Impacts of a weed biocontrol agent on recovery from water stress in a target and a non‐targetHypericumspecies

Summary 1. Biological control agents that impact on non‐target native species have limited desirability.Aculus hypericiwas introduced into Australia to help control the weed St John's wortHypericum perforatum, despite indications from pre‐release trials that this eriophyid mite could survive and reproduce on a non‐target native species,Hypericum gramineum. We used field experiments to explore the probability ofA. hypericicolonizing eitherH. gramineumorH. perforatumin sympatric populations of the alternative hosts, and to determine the impact ofA. hypericion the growth of field‐grownH. gramineum. 2. We used a glasshouse experiment to compare the impact ofA. hypericion the ability of each host to recover from water stress and to investigate how such recovery affects herbivore populations on the alternative hosts. As Australia is frequently affected by drought, this issue is critical to both the efficacy and safety of biological weed control. 3. The probability ofA. hypericicolonizingH. gramineumwas significantly lower than the probability of it colonizing the target weed. Despite the ability ofA. hypericito infestH. gramineum, the mite had negligible impacts on all measured indices of growth and reproduction. 4. During the water limitation experiment, the non‐target host supported populations ofA. hypericithroughout the experimental period, but population size was consistently low. In comparison, mite populations increased considerably on the target weed and were associated with significant decreases in growth. Water limitation compounded the adverse impacts of mites on the weed, but did not affect impacts on the non‐target native. 5. Despite the detrimental effects of mites on the growth ofH. perforatum, we found little evidence that they impaired the ability of eitherH. perforatumorH. gramineumto recover from water stress. 6. Synthesis and applications.Rigorous host‐specificity testing is necessary to ensure that biological weed control agents do not impact on non‐target species. Potentially, however, these tests may reject beneficial agents unnecessarily, because of low level feeding or reproduction on non‐target species. Although host‐specificity testing might indicate that non‐target native species may host biocontrol agents, our results demonstrate that such non‐target species are not, necessarily, damaged.