The most diverse and abundant family of termites, the Termitidae, evolved in African tropical forests. They have since colonised grassy biomes such as savannas. These open environments have more extreme conditions than tropical forests, notably wider extremes of temperature and lower precipitation levels and greater temporal fluctuations (both annual and diurnal variation). These conditions are challenging for soft-bodied ectotherms, such as termites, to survive in, let alone become as ecologically dominant as termites have.

Here, we quantified termite thermal limits to test the hypothesis that these physiological limits have widened in savanna termite species to facilitate their existence in savanna environments.

We sampled termites directly from mound structures, across an environmental gradient in Ghana, ranging from wet tropical forest through to savanna. At each location we quantified both Critical Thermal Maximum (CT_max) and Critical Thermal Minimum (CT_min) of all the most abundant mound-building Termitidae species in the study areas. We modelled the thermal limits in two separate mixed effects models against: canopy cover at the mound, temperature and rainfall, as fixed effects, with sampling location as a random intercept.

For both CT_max and CT_min savanna species had significantly more extreme thermal limits than forest species. Between and within environments, areas with higher amounts of canopy cover were significantly associated with lower CT_max values of the termite colonies. CT_min was significantly positively correlated with rainfall. Temperature was retained in both models, however it did not have a significant relationship in either.  Sampling location explained a large proportion of the residual variation, suggesting there are other environmental factors that could influence termite thermal limits.

Our results suggest there has been a widening of the thermal limits in termite savanna species. These physiological differences, in conjunction with other behavioural adaptations, are likely to have enabled termites to cope with the more extreme environmental conditions found in savanna environments and facilitated their expansion into open tropical environments.

Termites were sampled from four locations in Ghana across two different biomes: tropical rainforest and Guinea savanna. The three forest locations were: Bobiri Forest Reserve, a moist semi-deciduous forest in the Ashanti Region, a matrix of forest patches that have been, or currently are, logged (all colonies sampled from fragments that have not been logged for at least 20 years); the Forestry Research Institute of Ghana (FoRIG) Campus, a moist semi-deciduous forest patch located on the campus at Fumesua, near Kumasi, in the Ashanti Region; and the Assin-Attandanso Resource Reserve side of the Kakum Conservation Area, a moist evergreen tropical forest that has not been logged since 1990 (Wiafe, 2016), in the Central Region. Termites were sampled from a single savanna site, Mole National Park, a large unfenced protected area in the Savannah Region, between June and July 2019. Sampling in all locations was conducted during the end of the dry season and the start of the wet season. Termites were sampled during the end of the dry season and start of the wet season at all sites: between January and March 2019 for the three forest sites, and June and July 2019 for the savanna site.

Termite and Environmental Sampling

We sampled termites directly from termite mounds, identified to genus level (based on their diagnostic mound structures), which we located using directed searching, using the knowledge and experience of the field assistants. Two mounds were sampled per day, typically between 07h00 and 09h00. Target mounds were broken open and we collected termites, as well as some mound material, and placed them into a cool box to protect the termites from external conditions and reduce stress that could affect their thermal tolerances while being transported to the laboratory. The sample from each colony was placed in a separate container to prevent interaction between the two colonies. Where possible, both worker and soldier castes were sampled. The time between sampling and experimentation was reduced as much as possible (mean = 122.6 mins, sd = ±81.4 min). The large standard deviation is attributed to having a single collection, but running two experiments, per day. The same colonies that were sampled were used in two experiments, one testing thermal maximum and one testing thermal minimum (the order of which was decided randomly); hence, the individuals used in the second experiment were kept inside the cool box (which was out of direct sunlight) while the first experiment took place.

We photographed each mound to assist with species identification. Canopy cover above each sampled mound was quantified using a hemispherical fish-eye lens (Canon EOS 70D camera, Sigma 4.5mm f/1:2.8 Circular Fisheye Lens). For large mounds the image was taken directly above the apex of the mound, but with small mounds (<75 cm in height) the image was taken above the mound at a height of 75 cm due to the height of the stabilising monopod. Due to variable light conditions when the images were taken, we did not standardise ISO, shutter-speed or aperture, although variation in these settings were reduced as much as possible. Variable light conditions and camera settings were corrected using Gap Light Analyzer software (GLA; Frazer et al., 1999). The approximate leaf-area index above each mound was then calculated as a measure of canopy cover. Leaf-area index is a measurement of the quantity of leaf area in a canopy per unit area of ground, and ranges from 0 (no leaf cover) to 6 (complete leaf cover).

Thermal Tolerance Experiments

We used similar methods to those outlined in Bishop et al. (2017) to estimate the thermal tolerance of the termites. Fourteen individuals from each of two colonies (28 in total) were tested per experimental run. Only individuals of the helper castes, workers and soldiers, were used in the experiments Each termite was placed into a separate microcentrifuge tube. The tubes were then placed into a dry heat bath (Tropicooler 260014‐2, Boekel Scientific, Feasterville, PA, USA; temperature range -19°C to 69°C and an accuracy of ±1°C). The heat bath contained two aluminium inserts with each consisting of 14 wells, each of which held a single microcentrifuge tube. Individuals from two colonies and different castes were randomly allocated to different wells to remove potential equipment bias. The two colonies sampled each day were used in each experimental run (CT_max and CT_min) and each colony was tested for both CT_max and CT_min; however, each individual termite was only used in one experiment.

The temperature within each microcentrifuge tube was acclimated for 15 minutes. CT_max experiments had an acclimation temperature of 30°C, and CT_min had an acclimation temperature of 24°C. After the acclimation period, the temperature was changed by 1°C (raised for a CT_max experiment, and lowered for CT_min), and maintained at each new temperature for a further three minutes. Following this three-minute period, we checked each termite for the onset of a heat or chill coma. Heat coma onset was considered to have occurred when the individual lost muscle co-ordination; if this occurred it was recorded as the CT_max of that termite. A chill coma was considered to have occurred when there was a complete lack of movement despite flicking the microcentrifuge tube; if this occurred it was recorded as the CT_min of that termite. When we noticed the termite had lost muscle co-ordination (CT_max) or had stopped moving (CT_min) the temperature was recorded as the thermal limit of that individual termite. The experimental run was ended when the thermal limit of all 28 termites was recorded.

Termite Identification

Termite colonies were identified to genus level in the field based on their mound structure. All termites used in experimental runs were preserved in 70% ethanol and brought to the UK. We took an individual from each colony and DNA was extracted from each sample. The COII mitochondrial gene was amplified and sequenced using Sanger Sequencing at the Natural History Museum, London. Each sequence was matched with the most closely related sequence on NCBI using BLAST (Johnson et al., 2008). If the sequences had a unique high likelihood identity match (>98%) with a single species described on the NCBI database, that sample, and therefore the colony it was sampled from, was considered identified as that species. If there was not a unique high likelihood identity match, the sequences (and therefore the colonies from which each sample was taken) were grouped and labelled as an unspecified species within a genus (for example, Cubitermes sp. A). After identification, species that did not have at least three replicates per sampling location were removed from the dataset (19 colonies removed, present in online dataset). Since the genetic identification was conducted, the genus Cubitermes was split into several genera (Hellemans et al., 2020). However, there were limited sequence data for the new classifications so we have treated all the new genera as Cubitermes sensu lato.

Metadata tabs added in the datasets provide information as to the columns, their data type, units required, data format and a general description, which should provide the users with the necessary information to use these data.