To study intraspecific flexibility on a population-level, we worked with the endemic insectivorous Commerson’s leaf-nosed bat Macronycteris commersoni (Hipposideridae). This species is widespread across various vegetation formations and climatic regimes in Madagascar (only absent from the central highlands) and uses different types of diurnal roosts within its range (Goodman, 2011). For example, in Tsimanampetsotse National Park in south-western Madagascar, a dry spiny forest habitat, a population of M. commersoni roosts in a humid hot cave year-round (“cave habitat”; Reher et al., 2019), while another population roosts in the open vegetation among branches in dry deciduous forest ∼380 km further north, in Kirindy Forest/CNFEREF (“forest habitat”; Reher and Dausmann, 2021). The hot cave is well buffered and offers constant conditions of 32°C and 98% relative humidity at bat height (Reher et al., 2018), whereas the bats roosting in exposed foliage without any insulation have to cope with daily ambient fluctuations and weather extremes (Reher and Dausmann, 2021). Both habitats are situated in Madagascar’s drier western formations and highly seasonal, with a pronounced dry season without any rainfall for up to 9 months and reduced resource availability. In the dry spiny forest rain may even be absent for several years (Ratovonamana et al., 2013). T_a fluctuates on a daily basis between 12.6 and 32.5°C (mean daily minimum and maximum; cave habitat, outside the cave) as well as 13.3 and 34.0°C (forest habitat) in the dry season. During the wet season, resources are more abundant and particularly night-time T_a is higher, with daily fluctuations between 24.4 and 38.5°C (cave habitat) as well as 23.6 and 36.0°C (forest habitat).

We trapped bats in June/July 2016 (dry season) and February/March 2017 (wet season) in the cave habitat and in February/March as well as July/August 2018 in the forest habitat (wet season and dry season, respectively). Depending on habitat structure and bats’ roosting behavior, we used different trapping methods. In the cave habitat during the dry season, we trapped the bats with a hand net in the early morning hours between 07:00 and 09:00 because this species is inactive during this time of the year. In the wet season, we erected a harp trap (two-bank 4.2 m² harp trap; Faunatech∼Austbat, Bairnsdale, Australia) in front of the same cave’s entrance. In the forest habitat, two to three mist nets (3 m high × 6 m long, mesh size 19 mm; Ecotone, Sopot, Poland) per trapping night were installed blocking main flight corridors in the local grid systems N5, CS5, CS6, and CS7 in both seasons. The trap and nets remained open for the first 3 h after sunset and were checked every 10–20 min depending on season. We transferred a maximum of two bats into separate cloth bags and released additional adults and juveniles immediately at the capture site. None of the females were reproductively active during data collection.

We manipulated the bats at the point of capture. They were weighed to the nearest of 1 g, sexed and the forearm length was measured. We removed a patch of fur between the shoulder blades using a razor blade and attached a temperature-sensitive radio transmitter (∼ 0.9 g; Biotrack, Wareham, United Kingdom) with medical skin glue (Osto-Bond, Montreal, Canada; or Manfred Sauer GmbH, Lobbach, Germany), whereby the thermal sensor was placed directly on the skin. The proportion of transmitter mass to the lowest bat’s body mass was 2.7% (mean = 1.8 ± 0.43%, N = 70 bats) and thus well below recommended maxima of 5–10% (e.g., Aldridge and Brigham, 1988; Rojas et al., 2010). All transmitters were calibrated in a water bath at seven intervals over a temperature range of 3–45°C against a precision thermometer traceable to a national standard. Skin temperature (T_skin) obtained via external transmitters provide a non-invasive and reliable proxy of body temperature (T_b), particularly in small mammals (Audet and Thomas, 1996; Dausmann, 2005; Langer and Fietz, 2014; but see Willis and Brigham, 2003). We marked all bats with an individual, three-digit wing tattoo using non-toxic ink (Hauptner-Herberholz, Solingen, Germany) after locally anesthetizing the membrane (EMLA, AstraZeneca, Wedel, Germany). Animal handling lasted about 10 min per individual and never exceeded 15 min.

To study the effects of sudden environmental changes on the metabolic responses of the respective populations, we exposed individuals of each population to the respective contrasting roosting conditions, i.e., cave bats had to cope with fluctuating conditions outside the cave and the forest population with constant hot and humid conditions. In the cave habitat, measurements (N = 10 dry season, N = 13 wet season) were conducted in a shaded enclosure near the camp without any further buffer to ambient extremes. In the forest habitat, two modified incubators (Exo Terra PT2445; HAGEN, Holm, Germany) served as climate chambers to mimic cave conditions and were calibrated to maintain constant T_a = 32.1 ± 0.3°C and ambient relative humidity (RH_a) = 88.9 ± 2.4%. RH_a was increased by passing the air entering the climate chamber through a humidified sponge and by adding a dish with water to the floor of the climate chamber. Although 88.9% does not reflect cave relative humidity precisely (i.e., 98%), it never dropped below 82% and was as close as we could get under field conditions. We placed the respirometry chamber into the climate chamber and connected it to the oxygen analyzer (see below), whose pump drew the humidified air from the climate chamber through the respirometry chamber and finally the excurrent dried air to the analyzer. Since we did not trap any bats in the forest during the dry season, we only ran these measurements in the wet season (N = 10). The data from these experimental set-up were then compared to the physiological responses of bats resting in their native roost type, which are summarized and analyzed in an ecological context, i.e., focusing on habitat and roosting conditions but not acute microclimate changes, in Reher et al. (2022).

We used open-flow respirometry in pull mode to measure the rate of oxygen consumption (V̇O₂). We individually transferred the bats into 2 L plastic metabolic chambers, which were equipped with a net for the bats to hold onto, right after preparing them at their respective capture sites (i.e., attaching tag etc., see above). The chamber was connected to a portable gas analyzer (OxBox; T. Ruf and T. Paumann, University of Veterinary Medicine Vienna, Austria) with an integrated electro-chemical oxygen cell (Bieler & Lang, Achern, Germany). The oxygen sensor was calibrated in the laboratory before and after each field trip using room air and calibration gases generated by a gas mixing pump (19.9, 20.3, and 21% O₂ content; 2KM300/a, Wösthoff Messtechnik GmbH, Bochum, Germany). During the measurements, ambient air was pulled through the respirometry chamber at a constant flow of 50 L h^–1 using a diaphragm pump, dried with silica gel and filtered before passing through a flowmeter integrated in the OxBox. An aliquot thereof was then drawn through the gas analyzer. The oxygen content of the sample air was measured every 10 s for 55 min; reference air from outside the metabolic chamber was also dried with silica gel, filtered and its oxygen content was measured every 10 s for the remaining 5 min to control for drifts in the oxygen sensor. These data were used to correct the oxygen traces (Clampfit v10.3.1.4, Molecular Devices, Sunnyvale, United States).

The measurements started between 21:00 and 23:00 h in the wet season and 09:00 and 11:00 hr in the dry season owing to different trapping times. We recorded data for 20–32 h and ended the respirometry runs with the beginning of the bats’ usual active phase between 17:30 and 18:30 h to ensure access to immediate foraging possibilities upon release. We used the first MR values right after introducing the bats to the metabolic chambers as indicators of individual stress levels. If there was no marked decrease in these initial values within 30 or 60 min (depending on the inactive or active phase, respectively), the bats were immediately released at the capture site. Before we released them at their capture site, they were weighed a second time to calculate mean body mass (BM) and again offered food and fresh water (see Reher et al., 2018). During each respirometry run, we recorded T_a and RH_a inside each respirometry chamber (resolution: 0.0625°C, 0.04% RH; Hygrochron iButtons, Maxim integrated, San Jose, United States), and T_skin using a remote receiver/logger placed next to the setup (DataSika SRX-800-D, Biotrack, Wareham, United Kingdom). All temperature data were recorded at 5-min intervals.

In total, we analyzed data of 70 bats, none of which was a recapture (Table 1). We processed all data using Cran R (R Development Core Team, 2018) in the “RStudio” environment (RStudio Team, 2016) with support of the packages “plyr” (Wickham, 2011), “dplyr” (Wickham et al., 2020), “readxl” (Wickham and Bryan, 2017), “lubridate” (Grolemund and Wickham, 2011), “zoo” (Zeileis and Grothendieck, 2005), “stringr” (Wickham, 2019), “tidyr” (Wickham, 2020), “ggplot2” (Wickham, 2016), “ggpubr” (Kassambara, 2020a), and “cowplot” (Wilke, 2020).

Of all data sets, including those collected under natural conditions that sometimes lasted several days (Reher et al., 2022), only the first day of a measurement (from earliest 21:00 to ∼ 30 min before sunset) was analyzed. Moreover, because the forest environment is somewhat milder and both, warm (cloudy, rainy) and very hot days, can occur during the rainy season, we only included data from forest bats measured on hot days (N = 12 bats), which better reflected the southern regular weather.

We calculated individual MR (ml V̇O₂ h^–1), using the rate of oxygen consumption (V̇O₂) as ml O₂ h^–1 corrected to STPD with equation 11.2 given by Lighton (2008) specifically for this set-up, assuming an average respiratory quotient of 0.85 (oxidation of 50% fat and 50% carbohydrate, Dausmann et al., 2009). To account for effects of differences in BM among individuals on MR, we performed an initial regression analysis showing that resting MR (RMR) was better explained by BM as a linear function through the origin (likelihood ratio test: Chi² = −939.1, P = 0.023). For comparability in further analysis and data reporting, we therefore divided V̇O₂ by mean BM for mass-specific MR calculations (ml V̇O₂ h^–1 g^–1). We defined different physiological states via visual inspection of MR patterns following Reher and Dausmann (2021). Thereby, a drop in MR by at least 50% compared to RMR was considered to constitute torpor, which lies within the range of higher metabolic reductions seen during torpor in warm environments (25–84%; Song et al., 1997; Dausmann et al., 2009; Grimpo et al., 2013; Kobbe et al., 2014). We removed torpor arousal and torpor entry phases and continued with a subset of MR data to determine RMR and torpid MR (TMR). For RMR, we only included the lowest 50% of resting data per hour and per individual within the bats’ usual resting phase, i.e., from sunrise to 30 min before sunset to ensure that variations owing to disturbance or minor activity were excluded (Bethge et al., 2017; Reher and Dausmann, 2021; Rodgers and Franklin, 2021). For TMR, we included the lowest 70% of data per hour and per individual. Macronycteris commersoni is known to enter torpor bouts of varying lengths, including micro-torpor bouts lasting 12–20 min on average (Reher and Dausmann, 2021; Reher et al., 2022). However, the bats rarely entered micro-bouts in the manipulated conditions, and we therefore only included torpor bouts lasting ≥ 1 h for downstream analysis. We allocated RMR and TMR to different bins of T_a by rounding T_a to the nearest integer and assigning individual means of the different metabolic states to the respective T_a to avoid pseudo-replication.

To obtain T_skin, we converted the recorded transmitter signals from beat per minutes into temperature data using second-order polynomial regressions obtained from the calibration curves (all R² ≥ 0.99). T_skin was used as a proxy for T_b to complement already defined torpor states through MR.

The forest population naturally roosted in a fluctuating environment with daily variation in T_a between on average 23.6 ± 1.2°C and 36.0 ± 2.2°C (min/max: 21.2/41.7°C) as well as daily variation in RH_a between on average 47.6 ± 10.6% and 98.5 ± 2.2% (min/max: 29.0/101.2%). When exposing individuals from this population to constant 32.1 ± 0.3°C and 88.9 ± 2.4% RH_a (i.e., experimental cave conditions), all bats entered torpor within the first hour of measurement. Individuals were torpid for more than 70% of the total measurement duration (mean torpor bout duration: 263 ± 87 min; Table 1) despite several arousals within 24 h. They were euthermic for an average of only 19 ± 14 min at a time and never > 57 min (Figure 1A). This was in clear contrast to the torpor pattern observed in the native population as only 60% of native cave bats entered torpor and for a much smaller fraction of the total measurement duration (26%; Table 1; W = 1, P < 0.001). In comparison, native cave bats showed ∼17-fold longer euthermic phases (320 ± 143 min; W = 144, P < 0.001), usually entered torpor in the late morning and aroused shortly before their active phase (Figure 1C).

We found no difference in RMR between the two populations (Table 1; t₂₂.₈ = 0.497, P = 0.624) but TMR was lower in the forest bats exposed to cave conditions than the native cave population (Table 1; W = 91, P = 0.010). Despite lower TMR, forest bats had higher T_skin (Table 1; t₁₃ = −5.08, P < 0.001) and maintained a greater T_skin-T_a differential while torpid under cave conditions than native cave bats (2.3 ± 1.0°C vs. 4.2 ± 0.6°C; t₁₆.₂ = −5.21, P < 0.001). Simultaneously, during the short euthermic periods T_skin increased to 37.9 ± 0.7°C in the forest population, which was significantly higher than euthermic T_skin in the native cave population (35.1 ± 1.2°C; t₁₉.₃ = −4.65, P < 0.001), as was the T_skin-T_a differential (native cave population: 3.0 ± 1.0°C vs. experimental cave conditions: 5.2 ± 0.6°C; t₂₂.₃ = −4.72, P < 0.001).

The cave population roosts at constant 32 ± 0.1°C and 98 ± 0.5% RH_a at bat height (Reher et al., 2022). When we exposed cave bats to experimental forest conditions in the wet season, i.e., T_a fluctuating on a daily basis between on average 24.4 ± 1.7°C and 38.5 ± 3.1°C (min/max: 22.4/43.0°C) as well as RH_a fluctuating between on average 37.0 ± 8.4% and 85.6 ± 7.0% (min/max: 25.8/96.6%), all individuals entered torpor. One individual was torpid for almost the complete measuring duration while the other twelve bats remained euthermic during their usual active phase and entered torpor in the morning (Figure 1B). Similar to the experimental forest bats, all individuals from the native forest population entered torpor on hot days, i.e., days on which T_a exceeded its euthermic T_skin (Table 1 and Figure 1D; Reher and Dausmann, 2021). Individuals from the cave population in forest conditions entered torpor at high T_a earlier in the day (cave bats entry: 09:02 ± 0:40 h vs. native forest bats entry: 10:48 ± 1:06 h; x = 0.272, P < 0.01) and at lower T_a than individuals from the native forest population, at 32.1 ± 1.8°C and 34.8 ± 1.4°C, respectively (t₁₉.₄ = −3.98, P < 0.001).

We determined a TNZ for the cave bats in experimental forest between 31 and 34°C, which was slightly lower and narrower than the TNZ in the native forest population (32–36°C; Reher and Dausmann, 2021; Figure 2A). Within the TNZ, there was no significant difference in RMR between cave bats and native forest bats (0.67 ± 0.09 ml O₂ h^–1 g^–1 vs. 0.77 ± 0.20 ml O₂ h^–1 g^–1, respectively, W = 258, P = 0.057; Figure 2A). Below the respective TNZ, RMR increased with decreasing T_a (t = −10.51, P < 0.001), whereby the slope was steeper in cave bats measured under forest conditions (t = 2.41, P = 0.016; Supplementary Figure 1). In all bats, TMR increased with increasing T_a, which was especially apparent at higher T_as (t = 11.53, P < 0.001).

T_skin during both, torpor and euthermia increased with increasing T_a (torpor, t = 37.51, P < 0.001; euthermia, t = 12.74, P < 0.001; Figure 2B), but the slope of this increase was steeper in the cave population in experimental forest (torpor, t = −8.17, P < 0.001; euthermia, t = −2.28, P = 0.022; Supplementary Figure 2). At lower T_as, there was a noticeable difference in cooling rates in torpid individuals, with cave bats under forest conditions losing heat more quickly than their native forest conspecifics (Figure 2B).

In the dry season, the cave population roosted at constant 31.4 ± 0.1°C and 98.1 ± 2.8% RH_a at bat height. All individuals entered torpor when we exposed cave bats to experimental forest dry season conditions, i.e., fluctuating T_a between on average 16.3 ± 1.9°C and 32.5 ± 2.6°C (min/max: 6.9/36.4°C) as well as RH_a fluctuating between on average 31.5 ± 8.4% and 86.9 ± 7.6% (min/max: 17.8/99.1%). One individual was torpid for the entire measurement duration and did not arouse before it was returned to the cave, while all other individuals entered torpor either at night (02:18 ± 1 h 22 min, 14.9 ± 3.4°C; N = 4; Supplementary Figure 3A) or in the morning when T_a had increased (11:13 ± 32 min, 31.3 ± 1.0°C, N = 5). Only three bats were able to defend a low, but fairly stable T_skin when T_a decreased at night (T_skin: 29.3 ± 1.0°C; T_a: 17.5 ± 2.5°C; Figure 3A), with a daily amplitude of 27.7 ± 0.5–34.3 ± 2.4°C.

Mean euthermic T_skin in cave bats exposed to experimental forest dry season conditions was lower than that of bats under all other conditions (W = 2; P < 0.001; Table 1). In six bats, T_skin decreased with decreasing T_a (Figure 4B) although they increased maximum MR up to twofold RMR between T_a of 20–30°C (2.02 ± 0.60 ml O₂ h^–1 g^–1) and up to threefold RMR below T_a of 20°C (2.94 ± 0.15 ml O₂ h^–1 g^–1; Figure 4A), with an individual maximum of 6.46 ml O₂ h^–1 g^–1, i.e., > 6-fold RMR. Four of these six bats eventually entered torpor with a considerable drop in MR and reached a steady-state TMR within ∼1 h (55 min ± 17 min; Figure 3B). T_skin at torpor entry was 29.6 ± 1.1°C and torpid T_skin ranged from 7.8 to 38.4°C, depending on time of day and thus T_a. The other two bats slowly decreased MR and T_skin over several hours (3 h 38 min ± 49 min; Figure 3C), a pattern resembling hypothermia and not torpor as suggested by reversed hysteresis of MR and T_skin (Geiser et al., 2014; Supplementary Figure 3). Nonetheless, mean MR during torpor and potential hypothermia (excluding entry and arousal phases) were comparable (torpor 0.08 ± 0.05 ml O₂ h^–1 g^–1 vs. potential hypothermia 0.07 ± 0.04 ml O₂ h^–1 g^–1).

The cave bats exposed to experimental forest dry season conditions entered the longest torpor bouts (W = 282; P = 0.039; Table 1) with lowest mean TMR of all bats (W = 80; P < 0.001; Table 1).

Interestingly, roosting in unfamiliar environmental conditions did not lead to increased energy expenditure under any of the conditions (Table 1). In the wet season, both—cave bats exposed to the forest as well as forest bats exposed to experimental cave conditions—consumed less energy than their conspecifics in their natural roost (cave bats in forest microhabitat: W = 162, P = 0.002; forest bats in cave microhabitat: W = 1, P < 0.001). In the dry season, the cave bats had a similar energy expenditure when roosting in the experimental forest environment or in their native cave (W = 42, P = 0.842) but spent 84% of their energy budget at night defending euthermia, much more than they usually do in the cave during this time of the year (55%; t₁₇ = −5.276, P < 0.001; Supplementary Figure 4).

Bats usually roosting in constant cave conditions could not defend euthermia when T_a decreased below their TNZ under experimental forest conditions. While three individuals defended a stable, albeit low, T_skin (∼29°C), six bats failed maintaining it constant below a T_a of ∼30°C: T_skin decreased with decreasing T_a despite maximal MR increases of sixfold RMR. Most bats eventually entered torpor to compensate for increased energy demands during this extensive upregulation of metabolic heat production. Two cave individuals, however, showed hypothermia-like responses in forest conditions, indicated by a noticeable slow torpor entry with reversed hysteresis and negligible T_skin-T_a differentials. Additionally, although these bats later passively rewarmed to almost euthermic T_skin (∼34.9°C, Supplementary Figure 3), they did not arouse spontaneously but only upon disturbance. These patterns resemble hypothermia rather than controlled torpor (Tomlinson et al., 2007; Geiser et al., 2014) and thus, reflect the bats’ limited capacity to generate sufficient heat at T_as below the TNZ. Simultaneously though, MR was substantially reduced and as low as MR in bats that had entered torpor. While we cannot specify the exact mechanism, we suggest that these bats entered torpor involuntarily with a hypothermic descent and subsequent transition into torpor.

Small mammals usually acclimatize to cold conditions by increasing their capacity for non-shivering thermogenesis (NST) in skeletal muscle or brown adipose tissue (Cannon and Nedergaard, 2004; Bal et al., 2012, 2016; Nowack et al., 2017a). Given the lack of cold exposure though, the cave bats likely have only limited amounts of brown adipose tissue and only limited capacity for muscle NST (Heldmaier and Hoffmann, 1974; Bal et al., 2016). This population overwinters in a hot cave offering thermal neutral 32°C year-round and does not leave the cave for the entire winter season, when they potentially could face low T_a of ∼7°C outside the cave (Rakotoarivelo et al., 2007; Raharinantenaina et al., 2008; Goodman, 2011; Reher et al., 2019). In the wet season by contrast, T_a outside the cave may drop to ∼20°C, but bats only experience this comparatively low T_a while foraging. Flapping flight is costly, and MR may increase up to 8–15-fold during flight, thereby generating substantial amounts of heat, which helps to maintain a stable body temperature despite the cold air (Carpenter, 1986; Speakman and Thomas, 2003; Voigt and Lewanzik, 2011). T_b regulation during the active phase is further supported by digestive heat production (MacArthur and Campbell, 1994; Beale et al., 2018), and thus, the native cave individuals may rarely ever have to actively increase heat production for thermoregulatory purposes.

Importantly, the attempts to defend euthermia at low T_a enforced a shift in the allocation of energy. In experimental forest microhabitat, cave bats spent 84% of their energy expenditure at night, compared to 55% under natural cave conditions. Interestingly, cave bats under forest dry season conditions entered longer and deeper torpor bouts during the day than all other individuals, by which they compensated for the increased night-time energy expenditure and maintained a similar energy expenditure as under native cave conditions. Although flexible torpor use mitigated the energetic costs, the cave population severely struggled defending euthermic T_b when T_a dropped below their TNZ, suggesting adaption to high cave T_a.

All forest bats exposed to constant experimental cave conditions remained torpid for a large fraction of the measurement duration and the occasional euthermic periods were the shortest recorded within this study. RMR during those short euthermic bouts was comparable to RMR of the native cave individuals, although resting T_skin was higher in forest bats in cave conditions than in cave bats and increased to a hyperthermic maximum of 39.4°C in one individual. The forest population usually experiences considerable daily fluctuations in both T_a and RH_a (total range: 21.2–41.7°C, 29.0–101.2%; Reher et al., 2022), but in the forest highest T_a always coincides with lowest RH_a. Many endotherms inhabiting warm and dry environments have developed adaptations to conserve water (Heldmaier et al., 2013; Fuller et al., 2016; Clarke, 2017). For example, different bird (Muñoz-Garcia and Williams, 2005; Champagne et al., 2012) and bat species (Muñoz-Garcia et al., 2012; Ben-Hamo et al., 2016) from arid habitats have a modified skin lipid composition with higher proportions of waxy lipids compared to their conspecifics from mesic habitat, which reduces the skin’s permeability and thus the rate of cutaneous water loss (Haugen et al., 2003). If the forest population, which is roosting under rather dry conditions, did have such adaptations, its evaporative cooling capacity would be limited. In addition, high RH_a near saturation experienced by the bats under (experimental) cave conditions reduces the gradient that drives evaporation and thus, the amount of heat an animal can dissipate (Gerson et al., 2014). Cooling through evaporation is possible at high RH_a when the water vapor pressure (WVP), e.g., on the skin surface, is higher than that in the air (Schmidt-Nielsen, 1997; Withers et al., 2016). Indeed, individuals may lick their forearms when heat stressed (Reher and Dausmann, 2021). This observation, and the high T_skin we measured during short euthermic periods, suggest that the forest bats were attempting to maintain cooling mechanisms by increasing both skin temperature and WVP on the skin’s surface. A higher T_skin-T_a differential additionally allows for passive pathways of heat loss through convection, conduction and radiation (Tattersall et al., 2012; Mitchell et al., 2018). Interestingly, torpid forest bats in the experimental cave had a higher T_skin than their torpid native cave conspecifics despite lower TMR and thus, lower metabolic heat production. During torpor however, vasomotor regulation is considered less functional (Muñoz-Garcia et al., 2016), suggesting that the high T_skin was not necessarily actively upregulated by vasodilation but that the patterns reflect difficulties in efficient heat dissipation. Thus, the excessive use of torpor in forest bats measured under hot cave conditions was probably driven by heat stress, a response also observed in the same population in their natural forest environment but at higher T_a (Reher and Dausmann, 2021). Although T_a of the imitated cave was within the forest bats’ TNZ, we suggest that the combination of high T_a with high RH_a may have overwhelmed their thermoregulatory system (Cooper and Withers, 2008; Gerson et al., 2014). This particular population appears to rely on a certain water vapor gradient to dissipate excess heat, and this gradient was compromised when the bats were exposed to the highly humid conditions mimicking the experimental cave.

When we exposed the cave bats to experimental forest wet season conditions, we found a lower thermal inertia below the TNZ and higher cooling rates during torpor in cave bats than native forest bats. This is consistent with the dry season results discussed previously and indicates that this population lacks traits for coping with T_a below their commonly experienced 32°C in the cave. However, the bats’ responses to T_a exceeding their TNZ (up to 42.7°C) were surprisingly homogenous. At daytime T_a higher than usually experienced in the cave, all individuals responded by entering hyperthermic (“hot”) torpor, thus showing the same response that native forest bats use in their forest environment (Reher and Dausmann, 2021). Interestingly, bats from both populations did not use torpor in the early morning hours to benefit from lower T_a and enhance energetic savings like other tropical bats do (Geiser et al., 2000; Bondarenco et al., 2013), but entered torpor when T_a approached or exceeded their respective upper TNZ limit. Torpor at high T_a reduces metabolic heat production, allowing higher rates of heat from the environment to be stored in the body via facultative hyperthermia and reduces water lost during evaporative cooling when T_a exceeds T_b (Lovegrove et al., 2014; Welman et al., 2017; Mitchell et al., 2018; Reher and Dausmann, 2021). Although the general patterns were intriguingly similar, cave bats measured under forest conditions entered torpor at lower T_a, indicating that they were more heat stressed at lower T_a than the native forest bats, which is also reflected by their slightly lower TNZ. Since the cave bats roost in a very specific microclimate of high T_a and high RH_a year-round, it is likely that they are adapted to their roosting conditions rather than the larger-scale, dry and hot climate of southwestern Madagascar. Water is scarce here, which could entail that outside the cave, at a higher WVP gradient and without sufficient access to water, this population may not necessarily be able to uphold cooling mechanisms (Mitchell et al., 2018) and the bats’ water budget could deplete quickly. Thus, entering torpor may have been the only physiological option when roosting outside the cave (Bondarenco et al., 2014; Reher and Dausmann, 2021). However, tropical bats experiencing high T_a in their diurnal roosts appear to have a great capacity for coping with heat (Cory Toussaint and McKechnie, 2012; Bondarenco et al., 2013, 2014; Czenze et al., 2020, 2022; Noakes et al., 2021). Indeed, torpor combined with hyperthermia never was observed in the native cave population roosting under constant cave conditions, which indicates that this particular population has a broader physiological repertoire than it commonly uses in its natural environment and thus, certain phenotypic flexibility in the context of heat exposure.

Outside the respective populations’ commonly experienced environmental range, most bats showed patterns that aid overcoming acute perturbations. However, we have also seen responses suggesting limited physiological compensation of acute microclimate change. Forest bats, for example, had difficulties to dissipate heat at high RH_a to the point that they were hyperthermic during short euthermic periods, which may have severe physiological consequences when T_b exceeds certain thresholds during tropical cyclones or extreme rainfall events, adding humidity to high T_a (Schmidt-Nielsen, 1997; Lepock, 2003). The cave population exposed to T_a below their TNZ could not maintain a high and constant euthermic T_skin despite expending more than 80% of their energy budget and two individuals had intriguing torpor entries similar to hypothermia. Hypothermia is potentially fatal (Tomlinson et al., 2007; Geiser et al., 2014) and the high night-time energy consumption may become problematic in the long-term when torpor has to be traded off with social interactions like reproduction to avoid depleting energy reserves in the dry season. While the response to heat resembled the forest bats’ response, cave bats were heat stressed at lower T_a than forest bats. It remains to be confirmed whether longer term responses to changed conditions mirror those of our study with a time-limited scope. Endotherms may adapt to novel conditions within hours, weeks or generations depending on their scope for phenotypic plasticity (Menon et al., 1996; Piersma and Drent, 2003; Heldmaier et al., 2013; Noakes and McKechnie, 2020). To take advantage of these processes, however, acute coping mechanisms are crucial to overcome interim time spans. Accordingly, the narrow capacity of the cave population to cope with conditions unlike those specific to hot caves is of particular concern given that human-induced disturbance is already evident in this population (Goodman, 2006), and no further hot caves are currently known to exist in the region (Reher et al., 2018).

Interestingly, increased torpor use appeared to be the common response for coping with unfamiliar microclimatic conditions, as reflected in the high ratio of bats entering torpor in the experimental set-ups. In this particular bat species though, multi-hour torpor is usually avoided when the populations roost in their natural conditions unless energy savings are vital or the bats are heat stressed (Reher and Dausmann, 2021; Reher et al., 2022). This may be due to certain costs associated with torpor (see Landes et al., 2020 for review). Especially the ecological and behavioral consequences may be critical for M. commersoni. Reduced sensory and motor capabilities during torpor can increase predation risk and limit social opportunities (Choi et al., 1998; Mzilikazi and Lovegrove, 2002; Nowack et al., 2016b). Nonetheless, torpor is a powerful emergency response to withstand acute disturbances such as droughts, heatwaves, storms, fires and flooding (Doucette et al., 2012; Bondarenco et al., 2014; Nowack et al., 2015, 2016a; Stawski et al., 2015; Barak et al., 2019). Importantly, torpor allows individuals to drastically reduce internal heat production if the water budget does not allow for extensive evaporative cooling or to finely balance the energy budget if unanticipated environmental changes have necessitated a short-term increase in energy expenditure (Herreid and Schmidt-Nielsen, 1966; Geiser, 2004; Heldmaier et al., 2004; Cooper et al., 2005; Nowack et al., 2017b; Reher and Dausmann, 2021). This is advantageous to endure unfavorable environmental conditions in general, but potentially also to bridge the time needed for adaptive mechanisms to take effect. None of these are options for homeothermic species, which likely face dramatic challenges when their environment changes rapidly.