Basal metabolism in tropical birds: latitude, altitude, and the ‘pace of life’
Summary
- Life history varies across latitudes, with the ‘pace of life’ being ‘slower’ in tropical regions. Because life history is coupled to energy metabolism via allocation tradeoffs and links between performance capacity and energy use, low metabolic intensity is expected in tropical animals. Low metabolism has been reported for lowland tropical birds, but it is unclear if this is due to ‘slow’ life history or to a warm, stable environment.
- We measured basal metabolic rates (BMR) of 253 bird species across a 2·6 km altitude gradient in Peru. We predicted higher BMR at high altitude due to lower temperatures leading to elevated thermoregulatory costs. We also tested for BMR differences between widely separated tropical regions (Peru and Panama), and between tropical- and temperate-breeding birds.
- We found no effect of altitude on BMR in Peruvian species and no difference in BMR between Peruvian and Panamanian birds, suggesting that BMR in Neotropical birds is consistent and independent of environmental temperature. In a data set encompassing more than 500 species, tropical birds had significantly lower BMR than temperate-breeding birds.
- In contrast to several recent analyses, we found higher BMR in passerine birds than in non-passerines, independent of breeding latitude.
- Breeding latitude affects BMR, but diversity in avian life history within and between temperate and tropical regions may explain some of the residual variation in BMR after accounting for body mass and breeding latitude. Future studies of links between life history, metabolism and environmental factors might benefit from examining these variables within individual species as well as across broad geographic contrasts.
Introduction
Life histories are fitness-maximizing solutions to environmental challenges that evolve in response to numerous ecological factors, both abiotic (e.g. temperature, rainfall) and biotic (e.g. competition, predation, pathogens and parasites). Theory predicts that life history is largely driven by resource limitations and allocation tradeoffs (Roff 1992; Stearns 1992; Reznick 2014). This view stems from the assumption that food is limited (in acquisition or processing), and the largely mutually exclusive ways that animals can allocate energy: for maintenance (resting and activity metabolism, immunity), somatic growth or reproduction. These constraints, coupled with selection imposed by expected longevity on the urgency of reproduction, should generate a ‘slow-fast’ continuum of life histories ranging from long-lived species that invest heavily in maintenance but relatively little in fecundity to short-lived species that invest little in self-maintenance and produce numerous rapidly-growing offspring.
In addition to extrinsic ecological influences, life history is inescapably dependent on intrinsic organismal properties (physiology). The ‘slow-fast’ continuum is coupled to energy metabolism via allocation tradeoffs and by linkages between organismal performance and the physiological systems that support it (Ricklefs & Wikelski 2002; Williams et al. 2010). For example, high rates of offspring production require high rates of resource acquisition and hence intense foraging and rapid food processing; in turn, these require costly high-capacity musculoskeletal, cardiorespiratory and digestive systems. The flux of energy through the animal, sometimes called the ‘rate of living’ (Pearl 1928) or the ‘pace of life’ (Ricklefs & Wikelski 2002), is expected to be high at the ‘fast’ end of the life-history continuum and low at the ‘slow’ end. These predictions have been tested, with varying support, in species ranging from nematodes (Van Voorheis & Ward 1999) to mammals (Speakman et al. 2002; White & Seymour 2004).
Contrasts in the ‘pace of life’ between temperate and tropical birds are of special interest (we use latitude as a proxy for major environmental characteristics: e.g. the tropics have stable, warm temperatures, low seasonality and high biotic diversity). Compared to temperate birds, tropical birds – particularly those from humid forests – are thought to have ‘slow’ life histories: small clutches (Moreau 1944; Kulesza 1990), slow nestling growth and maturation (Ricklefs 1976; Russell, Yom-Tov & Geffen 2004; Schaefer et al. 2004; Martin et al. 2011) and low adult mortality (Fogden 1972; Snow & Lill 1974; Francis, Terborgh & Fitzpatrick 1999; Gill & Haggerty 2012). Tropical birds tend to be sedentary whereas many temperate breeders migrate over long distances, a behaviour requiring massive use of energy (Jetz, Freckleton & McKechnie 2008). Accordingly, if life history predicts the energetic ‘pace of life’, then tropical birds should have lower metabolic rates than temperate birds.
Early work on tropical bird energetics was ambivalent, with low metabolic rates reported in some species (Weathers 1979; Hails 1983; Tieleman et al. 2006) but not others (Scholander et al. 1950; Bennett & Harvey 1987; Vleck & Vleck 1979). Two recent studies support the ‘pace of life’ hypothesis. Wikelski et al. (2003) found lower resting metabolic rates in a tropical population of stonechats (Saxicola torquata) than in stonechats native to temperate areas (but see Tieleman 2007). An interspecific comparison of lowland tropical birds and north-temperate breeders found lower basal metabolic rate (BMR) and lower maximal metabolism during exercise and thermogenesis in tropical species (Wiersma, Chappell & Williams 2007; Wiersma et al. 2007). The low limits may be functionally related to small visceral organs (Wiersma, Nowak & Williams 2012) and differences at the cellular and subcellular levels (Jimenez et al. 2014). These authors interpreted low metabolic rates as manifestations of a ‘slow’ life history. However, the limited set of tropical species (primarily from lowland rain forest in Panama) does not rule out other explanations. For example, low metabolic rates might result from residence in a benign tropical climate (White et al. 2007; Wiersma, Chappell & Williams 2007; Jetz, Freckleton & McKechnie 2008).
Comparisons of bird species from adjacent habitats with stable but contrasting thermal regimes, such as along tropical altitude gradients, could offer insights into relationships between environment, life history and energy metabolism. We used BMR from a large sample of forest birds inhabiting a 2·6 km altitude gradient in south-eastern Peru to address four questions. First, we tested if differences in native altitude influenced BMR. Low ambient temperatures at high altitude should increase costs of thermoregulation and result in higher BMR according to the ‘pace of life’ hypothesis (higher maintenance costs and possible cold acclimatization; Cooper & Swanson 1994; McKechnie & Swanson 2010). Secondly, we asked if the low BMR of Panamanian birds is also characteristic of geographically distant Peruvian species, as expected if the tropical environment per se is important in determining BMR. Thirdly, we combined our results with published data to test whether BMR is lower in tropical birds than in temperate birds, another ‘pace of life’ prediction that to date has been confirmed for Panamanian lowland forest species. Finally, we used this large data set to explore a long-standing question in avian ecological and evolutionary physiology: whether BMR differs between passerines and non-passerines. We used both conventional and phylogenetically informed statistics, and this is the first major analysis of avian BMR based on the Hackett et al. (2008) phylogenetic tree rather than that of Sibley, Ahlquist & Monroe (1988). Given the many differences between the two phylogenies, our results have implications for understanding how BMR varies across avian orders, as well as how it is influenced by latitude and altitude.
Materials and methods
Field Sites
We worked at three field stations in south-eastern Peru. Two are in steep mountain terrain along the south-eastern edge of Parque Nacional del Manu. Wayqecha (altitude 2550–3200 m; 13°10′S, 71°35′W) is in cloud forest near tree line. San Pedro (1300–1600 m; 13°03′S, 71°32′W) spans montane rain forest and lower cloud forest (Jankowski et al. 2013). Pantiacolla (380–500 m; 12°39′ S, 71°13′W) is in lowland rain forest. The stations are roughly aligned along a 70 km northeast–southwest axis. Annual temperature means ranged from 11·2 °C (Wayqecha) to 23·2 °C (Pantiacolla), and for all stations, monthly means differed by 3 °C or less (Table S1, Supporting information).
Bird Capture and Handling
We measured metabolism between July and December 2011–2013. Birds were mist-netted in the afternoon (15:00–17:00 h local time); individuals with brood patches or other indications of active reproduction were released. Non-reproductive birds were held in cloth-covered cages in quiet rooms and provided with water but not food (except for hummingbirds, which were given 25% sugar water). After sunset (c. 19:00 h), birds were weighed (±0·05 g for birds <125 g; FlipScale F2, www.myweigh.com/pocket/; ±0·5 g for larger species; Pesola 300 g spring scale; Baar, Switzerland) and placed in the respirometry system. Measurements of BMR began several hours later, usually after 23:00 h, to ensure that birds were postabsorptive. All birds (except hummingbirds) had been fasting for at least 4–5 h prior to BMR measurements. Fast duration in hummingbirds was > 2 h, sufficient for absorption of sugar (Karasov et al. 1986). All individuals were tested for BMR at 30 °C and most were also tested at 32–34 °C; in addition, we measured metabolism at 10 and 20 °C for nearly all individuals. Body temperature (Tb) was obtained immediately after BMR measurements using fine-gauge thermistors inserted at least 5–10 mm into the cloaca. After the conclusion of measurements, birds were reweighed and the mean of initial and final weights was used as body mass. Birds were released the following morning at the site of capture.
In addition to Peruvian species, we report BMR measured in the laboratory of one of the authors (M.A. Chappell) at the University of California, Riverside (‘UCR’ data set). Specimens were wild-caught or captive-bred; all were housed in large outdoor flight cages.
Capture and handling procedures were approved by the Institutional Animal Care and Use Committees at the University of California, Riverside (protocol # 0408026, 20070023, 20100048) and the University of Florida (Protocol #: 201106068), and by permits from the government of Peru (0239-2013 MINAGRI-DGFFS/DGEFFS 2013).
Respirometry
) with open-circuit respirometry (identical equipment and procedures were used at all stations). Air flow (supplied with aquarium pumps) was dried with silica gel and divided among four metered channels (FlowBar; Sable Systems, Las Vegas, NV, USA) calibrated with a mass flow controller (Sensirion 4200, Staefa, Switzerland) verified against a dry volume meter (DTM115; Singer American, Stanford, CN, USA; Stromme & Hammel 1968). One channel was used for reference air; each of the others could supply continuous flow to a metabolism chamber, and we usually measured three birds per night. Chambers were constructed from opaque airtight plastic and equipped with thermistors to measure ambient temperature (Ta). We matched chamber volume (1, 2, or 4 L) and airflow (200–1200 mL min−1 STPD) to the size of tested birds, with larger chambers and higher flows used for larger species. Excurrent air was routed through a multiplexer (Sable Systems RM-8) that selected one flow at a time for analysis. The selected flow was subsampled (50–100 mL min−1) at constant pressure, scrubbed of CO2 and water vapour (soda lime and silica gel) and analysed for O2 content with a Sable Systems FoxBox. Flow rate, Ta and O2 content were recorded every 1 s (Sable Systems ExpeData or Warthog LabHelper; www.warthog.ucr.edu). The system monitored each bird for 15 min, switched to reference air for 2·5 min, then switched to the next bird in sequence. Accordingly, the fraction of time each bird's metabolism was monitored was 29% if three birds were being measured, 43% if two birds were measured, and 86% for single birds. Airflows were sufficient to keep O2 concentrations above 20·4%; subsampled air attained steady-state values within 1–2 min after switching from reference readings to chamber air. Chamber Ta was maintained ±0·5 °C with PELT-5 controllers and PTC-1 cabinets (Sable Systems). Data were baseline-corrected and converted to 
(mL O2 min−1) with LabAnalyst (www.warthog.ucr.edu) using the ‘Mode 1’ formula:
is flow rate (mL min−1 STPD), FiO2 is incurrent fractional oxygen concentration (0·2095), and FeO2 is excurrent fractional oxygen concentration. BMR was calculated as the lowest continuous average 
over at least 2 min (usually 5 min) during periods when 
was low and stable and birds had been at constant Ta for at least 1 h. We rejected data from birds that did not become quiescent or had Tb < 35·0 °C. For birds tested at both 30 and at 32–34 °C, we used the lowest of the two metabolic rates as BMR. In a few cases, the lowest 
occurred at 20 °C and we used that value as BMR. We converted 
to watts using a coefficient of 19·8 Joules mL−1 O2 (Gessaman & Nagy 1988).
Procedures for UCR data were similar, except for the use of a different oxygen analyzer (S-3A; Applied Electrochemistry, Sunnyvale, CA, USA) and flow controllers (Applied Materials or Tylan). Only one bird was tested per night, with baseline readings every 42–45 min.
Analyses and Statistics
We log10-transformed species means of body mass and BMR to account for nonlinear metabolic scaling. Several Peruvian species were captured at two stations; none were captured at three. For species with multiple captures at one station and a single capture at another (N = 7), the single value was discarded. For species with multiple captures at two stations (four tropical, one temperate migrant), we computed means for each station; no species showed between-station differences in BMR (P > 0·10). Because many species were represented by single individuals, we tested for sample size effects by performing duplicate analyses, one including all species and a second that excluded species with single measurements. Results of duplicate analyses were qualitatively identical and quantitatively similar. Accordingly, we present results for the entire data set only.
To compare between two Neotropical sites (Peru and Panama), and between tropical and temperate-breeding birds, we combined our measurements with data from three other studies (McKechnie & Wolf 2004; White et al. 2007; Wiersma et al. 2007). McKechnie & Wolf (2004) did not specify latitude and White et al. (2007) did not specify breeding habitats, so we categorized species from these studies from published distribution data. Multiple measurements were available for several species. Our conventional analyses included all independently-measured BMR. For phylogenetic statistics, species were represented once and we handled replicate measurements as follows. For species with more than one literature value, we used the first-published datum. For Peruvian species also represented in the literature, we used our data (in all such cases we measured more individuals). The pooled data (534 measurements of 176 temperate and 350 tropical species) include several species that breed in both temperate and tropical regions, and some tropical species inhabiting climatically variable regions such as deserts or seasonal grasslands. To emphasize possible contrasts in BMR between cool, variable temperate climates and warm, stable tropical climates, we analysed a subset of 496 measurements of 488 species that, to our knowledge, breed exclusively in either temperate regions or humid tropical forests.
Because there is uncertainty over differences in BMR among avian lineages (Lasiewski & Dawson 1967; Reynolds & Lee 1996; McKechnie & Wolf 2004), we used our pooled data set to test for differences in BMR among major avian groups.
We compared BMR with ordinary least squares analysis of covariance (ols ancova) where altitude (Peru species only) and breeding latitude were fixed factors and body mass was the covariate. To account for potential sampling bias or methodological differences, we also included the data source. These analyses were performed with jmp 10 (SAS Institute, Cary, NC, USA).
Equivalent tests were performed within a phylogenetic context using a phylogeny derived from the avian ‘tree of life’ (Jetz et al. 2012), using Hackett et al. (2008) as a backbone. We downloaded 5000 randomly-selected topologies that included all species in our data set (http://birdtree.org/; Hackett backbone, ‘genetic data only’ trees), and computed a maximum clade credibility consensus (mcc) tree (treeannotator v. 1.8.0; Drummond & Rambaut 2007). Subsequent calculations were performed using the mcc tree as a point estimate, and the entire sample of 5000 trees was used to evaluate how accurately the consensus tree represented the entire distribution (Tables S2, S3).
Forty-four species lacked molecular data. For tree construction, we substituted molecular data from a congener for 37 of these (Table S4); the rest were excluded from phylogenetic analyses. For Peruvian species with data from two altitudes, we removed data from the elevation with the lowest sample size for phylogenetic analysis (Table S3). Tree structures and raw data are available in http://datadryad.org.
Phylogenetic tests were performed in r (R Core Team 2013). To justify use of phylogenetically informed regressions, we computed Pagel's λ (Pagel 1999) to estimate phylogenetic signal in BMR, body mass and residuals of regressions between BMR and mass (‘phytools’ package; Revell 2010, 2012). We used phylogenetic independent contrasts (Felsenstein 1985; Garland et al. 1992) to estimate correlations between BMR and body mass (‘ape’ package; Paradis, Claude & Strimmer 2004). We performed phylogenetic generalized least squares (pgls) to examine relationships of BMR with body mass, altitude, breeding latitude and data set (‘nlme’ package; Pinheiro et al. 2013). For pgls, we compared five models of evolution (Table S5): Brownian Motion, Grafen (1989), Pagel (1999), Accelerated/Decelerated (Blomberg, Garland & Ives 2003), and Ornstein-Uhlenbeck (Martins & Hansen 1997). According to akaike information criterion tests (‘qpcR’ package; Ritz & Spiess 2008), Pagel's model best fit our data, indicating that the traits in question are best described as having evolved in a directional manner in which patterns of similarity reflect shared evolutionary histories (Pagel 1999). We used this model for all pgls analyses (Table S5).
Results
We measured BMR in 253 species of tropical breeders from Peru, and three species of temperate North American breeders wintering in Peru (Table S4). The body mass (Mb) range was 2·77–322·4 g; mass distributions were similar at the three stations (Fig. 1). The mean number of individuals sampled per species was 5·1 ± 6·7 (range = 1–41). The UCR data (Table S4) included BMR from 185 individuals of 27 species of tropical- and temperate-breeding birds (Mb range 13·8–1140 g).
In all comparisons, strong phylogenetic signal was present in Mb, BMR and the regression residuals (λ = 0·73–1·00, P < 0·0001).
Altitude
Ordinary least squares analysis of covariance (ols ancova; N = 253 species, three of which had data from two altitudes) showed no significant effect of altitude on BMR (F2,250 = 1·80, P = 0·17). BMR was highly correlated with Mb, but there was an altitude × Mb interaction (F2,250 = 7·0, P = 0·0011): highland natives had a larger mass exponent than low- and mid-altitude species, but had lower BMR at small Mb (Fig. 2). The overall relationship between Mb and BMR (watts) was BMR = 0·056 × 
(r2 = 0·80, t = 32·3, P < 0·0001).
The data set for phylogenetic analyses was reduced to 227 species, but pgls results were similar: BMR was not affected by altitude (F2,221 = 0·02, P = 0·98), but there was a significant Mb × altitude interaction (F2,221 = 5·07, P = 0·007). BMR was highly correlated with Mb, and the overall equation for BMR (watts) was BMR = 0·052 × 
(r2 = 0·74; P < 0·0001).
Peru vs. Panama
ols ancova revealed no difference in BMR between 64 Panamanian lowland forest species reported by Wiersma et al. (2007) and 253 Peruvian birds (F1,316 = 0·43, P = 0·56; Fig. 3). Thirteen species occurred in both locations; these were included in the analyses as independent measurements. There was no Mb × data set interaction (P = 0·18). The combined regression was BMR (watts) = 0·055 × 
(r2 = 0·81, P < 0·0001).
pgls analysis (278 species) also found no difference between Peruvian and Panamanian species (F1,274 = 1·81, P = 0·18), a strong effect of Mb on BMR (F1,274 = 629, P < 0·0001), and a marginally significant Mb × data set interaction (P = 0·052). The combined equation was BMR (watts) = 0·050 × 
(r2 = 0·75; P < 0·0001).
Tropical vs. Temperate Species
ols ancova (534 measurements of 350 tropical and 176 temperate species) showed a significant effect of breeding latitude, with the BMR of temperate breeders averaging 14·1% higher than that of tropical species (F1,533 = 16·0, P < 0·0001). However, there was an Mb × breeding latitude interaction (F1,533 = 28·7, P < 0·0001) and data set was a significant factor (F4,526 = 12·9, P < 0·0001), with BMR in the McKechnie & Wolf (2004) and UCR data sets averaging lower than the other data sets after correction for Mb and latitude. BMR scaled in proportion to 
(overall r2 = 0·92, P < 0·0001).
Exclusion of species with ambiguous distributions reduced sample size to 496 measurements of 488 species (331 tropical, 157 temperate) and removed the ostrich (Struthio camelus), by far the largest species. Nevertheless, results were similar. There was an effect of breeding latitude, with BMR of temperate breeders averaging 16·4% higher than that of tropical species (F1,495 = 18·6, P < 0·0001), and an Mb × breeding latitude interaction (F1,495 = 32·1, P < 0·0001). Data set was a significant factor (F4,488 = 8·8, P < 0·0001), with lower BMR in the McKechnie & Wolf (2004) and UCR data sets. Overall, BMR scaled in proportion to 
(r2 = 0·92, P < 0·0001).
Results from phylogenetic analyses were broadly similar: there was an effect of breeding latitude on BMR (F1,477 = 15·8, P = 0·0001) and interactions between Mb and both breeding latitude (F1,477 = 20·8, P < 0·0001) and data set (F4,477 = 6·06, P = 0·0001). pgls indicated that BMR was strongly affected by Mb (F1,477 = 2176, P < 0·0001), but not by data set (F4,477 = 1·01, P = 0·41). For tropical species, BMR (watts) = 0·031 × 
(r2 = 0·797, P < 0·0001); for temperate-breeding species, BMR = 0·023 × 
(r2 = 0·875, P < 0·0001).
pgls results were qualitatively identical when species with ambiguous distributions were excluded. There was an effect of breeding latitude on BMR (F1,457 = 25·5, P < 0·0001), no effect of data set (F4,457 = 1·91, P = 0·11), and interactions between Mb and both breeding latitude (F1,457 = 24·8, P < 0·0001) and data set (F4,457 = 5·20, P = 0·0004). For tropical species, BMR (watts) = 0·044 × 
(r2 = 0·745 P < 0·0001); for temperate-breeding species, BMR (watts) = 0·023 × 
(r2 = 0·867 P < 0·0001).
Comparisons Among Orders
In contrast with recent reports (McKechnie & Wolf 2004; White et al. 2007), phylogenetic analyses indicated that BMRs of passerine birds are higher than those of non-passerines (F1,485 = 54·6, P < 0·0001), but with an order × Mb interaction (F1,485 = 5·9, P = 0·02). Passerine BMR (watts) = 0·0448 × 
(r2 = 0·719; P < 0·0001); for non-passerines, BMR (watts) = 0·025 × 
(r2 = 0·88; P < 0·0001). ols results were similar, except that there was no order × Mb interaction (P = 0·84): BMR scaled to a × 
, where the mass coefficient a = 0·0401 for passerines and 0·0378 for non-passerines (i.e. passerine BMR averaged 12·2% higher than that of non-passerines; F1,533 = 15·6, P < 0·0001).
Because hummingbirds may have different metabolic constraints than other non-passerine orders, they could have a disproportionate influence on contrasts between passerines and non-passerines. Our data contained 34 Apodiformes (32 hummingbirds and two swifts; 2·8–44·9 g). Accordingly, we tested for differences in BMR between Apodiformes, Passeriformes and the pooled data from other non-passerine orders. pgls revealed differences among these groups (F2,483 = 70·0, P < 0·0001), and an order × Mb interaction (F2,483 = 6·1, P = 0·003). BMR differed between Apodiformes and Passeriformes (P < 0·0001), Apodiformes and non-passerines (P < 0·0001), and non-passerines and Passeriformes (P < 0·0001). For Passeriformes, BMR (watts) = 0·045 × 
(r2 = 0·719; P < 0·0001); for Apodiformes, BMR (watts) = 0·067 × 
(r2 = 0·554; P = 0·0033); for other non-passerine orders, BMR (watts) = 0·021 × 
(r2 = 0·89; P < 0·0001).
Discussion
Our latitudinal comparison supports the ‘slow pace of life’ concept for tropical species (Ricklefs & Wikelski 2002), as we found lower BMR in tropical-breeding birds compared to temperate-breeding birds. Contrary to predictions based on thermoregulatory costs, the BMR of Peruvian forest birds did not differ across a 2·6 km altitude gradient. We found no BMR differences between Neotropical birds from Peru and Panama (Wiersma et al. 2007), even though the sites are separated by 2500 km and are in opposite hemispheres. Together, these findings indicate consistently ‘slow’ energy metabolism in tropical birds across broad geographical regions and a wide range of altitudes (and therefore temperatures).
Two caveats must be kept in mind when considering the biological significance of these findings. First, BMR is easily and frequently measured, but it includes only a fraction of an endotherm's total energy expenditures; moreover, it probably has minimal ecological importance as it encompasses none of the costs of growth, activity, thermoregulation, reproduction or other critical life-history traits. Nevertheless, BMR is often used as a proxy for metabolic intensity. Several studies found positive correlations between BMR and maximal metabolism in exercise or thermogenesis (Dutenhoffer & Swanson 1996; Chappell, Bech & Buttemer 1999; Wiersma, Chappell & Williams 2007) although others have not (Hammond et al. 2000). Daily energy expenditures or field metabolic rates (FMR) – the most direct measure of the metabolic ‘pace of life’– often covary with BMR or resting metabolism, albeit with considerable variance (Daan et al. 1991; Ricklefs, Konarzewski & Daan 1996; White & Seymour 2004). The association of avian FMR with latitude is not clear: FMR does tend to increase at high latitudes as predicted by the ‘pace of life’ concept, but variation is large, particularly in the tropics (Anderson & Jetz 2005). That may be due in part to the paucity of FMR from tropical birds (<15 species) and particularly of FMR from species inhabiting humid forests (4–5, Anderson & Jetz 2005). In one comparison, the FMR of breeding house wrens (Troglodytes aedon) was lower in tropical Panama than in temperate Ohio (Tieleman et al. 2006).
The second caveat is that like most authors of comparative studies of energy metabolism, we interpret between-group contrasts as heritable products of evolutionary change. However, we cannot be sure that differences are not manifestations of phenotypic plasticity in response to dissimilar environmental conditions, which are well known in birds (Piersma et al. 1996; Kvist & Lindstrom 2001; McKechnie 2008; McKechnie & Swanson 2010). Rigorous demonstration of a genetic basis of between-group variation requires a level of experimental control (e.g. ‘common garden’ breeding) beyond the scope of most large comparative studies, including this one.
Those caveats aside, why, from a mechanistic perspective, might BMR be low in tropical birds? Some authors (Wikelski et al. 2003; Wiersma, Chappell & Williams 2007; Wiersma et al. 2007) interpret low metabolic rates of tropical birds as by-products of their ‘slow’ life histories. Tropical environments with intense competition and predation should favour reduced fecundity. As the maximal sustained metabolic rates in birds (other than migration flights) are assumed to occur during breeding as a consequence of high requirements for parental care (Drent & Daan 1980), small, slow-growing broods should reduce parental workloads. Because BMR derives mainly from the metabolism of visceral organs that are ‘scaled’ to maximal power production (Daan, Masman & Groenewold 1990; Piersma et al. 1996; Chappell, Bech & Buttemer 1999), reduced peak energy demands should permit reductions in organ size and hence BMR. This hypothesis is consistent with a recent study showing that low BMR in tropical birds from Panama is accompanied by reduced organ and skeletal muscle masses (Wiersma, Nowak & Williams 2012).
The findings of Wiersma, Chappell & Williams (2007) and Wiersma et al. (2007) support the ‘slow life history’ concept, but the geographically limited set of tropical species, mostly from a single habitat (lowland rain forest in Panama) does not rule out other explanations. For example, low BMR could be due to residence in a warm, stable climate that does not require high activity or thermoregulatory capacity (Anderson & Jetz 2005; Wiersma, Nowak & Williams 2012). Also, many high latitude breeders migrate over long distances (Jetz, Freckleton & McKechnie 2008), which requires massive energy turnover both during migration and in preparation for it (e.g. Kersten & Piersma 1987). In this alternate scenario, reduced demands for aerobic power output (thermogenesis, long-distance flight) in benign tropical habitats – not ‘slow’ life history – has led to lower capacities of muscle and visceral organs and hence low BMR. This model does not require or derive from differences in reproductive investment and related life-history traits and is consistent with two analyses showing negative associations between avian BMR and environmental temperature (White et al. 2007; Jetz, Freckleton & McKechnie 2008). It is also analogous to shifts in energy metabolism in cold acclimatization, which elicits elevated BMR as well as thermogenic capacity (Cooper & Swanson 1994; Liknes, Scott & Swanson 2002; McKechnie & Swanson 2010).
Peruvian birds provide a test of these two hypotheses. All our field sites have stable environmental temperatures, but the high-altitude site (Wayqecha) averages 12 °C colder than the lowland site (Pantiacolla; Table S1) and consequently requires higher thermoregulatory costs (M.A. Chappell and G. A. Londoño, unpublished data). Other factors being equal, daily energy requirements (DEE) should be concomitantly greater at high altitude, and if energy flux or cold acclimatization is primary ‘drivers’ of BMR, then BMR at Wayqecha should be higher than at Pantiacolla. An across-latitude analysis (Jetz, Freckleton & McKechnie 2008) found that a 20 °C decrease in temperature was associated with a 50% increase in avian BMR. If that relationship applies to Peruvian birds, the BMRs of highland species should be 30% greater than in lowland species. We found no difference in BMR across altitudes, which argues against the idea that warm, stable temperatures are responsible for low BMR in tropical birds. However, costs of thermoregulation are only one component of DEE, and by adjusting other aspects of the energy budget (e.g. reduced flight activity), highland birds could achieve DEE similar to that of lowland species.
White et al. (2007) suggested that low BMR is favoured by selection in tropical lowlands because it would reduce heat loads and risks of hyperthermia. Our Peru BMR is not consistent with that hypothesis: no species – even those from the cool highland site – showed indications of heat stress when tested at 33–34 °C, which approximates the hottest shade temperatures at Pantiacolla. Sunlight can increase heat loads, but the lowland species in our data set do not use open habitat, so prolonged unavoidable sun exposure is unlikely. Low BMR may be beneficial in more challenging habitats, such as hot deserts or open grasslands (Weathers 1979).
One other finding merits comment. Early analyses of avian BMR reported higher BMR in passerines than in non-passerines (e.g. Lasiewski & Dawson 1967), but recent studies using phylogenetically informed statistics generally found no difference (e.g. Reynolds & Lee 1996; McKechnie & Wolf 2004). Both ols and pgls analysis of our data set revealed higher BMR in passerines than in non-passerines, although by only 12% compared to the 50–60% difference reported by Lasiewski & Dawson. Additional clade comparisons were hampered by small sample sizes in most orders, but BMR differed between the best-represented non-passerine order (Apodiformes) and both passerines and other non-passerines. Larger sample size and a more recent and robust phylogenetic tree (Hackett et al. 2008) may explain why we found differences while earlier phylogenetic analyses did not. Interestingly, non-passerines tend to form a larger proportion of bird communities in tropical habitats than in temperate regions and also tend to be better represented on islands than on the mainland (Faaborg 1977). Perhaps their lower metabolic rates give them an advantage on islands, which are often relatively free of predators, and in the tropics where a slow life history seems to be favoured by selection.
Our data expand upon previous reports of low BMR in tropical birds. We found no effect of environmental temperature on BMR in Peruvian forest birds, consistent with the view that low BMR in tropical birds is mainly ‘driven’ by slow life history. Nevertheless, we urge due caution in interpreting these findings. Considerable variation in avian life history (clutch and egg size, incubation and nestling periods, mating systems) occurs in both temperate and tropical regions (Skutch 1985; Martin 1996; Robinson et al. 2010; G.A. Londoño, unpublished data). This spectrum of life histories probably explains some of the substantial BMR variation remaining after correction for mass, both within regions (Figs 2, 3) and across broad geographical contrasts (Fig. 4). Therefore, in addition to inferring physiological consequences of life history from large-scale comparisons as we have done here, a promising approach for testing links between life history, physiology and ecology might be to examine these variables in detail at a species-by-species level (e.g. Tieleman et al. 2006). We also note that daily energy consumption, such as is measured with the doubly-labelled water technique, is a more comprehensive and relevant index of metabolic intensity than BMR, and its use should be encouraged in future studies of relationships between life history and energy metabolism.
Acknowledgments
Over 100 field and laboratory assistants were essential in obtaining the data in this paper. We especially thank Sherrie Souza and Kyle van Dolah (UCR data); Andres Chinome, Laura Gómez, Santiago David-Rivera, Jenny Muñoz, Camilo Flórez, David Ocampo, Diego Rincon, Jaime Garizabal, Brandon McLaughlin, Callum Kingwell, Tim Forrester, Rachel Grey, Katherine Fielding, Kristie Hazelwood, Valerie Dean, Jessica Thompson, Nik Aspey, Catherine Klein, Wendy Vidal, Juliana Soto, Gliselle Marin, Mario Loaiza, Stephanie Levins, Colin Fagan, Miranda Ciotti, Justin Demianew, Nancy Chen, Justin Baldwin, Juliana Ceron, Dano Grayson, Thomas Martin, Juan Diego Castaño, Pauline Pearse, Torin Heavyside, Meghan Conway, Simon Nockold and Naman Goyal (Peru data). Liam Revell and Juan Pablo Gomez helped with phylogenetic analyses. The assistance of Marianne van Vlaardingen and the Pantiacolla Lodge, Daniel Blanco and the Cock-of-the-Rock Lodge, and the Wayqecha Cloud Forest Biological Station was invaluable. We thank SERNAP for permission to work in the Manu Park buffer zone. Joe Williams and another reviewer provided valuable feedback. Funding was provided by National Science Foundation Grant DEB-1120682 (to S.K. Robinson and M.A. Chappell), the University of California, Riverside Academic Senate, and the Florida Museum of Natural History.
Data accessibility
BMR data for this paper are deposited Dryad Digital Repository: http://doi:10.5061/dryad.vg313 (Londoño et al. 2015).
