Migrants vs non-migrants

A considerable component of body mass-independent variation in avian BMR can statistically be explained by migratory tendency: the minimum normothermic maintenance energy requirements of migrants are significantly higher than those of non-migrants, at least when phylogeny is not accounted for. This has implications for comparative studies, since comparisons of the energetic traits of non-migrants with those of migrants may lead to misleading conclusions regarding physiological adaptation. The physiological divergence we have identified between migrant and non-migrant birds is consistent with observations that several species of migrant shorebirds have higher BMR than expected on the basis of body mass [30]–[32].

Why should BMR be higher in migrants than in non-migrant species? One possibility is that compared to species with more sedentary life histories the metabolic machinery for long-distance migration involves elevated maintenance costs. Avian energy intake rates as well as maximum thermogenic metabolic rates are positively correlated with BMR [46], [47]. A mechanistic link between elevated BMR and the capacity for sustained activity is supported by the observation that in Rock Doves (Columba livia), the metabolic intensity (cytochrome c oxidase activities) of pectoral muscles is higher in active birds than in sedentary individuals. The elevated BMR of migrant species may also in part reflect the timing of metabolic measurements and the influence of flight muscle hypertrophy preceding long-distance migratory flights in species such as Red Knots Calidris canutus [48].

A second set of possible explanations for the higher BMR of migrants compared to non-migrants concerns their uneven latitudinal distribution. During the breeding season migrants generally occur in environments that are colder (Figure 2) and where the short time window available for breeding may impose particularly high energy demands. According to the former view, the elevated BMR of migrants may simply reflect the correlation between BMR and Temp avg. This idea is not new; Kvist and Lindström [49] showed that the BMR of migrant shorebirds is highest on the Arctic breeding grounds before migration, and lowest while on tropical wintering grounds. Lindström and Klaassen [32] confirmed the generality of elevated BMR in shorebirds while in the Arctic, and hypothesized that the reduction of BMR exhibited by migrants as they move from high to tropical latitudes reflects changing requirements for thermoregulatory heat production. Many non-migrant species rapidly up- or down-regulate BMR in response to thermal acclimation [22], [23], [50], with BMR up-regulation associated with cold air temperatures and vice versa. Intra-individual variation in BMR often reflect changes in organ size [51], [52], but may also reflect changes in the metabolic intensity of specific tissues [53].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Partial residual plots of core environmental correlates of BMR for the combined dataset of both migrant (black circles) and non-migrant birds (open circles).

Each panel shows the effect of a single predictor on BMR controlled for body size and Pass/non-p membership. Regression lines are those significant for combined non-migrant – migrant data (Table 2). Negative residual outlier (<−0.4 partial residual value) is the group-living Green Woodhoopoe (Phoeniculus purpureus).

https://doi.org/10.1371/journal.pone.0003261.g002

In our analysis, migrant data came from much higher latitudes and colder regions than data for non-migrants. In the combined non-migrant/migrant dataset several environmental variables exceed migratory tendency as predictor (Table 2): after accounting for either Temp avg, PET or Temp range there is no significant difference in BMR between migrants and non-migrants. In view of the considerable phenotypic flexibility in BMR exhibited by long-distance migrants [25], [32], [54], [55], and the consistent negative relationship between BMR and air temperature in laboratory acclimation studies [22], [23], [25], [50], it seems likely that the higher BMR of migrants we report here is determined in part by temperature effects. However, the BMR data currently available for long-distance migrants on their tropical wintering grounds are too few to rigorously test this hypothesis.

Environmental correlates of avian BMR

Our analyses confirm the previously observed considerable body mass-independent variation in avian BMR that can be attributed to several environmental variables. For interpretation of specific environmental effects we focus on single-predictor environmental relationships as high collinearity of environmental variables (Table 1) hampers the interpretation of multi-predictor models. Average temperature (Temp avg) emerges as the most significant single environmental predictor of BMR, confirming the findings of White et al. (2007). BMR is significantly lower in warmer environments among all species included in our analysis, as well as within non-migrant and migrant subsets. These observations are consistent with Weathers' [14]'s finding that avian BMR increases with increasing latitude, Wiersma et al.'s [12] observation that tropical birds have lower BMRs than their temperate counterparts, as well as with the negative correlation between temperature and BMR in mammals [10]. A negative effect of Temp avg on avian BMR has also been demonstrated to explain intraspecific variation [16], [56]. Moreover, numerous studies of thermal acclimation or acclimatization have found that birds adjust BMR in response to changing thermoregulatory demands [21]–[23], [25], [50], [57], [58]. In studies involving thermal acclimation under laboratory conditions, BMR can be adjusted by more than 20% over time scales of several weeks [22], [25], [50].

Within and among migrants and non-migrants, BMR exhibited no correlation or a negative relationship with net primary productivity (which correlates positively with average temperature), a similar observation to that of White et al. (2007). This conflicts with the pattern among five species of Peromyscus mice under common-garden conditions [59]. Furthermore, the lack of a positive effect of habitat energy availability on avian BMR indicates that the apparent independence of individual energy demand and supply at broad scales found for field metabolic rate [20] is manifested at the level of BMR. An increase in maintenance metabolism with decreasing NPP could also potentially reflect greater mobility among desert species that have to cover larger areas to acquire sufficient resources to breed. Unpredictable fluctuations in food availability, driven by erratic rainfall, have traditionally been viewed as one of the major factors driving the evolution of low BMR in desert endotherms [60], and more recently have been invoked as a major determinant of zoogeographical patterns of mammalian BMR [8], [10] Whereas precipitation variability has been found to be strongly negatively associated with the BMR of small mammals [10], a recent study found the opposite for birds (White et al. 2007). Here we were not able to confirm any significant effect of precipitation variability on BMR. We interpret our results as evidence that the low BMR of desert bird species is determined primarily by temperature effects, rather than magnitude and variability in energy availability.

We emphasize, however, that our findings are correlational. BMR has long been viewed as a fixed, taxon-specific parameter, with adaptation inferred from interspecific variation remaining after scaling and/or phylogenetic patterns of descent have been accounted for. However, there is increasing evidence for considerable phenotypic flexibility in BMR, with substantial within-individual adjustments occurring over short time scales [61]. Thus, interspecific variation in BMR reflects in part the conditions to which birds are acclimatized or acclimated at the time of metabolic measurements, and not necessarily genotypic divergence.

Disparate phylogenetic structure

Our results indicate a hitherto unappreciated complexity in the phylogenetic structure of the examined associations. Previous analyses of these types of data [40] and of BMR in birds in particular [13], [33] have assumed a common phylogenetic dependence across the whole data set analyzed. In this dataset migrants appear to have a much lower degree of phylogenetic dependence than non-migrants. This is consistent with the idea that in migrants post- and pre-migration changes in BMR within individuals cause intraspecific variation which then leads to BMR to appear phylogenetically more labile than in non-migrants (at least when measured in summer). Alternatively, it may also reflect a closer adaptation to more constant environmental conditions in migrants, but as we do not find much stronger environmental associations for them we believe this to be unlikely. These results suggest that future analyses may benefit from exploring more complex mosaic models of evolution and, ideally, datasets that include intraspecific genetic information.

Conclusions

By analyzing BMR of migrants and non-migrants separately for a large number of species we have shed further light on an apparent pattern of physiological divergence among birds. The higher summer BMR of many migrant populations confirms that the observations of several authors working on migrant shorebirds [30]–[32] represent a general pattern. We find that the higher BMR of migrants may at least partly be due to the latitudinal distribution of their breeding grounds at higher latitudes and thus colder climates. While ambient temperature exerts a strongly negative effect on both migrant and non-migrant BMR, our analysis reveals that overall different environmental variables are correlated with the variation of BMR in these two groups. Our results confirm the need to consider and, ideally, to quantify environmental effects when addressing topics such as the body size dependence of metabolic rate or when developing models of population energy fluxes.

In conclusion it appears that broad-scale climatic gradients constraints, specifically those related to temperature, present a stronger constraint on avian BMR than migratory tendency. But different climatic associations and a much weaker phylogenetic signal point to different control of BMR in migrants that appears to be characterized by strong phenotypic flexibility. Further empirical work on winter vs. summer BMR in both migrants and non-migrants across environments will help to establish the full extent of phenotypic flexibility in each group and provide a fuller picture of its environmental determinants. Ultimately, such work may be extended to help understand the shapes and phylogenetic inertia of reaction norms across different migratory and other behavioral strategies. In our study we were able to only indirectly draw inference about the exact pathway that causes the statistically strong association between climate and phenotypic variation. But especially for migratory tendency the phylogenetically labile environmental control of BMR emphasizes the significance of phenotypic plasticity. Promising additional insights are likely to be gained from any study that was able to explicitly and simultaneously address environment - BMR association within individuals, within (and ideally across) populations and across species. Such studies, logistically challenging they may be, may be able to reconcile broad-scale comparative/eco-geographic perspectives that are concerned with the broad interspecific patterns with ecophysiological viewpoints that have helped appreciate the importance of small scale and intraspecific variation arising from phenotypic flexibility.

BMR and body mass data

We obtained BMR (Watts) and body mass (M, g) data for wild-caught populations of 135 species from McKechnie et al. [35]. We included estimates of BMR irrespective of the sample size from which they were generated, but tested for bias resulting from this approach (see below). For each datum, we consulted the original source for the location at which the experimental individuals were captured (online supporting material for McKechnie et al. 2006 at http://dx.doi.org/10.1098/rspb.2005.3415). In cases where the co-ordinates of the capture site were not reported, we obtained these from the Alexandria Library Digital Gazetteer (http://www.alexandria.ucsb.edu/gazetteer/). We classified each species as non-migrant (71 species) or migrant (64 species) using secondary literature and further separated migrants into those performing long-distance (inter-continental, 29 species) or short distance (intra-continental, 35 species) seasonal movements. Data included only birds captured on their breeding grounds (i.e., data from winter quarters are excluded). For the environmental analyses, we excluded 38 species from the dataset of McKechnie et al. [35] for which not all the environmental variables (see below) were available. This yielded a final data set of 97 species.

Environmental predictor variables

For each BMR datum, we extracted the following environmental variables from the CRU CL2.0 data-set in 10-minute (0.167°) resolution and over the period 1961–1990 [37]: mean monthly temperature (Temp avg, °C), mean monthly temperature of hottest three months (Temp max, °C), mean monthly precipitation (Prec avg; mm/month), absolute difference between average January and July temperature (Temp range, °C), log₁₀-transformed difference between average maximum and minimum monthly precipitation (Prec range, mm), and mean coefficient of variation of the monthly precipitation estimates across all 30 years (Prec CoV). We obtained estimates of net primary productivity (NPP, t Carbon ha⁻¹ y⁻¹) estimates for the period 1960–90 from the DOLY model in 0.5° resolution [38] and evaluate effects of average annual NPP (NPP avg, t Carbon ha⁻¹ y⁻¹) and average NPP of the most productive three months (NPP max, t Carbon ha⁻¹ 3months⁻¹). We also obtained average potential evapotranspiration (PET avg, mm) data from a 0.5° gridded global dataset [39]. PET is the amount of moisture which, if available, would be removed from a given land area by soil evaporation and plant transpiration. We used these data to calculate an aridity index (Aridity Index) as average monthly precipitation/average monthly PET, where moist areas have high aridity index and arid areas a low index. Several other variables measuring environmental conditions or their seasonality were evaluated, but excluded from the final analysis for their poor predictive ability or co-linearity with already included predictors.

Model fitting and analysis – environmental effects on BMR

We initially analyzed the BMR data following McKechnie et al. [35], using a generalized least squares approach whereby covariance among species is accounted for using a phylogeny [40]–[42]. The strength of the phylogenetic signal in each model was assessed using the parameter λ which measures and controls for the degree of phylogenetic dependence in the model residuals [42], [43], estimated using a maximum-likelihood approach.

Our first set of analyses addresses the effects of single environmental predictors on avian BMR without distinguishing between migratory strategies. We fit general linear models (GLM) to the overall data set, using log₁₀-transformed BMR as a continuous dependent variable, and successively add log₁₀-transformed M (continuous) and passerine/non-passerine membership (Pass/non-p., categorical). The passerine/non-passerine variable was added since there are significant differences in M between passerines and non-passerines, and omitting this variable leads to an under-estimation of the scaling exponents relating BMR to M [35], [44].

Environmental correlates on avian BMR may arise via the different geographic distributions and environmental associations of migratory vs. non-migratory species with different BMR. In the next set of analyses we therefore sought to examine how in this dataset environmental variables and thus environmental niches of birds and their migratory strategy interact with each other, and how this is affected by phylogeny. To do this we fitted linear models in which the environmental variables were individually treated as response variables, and migratory strategy used as a predictor. These analyses therefore test whether there are significant differences between migrants and non-migrants in terms of the broad-scale environments they occupy. We controlled for phylogeny in fitting these models and estimated the λ statistic, as described above, to determine whether there was phylogenetic signal in the patterns of environment occupancy.

Model fitting and analysis – contrasting migrants and non-migrants

Next we repeated the previous analysis of single environmental correlates of BMR separately for migrants and non-migrants, with M and Pass/non-p as covariates and evaluated model fits using the Akaike Information Criterion [45]. Using the approach of analyzing single environmental predictors it was not clear whether the predictors that explain BMR variation are the same for migrants and non-migrants. This is because first, a large number of potential combinations of predictors exist many of which possess similar degrees of explanatory power; and second, the predictors are correlated with each other. Further complicating this analysis was the problem that the degree of phylogenetic dependence differs between the migrants and non-migrants. This makes it difficult to deal with both groups of species together in a single analysis. In the next stage of the analysis we therefore (i) reduced the set of predictors down to a core set of environmental variables measuring key aspects of the environment. Specifically we looked at the effects of average NPP, average temperature and average precipitation. (ii) We analyzed migrants and non-migrants separately to see whether the overall responses appeared different. (iii) We combined all variables into a single multiple regression analysis in order to examine the effects of all of these simultaneously and analyzed the effects using a sequential decomposition of variance. We introduced M and Pass/non-p as initial predictors. We then added Temp_avg as this was, according to the initial analysis, clearly the strongest environmental correlate of BMR. The other predictors were then added after this (note that the order in which these variables were added did not affect the results). We note that any broad-scale analysis of this sort is complicated by the overlap in information (collinearity) between alternative predictors such different environmental variables or environmental and behavioral-ecological determinants, and the cause - effect assumptions inherent in statistical modeling approaches.

Model fitting and analysis – effects of BMR sample size

In the above analyses, we included BMR data irrespective of the number of individuals per population sampled. To verify that the inclusion of data measured in only one or two individuals did not bias the results, we repeated the environmental variable analyses described above after weighting the BMR and M data for each species by the number of individuals used. In no instance did the significance, direction or strength of an effect change. At most, partial coefficient values altered slightly after data were weighted.