Temminck’s pangolins relax the precision of body temperature regulation when resources are scarce in a semi-arid environment

ABSTRACT Climate change is impacting mammals both directly (for example, through increased heat) and indirectly (for example, through altered food resources). Understanding the physiological and behavioural responses of mammals in already hot and dry environments to fluctuations in the climate and food availability allows for a better understanding of how they will cope with a rapidly changing climate. We measured the body temperature of seven Temminck’s pangolins (Smutsia temminckii) in the semi-arid Kalahari for periods of between 4 months and 2 years. Pangolins regulated body temperature within a narrow range (34–36°C) over the 24-h cycle when food (and hence water, obtained from their prey) was abundant. When food resources were scarce, body temperature was regulated less precisely, 24-h minimum body temperatures were lower and the pangolins became more diurnally active, particularly during winter when prey was least available. The shift toward diurnal activity exposed pangolins to higher environmental heat loads, resulting in higher 24-h maximum body temperatures. Biologging of body temperature to detect heterothermy, or estimating food abundance (using pitfall trapping to monitor ant and termite availability), therefore provide tools to assess the welfare of this elusive but threatened mammal. Although the physiological and behavioural responses of pangolins buffered them against food scarcity during our study, whether this flexibility will be sufficient to allow them to cope with further reductions in food availability likely with climate change is unknown.


Figure S2.
The 24h body temperature patterns of a pangolin, showing the estimated (Est.) and actual (Act.; camera trap data) burrow emergence and return times for two consecutive summer days (top) and two consecutive winter days (bottom) Pangolin 24h body temperature records showed a notch (indicated by an increase or decrease in body temperature by at least 0.5°C in less than one hour; 93% of notches were ≥0.5°C for data points for which we had camera trap data for validation) around the times of emergence from and return to their burrows.We tested the accuracy of the body temperatures deviations to identify the time of emergence and return to the burrow by matching 178 camera trap times of emergence and 65 camera trap times of return with 24h body temperature records.The success of using of a conspicuous body temperature notch to detect burrow emergence or return was calculated by counting the number of times (reported as a percentage of time) the notch was detectable to within one hour of the actual emergence and return using camera traps (Figure S2).The use of the body temperature notch to detect time of emergence (for when camera trap data were available) was successful for 89% of the time during autumn, for 89% of the time during spring, and for 100% of the time during winter.In other words, for only up to 11% of the time, depending on the season, the body temperature notch was not conspicuous enough to detect time of emergence from the burrow.During summer, however, the use of body temperature notches to detect time of emergence was only possible for 64% of the time because the notches were not as distinct during summer compared to the rest of the year.The low detection success during summer resulted in fewer estimated times of emergence being available for analysis during summer compared to the rest of the year.The use of the body temperature notch to detect time of return (for when camera trap data were available) was successful for 100% of the time during autumn, for 92% of the time during spring, for 100% of the time during winter, and for 93% of the time during summer.In other words, for only up to 8% of the time, depending on the season, the body temperature notch was not conspicuous enough to detect time of return to the burrow.
The time of emergence and return estimation error was determined by calculating the absolute difference between the estimated time of burrow emergence or return using 24h body temperature patterns and the actual time of burrow emergence or return using camera traps.On average, the time of emergence estimation error was 32 ± 28 (mean ± SD of total emergences) minutes for summer, 20 ± 22 minutes for autumn, 12 ± 11 minutes for winter, and 18 ± 15 minutes for spring.The time of return estimation error was 17 ± 18 (mean ± SD of total returns) minutes for summer, 13 ± 7 minutes for autumn, 11 ± 15 minutes for winter, and 21 ± 24 minutes for summer.
Frequency histograms created using activity obtained from camera traps only (n=243; emergences, 65 returns) and activity derived from 24h body temperature only (n=4998; emergences, 2254 returns) revealed that the overall distribution of the data was similar for the two methods (Figure S3).We were therefore confident that 24h body temperature notches could be used to accurately estimate time of emergence and return to the burrow to within one hour of actual activity.

Figure S3 .
Figure S3.Frequency distributions of time of emergence (A) and return (B) data obtained directly from camera traps only (black bars) and indirectly from 24h body temperature patterns only (grey bars) for which camera trap data were available.The hourly bins represent the time of day during which emergence from or return to the burrow occurred (for example, 1 = 01h00-01h59 and 20 = 20h00-20h59)

Figure S6 .
Figure S6.Estimate marginal means of daily minimum globe temperature by season, averaged over period.Error bars show the 95% confidence interval for the estimates

Figure S8 .Figure S9 .
Figure S8.Estimate marginal means of daily amplitude of globe temperature by season, averaged over period.Error bars show the 95% confidence interval for the estimates.

Figure S10 .Figure S11 .Figure S12 .
Figure S10.Raw monthly rainfall recordings at the study site for the period 1 November 2015 to 31 October 2017.

Figure S14 .
Figure S14.Estimated marginal means of prey abundance by season, averaged over period.Error bars show the 95% confidence interval for the estimates.Values are back-transformed from the log scale

Figure S15 .
Figure S15.Estimate marginal means of daily minimum body temperature by season, averaged over period.Error bars show the 95% confidence interval for the estimates

Figure S16 .
Figure S16.Estimate marginal means of daily maximum body temperature by season, averaged over period.Error bars show the 95% confidence interval for the estimates

Figure S17 .
Figure S17.Estimate marginal means for daily amplitude of body temperature by season, averaged over period.Error bars show the 95% confidence interval for the estimates

Figure S18 .
Figure S18.Estimate marginal means of daily mean body temperature by season, averaged over period.Error bars show the 95% confidence interval for the estimates

Table S1 .
Days of data per season in each period of study 3 Missing days due to weather station failure 8 Emergence times Raw data Figure S5.Distribution of burrow emergence times for each animal

Table S3 .
Linear regression model of seasonal and yearly (period) differences in daily minimum globe temperature (ºC)

Table S4 .
Pairwise contrasts of estimated marginal means of daily minimum globe temperature

Table S5 .
Linear Estimate marginal means of daily maximum globe temperature by season, averaged over period.Error bars show the 95% confidence interval for the estimates regression model of seasonal and yearly (period) differences in daily maximum globe temperature (ºC) 1 CI = Confidence Interval Figure S7.

Table S6 .
Pairwise contrasts of estimated marginal means of daily maximum globe

Table S7 .
Linear regression model of seasonal and yearly (period) differences in daily amplitude of globe temperature (ºC)

Table S10 .
Pairwise contrasts of estimated marginal means of daily mean globe temperature

Table S13 .
Generalised linear mixed-effects model (negative binomial link function) results of the interannual and seasonal differences in ant abundance

Table S14 .
Pairwise contrasts of estimated marginal means for prey abundance (ants/trap)

Table S16 .
Pairwise contrasts of estimated marginal means for minimum body temperature

Table S18 .
Pairwise contrasts of estimated marginal means for maximum body temperature

Table S20 .
Pairwise contrasts of estimated marginal means for daily amplitude of body

Table S22 .
Pairwise contrasts of estimated marginal means for daily mean body temperature