The Termite Wars: How tiny insect engineers shape the African savanna

Review written by Jarome Ali (EEB, G5)

Music by Three Music

On the savannas of Kenya, a battle has been waged for centuries. The landscape hints at how this battle has shaped the entire ecosystem, but it must be viewed from far above. From just a few meters above ground level, the telltale signs are still invisible. However, from the vantage point offered by drone photography or satellite imagery, a clear pattern emerges. Patches of vegetation are spotted across the savanna, in a regular hexagonal layout. This kind of order in the natural world fascinates biologists and begs for an explanation. Researchers at Princeton have been investigating how warring termite colonies (or insect colonies in general), and the underlying resource distribution can drive the emergence of order in the savanna landscape. 

When scientists first noted the presence of these vegetation patterns, two main explanations arose for their origin. The more popular explanation, backed by mathematical theory and computer simulations, was scale-dependent feedbacks (SDFs). In SDFs, the nature of interactions depend on the distances at which they take place. In the case of savannas, the proposed explanation is that plants near to each other facilitate each other’s growth by enhancing filtration of water into the soil, while at larger scales the dominating effect is competition for water. SDFs go back to Alan Turing’s reaction-diffusion models and can produce patterns that match everything from the reticulation of giraffe colors to the distribution of mussels on the seabed. But could other biotic forces be at work here? 

Another explanation was hiding in the literature for years–the patterns seen in the savanna might be driven by an unseen force, the humble termite. How could termites be involved in patterning the landscape? Surely, dynamics between plants and water were more important. The answer lay in a fundamental ecological interaction: competition, the termite wars. Territorial competition between adjacent termite colonies sets a limit on how close together colonies can be and could result in a regular pattern of termite nests. During nest construction, termites change the composition of the soil, for example, adding or removing sand. This action alters water penetration into the soil. As they bring food to the nest, the fungi that they farm break down this plant material, enriching the nutrients available in the soil surrounding the nest. This process creates an oasis for plants, and results in the patches of green seen in the savanna. 

SDFs and termite activity are not mutually exclusive explanations. Previous work by Princeton Associate Professor Corina Tarnita and colleagues sought to unify these explanations. By incorporating competition between termite colonies into a theoretical model for SDFs, they were able to produce patterns that were strikingly similar to the distribution of termite mounds in Kenyan savannas. 

While this was a satisfying theoretical finding, there were still open questions. Firstly, do the termite wars implemented in the model really occur? Another kink in the theory: the pattern wasn’t the same everywhere. In some places a near perfect hexagonal grid exists, but in others you really have to squint to see any degree of regularity. Recent work by Tarnita’s research group, led by Postdoctoral researcher Jessica Castillo Vardaro has addressed these questions. 

First, the researchers asked the simple question: do termites really fight? By placing wild-caught termites from different colonies into a mini battle arena, they found that termites do indeed fight, and to the death. This intense aggression was displayed to any non-nestmates, regardless of their species or whether the interloper was gathered from nearby or distant locations. 

Next, Castillo Vardaro et al. hypothesized that the distribution of resources across the landscape could make or break the regularity of the termite pattern. In the scenario they imagined, colony size at a given resource level is limited by territory size. This happens because the amount of resources available to grow the colony depends on the area available for foraging. If this is true, then a colony of a given size would require a smaller territory in a resource-rich area vs in a resource-poor area. The researchers set out to test these hypotheses by performing large-scale measurements of termite colony distribution and size at Mpala Research Centre in Kenya. At Mpala, farmers keep their livestock in corrals and abandoned corrals act as areas of elevated resource availability. These farming practices create a heterogeneous resource landscape, ideal for testing the researchers’ hypothesis. In line with their predictions, Castillo Vardaro et al. found that that colonies were further apart when they were larger, supporting the idea that large colonies need large foraging areas. They also found that colonies were packed more densely near the resource-rich abandoned corrals, suggesting that territory sizes do respond to resource levels.

Now that the natural world had given up its secrets, the next step was to take the newfound knowledge to the virtual world. The researchers built on their previous model by adding resource gradients and allowing termite territory sizes to vary depending on resource density. The updated model produced results that mirror what is seen in the real world. A uniform resource distribution produces colonies that are pleasingly spaced out in a hexagonal lattice. On the other hand, as the resource distribution becomes more patchy, the regularity of this pattern falls apart. 

In addition to explaining the fascinating natural phenomenon of hexagonally distributed termite colonies, this study has important implications for how landscapes are managed. Complementing previous empirical work, the researchers found that their model predicts increased ecosystem productivity when the termite mounds are regularly spaced. That is, regularly spaced termite colonies means that plant biomass is higher and herbivores are more abundant. The presence of areas of high resource availability, especially when these areas are regularly distributed, further heightens ecosystem productivity. This finding suggests that encouraging the regular distribution of corrals can have positive impacts on the landscape.

Castillo Vardaro et al.’s work shows how patterns that give the impression of being designed can arise from interactions between animals and their environment. This work also demonstrates that, by understanding the dynamics of landscape-level patterns, we can more efficiently manage landscapes that are increasingly subject to human influence.

Dr. Tarnita explains that this research has been something that she has been thinking about for a while, but it took much longer than expected. “It really took someone like Jess [Castillo Vardaro],” she explained. “We thought it was going to be so hard, that it wouldn't work out. But she said ‘I'm already doing it.’ ...It took that kind of determination to try to get to empirical bits of it.”


Castillo Vardaro, Jessica A., et al. “Resource Availability and Heterogeneity Shape the Self‐organisation of Regular Spatial Patterning.” Ecology Letters, edited by Vojtech Novotny, vol. 24, no. 9, Sept. 2021, pp. 1880–91.

Tarnita, Corina E., et al. “A Theoretical Foundation for Multi-Scale Regular Vegetation Patterns.” Nature, vol. 541, no. 7637, Jan. 2017, pp. 398–401.

Pringle, Robert M., and Corina E. Tarnita. “Spatial Self-Organization of Ecosystems: Integrating Multiple Mechanisms of Regular-Pattern Formation.” Annual Review of Entomology, vol. 62, no. 1, Jan. 2017, pp. 359–77.

Pringle, Robert M., et al. “Spatial Pattern Enhances Ecosystem Functioning in an African Savanna.” PLOS Biology, vol. 8, no. 5, May 2010, p. e1000377. PLoS Journals,