Skip to main content
Dryad logo

Arboreality drives heat tolerance while elevation drives cold tolerance in tropical rainforest ants


Leahy, Lily; Scheffers, Brett; Williams, Stephen; Andersen, Alan (2021), Arboreality drives heat tolerance while elevation drives cold tolerance in tropical rainforest ants, Dryad, Dataset,


Determining how species thermal limits correlate with climate is important for understanding biogeographic patterns and assessing vulnerability to climate change. Such analyses need to consider thermal gradients at multiple spatial scales. Here we relate thermal traits of rainforest ants to microclimate conditions from ground to canopy (microgeographic scale) along an elevation gradient (mesogeographic scale) and calculate warming tolerance in the Australian Wet Tropics Bioregion. We test the thermal adaptation and thermal niche asymmetry hypotheses to explain interspecific patterns of thermal tolerance at these two spatial scales. We tested CTmin, CTmax, and calculated CTrange using ramping assays for 74 colonies of 40 ant species collected from terrestrial and arboreal habitats at lowland and upland elevation sites and recorded microclimatic conditions for one year. Within sites, arboreal ants were exposed to hotter microclimates and on average had a 4.2°C (95% CI: 2.7 – 5.6°C) higher CTmax, and 5.3°C (95% CI: 3.5 – 7°C) broader CTrange than ground-dwelling ants. This pattern was consistent across the elevation gradient, whether it be the hotter lowlands or the cooler uplands. Across elevation, upland ants had significantly lower CTmin than lowland ants, whereas the change in CTmax was less pronounced, and CTrange did not change over elevation. Differential exposure to microclimates, due to localised niche preferences, drives divergence in CTmax while environmental temperatures along the elevation gradient drive divergence in CTmin. Our results suggest that both processes of thermal adaptation and thermal niche asymmetry are at play depending on the spatial scale of observation, and we discuss potential mechanisms underlying these patterns. Despite the broad thermal tolerance range of arboreal rainforest ants, lowland arboreal ants had the lowest warming tolerance and may be most vulnerable to climate change.


These datasets are ant thermal tolerance limits and environmental microhabitat temperatures from a tropical rainforest in the Australian Wet Tropics Bioregion. Ants were sampled and microclimate was recorded at ground and arboreal/canopy microhabitats at replicate trees at three sites along an elevation gradient, two lowland sites at 100 m asl, Carbine and Daintree, and a upland elevation site, 1200 m asl at Carbine. In each dataset, the latitude and longitude of sites are provided.  


Across all sites, we considered all individuals of a species from a single tree to be from the same colony and ants from separate trees as from different colonies. Multiple colonies of each species were collected where possible. We tested 20 species represented by one colony and the remaining 20 species by 2–5 colonies (mean ± SE: 1.85 ± 0.17). We aimed to test five individual worker ants from each colony. In some cases, there were more or less than five individuals available to test. Overall, the mean number tested per colony was 4.8 ± 0.3 SE. For the polymorphic genera Pheidole and Camponotus we tested minor workers only. Following thermal tolerance experiments (see below) ants were placed in ethanol and returned to the lab to confirm species identification and to measure body mass. Between 3–5 (mean ± SE: 4.1 ± 0.1) workers of each tested colony were oven dried for 24 hrs at 70°C using a Blue M Electric drying oven, and dry weight body mass was measured using a Satorius semi-microbalance scale with 0.01 mg accuracy. Voucher specimens were deposited in the ant collection held at CSIRO’s Tropical Ecosystems Research Centre in Darwin, Australia.

We took measures to limit sources of variation from any potential seasonal or daily thermal trait plasticity. Ants from different sites were tested within a three-month period of one year during the dry season when the weather is more stable, and ants within sites were tested within a period of seven to ten days. In addition, ants were tested for thermal tolerance as soon as possible after collection to avoid experimental acclimation. Average time between collection in the field and beginning testing was 3 hrs 22 min (± SD 2 hrs) and the maximum time was 8 hrs and 47 min.

To measure thermal limits, we used a custom digital dry bath with 20 individual wells set into an insulated aluminum block above a Peltier plate that was programmed to heat and cool at a set rate. Each single ant was placed in a 1.5 ml microcentrifuge vial capped with a firm piece of foam to prevent it from hiding in the lid space above the thermal block. We tested several species at one time and individual workers of different species were randomly placed in one of the 20 well positions to prevent any systematic biases. We reserved one random well position for a temperature test vial without an ant, which had a microcentrifuge vial with a thermocouple wire inserted and capped with firm foam. This ensured we were recording the temperature that an ant would experience inside the vial. Following Kaspari et al. (2015), we used a protocol that ramped temperatures down to CTmin, allowed ants to recover, and then ramped upwards to CTmax. Kaspari et al. (2015) did not find any evidence that ramping down before ramping upwards systematically biases the final estimate of CTmax. We placed ants in the thermal block for 15 min at 20°C to allow the resumption of normal behaviors. To test CTmin we ramped temperature down at 0.5°C per minute. This ramping rate was chosen as a careful balance between a slow ramping rate that is more likely to be experienced in nature and a faster ramping rate that avoids the potentially confounding effect of desiccation associated with long experimental times (Rezende et al. 2011). On ramping down, the ants were checked at 15 °C, 12°C, and then every 1°C until there was loss of muscle coordination or absence of any movement after flicking the vial, indicating CTmin (Lutterschmidt and Hutchison 1997). Upon reaching CTmin ants were removed from the experiment and left to acclimate back to room temperature. Once all ants were removed, they were left for 15 min, after which they were checked for survival and normal movement. Ants that did not recover were replaced by a worker from the same colony for the CTmax assay (15 out of 380 ants tested, 4% of tests). Ants were then acclimated for another 15 min at 20°C in the thermal block before temperature was ramped up. Temperature was raised 0.5°C per minute and ants were checked at every 2°C increment between 30–40°C and then every 1°C from 40°C until CTmax was reached. CTmax was determined as above for CTmin and ants were checked for recovery after 15 min. Ants that did not recover from CTmax were excluded from the analysis as they had reached a lethal rather than critical thermal limit. We calculated CTrange (CTmax – CTmin) for individuals which recovered from both CTmin and CTmax assays. Excluding failed tests, we analysed the test results for 40 species for each thermal trait with 355 individuals of 74 colonies for CTmin, 277 individuals of 73 colonies for CTmax, and 269 individuals of 73 colonies for CTrange.

The dataset provided is the colony average (averaged over the individuals tested) CT limits and body mass, and includes the species vertical niche (ground or arboreal) and site of capture. 


We recorded ambient temperature at 30 min intervals using HOBO Pro v2 (U23-001) temperature data loggers at ground (~0.5 m) and canopy (~20 m) vertical habitat at 100 m asl and 1200 m asl Carbine sites.  Loggers were placed for one year from March 2019 – March 2020 in the same year as ant sampling.

In the dataset provided, we have calculated daily minimum, daily maximum, and daily mean temperature for each vertical habitat and elevation site.

Usage Notes


There are missing values for one colony for CTmax and CTrange indicated by NA. Models in the manuscript were run for each thermal trait separately such that there were 74 colonies included in the model for CTmin, and 73 colonies included in the models for CTmax and CTrange. 


Explorers Club, Award: Student Grant

Wet Tropics Management Authority, Award: Student Research Grant

Skyrail Rainforest Foundation, Award: Research Grant

Holsworth Wildlife Research Endowment, Award: Research Grant

Wet Tropics Management Authority, Award: Student Research Grant