Top-down and bottom-up interactions determine tree and herbaceous layer dynamics in savanna grasslands

Introduction The savanna biome is defined, broadly, as comprising ecosystems in which trees and grass coexist, where the grass layer is approximately continuous and the tree layer discontinuous (Scholes and Archer, 1997; Bond, 2008). This definition encompasses a diverse range of vegetation types, from savanna woodlands at the woody extreme to open savannas or even grasslands at the other. A growing body of evidence from small (Hennenberg et al., 2006) to large scales (Staver et al., 2011a; 2011b) suggests that savanna is an ecologically meaningful biome categorization. Where savannas transition to forests, savanna boundaries are abrupt and discontinuous, marked by thresholds in canopy cover associated with grass persistence (Hennenberg et al., 2006; Lloyd et al., 2008) and fire spread (Archibald et al., 2009; Staver et al., 2011b). The degree to which savannas differ qualitatively from grasslands at the biome level – whether at their arid, flooding, or elevational extremes – is less clear. Some argue that the transition from savanna to grassland is also abrupt and discontinuous (Hirota et al., 2011), while others suggest, based on similarities in ecological processes, that grasslands may represent the treeless extreme of savannas (Bond, 2008; Wakeling et al., 2012). Here, on the basis of these similarities in trophic ecology, we discuss interactions between top-down and bottom-up ecological processes in savannas and grasslands together, and hereafter we refer to the entire gradient as savanna grassland (sensu Scholes and Walker, 1993). Savanna grasslands represent a fundamental challenge for ecology in two ways. First, the climatic and edaphic variables that have historically been used to map global biome distributions (Holdridge, 1947; Whittaker, 1975) do a uniquely poor job in describing the limits of savanna grasslands (Whittaker, 1975; Bond et al., 2005). Second, the long-term coexistence of two vastly different functional types – trees and grasses – presents a challenge not just for savanna ecology (Scholes and Archer, 1997), but for ecology more generally (Hutchinson, 1959; Tilman and Pacala, 1994). These same challenges make savanna grasslands of particular interest in thinking about interactions between top-down and bottom-up ecological processes.