SignificanceFire is crucial to maintaining modern subtropical grasslands, yet the geologic and ecological history of this association is not well constrained. Here, we test the role of fire during the expansion of C4 grassland ecosystems in the Mio-Pliocene through innovative molecular proxies from ancient soils in Pakistan. We produce a synoptic terrestrial record of fire and vegetation change in this region, which indicates that increased fire occurrence accompanied two stages of landscape opening. Proxy data confirm that a pronounced fire–grassland feedback was a critical component of grassland ecosystems since their origination and fostered the rise of C4-dominated grasslands. The approach presented here can be used to examine landscape-scale interactions between paleofire and vegetation for other geographic regions and climatic transitions.

No abstract

1. Herbivory is a key process structuring vegetation in savannas, especially in Africa where large mammal herbivore communities remain intact. Exclusion experiments consistently show that herbivores impact savanna vegetation, but effect size variation has resisted explanation, limiting our understanding of the past, present and future roles of herbivory in savanna ecosystems.2. Synthesis of vegetation responses to herbivore exclusion shows that herbivory decreased grass abundance by 57.0% and tree abundance by 30.6% across African savannas.3. The magnitude of herbivore exclusion effects scaled with herbivore abundance: more grazing herbivores resulted in larger grass responses and more browsing herbivores in larger tree responses. However, existing experiments are concentrated in semi-arid savannas (400-800-mm rainfall) and soils data are mostly lacking, which makes disentangling environmental constraints a challenge and priority for future research.4. Observed herbivore impacts were ~2.1× larger than existing estimates modelled based on consumption. Wildlife metabolic rates may be higher than are usually used for estimating consumption, which offers one clear avenue for reconciling estimated herbivore consumption with observed herbivore impacts. Plant-soil feedbacks, plant community composition, and the phenological or demographic timing of herbivory may also influence vegetation productivity, thereby magnifying herbivore impacts.5. Because herbivore abundance so closely predicts vegetation impact, changes in herbivore abundance through time are likely predictive of the past and future of their impacts. Grazer diversity in Africa has declined from its peak 1 million years ago and wild grazer abundance has declined historically, suggesting that grazing likely had larger impacts in the past than it does today. 6. Current wildlife impacts are dominated by small-bodied mixed feeders, which will likely continue into the future, but the magnitude of top-down control may also depend on changing climate, fire and atmospheric CO 2 . 7. Synthesis. Herbivore biomass determines the magnitude of their impacts on savanna vegetation, with effect sizes based on direct observation that outstrip | 2805

During the Cenozoic (66 Ma to present), grasslands and open habitats replaced forests across large swaths of the tropics and subtropics (Jacobs et al., 1999;Strömberg, 2011), reshaping the vegetation structure and faunal communities in terrestrial ecosystems around the world (Anderson, 2006). In many regions, the late Miocene (∼11-5 Ma) conclusion of this ecological trajectory was the establishment of grasslands dominated by grass species using the C 4 photosynthetic pathway (Edwards et al., 2010). A substantial number of studies have mapped out when C 4 ecosystems became established around the globe. The earliest onset of C 4 ecosystem expansion occurred around 10