Ecology, 96(3), 2015, pp. 642–653 Ó 2015 by the Ecological Society of America A late-Quaternary perspective on atmospheric pCO 2 , climate, and fire as drivers of C 4 -grass abundance MICHAEL A. URBAN, 1 DAVID M. NELSON, 2 F. ALAYNE STREET-PERROTT, 3 DIRK VERSCHUREN, 4 AND FENG SHENG HU 1,5,6,7 1 Program in Ecology, Evolution and Conservation Biology, University of Illinois, 505 South Goodwin Ave, Urbana, Illinois 61801 USA 2 University of Maryland Center for Environmental Science, Appalachian Laboratory, Frostburg, Maryland 21532 USA 3 Department of Geography, College of Science, Swansea University, Swansea SA2 8PP United Kingdom 4 Limnology Unit, Department of Biology, Ghent University, B-9000 Ghent, Belgium 5 Department of Plant Biology, University of Illinois, Urbana, Illinois 61801 USA 6 Department of Geology, University of Illinois, Urbana, Illinois 61801 USA Abstract. Various environmental factors, including atmospheric CO 2 ( pCO 2 ), regional climate, and fire, have been invoked as primary drivers of long-term variation in C 4 grass abundance. Evaluating these hypotheses has been difficult because available paleorecords often lack information on past C 4 grass abundance or potential environmental drivers. We analyzed carbon isotope ratios (d 13 C) of individual grains of grass pollen in the sediments of two East African lakes to infer changes in the relative abundance of C 3 vs. C 4 grasses during the past 25 000 years. Results were compared with concurrent changes in pCO 2 , temperature, moisture balance, and fire activity. Our grass-pollen d 13 C analysis reveals a dynamic history of grass-dominated vegetation in equatorial East Africa: C 4 grasses have not consistently dominated lowland areas, and high-elevation grasses have not always been predominantly C 3. On millennial timescales, C 4 grass abundance does not correlate with charcoal influx at either site, suggesting that fire was not a major proximate control of the competitive balance between C 3 and C 4 grasses. Above the present-day treeline on Mt. Kenya, C 4 grass abundance declined from an average of ;90% during the glacial period to less than ;60% throughout the Holocene, coincident with increases in pCO 2 and temperature, and shifts in moisture balance. In the lowland savanna southeast of Mt. Kilimanjaro, C 4 grass abundance showed no such directional trend, but fluctuated markedly in association with variation in rainfall amount and seasonal-drought severity. These results underscore spatiotemporal variability in the relative influence of pCO 2 and climate on the interplay of C 3 and C 4 grasses and shed light on an emerging conceptual model regarding the expansion of C 4 -dominated grasslands in Earth’s history. They also suggest that future changes in the C 3 /C 4 composition of grass-dominated ecosystems will likely exhibit striking spatiotemporal variability as a result of varying combinations of environmental controls. Key words: atmospheric pCO 2 ;C 3 grasses; C 4 grasses; carbon isotopes; climate change; East Africa; fire; grasslands. INTRODUCTION Grasslands and savannas cover about one-third of Earth’s land surface, influence global biogeochemical processes, and provide food and shelter for humans and herbivores (Jacobs et al. 1999, Gibson 2009). These ecosystems are projected to be highly sensitive to future changes in climate and atmospheric chemistry (IPCC 2014), partly because of the differential responses of C 3 and C 4 plants. Field observations and model simula- tions indicate that the partial pressure of atmospheric CO 2 ( pCO 2 ), temperature, rainfall, and fire all variably influence the proportions of C 3 and C 4 plants in these ecosystems (e.g., Ripley et al. 2010, Lehmann et al. 2011, Morgan et al. 2011, Staver et al. 2011, Higgins and Scheiter 2012). However, the relative importance of these factors is uncertain, and studies of present-day ecosystems are limited to providing a snapshot of vegetation–environment relationships (Bond and Midgley 2012). High-quality paleorecords with inde- pendent evidence for shifts in climate, vegetation, and fire have the potential to elucidate how various environmental factors influence the interplay between C 3 and C 4 plants under the wide range of conditions that have occurred in the past and are anticipated for the future. Because of the differential influence of pCO 2 on C 3 and C 4 quantum yields of photosynthesis (Ehleringer 1978), variation in C 3 - and C 4 -plant abundances in the paleorecord was widely thought to be driven primarily Manuscript received 30 January 2014; revised 29 July 2014; accepted 30 July 2014; final version received 2 September 2014. Corresponding Editor: F. A. Smith. 7 Corresponding author. E-mail: fshu@life.illinois.edu 642