| Summary: | Cost-effective reduction in the uncertainty surrounding global indicators of biodiversity change is a central goal of conservation. In this thesis, I identify and resolve the effects of IUCN Data Deficient species on the estimation of global patterns and levels of extinction risk. I show that gaps in our knowledge of species' conservation status are primarily driven by spatial patterns of ecological research (Chapter 2). Large numbers of species are extremely poorly known, highlighting the importance of basic taxonomic and natural history information in conservation assessments. Using sensitivity analyses (Chapter 3), I show that Data Deficient species contribute to considerable uncertainty in patterns of extinction risk in freshwater invertebrates, limiting our understanding of the factors influencing extinction risks and our capacity to design reliable conservation schemes. To determine the likely conservation status of Data Deficient species, I develop seven machine learning models based on species' life-history traits, niche and threat exposure (Chapter 4). I find that machine learning models accurately predict species conservation status and geographical patterns of threatened species richness. I predict 64% of Data Deficient mammals to be at risk of extinction, increasing the estimated proportion of threatened mammals from 22% to 27% globally. Finally, I use sampling theory to compare the cost-effectiveness of predictive models and IUCN Red List assessments in mammals, amphibians, reptiles and crayfish (Chapter 5). Double sampling with predictive models reduces the cost of determining the proportion of Data Deficient species at risk of extinction by up to 69%, and can be used to reduce the impact of uncertainty in the Red List and Red List Index. My thesis demonstrates how predictive models and decision theory can strengthen indicators of biodiversity change to monitor progress towards international biodiversity targets.
|
|---|
