The transmission of infectious disease agents inevitably have spatial geometry. In the case of the fungal pathogen of coffee, the coffee leaf rust (CLR), Hemelia vastarix, the spatial geometry of the pathogens underlying resource (coffee) is essential to understanding the dynamics of the system. We find it convenient to represent this phenomenon as a network model, with dynamic impositions creating the network structure. For the coffee plants themselves, it is a combination of intentional imposition of a regular lattice coupled with the pruning and natural deaths requiring replanting on a periodic basis. Subsequently, the CLR dynamics unfold on this dynamically constructed network, resulting in a new network. Finally, the consumer of the CLR, in this case the larva of a small fly Mycodiplosis engages with the network of the diseased plants. Thus, we have a multi-layer network, in which certain dynamic forces create a network structure on which the dynamic forces of disease transmission create a subsequent network of the coffee rust disease, which creates the ultimate pattern in which the consumer, Mycodiplosis, searches for its resource. We couple two years of field data with simulations to explore the usefulness of conceptualizing the system with this multi-layer spatial network approach.
Results/Conclusions
First, we quantify the spatial distribution of three empirical plots and compare them with a simple model that simulates the socio-ecological forces that determine the underlying plant geometry (planting, pruning and replanting coffee plants). We then investigate how the underlying spatial geometry of the plants influences the spreading dynamics of the pathogen with a simple model to simulate spread on the spatial networks. We find that more lattice-like geometries are far more sensitive to small changes in the scale of pathogen spread than geometries that approximate complete spatial randomness. Using two years of field data, we then develop statistical approach to approximate the spatial scale the pathogen spread in our empirical system and find that the scale of pathogen action is variable across plots. We suggest that the difference in the spatial scale of pathogen spread is likely shaped by management factors within the agroecosystem. Finally, we use data on the distribution of the consumer, Mycodiplosis, to approximate the spatial scale at which they navigate the pathogen in space. We find that their scale of response to the pathogen spatial network is consistent across our plots suggesting a fundamental spatial scale at which they interact with the pathogen.