Contrasting capabilities of two ungulate species to cope with extremes of aridity

Study area

The study took place in the south-western Kalahari region of Botswana, known as the Bakgalagadi Schwelle (S 24.35°, E 20.62°), including the Botswana side of the Kgalagadi Transfrontier Park. The vegetation forms an open savanna, overlying deep sandy substrate with limited free-standing water. There is an intermittent river, Nossob river, in the south, ~ 80 km from the centre of the study area. A characteristic of this area is the highly mineralized, clay-rich depressions called pans, which retain water for variable periods after rain⁶. Air temperatures exceed 40 °C in summer and fall below 1 °C in winter⁶. Rainfall is seasonal but erratic, falling primarily during short-duration, high-intensity thunderstorms between November and April⁶. Mean annual rainfall in the Schwelle region ranges between 250 and 350 mm¹³.

Climatic variables

A free-standing miniature black globe thermometer (“miniglobe”), identical to the collar miniglobe thermometer, was placed within the area used by the animals in direct sun, 1 m aboveground, and recorded temperature (°C) every hour (S 24.307°, E 20.745°; reference miniglobe). Dry-bulb air temperature (°C), wind speed (ms⁻¹), and solar radiation (Wm⁻²) data were obtained from the Agricultural Research Council (ARC) weather station located at the Nossob campsite (S 25.4°, E 20.6°). Normalised Difference Vegetation Index (NDVI) (MODIS Terra 16-day) and local rainfall (mm; CHIRPS) data covering the study area (S 24.434°, E 20.293°) were obtained from Google Earth Engine¹⁴.

Study species and data collection

In August 2013, eight individual female gemsbok and eight individual female wildebeest, each from separate herds, were darted by a veterinarian from a helicopter. Each dart consisted of Thiafentanil (gemsbok: 7–8 mg, wildebeest: 4–6 mg, Thianil, Kyron Laboratories, Johannesburg, South Africa), medetomidine hydrochloride (gemsbok: 3–6 mg, wildebeest: 2–4 mg, Kyron Laboratories, Johannesburg, South Africa) and ketamine (gemsbok: 75–150 mg, wildebeest: 50–150 mg Pfizer Animal Health, Sandton, South Africa). Each individual was fitted with a GPS collar (African Wildlife Tracking, Pretoria, South Africa) that supported a miniglobe attached to the top to record the thermal environment that the individual bearing it occupied¹⁵. Miniglobe temperatures and GPS locations were recorded hourly. In addition, each individual underwent surgery to implant miniature temperature-sensitive data loggers in the retroperitoneal space and had a motion-sensitive data logger tethered to the abdominal muscle wall (see⁷ for details). The data loggers were covered with biologically and chemically inert wax (Sasol, South Africa) and sterilised in instant sterilant (F10 Sterilant with rust inhibitor, Health and Hygiene (Pty) Ltd., Roodepoort, South Africa) before implantation. Once the individual animal was immobile, it was placed in sternal recumbency with its head elevated and supported by sandbags. Following intubation, anaesthesia was maintained with 2–5% isoflurane (Aerrane, Astra Zeneca, Johannesburg, South Africa), administered in 100% oxygen. Incision sites were shaved and sterilised with chlorhexidine gluconate (Hibitane, Zeneca, Johannesburg, South Africa). A local anaesthetic (3 ml subcutaneously (S.C.); lignocaine hydrochloride, Bayer Animal Health (Pty) Ltd., Isando, South Africa) was administered to the incision site. After placement of the loggers, the incision was sutured closed. Respiratory rate, heart rate, arterial oxygen saturation, and rectal temperature were monitored throughout the surgery, which lasted approximately 30–45 min. Each individual animal also received an antibiotic (~ 0.04 ml kg⁻¹, intra muscularly (I.M.), Duplocillin, Schering-Plough Animal Health Ltd., New Zealand), and anti-inflammatory (~ 0.5 mg kg⁻¹ I.M., Metacam, Meloxicam injectable solution, Boehringer Ingelheim Vetmedica, Inc, St. Joseph, U.S.A.) medication. Following surgery and termination of inhalation anaesthesia, the immobilization drugs were completely reversed by a combination of naltrexone (gemsbok: 75–120 mg, wildebeest: 60–100 mg, I.M. Naltrexone, Kyron Laboratories, Johannesburg, South Africa) and atipamezole (gemsbok: 10–20 mg; wildebeest: 10–15 mg, I.M. Antisedan, Orion Corporation, Orion Pharma, Finland).

The temperature-sensitive data loggers (DST centi-T, Star-Oddi, Iceland) recorded body temperature at 10-min intervals (Fig. 1a,b) and the motion-sensitive data logger recorded whole body movements (i.e., motion changes) as activity counts within the first 10 s of each 5-min interval. The motion-sensitive logger had a triaxial accelerometer (ADXL345, Sigma Delta Technologies, Australia) with equal sensitivity across three planes (resolution one-fourth 4 mg/least significant bit). We adjusted the activity units to be relative to the maximum activity count for the entire study period per logger, to account for differences in the sensitivity of the individual motion-sensitive loggers. The data loggers and the collar weighed less than 1% of the individual animal’s body mass and is unlikely to have adversely affected their behaviour.

Two wildebeest were never relocated, possibly as a result of collar failure or predation. Three gemsbok died in October 2013. The remaining 11 animals were recaptured in May 2015, and data loggers and collars were removed. Thereafter the animals were released. Because of the inability to relocate all animals, animal deaths, and technological failures, we recovered a sample of 11 internal body temperature loggers (five gemsbok and six wildebeest); eight internal motion-sensitive loggers (four gemsboks and four wildebeest); nine GPS units (five gemsboks and four wildebeest) and nine miniglobe temperature sensors from the collars (five gemsboks and four wildebeest).

All procedures were approved by the Animal Ethics Screening Committee of the University of the Witwatersrand (protocol no. 2012/24/04) and all experiment procedures were performed in accordance with relevant guidelines and regulations as well as the ARRIVE guidelines (https://arriveguidelines.org/). The Government of Botswana via the Ministry of Environment, Wildlife and Tourism and Department of Wildlife and National Parks granted approvals and permits [numbers EWT 8/36/4 XX (32), EWT 8/36/4 XXVII (15), EWT 8/36/4 XXIV (102)] to conduct the study.

Data analysis

During the study period, the first hot-dry season (September to November 2013, ‘drought’) occurred at the end of a prolonged dry period, whereas the second hot-dry season (September to November 2014, ‘non-drought’) followed more typical rainfall conditions (Fig. 1c). Miniglobe temperature (24 h mean, minimum and maximum) and dry-bulb air temperature (24 h minimum and maximum), as well as mean 24 h wind speed and solar radiation were similar between the two hot-dry periods (Table 1). We averaged 16-day NDVI composites per season as an index of vegetation greenness in response to prior rainfall. Rainfall during the wet season prior to the commencement of the study (December to May 2013) was less than 40% (< 132 mm) and outside of the 95% confidence interval of the long-term average (255 ± 56 mm between 1981 and 2017) for the study area¹⁴, whereas rainfall during the subsequent wet season (277 mm) was within the long-term range (Table 1). The lower rainfall preceding the drought period resulted in lower vegetation greenness (as indexed by NDVI) compared to the non-drought hot-dry season (Table 1; Fig. 1c).

For consecutive 24 h periods, we calculated minimum, maximum and amplitude (maximum minus minimum) of the 24 h rhythm of body temperature. To determine behavioural adjustments in heat load brought about by microclimate usage, we calculated the difference between miniglobe temperature on the collar and the reference miniglobe per hour per individual animal. We calculated the time spent in microclimates cooler than direct sun by summing the hours when collar miniglobe was more than 0.5 °C lower than the reference miniglobe during daylight hours. We calculated a cumulative measure of cool microclimate usage during daylight hours per day as the sum of the hourly differences between collar miniglobe and reference miniglobe temperatures when animals were in cool microclimates (i.e., collar miniglobe was more than 0.5 °C lower than the reference miniglobe). To assess time spent travelling, we calculated the number of hours per day in which animals traversed more than 1.6 km within an hour, a distance previously associated with at least 50% of the hour engaged in directed movements at maximum walking speeds of ~ 3 km h^−116,17. We also calculated total 24 h activity and the proportion of activity that took place during the heat of the day (between 10:00 and 16:00) per individual. We used a series of Generalised Linear Mixed Models (GLMMs; Gaussian error family for continuous data, Binomial error family for proportions and Poisson error family for counts) to test the association between season and species in terms of body temperature profiles, cumulative measure of cool microclimate use and time spent in cool microclimates per day, time spent travelling per day, total 24 h activity and the proportion of activity that took place during the heat of the day (10:00 to 16:00). Model assumptions of residual normality and homoscedasticity were assessed graphically within R software¹⁸, informing subsequent data transformations and model choices. Season, species and an interaction between season and species were represented as fixed effects with individual animals and date included as random effects to control for repeated measures per individual per season. The response values of the body temperature profiles were heteroscedastic, so we included a variance structure and correlation structure within the models to correct for heteroscedasticity¹⁹. The total 24 h activity response values were log transformed for normality. We first fitted the global model, containing all the explanatory variables and interactions, then fitted season or species on their own. For example, our activity models took the form: log (total 24 h activity) ~ season + species + interaction between season and species + individual animals (random effect) + date (random effect). The coefficients reported are for the non-drought hot-dry season, wildebeest, and the interaction for wildebeest in the non-drought hot-dry season compared to the reference levels: drought, gemsbok, and gemsbok in the drought, respectively. Using model selection procedures of Akaike’s Information Criterion, corrected for small sample bias (AICc) and model weighting (wi)²⁰, we established the importance of each fixed variable and identified which model best supported the data²¹. Of all the candidate models, all models with a difference in AICc value of < 2 were considered plausible and were used for making inferences²⁰. We report on the effects size estimates and their precision (95% confidence intervals (CI)) per model as well as the conditional R² for mixed-effect models. The conditional R² describes the proportion of variance explained by both the fixed and random factors²². We used the R statistical software¹⁸ with R packages nlme²³ and lme4²⁴ to perform the GLMM analysis, AICcmodavg²⁵ to perform the model selection, and emmeans²⁶ to estimate marginal means from models. All graphs were created in GraphPad Prism (version 6.00 for Windows, GraphPad Software, San Diego, CA, USA).

Source: Ecology - nature.com