Differential construction response to humidity by related species of mound-building termites

The two closely related termite species Macrotermes michaelseni and M. natalensis coexist sympatrically, are morphologically almost indistinguishable, and demonstrate congruent behavioural characteristics, including the construction of large mounds and the tending and harvesting of fungal symbionts deep inside their nests. Yet despite these similarities, the mound superstructures they create are markedly different. Macrotermes michaelseni build tall pillars that can stretch 2 to 3 m high (Fig. 1A) (Turner, 2001), while M. natalensis construct short, squat mounds less than 1 m high, without spires (Fig. 1B) (Harris, 1956). It has been hypothesized that some behavioural difference between individual termites of the two species must give rise to this discrepancy (Turner, 2011); however, identification of such has remained elusive (Green et al., 2017). In this paper, we show that although the two species display no significant differences in building activity under still-air conditions with high ambient humidity (>80% relative humidity), when humidity is lowered (<45% relative humidity) significant differences appear. This suggests some discrepancy in behavioural threshold or sensitivity to airborne moisture, which when extrapolated over a long time scale, may relate to the distinct macro-scale structures observed in the African bushland.

In natural conditions, air humidity is high inside the mound and low outside (Turner, 2001). Macrotermes spp., with a permeable cuticle, are vulnerable to desiccation and do not survive long outside the mound (Hu et al., 2012). When exposed to an external environment (and given a suitable substrate), they create tunnels and passageways that shelter them from the harsh, often dry conditions. By confining termites to a laboratory arena and providing them with a soil reservoir with a single restricted access point, we were able to provoke and monitor deposition activity without the potentially confounding effects of excavation (Green et al., 2017).

Experimental work was conducted at the Cheetah View Field Biology Station near Otjiwarongo, Namibia (20°25′S, 17°4′E)​. We took groups of 35 major worker termites from three different colonies each of Macrotermes michaelseni (Sjöstedt) and Macrotermes natalensis (Haviland 1898) and induced them to build on a flat, undifferentiated surface in a humidity-controlled environment (Fig. 1C,D). This environment was created using a feedback loop between three Sensirion SHT85 humidity sensors and an off-the-shelf misting humidifier (UrPower OD-101). Control and sensor recordings were facilitated by an I2C breakout circuit connected to an Intel NUC5i7RYH. The chamber contained a plastic cylinder completely filled with nest soil in its solid Atterberg state (approximately 25% by mass) (ASTM Standard D 4318 test) with an acrylic plate above it. The plate was sanded with a diagonal hatching of fine-grade sandpaper to enable the termites to move easily across the surface.

A 5-mm-diameter circular opening in the centre of the plate gave access to the cylinder of soil underneath, which provided both a source of soil for deposition on the surface and a space where termites could retreat via excavation. We monitored building activity above the surface over 4 h, recording 2D activity (using an RGB camera) and 3D soil deposition (using an infrared depth camera) at 1 Hz (Carey et al., 2017) (Fig. 2, Movie 1). In one set of trials, the ambient humidity inside the chamber was held high (81.3±3.2% relative humidity). A second set held the humidity at 39.9±5.2% relative humidity, close to the natural level of humidity of the outside environment, though not so dry that the termites would desiccate before building had begun. Each of the six colonies provided three groups of termites per treatment, giving n=9 trials for each combination of species and humidity condition.

Prior to the experiment, termites were stored in a closed container lined with wet paper towels. Soil was sieved to remove large particulates and all soil cylinders were prepared together, so that within the same colony, the soil texture and water content was consistent. Condensation from the humidifier was trapped by a condenser plate set just above the humidifier, underneath the build surface, ensuring the arena was not affected by large droplets or water spray (Fig. 1C). A 3D-printed ring structure confined the termites to the area of the build surface directly under the camera. At the end of each experiment, the mud cylinder was detached from the plate, any termites remaining on the plate were removed with forceps, and the resulting build structure was dried completely, then weighed. The area of the plate covered by deposition and the 3D surface map of the resulting build structure (as viewed from above) were also recorded. Custom analysis software used for 3D reconstruction is available from GitHub (https://github.com/niccarey/RGBD-Environment-Analysis).

Owing to the low sample size, significance testing was performed using an exact permutation test (https://www.github.com/lrkrol/permutationTest; MATLAB, Natick, MA, USA), a non-parametric test that is valid even with an unknown sampling distribution (Hesterberg et al., 2005). In high-humidity conditions, none of the measured parameters registered a statistically significant difference between species. However, in low-humidity (ambient) conditions, the area, volumetric envelope and mass transfer of soil all showed a significant (P<0.01) difference between M. michaelseni and M. natalensis (Fig. 3). At low humidity, the M. michaelseni groups transferred less soil, covered less of the plate with deposition, and built smaller-volume structures than the M. natalensis groups (Fig. 2; Table S2). Previous studies, which have largely taken place at high humidity and did not physically separate excavation from construction (e.g. Green et al., 2017), did not reveal this difference. We note that the density ratio of soil mass to enclosed volume was not significantly different among the different experimental conditions (Table S1), indicating there were no gross differences in the internal structure not visible to our sensors.

These results raise the possibility that sensitivity to air moisture content may drive the differences in large-scale mound morphology, and add an interesting perspective to prior work linking soil and air moisture to the building behaviour of termites and the evolving structure of the mound (Turner, 2011; Soar et al., 2019). The characteristic mound shapes and internal architecture could be an emergent function of the physics of airborne moisture and the termites' own sensory thresholding. If, for example, evaporative flow is constrained to a largely vertical rising water vapour column, vertical building would allow M. michaelseni (more sensitive to ambient humidity) to remove the requisite quantities of water-laden soil from the base of the nest without moving outside this environment, leading to tall, narrow mounds; M. natalensis, not confined to a central column of high humidity, could build more broadly outward horizontally.

Because both species rely on fungal harvesting for survival, it has been proposed that their mound structures are tuned towards producing optimal conditions for the health of the fungal colony (Wood and Thomas, 1989). In a similar manner to the observed effect of nest shape on temperature in M. bellicosus (Korb and Linsenmair, 1999), the tall towers and large internal pipe structures of M. michaelseni mounds could create a different internal humidity setpoint than the squat M. natalensis mounds. Whether these different conditions are preferential for cultivation of distinct symbionts is unknown, but M. natalensis is associated with a single lineage of Termitomyces (Nobre et al., 2011), even though in general fungal diversity in Macrotermes is high and likely dependent on horizontal transmission to new colonies (Osiemo et al., 2010). Additionally, the fungus combs of M. michaelseni produce mushrooms that erupt from the mound to release spores, whereas those of M. natalensis do not (Van Der Westhuizen and Eicker, 1991). It has been hypothesized that this is due to contrasting farming methods between Macrotermes species (de Fine Licht et al., 2006); however, it is possible, instead, that this discrepancy in the production of fruiting bodies might not result from direct termite husbandry, but rather from different nest humidity conditions arising from the mound structure. In turn, this could either promote or suppress different strains of Termitomyces, or even activate or suppress different stages of the fruiting cycle of a single strain. It is worth noting that M. michaelseni and M. subhyalinus are known to share Termitomyces strands (Osiemo et al., 2010; Tilahun et al., 2012), and both build tall mounds (>2 m) with spires or turrets.

It is clear from these experiments that humidity – itself a function of water table seepage, evaporative patterns, fungal metabolism and the actions of the termites themselves – is a key factor likely to influence termites in their role as ecosystem engineers. Another study from our laboratory analyzes the differential response of M. michaelseni alone to humidity conditions (P.B., N.E.C., D.S.C., R. Soar, J.S.T., R.N. and J.W., unpublished data). Whether these observed microscale building differences can be directly linked to macroscale structural features of mounds has not yet been demonstrated, but these experiments represent a promising direction for further investigation. We note in particular that other ‘pairs’ of related termite species occupying the same geography but building dissimilar mounds have been observed on multiple continents (Darlington, 1997; Araújo et al., 2017; Coles de Negret and Redford, 1982); investigating behavioural thresholding between these species may yet give us more insights into the design and function of mound morphology.

We thank the Institute of Quantitative Social Sciences at Harvard for consultation on statistical methodology, and the Cheetah Conservation Fund for use of their facilities.