Print PageMounting evidence indicates that biodiversity loss frequently increases disease transmission. Although effects of species diversity on disease risk have been reasonably well studied in a range of disease systems, few studies have examined multihost plant pathogens occurring in naturally-growing, non-experimental communities. We conducted an observational, landscape-level analysis to assess the relationship between species diversity and disease risk in the exotic plant pathogen, Phytophthora ramorum, across the Big Sur ecoregion of coastal California. This system provides a unique model for studying diversity-disease relationships because the pathogen infects almost all woody plant species throughout the area, yet hosts exhibit asymmetric transmission and susceptibility. We examine two hypotheses regarding the effect of plant species diversity in this system: (1) an amplification effect where disease risk is greater in areas with higher diversity due to the generalist properties of P. ramorum, or in contrast, (2) higher diversity could lead to reduced disease risk due to a dilution effect of the two most competent host species (bay laurel and tanoak). To accommodate inherent spatial effects of the invasion process, we compare inference among a set of Bayesian hierarchical models with varying complexity: (1) a binomial generalized linear model (GLM), (2) a zero-inflated binomial GLM, and (3) a zero-inflated binomial GLM with a spatial random effect.
Results/Conclusions
A total of 10152 hosts among 279 plots were assessed for P. ramorum symptoms. Of these, 23% of hosts across 152 plots were considered infected following laboratory confirmation of P. ramorum in the respective plot. All three Bayesian models showed a negative relationship between species diversity and disease risk (a dilution effect), after accounting for the potentially confounding effects of host density and landscape context effects. The zero-inflated binomial GLM with the spatial effect had better model fit based on DIC criterion compared to the simpler zero-inflated binomial GLM and binomial GLM. Our finding of a dilution effect in the multihost P. ramorum plant pathogen system is in accordance with the overwhelming majority of diversity-disease risk studies on plant pathogens in agricultural and experimental grassland settings, and suggests that similar mechanisms may be occurring in more complex, natural ecosystems.
