Plant-herbivore interactions are the foundation of terrestrial food webs and are among the most ubiquitous of ecological associations. While plant diversity is positively correlated with insect diversity and affects insect community composition, changes in the structure of interactions between insects and their host plants across a plant diversity gradient are expected but not well understood. An ecological network approach examines relationships between community-level species interactions and ecosystem stability, and is a suitable framework for studying consequences of changing ecological interaction structure. The relationship between network structure and biodiversity is a point of longstanding debate in network studies, but has rarely been evaluated empirically. We sample arthropods and plants across three years, beginning in the fourth year of an experimental prairie restoration containing three treatment levels of initial plant seeding diversity (high diversity, low diversity and monoculture). Hemipteran and orthopteran herbivores were identified to species and an extensive literature review was used to identify feeding interactions and construct plant-herbivore networks. We then calculated four measures of network structure: connectance, modularity, generality of herbivore diets, and herbivore robustness to simulated plant extinction, and asked how these network structures changed with plant seeding diversity treatment and time.
Seeding diversity treatment strongly affected plant-herbivore interaction structure; the increased seeding diversity treatment decreased plant-herbivore connectance (n=36, F2,31=22.1, P<0.001), increased average herbivore diet generalism (n=36, F2,31=17.8, P<0.001), and increased herbivore robustness to plant species loss (n=36, F2,31=13.0, P<0.001). Treatments diverged over time with the last year of sampling showing the greatest difference between treatments in plant and arthropod species richness, and plant-herbivore network structure. The high seeding diversity treatment was the least variable temporally in network structure. However, this divergence between diversity treatments was not found when using traditional ordinations of community composition. While previous comparisons of plant-herbivore interaction networks across ecological gradients treated network size (biodiversity) as a confounding factor, we demonstrate that plant diversity directly shapes network structure in this grassland, and higher plant diversity maintains less temporally variable ecological networks of herbivores and their host plants.