In grasslands and forests worldwide, experiments show that local species loss decreases ecosystem functioning and stability. Yet, as Earth enters its sixth mass extinction event, it remains difficult to quantitatively predict the consequences of biodiversity loss at the large spatial and temporal scales most relevant to conservation and policy. In this talk, I will present two complementary research themes that help bridge this gap. First, I will show how we can scale up insights from local biodiversity experiments to landscapes, regions, and global patterns. This work demonstrates that the impacts of biodiversity loss on ecosystem functioning can intensify at larger scales, due to emergent effects of beta diversity. Second, I will show how temporal stability can be predicted from its resistance and recovery components at the ecosystem and species levels. These results suggest that temporal stability can often be predicted from resistance alone, and that resistance to drought can be forecasted by monitoring temporal stability. Together, these findings help build a predictive biodiversity science that informs conservation and policy in a rapidly changing world.