Strengthening the application of species distribution models for three invasive insect species of trade importance

Xingyu Li

Growth of international trade in agricultural and horticultural products has seen an increasing risk of invasive insect introduction to Australia. These invasions pose a significant threat to Australian native ecosystems and agriculture and may result in substantial economic losses. Species Distribution Models (SDMs) have been widely used in biosecurity to determine potential areas that invasive pests might establish and thrive. As such, this thesis used several SDM approaches to predict potential suitable habitats for the internationally recognised insect pest Thaumatotibia leucotreta (Meyrick) (False codling moth) in Australia. Further, this work demonstrated that Fuzzy Logic was beneficial in highlighting favourable areas in Australia for a species complex of two invasive wood boring beetles, Euwallacea fornicatus (Eichhoff 1877) (Polyphagous shot hole borer, PSHB) and Euwallacea perbrevis (Schedl 1951) (Tea shot hole borer, TSHB), allowing a biogeographical range and relationship comparison between these two species with different prevalence. All these approaches can contribute to various aspects of biosecurity pest management, including area-freedom, surveillance, early detection and eradication. Thaumatotibia leucotreta is a polyphagous insect that is a serious pest in some African countries and could be introduced to Australia through trade in cut flowers, threatening many plants and commercial horticulture and agriculture. This thesis initially utilised Maxent, a popular SDM, to predict climatic suitability of Australian regions for T. leucotreta. Analysis considered not only widely used bioclimatic variables, but also temperature degree days of T. leucotreta. Moreover, this study identified habitat suitability in the areas of interest using overlay maps that combined climatic suitability and cropping land use. The results indicate that Australia is vulnerable to invasion and establishment by T. leucotreta, with high habitat suitability in the areas of western and southern Australia. These results provided important insights into the potential threat posed by the invasive pest T. leucotreta and can be used to enhance future biosecurity activities. This thesis strengthened SDM utility by evaluating the performance of six SDMs with T. leucotreta as an exemplar pest species. ‘Ensemble Modelling’ was also utilised, which weighted and combined the individual models based on their performance in order to further enhance prediction accuracy and reliability. Random Forest performed the best among the six SDMs investigated and even outperformed the Ensemble Model. These improved prediction maps indicated varying and overlapping habitat suitability across coastal regions of Australia through the utilization of different statistical methodologies in the modelling process. This highlights the necessity for appropriate SDM choice when investigating potentially invasive species, while also urging circumspection in the thoughtful application of the Ensemble Model. These model outcomes will direct decision-making by biosecurity policy-makers and managers regarding biosecurity activities and efficient allocation of resources. Euwallacea fornicatus and E. perbrevis are examples of cryptic species with complex life histories and taxonomy. E. perbrevis has been present in Queensland, Australia for more than ten years and E. fornicatus is a recent arrival in Western Australia. Both species are of significant biosecurity concern. Application of Fuzzy Logic to SDM is a useful approach for modelling multiple related species that may interact in positive or negative ways. This allows more effective comparison of distributional relationships between different species with varying prevalence, thereby facilitating determination of environmental favourability patterns for pest presence. This thesis applied the environmental favourability function to generated environmental favourability maps to analyse the biogeographic relationship between the two species. These analyses highlighted the differing potential threats to Queensland from E. fornicatus, as well as the lesser threat to Western Australia posed by E. perbrevis. Model performance was assessed using commonly used evaluation metrics, as well as ‘fuzzy entropy value’ and ‘Hosmer-Lemeshow test’, to provide a more reliable assessment. The research confirmed the benefits of utilising ‘environmental favourability function’ based on Fuzzy Logic with enhanced mapping standards and more informative predictions. The outcomes of this approach provided a straightforward and concise prediction and comparison of these two wood borers of biosecurity concern. Application of these approaches to other globally-important pest species could have far-reaching implications for management of inter-related species for biological control and biosecurity pest management.

Strengthening the application of species distribution models for three invasive insect species of trade importance

Files and links (1)

Abstract

Details

UN Sustainable Development Goals (SDGs)

Metrics