Taxonomic uncertainty and decision making for biosecurity: spatial models for myrtle/guava rust

The causal agent of myrtle rust, initially described as Uredo rangelii, was recorded in Australia for the first time in 2010. Much of the monitoring effort in Australia and elsewhere is driven by existing understanding of Puccinia psidii sensu lato (guava rust), because U. rangelii is part of the guava rust complex. Bioclimatic analyses for guava rust in Australia indicate highest risk along the eastern coast, from northern Queensland to the south coast of NSW. These analyses rely on native and invaded range records for Puccinia psidii sensu lato. However, models that are instead fitted to records representing U. rangelii emphasise different risk areas, with less focus on northern coastal Queensland and new predictions into NSW tablelands and north-eastern Victoria. These differences have important implications for biosecurity containment and monitoring efforts. Here we use this example as a case study to explore the implications of taxonomic uncertainty for predictions of the potential distribution of a species of biosecurity concern in Australia. The important message is that the different patterns implied by different taxonomic assumptions may have very different management implications and taxonomic uncertainty should be evaluated alongside other sources of uncertainty. We could find no published example where taxonomic uncertainty was evaluated in biosecurity planning, despite the fact that taxonomic uncertainties are common. We outline how to model such uncertainties and discuss methods for exploring them and their impacts on predicted distributions.