Diet specialization mediates drivers of herbivory in a semi-arid agroecosystem.

Herbivory is a major fitness pressure for plants and a key driver of crop losses in agroecosystems. Dense monocultures are expected to favor specialist herbivorous insects, particularly those who primarily consume crop species; yet, levels and types of herbivory are not uniform within regional cropping systems. It is essential to determine which local and regional ecological factors drive variation in herbivory in order to support functional agroecosystems that rely less on chemical inputs. Crops in the genus host a suite of both generalist and specialist herbivores that inflict significant damage, yet little is known about the relative contribution of these herbivores to variation in herbivory and how local- and landscape-scale resource concentrations, management practices, and natural enemies mediate this relationship. In this study, we tested whether three foundational ecological hypotheses influenced herbivory across 20 pumpkin fields in the semi-arid Southern High Plains Region of Texas. We used generalized linear mixed models and confirmatory path analysis to assess whether the Density-dependent Herbivory Hypothesis, Resource Concentration Hypothesis, or the Natural Enemies Hypothesis, could explain variation in herbivory and insect dynamics in the context of conventional agronomic practices. We found that herbivory increased over time, indicating that herbivores were causing sustained damage throughout the growing season. We also found that fields with higher local resources had lower herbivory, suggesting a resource dilution effect. Natural enemy communities were more abundant and taxonomically rich in sites with greater generalist herbivore abundance, though predator abundance declined over time, indicating that late-season crop fields are most at risk given high herbivory and low natural enemy-based control. Our findings also suggest that while local resource availability may drive the abundance and richness of arthropod communities, additional agronomic and phenological information is needed to anticipate herbivory risk in an agriculturally dominated landscape.