IPCC Fourth Assessment Report: Climate Change 2007
Climate Change 2007: Working Group II: Impacts, Adaptation and Vulnerability

4.4.3 Grasslands and savannas

Properties, goods and services

Dominated by a spatially and temporally variable mix of grass and tree-growth forms (Sankaran et al., 2005), grasslands and savannas include tropical C4 grasslands and savannas (C4 grass-dominated with 10-50% tree cover, about 28 Mkm2) and temperate C4 and/or C3-grass and herb-dominated grasslands (15 million km2; Bonan, 2002). Generally rich in grazing, browsing and other fauna (especially but not only in Africa), these systems are strongly controlled by fire (Bond et al., 2005) and/or grazing regimes (Scholes and Archer, 1997; Fuhlendorf et al., 2001). Disturbance regimes are often managed (e.g., Sankaran, 2005), although fire regimes depend also on seasonality of ignition events and rainfall-dependent accumulation of flammable material (Brown et al., 2005b). Temperate and tropical systems provide somewhat distinct goods and services. Temperate grasslands contain a substantial soil carbon pool, are important for maintaining soil stability and provide fodder for wild and domestic animals. Tropical savanna systems possess significant wild faunal diversity that supports nature-based tourism revenue (both extractive and non-extractive) and subsistence livelihoods (food, medicinal plants, and construction material), in addition to cultural, regulating and supporting services.

Key vulnerabilities

The structure, productivity and carbon balance of these systems appear more sensitive than indicated in the TAR to variability of, and changes in, major climate change drivers. The direct CO2-fertilisation impact and warming effect of rising atmospheric CO2 have contrasting effects on their dominant functional types (trees and C3 grasses may benefit from rising CO2 but not from warming; C4 grasses may benefit from warming, but not from CO2-fertilisation), with uncertain, non-linear and rapid changes in ecosystem structure and carbon stocks likely. Carbon stocks are very likely to be strongly reduced under more frequent disturbance, especially by fire, and disturbance and drought impacts on cover may exert regional feedback effects. On balance, savannas and grasslands are likely to show reduced carbon sequestration due to enhanced soil respiratory losses through warming, fire regime changes and increased rainfall variability, but possible regional gains in woody cover through direct CO2-fertilisation, and increased plant carbon stocks, cannot be excluded. Scientific predictive skill is currently limited by very few field-based, multi-factorial experiments, especially in tropical systems. Projected range shifts of mammal species will be limited by fragmented habitats and human pressures, as suggested in the TAR, with declines in species richness likely, especially in protected areas. Because of the important control by disturbance, management options exist to develop adaptive strategies for carbon sequestration and species conservation goals.


Ecosystem function and species composition of grasslands and savanna are likely to respond mainly to precipitation change and warming in temperate systems but, in tropical systems, CO2-fertilisation and emergent responses of herbivory and fire regime will also exert strong control. Very few experimental approaches have assessed ecosystem responses to multi-factorial treatments such as listed above (Norby and Luo, 2004), and experiments on warming, rainfall change or atmospheric CO2 level are virtually absent in savannas, with many ecosystem studies confined mainly to temperate grasslands (Rustad et al., 2001).

Rainfall change and variability is very likely to affect vegetation in tropical grassland and savanna systems with, for example, a reduction in cover and productivity simulated along an aridity gradient in southern African savanna in response to the observed drying trend of about 8 mm/yr since 1970 (Woodward and Lomas, 2004a). Sahelian woody plants, for example, have shown drought-induced mass mortality and subsequent regeneration during wetter periods (Hiernaux and Turner, 2002). Large-scale changes in savanna vegetation cover may also feed back to regional rainfall patterns. Modelled removal of savannas from global vegetation cover has larger effects on global precipitation than for any other biome (Snyder et al., 2004) and, in four out of five savannas studied globally, modelled savanna-grassland conversion resulted in 10% lower rainfall, suggesting positive feedback between human impacts and changing climate (Hoffmann and Jackson, 2000). At the continental scale, modelled forest-savanna conversion reduced rainfall in tropical African regions, but increased it in central southern Africa (Semazzi and Song, 2001).

Changing amounts and variability of rainfall may also strongly control temperate grassland responses to future climate change (Novick et al., 2004; Zha et al., 2005). A Canadian grassland fixed roughly five times as much carbon in a year with 30% higher rainfall, while a 15% rainfall reduction led to a net carbon loss (Flanagan et al., 2002). Similarly, Mongolian steppe grassland switched from carbon sink to source in response to seasonal water stress, although carbon balance was neutral on an annual basis (Li et al., 2005). Non-linear responses to increasing rainfall variability may be expected, as ecosystem models of mixed C3/C4 grasslands show initially positive NPP relationships with increasing rainfall variability, but greater variability ultimately reduces both NPP and ecosystem stability even if the rainfall total is kept constant (Mitchell and Csillag, 2001). Empirical results for C4 grasslands confirm a similar monotonic (hump-backed) relationship between NPP and rainfall variability (Nippert et al., 2006). Increased rainfall variability was more significant than rainfall amount for tall-grass prairie productivity (Fay et al., 2000, 2002), with a 50% increase in dry-spell duration causing 10% reduction in NPP (Fay et al., 2003) and a 13% reduction in soil respiration (Harper et al., 2005).

The CO2-fertilisation and warming effect of rising atmospheric CO2 have generally opposite effects on savanna- and grassland-dominant functional types, with CO2-fertilisation favouring woody C3 plants (Ainsworth and Long, 2005), and warming favouring C4 herbaceous types (Epstein et al., 2002). Simulated heat-wave events increased C4 dominance in a mixed C3/C4 New Zealand grassland within a single growing season, but reduced productivity by over 60% where C4 plants were absent (White et al., 2000b). Some African savanna trees are sensitive to seasonal high air temperature extremes (Chidumayo, 2001). North American forest vegetation types could spread with up to 4°C warming; but with greater warming, forest cover could be reduced by savanna expansion of up to 50%, partly due to the impacts of fire (Bachelet et al., 2001).

Elevated CO2 has important effects on production and soil water balance in most grassland types, mediated strongly by reduced stomatal conductance and resulting increases in soil water (Leakey et al., 2006) in many grassland types (Nelson et al., 2004; Niklaus and Körner, 2004; Stock et al., 2005). In short-grass prairie, elevated CO2 and 2.6°C warming increased production by 26-47% , regardless of grass photosynthetic type (Morgan et al., 2001a). In C4 tropical grassland, no relative increase in herbaceous C3 success occurred in double-ambient CO2 (Stock et al., 2005). Regional climate modelling indicates that CO2-fertilisation effects on grasslands may scale-up to affect regional climate (Eastman et al., 2001).

Differential effects of rising atmospheric CO2 on woody relative to herbaceous growth forms are very likely (Bond and Midgley, 2000). Trees and shrubs show higher CO2 responsiveness than do herbaceous forms (Ainsworth and Long, 2005). Savannas may thus be shifting towards greater tree dominance as atmospheric CO2 rises, with diminishing grass suppression of faster-growing tree saplings (Bond et al., 2003). Simulations suggest that rising CO2 may favour C3 forms at the expense of African C4 grasses (Thuiller et al., 2006b), even under projected warming. Continuing atmospheric CO2 rise could increase the resilience of Sahelian systems to drought (Wang and Eltahir, 2002). However, without definitive tests of the CO2-fertilisation effect on savanna trees, other factors can be invoked to explain widely observed woody plant encroachment in grassland systems (Van Auken, 2000).

Above-ground carbon stocks in savannas are strongly contingent on disturbance regimes. Australian savanna systems are currently a net carbon sink of 1-3 t C/ha/yr, depending on fire frequency and extent (Williams et al., 2004b). Fire exclusion can transform savannas to forests (e.g., Bowman et al., 2001), with an upper (albeit technically unfeasible) global estimate of potential doubling of closed forest cover (Bond et al., 2005). Thus savanna structure and carbon stocks are very likely to be responsive to both individual and interactive effects of the disturbance regime (Bond et al., 2003; Sankaran et al., 2005) and atmospheric CO2 change (Bond and Midgley, 2000).

There are few factorial experiments on multiple changing factors, but they suggest interactions that are not predictable from single factor experiments – such as the dampening effect of elevated CO2 on California C3 grassland responses to increased rainfall, nitrate and air temperature (Shaw et al., 2002). Increasing temperature and rainfall changes are seen to override the potential benefits of rising CO2 for C3 relative to C4 grasses (Winslow et al., 2003), and European C3 grassland showed minor responses to a 3°C rise in temperature, possibly due to concomitant drying impacts (Gielen et al., 2005). Elevated CO2 impacts on grassland carbon sequestration also seem to be dependent on management practices (Harmens et al., 2004; Jones and Donnelly, 2004), and are complicated by being species- but not functional-type specific (Niklaus et al., 2001; Hanley et al., 2004).

Soil-mediated responses are important in biogeochemical controls of vegetation response. Long-term CO2 enrichment of southern African C4 grassland revealed limited impacts on nitrogen cycling and soil C sequestration (Stock et al., 2005), in contrast to greater C sequestration in short-term studies of grassland ecosystems (e.g., Williams et al., 2004a). Likewise, elevated CO2 impacts on litter decomposition and soil fauna seem species-specific and relatively minor (Ross et al., 2002; Hungate et al., 2000). Warming of a tall-grass prairie showed increased plant growth that supported enhanced soil fungal success (Zhang et al., 2005). However, complex interactions between plants and fungal symbionts showed potential impacts on soil structure that may predispose them to accelerated erosion (Rillig et al., 2002). Soil respiration shows approximately 20% increase in response to about 2.4°C warming (Norby et al., 2007), although acclimatisation of soil respiration (Luo et al., 2001) and root growth (Edwards et al., 2004) to moderate warming has also been observed. Soil carbon loss from UK soils, many in grasslands, confirm carbon losses of about 2% per annum in carbon-rich soils, probably related to regional climate change (Bellamy et al., 2005). In an African savanna system, rainfall after a dry spell generates substantial soil respiration activity and soil respiratory carbon losses (Veenendaal et al., 2004), suggesting strong sensitivity to rainfall variability.

Climate change impact studies for savanna and grassland fauna are few. The proportion of threatened mammal species may increase to between 10 and 40% between 2050 and 2080 (Thuiller et al., 2006a). Changing migration routes especially threaten migratory African ungulates and their predators (Thirgood et al., 2004). Observed population declines in three African savanna ungulates suggest that summer rainfall reductions could result in their local extirpation if regional climate change trends are sustained (Ogutu and Owen-Smith, 2003). For an African arid savanna raptor, population declines have been simulated for drier, more variable rainfall scenarios (Wichmann et al., 2003). A 4 to 98% species range reduction for about 80% of mainly savanna and grassland animal species in South Africa is projected under an IS92a emissions scenario (Erasmus et al., 2002).