Environmental and Evolutionary Correlates of Extinction Risk in a Global Biodiversity Hotspot.

Macro-evolutionary and ecological drivers of extinction risk in different organismal groups are not well understood. Some studies have indicated that certain clades may be predisposed to extinction. Through a comparative phylogenetic framework, we assessed whether there are significant evolutionary and environmental correlates of extinction risk for a species rich plant group (Pomaderreae tribe; Rhamnaceae). We focused on the southwest Australian global biodiversity hotspot, where most species of Pomaderreae occur. We also applied 10 different machine learning models to predict and impute data for species that lack flowering time information (18/87 taxa). The k-nearest neighbours algorithm was selected as the best model among the evaluated machine learning approaches because it demonstrated competitive predictive performance, strong recall performance and consistent monthly accuracy for our dataset. We show that larger genera have more threatened species but these species are not over-represented in larger genera. We also show that extinction risk is uncoupled with evolutionary history (diversification, speciation, extinction rates and lineage age). Shorter flowering periods and temperature variables (WorldClim) were shown to be significant correlates to extinction risk, but polyploidy and precipitation variables were not, even when accounting for phylogenetic relatedness. These findings suggest that threatened species would be more vulnerable to future climate change from their limited ability to respond. The decoupling of diversification rates and lineage age with current extinction risk suggests drivers of extinction risk are multifaceted and are likely predominantly driven by present-day threats rather than from the past.