Data from: Nontarget herbivory by a weed biocontrol insect is limited to spillover, reducing the chance of population-level impacts

Insects approved for classical biocontrol of weeds are often capable of
using close relatives of their target weed for feeding, oviposition, or
larval development, with reduced preference and performance. When
nontarget herbivory occurs and is suspected to reduce survival, growth, or
fecundity of individual plants, and insects are capable of reproducing on
their nontarget host, characterization of spatial and temporal patterns of
the occurrence and intensity of herbivory is valuable for predicting
potential population-level effects. Here, we perform a novel post-release
manipulative field experiment with a root-feeding biocontrol weevil,
Mogulones crucifer, released in Canada to control the rangeland weed
Cynoglossum officinale, to test for its ability to establish on the
nontarget plant Hackelia micrantha. After Cynoglossum, M. crucifer
exhibits its highest preference for and performance on Hackelia spp. We
released M. crucifer on Canadian rangeland sites with naturally occurring
populations of H. micrantha growing interspersed with the target weed or
in the near absence of the target weed. Adult weevil feeding on
surrounding plants was monitored for three summers after release (years 0,
1, and 2), and, subsequently, subsets of plants were destructively sampled
to determine M. crucifer oviposition levels. Additional oviposition and
larval development data were obtained from seven non-experimental sites
where weevils were released zero, three, or four years earlier. M.
crucifer was not detected on experimental sites without C. officinale
after two years, and nontarget herbivory was restricted to rare, low-level
spillover. Visible evidence of adult herbivory (i.e., scars on shoots) was
associated with oviposition in 90% of targets but only 30% of nontarget
plants. We infer, through ecological refuge theory, that nontarget
population-level impacts from M. crucifer spillover are unlikely because
of temporal, spatial, and probabilistic refuges from herbivory, and make
recommendations for monitoring and management of biocontrol systems with
similar attributes, such as removing target plants around nontarget
populations of interest. Because M. crucifer is among the least
host-specific of the modern weed biocontrol agents, and H. micrantha is
likely one of its most highly preferred nontargets, these conclusions are,
arguably, generally applicable to other nontarget plants and biocontrol
systems.