Non-target impacts of introduced parasitoids and validation of probabilistic risk assessment for biological control introductions

Abstract

Field surveys and partial life table studies were conducted to assess apparent mortality and marginal attack rate, respectively. Field surveys of larvae were conducted at eight different sites throughout the Hawaiian Islands, parasitism rates by individual parasitoid species varied significantly among study sites of varying ecological conditions. Adventive parasitoids rather than purposely introduced ones were responsible for the majority of U. stellata apparent mortality. Results from the life table studies showed that predation) was the major larval mortality factor at all study sites and that parasitism contributed minimally to total mortality. The two purposely introduced parasitoids were present at high altitude, in relatively undisturbed sites.Multivariate analyses were used to detect patterns in species assemblage among sites. Udea stellata density, elevation, and level of habitat disturbance significantly explained variability in the parasitoid assemblage among sites. Most species increased in abundance with higher densities of U. stellata and some were restricted to less disturbed sites.This dissertation addressed non-target parasitism of the endemic Hawaiian moth Udea stellata (Butler) and validated a probabilistic risk assessment approach for biological control introductions. Udea stellata is distributed across a wide gradient of environmental conditions, which allowed an assessment of non-target effects under a range of ecological conditions. Seven parasitoid species were associated with U. stellata larvae. Trathala flavoorbitalis, Casinaria infesta and Triclistus nr. aitkeni are of adventive origin; Cotesia marginiventris and Meteorus laphygmae were purposely introduced to Hawaii; and Diadegma blackburni and Pristomerus hawaiiensis are of unknown origin. By comparing the use of single point estimates versus probability distributions in quantitative risk assessment modelling, it was demonstrated that the use of point estimates can hide important variability and significantly impact the estimates of risk. It was also demonstrated that, at least in this study system, the use of apparent mortality significantly increased the estimate of risk compared to the use of marginal attack rate.