); covariate model: S(m + fa1 + fa2 + NAO) P(t3)]. We note that this test is based on a single test of the null hypothesis. The final covariates model, that included NDVI for the six areas, explained 94.8% of the variance in annual survival estimates during the study period 1985–2002. In the remaining part of the paper, the areas of importance for survival are for simplicity referred to by the name of the country.

image

Figure 1. Areas of importance for adult survival of Danish barn swallows: A1, Spain (March–May); A2, Tunisia (September–November); A3, Egypt (September–November); A4, South Africa (March–May); A5, Angola (September–November); A6, Algeria (March–May).

Download figure to PowerPoint

We cross-validated the areas identified as being of potential importance for survival by comparing these areas with the locations of banding recoveries of breeding birds from Denmark. A total of 73% of the 15 winter recoveries of banded barn swallows from Denmark were from Angola, Namibia and South Africa within or close to the identified areas of importance for survival, whereas only 11% would be predicted by chance based on the area of Africa. Likewise, all 13 recoveries from North Africa during spring migration were from Morocco, Algeria and Tunisia within or close to the identified area of importance for survival in Algeria, whereas only 10% would be predicted by chance based on area alone. These distributions differ significantly from random expectations. Thus, there was a high degree of coincidence between the identified areas of importance for survival in South Africa, Angola, and Algeria and the known recovery distribution of barn swallows from Denmark. Szép & Møller (2004) provide additional evidence for the location of the selected areas being highly unlikely to be nonrandom, based on the location of the selected areas relative to the known geographical distribution of barn swallows in Africa. This evidence is consistent with our suggestion that environmental conditions in these areas, as reflected by NDVI, were directly affecting survival rates of adult barn swallows.

Annual survival rate of males was strongly positively related to NDVI in the identified area in Algeria in spring (Fig. 2). NDVI for the other five identified areas did not significantly predict annual survival rate of males (all P > 0.20). Annual survival rate of females was also significantly positively correlated with NDVI in Algeria in spring, although more weakly than for males [F = 4.97, d.f. = 1,16, r2 = 0.24, P = 0.04, slope (SE) = 0.960 (0.431)].

image

Figure 2. Survival rate of adult male barn swallows of the cohorts 1984–2002 in relation to Normalized Difference Vegetation Index in Algeria in March–May. The linear regression is significant [F = 9.26, d.f. = 1,16, r2 = 0.37, P = 0.008, slope (SE) = 1.38 (0.45)].

Download figure to PowerPoint

Temporal change in tail length

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References

Tail length of male and female barn swallows and sexual size dimorphism changed during 1984–2002 (Table 2). Mean tail length of yearling males increased by on average 11.4 mm or 1.3 standard deviations [Fig. 3a; slope (SE) = 0.53 mm (0.09), P < 0.0001], whereas mean female tail length only increased by on average 3.3 mm or 0.5 standard deviations during the same period (Fig. 3a, Table 2). Therefore, sexual size dimorphism increased during the study period (Table 2). In contrast, there was no significant change in the length of the sexually monomorphic central tail feathers or their dimorphism (Table 2). Restricting the analysis for outermost tail feathers to the same years as those for which information existed for central tail feathers did not change the conclusions [1989–2003: male tail length: F = 9.39, d.f. = 1,13, r2 = 0.42, P = 0.009, slope (SE) =0.271 (0.088); female tail length: F = 6.47, d.f. = 1,13, r2 = 0.33, P = 0.03, slope (SE) = 0.146 (0.057); sexual size dimorphism: F = 6.81, d.f. = 1,13, r2 = 0.34, P =0.02, slope (SE) = 0.110 (0.042)].

Table 2.  Linear regressions of temporal change in size and sexual size dimorphism in length of the outermost (tail length) and the central (central tail length) tail feathers of adult barn swallows
VariableFd.f.r2PSlope (SE)
Male tail length27.6711,180.606<0.00010.354 (0.067)
Female tail length24.8381,180.580<0.00010.165 (0.033)
Sexual size dimorphism in tail length8.2781,180.3150.0100.189 (0.066)
Male central tail length0.6531,130.0480.434 
Female central tail length1.7291,130.1170.211 
Sexual dimorphism in central tail length0.5261,130.0390.481 
image

Figure 3. (a) Mean tail length (mm, +SE) of yearling male and female barn swallows of the cohorts 1984–2002. The increase in tail length in male and female barn swallows is statistically significant (see Table 2). (b) Survival rate of male barn swallows of the cohorts 1984–2002. Values are estimates (±SE). (c) Normalized Difference Vegetation Index (NDVI) in Spain (March–May), Tunisia (September–November), Egypt (September–November), South Africa (March–May), Angola (September–November) and Algeria (March–May). The linear regression line for Algeria is shown (the only significant relationship). See Fig. 1 for location of areas.

Download figure to PowerPoint

Survival rate of males decreased consistently during the period 1984–2001 (Fig. 3b). The decrease was statistically highly significant [F = 6.01, d.f. = 1,17, r2 =0.26, P = 0.025, slope (SE) = −0.004 (0.002)]. In contrast, survival rate of females did not change significantly (F = 0.33, d.f. = 1,17, r2 = 0.02, P = 0.57).

We should expect a clear temporal trend in NDVI in the area that predicted change in tail length, because of the temporal increase in tail length during the period 1984–2002, but not in other selected areas where NDVI did not predict tail length. Indeed, only NDVI in Algeria in spring showed a significant temporal pattern with low values in recent years [Fig. 3c; F = 5.63, d.f. =1,16, r2 = 0.26, P = 0.031, slope (SE) = −0.003 (0.001)]. None of the other five identified areas of importance for survival showed a significant temporal trend in NDVI (Fig. 3c; F < 1.01, d.f. = 1,16, r2 < 0.06, P > 0.33). This finding is consistent with the observation that only NDVI in Algeria predicted the temporal increase in tail length in male barn swallows, whereas NDVI of the other areas did not enter as significant predictors.

Phenotypic plasticity in tail length of males with respect to age, calculated as the increase in tail length from 1 to 2 years of age decreased significantly during the study (Fig. 4). The decrease was statistically highly significant [F = 12.27, d.f. = 1,16, r2 = 0.43, P = 0.003, slope (SE) = −0.306 (0.087); there was no significant relationship for the period 1989–2003: F = 0.38, d.f. =1,12, r2 = 0.03, P = 0.55]. There was no significant change in phenotypic plasticity of central tail feathers (F = 0.41, d.f. = 1,11, r2 = 0.03, P = 0.54). This implies that phenotypic plasticity in tail length actually decreased as tails became longer.

image

Figure 4. Phenotypic plasticity in tail length (mm) of male barn swallows, measured as the increase in tail length for the same individuals between 1 and 2 years of age, during the period 1985–2002. Values are mean (SE). Some standard errors are two small for being shown on the graph.

Download figure to PowerPoint

Selection on tail length

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References

Directional selection differentials for tail length in males were calculated for four different selection episodes. Mating selection differentials were positive in all 20 years, and 14 of these were statistically significant. There was variation in intensity of mating selection during the study, although no consistent pattern was obvious (Fig. 5a). For length of central tail feathers, 13 of 15 selection differentials were positive, and two of these were statistically significant. There was no significant temporal trend in mating selection acting on central tail feathers (F = 0.21, d.f. = 1,13, r2 = 0.02, P = 0.66).

image

Figure 5. Directional selection differentials for (a) mating selection, (b) breeding date selection, (c) fecundity selection, (d) survival selection and (e) total selection during 1984–2003. Survival selection and total selection were only calculated for the period 1984–2002, as the survival of the cohort from 2003 was still unknown. The linear regressions were as follows: mating selection: F = 0.94, d.f. = 1,18, r2 = 0.05, P = 0.35; breeding date selection: F = 7.53, d.f. = 1,18, r2 = 0.30, P = 0.01, slope (SE) = −0.005 (0.002); fecundity selection: F = 2.09, d.f. = 1,18, r2 = 0.10, P = 0.17; survival selection: F = 4.47, d.f. = 1,17, r2 = 0.21, P = 0.049, slope (SE) = −0.019 (0.009); total selection: F = 7.29, d.f. = 1,17, r2 = 0.30, P = 0.02, slope (SE) = −0.031 (0.011).

Download figure to PowerPoint

Breeding date selection on tail length was positive in 19 of 20 years, and nine of these selection differentials were statistically significant. There was a significant decrease in the intensity of selection during the study period, with values changing from large and positive to being close to zero in later years (Fig. 5b). For central tail feathers six of 15 selection differentials were positive, but none were significant. There was no significant temporal trend in mating selection acting on central tail feathers (F = 1.23, d.f. = 1,13, r2 = 0.09, P = 0.29).

Fecundity selection on tail length was positive in 18 of 20 years and four of the selection differentials were statistically significant. There was no clear pattern of temporal change in fecundity selection (Fig. 4c). For central tail feathers 11 of 15 selection differentials were positive and four were statistically significant. There was no significant temporal trend in mating selection acting on central tail feathers (F = 0.46, d.f. = 1,13, r2 = 0.03, P = 0.51).

Survival selection was positive in eight of 19 years and three values were statistically significant. There was a significant decrease in survival selection differentials during the study period from positive values in the mid 1980s to negative values in recent years (Fig. 5d). This result is consistent with a decrease in survival estimates based on mark–recapture analyses being associated with an increase in tail length of males (Møller & Szép, 2002). Survival selection on tail length was significantly positively related to NDVI in Algeria [F = 9.89, d.f. =1,16, r2 = 0.38, P = 0.006, slope (SE) = 4.37 (1.39)]. For central tail feathers eight of 14 selection differentials were positive, but none were significant. There was no significant temporal trend in mating selection acting on central tail feathers (F = 1.03, d.f. = 1,12, r2 = 0.08, P = 0.33). Similarly, there was no significant relationship between survival selection on central tail feathers and NDVI in Algeria (F = 0.45, d.f. = 1,11, r2 = 0.04, P =0.52).

Finally, overall directional selection on male tail length was positive in 14 of 19 years, with a mean value of 0.223 (SE = 0.072). Overall selection decreased significantly during the study period from large positive values in the mid 1980s to negative values in recent years (Fig. 5e). Total selection was significantly positively related to NDVI in Algeria [F = 6.42, d.f. = 1,16, r2 =0.29, P = 0.022, slope (SE) = 4.89 (1.93)]. Overall directional selection on length of central tail feathers was positive in nine of 14 years, with a mean value of 0.080 (SE = 0.068). There was no temporal trend in total selection on length of central tail feathers (F = 1.03, d.f. = 1,12, r2 = 0.08, P = 0.33). Likewise, there was no significant relationship between total selection on length of central tail feathers and NDVI in Algeria (F = 1.11, d.f. = 1,11, r2 = 0.09, P = 0.31).

Change in tail length in relation to environmental conditions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References

We determined the best predictor of temporal change in tail length in male barn swallows, using NDVI and NAO as predictor variables in forward stepwise regression models. The best model only included NDVI in Algeria in spring as a significant predictor of tail length in males (Fig. 6a). NDVI in the other seven areas or NAO did not enter as significant predictors. The negative relationship implies that as environmental conditions deteriorated, tails became longer. Importantly, NDVI in Algeria did not significantly predict female tail length (Fig. 6b) or length of the central tail feathers (F = 0.48, d.f. = 1,11, r2 = 0.04, P = 0.50). An ancova with NDVI, sex and NDVI by sex interaction showed that the interaction was statistically significant (F = 7.25, d.f. = 1,32, r2 = 0.19, P = 0.011), implying that the relationship between NDVI and tail differed between the two sexes. Thus, male tail length decreased significantly with decreasing NDVI in Algeria, whereas female tail length did not.

image

Figure 6. Mean tail length (mm, ±SE) of yearling (a) male and (b) female barn swallows of the cohorts 1984–2002 in relation to Normalized Difference Vegetation Index in Algeria in March–May. The linear regression for males was significant [F = 13.33, d.f. = 1,16, r2 = 0.46, P = 0.002, slope (SE) = −73.24 (20.06)], although this is not the case for females (F = 3.49, d.f. = 1,16, r2 = 0.18, P = 0.08).

Download figure to PowerPoint

Change in tail length from one cohort to the next was also predicted by NDVI in Algeria in male but not in female barn swallows. Change in tail length between cohorts was measured in units of standard deviations (standardized change in phenotype). The relationship between standardized change in phenotype and NDVI was significant in males (Fig. 7a), but not in females (Fig. 7b). There was no significant relationship between standardized change in central tail feather length and NDVI (F = 0.02, d.f. = 1,10, r2 = 0.00, P = 0.89). These findings imply that at high values of NDVI in Algeria tail length of males tended to decrease across generations and to increase at low values, whereas there were no similar effects for monomorphic central tail feathers.

image

Figure 7. Change in tail length of yearling (a) male and (b) female barn swallows estimated in units of standard deviations of the cohorts 1984–2002 in relation to Normalized Difference Vegetation Index in Algeria in March–May. The linear regression for males was significant [F = 4.94, d.f. = 1,16, r2 = 0.24, P = 0.04, slope (SE) = −3.22 (1.45)], although this was not the case for females (F = 1.47, d.f. = 1,16, r2 = 0.08, P = 0.24).

Download figure to PowerPoint

Heritability of tail length was estimated from son–father regressions. For recruits in the Danish population, the resemblance between sons and fathers was statistically significant [F = 9.38, d.f. = 1,128, r2 =0.08, P =0.002, slope (SE) = 0.21 (0.07)], giving a heritability estimate of 0.42 (0.14). A similar estimate was obtained if based on residual tail length from a regression of tail length on year in both sons and fathers [F = 7.21, d.f. =1,128, r2 = 0.05, P = 0.008, slope (SE) = 0.21 (0.06)]. We investigated the stability of this estimate using a range of phenotypic characters of the parent or the nesting event as additional independent variables. The heritability estimate was 0.42 (0.17) when clutch size was controlled statistically [multiple F = 3.61, d.f. = 2,81, r2 = 0.08, P = 0.03; partial regression for brood size: slope (SE) = 0.51 (1.04), t = 0.49, P = 0.63], 0.45 (0.22) when brood size was controlled [multiple F = 3.93, d.f. = 2,81, r2 = 0.09, P = 0.02; partial regression for brood size: slope (SE) = 0.64 (0.70), t = 0.91, P = 0.37], and 0.49 (0.14) when laying date was controlled [multiple F =6.19, d.f. = 2,110, r2 = 0.10, P = 0.003; partial regression for brood size: slope (SE) = 0.03 (0.04), t = 0.65, P = 0.52]. Similarly, heritability estimate was 0.43 (0.16) when the abundance of tropical fowl mites Ornithonyssus bursa in the nest was controlled statistically [multiple F = 3.96, d.f. = 2,81, r2 = 0.09, P =0.02; partial regression for brood size: slope (SE) = 0.79 (0.85), t = 0.94, P = 0.35], 0.39 (0.15) when the abundance of chewing lice Hirundoecus malleus on the father was controlled [multiple F = 3.76, d.f. = 2,83, r2 = 0.08, P = 0.03; partial regression for brood size: slope (SE) = 1.87 (1.34), t = 1.40, P = 0.16], and 0.40 (0.14) when the abundance of feather mites of the father was controlled [multiple F = 3.61, d.f. = 2,81, r2 = 0.08, P = 0.03; partial regression for brood size: slope (SE) = −0.01 (0.03), t = 0.23, P = 0.82]. The mean of these estimates was 0.42, which is very similar to the estimate listed above.

The genetic correlation between lengths of the outermost tail feathers in the two sexes was weak, but statistically significant (mean estimate of rG = 0.22; rG = 0.22; F = 8.96, d.f. = 1,169, r2 = 0.05, P = 0.003; rG = 0.22; F = 2.05, d.f. = 1,39, r2 = 0.05, P = 0.18). The genetic correlation between length of outermost and length of central tail feathers was weak (mean estimate of rG = 0.02) and not statistically significant for males (rG = −0.08; F = 0.06, d.f. = 1,104, r2 = 0.01, P = 0.45; rG = 0.12; F = 1.27, d.f. = 1,89, r2 = 0.01, P = 0.26).

Average selection on tail length was 0.223, and the predicted response to selection is therefore 0.223 × 0.42 (heritability of tail length) × 13.141 (number of generations) = 1.231 standard deviation units. During the period 1984–2003 tail length increased by 11.41 mm, and with an overall standard deviation of 9.022, this gives an increase by 1.265 standard deviation units. This value is very close to the value of 1.231 predicted from the breeder's equation.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References

The ability of extant populations to respond evolutionarily to rapid climatic change is of great current interest (Lynch & Lande, 1993; Hoffmann & Parsons, 1997). However, few examples of such responses are available (Bradshaw & Holzapfel, 2001; Etterson & Shaw, 2001; Meriläet al., 2001; Réale et al., 2003). Secondary sexual characters have a high potential for evolutionary change (Pomiankowski & Møller, 1995), because such characters are commonly subject to intense directional selection, but also have high evolvabilities estimated in terms of the additive genetic coefficient of variation (Darwin, 1871; Andersson, 1994; Pomiankowski & Møller, 1995). The present study suggested that the length of the outermost tail feathers of cohorts of male barn swallows changed rapidly in size during a few generations, with an average increase of more than 1 standard deviation or 11 mm (Fig. 3a). Females showed a much weaker temporal trend of 3.3 mm or half a standard deviation (Fig. 3a), giving rise to an increase in sexual size dimorphism during a period of less than 20 years. As there is a genetic correlation for tail length between the sexes (Møller, 1991b; this study), we should expect a response to selection in females due to indirect selection via males, even when direct selection acted on males alone. The temporal change in tail length was much more pronounced than the change in length of the central tail feathers, which did not show any significant temporal change (Table 2). As the genetic correlation between length of outermost tail feathers and length of central tail feathers was weak and nonsignificant, direct selection on tail length should not result in a response to selection for central tail feathers due to indirect selection. In fact, this was what we observed. The observed changes in tail length of males of more than 1 standard deviation are large compared with the maximum micro-evolutionary changes reported in any organisms. Reported records include 0.50–0.75 standard deviations in Darwin's finches (Grant & Grant, 1995) and 0.85–1.10 standard deviations in cliff swallows Petrochelidon pyrrhonota during a single selective event (Brown & Brown, 1998). However, both these responses to selection were followed by an evolutionary response in the opposite direction. Two recent reviews of rates of micro-evolutionary change suggested that changes as large as those reported here for equally short time periods have only rarely been recorded, and when they have, this was due to artificial selection experiments (Hendry & Kinnison, 1999; Stockwell et al., 2003). Therefore, the values reported here for barn swallows are large compared with other values for other organisms during similar time spans (see Stockwell et al., 2003, Fig. 1).

Tail length of male barn swallows was negatively related to environmental conditions in Algeria during spring migration, as indexed by NDVI (Fig. 6a), whereas no significant pattern was obvious in females (Fig. 6b). Although the correlation between male tail length and NDVI in Algeria during spring is nothing but a correlation, this area coincides nonrandomly with the known distribution of banded and recovered Danish barn swallows during spring migration towards their European breeding grounds. Thus, when conditions were favourable during spring migration, tails of males were on average shorter, whereas adverse conditions had the opposite effect. The simplest interpretation is that as environmental conditions deteriorated, selective mortality eliminated an increasing fraction of males in poor condition, and such individuals tend to be short-tailed males. The mechanism underlying these hypothesized effects is unknown. High values of NDVI imply that precipitation is abundant, causing an increase in greenness of the vegetation (Prince & Justice, 1991). In such years, we suggest that insects are abundant, and that barn swallows enjoy superior survival after crossing the Sahara desert. In years with low NDVI values we hypothesize that barn swallows have difficulty recovering after the costly crossing of the Sahara, and that subsequent survival is reduced. This interpretation is supported by the observation that survival of barn swallows decreased in years when NDVI was low (Fig. 2).

An alternative interpretation is that density-dependent effects on the expression of tail length have resulted in tails being particularly long in years when mortality was high. This effect seems unlikely for several reasons. First, population size during the period 1984–2002 was only 4.7–29.8% of the maximum population level in 1977 (Engen et al., 2001). Thus, population size was persistently lower throughout the study period compared with the most recent maximum. Therefore, it is difficult to imagine that any effects of density-dependence would have been important during the study period. Secondly, although the tail feathers are grown in the winter quarters in South Africa during December–March (Møller, 1994), the most important mortality during the annual cycle first takes place during spring migration in Northern Africa (Fig. 1). Thus, any density-dependent effects on tail length cannot act until the following winter when the yearlings grow their first set of long tail feathers. Therefore, any density-dependent effects will be delayed in time and first occur in the next generation compared with the immediate effects of NDVI in Algeria on mean tail length of males. Third, barn swallows are colonial and also highly gregarious outside the breeding season (Møller, 1994). However, there is little evidence of density-dependent effects on survival or reproduction in the Danish barn swallow population (Møller, 1989). Finally, NDVI in South Africa, where the tail feathers are moulted, and NDVI in Algeria, where selective mortality occurs, are only weakly and nonsignificantly correlated (Pearson r = 0.12, n = 19 years, t = 0.49, P = 0.63). This implies that environmental conditions during winter are independent of conditions during spring migration, when mean tail length of yearlings as estimated in the breeding sites is determined.

Change in tail length across generations was quantified in terms of phenotypic change measured in units of standard deviations. The change in tail length in males across cohorts was significantly related to environmental conditions in Algeria during spring migration (Fig. 7a). When environmental conditions were benign, as shown by high values of NDVI, tail length decreased across cohorts, whereas the pattern was the opposite for adverse environmental conditions. Females did not show any significant relationship between change in phenotype across generations and NDVI (Fig. 7b), nor did the length of the central tail feathers. Such change may be due to a micro-evolutionary change or phenotypic plasticity. Mean change in tail length of individual male barn swallows from 1 year to the next is only 2 mm (Møller, 1994), whereas the recorded change reported here exceeds 11 mm. Thus, it seems unlikely that such a cross-generational mean change of more than five times the mean degree of phenotypic plasticity between moults can be accounted for by phenotypic plasticity. In addition, the degree of phenotypic plasticity reflected by the mean increase in tail length for the same individuals between the age of 1 and 2 years decreased significantly during the study (Fig. 4). This pattern is opposite to what would be expected, if phenotypic plasticity was the cause of the increase in tail length. Alternatively, environmental conditions during early development can have strong effects on adult phenotype (e.g. Mousseau & Fox, 1998). Maternal effects in the barn swallow do depend on weather conditions during the period of egg formation (Saino et al., 2004). However, there is only little evidence of a temporal trend in weather during the breeding season in the Danish study site during 1984–2002 (Møller, 2002), as only temperature in April increased significantly by 2.23 °C during this period (before more than 90% of all barn swallows have arrived to the breeding sites) (A. P. Møller, unpublished data). The interpretation that the change in mean phenotype of male barn swallows was a true micro-evolutionary change was supported by analyses of predicted change in tail length based on the breeder's equation. The observed change in tail length of males by 1.265 standard deviation units was very similar to the predicted change of 1.231 standard deviation units. This suggests that the change in tail length could be predicted as a micro-evolutionary response to selection, given the best available estimates of heritability and intensity of selection.

The intensity of selection on tail length decreased significantly for several different components of selection. Selection with respect to breeding date decreased significantly (Fig. 5b) as did viability selection (Fig. 5d), causing overall intensity of selection to decrease significantly (Fig. 5e). It may seem surprising that the intensity of selection decreased, if environmental conditions deteriorated, as indicated by NDVI in Fig. 3c. However, if tail length of male barn swallows reflects the phenotypic quality of males, as shown by a number of different studies (review in Møller, 1994), then we should expect that the remaining part of the population after several episodes of selection to be composed of individuals of higher phenotypic quality. This may result in their ability to survive and reproduce successfully to be superior to that of the population before the environment deteriorated. Alternatively, as male barn swallows now arrive earlier than during the 1980s (A. P. Møller, unpublished data), this may reduce the intensity of selection on breeding date and survival. The duration of the breeding season provides a constraint in the double-brooded barn swallow, because late arriving individuals cannot produce a second brood that will interfere with fall migration, which starts already in August–September. Advanced arrival in the barn swallow due to climate change would reduce constraints on breeding date because a longer breeding season would provide better opportunities for terminating reproduction without interfering with fall migration.

The difference in patterns of survival and phenotypic change between the sexes needs an explanation. Male barn swallows arrive to the breeding grounds before females, on average by 4 days (Møller, 1994). Although this mean difference appears small for a journey exceeding 12 500 km and lasting a minimum of 36–50 days (Glutz von Blotzheim, 1985), cutting off 4 days, which equal 9.3%, from the total duration of migration may be very difficult. Barn swallows appear already to be migrating at maximum speed, which is 250–350 km per day for a 1-month period, according to recoveries of banded birds (Glutz von Blotzheim, 1985). The faster speed of migration by males compared with females may put an extra cost on migration, and we hypothesize that this cost is particularly severe during adverse environmental conditions. Therefore, we expect males to suffer more than females from adverse environmental conditions during spring migration, because females do not pay this extra cost due to their migration being slower than that of males.

Barn swallows breed in northern temperate and subtropical zones, have trans-equatorial migration and winter in sub-Saharan Africa and parts of South Asia and South America (Møller, 1994). Therefore, individuals from particular breeding populations are exposed to environmental conditions in several different areas during the annual cycle. All of these areas must provide suitable conditions for survival in order to maintain a viable population. Where barn swallow populations and populations of other migratory birds are regulated is of great scientific and conservation interest. Current evidence suggests that most mortality of migratory birds occurs outside the breeding season (Sillett & Holmes, 2002; Szép & Møller, 2004). Given that tail length of male barn swallows shows a temporal trend (Fig. 3a), we can cross-validate the area identified as being important for survival, because environmental conditions in that area must show a temporal trend similar to that for the changing phenotypic character. We tested for temporal trends in NDVI for the six identified areas of importance for survival. Only NDVI in Algeria showed a consistent decrease during the study period (Fig. 3c), as predicted if NDVI was the cause of the change in phenotype. This finding supports our conclusion that it was environmental conditions on spring migration in Northern Africa that was the cause of the altered phenotype in male barn swallows.

In conclusion, this study of micro-evolutionary response to selection arising from changing climatic conditions in areas visited by barn swallows during their spring migration suggests that the size of secondary sexual characters can change very rapidly. Such changes can be specific to phenotypic characters in one sex, even when the characters are expressed in both sexes. These findings have the additional implication that the expression of secondary sexual characters can be considered early warning signals of marked environmental change.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study area
  6. North Atlantic Oscillation
  7. Normalized Difference Vegetation Index
  8. Survival analysis
  9. Selection analysis
  10. Statistical analysis
  11. Results
  12. Survival estimates and environmental conditions
  13. Selection of areas of importance for survival
  14. Survival and environmental conditions
  15. Temporal change in tail length
  16. Selection on tail length
  17. Change in tail length in relation to environmental conditions
  18. Discussion
  19. Acknowledgments
  20. References
  • Anderson, D.R. & Burnham, K.P. 1999. Understanding information criteria for selection among capture–recapture or ring recovery models. Bird Study 46: S14S21.
  • Anderson, D.R., Burnham, K.P. & White, G.C. 1994. AIC model selection in overdispersed capture–recapture data. Ecology 75: 17801793.
  • Andersson, M. 1994. Sexual Selection. Princeton University Press, Princeton, NJ.
  • Arnold, S.J. & Wade, M.J. 1984. On the measurement of natural and sexual selection: theory. Evolution 38: 709719.
  • Badyaev, A.V. & Martin, T.E. 2000. Sexual dimorphism in relation to current selection in the house finch. Evolution 54: 987997.
  • Badyaev, A.V., Hill, G.E., Stoehr, A.M., Nolan, P.M. & McGraw, K.J. 2000. The evolution of sexual size dimorphism in the house finch. II. Population divergence in relation to local selection. Evolution 54: 21342144.
  • Badyaev, A.V., Hill, G.E., Beck, M.L., Dervan, A.A., Duckworth, R.A., McGraw, K.J., Nolan, P.M. & Whittingham, L.A. 2002. Sex-biased hatching order and adaptive population divergence in a passerine bird. Science 295: 316318.
  • Bartlett, M.S. 1952. On the theoretical specification of sampling properties of autocorrelated time series. J. R. Statist. Soc. Suppl. 8: 27411.
  • Boone, R.B., Galvin, K.A., Smith, N.M. & Lynn, S.J. 2000. Generalizing El Niño effects upon Maasai livestock using hierarchical clusters of vegetation patterns. Photogramm. Eng. Remote Sens. 66: 737744.
  • Bradshaw, W.E. & Holzapfel, C.M. 2001. Genetic shift in photoperiodic response correlated with global warming. Proc. Nat. Acad. Sci. USA 98: 1450914511.
  • Brown, C.R. & Brown, M.B. 1998. Intense natural selection on body size and wing and tail asymmetry in cliff swallows during severe weather. Evolution 52: 14611475.
  • Brown, J.L., Li, S.H. & Bhagabati, N. 1999. Long-term trend toward earlier breeding in an American bird: a response to global warming? Proc. Nat. Acad. Sci. USA 96: 55655569.
  • Burnham, K.P., Anderson, D.R., White, G.C., Brownie, C. & Pollock, K.H. 1987. Design and Analysis Methods for Fish Survival Experiments Based on Release–Recapture. American Fisheries Society Monographs 5, Bethesda, MD.
  • Chen, D. & Brutsaert, W. 1998. Satellite-sensed distribution and spatial patterns of vegetation parameters over a tallgrass prairie. J. Atmos. Sci. 55: 12251238.
  • Clobert, J. & Lebreton, J.-D. 1987. Recent models for mark-recapture and mark resighting data: a response to C. Brownie. Biometrics 43: 10191022.
  • Crick, H.Q.P., Dudley, C., Glue, D.E. & Thomson, D.L. 1999. Climate change related to egg-laying trends. Nature 399: 423424.
  • Darwin, C. 1871. The Descent of Man and Selection in Relation to Sex. John Murray, London, UK.
  • Dunn, P.O. & Winkler, D.W. 1999. Climate change has affected the breeding date of tree swallows throughout North America. Proc. R. Soc. Lond. B 266: 24872490.
  • Engen, S., Sæther, B.-E. & Møller, A.P. 2001. Stochastic population dynamics and time to extinction of a declining population of barn swallows. J. Anim. Ecol. 70: 789797.
  • Etterson, J.R. & Shaw, R.G. 2001. Constraint on adaptive evolution to global warming. Science 294: 151154.
  • Evans, M. R.. & Thomas, A.L.R. 1997. Testing the functional significance of tail streamers. Proc. R. Soc. Lond. B 264: 211217.
  • Falconer, D.S. & Mackay, T.F.C. 1996. Introduction to Quantitative Genetics, 4th edn. Longman, New York, NY.
  • Glutz von Blotzheim, U.N. (ed.) 1985. Handbuch der Vögel Mitteleuropas, Vol. 10. AULA-Verlag, Wiesbaden, Germany.
  • Grant, P.R. & Grant, B.R. 1995. Predicting microevolutionary responses to directional selection on heritable variation. Evolution 49: 241245.
  • Hendry, A.P. & Kinnison, M.T. 1999. The pace of modern life: measuring rates of contemporary microevolution. Evolution 53: 16371653.
  • Hoffmann, A.A. & Parsons, P.A. 1997. Extreme Environmental Change and Evolution. Cambridge University Press, Cambridge, UK.
  • Hurrell, J.W. 1995. Decadal trends in the North Atlantic Oscillation: regional temperatures and precipitation. Science 269: 676679.
  • Jennions, M.D., Møller, A.P. & Petrie, M. 2001. Sexually selected traits and adult survival: A meta-analysis of the phenotypic relationship. Q. Rev. Biol. 76: 336.
  • Lebreton, J.-D., Burnham, K.P., Clobert, J. & Anderson, D.R. 1992. Modeling survival and testing biological hypotheses using marked animals: a unified approach with case studies. Ecol. Monogr. 62: 67118.
  • Lynch, M. & Lande, R. 1993. Evolution and extinction in response to environmental change. In: Biotic Interactions and Global Change (P. M.Kareiva, J. G.Kingsolver & R. B.Huey, eds), pp. 234250. Sinauer, Sunderland, MA.
  • Lynch, M. & Walsh, B. 1998. Genetics and Analysis of Quantitative Traits. Sinauer, Sunderland, MA.
  • Møller, A.P. 1988. Female choice selects for male sexual tail ornaments in the monogamous swallow. Nature 332: 640642.
  • Møller, A.P. 1989. Population dynamics of a declining swallow Hirundo rustica L. population. J. Anim. Ecol. 58: 10511063.
  • Møller, A.P. 1990. Effects of an haematophagous mite on the barn swallow (Hirundo rustica): a test of the Hamilton and Zuk hypothesis. Evolution 44: 771784.
  • Møller, A.P. 1991a. Viability is positively related to degree of ornamentation in male swallows. Proc. R. Soc. Lond. B 243: 145148.
  • Møller, A.P. 1991b. Sexual selection in the monogamous barn swallow (Hirundo rustica). I. Determinants of tail ornament size. Evolution 45: 18231836.
  • Møller, A.P. 1991c. Sexual ornament size and the cost of fluctuating asymmetry. Proc. R. Soc. Lond. B 243: 5962.
  • Møller, A.P. 1992. Parasites differentially increase the degree of fluctuating asymmetry in secondary sexual characters. J. Evol. Biol. 5: 691699.
  • Møller, A.P. 1993. Morphology and sexual selection in the barn swallow Hirundo rustica in Chernobyl, Ukraine. Proc. R. Soc. Lond. B 252: 5157.
  • Møller, A.P. 1994. Sexual Selection and the Barn Swallow. Oxford University Press, Oxford, UK.
  • Møller, A.P. 2002. North Atlantic Oscillation (NAO) effects of climate on the relative importance of first and second clutches in a migratory passerine birds. J. Anim. Ecol. 71: 201210.
  • Møller, A.P. & Szép, T. 2002. Survival rate of adult barn swallows Hirundo rustica in relation to sexual selection and reproduction. Ecology 83: 22202228.
  • Møller, A.P., Barbosa, A., Cuervo, J.J., de Lope, F., Merino, S. & Saino, N. 1998. Sexual selection and tail streamers in the barn swallow. Proc. R. Soc. Lond. B 265: 409414.
  • Maurer, B.A. 1994. Geographical Population Analysis Tools for the Analysis of Biodiversity. Blackwell, Oxford, UK.
  • Merilä, J., Kruuk, L.E.B. & Sheldon, B.C. 2001. Cryptic evolution in a wild bird population. Nature 412: 7679.
  • Moss, R., Oswald, J. & Baines, D. 2001. Climate change and breeding success: decline of the capercaillie in Scotland. J. Anim. Ecol. 70: 4761.
  • Mousseau, T.A. & Fox, C.W. (eds) 1998. Maternal Effects as Adaptations. Oxford University Press, New York, NY.
  • Norberg, R. Å. 1994. Swallow tail streamer is a mechanical device for self-deflection of tail leading edge, enhancing aerodynamic efficiency and flight manoeuvrability. Proc. R. Soc. Lond. B 257: 227233.
  • Osborne, P.E., Alonso, J.C. & Bryant, R.G. 2001. Modelling landscape-scale habitat use using GIS and remote sensing: a case study with great bustards. J. Appl. Ecol. 38: 458471.
  • Paruelo, J.M., Epstein, H.E., Lauenroth, W.K. & Burke, I.C. 1997. ANPP estimates from NDVI for the Central Grassland Region of the United States. Ecology 78: 953958.
  • Pomiankowski, A. & Møller, A.P. 1995. A resolution of the lek paradox. Proc. R. Soc. Lond. B 260: 2129.
  • Prince, S.D. & Justice, C.O. (eds) 1991. Coarse resolution remote sensing of the Sahelian environment. Int. J. Remote Sens. 12: 11331421.
  • Réale, D., McAdam, A.G., Boutin, S. & Berteaux, D. 2003. Genetic and plastic responses of a northern mammal to climate change. Proc. R. Soc. Lond. B 270: 591596.
  • Sæther, B.-E., Tufto, J., Engen, S., Jerstad, K., Røstad, O.W. & Skåtan, J.E. 2000. Population dynamical consequences of climate change for a small temperate songbird. Science 287: 854856.
  • Saino, N., Primmer, C.R., Ellegren, H. & Møller, A.P. 1997. An experimental study of paternity and tail ornamentation in the barn swallow (Hirundo rustica). Evolution 51: 562570.
  • Saino, N., Romano, M., Ambrosini, R., Ferrari, R.P. & Møller, A.P. 2004. Timing of reproduction and egg quality covary with temperature in the insectivorous Barn Swallow (Hirundo rustica). Funct. Ecol. 18: 5057.
  • Schmidt, H. & Karnieli, A. 2000. Analysis of the temporal and spatial vegetation patterns in a semi-arid environment observed by NOAA AVHRR imagery and spectral ground measurements. Int. J. Remote Sens. 23: 39713990.
  • Sillett, T.S. & Holmes, R.T. 2002. Variation in survivorship of a migratory songbird throughout its annual cycle. J. Anim. Ecol. 71: 296308.
  • Skalski, J.R., Hoffmann, A. & Smith, S.C. 1993. Testing the significance of individual- and cohort-level covariates in animal survival studies. In: Marked Individuals in the Study of Bird Population (J.-D.Lebreton & P. M.North, eds), pp. 928. Birkhäuser Verlag, Basle, Switzerland.
  • Stockwell, C.A., Hendry, A.P. & Kinnison, M.T. 2003. Contemporary evolution meets conservation biology. Trends in Ecol. Evol. 18: 94101.
  • Szép, T. & Møller, A.P. 2004. Using remote sensing data to identify migration and wintering areas, and to analyze effects of environmental conditions on migratory birds. In: Birds of Two Worlds (P.Marra & R.Greenberg, eds). Smithsonian Institution Press, Washington, DC.
  • Tucker, C.J., Dregne, H.E. & Newcomb, W.W. 1991. Expansion and contraction of the Sahara Desert between 1980 and 1990. Science 253: 299301.
  • Wallin, D.O., Elliott, C.C.H., Shugart, H .H., Tucker, C.J. & Wilhelmi, F. 1992. Satellite remote sensing of breeding habitat for an African weaver-bird. Landscape Ecol. 7: 8799.
  • White, G.C. & Burnham, K.P. 1999. Program mark: Survival estimation from populations of marked animals. Bird Study 46: S120S138.
  • Winkel, W. & Hudde, H. 1997. Long-term trends in reproductive traits of tits (Parus major, P. caeruleus) and pied flycatchers Ficedula hypoleuca. J. Avian Biol. 28: 187190.