How birds dissipate heat before, during and after flight

Abstract

Animal flight uses metabolic energy at a higher rate than any other mode of locomotion. A relatively small proportion of the metabolic energy is converted into mechanical power; the remainder is given off as heat. Effective heat dissipation is necessary to avoid hyperthermia. In this study, we measured surface temperatures in lovebirds (Agapornis personatus) using infrared thermography and used heat transfer modelling to calculate heat dissipation by convection, radiation and conduction, before, during and after flight. The total non-evaporative rate of heat dissipation in flying birds was 12× higher than before flight and 19× higher than after flight. During flight, heat was largely dissipated by forced convection, via the exposed ventral wing areas, resulting in lower surface temperatures compared with birds at rest. When perched, both before and after exercise, the head and trunk were the main areas involved in dissipating heat. The surface temperature of the legs increased with flight duration and remained high on landing, suggesting that there was an increase in the flow of warmer blood to this region during and after flight. The methodology developed in this study to investigate how birds thermoregulate during flight could be used in future studies to assess the impact of climate change on the behavioural ecology of birds, particularly those species undertaking migratory flights.

1. Introduction

In many endotherms, the regulation of heat exchange with the environment allows them to maintain a stable high body temperature. Increased heat production and retention serve to avoid hypothermia at ambient temperatures below the lower critical temperature, while at ambient temperatures above the upper critical temperature and/or during intense exercise, heat dissipation serves to avoid hyperthermia. The heat dissipation limitation theory [1] hypothesizes that constraints on the maximal capacity to dissipate body heat sets a limit on an animal's energetic expenditure (especially during locomotion) to avoid hyperthermia. This theory is supported by the reduction in the metabolic rate in warmer conditions in birds living in arid environments (e.g. [2,3]) and hot acclimatized birds [4].

During flight, birds exhibit the highest rate of metabolic energy expenditure compared with other modes of locomotion [5,6] with heat production 10–19× higher than at rest [7,8]. Heat dissipation areas (HDAs) or ‘thermal windows’, are areas of the body where the surface temperature (T_s) is elevated well above ambient temperature (T_a) (e.g. T_s − T_a = 5–8°C in hummingbirds; [9]) and other areas of the body. Three main HDAs have been identified during flight in birds: (i) the head, in particular the unfeathered hot spot eye area, (ii) the feet, and (iii) the proximal wing with the hot spot around the shoulder area [9–11]. In addition, it appears that in species with relatively large bills, the bill is an important HDA after flight, accounting for 35–60% of total heat exchange in toco toucans (Ramphastos toco; [12]), 10–18% in tufted puffins (Fratercula cirrhata; [13]) and 1.4–19.9% in southern yellow-billed hornbills (Tockus leucomelas; [14]). The relative importance of different HDAs in heat dissipation varies. For example, in flying European starling (Sturnus vulgaris), 55% of the overall heat loss was dissipated by the wing with 19% dissipated by the shoulder area hot spot [10]. In addition, heat dissipation varies with flight speed. During flight, overall metabolic rate typically exhibits a U-shaped relationship with speed [15,16]. The overall metabolic rate is equal to the sum of the rate of heat production and mechanical power. Since mechanical power also exhibits a U-shaped relationship with speed [16,17], it is, therefore, expected that the rate of heat loss will also follow a similar U-shaped relationship with flight speed. Heat dissipation during flight has only been quantified in European starling where, contrary to the expectation of U-shaped relationship, heat transfer increased linearly with flight speed, between 6 and 14 m s⁻¹ [15]. In flying calliope hummingbirds (Selasphorus calliope), the temperature of the shoulder region shows a U-shaped relationship with speed [9], whereas the eyes, feet and the mean body surface temperatures decreased with increasing flight speed [18]. However, since heat loss from a body region is dependent on several factors, including its surface area, air speed and the difference between T_s and T_a, the relative importance of each HDA in dissipating heat at rest, during and after flight, is currently unclear, having never been quantified.

The aim of the study was to quantify regional variation in T_s and to calculate whole-body and regional heat dissipation before, during and after flight in lovebirds (Agapornis personatus; [19]). We hypothesized that during flight, heat dissipation would increase compared with rest due primarily to the need to dissipate the heat generated by the locomotory muscles as they convert metabolic energy into mechanical work, and the relatively low efficiency of this process (e.g. [16]). It was also hypothesized that the elevated metabolism that occurs following exercise [20] may result in increased heat dissipation immediately following flight, but that this would decrease to pre-flight levels over time. Moreover, the change in the bird's posture, with closed wings after landing [12], may impact upon the relative importance of the HDAs in perched birds. We additionally investigated the effects of the preceding flight duration on T_s and heat dissipation with the hypothesis that birds that performed relatively longer flights, or flights at more energetically demanding higher or lower speeds, would have higher post-flight T_s and heat loss.

2. Material and methods

Between May and June 2018, seven masked lovebirds (body mass 50.17 ± 0.21 g; mean ± standard error (s.e.)) were trained to fly in an open circuit Eiffel-pattern wind tunnel with a working section with dimensions 0.52 × 0.52 × 1.00 m (width × height × length), 5 days per week at the University of Leeds (Leeds, West Yorkshire, UK; 53°48′ N, 1°33′ W). Airspeed was controlled by setting the speed of the fan; the relationship between fan speed and airspeed was determined on the day and prior to data collection using a Pitot-static probe and differential pressure transmitter (FCO332, Furness Controls, Bexhill-on-Sea, East Sussex, UK). Upstream and downstream of the bird a series of vertical, nylon cord restricted the bird to the working section of the wind tunnel. The RMS turbulence in the centre of the working section was less than 0.35% [21]. Birds were maintained in an indoor aviary with access to water, food and cuttlefish ‘bones’ ad libitum. Birds were flown at up to two randomly selected speeds per day (selected from 5, 6, 7, 8, 9, 10, 11, 12, 13, 14 and 15 m s⁻¹ which covered speeds both below and above the minimum power speed (A Lewden, CM Bishop and GN Askew 2018, unpublished data)), for a mean duration of 179 ± 9 s (± s.e. with a minimum flight duration of 144 s and a maximum flight duration of 547 s). Birds performed between 2 and 16 flights at different speeds with a mean of 11 flights per bird and there were no differences in flight speeds between birds (ANOVA; p = 0.791) and no individual flew longer than the other (ANOVA; p = 0.115).

Birds were recorded continuously throughout these flights from the initial period when they were sitting on a wooden perch in the working section of the wind tunnel before flight (hereafter pre-flight), during the flight (hereafter flight) and once they had landed and were once more sitting on the perch in the working section of the wind tunnel (hereafter post-flight) using a thermal camera (FLIR A65 with a 13 mm lens; accuracy ± 5% of reading; FLIR Systems AB, Sweden). The thermal camera was calibrated using a blackbody source (model 989, Isothermal Technology Ltd, Southport, Merseyside, UK) at a distance of 0.4 m, allowing for the correction of non-uniformity across the field of view. The wind tunnel was switched on and off immediately before and after the flight, respectively. T_s was measured on perched birds in still air during a period of 14 ± 2 s (mean ± s.e.) immediately preceding the flight (i.e. pre-flight) and on perched birds in still air immediately after the end of the flight (i.e. post-flight) during a period of 39 ± 4 s (mean ± s.e.). Relative humidity (mean of 47.36 ± 0.20% (± s.e.)) and ambient temperature (mean of 23.69 ± 0.02°C (± s.e.); range 23–24.5°C) were recorded.

Thermal videos of the bird were recorded in the lateral view through a hole cut in the side of the wind tunnel at a constant distance (0.4 m) between the thermal imaging camera and the bird. Thermal images were analysed using FLIR Tools software (v. 6.4, FLIR systems, Wilsonville, Oregon, USA). Images were selected where the bird was sharply focused and where its lateral view was parallel to the recording plane of the thermal imaging camera. In the analysis software, emissivity was set to 0.95 [10] and distance to the subject to 0.4 m; T_a and relative humidity were set accordingly to the values recorded on the day of the experiment. At least two images per bird were selected in pre-flight before each flying speed with a mean of 3.4 ± 0.5 images and a maximum of 13 images.

At least one image per speed per bird was selected during flight and on six individuals in post-flight, with a mean of 3.9 ± 0.3 images and a maximum of nine images selected in flight and a mean of 4.7 ± 0.7 and a maximum of 16 images in post-flight.

For the images selected of the non-handled bird sitting on the perch, polygons were fit to three body areas and the mean T_s determined. The body areas were: the head (hereafter head) including eye and bill hot spots, the legs (hereafter legs) including tarsi and feet, and the trunk and the dorsal surface of the folded wing (hereafter trunk) in lateral view including a part of the back, the upper surface of the wing and a part of the ventral area (figure 1). Each body area measured was larger than three times the spot size of 0.68 mm with our thermal camera [22,23]. Body surface areas were measured after the birds were euthanized following another study. For each bird, a set of digital images was taken in upstroke lateral position and included a calibration scale. The calibrated images were then analysed using ImageJ software (US National Institutes of Health) to determine the areas of each region of interest. Mean temperature of the wooden perch was determined using a standardized square measuring the size of the width of lovebird's feet. During flight, we used the mean T_s of similar body areas recorded in a lateral view, at the transition between the upstroke and the downstroke: the head, the trunk (completely visible during late upstroke), legs and the ventral proximal wing area in upstroke (hereafter wing; figure 1c,d), including the axillary region (armpit) and the region of the lower surface of the wing held above T_a (i.e. excluding the flight feathers). Only surface temperatures of sharply focused body areas were used for thermal imaging.

Figure 1.

Thermal images of two different perched lovebirds (a–c–e Individual 1 and b–d–f Individual 2) within the wind tunnel working section in pre-flight (a,b), in flight (c,d) and in post-flight (e,f).

Convective (q_conv) heat loss (W) was calculated for each body area as follows:

$$q_{\mathrm{conv}}=h_{\mathrm{c}}A(T_{\mathrm{S}}+T_{\mathrm{a}}),$$ 2.1

where h_c is the convective heat transfer coefficient, determined using the Nusselt number calculated following: method 1 in [10] for the wing (wing represented by a flat plate); following method 2 [10] for the head (section represented as a sphere) and trunk (section represented as cylinder); and following method 3 [10] for the legs (section represented as a cylinder). During the pre- and post-flight periods when the airflow in the wind tunnel was still, h_c for free convection was used. During flight period, h_c for forced convection was used to account for the effects of airflow.

Radiative (q_rad) heat loss (W) was calculated for the pre-flight, flight and post-flight periods using the following relationship:

$$q_{\mathrm{rad}}=\sigma A{\epsilon }_{\mathrm{h}}{\epsilon }_{\omega }(T_{\mathrm{s}}^4-T_{\mathrm{a}}^4),$$ 2.2

where σ is the Stefan–Boltzmann constant, A is the surface area of the body region, ε_h represents the bird emissivity (assumed to be 0.95; [10,24], ε_ω is the wall emissivity (assumed to be 1), T_s and T_a in K.

Conductive heat loss (q_cond) was calculated for the area of the feet that was in contact with the wooden perch (following [25]) during pre-flight and post-flight as follows:

$$q_{\mathrm{cond}}={\displaystyle \frac{k_{\mathrm{s}}\hspace{0.17em}(T_{\mathrm{s}}-T_{\mathrm{p}})(1-\lambda )A_{\mathrm{f}}}{\sqrt{(K_{\mathrm{s}}\pi t)}\hspace{0.17em}}\sum _{n=0}^{\mathrm{\infty }}{(-\lambda )}^n{\mathrm{e}}^{-{(n+(1/2))}^2{d}^2/(K_{\mathrm{s}}t)},}$$ 2.3

where k_s is thermal conductivity of skin (0.284 W m⁻¹ K⁻¹, value for heel assumed; [26]), T_p is the temperature of the perch, K_s is the thermal diffusivity of skin (1.41 × 10⁻⁷ m² s⁻¹, value for heel assumed; [26]), A_f is the area of the feet in contact with the perch and t is the time duration spent on the perch before the thermal image was recorded. Finally, λ was calculated as follows:

$$\lambda =\left({\displaystyle \frac{k_{\mathrm{s}}}{\sqrt{K_{\mathrm{s}}}}-{\displaystyle \frac{k_{\mathrm{p}}}{\sqrt{K_{\mathrm{p}}}}}}\right)÷\left({\displaystyle \frac{k_{\mathrm{s}}}{\sqrt{K_{\mathrm{s}}}}+{\displaystyle \frac{k_{\mathrm{p}}}{\sqrt{K_{\mathrm{p}}}}}}\right),$$ 2.4

where k_p is thermal conductivity of the perch (0.093 W m⁻¹ K⁻¹, value for softwood assumed; [27]) and K_p corresponds to the thermal diffusivity of ground surface (1.44 × 10⁻⁷ m² s⁻¹, value for softwood assumed; [27]).

Evaporative heat losses could not be quantified with the current experimental set-up; however, the magnitude of these losses has previously been shown to represent only approximately 7–12% of the total heat dissipation (e.g. 7% in herring gulls (Larus argentatus) at 25°C [28]; 11.6% in European starlings [15]).

The sum of convective, radiative and conductive heat loss (representing an estimated 88.4% of the total heat lost), will be called hereafter non-evaporative heat dissipation. Measurements of surface temperature were made unilaterally and were assumed to be bilaterally identical. When calculating regional and whole-body non-evaporative heat dissipation, unilateral surface areas were doubled to account for the two sides of the body.

2.1. Statistical analysis

To describe the mean T_s of the three common body areas (i.e. head, trunk and legs) measured during pre-flight, flight and post-flight, we used a general linear mixed model (GLMM) including T_a, period (i.e. pre-flight, flight and post-flight), and timing within the period (min) nested in period as independent variables and bird ID as random effect. Significance between periods was assessed using a Tukey's honestly significant difference (HSD) post hoc test. The same GLMM was used to describe wing surface temperature but without the period factor and interaction as wing was only measured during flight. To study the effects of flying condition on T_s during post-flight, we used a GLMM including T_a, timing in post-flight (min), flying speed as binomial factor and flight duration (min) as independent variables and bird ID as random effect. To investigate the non-evaporative heat dissipation of each body area, we used the same GLMM to describe the mean T_s and then compare the effect of mean T_s on the non-evaporative heat dissipation during the three periods. We performed the statistical analysis using JMP v. 13 (SAS Institute Inc., Cary, North Carolina, USA). Finally, we compared the total non-evaporative heat dissipation during the three periods using ANOVA. Results are reported as means ± s.e.

3. Results

3.1. Surface temperature

The mean T_s of the head (T_head), trunk (T_trunk) and legs (T_legs) areas were positively related to T_a (table 1) and were related to the interaction between category and time within period (table 1). Specifically, T_head remained stable in pre-flight as well as during flight but significantly increased in post-flight (table 1; figure 2a). Similarly, T_trunk remained stable in pre-flight, slightly increased during flight and increased in post-flight (figure 2b). T_legs remained stable in pre-flight, and significantly increased during flight and in post-flight (figure 2c). Finally, we measured a higher mean T_head in pre- and in post-flight (which were not significantly different from each other (p = 0.9780)) than in flying birds (p < 0.0001; figure 2a). Similarly, mean T_trunk was also higher in pre- and in post-flight (which are not significantly different from each other (p = 0.9314)) than in flying birds (p = 0.0003 and p < 0.0001 respectively; figure 2b). Mean T_legs was higher in post-flight (compared with in pre-flight (p = 0.0005) and in flight (p < 0.0001)) but was similar in pre-flight and in flight (p = 0.9781) (figure 2c). Finally, the mean T_s of the wing during flight was positively correlated to T_a (estimate of 0.76 ± 0.24; p = 0.0020) but remained stable throughout the flying period (p = 0.7092) (n = 224 N = 7, R² = 0.14).

Table 1.

Summary of general linear mixed models (GLMMs) investigating the determinants of head, trunk and legs surface temperatures. Values not sharing a common letter are significantly different for p < 0.05 using Tukey's HSD post hoc test. Bold values denote statistical significance at the p < 0.05 level.

	head (°C)					trunk (°C)					legs (°C)
		n = 403	R2 = 0.76				n = 419	R2 = 0.69				n = 404	R2 = 0.62
	Vresidual = 2.63 (n = 403 pictures), VID = 0.13 (N = 7 individuals)					Vresidual = 1.90 (n = 419 pictures), VID = 0.31 (N = 7 individuals)					Vresidual = 4.59 (n = 404 pictures), VID = 0.64 (N = 7 individuals)
	d.f.	F	prob. > F			d.f.	F	prob. > F			d.f.	F	prob. > F
Ta	1,394	29.96	<0.0001			1,409	14.70	0.0001			1,393	5.07	0.0249
Period	2,395	185.93	<0.0001	estimates (°C)	post-hoc	2,409	409.00	<0.0001	estimates (°C)	post-hoc	2,394	119.58	<0.0001	estimates (°C)	post-hoc
pre-flight				33.67 ± 1.30	a				30.31 ± 1.03	a				27.35 ± 1.49	b
flight				27.26 ± 0.17	b				26.36 ± 0.24	b				27.01 ± 0.35	b
post-flight				33.50 ± 0.34	a				29.93 ± 0.33	a				33.30 ± 0.49	a
Timing[period]	3,394	45.01	<0.0001	estimates (°C min−1)		3,409	408.70	<0.0001	estimates (°C min−1)		3,393	82.82	<0.0001	estimates (°C min−1)
pre-flight			p = 0.9187	-				p = 0.9187					p = 0.2203	-
flight			p = 0.5580	-				p = 0.0112	0.10 ± 0.04				p < 0.0001	0.91 ± 0.06
post-flight			p < 0.0001	2.51 ± 0.22				p < 0.0001	2.53 ± 0.18				p < 0.0001	1.46 ± 0.27

Figure 2.

Surface temperature (a–c) and heat dissipation (d–f) of body areas: head (a,d) trunk and wing (b,e) and legs (c,f) measured in pre-flight (empty square) in flight (full circle) and in post-flight (empty triangle). The dashed line denotes no exchange with environment, whereas negative values correspond to heat gain and positive value to heat lost. Each dot corresponds to one thermal image and several images have been analysed per flight per period (see text for statistical details). Note that the flight data and the post-flight data have each been aligned to the same start time (because the duration of the pre-flight and flight varied).

T_s measured in post-flight was positively correlated to T_a and flight duration, independently to the HDA considered (table 2). Indeed, we measured a significant but slight increase in T_s in post-flight as a function of flight duration in head (estimate of 0.009 ± 0.003°C), trunk (estimate of 0.008 ± 0.002°C) and in legs (estimate of 0.01 ± 0.003°C) (table 2), with T_s increasing with increasing flight duration. However, flight speed during flight only impacted the T_legs in post-flight (table 2).

Table 2.

Summary of general linear mixed models (GLMMs) investigating the effect of flying condition on mean surface temperature in post-flight in head, trunk and legs. Bold values denote statistical significance at the p < 0.05 level.

in post-flight:	head (°C)			trunk (°C)			legs (°C)
	n = 104		R2 = 0.46	n = 114		R2 = 0.67	n = 105		R2 = 0.48
	Vresidual = 4.9 (n = 104 pictures), VID = 0.06 (N = 7 individuals)			Vresidual = 2.64 (n = 114 pictures), VID = 0.78 (N = 7 individuals)			Vresidual = 4.09 (n = 105 pictures), VID = 2.20 (N = 7 individuals)
	d.f.	F	prob. > F	d.f.	F	prob. > F	d.f.	F	prob. > F
Ta	1,98	23.77	<0.0001	1,105	16.31	0.0001	1,94	10.67	0.0015
period	1,1	17.46	<0.0001	1,106	106.44	<0.0001	1,96	25.34	<0.0001
speed			n.s.	1,106	3.61	0.0603	1,95	9.38	0.0029
speed2			n.s.	1,106	2.84	0.0949	1,95	6.95	0.0098
flight duration	1,1	14.17	0.0003	1,108	10.26	0.0019	1,95	9.50	0.0027

3.2. Mechanism of heat dissipation

Heat was mainly lost by radiation during pre- and post-flight with a mean of 80.73% and 76.42% of the total non-evaporative heat dissipation, respectively (figure 3a). However, during flight birds lost only 17.16% of the total non-evaporative heat dissipation by radiation and the remaining 82.84% by convection (figure 3a). Conductive heat lost represented 6.98% and 11.10% of the total non-evaporative heat dissipation during pre- and post-flight, respectively.

Figure 3.

Comparison of the per cent of the total heat dissipation (%) lost by convection, radiation and conduction (a), the amount of heat (W) dissipated by the head, trunk and wing in function of experimental periods (b) and the representing per cent of the total heat dissipation (%) dissipated by the three body areas in pre- and post-flight and in four body areas during flight (c).

3.3. Non-evaporative heat dissipation

The total non-evaporative heat dissipation was significantly higher during flight (4.46 ± 0.05 W) than in pre-flight (0.36 ± 0.01 W) and in post-flight (0.23 ± 0.01 W; paired t-test p < 0.0001; figure 3b) with similar heat dissipation in pre- and post-flight (paired t-test p = 0.1831; figure 3b). In perched individuals during pre-flight, the total amount of heat was mainly dissipated by trunk (56.6%) and head (34.7%) with the remainder being dissipated by the legs (8.7%; figure 3c); these HDA represent, respectively, 25%, 72% and 3% of the whole-body surface area. Thus, when normalized to surface area (i.e. heat flux), the legs had the highest heat flux, 0.026 ± 0.004 W cm⁻², compared with the head with 0.014 ± 0.001 W cm⁻² (p = 0.0041) and trunk with 0.008 ± 0.001 W cm⁻² (p < 0.0001), which both had similar heat flux (p = 0.0652). In pre-flight, heat flux was 1.85- and 3.25-fold higher in legs than in head and in trunk, respectively. During flight, the wing dissipated 85.9% of the total amount of heat (figure 3c), but represented only 26% (i.e. ventral proximal area) of the whole-body surface area. In comparison, the head dissipated 7.8% of the heat, with the area representing 18% of the whole-body surface, the trunk dissipated 4.5% of the heat and represented 54% of the body surface, and the legs dissipated 1.8% of the heat and represented 2% of the whole-body surface area (figure 3c). During flight, when normalized to surface area, the wing had the highest heat flux (0.32 ± 0.02 W cm⁻²) compared with the legs (0.062 ± 0.08 W cm⁻²; p < 0.0001), head (0.042 ± 0.005 W cm⁻²; p < 0.0001) and trunk (0.008 ± 0.001 cm⁻²; p < 0.0001), with the legs exhibiting a higher heat flux than the trunk (p < 0.0001) and head (p = 0.0541), and the head a higher heat flux than the trunk (p = 0.0027). During flight, heat flux was 5.16-, 7.62- and 40-fold higher in wing than in legs, head and in trunk, respectively. In post-flight, when normalized to surface area, perched birds dissipated heat mainly by the trunk (45.7%) and head (40.6%), whereas the legs dissipated 13.7% of the total amount of heat (figure 3c). During post-flight, when normalized to surface area, legs had the highest heat flux 0.03 ± 0.007 W cm⁻² compared with the trunk with 0.003 ± 0.02 W cm⁻² (p < 0.0001) and head with 0.009 ± 0.004 cm⁻² (p < 0.0001), and head exhibiting a higher heat flux than the trunk (p = 0.0135). In post-flight, heat flux was 10- and 3.33-fold higher in legs than in trunk and head, respectively. In the three common HDA measured at each period, non-evaporative heat dissipation varied differently as a function of period and time within the period (table 3). The rate of change of non-evaporative heat dissipation by the head was constant in pre-flight and in flight but increased in post-flight (table 3, figure 2d). Heat dissipated by the trunk remained stable in pre-flight, slightly increased during flight and increased in post-flight (table 3, figure 2e). Finally, heat dissipated by the legs decreased in pre-flight increased during flight (figure 2f) and remained stable in post-flight (table 3). Moreover, independently of T_a (p = 0.3346) heat dissipated by the wing area was negatively correlated with flying time (slope −0.15 ± 0.02 W min⁻¹; n = 224 N = 7, R² = 0.40).

Table 3.

Summary of general linear mixed models (GLMMs) investigating the determinants of head, trunk and legs dissipation. Values not sharing a common letter are significantly different for p < 0.05 using Tukey's HSD post hoc test. Bold values denote statistical significance at the p < 0.05 level.

	head dissipation (W)					trunk dissipation (W)					legs dissipation (W)
	n = 403	R2 = 0.57				n = 396	R2 = 0.41				n = 405	R2 = 0.70
	Vresidual = 0.01 (n = 403 pictures), VID = 0.001 (N = 7 individuals)					Vresidual = 0.005 (n = 396 pictures), VID = 0.0005 (N = 7 individuals)					Vresidual = 0.0009 (n = 405 pictures), VID = 0.0001 (N = 7 individuals)
	d.f.	F	prob. > F			d.f.	F	prob. > F			d.f.	F	prob. > F
Ta	1,393	0.79	0.3752			1,386	3.35	0.0681			1,394	0.84	0.3600
Period	2,394	48.21	<0.0001	estimates (W)	post-hoc	2,387	0.09	0.9172	estimates (W)	post-hoc	2,395	33.87	< 0.0001	estimates (W)	post-hoc
pre-flight				0.08 ± 0.09	b					-				−0.03 ± 0.02	c
flight				0.35 ± 0.01	a				0.18 ± 0.03	-				0.06 ± 0.005	a
post-flight				0.13 ± 0.03	b					-				0.02 ± 0.006	b
Timing[period]	3,394	3.25	0.0219	estimates (W min−1)	post-hoc	3,386	27.93	<0.0001	estimates (W min−1)		3,395	154.56	<0.0001	estimates (W min−1)
pre-flight			p = 0.6987	-				p = 0.6845	-				p = 0.0072	−0.03 ± 0.002
flight			p = 0.1452	-				p < 0.0001	0.008 ± 0.002				p < 0.0001	0.02 ± 0.00
post-flight			p = 0.0066	0.04 ± 0.01				p = 0.0072	0.07 ± 0.01				p = 0.6790	-

Globally, the head dissipated more heat during flight than during any other periods (p < 0.0001; table 3, figure 4a). The trunk dissipated the same amount of heat during the three periods (table 3, figure 4b) and the legs dissipated the most of heat during flight (p < 0.0001 and p < 0.0001 compared with pre- and post-flight, respectively), and more heat in post-flight than in pre-flight (p = 0.017) with a slight heat gain (table 3, figure 4c).

Figure 4.

Least-square means ± s.e. are presented from the final statistical model of surface temperature (circles—table 2) and non-evaporative heat dissipation (columns—table 4) analysis in pre-flight, flight and post-flight in head (a), in trunk (b; yellow bar indicates heat dissipation by wing area during flight) and in legs (c). During flight, the filled and open circles are the surface temperature of the trunk and wing, respectively. Different letters underline a significant difference according to Tukey's HSD post hoc test.

Within the first 30 s after landing, the trunk (n = 8 T_trunk recorded after seven different flights led by N = 4 individuals) and legs (n = 1 T_legs) areas both showed a net heat gain with negative heat dissipation due to the mean T_trunk (23.00 ± 0.20°C) and T_legs (21.70°C) being lower than the mean T_a (23.78 ± 0.13°C; figure 5). Based on a measurement in one bird after it had been sitting on the perch for 256 s during the post-flight period, it appears that eventually the rate of heat dissipation starts to decrease sometime after approximately 150 s for both the head and trunk regions as well as for the whole-body (significant binomial relationship in head, trunk and whole-body non-evaporative heat dissipation (p < 0.0001) but not in feet dissipation (p = 0.117; figure 5)).

Figure 5.

Non-evaporative heat dissipation of the head (purple triangles), trunk (grey triangles) legs (orange triangles) and the sum (red triangles) of these three body areas measured during post-flight in seven lovebirds. Positive values represent heat loss and negative ones correspond to heat gain.

Non-evaporative heat dissipation measured in post-flight in the three HDA was positively correlated with T_a (table 4). Moreover, heat dissipated was positively correlated with the timing after landing in the head and trunk (table 4). The heat dissipated by the three HDA increased with increased duration of the preceding flight (table 4). Finally, the preceding flight speed influenced the heat dissipated by the legs in post-flight with a slight decrease in non-evaporative heat dissipation when the preceding flight was at low speed (below 7 m s⁻¹; slope −0.01 ± 0.004 W m s⁻¹) but increased slightly when the preceding flight was at high speed (above 7 m s⁻¹; 0.001 ± 0.00 W m s⁻¹).

Table 4.

Summary of general linear mixed models (GLMMs) investigating the effect of flying condition on heat dissipation in post-flight in head, trunk and legs. Bold values denote statistical significance at the p < 0.05 level.

in post-flight:	head dissipation (W)			trunk dissipation (W)			legs dissipation (W)
	n = 100	R2 = 0.64		n = 103	R2 = 0.64		n = 105	R2 = 0.39
	Vresidual = 0.001 (n = 100 pictures), VID = 0.00002 (N = 6 individuals)			Vresidual = 0.002 (n = 103 pictures), VID = 0.001 (N = 6 individuals)			Vresidual = 4.15 (n = 105 pictures), VID = 0.00 (N = 5 individuals)
	d.f.	F	prob. > F	d.f.	F	prob. > F	d.f.	F	prob. > F
Ta	1,92	8.98	0.0035	1,94	4.13	0.0450	1,96	7.38	0.0078
period	1,93	69.89	<0.0001	1,96	86.93	<0.0001			n.s.
speed	1,92	3.22	0.0757	1,95	2.94	0.0895	1,97	9.35	0.0029
speed2	1,92	2.98	0.0879	1,95	2.33	0.1305	1,97	6.75	Vresidual
flight duration	1,94	9.10	0.0033	1,96	10.01	0.0021	1,97	4.82	Vresidual

4. Discussion

Quantification of regional surface temperature has given us insight into the dissipation of heat across different regions of the body in lovebirds, before, during and after flight. According to our hypothesis, at the whole-body scale we demonstrated that total non-evaporative heat dissipation during flight was 12× higher than in pre-flight and 19× higher than in post-flight (figure 3b). The highest rate of non-evaporative heat dissipation occurred during flight, simultaneously lowering T_head, T_trunk and T_legs compared with in the pre- and post-flight periods (figure 2a–c). During flight, convection, facilitated by the constant airflow over the bird when flying (i.e. forced convection; see below; figure 3a), was the primary mechanism for heat dissipation. Non-evaporative heat dissipation in perched birds during the pre-flight period (i.e. without airflow) remained stable in the head and trunk areas and slightly decreased in the legs area as a result of the largely stable surface temperatures (figure 2a–c), confirming that our birds were in resting state before flight. On the contrary, during the 150 s after the end of the flight (i.e. post-flight) when the bird was perching, there was a steady increase in heat dissipation in the head and trunk areas, but heat dissipation by the legs remained stable (figure 5). During flight, the head and legs dissipated more heat (0.35 and 0.08 W, respectively) than in pre-flight, whereas the trunk dissipated a similar amount of heat (0.18 W) during pre-flight and flight. However, the newly exposed ventral, proximal region of the wings in flying birds, while representing only 26% of the whole-body surface area, dissipated the highest, 85.9%, amount of heat (3.84 W of the total 4.47 W during flight; figure 3c) at the highest area-specific rate (0.32 W cm⁻²) of any region of the body.

The total metabolic rate during flight encompasses both the heat dissipated and the mechanical power generated by both the flight muscles and the circulatory and breathing systems. Therefore, the relative increase in metabolic rate is expected to be slightly higher than the increase in heat dissipation, approximately 13–14× resting metabolic rate (estimated using mechanical and metabolic power in cockatiels [16] and assuming evaporative heat loss represents 12% of the total metabolic rate [15]). The increase in metabolic rate during flight in other species ranges from 12× to 19× that at rest (13.8× in cockatiel Nymphicus hollandicus; 19.0× in budgerigars Melopsittacus undulates [29]; 12.0× in barnacle goose Branta leucopsis and bar-headed goose Anser indicus [30]; 13.5× in Europeans starlings S. vulgaris [31]), which is similar to the range estimated from the change in non-evaporative heat dissipation during flight in lovebirds. During flight, the overall non-evaporative heat dissipation by radiation and convection was 4.5 W. In a separate study on the same individuals (A Lewden, CM Bishop and GN Askew 2018, unpublished data), the gross metabolic rate measured using respirometry ranged from 7.5 to 8.2 W across the same speed range. Assuming an overall flight efficiency of 16.5% [30], the expected heat dissipation is 6.3–6.8 W, of which we estimate 88% is lost via radiation and convection (12% being lost via evaporative water loss [30]) giving 5.5–6.0 W of sensible heat loss. Our estimates from infrared thermography, 4.5 W, are a little lower than this estimate of the heat dissipated based on respirometry measurements. There are several possible reasons for this discrepancy: (i) during respirometry the bird flew while equipped with a respirometry mask and tubing, which may increase the metabolic cost of flight (and therefore heat dissipation), whereas the infrared thermography flights were performed without additional equipment; (ii) heat dissipation may have been underestimated because some areas of the body were not visible in the lateral view (e.g. the back) or because of errors introduced through the assumptions in the heat transfer model; (iii) the overall flight efficiency used to predict the sensible heat loss is uncertain given that it is based on an aerodynamic estimate of mechanical power, which has inherent uncertainty [17], and metabolic power obtained using the doubly labelled water technique, which lacks precision [30]; (iv) the assumption that evaporative water loss accounts for 12% of the dissipated heat may not be correct in lovebirds; and (v) not all heat may have been dissipated, some may have been stored.

Flapping flight is the most costly mode of locomotion in terms of the rate of metabolic energy expenditure [5], and is associated with the highest known rate of heat production among endotherms [32]. In our study, the elevation of non-evaporative heat dissipation during flight reflected this increased heat production [7,8]. However, the larger amount of heat dissipated was not due to a larger temperature gradient between T_s and T_a in the flying bird. On the contrary, we measured a decrease in T_s for each region, resulting in lower global T_s during flight compared with either pre- or post-flight (figure 2a–c), and a lower overall temperature gradient (T_s–T_a) maintained during flight at +3.77°C in trunk and at +14.04°C in the eye region in our lovebirds, which is similar to the temperature gradient measured in the trunk (5°C) and in eye region (15°C) in both hummingbirds [9] and pigeons [33]. In addition to the temperature gradient, the effectiveness with which heat is lost during flight was determined by the airflow over the bird during flight, which results in heat dissipation by forced convection. In the absence of airflow when the bird is perched, heat lost by free convection represents a mean of 12% of the total heat lost in pre- and post-flight (figure 3a), whereas during flight heat lost by forced convection rose to a mean of 83% of the total heat lost (figure 3a).

In post-flight, after the bird had landed, we measured a decrease in non-evaporative heat dissipation to a level that was 19× lower than during flight (figure 3). We also found that T_s was related to the preceding flying condition with a higher T_s after longer flights and at low and high flight speeds (table 2), where the mechanical power requirements are expected to be higher ([34]; e.g. in budgerigar [7,29]; e.g. in cockatiel [16,17]). This probably indicates that some heat was being stored early on in flight, with a consequent rise in body temperature. As a result, there was also a higher heat loss while perching after longer flights in the head, trunk and legs areas and at low and high flight speeds by the legs (table 4). This relationship supports the idea that non-evaporative heat dissipation was adjusted as a function of flight metabolic rate (e.g. flight speed) and was dependent on the body areas (e.g. higher in the unfeathered legs compared with the feathered areas of the trunk). Moreover, even with a small range of T_a variation (between 23 and 24.5°C), we measured a significant effect of T_a on T_s, suggesting that environmental conditions could impact the thermoregulatory response and homeostasis. Supporting this suggestion are measurements made on wild birds. For example, at higher air temperatures pink-footed geese (Anser brachyrhynchus; [35]) are less likely to fly, godwits (Limosa limosa; [36]), great reed warblers (Acrocephalus arundinaceus; [37]) and great snipe (Gallinago media; [38]) fly at higher altitudes and hyperthermia in common eiders (Somateria mollissima), has been suggested to explain their use of stopovers [39].

Excess post-exercise oxygen consumption in humans following cycling increases with exercise duration and intensity [40]. The mechanisms underlying the increased heat dissipation and oxygen consumption following activity are multifactorial, and include increased ventilation, circulation and muscle temperature, as well as replenishment of muscle and liver glycogen stores, resynthesis of creatine phosphate and ATP [40]. Therefore, it is not surprising to observe that post-flight heat dissipation was higher than before the flight, though heat dissipation remained at a lower rate post-flight than during flight (figure 2d–f), primarly due to the loss of forced convective cooling via the extended wings. Surprisingly, we found that the body surface temperature was lower immediately after flight compared with immediately before flight (figure 2a–c and see head area on the electronic supplementary material, figure S1). Moreover, our results also show a switch in the importance of the main HDAs in post-flight compared with those involved during flight. These changes are related to differences in the bird's posture and activity. During flight, the proximal wing, which was exposed when the wing was extended during flight, dissipated heat at a rate 11× higher than that lost by the head, 19× higher than the heat lost by the trunk and 48× higher than the heat lost by the legs, representing 86% of the total non-evaporative heat dissipation (figure 3c). Whereas when the birds were perched with their wings folded, the trunk and head assumed the major sites of heat dissipation, accounting for 91% in pre-flight and 86% in post-flight of the total non-evaporative heat dissipation (figure 3c). However, it is interesting to note that the increase in T_trunk of +3.58°C in post-flight compared with that in flight could be mainly due to a higher T_s in the shoulder area as observed in post-flight (figure 1).

The similarity in T_legs between pre-flight and flight, despite the increase in airflow over the legs during flight, indicates that peripheral blood flow to the legs is higher during flight compared with pre-flight. The increase in T_legs post-flight (32.28°C), compared with pre-flight (29.69°C; figures 1, 2c and 4c), suggests that peripheral blood flow to the legs is higher during post-flight compared with pre-flight. In herring gulls, it has been demonstrated that blood flow to the legs increases during flight compared with at rest, and following sustained flights the feet were warm to touch [28]. Blood flowing to the legs may also be at a higher temperature during post-flight compared with pre-flight as the heat production during activity probably induces an increase in core temperature during flight compared with at rest [28]. However, the relative increase in flow is actually slightly lower than the relative increase in metabolic rate (4.5× increase in blood flow compared with 7.5× increase in metabolic rate in flight compared with rest at 20°C [28]), which is consistent with the majority of the increased cardiac output during exercise being directed to the locomotory muscles [41]. Increasing peripheral blood flow post-flight and during flight is likely to be a mechanism to increase heat dissipation by the legs (figure 3c). In herring gulls heat dissipation by the legs accounts for 37–52% of the heat dissipation at rest. This is considerably higher than the proportion of heat dissipated by the legs in lovebirds (9% pre-flight and 13.7% post-flight), which is likely to be due to the larger relative surface area of herring gull's webbed feet compared with the unwebbed feet of lovebirds. The increase in T_legs post-flight compared with during flight is likely to reflect the more effective heat dissipation that occurs during flight (lowering T_legs), largely as a result of forced convection during flight. It is possible that the increased blood flow to the legs during flight continues into post-flight, resulting in the rapid increase in T_legs, since airflow and forced convective heat loss over the legs has ceased. Moreover, the large range of leg temperature measured in post-flight (figure 2c) may suggest a difference in the need to dissipate heat between birds that could be related to the core temperature attained during exercise. After flight when birds were perched and in the absence of airflow, we measured a mean increase in all T_s with +6.13°C in head, +3.59°C in trunk and +6.29°C in leg areas (figures 2a–c and 4) compared with in flight. The highest increase measured in the legs confirm that legs play an important role in avian thermoregulation [42] and we measured that independently to the surface area, the legs in post-flight showed a rate of dissipation (0.03 W cm⁻²) 10x higher than the feathered trunk (0.003 W cm⁻²) highlighting the advantage of exposing this area for the benefit of heat exchange. However, the initial increase in non-evaporative heat dissipation immediately upon landing followed by a decrease initiated after approximately 150 s (figure 5), suggests that muscle/body temperatures, metabolite levels and fuel reserves are returning to pre-exercise levels and the need to dissipate heat is diminishing—i.e. excess post-exercise oxygen consumption is reducing. Although T_s increased post-flight compared with flight in the trunk, head and legs, non-evaporative heat dissipation in all regions was lower upon landing (figure 4) due to the absence of airflow and the resulting large decrease in convective heat loss (figure 3a).

Comparing feathered and partially feathered areas, we found that the head (including eye region (i.e. periorbital region) and bill hot spots; electronic supplementary material, figure S2A), dissipated heat at 1.75× faster rate than trunk in pre-flight and at 3× faster rate in post-flight (figure 3c). Hence the lovebird's bill, which accounted for only 2.1% of the whole-body surface area, played an important role dissipating 8.4% of the total head heat loss in post-flight (electronic supplementary material, figure S2) and represents 3.4% of the total non-evaporative heat dissipation in post-flight; this is a smaller relative heat loss than has been measured in any other species to date. For example, in toco toucans, the bill accounted for 30–50% of the whole-body surface area and was responsible for 35–60% [12] of the total heat dissipation; in tufted puffin, the bill represented 6% of the total surface area and accounted for 18% of the total non-evaporative heat loss [13]. However, the increase in T_head, driven by the increase in bill and eye T_s in lovebirds (electronic supplementary material, figure S2B), contrasts with a global decrease of bill T_s measured within the first 35 min after landing in tufted puffin [13]. In addition to interspecific variation in heat dissipation mechanisms, differences in the environmental conditions and the time at which the measurements were made could also explain the discrepancy between the studies. In our experimental set-up birds were perched in still air, whereas puffins were studied in their natural habitat on the cliffs of Alaska, where it is likely that the birds were still subjected to air currents and where the air temperature was lower (11.1–13.9°C compared with 23.7°C in this study), which probably accounts for the more rapid cooling of the bill in puffins compared with lovebirds. Moreover, the timescale could explain the two different patterns. In this study, as well as in puffins [13], the relationship between T_s and time in post-flight was analysed using linear regression. However, during post-flight (figure 5) the initial increase in T_s was followed by a tendency for T_s to decrease, though this is only supported by one measure (4.15 min after landing), which does not allow us to run an alternative statistical analysis and was excluded in our previous analysis. More data after a longer recovery period in post-flight would allow us to study the nature of the relationship using binomial regression; we hypothesize that bill T_s would initially show an increase followed by a decrease, the latter being consistent with the low bill temperature measured in puffin 35 min after landing [13].

In this study, we analysed thermal images of the birds in a lateral view in order to standardize the orientation and exposure of the three common HDA during the three periods [22,23]. However, not all regions of the body were visible in this view. For example, the ventral surface of the proximal wing was not exposed in perched birds; therefore, any heat dissipation from this region in pre- and post-flight was ignored. We noted that in post-flight some birds show behaviour such as, holding their wings away from their trunk (i.e. wing drooping), which may facilitate heat dissipation; however, we were unable to quantify the heat loss resulting from these postural adjustments in post-flight as the T_s of this region could not be determined in a lateral view. In addition, we observed that birds sometimes panted and used gular fluttering post-flight, behaviours that increase heat loss by evaporative cooling in the buccal cavity [43]. In the same way, we were unable to quantify dangling feet during flight as has been observed in hummingbirds and herring gulls [9,28,44] due to the small set of thermal images in which this behaviour was observed, but we clearly observed a larger part of the legs (i.e. tarsus) exposed in perched birds immediately after the end of flight than later and especially after long flights.

5. Concluding remarks

Understanding the mechanisms through which birds dissipate heat both during and after flight is important in enabling biologists to predict how birds may be physiologically challenged as a result of global warming, which may affect their behaviour and their survival. For example, with an increase in atmospheric temperatures, longer migratory flights could be limited by a bird's heat dissipation capacity and would require longer or more frequent stopovers, flying at higher altitudes [37,38] or changing their migration routes. In future work, using the methodological framework that we have developed in this study, investigating how birds thermoregulate across a range of ambient conditions could prove to be insightful in assessing the impact of climate change on this ecologically important group of animals.

Ethics

Experiments were carried out in accordance with the legal requirements of the Animals (Scientific Procedures) Act 1986 and were conducted under a project license (grant no. PP3300971) issued by the UK Home Office, and were approved by the University of Leeds Animal Welfare and Ethical Review Committee.

Data accessibility

The raw data are available from the Research Data Leeds Repository: https://archive.researchdata.leeds.ac.uk/1192/ [45].

Supplementary material is available online [46].

Declaration of AI use

We have not used AI-assisted technologies in creating this article.

Authors' contributions

A.L.: conceptualization, data curation, formal analysis, investigation, methodology, resources, software, validation, visualization, writing—original draft, writing—review and editing; C.M.B.: funding acquisition, writing—review and editing; G.N.A.: conceptualization, data curation, investigation, methodology, funding acquisition, project administration, resources, supervision, validation, writing—review and editing.

All authors gave final approval for publication and agreed to be held accountable for the work performed therein.

Conflict of interest declaration

We declare we have no competing interests.

Funding

This work was supported by the Biotechnology and Biological Sciences Research Council (grant nos. BB/P020933/1 and BB/P020461/1 to the authors (G.N.A. and C.M.B.)).