Avian thermoregulation in the heat: resting metabolism, evaporative cooling and heat tolerance in Sonoran Desert doves and quail

Birds in hot environments face significant physiological challenges when environmental temperatures exceed body temperature, a situation where evaporation is the only avenue for heat dissipation (Calder and King, 1974). Increasing air temperatures drive increases in avian body temperature, metabolism and evaporative water loss in an effort to offset increasing heat loads, but the costs incurred can affect survival and fitness (McKechnie and Wolf, 2010; Cunningham et al., 2013a,b). Warmer and more frequent hot days over land are virtually certain by the late 21st century (IPCC, 2013) and will be accompanied by hotter, longer and more frequent heat waves (Meehl and Tebaldi, 2004; IPCC, 2011; Rahmstorf and Coumou, 2011). These rapid increases in environmental temperatures over the coming century will increasingly tax the abilities of animals to thermoregulate effectively and may dramatically influence community structure and the distribution of animals (McKechnie et al., 2012; Diffenbaugh and Field, 2013; Quintero and Wiens, 2013).

The high solar heat loads and high air temperatures of the world's deserts already routinely expose birds to environmental or operative environmental temperatures (T_e) that greatly exceed avian body temperature (Bakken, 1976; Robinson et al., 1976). Gambel's quail (Callipepla gambelii Gambel 1843) foraging on a typical hot summer day in the Sonoran Desert, for example, experience T_e approaching 50°C (Goldstein, 1984). Desert-nesting doves often place their nests in sites with high solar heat loads, where the nest environment can reach T_e of 50–60°C, which demands that incubating birds cool their eggs to maintain viable egg temperatures (B.O.W., personal observation; Russell, 1969; Walsberg and Voss-Roberts, 1983; Marder and Gavrieli-Levin, 1986).

Despite decades of research on thermoregulation in birds, our information on the thermoregulatory performance of wild birds exposed to air temperatures above body temperature is limited. Early research on avian thermoregulation in the heat focused on passerines, which evaporate water from their respiratory surfaces by panting (Dawson and Bartholomew, 1968). Heat loss through panting is an active process, requiring muscular movement of the respiratory apparatus, a metabolically costly effort that produces heat. Rates of water loss through panting are very high and in small passerines may exceed 5% of body mass per hour (Wolf and Walsberg, 1996). A lack of sweat glands and the thick plumage were thought to impede significant evaporation from the skin in birds (Calder and Schmidt-Nielsen, 1967). Smith and Suthers (1969) showed that pigeons and doves dissipate large heat loads by cutaneous evaporation with little increase in metabolism, allowing rock doves (Columba livia) to raise young at air temperatures as high as 60°C (Marder and Arieli, 1988). The strong reliance on cutaneous evaporation in the heat is apparently widespread in columbiform birds, where it has been observed in two of this study's species – mourning doves (Zenaida macroura Linnaeus 1758) and white-winged doves (Zenaida asiatica Linnaeus 1758) (Hoffman and Walsberg, 1999; McKechnie and Wolf, 2004) – but not in galliform birds (Marder, 1983; Marder and Ben-Asher, 1983). Other mechanisms that enhance evaporative heat loss such as gular flutter, deep esophageal pulsation and cloacal evaporation remain poorly understood (Calder and King, 1974; Gaunt, 1980; Dawson, 1982; Baumel et al., 1983; Schleucher et al., 1991; Hoffman et al., 2007).

Among the modest number of studies that have examined heat tolerance of birds, differences in methodologies make it difficult to draw broad conclusions. These studies often relied on birds raised or held in captivity for long periods at standard temperatures (e.g. 25–30°C; see Hudson and Brush, 1964; Lasiewski and Seymour, 1972; Weathers, 1981; Hoffman and Walsberg, 1999; Hoffman et al., 2007). Because heat-acclimated birds can tolerate significantly higher temperatures than birds acclimated to lower standard temperatures (Marder and Arieli, 1988; McKechnie and Wolf, 2004), prior heat stress studies may have greatly overestimated the susceptibility of subjects to elevated temperatures.

Here, we present data on the thermoregulatory performance of the Gambel's quail, a galliform, and two species of columbiforms, the mourning dove and white-winged dove, with the aim of quantifying responses to heat stress in birds summer-acclimatized to a hot subtropical desert. We continuously measured body temperature (T_b, °C), whole animal rates of evaporative water loss (EWL, g H₂O h⁻¹) and daytime resting metabolism (RMR, W) in response to chamber air temperature (T_air, ranging from 30°C to 66°C) in birds on the day they were captured. We hypothesized that like rock doves, white-winged doves and mourning doves would demonstrate high tolerances for heat, and nearly constant T_b and RMR with increasing T_air. Because galliforms lack the cutaneous evaporative cooling of columbiforms, we anticipated that Gambel's quail would demonstrate lower tolerances for heat and show more rapid increases in T_b and RMR with increasing T_air. We present these species together because of their ecological similarities as large (>100 g), desert-nesting granivores with overlapping distributions, but make no inferences about adaptive differences. Thus, we examine their relative thermoregulatory performance without having a sufficient number of species to consider their evolutionary histories (Garland et al., 1992, 1993, 2005; Blomberg et al., 2003).

Birds were captured in June and July of 2012 (white-winged and mourning doves) and in July of 2013 (Gambel's quail) in northwest Tucson, Arizona, USA. Experiments were conducted on the same day as capture. During June and July of 2012, temperatures at the Tucson AZMET weather station ranged from 14 to 42°C, daily maximum temperatures ranged from 31 to 42°C and the dew point ranged from −15 to 22°C. During July of 2013, temperatures ranged from 21 to 41°C, daily maximum temperatures ranged from 32 to 41°C, and the dew point ranged from 7 to 23°C.

The birds were captured using standard walk-in traps baited with seed and held outdoors in shaded, darkened screen cages [model 1450B, 12″ (30.5 cm) cubed, BioQuip, Rancho Dominguez, CA, USA]. None of the birds received supplemental food after capture. However, they probably continued digesting seed in their crops and were not post-absorptive; thus, we assumed for metabolic measurements an RER of 1 and 20.9 J ml⁻¹ CO₂ produced (Gessaman and Nagy, 1988; Walsberg and Wolf, 1995). Free water was provided to the quail but they were not observed to drink. Thus, prior to each experimental run, the quail were provided between 3 and 6 ml of tap water via oral gavage. Five white-winged doves were also given additional water by gavage prior to high-temperature runs and no differences in thermal tolerance were observed compared to un-watered individuals. Prior to each experimental run, a temperature-sensitive PIT (passive integrated transponder) tag (model TX1411BBT, Biomark, Boise, ID, USA) was injected into each bird's abdominal cavity. Use and accuracy of these tags for T_b measurement is detailed in Whitfield et al. (2015). Following experimental runs, where birds were exposed to high temperatures, birds were placed in front of cool air flow and T_b was monitored until it returned to resting levels (∼40°C); Gambel's quail were also gavaged with tap water post-exposure to ensure rehydration. Birds were monitored for 1–5 h after exposure, after which the doves were released at the site of the experiments and quail from their site of capture. Animal care protocols were approved by the Institutional Animal Care and Use Committee of the University of New Mexico (protocol no. 12-100537-MCC). Birds were captured under permits from the US Fish and Wildlife Service and the Arizona Game and Fish Department.

Measurements of RMR and EWL were made using a flow-through respirometry system similar to Whitfield et al. (2015). The respirometry chamber consisted of a transparent plastic container (5 litre, 22 cm×25 cm×12 cm, Rubbermaid, Atlanta, GA, USA) sealed with an opaque lid, modified by addition of inlet and outlet ports, and containing a plastic mesh platform above a 2 cm layer of medium weight mineral oil to trap excreta. The chamber was placed inside a modified ice chest where temperature was controlled to ±0.5°C. Mass-flow controllers (Alicat, Tucson, AZ, USA) provided dry air to the chamber from a pump through columns of silica gel and drierite connected in series (2012) or from a compressor through a membrane dryer (2013: Champion^® CMD3 air dryer and filter, Champion Pneumatic, Quincy, IL, USA). Excurrent air was sampled at 250 ml min⁻¹ and directed to a CO₂/H₂O gas analyzer (model LI-840A, LICOR, Lincoln, NE, USA). Gas analyzer outputs were sampled once per second by Expedata (version 1.4.15, Sable Systems, Las Vegas, NV, USA).

Following PIT tag insertion, a bird was weighed (model V31XH2, accuracy=0.1 g, Ohaus, Parsippany, NJ, USA) and placed in the darkened respirometry chamber, where an infrared light and video camera enabled continuous viewing. The bird was exposed to one or more thermoneutral temperatures (30 or 35°C) and one or more higher temperatures (40–66°C in 2°C increments) in a stepped pattern of temperature trials over the span of the 1–3 h experimental run. In order to keep H₂O content of the respirometry chamber at values that would not impede evaporation (dew point <5°C), flow rates were maintained between 5 and 40 litre min⁻¹ depending on T_air and the bird's evaporative rate. The initial thermoneutral temperature trial allowed a bird to calm from handling; H₂O and CO₂ production were monitored and observed to drop to resting levels (typically within 30 min). T_air was then increased to higher trial temperatures and birds were allowed to adjust to each temperature for 10–20 min and stabilize before moving to the next temperature. Most birds remained calmer when flow rates were higher and chamber humidity was lower (<5 ppt water vapor).

As T_air was increased above thermoneutrality, T_b and activity were monitored closely to prevent mortality. A trial was ended and the bird recorded as reaching its thermal limit if it: (1) remained continuously active for 5–10 min, (2) showed a T_b approaching or exceeding 45°C, or (3) showed a loss of balance or righting response (extremely rare). In some trials, birds were exposed to T_air levels that were at or near their thermal limits, but these measurements were not designed to elicit mortality. Unlike critical thermal maxima trials where mortality is common (reviewed in Lutterschmidt and Hutchison, 1997), our procedure resulted in rates of mortality of less than 1%. In addition, the experimental runs were timed to end before a bird had lost more than 10% of M_b during a trial (including any fecal losses). Approximately 5 min of baseline data were collected following each temperature trial.

We selected the lowest 1 min (doves) or 5 min (quail) average of CO₂ ppm readings less baseline values for each trial temperature. Birds noted as active or failing or having mean T_b>44.7°C or slope of T_b>0.1°C min⁻¹ during a temperature trial were not included in the analysis. Using eqn 10.5 of Lighton (2008), we estimated the rate of CO₂ production in ml CO₂ min⁻¹ and whole-animal RMR in watts (W), assuming an RER of 1 (see Animal capture and handling). Rates of whole-animal EWL (g H₂O h⁻¹) were calculated from the H₂O ppt readings (less baseline values) for the same data points using eqn 10.9 of Lighton (2008).

Statistical analyses and graphs were produced using R (v3.1; R Development Core Team, 2011) inside RStudio (v0.98.932). We used the linear mixed effects model from the nlme package and individual as a random factor (Pinheiro et al., 2014) due to the repeated measurement of individuals in an unbalanced design. We used the segmented package (Muggeo, 2008) to determine inflection points in the metabolic data. The main effects examined in all analysis were T_air and species. M_b was used as a covariate in all analysis. Backwards-stepwise model selection was used where the initial model included all covariates, random factors and main effects, including two-way interactions between main effects.

Initial body mass (M_b) averaged 160.7±11.1 g (mean±s.d.; N=19) for Gambel's quail, 104.0±10.2 g (N=49) for mourning doves and 147.3±17.7 g (N=52) for white-winged doves (Table 1). M_b (P<0.001), species (P=0.002) and T_air (P<0.001) all significantly predicted RMR and EWL in the species tested.

A significant relationship between RMR and body mass was found among species (F_1,107=52.1, P<0.001). Thus, mass-independent RMR residuals were calculated to investigate the mass-independent responses to air temperature among species (Packard and Boardman, 1999). Segmented regressions of these RMR residuals against air temperature allowed us to identify inflection points of 41.1°C in Gambel's quail, 45.9°C in mourning doves and 46.5°C in white-winged doves (Fig. S1). These inflection points represent the upper critical temperatures (T_uc) above which daytime RMR increases sharply (Kendeigh, 1969; Table 1). The 95% CI of these T_uc values overlapped considerably between the two dove species, did not overlap between Gambel's quail and mourning doves, and slightly overlapped between Gambel's quail and white-winged doves – the upper bound of the former being 43.1°C and the lower bound of the latter being 42.8°C (Table 1). A z-score comparing the mean T_uc values of Gambel's quail and white-winged doves nonetheless revealed their significant difference (z=2.55; P<0.01). Thus, while T_uc values did not differ among dove species, T_uc values for Gambel's quail were significantly lower than those for both dove species.

For T_air between 30°C and the T_uc, RMR decreased as T_air increased (F_1,44=24.3, P<0.001), but this response differed among species (F_1,88=21.9, P<0.001) and was significantly influenced by M_b (F_1,88=30.8, P<0.001). Above the T_uc, RMR increased as T_air increased (F_1,65=43.4, P<0.001); this response also differed among species (F_1,92=11.3, P=0.001) and was significantly influenced by M_b (F_1,92=48.1, P<0.001). Thus, the effects of T_air and M_b were assessed for each species independently.

For Gambel's quail exposed to T_air below their T_uc of 41.1°C, RMR decreased with increasing T_air (F_1,5=15.5, P<0.05; Fig. 1) but was unaffected by M_b (F_1,15=0.04, P=0.85). RMR increased with T_air above the T_uc (F_1,11=24.6, P<0.001; Fig. 1) with a slope of 0.022 W °C⁻¹ (Table 2), and M_b was a significant covariate (F_1,16=8.03, P<0.05). RMR was lowest at the T_uc and averaged 0.71 W or 4.55 mW g⁻¹ (Table 1) and RMR averaged 0.75 W at T_air of 42°C and 0.79 W at 48°C (Table 3).

For mourning doves at T_air below their T_uc of 45.9°C, RMR decreased with increasing T_air (F_1,16=10.1, P<0.01; Fig. 1) and M_b was a significant covariate (F_1,40=8.42, P<0.01). RMR increased with T_air above the T_uc (F_1,27=22.3, P<0.001; Fig. 1) with a slope of 0.017 W °C⁻¹ (Table 2) and M_b was a significant covariate (F_1,37=9.66, P<0.01). RMR averaged 0.64 W at T_air of 42°C, 0.66 W or 6.07 mW g⁻¹ at the T_uc and 0.58 W at 48°C (Tables 1 and 3).

For white-winged doves at T_air below their T_uc of 46.5°C, RMR decreased with increasing T_air (F_1,21=7.18, P<0.05; Fig. 1) and M_b was a significant covariate (F_1,30=18.8, P<0.001). RMR increased with T_air above the T_uc (F_1,25=12.3, P<0.01; Fig. 1) with a slope of 0.019 W °C⁻¹ (Table 2) and M_b was a significant covariate (F_1,36=5.73, P<0.05). RMR averaged 0.99 W at T_air of 42°C, 0.94 W or 6.49 mW g⁻¹ at the T_uc, and 0.97 W at 48°C (Tables 1 and 3).

For T_air below the T_uc, there was a significant effect of T_air (F_1,43=197.9, P<0.001), species (F_1,88=28.0, P<0.001), and M_b (F_1,88=32.2, P<0.001) – as well as a significant interaction of T_air and species (F_1,43=22.2, P<0.001) – on EWL (g H₂O h⁻¹) indicating differences in the slope of this relationship among species. For T_air above the T_uc, EWL was significantly affected by T_air (F_1,64=475.6, P<0.001), species (F_1,92=28.8, P<0.001), M_b (F_1,92=67.9, P<0.001) and an interaction of T_air and species (F_1,64=30.8, P<0.001). Thus, there was a significant difference in the response of EWL to T_air among species both above and below the respective T_uc.

For Gambel's quail at T_air below their T_uc of 41.1°C, EWL increased with T_air (F_1,5=20.1, P<0.01; Fig. 2) with a slope of 0.035 g H₂O h⁻¹ °C⁻¹ (Table 2) but was unaffected by M_b (F_1,15=0.00, P=0.98). EWL increased more steeply with T_air above the T_uc (F_1,11=176.3, P<0.001; 0.196 g H₂O h⁻¹ °C⁻¹; Fig. 2; Table 2). EWL averaged 0.67 g H₂O h⁻¹ at T_air of 35°C, 1.71 g H₂O h⁻¹ at 42°C and 2.46 g H₂O h⁻¹ at 48°C (Table 3).

For mourning doves at T_air below their T_uc of 45.9°C, EWL increased with T_air (F_1,16=63.6, P<0.001; Fig. 2) with a slope of 0.069 g H₂O h⁻¹ °C⁻¹ (Table 2) and M_b was a significant covariate (F_1,40=14.2, P<0.001). EWL increased more steeply with T_air above the T_uc (F_1,27=286.2, P<0.001; 0.210 g H₂O h⁻¹ °C⁻¹; Fig. 2; Table 2) and M_b was a significant covariate (F_1,37=11.6, P<0.01). EWL averaged 0.74 g H₂O h⁻¹ at T_air of 35°C, 1.27 g H₂O h⁻¹ at 42°C and 1.96 g H₂O h⁻¹ at 48°C (Table 3).

For white-winged doves at T_air below their T_uc of 46.5°C, EWL increased with T_air (F_1,21=87.9, P<0.001; Fig. 2) with a slope of 0.140 g H₂O h⁻¹ °C⁻¹ (Table 2) and M_b was a significant covariate (F_1,30=12.9, P<0.01). EWL increased more steeply with T_air above the T_uc (F_1,25=174.5, P<0.001; 0.383 g H₂O h⁻¹ °C⁻¹; Fig. 2; Table 2). EWL averaged 0.99 g H₂O h⁻¹ at T_air of 35°C, 1.95 g H₂O h⁻¹ at 42°C and 3.41 g H₂O h⁻¹ at 48°C (Table 3).

We converted EWL to rates of whole-animal evaporative heat loss (EHL, W) using a latent heat of vaporization of water of 2.26 J (mg H₂O)⁻¹ and define the ratio EHL/MHP – a dimensionless metric for evaporative cooling capacity – as EHL relative to metabolic heat production (MHP=RMR). For T_air below the T_uc, there was a significant effect of T_air (F_1,43=374.7, P<0.001) and species (F_1,89=6.50, P<0.05), as well as a significant interaction of T_air and species (F_1,43=14.9, P<0.001) on EHL/MHP. For T_air above the T_uc, EHL/MHP showed a significant effect of T_air (F_1,64=282.6, P<0.001) and species (F_1,93=6.01, P<0.05) and a significant interaction of T_air and species (F_1,64=10.6, P<0.01). Thus, there was a significant difference in the response to T_air among species both above and below the respective T_uc. There were no significant effects of M_b on EHL/MHP (P>0.05), either above or below the T_uc.

In Gambel's quail EHL/MHP increased with T_air both below and above the T_uc (below: F_1,5=57.4, P<0.001; above: F_1,11=33.9, P<0.001; Fig. 3). EHL/MHP averaged 0.55 at T_air of 35°C, 1.42 at 42°C and 1.97 at 48°C, increasing 3.6-fold between 35 and 48°C (Table 3). At T_air of 50°C, just below the HTL (heat tolerance limit, see below), Gambel's quail reached a maximum EHL/MHP of 2.14 (Table 1).

In mourning doves EHL/MHP increased with T_air both below and above the T_uc (below: F_1,16=123.8, P<0.001; above: F_1,27=159.1, P<0.001; Fig. 3). EHL/MHP averaged 0.65 at T_air of 35°C, 1.26 at 42°C and 2.19 at 48°C, increasing 3.4-fold between 35 and 48°C (Table 3). At T_air of 56°C, just below the HTL, mourning doves reached a maximum EHL/MHP of 3.08 (Table 1).

In white-winged doves EHL/MHP increased with T_air both below and above the T_uc (below: F_1,21=122.9, P<0.001; above: F_1,25=84.5, P<0.001; Fig. 3). EHL/MHP averaged 0.62 at T_air of 35°C, 1.22 at 42°C and 2.27 at 48°C, increasing 3.7-fold between 35 and 48°C (Table 3). At T_air of 58°C, just below the HTL, white-winged doves reached a maximum EHL/MHP of 3.69 (Table 1).

For T_air below 39°C there were no significant effects of M_b, species, T_air or the interaction of T_air and species on T_b (P>0.05) and thus T_b in this range were considered normothermic. For T_air above 39°C, however, T_b increased with T_air differentially among species (T_air: F_1,148=79.2, P<0.001; species: F_1,107=4.9, P<0.05; T_air×species: F_1,148=4.9, P<0.05). Thus, above T_air of 39°C, there was a significant difference in the response of T_b to T_air among species.

T_air below 39°C did not affect T_b in Gambel's quail (F_1,14=0.08, P=0.78) and normothermic T_b averaged 41.1°C (Table 1). T_b increased with T_air above 39°C (F_1,17=67.8, P<0.001; Fig. 4) with a fairly steep slope of 0.170°C °C⁻¹ (Table 2). T_b averaged 41.8 and 42.5°C at T_air of 42 and 48°C, respectively (Table 3) and 43.6°C at T_air of 50°C (Table 1). Thus at T_air of 50°C, just below the HTL, Gambel's quail maintained a T_a–T_b gradient of 6.4°C.

T_air below 39°C did not affect T_b in mourning doves (F_1,9=0.01, P=0.93) and normothermic T_b averaged 41.0°C (Table 1). T_b increased with T_air above 39°C (F_1,68=42, P<0.001; Fig. 4) with a fairly shallow slope of 0.041°C °C⁻¹ (Table 2). T_b averaged 42.0 and 42.1°C at T_air of 42 and 48°C, respectively (Table 3) and 41.9°C at T_air of 56°C (Table 1). Thus at T_air of 56°C, just below the HTL, mourning doves maintained a T_a–T_b gradient of 14.1°C.

T_air below 39°C did not affect T_b in white-winged doves (F_1,9=0.40, P=0.54) and normothermic T_b averaged 41.4°C (Table 1). T_b increased with T_air above 39°C (F_1,62=25, P<0.001; Fig. 4) with a fairly shallow slope of 0.056°C °C⁻¹ (Table 2). T_b averaged 41.9 and 41.7°C at T_air of 42 and 48°C, respectively (Table 3) and 42.7°C at T_air of 58°C (Table 1). Thus, at T_air of 58°C, just below the HTL, white-winged doves maintained a T_a–T_b gradient of 15.3°C.

The results presented thus far pertained to data collected from birds that were inactive and maintaining stable T_b (see Materials and methods). To investigate heat tolerance, we examined the final 5 min of all temperature trials above 39°C with regard to T_b maximum and T_b slope (Fig. 5). The T_b (maximum, slope) for Gambel's quail (45.1°C, 0.27°C min⁻¹) and mourning doves (46.1°C, 0.24°C min⁻¹) resulted from individuals in 52 and 54°C temperature trials, respectively; the values for white-winged doves (46.0°C, 0.39°C min⁻¹) were recorded by different individuals in trials of 54°C and above. Individual birds passed a given trial if T_b and slope of T_b remained less than or equal to 44.7°C and 0.1°C min⁻¹, respectively; individuals that exceeded either of these limits were classified as failing, indicating that the heat load experienced by an individual exceeded its thermoregulatory capacity and it had reached its heat tolerance limit (HTL). Once the majority of individuals failed in a trial at a given T_air, a species was considered to have reached its HTL. Thus, the HTL of Gambel's quail was 52°C (1/9 pass), of mourning doves was 58°C (3/8 pass) and of white-winged doves was 60°C (1/4 pass; Fig. 5; Table 1).

In this study, we examined the thermoregulatory ability of wild doves and quail under conditions that normally persist during the summer in hot subtropical deserts. We maintained low chamber humidities (dew point <5°C) to maximize the evaporative gradient and used summer-acclimatized birds in order to make measurements under ecologically relevant conditions. White-winged and mourning doves demonstrated a heat tolerance and ability to thermoregulate at T_air as high as 60°C, while Gambel's quail did so successfully up to 50°C.

Earlier studies often did not provide conditions conducive for examining natural responses to heat stress. Hudson and Brush (1964), for example, found that mourning doves died at T_air above 41°C and that California quail (Callipepla californica) could not survive T_air above 44°C. Birds exposed to very high environmental temperatures in nature are typically subject to very low atmospheric humidity. High chamber humidity impeded the birds' ability to evaporate water and severely limited their ability to cope with heat stress (see Lasiewski et al., 1966, and Whitfield et al., 2015, for more discussion of these issues).

In addition, most studies of thermoregulation in these taxa have focused on responses to T_air below normothermic T_b. Prior to this study, thermoregulatory performance at T_air above 40°C had only been quantified in 9 of the 300+ species of columbiform birds and 7 of the 170+ species of galliform birds (see Figs 6 and 7). Of these, only 4 columbiform and 3 galliform species were exposed to T_air above 45°C. Below, we discuss the thermoregulatory performance of each study species in detail and compare our observations to earlier relevant studies in galliform and columbiform birds as well as speculate on their future performance in a warmer and dryer environment.

Understanding changes in RMR with changes in T_air is critical to quantifying thermoregulatory capacity in the heat. We found similar slopes in doves and quail for changes in RMR and mass independent RMR at T_air above the T_uc (Table 2, Table S1). However, the T_uc – the T_air at which heat dissipation becomes an active process and marked increases in daytime RMR occur (Kendeigh, 1969) – was significantly lower in the quail (41.1°C) than observed in both dove species (45.9–46.5°C) (Table 1). These differences in the T_uc among doves and quail are importantly related to pathways of evaporative heat dissipation. At a T_air of 30°C, evaporative losses in most birds are more or less evenly distributed between respiratory and cutaneous pathways (Bernstein, 1971; Ro and Williams, 2010). At a T_air≥40°C, the primary evaporative pathway is respiratory in passerines (≥75%, Wolf and Walsberg, 1996) and galliforms (75%, Richards, 1976) and cutaneous in doves (≥70%, McKechnie and Wolf, 2004). Since metabolic heat contributes to total heat load, birds that actively dissipate heat via panting (quail, passerines) tend to have lower limits of heat tolerance and relatively high rates of water loss compared with doves, which dissipate most of their heat load using cutaneous evaporation.

Other studies of columbiforms have reported responses of metabolic rate to temperature similar to those reported in the current study (Fig. 6). In rock doves acclimated to very high T_air (up to 62°C) metabolic rate remained constant from 30 to 60°C and no T_uc is apparent (Marder and Arieli, 1988). White-winged doves acclimated to 43°C showed no significant difference in RMR between 35 and 45°C (McKechnie and Wolf, 2004), but these T_air were below the T_uc found in this study. Spinifex Pigeons (89 g, Geophaps plumifera; Withers and Williams, 1990) and Crested Pigeons (186 g, Ocyphaps lophotes; B.O.W., unpublished results), like the two species of dove tested here, show T_uc of 46°C. However, three columbiforms, including the diamond dove (∼40 g, Geopelia cuneata), crested pigeon (174 g) and brush bronzewing (224 g, Phaps elegans), show T_uc close to normothermic T_b, which is contrary to the pattern we observed and may be due to acclimation history or measurement dates or conditions (Schleucher, 1999; Larcombe et al., 2003). In each of these cases, the experimental birds were housed outdoors at T_air below 25°C, suggesting a seasonal adjustment of T_uc like that of BMR (McKechnie, 2008). Given their capacities for cutaneous heat dissipation, doves in the wild may only rarely experience heat loads that require panting and an increase in RMR.

Galliform birds such as quail, pheasants and partridges, in contrast, show clear well-defined T_uc where metabolic rate increases markedly with increasing heat stress. T_uc values for galliforms in the heat tend to cluster around normothermic T_b and show only modest variability (Fig. 6). Data from eight studies encompassing six species show that T_uc range from 35 to 42°C with only the sand partridge (Ammoperdix heyi) showing no increase in metabolic rate with T_air ranging from 35 to 50°C. Sand partridges, thus, appear more similar to columbiform birds in that their metabolic rate does not increase steeply at any point (Frumkin et al., 1986), which suggests that an elaborated gular mechanism or enhanced cutaneous evaporation may occur in this species. Our T_uc value for Gambel's quail is similar to other studies for the same species (Goldstein and Nagy, 1985; Weathers, 1981), indicating that T_uc in galliform birds tends to fall just below or slightly above T_b.

Given the differences in individual birds, their acclimation history, the time of year and experimental conditions, the lowest metabolic rates we measured are similar to previous measurements of RMR and BMR in these species. For Gambel's quail in this study, RMR was 4.6 mW g⁻¹ at the T_uc (Table 1), which is less than the BMR of 6.0 mW g⁻¹ found for birds acclimated to 25°C (Weathers, 1981) and comparable with an RMR of 5.0±1.7 mW g⁻¹ (n=24; 33°C≤T_air≤47°C) found in wild birds – as in this study – acclimatized to the Sonoran Desert summer (Goldstein and Nagy, 1985). For mourning doves in this study, RMR was 6.1 mW g⁻¹ at the T_uc, which is less than the BMR of 7.1 mW g⁻¹ observed in birds at higher humidities and acclimated to 25°C (Hudson and Brush, 1964; appendix A of McNab, 2009). For white-winged doves in this study, RMR was 6.5 mW g⁻¹ at the T_uc, which compares favorably with the metabolic rate (6.3±0.8 mW g⁻¹; n=15; 35°C≤T_air≤45°C) found for birds acclimated to 43°C (McKechnie and Wolf, 2004).

Gambel's quail showed a significant increase in EWL at the T_uc of 41.1°C (Table 2, Fig. 2). The marked increase in RMR and EWL together indicate an increase in respiratory evaporation at a T_air coinciding with their normothermic T_b. Doves also increased EWL at their T_uc, but this value was ∼46°C. Their reliance on cutaneous evaporation, which has negligible metabolic costs (Marder and Ben-Asher, 1983; Marder and Gavrieli-Levin, 1987), allowed these doves to defer active heat dissipation via panting and/or gular flutter to a T_air approximately 5°C higher than normothermic T_b.

Despite their differences in M_b, mourning doves and Gambel's quail showed similar responses of EWL above their respective T_uc (Table 2, Table S1). However, as a percentage of M_b, EWL was 30% greater in mourning doves (2.1% M_b h⁻¹) and 45% greater in white-winged doves (2.3% M_b h⁻¹) than that observed in Gambel's quail (1.6% M_b h⁻¹) at a T_air of 48°C (Table 3). White-winged doves increased their EWL with T_air more steeply than the other two species, both above and below T_uc (Table 2, Table S1). Compared with similarly sized Gambel's quail, white-winged doves evaporated water at higher rates over a range of high T_air, which provided a greater cooling capacity, but resulted in greater rates of water loss.

Evaporative water loss generally scales with M_b with increasing T_air in both galliforms and columbiforms (Fig. 7). At the highest T_air measured among columbiforms, ∼55–60°C, EWL ranges from 2–3% M_b h⁻¹ in rock doves to 4–5% M_b h⁻¹ in the smaller dove species (39–147 g). At the highest T_air frequently measured among galliforms, ∼44–45°C, EWL ranges from 0.7% M_b h⁻¹ in the large bedouin fowl (∼1400 g, Gallus domesticus) to 3% M_b h⁻¹ in the small king quail (40 g, Coturnix chinensis). In this study EWL in Gambel's quail at T_air of 50°C measured about 2% M_b h⁻¹.

At T_air just below their respective HTL we found that Gambel's quail are able to produce EHL/MHP of 2.14 at 50°C, with mourning doves producing ratios of 3.08 at 56°C and white-winged doves showing ratios of 3.69 at 58°C (Table 1). In prior studies, only the larger rock doves exhibit ratios approaching 3.0 at 60°C (Marder and Arieli, 1988). The smaller spinifex pigeons had EHL/MHP ratios of 1.25–1.5 at 45°C and 1.75 at 50°C (Dawson and Bennett, 1973; Withers and Williams, 1990). Data for small doves (∼40 g) show that these species can produce EHL/MHP ratios of 1.0 (diamond dove), 1.1–1.2 (Inca dove, Columbina inca) and 1.6 (Namaqua dove, Oena capensis) at 44°C (Schleucher et al., 1991; Lasiewski and Seymour, 1972; MacMillen and Trost, 1967; Schleucher, 1999; Gerson et al., 2014). Among these smaller doves, the Namaqua doves appear to be more heat tolerant than the other smaller species, although this could be an artifact of acclimation or measurement conditions, producing EHL/MHP of 3.55 at 56°C (Gerson et al., 2014), which is comparable to the value we found for white-winged doves at 58°C. Overall, heat-acclimatized doves exposed to T_air of 55–60°C dissipate heat through evaporation at 3- to 3.7-fold their rates of metabolic heat production and thus are able to cope with extremely high environmental heat loads.

Excluding the Gambel's quail used in this study, galliforms ranging in size from 40 to 1400 g have not been shown to produce EHL/MHP ratios >2.0 or tolerate T_air above 48°C. Large Sinai and bedouin fowls (∼1300–1400 g) produce a maximum EHL/MHP of 1.65 at 44°C and apparently tolerate maximum T_air of 48°C (Marder, 1973; Arad and Marder, 1982). Chukar partridge (475 g, Alectoris chukar) are able to produce EHL/MHP of 1.82 at 48°C and reached their heat tolerance limits in trials >48°C (see below; Marder and Bernstein, 1983). Japanese quail (∼100 g, Coturnix japonica) produce a maximum EHL/MHP of 1.12 at a T_air of 43°C. The smallest galliform that has been studied, king quail (∼40 g), produced a EHL/MHP ratio of 1.17 at 45.2°C (Lasiewski et al., 1966) and 1.23 at 44°C (Lasiewski and Seymour, 1972). In general, it appears that galliforms are able to tolerate maximum T_air of ∼48–50°C and are able to evaporate water at rates that can account for slightly more than two times their RMR at the highest T_air.

Facultative hyperthermia represents a means of partially offsetting heat gain in hot environments and reducing EWL and, as a consequence, is a well-developed response in heat stressed and exercising birds. We found that the normothermic T_b for mourning and white-winged doves averaged 41.0 and 41.4°C, respectively (Table 1). The T_b of both species increased linearly with T_air above this range, staying well below lethal temperatures and reaching a value of 41.9°C for T_air of 56°C in mourning doves and 42.7°C for T_air of 58°C in white-winged doves (Table 1). This increase of 0.9°C for mourning and 1.3°C for white-winged doves from normothermic T_b falls below the 1.6 to 4.2°C range of T_b increases found for other columbiform birds exposed to lower T_air of 44–45°C (Lasiewski and Seymour, 1972; MacMillen and Trost, 1967; Schleucher, 1999; Withers and Williams, 1990; Larcombe et al., 2003), although it exceeds the 0.8°C variation in T_b found for heat-acclimated rock doves exposed to T_air of 30–60°C (Marder and Arieli, 1988). At ∼41°C, the normothermic T_b of Gambel's quail and the dove species were very similar. As T_air increased, however, T_b of Gambel's quail had a significantly steeper slope and lower intercept than that of both dove species (Fig. 4, Table 2), reaching 43.6°C at T_air of 50°C (Table 1). Thus, just below the HTL, Gambel's quail maintained a T_air−T_b gradient of just 6.4°C, compared with 14.1 and 15.3°C, respectively, for mourning and white-winged doves.

Establishing the thermal tolerance of animals to heat stress is a prerequisite for understanding how heat exposure and its severity may impact wild animal populations. Historically, these determinations have taken a variety of forms, but the assays typically produced high mortality in the subjects and most have tended to focus on ectotherms (98% of 604 animal taxa reviewed by Lutterschmidt and Hutchison, 1997). In this study, we measured core T_b continuously, rather than recording a single T_b measurement at the end of a temperature trial, which allowed us to determine both maximum T_b of individuals as well as the response of T_b to changes in T_air. Thus, we could determine real time if an individual had reached its thermal limits and terminate the trial without harming the bird. The HTL metric provided a clear T_air above which birds could not maintain a constant T_b. Using a protocol similar to ours, Marder and Bernstein (1983) found that chukar partridges exposed to T_air of 48°C maintained a stable T_b for 2–4.5 h, but could not maintain T_b below lethal levels at T_air of 52°C. Thus, like the quail observed in our study, this large galliform (475 g) reached its limit of heat tolerance at a T_air of 52°C. Our findings and those of others provide strong evidence that desert-dwelling columbiform birds, primarily because of their high rates of cutaneous evaporation, cope better with high heat loads than galliforms in the same environment.

Williams et al. (2012) has suggested that desert birds need the capacity to increase EWL during periods of extreme heat in order to maintain sub-lethal T_b, while minimizing dehydration over the long term. In this study, we found that the thermoregulatory performance and water balance of mourning and white-winged doves differed greatly from that of the sympatric Gambel's quail. The quail showed lower RMR and lower mass-specific EWL at T_air above normothermic T_b, which serves to conserve water and energy, but also potentially makes them vulnerable to increasing temperatures, especially during heat waves. Doves, in contrast, support a high water use ‘lifestyle’ when environmental temperatures exceed normothermic T_b. Thus, at a T_air of 48°C, mourning and white-winged doves lose 10% M_b in water over a period of 4–5 h, while this takes 6 h or more in Gambel's quail. Greater rates of EWL in doves could reduce their abundance if water were unavailable, for example, because of longer or more intense droughts (IPCC, 2013; Albright et al., 2010); but it also allows them to survive at temperatures of 60°C or more. With their more constrained EWL and greater reliance on respiratory EWL, Gambel's quail appear to have heat tolerance limits that are significantly lower than those observed in doves – but may persist without water for longer periods in moderate heat.

The differing thermoregulatory responses among the three species to high T_air, in part, may reflect their differing ecologies within the Sonoran Desert. Mourning and white-winged doves often nest in exposed locations with significant solar heat loads, where the nest environment can reach T_e of 50–60°C and incubating birds must cool rather than warm their eggs to maintain viability (B.O.W., personal observation; Russell, 1969; Walsberg and Voss-Roberts, 1983). Doves require access to preformed water almost daily and travel up to 20 km or more access water holes (Gubanich, 1966; Brown, 1989); their strong flight capabilities allow them to nest many miles from water sources (Bartholomew and Dawson, 1954). In addition, doves are able to rapidly drink by suction, placing their bills in water sources and drinking continuously. Mourning doves deprived of water for 24 h can drink 17% of their M_b in water in 10 min (Bartholomew and Dawson, 1954). Gambel's quail are much less mobile and typically range about 0.5 km from their roost sites (Gorsuch, 1934). During the Sonoran Desert summer, they constrain their activities to shaded microsites during the heat of the day, with only intermittent exposure to the sun when moving between shaded sites or flying short distances (Goldstein, 1984; Goldstein and Nagy, 1985). Gambel's quail often live independent of surface water resources and rely on vegetation and insects in desert valleys far from water (Lowe, 1955; Hungerford, 1960; Bartholomew and Cade, 1963). Very high air temperatures, however, may constrain the ability of quail to forage and potentially limit their distribution in future climates (Goldstein, 1984; McKechnie and Wolf, 2010).

It is critical that we understand the thermoregulatory and water balance challenges that birds face in a rapidly warming world. Such understanding might allow us predict and perhaps mitigate such events; it is a prerequisite for understanding avian distributions now and in a warmer future. This study is a first step in examining the relative responses of birds to higher temperatures and gives us some baseline measure of their ecological performance. Given the apparent differences among species in physiological responses to heat stress, examining the tolerances of naturally occurring avian communities in similar ways will be critical for understanding their composition and persistence.

We thank the Cadden and Wolf families for allowing us to conduct this research on their properties. Matt Baumann, Corrie Borgman, Jennifer Clark, Mateo Garcia, Michael Griego, Chuck Hayes, Alex Kusarov, Ben Smit and Bill Talbot provided invaluable assistance in the field and laboratory. We thank Christian Giardina and Andrew McKechnie for helpful comments on the manuscript.