Although we acknowledge this system as complex, it makes some sense to try and simplify it a little to perhaps gain some deeper insight into its operation. A glance at figure 12 suggests that there are two generalized groups of ant predators on the berry borer: the big ants and the small ants . So we can think of it as a two-predator, one-prey system. But there is an obvious indirect effect because the bigger ants negatively affect the ability of the smaller ants to be predators, as was discussed above. Depending on nest densities, it would seem that because the smaller ants prey on all three life stages of the borer and throughout the year for those that forage in infected berries on the ground, they might, in the end, be more efficient predators than the larger ants. Furthermore, the larger ants have an indirect trait-mediated effect on the smaller ants, reducing their effectiveness. One might argue that it is the existence of the smaller ants that potentially regulates the borers over the long run. Unpublished evidence even supports the idea that the berry borer actually seeks areas that are under protection from the ants, supporting the speculation of Gonthier and colleagues that the berry borer gains protection from smaller predators by preferring to attack berries under the protection of larger ants. This basic speculation was put to an unintended test in 2012. Because of a major outbreak of the coffee rust disease , the coffee landscape where we work was heavily sprayed with a combination of calcium carbonate and copper sulfate, hydroponic bucket a permitted activity for organic agriculture. In figure 13, we show the distribution of P. synanthropica as a heat map based on how many tuna fish baits placed in coffee bushes had swarms of P. synanthropica after about 30 minutes.
It is clear that in a single year, a population of thousands of nests of P. synanthropica simply disappeared. Surveys in subsequent years indicated that the smaller ants in the system began to recuperate from the reduced state they had been in, apparently because of the indirect effects of P. synanthropica . Those small ants that attack the borer within the seed, both on the tree and on the ground, especially increased over the next few years. In figure 14, we show the distribution of several of the species in a 50 × 50 meter subplot within the 45-hectare plot. Note how, during the years 2009–2012, the distribution of P. synanthropica remained relatively constant, perhaps slowly increasing in its area of dominance, at the expense of P. protensa on the ground and S. picea arboreally. Then, after the collapse of P. synanthropica between 2012 and 2013, both of those smaller species began to move into the area previously dominated by P. synanthropica. If the above speculations about how the ant community affects the borer are true, we might expect that the elimination of one of the borer’s predators would result in better overall control of the berry borer. In surveys of the berry borer in 2005 and then repeated in approximately the same area in 2018, the attack rate of the borer went from an average of about 15% of berries infected with borers to less than 1%. Insect populations are notoriously variable and respond to many cues in the environment by increasing and decreasing population numbers, frequently in unpredictable ways. Therefore, although this dramatic decline in borer numbers cannot be directly linked to the change in the ant community structure, it is nevertheless worth noting that the underlying narrative of how that community functions as a system of biological control concords perfectly with the changes observed.In the early 1980s, a specter haunted the coffee growing regions of Central America.
The infamous coffee rust disease had arrived in Brazil, and its eventual spread all the way to Mexico was expected, causing extreme worry among farmers and technical advisors. This worry was certainly justified on the basis of the history of the coffee rust disease in Asia . Great Britain’s expansion in what was then called Ceylon was qualitatively distinct from many of its previous imperial adventures. Planting what was effectively a monoculture of coffee, along with a great deal of infrastructure for the time, it was a remarkable centrally planned agricultural development plan. However, the plan effectively created ideal conditions for any disease that could get a foothold, with its virtually shadeless monoculture and networks of roads and railroads that could help distribute the fungal spores widely. When the disease arrived, it took hold and spread throughout the entire island, eventually causing a complete loss of coffee production . However, the rust scare of the 1980s Mesoamerica turned out to be a bit of a false alarm, at least until 2012. Before that year, the rust was always an irksome constraint on production, but the complete devastation that had been feared when it was discovered in the early 1980s never came to pass; it was a problem, to be sure, but not one to get overly agitated about. But then, without much warning, there was an explosion of coffee rust in the 2012–2013 cycle. Countries in the zone declared emergencies as one of their main sources of income seemed to be threatened with severe disruption. Local governments throughout the affected area provided emergency support to coffee producers and both the United Kingdom and the United States came up with significant international aid, specifically for what rapidly came to be called the most devastating emergency in the history of coffee production throughout the region . There are two ecological questions associated with this episode. First, why did the disease not become rampant for approximately 30 years after its introduction, and, second, what caused the very sudden explosion?
Although the answers to either of these questions remain enigmatic, using tools from complexity science provides us with some ideas. The relevant biology of the rust disease is well known. A windblown spore adheres to the under surface of a leaf and encounters a small amount of moisture, causing germination directly into a stoma. The mycelia grow intercellularly and produce haustoria, which penetrate into the plant cell and absorb nutrients, effectively killing the cell. As the fungus grows within the leaf tissue it eventually forms uridia that contain new spores, exiting the leaf from other stomata, causing the characteristic yellow spots on the under surface of the leaves . The transmission dynamics of the disease are dual , stackable planters with some close plant-to-plant dispersion of spores , especially when plants are close enough to touch one another, along with propagule rain from the general spore load that exists in the atmosphere, especially in areas of high concentration of coffee production, when that coffee is attacked by the rust. From the perspective of an individual coffee plant, there are two sources of rust spores: its local neighbors and the general accumulation of spores in its region—that is, from the overall spore load in the atmosphere. But it is also the case that this coffee plant and all others over a very large region contribute to the spores in the overall spore rain from the atmosphere. Given this narrative, it is easy to imagine a situation in which a generally traditional shade coffee landscape would receive a particular rate of spore rain each year and would contribute a bit to the general pool, but because the shade trees act as windbreaks, much of the spore load is never delivered to the coffee plants. One could imagine an equilibrium in which the rust disease would be endemic but not severe, partly because the wind-borne spores have limited access to the coffee trees, meaning that the increment of spore load in the general atmosphere would be limited. Focusing on the large landscape level, if the abundance of spores in the atmosphere is low, it is likely that the incidence of the disease will also be low. But each epidemic will increase the spore density in the atmosphere. The probability that a given farm will become epidemic is a function of both the spore density in the atmosphere and the dispersion rate from the atmosphere to the farm. Changing focus to the local level, the rate of spread of spores from coffee bush to coffee bush on an average individual farm will partially determine whether the rust within that farm will become epidemic. From the point of view of an individual coffee bush, the danger of being infected by a spore comes from two sources: the atmosphere and neighboring infected plants—a regional source and a local source. Imagine that a forested ecosystem is gradually deforested of both shade trees in the coffee farms and the trees in the natural forest around them, and ask what proportion of the farms could be susceptible to an epidemic of coffee rust? According to a simple model that incorporates both regional and local dispersal , the initial deforestation will generate an increase in the number of farms experiencing an epidemic. That increase is likely to be slow and steady at first, but there will be a specific point at which a critical transition will occur and a large number of farms will suddenly become highly infected. This will happen in the complete absence of any other environmental driver, such as climate change or a new more virulent strain of the disease. Indeed, one study in Costa Rica showed that the incidence of rust disease was correlated with the amount of sun coffee and pasture in the surrounding landscape. It could very well be that the sudden outbreak of coffee rust in 2012 is an example of the inevitability of surprise arising from the formality of a critical transition that we have come to associate with highly nonlinear complex systems .
A cartoon version of this theoretical process is presented in figure 16. An important component of the rust disease system, not yet completely understood, is the existence of several natural enemies of the rust . Providing an example of the sorts of ecological complexity of popular literature, the fungal disease of the first pest we discussed, the green coffee scale , is caused by the same species of fungus that, when given the chance, attacks the coffee rust fungus. That same white halo fungus that attacks the green coffee scale, now acts as a mycoparasite . Because this natural enemy is also a natural enemy of the green coffee scale, the connection to the Azteca ant became obvious early on ; Azteca creates conditions under which the scale insect becomes highly concentrated locally, which attracts the infestation of the white halo fungus and creates local hot spots of spores that disperse locally and attack the rust. Correlative evidence for this hypothesis, prior to the 2012 epidemic of rust, comes from multiple sources . Indeed, there has been considerable discussion at international conferences on the potential of L. lecanii as a spray for the rust disease. Our work suggests that partial control of the rust may naturally occur through this and other agents , although the epidemic throughout Mesoamerica in the 2012–2013 growing season shows the potential for the disease to escape such control, if, in fact, it did exist before that. It is quite a remarkable qualitative impression one gets when examining the rust and its control comparatively. It is endemic but rarely epidemic in Puerto Rico but has maintained a relatively severe status in much of Mesoamerica since 2012 . Examining coffee leaves in Mexico easily reveals the presence of L. lecanii but only after considerable searching effort, whereas in Puerto Rico, it is almost inevitable that, if one encounters the rust on a leaf, it is almost certain that one encounters L. lecanii also. What seems epidemic in Puerto Rico is the L. lecanii that seems to keep the rust under control. In addition to the white halo fungus, the larval form of a small fly, Mycodiplosis hamaelae, preys on the spores directly on the coffee leaf but probably also acts as a dispersal agent, at least locally . The coffee leaf rust continues to plague Latin American coffee farmers.