香京julia种子在线播放

The study was undertaken in semi-arid mallee woodland at Calperum Station, South Australia (34°03′S 140°38′E) during the austral summer (January–February) of 2020. The vegetation at the study site is comprised of low, multi-stemmed mallee eucalypts (Eucalyptus oleosa, E. dumosa, E. incrassata, E. socialis) and sugarwood (Myoporum platycarpum) above a sparse understorey of Senna and Acacia shrubs and porcupine grass (Triodia scariosa). All study groups lived in woodland that had not been burnt within the last 50 years. Mean annual rainfall is 251 mm, distributed irregularly across the year, and mean daily maximum temperatures in January and February are 33.8 and 32.7°C, respectively [data for Renmark, 1996–2020; Bureau of Meteorology, Australia (BOM)]. The frequency of extreme heat events at this location has been increasing, with the annual number of days ≥ 42°C more than doubling in the last 25 years (Figure 1A; Renmark, BOM).

The study species—the Jacky Winter (Microeca fascinans)–is a small (14–20g) Australasian robin (Petroicidae) found in woodlands throughout Australia (Higgins and Peter, 2002). It is a “perch and pounce” predator, snatching tiny to small arthropods, either on the wing, from the ground surface or—less frequently—from vegetation (Recher and Davis, 1998). Although not a desert-adapted species, Jacky Winters rely entirely on dietary water, and surface water was unavailable to the study population during the data collection period. Pairs are territorial year-round (territory size is 1.7 hectares in southern Queensland; Wood et al., 2008) although they occasionally form cooperative groups of up to six birds (Higgins and Peter, 2002). Breeding occurs from September to January, with multiple clutches (comprised of two eggs) laid in open cup nests (Donaghey and Donaghey, 2017). At high temperatures, Jacky Winters use wing-spreading to facilitate convective heat loss and panting for evaporative cooling (Sharpe et al., 2019, 2021).

Our color banded population of 40 breeding groups of Jacky Winters has been studied since May 2018. A subset of 16 groups have been habituated to the presence of an observer accompanying the birds at 3–5 m to allow accurate and unimpaired behavioral observations.

Behavioral data were collected from non-breeding birds between 8:00 and 16:00 during January and February 2020. Air temperatures during observation periods ranged from 15 to 43°C (Figure 1B; see “Weather and surface temperature data” for measurement details). We undertook 20-minute focal observation sessions, during which an observer and a scribe documented the following behaviors of the focal bird: all predator scares, social interactions (intra and intergroup squabbles and chases), non-foraging flights > 5 m and prey catching attempts (distance from perch to location of prey; location of prey: ground, air, leaves, trunk). Due to the very small size of the Jacky Winters’ prey, we were unable to ascertain whether prey catching attempts were successful. We also documented the start and stop times of all bouts of preening, panting and wing-spreading (bouts were considered ended if the behavior ceased for > 10 s or was replaced by a different behavior). We recognized three wing-spreading postures: wing-raising (wings held slightly away from the sides of body), wing-drooping (wings held further away from body and wing-tips lowered beneath the level of the tail) and wind-surfing (wings held out horizontally, similar to flying position). We also noted changes to body posture and ptiloerection but did not quantify these due to the difficulty of defining them objectively. Additionally, every 2 min throughout the 20-min observation period, we recorded details of the focal individual’s perch, estimating height above ground, degree of sun exposure (full sun, shade, dappled or overcast), direction faced (to nearest 45° compass point, using a Garmin Etrex 30x GPS), whether the bird was > 0.5 m from a living, naked (i.e., without dead fibrous bark) trunk/branch > 8 cm in diameter, and size of tree (diameter of the tree’s largest trunk at 1.5 m). Prior to data collection, all observers were trained to estimate distances and heights accurately using a Bosch digital laser distance measure. During observation periods, we paused all data collection if the focal individual ceased to be visible and resumed when the bird could be seen again. If the focal bird could not be relocated within 10 min-, we aborted the observation session. Sessions of < 5 min were discarded.

We collected behavioral observations for 17 adult, non-breeding Jacky Winters (nine males, eight females) from 10 habituated groups. To ensure we obtained a balanced sample across all individuals, time of day and temperatures, we divided the 8-h data collection period into four 2-h blocks and classified days as either mild (<35°C), warm (35–38°C) or hot (>38°C) based on the maximum temperature for that day. To select which bird to sample, we chose individuals randomly from the pool of unsampled birds for that specific temperature/time of day category. Every individual was sampled during all four time periods and on cool, warm and hot days, with an average of 14.4 ± 0.8 observation sessions obtained for each bird (range 11–21). In total, we collected 79 h of behavioral data and 2,379 records of perch choice.

To assess how weather variables influenced the birds’ behavior we used weather data collected every 30 min within 6 km of our study site by the Calperum-Chowilla OzFlux land-atmosphere observatory, operated by the Terrestrial Ecosystem Research Network (TERN). Windspeed and wind direction were measured 2 m above ground by a RM-Young Wind Sentry. At a height of 20 m, vapor pressure was measured by a LI-COR LI-7500 infra-red gas analyzer, and incoming shortwave radiation, by a large Kipp and Zonen CNR4. Air temperature in the shade was measured at a height of 2 m by a Vaisala HMP-45C temperature probe. We used this OzFlux weather data for all analyses of the birds’ behavior. Because our behavioral observation sessions were shorter than the weather logging interval (20 versus 30 min), when allocating weather data to a session that occurred between weather readings, we selected the weather reading closest to the observation period (i.e., always within 10 min).

To obtain an indication of the thermal profile of the bird’s habitat, we measured surface temperatures at 83 trees (all mallee eucalypts) within 10 Jacky Winter territories. All territories were within areas that had remained unburnt for at least 50 years. Temperatures were measured on clear afternoons during January 2020 (16:00 −19:30) when air temperature was ≥ 35°C (mean 39.5°C; range 35–47°C). We used a Digitech (model QM7221) infrared laser digital thermometer gun at a distance of 2cm, to measure all surface temperatures.

When choosing which trees to sample, we randomly selected up to five GPS waypoints from within each group’s territory (waypoints represent a location where the territory-owners have been observed, and a pool of approximately 35 waypoints were available for each group). At the waypoint location, we measured air temperature at a height of 1.5 m using a Falcon Kestrel 2000 pocket wind and temperature meter (0.5°C accuracy) positioned within our own shade. We also took three ground surface temperature measurements (sun-exposed sand, shaded sand and shaded leaf litter), all sampled > 2 m from a tree trunk. We then selected two trees: the nearest to the waypoint and the largest (based on trunk circumference) within a 20 m radius of the waypoint. At each tree, we identified the thickest trunk (because mallee eucalypts are multi-stemmed) and measured the shaded surface temperature of this trunk at its base and at a height of 1.5 and 2.5 m. We also recorded its circumference at 1.5 m. We identified any potential refuges (crevices, spouts, hollows, etc.) at the tree’s base and measured the surface temperature within these refuges at the deepest/shadiest location. We estimated the extent of leaf litter cover beneath the tree (the percentage of ground covered by leaf litter within a circle, 1 m in diameter, centered on the largest trunk) and measured the shaded surface temperature of this leaf litter.

We constructed a series of linear and zero-inflated mixed models to explore relationships between behavior, habitat characteristics and weather. Explanatory variables were chosen based on hypothesized drivers, including potential non-linear effects and interactions.

For each model we fitted all potential explanatory variables (as described above), then removed non-significant terms sequentially, starting with interactions, to arrive at a final model with only significant or near-significant terms. The model with the lowest AIC score was chosen, which always only included significant or near-significant variables. To avoid multicollinearity among explanatory variables we first estimated pair-wise Pearson’s correlation coefficients and confirmed that correlations were not high (|r| < 0.7). Model fit was checked using diagnostic plots to confirm model residuals were distributed around zero. For linear models we used residual plots and normality plots to check for deviations from normality among residuals, and for zero-inflated models we used DHARMa package (Hartig, 2020) to check for overdispersion.

Linear mixed models were fitted with REML (Restricted maximum likelihood) using the lmer function in the lme4 package (Bates et al., 2015) and generalized linear mixed models (GLMM) were fitted using the glmer function in the lme4 package (Venables and Ripley, 2002). Zero-inflated models (used for heat dissipation models) were fitted using the glmmTMB package with beta-binomial distribution. All models were fitted in R 3.3.2 (R Development Core Team).

To compare the surface temperatures of different microsites within mallee eucalypt trees (except for potential refuges), we used the full data set and conducted a Repeated One-way ANOVA and Tukey tests for all pairwise comparisons, using R 3.3.2. We also undertook a second Repeated ANOVA to compare the surface temperatures of potential refuges with other microsites, restricting the analysis to trees that contained refuges. To test whether the difference between surface temperatures and air temperature changed with air temperature, we used Pearson correlation coefficients to compare the variable “surface temperature minus air temperature” with air temperature. These tests, and all other non-modeling statistical tests were conducted in Microsoft Excel 2016 v2111. Means are presented ± 1 SE.

Although there was no overall effect of either air temperature or solar radiation on foraging effort, when both these variables were simultaneously high, Jacky Winters substantially reduced their foraging effort, after controlling for time of day (Table 1A: air temperature × solar radiation interaction; Figure 2A). At >35°C, the birds averaged only 12 (±2) prey capture attempts per hour compared with 35 (±2) at < 28°C. Hourly rate of attempts was also negatively related to perch height and showed a negative trend with the proportion of a bird’s perches that were in the shade (Table 1A).

To compensate for reduced foraging effort at high temperatures, individuals modified their time budget. On days > 38°C, they increased their rate of foraging attempts in the mornings (8–10 am) by 30% and reduced afternoon foraging rates by 70%, (as compared with days < 35°C) (Table 1B and Figure 3). To test if this increase in foraging effort was genuinely pre-emptive and that the birds were not simply compensating for a deficit of foraging on the previous afternoon, we focused on the hourly rate of foraging attempts at 8–10 am on days with a maximum temperature > 38°C, comparing days that were preceded by a warm day (35–38°C), when the birds foraged normally, and days preceded by a > 38°C day, when they abstained from afternoon foraging. There was no significant difference between the two (t-test: t₁₁ = 0.77, p = 0.46; hourly attempts: warm: 50 ± 5; hot: 42 ± 9).

Jacky Winters did not modify the distance they flew to catch prey in response to air temperature, windspeed or level of solar radiation, although they flew further at high temperatures when the wind was strong (Table 1C). The distance the birds flew to obtain prey was, however, positively correlated with perch height, but unrelated to whether the bird was perched in the shade. Males flew further than females. Flight distance also varied with the prey’s location, with birds flying furthest for aerial or ground prey but shorter distances for prey on leaves or tree trunks (Table 1C).

The hourly rate of non-foraging flights (>5 m) was negatively related to air temperature, after controlling for time of day (Figure 2B). This negative effect of increasing temperature on non-foraging flights tended to be exacerbated as solar radiation and windspeed increased (Table 2A: air temperature × solar radiation interaction, air temperature × windspeed interaction, respectively). Males moved more frequently than females (Table 2A). Hourly rates of social interactions were also negatively related to air temperature and to windspeed (Table 2B). However, the proportion of time spent preening showed a non-linear relationship with air temperature, increasing with temperature until the mid 30°Cs, then decreasing. Preening was also positively related to solar radiation, and this relationship was stronger as temperature increased (Table 2C: air temperature × solar radiation interaction).

The proportion of time spent wing-spreading (i.e., wing-raising, wing-drooping or wind-surfing) was positively associated with air temperature, although the relationship had a non-linear component (Table 3A and Figure 2C). The birds began wing-raising at air temperatures as low as 26°C and wing-drooping at about 32°C. Wing-raising was the most prevalent form of wing-spreading until around 35°C, when wing-drooping became more prevalent (Figure 2D). Proportion of time spent wing-spreading declined as windspeed increased, but the declining effect of windspeed on wing-spreading was reduced as air temperatures increased (Table 3A: air temperature × windspeed interaction).

Birds began panting at air temperatures in the 30°Cs, and proportion of time spent panting increased with air temperature (Figure 2E). Vapor pressure was also positively associated with time panting, but its effect on panting was less pronounced as air temperature increased (Table 3B: air temperature × windspeed interaction). Conversely, windspeed was negatively related to time spent panting: birds panted less when the wind was strong (Table 3B and Figure 4). Solar radiation was also negatively related to time panting.

The likelihood of a bird choosing to perch in direct sunlight was negatively correlated with both air temperature and level of solar radiation (Table 4A). In sunny conditions, birds began using shaded perches at air temperatures > 28°C. Only 1% of perches were in sun at temperatures > 35°C (compared with 47% at < 28°C; Figure 2F). Birds also shifted to shaded perches in windy conditions (Table 4A).

There was a non-linear relationship between the height of a bird’s perch and air temperature, with the birds switching to elevated perches at temperatures > 35°C, when mean perch height almost doubled (from 2.7 to 5.1 m; Figure 5A and Table 4B). Perch height was negatively related to windspeed, with birds perching lower in high winds (Table 4B). Birds choosing high perches favored larger trees, with a positive correlation between perch height and perch tree size (Pearson’s correlation: r₁₉₇₈ = 0.34, p < 0.001). The size of the perch tree was also related to air temperature, with birds favoring large trees at the highest temperatures (Figure 5B and Table 4C). Air temperature was also positively related to the likelihood of the birds perching within 0.5 m of a naked trunk or large branch (Table 4D).

Birds perched facing into the wind more often than expected by chance (44 versus 25%; Chi² goodness of fit: X² = 434.38, n = 2,322, p < 0.001) and this became more pronounced at high temperatures (67% at > 40°C; Table 4E). Facing into the wind was also positively related to windspeed, but negatively associated with level of solar radiation (Table 4E). Birds were also twice as likely to face into the wind if it was coming from the North (i.e., from the hot central Australian desert) as compared with other directions (78 versus 38%; Figure 6).

Once air temperature reached 39–40°C, Jacky Winters usually moved to the ground, sheltering in shaded crevices and depressions at the base of large trees. They did not perch on the ground under any other circumstances. At ≥ 39°C, 39% of our focal individuals retreated to a tree-base refuge, but this is a recognized underestimate because, once birds were within a refuge, we were unable to locate them to begin focal observations, so our sample was entirely biased to individuals that remained out of refuges. In total, we observed nine instances of Jacky Winters moving into tree-base refuges (eight individuals from four groups) but, at air temperatures ≥ 41°C, we observed only two individuals (from one group) remaining aloft.

There was considerable thermal heterogeneity in the birds’ habitat at air temperatures > 35°C (Figure 7). Sun exposure greatly increased surface temperatures, with sun-exposed sand 15.9 (±1.2)°C hotter than air temperature and 7.3 (±0.9)°C hotter than shaded sand, at an average air temperature of 39.5°C (Repeated one-way ANOVA for surface temperatures, excluding refuges: F_6,81 = 189.1, p < 0.001; Tukey: p < 0.001). The temperature difference between sun-exposed sand and air temperature increased as air temperature increased (Pearson’s correlations: r₄₂ = 0.442, p = 0.003).

Trees reduced ground surface temperatures, with shaded leaf-litter at the base of a tree 3.6 (± 0.4)°C cooler than shaded leaf-litter > 2 m from a tree (Tukey: p < 0.001). The average temperature of shaded, tree-base litter was the same as air temperature at air temperatures ≤ 38°C but rose to 6°C above air temperature at air temperatures ≥ 45°C (Pearson’s: r₇₆ = 0.481, p < 0.0001). The size of the tree (circumference of the tree’s largest trunk) was negatively related to surface temperatures beneath the tree, with larger trees being cooler at all measured locations, except for the trunk at 2.5 m (Tables 5A–D and Figure 8).

Leaf-litter also moderated ground surface temperatures with shaded leaf-litter 1.4 (±0.3)°C cooler than shaded sand (both measured > 2 m from a tree, at an air temperature of 39.5°C; Tukey: p = 0.047). The greater the proportion of ground covered by leaf-litter (within 1 m of the tree base), the cooler were the surface temperatures at the base of the tree (Tables 5A,B).

We documented two types of potential thermal refuge at the base of mallee eucalypts: tree hollows and spouts (holes within a dead or living trunk < 50 cm from ground; n = 20; Figure 9a) and small crevices that provided partial shelter (nooks among the tree-base roots or depressions on the basal lignotuber; n = 11; Figure 9b). We included crevices in our sample because Jacky Winters used this refuge type most often.

The surface temperature within potential tree-base refuges (crevices and hollows) was significantly cooler than all other locations (Repeated one-way ANOVA: F_7,31 = 189.1, p < 0.001; Figure 7). Tree-base hollows were 2.0 (± 0.8)°C cooler than air temperature, but tree-base crevices did not differ significantly from air temperature (Paired t-tests: hollows: t₁₉ = 2.368, p = 0.029; crevices: t₁₀ = −1.595, p = 0.142). Tree size affected the thermal benefit of tree-base crevices and hollows, with bigger trees providing cooler refuges (Table 5D and Figure 9c).

Large trees were also more likely to contain tree-base refuges (Figure 9d): refuge-bearing trees were significantly larger (trunk circumference) than those without (t-test: t₂₅ = 3.016, p = 0.006). Tree-base hollows did not occur in mallees with a trunk circumference < 30 cm (at a height of 1.5 m). Overall, refuges were rare: 63% of trees offered no potential refuges, even though we biased the sample toward large trees (by sampling the largest tree within 20 m at each sampling location).

Jacky Winters exhibited a broad repertoire of thermoregulatory behaviors, which was nuanced, flexible and highly responsive to both air temperature and other abiotic factors, such as wind velocity and solar radiation. The sequence of their behavioral responses to heat is shown in Figure 10. At air temperatures > 28°C, the birds withdrew to shade and reduced physically demanding activities such as non-foraging flights and social interactions. Air temperature per se did not reduce foraging activity (except at temperatures ≥ 38°C), suggesting that energy and water gains from feeding outweighed the costs of increased thermogenesis (Wolf and Walsberg, 1996). However, this balance shifted when air temperature and solar radiation were high simultaneously, and the birds curtailed foraging in response to this greater heat load. This reduction in foraging may have been influenced by heat-related changes to prey availability (Holm and Edney, 1973), but experimental studies show that birds given unlimited food still do not forage at high temperatures (digestion generates metabolic heat; Beale et al., 2018), suggesting that heat, not food availability, is responsible (Ricklefs and Hainsworth, 1968; Olinger, 2017).

The Jacky Winters also began adopting behaviors that facilitated convective heat loss at air temperatures in the high 20°Cs. In addition to wing-raising, to increase surface area to mass ratio, the birds flattened their plumage to reduce its insulative properties, and adopted a tall, upright posture to elongate body shape (Figure 10). As air temperature rose, wing-spreading became more exaggerated (becoming wing-drooping; Figure 10) and, when > 35°C, the birds moved to elevated perches in the largest trees. This preference for high perches appeared to be related to wind velocity, which increases with elevation (Mcilveen, 1992). Average windspeed at our site doubled between the heights of 2 and 10 m. Choosing the windiest microsites allowed the birds to maximize “forced” convective cooling, and it is notable that they did not use elevated perches if the wind was already strong. The thermal benefit of higher wind velocity is evidenced by our finding that birds devoted less time to heat-dissipation behaviors (e.g., wing-spreading and panting) with increasing windspeed, which is consistent with laboratory findings (Wolf and Walsberg, 1996; Wolf et al., 2000). Jacky Winters were also three times more likely to perch facing into the wind at high air temperatures (>40 versus <20°C), probably to enhance convective heat loss, because narrow objects, pointing into the wind, lose heat more rapidly (Mitchell et al., 2018). Facing into the wind also facilitated wind-surfing (i.e., allowing the wind to lift the wings into flying position; Figure 10), enabling air flow over a greater surface area. By seeking out windy microsites and maximizing convective cooling, birds can reduce their heat load without relying on more costly evaporative cooling.

At air temperatures in the high 30°Cs, the birds also favored perches close to naked, living branches or trunks, and some individuals were seen lying with their bellies against the wood (a posture only adopted during extreme heat or incubation; Sharpe et al., 2021). This presumably facilitated conductive cooling because the surface temperature of living timber has been shown to be significantly cooler than air temperature, due to water flow in the tree’s phloem tissue and thermal inertia (Briscoe et al., 2014). The birds may also have favored perches close to large branches to capitalize on the dense shade cast by large limbs, because unbroken shade is rare in this habitat.

As air temperature approached body temperature (approximately 40°C; Boyles et al., 2011), the birds began panting intermittently to facilitate evaporative cooling. Time spent panting increased with air temperature but was relatively low (averaging 13 min per hour at air temperatures ≥ 40°C) presumably reflecting the costs associated with this behavior. Panting uses body water, with rates of water loss increasing linearly with temperature (McKechnie et al., 2017). At air temperatures of 40°C, a small bird loses about 3% of its body mass per hour (e.g. the Lesser Goldfinch, Spinus psaltria; Albright et al., 2017), so the risk of lethal dehydration during prolonged heat exposure is extreme for birds without access to water. Evaporative cooling is also less efficient in humid conditions (Mitchell et al., 2018) and Jacky Winters spent more time panting (an autonomic response to body temperature) when vapor pressure was high.

Once temperatures rose into the 40°Cs, and dry (sensible) cooling was impossible, most of the birds retreated to the ground, taking refuge in shaded crevices or depressions at the base of large trees. These crevices were significantly cooler than any other location we measured (Figure 7). Inside refuges, the birds huddled together and lay with their bellies contacting the sand to facilitate conductive cooling. Jacky Winters do not normally perch within 5cm of conspecifics (Sharpe LL, pers. obs.), yet we observed one pair huddling with a territorial intruder they had been chasing earlier in the day. A similar breakdown of territoriality occurs in Arizona Grasshopper Sparrows (Ammodramus savannarum ammolegus) at high temperatures when the birds congregate in dense shrubs (Ruth et al., 2020). Huddling reduces the amount of surface area an animal exposes to the environment (Gilbert et al., 2010), which lessens heat gain when environmental temperature exceeds body temperature. Huddling inside a refuge also reduces evaporative water loss, due to increased humidity within the refuge (Luna et al., 2020), which is potentially critical for survival during prolonged heat exposure. The thermal and water-saving benefits of huddling may explain why, on the hottest days, we repeatedly observed large, mixed-species aggregations of small passerines sheltering together within single tree-base refuges.

Most Jacky Winters modified their behavior and microsite use to minimize convective heat transfer once air temperatures rose above body temperature. This behavioral transition, when heat gradients reverse, may prove to be critical for endotherms now experiencing unprecedented heat. Although we expect organisms to adopt behaviors that maximize their fitness, this is not always the case when individuals encounter novel conditions (e.g., responses to novel predators, artificial light, anthropogenic food; Tuomainen and Candolin, 2011). Although some species may have sufficient cognitive and behavioral flexibility to cope with this transition point (Ghalambor et al., 2007; Mills and McGraw, 2021), for others, air temperatures above body temperature may act as an insurmountable barrier to effective behavioral thermal regulation. Australian flying foxes (Pteropus spp.), for example, exhibit a suite of thermoregulatory behaviors (e.g., wing-fanning to maximize convective and radiant cooling, shade-use, panting and saliva-spreading; Welbergen et al., 2008) but in extreme heat they do not seek cooler microsites or adopt behaviors that minimize dry heat exchange. As a consequence, they succumb to lethal heat stress at air temperatures > 42°C, and 45,500 (in 600 colonies in SE Queensland) died on a single day in 2014 (Welbergen et al., 2014). This vulnerability to extreme heat may be much more widespread than we recognize, particularly among small endotherms (McKechnie et al., 2021) because few species are as readily observable as flying foxes, which roost in large, noisy colonies, and attempts to quantify individual responses to extreme heat are rare (Buchholz et al., 2019).

Even when lethal stress is avoided, thermoregulatory behaviors carry opportunity costs, with organisms known to reduce sociality, territorial defense, food intake or offspring care (Cunningham et al., 2021). Panting, for example, reduces foraging success in birds that excavate subterranean prey with their beak (du Plessis et al., 2012; van de Ven et al., 2019) and the utilization of cooler microhabitats is often associated with a reduction in forage quality or foraging success (Mason et al., 2017; van de Ven et al., 2019). With greater heat exposure, these costs may become unsustainable. For example, in Mexico’s Sceloporus lizards, the growing need for extended periods of inactivity during the breeding season, to avoid lethal heat stress, has resulted in the extinction of 12% of monitored populations (Sinervo et al., 2010). We found that the Jacky Winters’ use of elevated perches carried foraging costs, (after taking into account the effect of temperature), with birds making fewer prey catching attempts and flying further to capture prey. Perching in shade may also impair the birds’ ability to detect prey because they tended to make fewer foraging attempts from shaded perches. Southern Fiscals (Lanius collaris)—an African “perch and pounce” avian predator—consume 50% less prey when hunting from shaded perches (Cunningham et al., 2015), although this could be due to decreased foraging effort associated with temperature or time of day.

Reduced food intake—due to the necessity for inactivity and voluntary fasting at high temperatures—is a major cost of thermoregulatory behaviors in endotherms (Youngentob et al., 2021). By mounting video cameras at refuges, we found that Jacky Winters remained within refuges for up to 8 h and only emerged just before sunset (when air temperature dropped to about 38°C), precluding any afternoon foraging. Under these conditions, the birds lose > 2% of body mass per day (Sharpe et al., 2019), and similar losses occur in other wild avian species (du Plessis et al., 2012; Gardner et al., 2016; van de Ven et al., 2019). Such loss of body condition is likely to have long-term consequences for individual health, fitness and survival (Cunningham et al., 2021; Youngentob et al., 2021). Although Jacky Winters can partially recoup lost body condition during the interval between heat waves (Sharpe et al., 2019), the predicted increase in the frequency and duration of heat wave events is likely to have a detrimental impact.

In an effort to compensate for lost foraging time on very hot days, the Jacky Winters appeared to pre-emptively increase their foraging effort on the mornings of extremely hot days (Figure 3). Given that this species is reliant solely on dietary water, obtaining sufficient food prior to prolonged heat exposure will be critical for avoiding lethal dehydration. Pre-emptive foraging has also been observed in other avian species (Carroll et al., 2015; Cooper et al., 2019) but is reliant upon the birds’ ability to anticipate forthcoming conditions. Overnight temperature seems to be the most likely cue, as evidenced by an avian mass mortality event that occurred on a single, isolated hot day in Western Australia in 2010, when early morning temperatures were indistinguishable from those of previous, much cooler days (Cooper et al., 2019). The short-term predictability of heat waves may be critical to the survival of arid zone birds.

The Jacky Winters’ use of different microhabitats altered markedly as temperatures increased and at high temperatures the birds used the largest trees available to them. Surface temperatures at different microsites varied by up to a 35°C on the hottest days (air: > 44°C; sun-exposed sand: 77°C; tree-base refuge: 42°C) but thermally buffered microsites were rare. Tree-base refuges, the coolest sites, were only, on average, 2°C cooler than air temperature. The presence of trees and leaf-litter ground cover reduced surface temperatures, but tree size had the greatest influence: the larger the mallee eucalypt, the cooler the surface temperatures at its base, including within refuges (Figures 8, 9), presumably due to their larger, multitiered canopies (Lindenmayer et al., 2014). This greater leaf area not only increases shade below the tree but amplifies levels of evapotranspiration (Sanusi et al., 2017). Transpiration from leaves directly reduces air temperature below a trees’ canopy by 1 to 8°C, depending on species and environmental conditions (Rahman et al., 2020).

Large, fully mature mallee eucalypts also provided the birds with the opportunity to perch higher (giving them access to higher wind velocities) and were significantly more likely to contain potential tree-base refuges. In fact, tree-base hollows, the best quality refuges, were found only in the largest mallees (trunk circumference > 30 cm). The importance of large trees to the survival of Jacky Winters became apparent when air temperatures rose into the high 40°Cs. Although only one individual succumbed to acute heat stress/dehydration during the data collection period, 29% of our habituated adults died during a severe heat wave 3 weeks prior to data collection, when air temperature attained a record-breaking 48.8°C. These deaths appeared to be concentrated in parts of our study area that did not contain large trees, and although we did not verify this quantitatively, it underscores the importance of thermally buffered refuges for organisms facing unprecedented heat exposure.

Large, old trees act as keystone structures in many ecosystems, providing cavity nest-sites and unique microhabitats, facilitating habitat regeneration, storing large quantities of carbon and modifying ecosystem hydrology and water table level (Lindenmayer et al., 2014). However, the potential thermal benefits provided by old, large trees have yet to be quantified and we could find only one published study: in Canada’s boreal forests, stands of mature conifers were significantly cooler than stands of younger trees or shrubland when ambient temperatures were highest, and grizzly bears (Ursus arctos horribilis) favored these old stands at high temperatures (Pigeon et al., 2017). Interspecific comparisons of urban trees reveal that size per se is not a determinant of a species’ thermal buffering properties because small tree species often have denser canopies, and it is leaf area—and rate of evapotranspiration—that dictate cooling effect (Sanusi et al., 2017; Rahman et al., 2020). Arid adapted species provide less thermal benefit because they have lower transpiration rates. Although the thermal buffering provided by mallee eucalypts at our semi-arid site would be constrained by their restricted evapotranspiration rates and limited shade generation (the leaves of arid-adapted eucalypts are vertically aligned to minimize sun exposure at the hottest time of day; King, 1997), the largest trees at our site may have accessed ground water unavailable to smaller conspecifics (Lindenmayer et al., 2014), potentially augmenting their value as thermal refuges.

Our finding that the availability and value of thermal refuges increases with the size of mallee eucalypts has important management implications. Mallee stems are usually killed by wildfire and high intensity planned burns, hence trunk size and tree height increase with time since fire (Yates et al., 2017; Clarke et al., 2021). Indeed, it takes 60 years for mallees to attain full canopy height, and 40–60 years to develop hollows in living timber (Haslem et al., 2011; Clarke et al., 2021), likely reflecting their value as thermal refuges. Climate change is expected to influence the incidence of wildfires in mallee habitat; in mallee regions of Victoria, for example, the number of days of extreme fire danger and the occurrence of dry lightning are predicted to increase (Clarke et al., 2021). Whether this leads to increased fire incidence depends on their interaction with reductions in fuel biomass and continuity, associated with increasing aridity (Bradstock, 2010). Nevertheless, any increase in fire frequency would prevent mallee eucalypts from attaining the large size required for thermal buffering.

The world’s dryland areas are predicted to become increasingly arid, with worsening droughts (IPCC, 2021), which reduce vegetation cover and can permanently alter habitat structure (Pritzkow et al., 2021). At our study site, densely leaved, green shrubs—identified as critical thermal refuges in other terrestrial habitats (Carroll et al., 2015; Anthony et al., 2020; Ruth et al., 2020)—were simply not present during the data collection period due to 2 years of severe drought. In previous summers, we had observed many small passerines (although not Jacky Winters) sheltering within such shrubs at high temperatures. The loss of this microhabitat, combined with the reduction in leaf area that eucalypts suffer during drought (Pritzkow et al., 2021), is likely to seriously impair the ability of avian species to survive heat wave events.

Options to help biodiversity adapt to climate change are limited, but the relationship between fire interval, tree size and thermal refuges offers an opportunity for “low regrets” intervention (Prober et al., 2019) in mallee ecosystems. Managing fire to maintain and increase the area of mallee in older age classes is likely to maximize the availability of thermal refuges, thereby reducing mortality during heat waves. Later seral stages in mallee also provide significant habitat structural elements that support a diversity of reptiles, birds and mammals (Yates et al., 2017; Clarke et al., 2021). Another option worth considering relates to habitat structures known to act as critical refuges in other hot, arid ecosystems. Underground burrows are exploited by a huge array of species during extreme heat: four lark species use lizard burrows in the Arabian Desert (Williams et al., 1999) and more than 50 vertebrate species shelter in gopher tortoise (Gopherus polyphemus) burrows in south-eastern United States (Walde et al., 2009). However, burrows are no longer present in SE Australia’s mallee woodlands due to the extirpation of at least three species of medium-sized, burrowing marsupial (Fleming et al., 2014). The reintroduction of burrowing species could greatly enhance the availability of thermally buffered microsites within this habitat.

Jacky Winters exhibited a broad and nuanced repertoire of thermoregulatory behaviors, which appeared “appropriate” to environmental conditions, based on principles of heat transfer. However, the effectiveness of their behavioral strategies was dependent on the quality and availability of specific microhabitats. There is an urgent need for studies that quantify how wild individuals respond behaviorally to heat waves and identify which fine-scale components of their habitat serve as thermal refuges (Buchholz et al., 2019). Only once we have this information can we hope to develop effective conservation strategies to help to mitigate the adverse impact of rising global temperatures on biodiversity.