How many hosts? Modelling host breadth from field samples

Abstract

Summary 1. Host‐specificity is important in fundamental and applied studies of biodiversity, but it is difficult to generate definitive data for many species. 2. Here, we present a model framework for estimating the realized host breadth of dependent fauna from samples collected from replicate host individuals. We use Bayesian, Zero‐inflated Poisson regression to model the abundance conditional upon host use – our focal parameter. We illustrate the model using a data set of insects on threatened and non‐threatened plant species from eastern Australia. Estimating co‐extinction risk through threats to host organisms is a key application for the model, requiring assessment of the host‐specificity of biota, whilst incorporating uncertainty. 3. The host breadth model indicated that several insect species had high probability of narrow host breadth, including several species that appeared to only have one host. Furthermore, the model distinguished polyphagous, oligophagous and monophagous insect species. Modification of the model to account for false positives also identified tourist occurrences. 4. The host breadth model can be utilized not only towards focusing conservation actions on the most co‐threatened dependent invertebrate species, but also for other projects involving host‐specificity, such as estimating the number of insects globally, and selecting the most likely candidates for biological control of invasive species. Further uses are likely in various data sets of associations between species.