<<

1 Thermoregulatory windows in Darwin’s Finches

2

3

4 Glenn J. Tattersall*a, Jaime A. Chavesb,c, Raymond M. Dannerd

5

6 aDepartment of Biological Sciences, Brock University, St. Catharines, ON, L2S3A1, Canada

7 bUniversidad San Francisco de Quito, Colegio de Ciencias Biológicas y Ambientales,

8 Extensión Galápagos, Campus Cumbayá, Quito, Ecuador

9 cGalápagos Science Center, Universidad San Francisco de Quito and The University of North

10 Carolina at Chapel Hill, San Cristóbal Island, Galápagos, Ecuador

11 dDepartment of Biology and Marine Biology, University of North Carolina Wilmington, 601

12 S. College Rd, Wilmington, NC, USA 28403

13 *Corresponding author: [email protected]

14

15 Accepted: August 31, 2017

16 Journal: Functional Ecology

17 Article ID: FEC12990

18 Article DOI: 10.1111/1365-2435.12990

19 Internal Article ID: 14640567

20

21

1

22 Summary

23 1. Darwin’s finches have been the focus of intense study demonstrating how climatic

24 fluctuations coupled with resource competition drive the of a variety of bill

25 sizes and shapes. The bill, as other peripheral surfaces, also plays an important role in

26 thermoregulation in numerous . The avian bill is vascularized, while limbs

27 have specialized vasculature that facilitate heat loss or heat conservation (i.e., they are

28 thermoregulatory windows).

29 2. The Galápagos Islands, home to Darwin’s finches, have a hot and relatively dry climate

30 for approximately half of the year, during which thermoregulatory windows (i.e.

31 surfaces) could be important for thermoregulation and the linked challenge of water

32 balance.

33 3. We hypothesized that Darwin’s finch bills have evolved in part for their role in

34 thermoregulation, possibly co-opted, following adaptation for other functions, such as

35 foraging. We predicted that bills of Darwin’s finches are effective thermoregulatory

36 windows, and that species differences in bill morphology, along with physiology and

37 behavior, lead to differences in thermoregulatory function.

38 4. To test these hypotheses, we conducted a field study to assess heat exchange and

39 microclimate use in three ground finch species and sympatric cactus finch (Geospiza

40 spp.). We collected thermal images of free-living during a hot and dry season and

41 recorded microclimate data for each observation. We used individual thermographic

42 data to model the contribution of bills, legs, and bodies to overall heat balance and

43 compared surface temperatures to those from dead birds to test physiological control

44 of heat loss from these surfaces. We derived and compared species-specific threshold

2

45 environmental temperatures, which are indicative of a species’ thermally neutral

46 temperature.

47 5. In all species, the bill surface was an effective heat dissipater during naturally

48 occurring warm temperatures. As expected, we found that finches controlled surface

49 temperatures through physiology and that young birds had higher surface

50 temperatures than adults. Larger bills contributed proportionally more to overall heat

51 loss than smaller bills.

52 6. We demonstrate here that related, sympatric species with different bill sizes exhibit

53 different patterns in the use of these thermoregulatory structures, supporting a role for

54 thermoregulation in the evolution and ecology of Darwin’s finch morphology.

55

56 Key-words: bill size, heat exchange, heat stress, operative temperature, thermography,

57 thermal niche, thermoneutrality, critical temperature

58

3

59 Introduction

60 The avian bill is the archetypal example for how evolution shapes morphology in

61 response to changing environments (Symonds & Tattersall 2010; Tattersall, Arnaout &

62 Symonds 2016). Variation within and between bird species’ bills have been interpreted in

63 light of differences in foraging behavior and diet, and studies of the avian bill provide some

64 of the strongest evidence of the effects of limiting food supply and competition on a

65 morphological feature (Bowman 1961; Smith 1990; Remsen 1991; Benkman 1993; Matessi,

66 Griggio & Pilastro 2002; Herrel et al. 2005; Badyaev et al. 2008). Changes in average bill

67 size have also been associated with climate-induced changes in resource availability (Boag

68 & Grant 1981), and the addition and subtraction of potentially competitive species (Grant &

69 Grant 2006).

70 Avian bills also play an important role in body temperature (Tb) regulation

71 (Tattersall, Arnaout & Symonds 2016), and thus, energetics. Birds make use of their

72 exposed bills and limbs to dump body heat as their local ambient temperature approaches

73 Tb (Martineau & Larochelle 1988; Maloney & Dawson 1994; Wilson, Adelung & Latorre

74 1998; Tattersall, Andrade & Abe 2009; Greenberg et al. 2012); that is, these structures are

75 thermal windows. Endothermic adjust peripheral blood flow to their uninsulated

76 appendages in response to heat stress (Hill & Veghte 1976; Hagan & Heath 1980; Hill,

77 Christian & Veghte 1980; Buchholz 1996); as blood flow increases to appendages, the

78 transfer of internal body heat to the periphery increases, facilitating greater heat loss to a

79 cooler environment. The ramphotheca (horny bill covering) is vascularized (Midtgård

80 1980; Midtgård 1984; Van Hemert et al. 2012), and heat loss from bills can be substantial

81 (Hagan & Heath 1980; Scott et al. 2008; Tattersall, Andrade & Abe 2009). In the toco

4

82 toucan, the bill can act as an adjustable thermal radiator of up to 400% of resting heat

83 production (Tattersall, Andrade & Abe 2009). The bill of song sparrows can account for 6–

84 8% of total body heat loss, and a subspecies that lives in hot and water limited sand dunes

85 has a larger bill that loses over 30% more heat than a subspecies found in mesic

86 environments. Furthermore, bill size correlates with summer temperatures, both within

87 (Greenberg & Danner 2012) and between species of sparrows (Greenberg et al. 2011),

88 suggesting that larger bill size may be selected to better dissipate heat in warm, water-

89 limited environments.

90 If appendages are important for heat and water balance, but critical for foraging and

91 resource acquisition, then these traits should be prone to evolutionary trade-offs or

92 reinforcement (Tattersall, Arnaout & Symonds 2016). To examine the interactions between

93 these selection pressures on morphological trait evolution, we chose a study system where

94 the variability in bill size is known to be critical to the acquisition of resources, but where

95 heat stress and water limitation are likely operating. Darwin’s finches are ideal for

96 comparative studies because they include several closely related, incipient species, with

97 vastly different bill sizes, and which have radiated to fill different niches. Occupancy of

98 different niches allowed us to compare the thermal biology of ecologically differentiated

99 taxa. The Galápagos Islands are characterized by variable and seasonal rainfall which

100 influence seed availability, but also introduce periodic heat and water stress to the

101 inhabitants (Grant 1986). Although the maximum temperatures in the Galápagos rarely

102 rise above 32–35°C (Trueman & d’Ozouville 2010), solar radiation is high, which can cause

103 high heat loads (Wolf & Walsberg 1996b). In water limited environments like the lowland

104 regions of the Galápagos (Fig. S1), costly evaporative mechanisms may drive selection not

5

105 only on morphology, but also behavioral efficiency or physiological economy; verdins, for

106 example, conserve up to 75% of their evaporative water loss by avoiding sun and wind, and

107 thereby exhibit microsite selection to achieve favorable operative temperatures (Wolf &

108 Walsberg 1996b).

109 Operative temperature (Te) is often used to describe the thermal environment

110 because it represents an ’s net heat transfer potential at the body surface, which is

111 the result of several factors in ecological settings (Porter and Gates 1969, Bakken and Gates

112 1975). Typically, Te is measured using artificial or specimen mounts of similar size and

113 physical dimensions to the live animal, although Te can be estimated from physical

114 equations that model heat transfer (Dzialowski 2005; Angilletta 2009). Formally, Te is the

115 realized temperature that produces the same level of heat or cold stress that the combined

116 effects of a specific air temperature, wind speed (which causes convection), and solar heat

117 load would produce (Angilletta 2009). For example, a low Te is associated with higher

118 wind speeds and lower solar heat loads, whereas a high Te is associated with low wind

119 speeds and high solar heat loads. Since the interface of heat transfer between the animal

120 and the environment is the body surface, whose temperature represents an averaging of

121 the core body temperature, combined with external heat loads, high Te values lead to high

122 body surface temperatures (see Fig. 1 for variation in finch body surface temperatures).

123 Birds therefore must balance their heat loads with changes in internal heat production,

124 evaporative heat dissipation, and cardiovascular changes to bring warm, core blood to the

125 surface. Although avian body temperatures typically range from 40 to 45 °C, they are also

126 labile in response to heat loading (Whitfield et al. 2015; McKechnie et al. 2016). Finally,

127 insulating surfaces like the serve to insulate the body from losing heat in the cold,

6

128 as well as to mitigate external heat loads being transferred to the skin at high Te. Since

129 plumage covers much of the body surface area, thermal windows such as bills and limbs

130 serve as a means to by-pass the plumage to dump excess heat and thereby reduce the

131 requirement for evaporative heat loss.

132 Assessing the functional role of the body surfaces in dynamic heat exchange is

133 challenging, especially in the field. This challenge can be overcome with infrared thermal

134 imaging, which allows for simultaneous temperature information to be obtained from all

135 surfaces of the body (McCafferty 2013; Tattersall 2016a). In combination with biophysical

136 models that incorporate simultaneous microhabitat parameters, the rate, direction, and

137 spatial contributions of heat transfer from the animal to the environment can be estimated.

138 Species with differently sized bills and limbs may have commensurately different heat

139 exchange patterns, which may in turn determine their thermal niche (Porter & Kearney

140 2009) and how traits related to thermoregulation have evolved. Within this context, we

141 hypothesized that the strikingly divergent bill morphology among Darwin’s finches

142 provides thermoregulatory windows that are of sizes consistent with the thermoregulatory

143 challenges experienced by each species in the wild. Darwin’s finches provide an ideal

144 system for studying hypotheses about trait evolution because they have experienced recent

145 and ongoing evolutionary radiations (Grant, 1986; Chaves et al. 2016), and they live in hot

146 and dry environments. We predicted that: i) bills in Darwin’s finches serve as

147 thermoregulatory windows to facilitate heat loss and gain; and ii) the thermoregulatory

148 function of the bill differs among species. To test these predictions, we measured surface

149 temperatures of free-living Darwin’s finches with non-invasive thermal imaging (Tattersall,

150 Andrade & Abe 2009; McCafferty 2013; Tattersall 2016a) and measured the birds’

7

151 microclimates. First, we measured surface temperatures and estimated the amount of heat

152 lost from bills and other body surfaces and tested if these values differed among species.

153 To test if surface temperatures change because of physiological processes rather than

154 passive heating and cooling from the environment, we measured the differences between

155 surface temperatures in living and untreated, freshly deceased specimens in a range of

156 natural environmental conditions. Finally, we estimated the range of environmental

157 temperatures at which these surfaces lose heat (i.e. are effective thermoregulatory

158 windows) and tested if these values differed among species.

159 Material and Methods

160 Study Species and Location

161 Populations of the sympatric small (Geospiza fuliginosa, Gould 1837), medium (G.

162 fortis, Gould 1837), and large ground finches (G. magnirostris, Gould 1837) as well as the

163 cactus finch (G. scandens, Gould 1837) were studied at two sites on Santa Cruz Island,

164 Galápagos: El Garrapaterro (EG) and the Charles Darwin Research Station (CDRS).

165 Environmental variables

166 Environmental variables were collected during thermal image acquisition (see

167 below). Every bird (i.e., image) observation was accompanied by the following

168 environmental parameters: ambient temperature (Ta, °C), ground temperature (Tg, °C),

169 direct short-wave solar radiation or energy experienced by the bird (SE, W/m2), wind

170 speed (WS, m/s), and relative humidity (RH, %). Ta, WS, and RH were collected using a

171 portable weather device (Kestrel 4000, Kestrel Instruments) at or near (2-5 m) the position

8

172 of the bird, and under similar shade conditions. Tg was obtained from each thermal image

173 (see below). SE was collected in a similar manner using a solar energy meter (DBTU1300

174 digital solar power meter, General Tools and Instruments) pointed vertically toward the

175 sky. RH values were converted to water vapor pressure (WVP, kPa) using standard

176 psychrometric equations (Parish & Putnam 1977), since RH was an inverse, highly

177 correlated function (r ~ -0.9) of Ta.

178 Principal Component Analysis of Environmental Variables

179 A data reduction approach to minimize multicollinearity, especially between Ta and

180 SE, was achieved using Principal Components Analysis (PCA) with the FactoMineR package

181 (Husson et al. 2016). All variables (Ta, Tg, SE, WS, and WVP) were centered and scaled to

182 unit variance prior to the PCA. The predominant axis (PC1) described a measure of heat

183 (temperature, solar energy), while the second axis (PC2) accounted for non-thermal

184 variables that changed throughout the day, namely water vapour and wind speed. With this

185 data reduction approach, PC1 explained >50% of the variation in the environmental

186 variables, allowing for surface temperatures to be assessed primarily using PC1.

187 Collection and Measurement of Thermal Images

188 Our goal was to collect thermal images of all species under the variety of

189 environmental conditions during which they were active. We imaged birds between the

190 hours of 0600 to 1700 over the course of 18 days. Thermal images of birds were collected

191 either by approaching birds just outside of their flight initiation distances (~2-3 m) or by

192 setting up thermal imaging cameras at ground sites where we had previously observed

193 finches to forage (Fig. 1). We have observed birds to forage at the same patch of ground

9

194 (±2m) repeatedly over the course of many days. We were primarily constrained by the

195 requirement to obtain an unfettered view of an individual bird foraging in a readily

196 accessible habitat. As a result, most birds were imaged at openly accessible sites. Sample

197 effort did not differ throughout the active times of day: early morning (0600–0830), mid-

198 morning (0900–1200), and late afternoon (1400–1700) (hours spent per period, mean ±

199 95% CI: 2.1 ± 0.4, F2,6=0.38; P=0.7) and did not differ across days (F15,6=0.38; P=0.94). Birds

200 were less active between approximately noon and 1400 hours. During the sampling

201 periods, we covered each study location broadly and avoided capturing images of the same

202 bird multiple times by recognition of location, species, sex, and age in combination. Of the

203 543 images obtained, 47 (8.7%) were of birds banded in other studies. Of these 47 images,

204 13 were of a previously imaged individual, suggesting that we obtained repeated images

205 (i.e. repeated measures) for 27.7% of the banded population. However, recapture rates are

206 very low at CDRS (Hendry et al. 2009), suggesting that repeated measures for individuals

207 would be rare. The mean (± 95% CI) time interval between known repeats was 18.6 ± 12.7

208 hours, compared to a mean interval of 2.4 ± 0.2 minutes between all successive images,

209 suggesting independence in the physical parameters measured. Therefore, we treated each

210 image as an independent measurement. For each image, we recorded simultaneous

211 measurements of all environmental variables, species, sex (except in immatures), plumage

212 (0-5; ref Bowman 1961) and age (adult vs. immature).

213 Thermal images were captured directly to raw format (FLIR JPG) using two different

214 thermal imaging cameras (FLIR T-300 with telephoto, resolution 320x240, and FLIR SC 660

215 with resolution 640x480, FLIR Systems). To allow for selection of ‘in-focus’ birds in flocks

216 and to minimize disturbance by observers, short videos (<10 seconds long) were collected

10

217 directly to computers at ~10 frames/s using an acquisition program (ExaminIR, FLIR

218 Systems). Images of focal birds (identifiable from field notes) were subsequently extracted

219 from these raw videos. Image analysis followed that described previously (Tattersall,

220 Andrade & Abe 2009; Greenberg et al. 2012; McCafferty et al. 2013; Tattersall 2016a).

221 Emissivity of bird surfaces was assumed to be 0.96 (Tattersall 2016a), atmospheric

222 temperature set to Ta, reflected environmental temperature (required for temperature

223 estimation in thermal imaging software) was assumed to be the average of Tg and Ta, and

224 the object distance set to 3 meters. Using specialized software (ExaminIR, FLIR Systems),

225 regions of interest were digitally drawn to obtain the average surface temperature of the

226 bill (Tbill; the portion of the bill protruding from the face), the tarsi (Tlegs; we chose to

227 analyze the warmest leg, toes were not included because they could not be seen in the

228 images), the body (Tbody; the surface of the largest portion of the body visible in the image),

229 and the cheeks (Tcheeks; a crescent shape, concave side facing up, under the eye). We

230 calculated radiative and convective heat exchange from each surface (W/m2) as previously

231 described (Tattersall, Andrade & Abe 2009); our methods are detailed in the supporting

232 information.

233 To demonstrate evidence for physiological regulation of surface temperatures, we

234 also imaged 3 untreated, recently deceased specimens (one juvenile G. fortis, one female G.

235 fuliginosa, and one mature male G. fortis collected as fresh roadkill under GNP Permit No.

236 PC-05-13) positioned in natural postures, under a range of solar radiation and ambient

237 temperature conditions for comparison to live bird surface temperatures.

238 Statistical Analysis of Surface Temperatures and Heat Exchange in Relation to Environmental

239 Conditions (PC Scores) 11

240 To quantify the temperature of each body surface, and to test for differences among

241 species, we designed and fit linear models. We selected models that best fit the data by

242 using an information-theoretic approach (Burnham & Anderson 2002) based on AICc

243 (Akaike 1973). Candidate linear models included species, environmental variables PC1 and

244 PC2, interactions between species and environmental variables, and additive effects of

245 plumage class, reflective of our hypotheses and biological expectations. We excluded sex

246 because we were unable to identify the sex of all individuals. We chose the most

247 comprehensive model with a DAICc<2 or averaged across those models with DAICc<2. We

248 included null (i.e. intercept-only) models in each set of candidate models. We performed all

249 analyses with R (R Core Team 2016), and used the MuMIn package (Bartoń 2016) for

250 model selection and information-theoretic approach. Residuals were verified for normality

251 and homoscedasticity. We present DAICc values, model weights, effect sizes (r2 or partial

252 eta2), model coefficients (B), or marginal mean values (± model SE or 95% CI where

253 appropriate) as measures of support. We used confidence limits to assess model

254 parameters (Zuur 2009; Bates et al. 2015), and P values (Holm adjusted for multiple

255 comparisons) from targeted post-hoc tests of interaction contrasts using the phia package

256 (Rosario-Martinez 2015).

257 Calculation of Heat Exchange by Species

258 Heat exchange estimates (see supplementary methods for detailed calculations)

259 were conducted according to previous published studies (Tattersall, Andrade & Abe 2009;

260 McCafferty et al. 2013), and using the Thermimage package in R (Tattersall 2016b). We

261 present estimates of heat exchange (loss=negative, gain=positive) that can be attributed to

262 convection and radiation for the whole-body each of species. To assess the proportional

12

263 role of the bill or legs in heat loss and gain, we fit linear mixed models of QBill or QLeg as a

264 function of QTot by species, and used the slope estimates as indicators of the proportion

265 heat exchanged by the appendage.

266 Calculation of Heat Exchange in Relation to Te and Threshold Te

267 For each image, we calculated Te according to Angilletta (2009; details in the

268 supporting information). We estimated the amount of heat lost or gained by each body

269 region in relation to Te by fitting separate linear models for each body region. In each

270 model, the response variable was heat loss from that region (qregion) and the predictor

271 variables included an interaction between Te and species. To determine the temperatures

272 at which thermoregulatory windows are effective at dissipating heat, we calculated

273 threshold environmental temperatures (threshold Te) for each body surface of each

274 species. At threshold Te, heat loss is equal to heat gain (i.e. Qtotal = 0 Watts; see

275 supplementary material for calculation of total heat exchange). Below threshold Te, the

276 surface releases heat to the environment (i.e. “negative heat gain”), and above the

277 threshold Te, the surface gains heat from the environment. We calculated threshold Te

278 based on predicted values of the linear models described above. To visually inspect

279 differences among species, we plotted estimates of heat exchange at three values of Te

280 spanning the ranges of experienced Te values.

281 Results

282 Environmental Variables

13

283 Thermal variables, Ta, Tg, and SE loaded heavily onto PC1 (~58% of variance

284 explained), while WVP was the primary variable loading onto PC2 (~18% of explained

285 variance; Table 1). PC1 was also strongly related to the hour of day, following a quadratic

286 relationship (PC1 = -14.3 + 2.51*Hour – 0.100*Hour2, r2=0.504), peaking at midday.

287 Surface Temperatures in Relation to Environmental Conditions (PC Scores)

288 Surface temperatures of all body regions rose in relation to PC1 (i.e. increasing

289 ambient and ground temperatures, and solar energy), and the slope varied among species

290 (Fig. 2). For each body region, top ranked models included an interaction between PC1 and

291 Species (Tables S2–S5, top model weights ranged from 0.3 to 0.7). Partial eta squared (an r2

292 equivalent) values for PC1 were 0.67, 0.65, 0.81, and 0.61 for Tbill, Tleg, Tbody, and Tcheek,

293 respectively, whereas other parameters (PC2 and interactions) always exhibited partial eta

294 squared values of <0.1, demonstrating that PC1 robustly explained the majority of variance

295 in surface temperatures (Fig. 2). The PC1 by species interactions for all surfaces were

296 driven largely by lower slopes in G. fuliginosa (Tables S2–S5). Specifically, for bill

297 temperatures, the slope for G. fuliginosa was significantly lower than G. magnirostris, G.

298 scandens (post-hoc test P<0.001) and, nearly so with respect to G. fortis (P=0.06). For leg

299 surface temperatures, slopes differed between all species pairs (P<0.01) except for G. fortis

300 vs. G. magnirostris, G. fortis vs. G. scandens, and G. magnirostris vs. G. scandens. For body

301 surface temperatures, slopes differed between G. fuliginosa vs. G. fortis (P=0.029), G.

302 fuliginosa vs. G. magnirostris (P=0.0036) and G. fuliginosa vs. G. scandens (P<0.001). Cheek

303 surface temperature slopes in G. fuliginosa were lower than in G. scandens (P=0.0055),

304 while the remaining Species by PC1 slopes were all non-significant (P>0.05).

14

305 Young birds had higher surface temperatures than adults, as evidenced by plumage

306 scores. Plumage class was observed in the top-ranking models for surface temperatures,

307 and was driven by elevated surface temperatures in young birds with plumage class 1

308 (Tables S2–S5; post-hoc comparisons to class 1 for all surfaces exhibited P<0.001).

309 Demonstration of Physiological Control of Thermoregulatory Windows

310 The relationships between the surface temperatures of the bill, legs, and plumage of

311 the dead birds and PC1 were highly linear (r2>0.9) with slopes that were ~2 times higher

312 than observed in live birds. As a result, the maximum surface temperatures in the

313 specimens for bill surface was 55.2°C, max leg surface: 49.9°C and max body surface:

314 57.3°C, compared to 44, 46, and 53°C in live birds. These results indicate differences in heat

315 transfer between live and dead birds, and are consistent with hypothesized physiological

316 control of thermoregulatory windows in live birds.

317 Heat Exchange by Species

318 QBill as a proportion of QTotal varied considerably among species (i.e. no overlap in

319 95% density limits), from 2.0% in G. fuliginosa, 3.2% in G. fortis, 4.3% in G. magnirostris,

320 and 3.0% in G. scandens (Table 2). QLeg was elevated compared to bills and was similar

321 across all 4 species (~6.2 to 6.9%). Overall, species typically gained radiative heat at a rate

322 of less than 2 Watts, and typically lost convective heat at a rate of less than 2 Watts (Fig.

323 S2). G. magnirostris were unique in never experiencing convective heat gain, and in

324 exhibiting more substantial radiative heat loss.

325 Heat Exchange in Relation to Te and Threshold Te

15

2 326 Heat loss increased linearly with Te (r >0.6) and the slopes differed among species

327 (all linear models with Te by species interactions had P<0.0001). All body surfaces lost

328 heat at lower Tes and gained heat at elevated Tes. Threshold Tes followed a general pattern

329 for each species where the body showed the lowest threshold Te, followed by the legs, and

330 then by the bill (Fig. 3). For the bill, G. scandens stood out with the highest threshold Te

331 (40.6°C), and the three ground finches were very similar (36.6 to 38.8°C). Threshold Te for

332 legs were similar among all species (33.6 to 35.8°C). G. magnirostris showed the lowest

333 threshold Te for the body (29.2°C compared to 31–33°C).

334 For clarity, we present species differences in heat exchange (per m2) at 3 different

335 values of Te: 35°C (near the average Te), 40°C (near Tb), and 45°C (well above most Tes,

336 hypothetically causing thermal stress; Fig. 4). G. fuliginosa exhibited the highest area

337 specific heat exchange in absolute terms, with the bill releasing the most heat at low Te, and

338 the legs and body gaining the most heat at Te=45°C. Heat loss and gain was similar among

339 G. fortis, G. magnirostris, and G. scandens for all body regions. Overall, appendages are more

2 340 efficient (per m ) sources of heat loss at low Te, whereas the body surface is a large

2 341 potential absorber (per m ) of heat at high Te.

342 Discussion

343 Much of our current knowledge of avian thermoregulatory control derives from

344 laboratory studies (Wolf & Walsberg 1996b; Wolf & Walsberg 1996a; Tieleman,

345 Williams & Bloomer 2003; Wiersma et al. 2007; Whitfield et al. 2015), so it is

346 important to ask whether birds show evidence of active heat exchange regulation in

347 the field, since behavioral thermoregulation through microhabitat choice may be less

16

348 costly and more effective (Angilletta 2009). Given previous research, we predicted that

349 bills of Darwin’s finches would be effective thermoregulatory windows and

350 demonstrate evidence of physiological control. Comparing surface temperatures of live

351 birds to recently killed specimens supports our first hypothesis that Darwin’s finches

352 actively regulate surface temperatures, especially in thermoregulatory windows like

353 the bills and limbs (Hagan & Heath 1980; Tattersall, Andrade & Abe 2009; Greenberg et

354 al. 2012). Birds are clearly experiencing high Te values that are reflected in the

355 measurements made on the specimens that could not escape the sun; the surface

356 extremities of live birds at high Te, however, were always substantially cooler than

357 those seen in dead birds. Typically, the leg, cheek, and bill temperatures did not rise

358 much above mid 40s (°C), which is only a few degrees above the expected Tb for birds,

359 whereas the dead birds showed surface temperatures well into the high 50s.

360 Thermoregulatory windows in Darwin’s finches vary in function depending on the

361 body region. Legs and bill surfaces warmed at steep slopes in relation to environmental

362 conditions (i.e. PC1), suggesting that they respond to altered heat loads through

363 physiological control more readily than other regions. In contrast, the shallow slope of

364 cheek temperature in relation to ambient (Fig. 2) suggests less physiological control

365 resulting from continuous rates of blood flow to a surface with low plumage thickness

366 (Klir, Heath & Bennani 1990; Tattersall 2016a), which we expect to be important for

367 supplying muscles of the jaw and face. In other words, highly vascularized surfaces

368 have surface temperatures closer to Tb, regardless of the ambient temperature or heat

369 load (Jessen 2001; Tattersall et al. 2012; Tattersall 2016a). Interestingly, the heavily

370 feathered body surface warmed at a steep slope in relation to PC1, which may result

17

371 from the fact that plumage has a low solar reflectance and thermal inertia, and thus

372 would reach higher temperatures in the sun because little body heat will transfer

373 through the plumage (Wolf & Walsberg 1996b). Indeed, the maximum plumage

374 temperature observed among species was 44–53°C, compared to the lower range of

375 42–46°C in the legs, bill, and cheeks.

376 Young finches had overall warmer surfaces than adults. In particular, male birds

377 with juvenal plumage (plumage score 1) had consistently higher surface temperatures than

378 adults. This may result from active growth at that age, in which appendages require

379 continuous blood flow, thus bringing core body heat to the surface (Tattersall, Andrade &

380 Abe 2009). , as well, are newly matured at that age, and the insulating properties

381 not fully developed (Prum 1999), which may provide less resistance to heat transfer and

382 increase surface temperatures on feathered surfaces.

383 Our second prediction was that species differences in bill morphology, along with

384 physiology and behavior, would lead to differences in thermoregulatory function of the

385 bills. As thermoregulatory windows, Darwin’s finch bills appear to exchange from 2–4.5%

386 of total body heat exchange, although these proportions are underestimates based on the

387 presumed high thermal conductivity of non-insulated regions compared to the plumage

388 (i.e. plumage estimates of heat transfer are over estimated and thus, so are the total heat

389 exchange rates). Overall, appendages were more efficient sources of heat loss at low Te,

390 whereas the plumage (body surface) is a large potential absorber of heat at high Te, with

391 the caveat that plumage thermal resistance mitigates the absorption of heat to the skin. The

392 bill remains an effective heat dissipater at a Te of 35°C, while other regions of the body start

393 to become sources of heat gain. Bill surfaces also gain the smallest amount of heat (per

18

394 surface area) at higher temperatures. These results support the hypothesis that the bill

395 behaves as a thermoregulatory device (Tattersall, Arnaout & Symonds 2016). Relatively

396 large bills did not seem to present greater challenges at low (i.e., 35°C) or high (45°C) Te.

397 That is, the species with the largest bills lose less heat per unit area at low temperatures

398 and gain less heat per unit area at the higher temps. Therefore, overall the proportional

399 amount of heat transferred across the bills is mostly a function of bill surface area relative

400 to body surface area.

401 The small ground finches (G. fuliginosa) stood apart from the remaining species,

402 since their surface temperatures were clearly cooler than expected at higher heat loads,

403 which may be due in part to capitalizing on higher wind speeds. In terms of bill surface

404 temperatures, G. fuliginosa may exhibit higher rates of respiratory water loss, which could

405 cool their non-insulating bill surfaces more effectively. If true, this has implications for

406 water balance, and may be one means by which the small ground finches expand or exploit

407 thermally challenging environments.

408 We subsequently calculated threshold Te values to define species-specific critical

409 temperatures at which the surface switches from net heat loss to heat gain. In principle,

410 threshold Te should conform to a combination of surface area:volume constraints (i.e.

411 shape) and the degree to which the bird exhibits control over blood flow to the surface (i.e.

412 higher threshold = higher thermal tolerance). For the bill, threshold Te was highest in G.

413 scandens, underscoring a wider thermal scope for the bill to lose heat, but within the 3

414 ground finches was nearly identical. Threshold Te for the body much lower, reflective of

415 surface area:volume constraints and ranged from ~29 to 32°C, being lowest for the largest

416 species (G. magnirostris), and highest for the smallest species (G. fuliginosa). These

19

417 threshold values provide insight into how regional heat exchange operates, and suggest

418 there is scope in each species for the recruitment of heat loss from the legs and the bills in

419 nature, since the threshold temperatures for the appendages are much higher than that for

420 the body.

421 That the threshold Te values for the bills were similar among the 3 ground finches,

422 but higher in the cactus finch, suggests that putative selection on bill form and function as a

423 radiator is phylogenetically constrained. Bill and skull shapes are constrained by and

424 strongly coupled to size (Bright et al. 2016), and thus, novelty in function would be

425 required to break this genetic “lock”. This likely applies to our study group where the three

426 sister species of ground finches in Santa Cruz show a high correlation between body and

427 size, genetically controlled by a few similar loci (Chaves et al. 2016). It is plausible

428 that the different foraging behavior of the cactus finch has “unlocked” regulatory control

429 over bill length and extended potential thermoregulatory function of the bill. How do we

430 interpret these results with respect to activity or habitat use? The cactus finch for example,

431 appears truly more tolerant of heat given their higher threshold Te. These observations are

432 consistent with G. scandens’ lower dependence on rainfall (Boag & Grant 1984) and

433 tendency to forage in the open, suggesting that different physiological tolerances to heat or

434 water stress have evolved. Given their exploitation of a liquid-rich food resource (e.g.

435 nectar), it is expected that they would not be as water limited, and may be able to forage

436 under higher heat loads while relying on evaporative cooling. The large ground finch,

437 which has a lower threshold body Te may have a greater ability to dominate water

438 resources that would allow this species to forage in the heat using evaporative heat

20

439 dissipation, or have altered foraging behavior (Schluter 1982) that minimizes time spent at

440 high heat loads.

441 Conclusions

442 Due to their recent and ongoing radiation, Darwin’s finches are an exemplary study

443 system for exploring the role of morphological trait evolution on physiological function. We

444 demonstrate here that closely related species that live in sympatry exhibit different

445 patterns in the use of bills as thermoregulatory structures, supporting the growing

446 evidence that avian bills contribute significantly to the evolution of thermal balance

447 (Tattersall, Andrade & Abe 2009; Symonds & Tattersall 2010; Greenberg et al. 2012;

448 Tattersall, Arnaout & Symonds 2016; van de Ven et al. 2016). We also demonstrate how a

449 threshold temperature for each species can be estimated using steady state modelling of

450 heat exchange. The threshold Te would relate to critical temperatures commonly associated

451 with physiological thresholds measured in laboratory studies, while benefiting from being

452 associated with natural microhabitat selection. In spite of their different sizes, the three

453 ground finches exhibited similar critical temperatures for bill radiator function suggesting

454 that vascular control within the bills is stabilized to match the optimal temperature of birds

455 under their natural environmental conditions. These results are consistent with the

456 evolution of Darwin’s finch bills for thermoregulation, possibly through co-option

457 (Tattersall, Arnaout & Symonds, 2016) following divergent selection for foraging.

458

21

459 Authors Contributions

460 GJT, RMD, and JAC conceived the ideas and designed the methodology. GJT, RMD, and JAC

461 collected the data. GJT and RMD analyzed the thermal images. GJT conducted the

462 statistical analysis. GJT and RMD led the writing of the manuscript. All authors contributed

463 critically to the drafts and gave final approval for publication.

464

465 Acknowledgments

466 We would like to thank and acknowledge Dr. Russell Greenberg for initiating and

467 facilitating many of the ideas presented in this study, and who sadly passed away before

468 the manuscript was written. Research funding for this study was kindly provided by the

469 National Geographic Society, the Smithsonian Migratory Bird Center, the Galápagos

470 Institute for the Arts and Sciences-Universidad San Francisco de Quito Grant, and the

471 Natural Sciences and Engineering Research Council of Canada (RGPIN-2014-05814).

472 Logistical support was kindly provided by the Charles Darwin Research Station and the

473 Galápagos National Park. Permits to conduct research were provided by the Galápagos

474 National Park Service (Authorization No. PC-05-13). Ethical oversight and approval for the

475 fieldwork was provided by the Smithsonian’s National Zoological Park IACUC (ACUC No.

476 NZP 13-04).

477

478 Data Accessibility

479 Data will be made available on Data Dryad: Dryad entry doi:10.5061/dryad.t4k41

22

480 References

481 Akaike, H. (1973) Information theory as an extension of the maximum likelihood principle. 482 Second International Symposium on Information Theory (eds B.N. Petrov & F. Csaki), 483 pp. 267-281. Akademiai Kiado, Budapest.

484 Angilletta, M.J. (2009) Thermal adaptation: A theoretical and empirical synthesis. Oxford 485 University Press, Oxford, UK ; New York.

486 Badyaev, A.V., Young, R.L., Oh, K.P. & Addison, C. (2008) Evolution on a local scale: 487 developmental, functional, and genetic bases of divergence in bill form and associated 488 changes in song structure between adjacent habitats. Evolution, 62, 1951-1964.

489 Bakken G.S. and D.M. Gates. 1975. Heat transfer analysis of animals: some implications on 490 field ecology, physiology and evolution. Pp. 255–290 in D.M. Gates and R.B. Schmerl, 491 eds. Perspectives of Biophysical Ecology, Springer, New York. 492 493 Bartoń, K. (2016) MuMIn: Multi-Model Inference. R package version 1.15.6.

494 Bates, D., Maechler, M., Bolker, B. & Walker, S. (2015) Fitting linear mixed-effects models 495 using lme4. Journal of Statistical Software, 67, 1-48.

496 Benkman, C.W. (1993) Adaptation to Single Resources and the Evolution of Crossbill (Loxia) 497 Diversity. Ecological Monographs, 63, 305-325.

498 Boag, P.T. & Grant, P.R. (1981) Intense Natural Selection in a Population of Darwin's Finches 499 (Geospizinae) in the Galapagos. Science, 214, 82-85.

500 Boag, P.T. & Grant, P.R. (1984) Darwin Finches (Geospiza) on Isla Daphne Major, Galapagos - 501 Breeding and Feeding Ecology in a Climatically Variable Environment. Ecological 502 Monographs, 54, 463-489.

503 Bowman, R.I. (1961) Morphological differentiation and adaptation in the Galápagos finches. 504 University of California Publications in Zoology., 58, 1-302.

505 Bright, J.A., Marugan-Lobon, J., Cobbe, S.N. & Rayfield, E.J. (2016) The shapes of bird 506 are highly controlled by nondietary factors. Proceedings of the National Academy of 507 Sciences of the United States of America, 113, 5352-5357.

508 Buchholz, R. (1996) Thermoregulatory role of the unfeathered head and neck in male wild 509 turkeys. Auk, 113, 310-318.

510 Burnham, K.P. & Anderson, D.R. (2002) Model selection and multi-model inference: a practical 511 information-theoretic approach. Springer, New York, NY.

23

512 Chaves, J.A., Cooper, E.A., Hendry, A.P., Podos, J., De León, L.F., Raeymaekers, J.A.M., 513 MacMillan, W.O. & Uy, J.A.C. (2016) Genomic variation at the tips of the adaptive 514 radiation of Darwin's finches. Molecular Ecology, 25, 5282-5295.

515 Dzialowski, E.M. (2005) Use of operative temperature and standard operative temperature 516 models in thermal biology. Journal of Thermal Biology, 30, 317-334.

517 Grant, P.R. (1986) Ecology and evolution of Darwin’s finches. Princeton University Press, 518 Princeton, NJ.

519 Grant, P.R. & Grant, B.R. (2006) Evolution of character displacement in Darwin's finches. 520 Science, 313, 224-226.

521 Greenberg, R., Cadena, V., Danner, R.M. & Tattersall, G. (2012) Heat loss may explain bill size 522 differences between birds occupying different habitats. PLoS ONE, 7, e40933.

523 Greenberg, R., Danner, R., Olsen, B. & Luther, D. (2011) High summer temperature explains bill 524 size variation in salt marsh sparrows. Ecography, 35, 146-152.

525 Greenberg, R. & Danner, R.M. (2012) The influence of the California marine layer on bill size in 526 a generalist songbird. Evolution, 66, 3825-3835.

527 Hagan, A.A. & Heath, J.E. (1980) Regulation of heat loss in the by vasomotion in the bill. 528 Journal of Thermal Biology, 5, 95-101.

529 Hendry, A.P., Huber, S.K., De Leon, L.F., Herrel, A. & Podos, J. (2009) Disruptive selection in a 530 bimodal population of Darwin's finches. Proceedings of the Royal Society B-Biological 531 Sciences, 276, 753-759.

532 Herrel, A., Podos, J., Huber, S.K. & Hendry, A.P. (2005) Bite performance and morphology in a 533 population of Darwin's finches: implications for the evolution of beak shape. Functional 534 Ecology, 19, 43-48.

535 Hill, R.W., Christian, D.P. & Veghte, J.H. (1980) Pinna temperature in exercising jackrabbits, 536 Lepus californicus. Journal of Mammalogy, 61, 30-38.

537 Hill, R.W. & Veghte, J.H. (1976) Jackrabbit ears - surface temperatures and vascular responses. 538 Science, 194, 436-438.

539 Husson, F., Josse, J., Le, S. & Mazet, J. (2016) FactoMineR: Multivariate Exploratory Data 540 Analysis and Data Mining. R Package version 1.32.

541 Jessen, C. (2001) Temperature Regulation in Humans and Other Animals. Springer-Verlag, 542 Berlin, Germany.

543 Klir, J.J., Heath, J.E. & Bennani, N. (1990) An infrared thermographic study of surface 544 temperature in relation to external thermal stress in the Mongolian gerbil, Meriones 545 unguiculatus. Comparative Biochemistry and Physiology A, 96, 141-146.

24

546 Maloney, S.K. & Dawson, T.J. (1994) Thermoregulation in a large bird, the Emu (Dromaius 547 novaehollandiae). Journal of Comparative Physiology B-Biochemical Systemic and 548 Environmental Physiology, 164, 464-472.

549 Martineau, L. & Larochelle, J. (1988) The cooling power of pigeon legs. Journal of 550 Experimental Biology, 136, 193-208.

551 Matessi, G., Griggio, M. & Pilastro, A. (2002) The geographical distribution of populations of 552 the large-billed subspecies of reed bunting matches that of its main winter food. 553 Biological Journal of the Linnean Society, 75, 21-26.

554 McCafferty, D.J. (2013) Applications of thermal imaging in avian science. Ibis, 155, 4-15.

555 McCafferty, D.J., Gilbert, C., Thierry, A.M., Currie, J., Le Maho, Y. & Ancel, A. (2013) 556 Emperor penguin body surfaces cool below air temperature. Biology Letters, 9.

557 McKechnie, A.E., Smit, B., Whitfield, M.C., Noakes, M.J., Talbot, W.A., Garcia, M., Gerson, 558 A.R. & Wolf, B.O. (2016) Avian thermoregulation in the heat: evaporative cooling 559 capacity in an archetypal specialist, Burchell's ( burchelli). 560 Journal of Experimental Biology, 219, 2137-2144.

561 Midtgård, U. (1980) Arteriovenous Anastomoses and Vascularity in the Feet of Eiders and Gulls 562 (Aves). Zoomorphology, 96, 263-270.

563 Midtgård, U. (1984) The blood vascular system in the head of the herring gull (Larus 564 argentatus). Journal of Morphology, 179, 135-152.

565 Parish, O.O. & Putnam, T.W. (1977) Equations for the determination of humidity from dewpoint 566 and psychrometric data. NASA Technical Note, D-8401, 1-23.

567 Porter, W. P., & Gates, D. M. 1969. Thermodynamic equilibria of animals with 568 environment. Ecological monographs, 39(3), 227–244.

569 Porter, W.P. & Kearney, M. (2009) Size, shape, and the thermal niche of endotherms. 570 Proceedings of the National Academy of Sciences of the United States of America, 106, 571 19666-19672.

572 Prum, R.O. (1999) Development and evolutionary origin of feathers. Journal of Experimental 573 Zoology, 285, 291-306.

574 R Core Team (2016) R: A language and environment for statistical computing. R Foundation for 575 Statistical Computing, Vienna, Austria.

576 Remsen, J.V.J. (1991) Community ecology of Neotropical kingfishers. University of California 577 Publications in Zoology, Berkeley and Los Angeles.

578 Rosario-Martinez, H. (2015) phia: Post-Hoc Interaction Analysis. R package version 0.2-1. 579 CRAN, https://cran.r-project.org/package=phia.

25

580 Schluter, D. (1982) Seed and patch selection by Galapagos ground finches - relation to foraging 581 efficiency and food-supply. Ecology, 63, 1106-1120.

582 Scott, G.R., Cadena, V., Tattersall, G.J. & Milsom, W.K. (2008) Body temperature depression 583 and peripheral heat loss accompany the metabolic and ventilatory responses to hypoxia in 584 low and high altitude birds. Journal of Experimental Biology, 211, 1326-1335.

585 Smith, T.B. (1990) Resource use by bill morphs of an African finch: Evidence for intraspecific 586 competition. Ecology, 71, 1246-1257.

587 Symonds, M.R.E. & Tattersall, G.J. (2010) Geographical variation in bill size across bird species 588 provides evidence for Allen's rule. American Naturalist, 176, 188-197.

589 Tattersall, G., Sinclair, B., Withers, P., Fields, P., Seebacher, F., Cooper, C. & Maloney, S. 590 (2012) Coping with thermal challenges: Physiological adaptations to environmental 591 temperatures. Comprehensive Physiology, 2, 2151-2202.

592 Tattersall, G.J. (2016a) Infrared thermography: A non-invasive window into thermal physiology. 593 Comp Biochem Physiol A Mol Integr Physiol, 202, 78-98.

594 Tattersall, G.J. (2016b) Thermimage: Thermal Image Analysis. R package version 2.1 595 http://cran.r-project.org/package=Thermimage.

596 Tattersall, G.J., Andrade, D.V. & Abe, A.S. (2009) Heat exchange from the toucan bill reveals a 597 controllable vascular thermal radiator. Science, 325, 468-470.

598 Tattersall, G.J., Arnaout, B. & Symonds, M.R.E. (2016) The evolution of the avian bill as a 599 thermoregulatory organ. Biological Reviews, doi:10.1111/brv.12299.

600 Tieleman, B.I., Williams, J.B. & Bloomer, P. (2003) Adaptation of metabolism and evaporative 601 water loss along an aridity gradient. Proceedings of the Royal Society of London. Series 602 B: Biological Sciences, 270, 207-214.

603 Trueman, M. & d’Ozouville, N. (2010) Characterizing the Galapagos terrestrial climate in the 604 face of climate change. Galapagos Research, 67, 26-37.

605 van de Ven, T.M.F.N., Martin, R.O., Vink, T.J.F., McKechnie, A.E. & Cunningham, S.J. (2016) 606 Regulation of Heat Exchange across the Hornbill Beak: Functional Similarities with 607 Toucans? PLoS ONE, 11.

608 Van Hemert, C., Handel, C.M., Blake, J.E., Swor, R.M. & O'Hara, T.M. (2012) Microanatomy 609 of passerine hard-cornified tissues: Beak and claw structure of the black-capped 610 chickadee (Poecile atricapillus). Journal of Morphology, 273, 226-240.

611 Whitfield, M.C., Smit, B., McKechnie, A.E. & Wolf, B.O. (2015) Avian thermoregulation in the 612 heat: scaling of heat tolerance and evaporative cooling capacity in three southern African 613 arid-zone passerines. Journal of Experimental Biology, 218, 1705-1714.

26

614 Wiersma, P., Muñoz-Garcia, A., Walker, A. & Williams, J.B. (2007) Tropical birds have a slow 615 pace of life. Proceedings of the National Academy of Sciences, 104, 9340-9345.

616 Wilson, R.P., Adelung, D. & Latorre, L. (1998) Radiative heat loss in gentoo penguin 617 (Pygoscelis papua) adults and chicks and the importance of warm feet. Physiological 618 Zoology, 71, 524-533.

619 Wolf, B.O. & Walsberg, G.E. (1996a) Respiratory and cutaneous evaporative water loss at high 620 environmental temperatures in a small bird. Journal of Experimental Biology, 199, 451- 621 457.

622 Wolf, B.O. & Walsberg, G.E. (1996b) Thermal effects of radiation and wind on a small bird and 623 implications for microsite selection. Ecology, 77, 2228-2236.

624 Zuur, A.F. (2009) Mixed effects models and extensions in ecology with R. Statistics for biology 625 and health, pp. 1 online resource. Springer, New York, NY. 626

27

627 SUPPORTING INFORMATION

628 Additional supporting information may be found in the online version of this article.

629

630 Appendix S1 Supporting Methods

631 Figure S1 Galápagos weather patterns

632 Figure S2 Total heat exchange rates by species

633 Table S1 Morphological parameters

634 Table S2 Statistical tables for bill surface temperatures

635 Table S3 Statistical tables for leg surface temperatures

636 Table S4 Statistical tables for body surface temperatures

637 Table S5 Statistical tables for cheek surface temperatures

28

638 Table 1. Loading scores onto the first 4 dimensions (PC1-PC4) for the principal 639 components analysis of the environmental variables from the field thermography 640 measurements. 641 Parameter PC1 PC2 PC3 PC4

Ta (°C) 0.462 0.3169 -0.4864 0.5115

Tg (°C) 0.542 -0.0816 -0.3058 -0.0725 SE (W/m2) 0.483 -0.3673 -0.0534 -0.6475 WVP (kPa) 0.317 0.7638 0.4978 -0.2598 WS (m/s) 0.399 -0.4179 0.6475 0.4963 Eigenvalue 2.890 0.8822 0.6726 0.4202 Percent Variance 57.793 17.6447 13.4519 8.4031 642

29

643 Table 2. Bill and leg heat exchange as proportions of total heat exchange.

Species Region B SE LDL UDL G. fuliginosa Bill 0.02002 0.0003150 0.01941 0.02064 G. fortis Bill 0.03233 0.0004098 0.03151 0.03310 G. magnirostris Bill 0.04257 0.0005042 0.04162 0.04355 G. scandens Bill 0.03013 0.0004395 0.02930 0.03099 G. fuliginosa Leg 0.06502 0.0005795 0.06389 0.06616 G. fortis Leg 0.06556 0.0007426 0.06414 0.06693 G. magnirostris Leg 0.06235 0.0009146 0.06052 0.06415 G. scandens Leg 0.06938 0.0008182 0.06787 0.07096 644 Parameter estimates (B) ± unconditional standard errors (SE) from the linear models of 645 the regional heat exchange as a function of total heat exchange. The 95% density limits for 646 each parameter are indicated by LDL and UDL.

30

647 Figure Captions

648

649 Figure 1. Sample of thermal images of Darwin’s finches in the field, demonstrating the

650 array of surface temperature responses. Each image has been scaled to different midpoint

651 temperatures but similar ranges (ΔT=20°C) from lowest to highest temperature (color bar

652 on right). Air temperature (Ta) and solar radiation (SE) are noted for each image, along

653 with mean values for leg, bill, and dorsal surface. Depicted are: a) G. fuliginosa in

654 shade; b) G. fortis with one leg vasodilated; c) G. fuliginosa with one leg vasodilated and

655 warm bill; d) G. fortis choosing cooler ground temperature under high SE conditions; e) G.

656 fortis with both legs vasodilated; f) G. fortis landed on hot rock; g) G. magnirostris resting on

657 a branch; h) G. fuliginosa foraging in the shade and avoiding the sun at peak heat; i) G.

658 fuliginosa (drinking from artificial water source) with Pox infection on legs showing

659 intense vasodilation; j) G. scandens in full sun.

660

661 Figure 2. Surface temperatures and model fit (± 95% CI) for four species of Darwin’s

662 finches obtained using infrared thermal imaging in the field. Regions of interest (bill, legs,

663 body, and cheek) were measured using thermal imaging software to estimate surface

664 temperature as a function of PC scores of independently measured environmental

665 variables. PC1 was associated positively with Ta, Tg and SE. Where appropriate, PC2 was

666 set to 0 (mean value) to construct the model fits.

667

668 Figure 3. Threshold Te (model fit ± 95% CI) where bill, legs, and body respectively

669 exchange zero heat with the environment, for all four species of Darwin’s finches examined.

31

670 Below threshold Te, the body loses heat to the environment, and above threshold Te, the

671 body region is a net absorber of heat from solar and ground radiation.

672

673 Figure 4. Area specific heat exchange (W/m2) for three difference regions (bill, legs, and

674 body) in the four Darwin’s finches examined. Negative values for heat exchange represent

675 heat loss and positive values represent heat gain. Values are model fits (± 95% CI)

676 calculated for three different Te values (35, 40, 45°C, listed above each facet) derived from a

677 linear model.

678

679

32

a Ta=26.2 b Ta=25.6 c Ta=26.2 SE=113 SE=84 SE=61 37.7

30.5 33.5 31.7 31.6

36.4 36.5 36.6 Ta=27.2 36.5 e SE=73

37.5 31.7 d g Ta=30.3 SE=270

39.6 38.1 36.6 34.2 T =32.8 36.1 a f SE=390

40.1 40.6 37.1 35.9 Ta=31.3 SE=810 40.9 h j

i 43.2 38.9

37.5 37.3 37.6 31.6

37.9 39.8 40.1 T =28.4 Ta=32.1 Ta=27.1 a 680 SE=1100 (80 in shade) SE=50 SE=450 681

682 Figure 1.

33

Bill Leg

50

40

30

20 Species G. fuliginosa

Body Cheek G. fortis G. magnirostris G. scandens 50 Surface Temperature (°C) Temperature Surface 40

30

20

−3 0 3 6 −3 0 3 6

PC1 (57.6%) [~ Ta, Tg, SE (+ve) ] 683 684 Figure 2.

34

Species G. fuliginosa G. fortis 40 G. magnirostris G. scandens

36

32

28 Threshold Operative Temperature (°C) Temperature Threshold Operative Bill Legs Body

685 686 Figure 3.

35

Te = 35°C Te = 40°C Te = 45°C

+ve = Heat Gain ) 2 200 Species G. fuliginosa 0 G. fortis G. magnirostris G. scandens

−200 Heat Exchange (W/m −ve = Heat Loss Bill Leg Body Bill Leg Body Bill Leg Body

687 688 Figure 4.

36

689 Electronic Supporting Materials

690

691 Thermoregulatory Windows in Darwin’s Finches

692

693 Glenn J. Tattersall*a, Jaime A. Chavesb,c, Raymond M. Dannerd

694

695 aDepartment of Biological Sciences, Brock University, St. Catharines, ON, L2S3A1, Canada

696 bUniversidad San Francisco de Quito, Colegio de Ciencias Biológicas y Ambientales,

697 Extensión Galápagos, Campus Cumbayá, Quito, Ecuador

698 cGalápagos Science Center, Universidad San Francisco de Quito and The University of North

699 Carolina at Chapel Hill, San Cristóbal Island, Galápagos, Ecuador

700 dDepartment of Biology and Marine Biology, University of North Carolina Wilmington, 601

701 S. College Rd, Wilmington, NC, USA 28403

702

703 This document includes:

704 Supporting Methods

705 Supporting Tables

706 Supporting Figures

37

707 Supporting Methods

708 Study Species and Location

709 Populations of the small sympatric (Geospiza fuliginosa), medium (G. fortis), and

710 large ground finches (G. magnirostris) as well as the cactus finch (G. scandens) were studied

711 at two sites on Santa Cruz Island, Galápagos: El Garrapaterro (EG, study site 1) beach (Lat:

712 0° 41’ 39”S, Long: 90° 13’ 18”W) and the Charles Darwin Research Station (CDRS, study site

713 2) reserve area (Lat: 0° 44’ 27”S, Long: 90° 18’ 10”W). Only two species (G. fuliginosa and G.

714 fortis) were observed at the EG site, although all four species were found at the CDRS site.

715 Both sites were chosen for their lowland location (<10 m above sea level), and although

716 long-term climatic records are only available for the CDRS area, both have low annual

717 rainfall (<300 mm/year) in non-El Niño years (Trueman & d’Ozouville 2010), classifying

718 them as an arid tropical environment (Meigs 1952; Food and Organization of

719 the United Nations. 1989). Bird were studied over a period of 5 weeks in April-May of

720 2013, 1-2 months after they had reproduced and young had fledged and during a period of

721 low precipitation. Most birds appeared to have ceased breeding, based on limited singing,

722 presence of fledged, adult-sized juvenile birds, and lack of evident brood patches. No

723 molting was evident during the study.

724 Galápagos Historical Climate and Evapotranspiration Analysis

725 Fifty years (1964-2014) of weather data (monthly mean, min, and max air

726 temperature, mean relative humidity, and total precipitation) from Puerto Ayora,

727 Galápagos were obtained from the Charles Darwin Research Station (Charles Darwin

728 Foundation 2016) and summarized for monthly trends in temperature and precipitation.

38

729 These data were further explored in to estimate seasonal changes in water stress by

730 calculating the potential evapotranspiration (ETo; reference evapotranspiration) according

731 to the Penman Monteith method (Zotarelli et al. 2010) using the SPEI package in R

732 (Beguería & Vicente-Serrano 2013). The ETo provides an estimate for a hypothetical crop

733 of 0.12 m in height and is expressed as the monthly rainfall that would be required to offset

734 offset evapotranspiration. The difference between ETo and actual precipitation is the

735 estimated precipitation deficit due to evaporation and transpiration from the surface.

736 Average wind speed was assumed to be 1.5 m/s and an average daytime solar radiation

737 value of 350 W/m2 (30.24 MJ/m2/day) was assumed to occur for 12 hours a day. This

738 reduced (from maximal possible levels of ~1500 W/m2) level was similar to the average

739 levels witnessed according to bird activity and acted as a conservative estimate to account

740 for cloud cover (estimated at 25%).

741 Heat Exchange Calculations

742 We estimated steady state heat exchange (Q; Watts) across each major body surface

743 (Qbill, Qlegs and Qbody), and total heat exchange (Qtotal) by adding values for all body surfaces.

744 Positive values for Q indicate heat gain, negative values indicate heat loss. Heat exchange

745 was initially assessed as area-specific heat exchange (q; W/m2) for radiation and

746 convection separately for each body region, and total heat exchange determined as the sum

747 of both modes of heat exchange as follows:

748 �, = �, + �, [1]

749 Area-specific radiative heat exchange was assessed as the difference between

750 absorbed radiation (qabs) and emitted radiation as follows:

751 � = � − �� � + 273.15 [2] 39

752 where e is the animal tissue emissivity (assumed to be 0.96), σ is the Stephan-Boltzman

753 constant, and Ts is the surface temperature of the region of interest (bill, leg, body).

754 Absorbed radiation represented the amount of short and long wave radiation predicted as

755 follows:

756 � = 1 − � ∙ �� + � [3]

757 where ρ is surface reflectivity (~0.05, estimated from museum specimens using a

758 reflectance meter, Danner and Tattersall, unpublished), SE is the solar radiation (W/m2) as

759 measured at image capture, αl is the long-wave absorptivity (=e, according to Kirschoff’s

760 law of thermal radiation), and Lu and Ld represent long-wave ground radiation (u=up) and

761 sky radiation (d=down), respectively (units: W/m2). These latter terms were divided by

762 two to obtain the average value for incoming long-wave radiation. Lu was estimated from

4 763 knowledge of ground temperature and Stefan-Boltzman law (eσT ), and Ld was similarly

764 estimated from Ta, RH and fractional cloud cover (n), according to equations for sky

765 emissivity from Konzelmann (1994). Fractional cloud cover (constrained to be from 0 to 1)

766 was estimated from solar radiation (SE; W/m2), time of day (24 hour), and humidity (RH)

767 based on an empirically derived relationship of 100 weather observations spanning a

768 similar time frame to this study (r2=0.74):

769 � = −0.000405 ∗ �� − 0.0797 ∗ �� + 0.0496 ∗ ���� [4]

770 Cloud cover (n) was typically ~0.25 throughout the study. No profile effect (i.e., angle of

771 radiation absorption) was modeled for short-wave absorption; rather we assumed that the

772 measured SE was equivalent to all of the solar radiation arriving at the bird’s surface.

773 Area specific convective heat exchange for each surface region was estimated as:

774 � = ℎ � − � [5]

40

775 where Ts is the surface temperature of the respective body region (°C), Ta is the ambient

776 temperature (°C), and hconv is the convective heat transfer coefficient for that particular

777 body region (W/m2/°C) calculated as:

778 ℎ = �� ∙ [6]

779 where L is the critical dimension of the body region (i.e., height of bill, tarsus length, body

780 diameter, units=m), k is the thermal conductivity of air (W/m/K), determined for each Ta:

781 � = 0.0241 + 7.5907� � [7]

782 and Nu is the Nusselt number determined for each region surface assuming forced

783 convection as:

784 �� = �� [8]

785 using c and n values specific to the object shape and Reynolds number (Blaxter 1989; Gates

786 2003). Re was determined from the following relationship:

787 � = � [9]

788 where V is air velocity (m·s-1), L is the critical dimension (sometimes referred to as D; Gates

2 -1 789 2003), and u is the kinematic viscosity of air (m ·s ), determined from Ta (°C):

790 � = −1.088� + 8.85� � [10]

791 For the assumptions inherent to convective heat exchange estimates certain shape

792 parameters were considered; the bill was modeled as a horizontal cylinder, the legs as a

793 series of vertical cylinders, and the feathered body as a sphere. Body dimensions (bill

794 depth, bill length, bill width, tarsus and toe lengths and widths, body mass) were obtained

795 from birds that were mist-netted as part of ongoing research at the same study site (see

796 Table S1). Area estimates for bills were obtained from mean species’ values using the

41

797 equation for the area of a cone, without the base area, from measurements of depth, length,

798 and width (π·0.5·Depth·Bill hypotenuse length). Area estimates for the legs were obtained

799 assuming the tarsus and toes were a series of connected single base cylinders of separate

800 radii and heights (π·2r·h + π·r2). Body surface areas were estimated from scaling equations

801 using mass to estimate bird external surface area (Walsberg & King 1978). Due to lack of

802 precise morphological data on head sizes, exchange rates were not incorporated into the

803 final heat exchange calculation, although the surface temperatures are reported for

804 comparison.

805 After all region specific values were obtained, multiplying by the species’ mean area

806 estimates provided the regional heat exchange (Watts):

807 � = � ∙ � [11]

808 Total body heat exchange was then the sum of all regional estimates of heat exchange. All

809 of the calculations above were conducted using equations provided in the Thermimage

810 package in R (Tattersall 2016).

811 Environmental Temperature Calculations

812 Environmental Temperature (Te), also known as operative temperature, provides an

813 estimate of the predicted temperature of an object under known thermal, radiative heat

814 loading, and convective heat loss conditions, in the absence of metabolic heat production or

815 evaporative heat loss, and thus more appropriately assesses an animal’s perceived thermal

816 environment. In other words, it is the predicted temperature of an inert object of similar

817 physical dimensions and properties to an animal (Angilletta 2009), and can be calculated

818 as:

42

. 819 � = � + [12] .

2 820 where qabs is the absorbed radiation (W/m ), and hconv is the convective heat coefficient

2 821 (W/m /°C). For the purposes of this study, we use Te as a way to assess species’

822 differences in realized thermal niche as well as to provide an objective measure of the

823 relative degree of thermal load experienced by the birds, normalized to standard physical

824 relationships. To estimate Te, we used whole bird dimensions (primarily height), and

825 where appropriate, modeled hconv assuming the bird approximated a sphere (Gates 2003).

826 We also used the equation for equivalent temperature, Teq (Mahoney & King 1977), which

2 827 led to almost identical values to Te above (Teq = 0.9895 Te; r =0.99), so we report only Te.

828

43

829 References 830 Angilletta, M.J. (2009) Thermal adaptation: A theoretical and empirical synthesis. Oxford 831 University Press, Oxford, UK ; New York.

832 Beguería, S. & Vicente-Serrano, S.M. (2013) SPEI: Calculation of the Standardised 833 Precipitation-Evapotranspiration Index version 1.6 https://cran.r- 834 project.org/package=SPEI.

835 Blaxter, K. (1989) Energy metabolism in animals and man. Cambridge University Press, 836 Cambridge.

837 Charles Darwin Foundation (2016) CDF Meteorological Database - Base de datos meterologico 838 de la FCD. Online data portal - portal de datos en linea: 839 http://www.darwinfoundation.org/datazone/climate/, Access Date: June 1, 2016.

840 Food and Agriculture Organization of the United Nations. (1989) Arid zone forestry : a guide for 841 field technicians. Food and Agriculture Organization of the United Nations, Rome.

842 Gates, D.M. (2003) Biophysical Ecology. Dover Books on Biology Ser, pp. 656. Dover 843 Publications, Incorporated, Mineola.

844 Konzelmann, T., Vandewal, R.S.W., Greuell, W., Bintanja, R., Henneken, E.A.C. & Abe-Ouchi, 845 A. (1994) Parameterization of global and longwave incoming radiation for the Greenland 846 ice-sheet. Global and Planetary Change, 9, 143-164.

847 Mahoney, S.A. & King, J.R. (1977) Use of Equivalent Black-Body Temperature in Thermal 848 Energetics of Small Birds. Journal of Thermal Biology, 2, 115-120.

849 McNab, B.K. (2002) The Physiological Ecology of Vertebrates: A View from Energetics. Cornell 850 University Press.

851 Meigs, P. (1952) World distribution of arid and semi-arid homoclimates. Reviews of research on 852 arid zone hydrology, pp. 203-209. United Nations Educational, Scientific, and Cultural 853 Organization, Paris.

854 Tattersall, G.J. (2016) Thermimage: Thermal Image Analysis. R package version 2.1 855 http://cran.r-project.org/package=Thermimage.

856 Trueman, M. & d’Ozouville, N. (2010) Characterizing the Galapagos terrestrial climate in the 857 face of climate change. Galapagos Research, 67, 26-37.

858 Walsberg, G.E. & King, J.R. (1978) Relationship of external surface area of birds to skin surface 859 area and body mass. Journal of Experimental Biology, 76, 185-189.

860 Zotarelli, L., Dukes, M.D., C. Romero, C.C., Migliaccio, K.W. & Morgan, K.T. (2010) Step by 861 Step Calculation of the Penman-Monteith Evapotranspiration (FAO-56 Method). pp. 1- 862 10. University of Florida, UF/IFAS Extension. 863 44

864 Table S1. Morphological parameters of four species of Darwin’s finches used in heat

865 exchange modeling.

866 Body Body Tarsus Bill Mass Leg SA Bill SA Species Height SA Length Depth (g) (cm2) (cm2) (cm) (cm2) (cm) (cm) G. fuliginosa 14.5 11.4 48.6 1.71 3.51 0.69 0.936 G. fortis 22.6 12.7 65.5 1.87 4.51 1.1 2.05 G. magnirostris 31.0 16.5 80.9 2.08 5.33 1.48 3.45 G. scandens 21.4 14.0 63.2 1.89 4.76 0.85 1.82 867

868 869 870

45

871 Table S2. Comparison of the candidate set of regression models described by the global 872 model: Billmean ~ PC1 * Sp + PC2 * Sp + Plumage * Sp + sinHourRadian + cosHourRadian 873 Model r2 K AICc ΔAICc ω ~PC1*Sp + PC2 + Plumage + 0.700 17 2344 0.000 0.435 cosHourRadian + sinHourRadian + 1 ~PC1*Sp + Plumage + cosHourRadian + 0.699 16 2345 0.374 0.361 sinHourRadian + 1 ~PC1*Sp + PC2*Sp + Plumage + 0.703 20 2346 1.518 0.204 cosHourRadian + sinHourRadian + 1

~1 0 2 2971 626.416 0.000

Parameter B SE Prob Lwr2.5 Upr97.5 (Intercept) 32.986 0.438 0.0000 32.1265 33.8462 cosHourRadian -1.376 0.475 0.0038 -2.3088 -0.4440 PC1 1.000 0.101 0.0000 0.8018 1.1972 PC2 0.200 0.142 0.1587 -0.0783 0.4788 Plumage1 1.415 0.289 0.0000 0.8467 1.9832 Plumage2 -0.247 0.321 0.4431 -0.8778 0.3840 Plumage3 -0.542 0.251 0.0309 -1.0349 -0.0498 Plumage4 -0.653 0.351 0.0636 -1.3430 0.0369 Plumage5 -0.754 0.314 0.0166 -1.3700 -0.1373 sinHourRadian -0.971 0.231 0.0000 -1.4246 -0.5166 Spfort 0.435 0.241 0.0722 -0.0392 0.9085 Spmagn 1.404 0.292 0.0000 0.8303 1.9783 Spscan 1.010 0.266 0.0002 0.4869 1.5336 PC1:Spfort 0.350 0.153 0.0221 0.0504 0.6503 PC1:Spmagn 0.948 0.174 0.0000 0.6057 1.2900 PC1:Spscan 0.747 0.146 0.0000 0.4601 1.0330 PC2:Spfort 0.176 0.283 0.5363 -0.3809 0.7320 PC2:Spmagn -0.343 0.303 0.2586 -0.9392 0.2524 PC2:Spscan -0.374 0.254 0.1419 -0.8737 0.1251 874 875 Parameter estimates (B) ± unconditional standard errors (SE) represent the coefficients 876 from the weighted model averages. The 95% confidence limit for each parameter is 877 indicated by Lwr2.5 and Upr97.5. 878

46

879 Table S3. Comparison of the candidate set of regression models described by the global 880 model: Legmean ~ PC1 * Sp + PC2 * Sp + Plumage * Sp + sinHourRadian + cosHourRadian 881 Model r2 K AICc ΔAICc ω ~PC1*Sp + Plumage + sinHourRadian + 1 0.679 15 2831 0.000 0.3212 ~PC1*Sp + PC2 + Plumage + 0.680 16 2831 0.427 0.2594 sinHourRadian + 1 ~PC1*Sp + Plumage + cosHourRadian + 0.680 16 2831 0.798 0.2155 sinHourRadian + 1 ~PC1*Sp + PC2 + Plumage + 0.681 17 2832 1.350 0.1636 cosHourRadian + sinHourRadian + 1 ~1 0.000 2 3401 570.037 0.0000

Parameter B SE Prob Lwr2.5 Upr97.5 (Intercept) 32.334 0.692 0.0000 30.975 33.694 PC1 1.874 0.168 0.0000 1.545 2.204 Plumage1 1.644 0.514 0.0014 0.635 2.654 Plumage2 -0.319 0.563 0.5718 -1.424 0.786 Plumage3 -0.696 0.438 0.1132 -1.557 0.165 Plumage4 -1.525 0.632 0.0160 -2.766 -0.284 Plumage5 -1.530 0.561 0.0065 -2.631 -0.429 sinHourRadian -1.464 0.418 0.0005 -2.286 -0.643 Spfort 0.519 0.423 0.2205 -0.311 1.350 Spmagn 2.301 0.510 0.0000 1.298 3.303 Spscan 2.169 0.465 0.0000 1.255 3.083 PC1:Spfort 1.046 0.266 0.0001 0.524 1.567 PC1:Spmagn 1.628 0.297 0.0000 1.044 2.212 PC1:Spscan 1.082 0.250 0.0000 0.592 1.573 PC2 0.231 0.182 0.2058 -0.127 0.590 cosHourRadian -0.932 0.836 0.2662 -2.575 0.711 882 883 Parameter estimates (B) ± unconditional standard errors (SE) represent the coefficients 884 from the weighted model averages with z statistics. The 95% confidence limit for each 885 parameter is indicated by Lwr2.5 and Upr97.5. 886

47

887 Table S4. Comparison of the candidate set of regression models described by the global 888 model: Bodymean ~ PC1 * Sp + PC2 * Sp + Plumage * Sp + sinHourRadian + cosHourRadian 889 Model r2 K AICc ΔAICc ω ~PC1*Sp + Plumage + cosHourRadian + 1 0.778 15 2406 0.00 0.5383 ~PC1*Sp + Plumage + cosHourRadian + 0.778 16 2408 1.93 0.2051 sinHourRadian + 1 ~PC1*Sp + PC2 + Plumage + cosHourRadian + 0.778 16 2409 2.12 0.1860 1 ~1 0.000 2 3181 774.88 0.0000

Parameter B SE Prob Lwr2.5 Upr97.5 (Intercept) 30.0267 0.450 0.0000 29.144 30.910 cosHourRadian -2.8360 0.515 0.0000 -3.849 -1.823 PC1 1.6312 0.108 0.0000 1.420 1.842 Plumage1 1.7182 0.334 0.0000 1.063 2.373 Plumage2 -0.0821 0.359 0.8196 -0.788 0.623 Plumage3 -0.7283 0.283 0.0101 -1.283 -0.173 Plumage4 -0.0440 0.400 0.9126 -0.829 0.741 Plumage5 0.0456 0.350 0.8966 -0.641 0.732 Spfort 0.0143 0.273 0.9584 -0.521 0.550 Spmagn 0.7179 0.331 0.0304 0.068 1.368 Spscan 0.8993 0.301 0.0029 0.308 1.491 PC1:Spfort 0.4668 0.172 0.0069 0.128 0.805 PC1:Spmagn 0.6435 0.189 0.0007 0.272 1.015 PC1:Spscan 0.8699 0.156 0.0000 0.563 1.177 sinHourRadian -0.1116 0.256 0.6634 -0.615 0.391 890 891 Parameter estimates (B) ± unconditional standard errors (SE) represent the coefficients 892 from the weighted model averages with z statistics. The 95% confidence limit for each 893 parameter is indicated by Lwr2.5 and Upr97.5. 894 895

48

896 Table S5. Comparison of the candidate set of regression models described by the global 897 model: Cheeksmean ~ PC1 * Sp + PC2 * Sp + Plumage * Sp + sinHourRadian + 898 cosHourRadian 899 Model r2 K AICc ΔAICc ω ~PC1*Sp + Plumage + cosHourRadian + 0.618 16 2080 0.00 0.454 sinHourRadian + 1 ~PC1*Sp + Plumage + cosHourRadian + 1 0.616 15 2081 0.91 0.288 ~1 0.000 2 2576 495.41 0.000

Parameter B SE Prob Lwr2.5 Upr97.5 (Intercept) 36.4040 0.3553 0.0000 35.7062 37.1017 cosHourRadian -1.1113 0.3743 0.0030 -1.8465 -0.3762 PC1 0.8293 0.0741 0.0000 0.6837 0.9749 Plumage1 1.2585 0.2263 0.0000 0.8139 1.7031 Plumage2 0.0314 0.2532 0.9016 -0.4661 0.5289 Plumage3 -0.2977 0.1978 0.1333 -0.6863 0.0909 Plumage4 -0.2509 0.2771 0.3663 -0.7952 0.2934 Plumage5 -0.3965 0.2468 0.1089 -0.8812 0.0882 sinHourRadian -0.3095 0.1799 0.0862 -0.6630 0.0440 Spfort 0.2031 0.1900 0.2861 -0.1701 0.5763 Spmagn 1.1292 0.2308 0.0000 0.6757 1.5827 Spscan 0.7534 0.2099 0.0003 0.3411 1.1658 PC1:Spfort 0.1272 0.1190 0.2863 -0.1066 0.3610 PC1:Spmagn 0.1711 0.1315 0.1944 -0.0873 0.4295 PC1:Spscan 0.3684 0.1090 0.0007 0.1543 0.5825 900 901 Parameter estimates (B) ± unconditional standard errors (SE) represent the coefficients 902 from the weighted model averages with z statistics. The 95% confidence limit for each 903 parameter is indicated by Lwr2.5 and Upr97.5. 904

49

a b c 32 300 200

150 28 200 Max ETo

100 24 100

Temperature (°C) Temperature 50 Precip 20 Min Precipitation (mm/month)

0 Precipitation Deficit (mm/month) 0

J FMAM J J A SOND J FMAM J J A SOND J FMAM J J A SOND Month Month Month 905 906 Figure S1. Monthly weather patterns (minimum and maximum temperatures in panel a, 907 monthly precipitation in panel b, and precipitation deficit in panel c) depicted as box- 908 whisker plots (25 and 75% quartiles with range and outliers) from 50 years (1964-2014) 909 of weather station data obtained from the Charles Darwin Research Station website. 910 Potential evapotranspiration (ETo) was calculated from the weather data using the SPEI 911 package in R (Beguería & Vicente-Serrano 2013) using the Penman-Monteith calculations; 912 precipitation deficit (reported as a positive number for simplicity ) is calculated as the 913 difference between ETo and precipitation. 914 915

50

916

Heat Gain

2

● ● Mode 0 ● ● Convection ● ● ● ● Radiation Total Heat Exchange (W) Total

−2

Heat Loss

G. fuliginosa G. fortis G. magnirostris G. scandens

917 918 Figure S2. Total estimated heat exchange in the four species of Darwin’s finches examined, 919 grouped according to mode of heat transfer (convection vs. radiation), and displayed as a 920 density violin plot. G. magnirostris exhibited no convective heat gain (no positive values), 921 and exhibited periods of radiative heat loss (i.e., more extensive negative values). Open 922 circles depict the median values (ranging from -0.060 to -0.12 Watts), and grey filled circles 923 the mean values (ranging from 0.054 to 0.27 Watts). Mean values were higher due to

924 extreme values (not shown) of heat absorption (QTot > 4 Watts primarily via radiation). 925 The modelling provided an adequate estimate of heat transfer, with total values for heat 926 exchange falling close to the estimated rates of heat production (~0.36 Watts) for a 20 927 gram passerine bird (McNab 2002).

51