Interplay between harvesting, planting density and ripening time affects coffee leaf rust dispersal and infection

Abstract

The relationship between the spatial structure of hosts and the individual movement of vectors is central to developing a comprehensive representation of pathogen dynamics. Here we focus on the dynamics of coffee leaf rust (CLR; Hemileia vastatrix), which is one of the principal coffee diseases around the globe. This fungus disperses between plants or regions through direct contact, water splash, or local turbulent wind conditions. Some studies have proposed that coffee harvesters can also bear CLR during harvesting. However, it is not clear how relevant these movements are for CLR dispersal and plot-level infection in plots with different planting densities. Besides, there are bidirectional interactions between the plantation characteristics (e.g. planting density and ripening synchronicity) and the movement of harvesters that might modify their dispersal potential. Overall, there is a lack of mechanistic models to represent these interactions. Here we present a computational model to explore the role of coffee ripening synchronicity in reproducing the different movement trajectories observed in the field. We then evaluate how these harvesting trajectories modify rust dispersal and change CLR infection in plots with increasing planting densities and a constant contact-mediated rust dispersal mechanism. For both synchronization scenarios, the distribution of step lengths is positively skewed. Interplant asynchronous ripening and scarcity of trees with berries generate trajectories with medium to long steps contrary to scenarios with synchronous ripening and without scarcity. The harvest dispersal significantly increases coffee plot infection (up to 15%) compared to scenarios where only local mechanisms for CLR dispersal are present. This effect is maximal for trajectories with medium to long steps from the asynchronous scenario and in plantations with medium planting densities. In particular, medium-sized steps are more likely to create new foci and networks of contact-mediated infected plants. The interaction between the proportion and the size of infected networks results in a nonlinear behavior of rust increase due to harvest. Our results aim to spur discussion on practices that could reduce the impact of harvesting and specific trajectories, in scenarios where one can benefit from asynchronous maturation of berries and shaded plantations. Some examples of practices include reducing the medium to long-distance movements during the same day, for instance, by avoiding harvesting at the end of the harvest season when trees ripening asynchronicity and CLR levels are higher.