Consistency in impact assessments of invasive species is generally high and depends on protocols and impact types

Abstract

Impact assessments can help prioritising limited resources for invasive species management. However, their usefulness to provide information for decision-making depends on their repeatability, i.e. the consistency of the estimated impact. Previous studies have provided important insights into the consistency of final scores and rankings. However, due to the criteria to summarise protocol responses into one value (e.g. maximum score observed) or to categorise those final scores into prioritisation levels, the real consistency at the answer level remains poorly understood. Here, we fill this gap by quantifying and comparing the consistency in the scores of protocol questions with inter-rater reliability metrics. We provide an overview of impact assessment consistency and the factors altering it, by evaluating 1,742 impact assessments of 60 terrestrial, freshwater and marine vertebrates, invertebrates and plants conducted with seven protocols applied in Europe (EICAT; EPPO; EPPO prioritisation; GABLIS; GB; GISS; and Harmonia+). Assessments include questions about diverse impact types: environment, biodiversity, native species interactions, hybridisation, economic losses and human health. Overall, the great majority of assessments (67%) showed high consistency; only a small minority (13%) presented low consistency. Consistency of responses did not depend on species identity or the amount of information on their impacts, but partly depended on the impact type evaluated and the protocol used, probably due to linguistic uncertainties (pseudo-R2 = 0.11 and 0.10, respectively). Consistency of responses was highest for questions on ecosystem and human health impacts and lowest for questions regarding biological interactions amongst alien and native species. Regarding protocols, consistency was highest with Harmonia+ and GISS and lowest with EPPO. The presence of few, but very low, consistent assessments indicates that there is room for improvement in the repeatability of assessments. As no single factor explained largely the variance in consistency, low values can rely on multiple factors. We thus endorse previous studies calling for diverse and complementary actions, such as improving protocols and guidelines or consensus assessment to increase impact assessment repeatability. Nevertheless, we conclude that impact assessments were generally highly consistent and, therefore, useful in helping to prioritise resources against the continued relentless rise of invasive species.