Agricultural landscape simplification affects wild plant fitness indirectly through herbivore-mediated changes in floral display

As natural landscapes are modified and converted into simplified agricultural landscapes, the community composition and interactions of organisms persisting in these modified landscapes are altered. While many studies examine the consequences of these changing interactions for crops, few have evaluated the effects on wild plants. Here, we examine how pollinator and herbivore interactions affect fitness for wild resident and phytometer plants at sites along a landscape gradient ranging from natural to highly simplified. We tested the direct and indirect effects of landscape composition on plant traits and fitness mediated by insect interactions. For phytometer plants exposed to herbivores, we found that greater landscape complexity corresponded with elevated herbivore damage, which reduced total flower production but increased individual flower size. Though larger flowers increased pollination, the reduction in flowers ultimately reduced plant fitness. Herbivory was also higher in complex landscapes for resident plants, but overall damage was low and therefore did not have a cascading effect on floral display and fitness. This work highlights that landscape composition directly affects patterns of herbivory with cascading effects on pollination and wild plant fitness. Further, the absence of fitness consequences for resident plants suggests that they may be adapted to their local insect community. 

Methods
This study was conducted in upstate NY, USA. Experimental sites were in field margins of farms that were selected along a gradient ranging from landscapes dominated by natural land cover to landscapes dominated by agriculture. Phenotypic data, pollinator visitation, herbivore damage, and plant fitness was measured on Barbarea vulgaris plants existing naturally at each site (resident plants) and on sentinel plants grown from a single parent plant (phytometer plants). Phytometer plants provide a useful tool for evaluating changing interactions for wild plants by minimizing the contribution of adaptive trait variation which, in resident populations, may be compensating for landscape mediated changes in insect interactions. To analyze the data collected we examined how insect interactions and plant traits varied along the landscape gradient and how these variables interacted using structural equation models.