Explaining variation in plant-herbivore associational effects in a tree biodiversity experiment

Within biodiversity-ecosystem function research, a major outstanding question is how herbivory, a critical ecosystem function at the base of the food web, changes along gradients of plant biodiversity. Neighborhood-level associational effects are hypothesized to be a strong driver of biodiversity-herbivory relationships, but we lack a successful framework that explains the wide variation observed in the sign and magnitude of plant-herbivore associational effects, particularly in systems with mainly generalist herbivores. In this study, we combine measurements from a tree biodiversity field experiment with simulation to provide a framework for explaining variation in plant-herbivore associational effects, particularly when herbivores that feed on many different species (e.g., generalists) cause most damage. We show that monoculture herbivory levels of focal species and their neighbors predict the direction and strength of associational effects. We provide evidence that this may be due to a “spillover effect”, in which some insect herbivores attracted to focal individuals ultimately end up feeding on neighboring individuals. With an empirically parameterized simulation, we explain how spatial organization modifies biodiversity-ecosystem function relationships when associational effects operate. We suggest a set of experiments to test the generality of our conceptual framework, to elucidate the underlying mechanisms that produce the patterns we find, and to ultimately increase the predictability of plant-herbivore associational effects. We conclude by discussing how our results might inform pest management in diversified agroecosystems and reforestation sites.

Synthesis

Our results provide a potential framework for explaining why positive and negative plant-herbivore associational effects are often balanced in systems with primarily generalist herbivores and point to a path forward for predicting when increased plant biodiversity will be associated with increased, decreased, or unchanged levels of insect herbivory on individual plant species in such systems.

Methods
We include here six datasets. Three (mini_mydiv_herbivory.csv, mydiv_herbivory_may_2021.csv, and mydiv_herbivory_aug_2021.csv) contain leaf-level herbivory data. Columns in this data show the block, plot, diversity of the plot (1, 2, or 4), mycorrhizal characteristics of the plot (A for only arbuscular mycorrhizae, E for only ectomycorrhizae, and AE for both types), tree species, position of the individual within the plot, and branch on the individual. Additional columns show presence/absence of herbivory on each leaf due to leaf chewers, holefeeders, leafminers, gallers, leafrollers, and skeletonizers, as well as the presence of infection (rust or mildew). Additionally, in the August 2021 dataset, an additional column gives a quantitative herbivory index for each leaf, and the "perc_mean" column gives the midpoint of the % leaf area removed category to which that index corresponds to. 

sem_14.csv and sem_nopalat.csv contain leaf trait data collected in July and August 2021, averaged at the level of the tree individual (i.e., as the data was used in SEM's described in the manuscript). They show for each tree individual, the plot in which it was located, the species, the position within the plot, the pubescence on a 1 cm diameter disc on each side of the leaf and the sum of these two measurements, the average toughness across multiple leaves. Additionally, sem_14.csv contains data on leaf palatability derived from feeding trials with caterpillars. Since palatability trials were only conducted for leaves from diversity levels 1 and 4, this dataset only contains data from those diversity levels. sem_nopalat.csv omits palatability data, but contains data on the other traits for all three diversity levels.

Finally, tree_inventory_2021.csv contains dbh and height data (along with plot, position, and species identifiers) for trees across the MyDiv experiment in 2021. 

Additional methods on data collection can be found in the manuscript main text and supporting information.