SCOPE 56 - Global Change: Effects on Coniferous Forests and Grasslands

4.1 INTRODUCTION

The extensive zone between the evergreen tropical forests and the mid-latitude deserts is occupied by vegetation types containing both trees and grasses, in varying proportions (Figure 4.1). The significance of this zone to the global environment is based on five factors:

1. The vast extent of the area (a conservative estimate is 16.1 million km², or 11.5% of the global land surface) makes it a significant component of global element budgets and gaseous fluxes;

2. The seasonally dry climate permits extensive annual biomass burning;

3. The proportion of trees and grass is inherently unstable, and small changes in climate or land-use practice can lead to rapid changes in biomass and soil properties;

4. It lies mostly within the area of the developing world where population pressure and land-use changes will be greatest in the next few decades. The agro-pastoral communities in savannas are highly dependent on the natural vegetation, which is vulnerable to degradation;

5. The biodiversity of plants and animals in these areas is greater than is usually recognized.

4.1.1 Definitions of grasslands, savannas and related formations

The importance of the tropical tree-grass systems has been obscured in the past by the vagueness of vegetation type definitions. The woodier savanna variations are often grouped with tropical forests (for instance, in the FAO deforestation statistics), while the grassier forms are lumped with temperate grasslands. The problem arises because tropical drought-deciduous woodlands, savannas, shrub lands, thickets and grasslands form a seamless continuum which can only be divided into distinct structural types by applying arbitrary limits (Figure 4.2). In this chapter we have followed the following naming conventions:

1. Forests have complete tree canopy cover and three or more overlapping vegetation strata;

2. Woodlands have 50-100% tree canopy cover by trees, and a sometimes sparse, but always significant gramineous layer;

3. Savannas have 10-50% cover by woody plants, and in the unexploited state, a well-developed grass layer;

4. Grasslands have less than 10% tree cover.

Figure 4.1 The approximate global distribution of tropical grasslands, savannas and woodlands, derived from Olson et al. (1983). The black areas represent savannas and are a combination of Olson categories: 32 (drought-deciduous forest), 43 (savannas), 48 (primarily Eucalyptus) and 59 (thorn and succulent bush). The details of this map are far from accurate, but there is no consistently better map at a global scale. The Olson et al. (1983) data have been used throughout this chapter to maintain internal consistency 

There are also a variety of other formations which are related to, or subsets of, the above broad categories. Grasslands with 2-10% tree cover are often referred to as wooded grasslands. Low-statured ( < 3 m tall) savannas are often called bushlands or shrublands, while low-statured multi-stemmed woodlands or forests are called thicket or scrub forest. If the woody plants are very short ( < 1 m) the formation is usually called a dwarf shrubland.

Of the non-forest types, tropical grasslands have the smallest extent (probably less than 10% of the total area). Completely treeless tropical grasslands are confined to the following situations: soils which are intermittently waterlogged; areas with frequent (annual or biannual) intense fires, typically on fertile soils; soils which contain elements toxic to trees (typically metals); areas subject to frequent frosts (i.e. high altitudes); areas receiving very low rainfall ( < 100 mm yr^-1); and areas cleared of trees by human activity.

Figure 4.2 Grasslands, savannas and woodlands are part of a continuum of vegetation types on moisture and temperature gradients. The divisions on this continuum are arbitrary, which leads to great variations in the estimates by different authorities of the area occupied by each type.

Woodlands and savannas are characterized by having an annual dry season of sufficient duration and intensity that most of the woody plants shed their leaves, and the grasses dry out. This accumulation of dry fuel permits fires every few years; and the fires help prevent complete domination by the woody plants. Tropical forests are generally evergreen and not prone to fires.

We used 30° N and 35° S as the limits of the tropical formations, based on the latitudinal distribution of woodlands, savannas and grasslands. The latitudinal distribution of these classes, inferred from the global vegetation data set of Olson et al. (1983) is illustrated in Figure 4.3. The bulk of the tropical grasslands, savannas and woodlands fall between 30° S and 20° N. The strict geographical definition of 'the tropics', being the area between 23.5° N and 23.5° S, has little biological meaning. The key difference between tropical ecosystems and their temperate analogues is the degree to which processes such as primary production and litter decay are constrained by low temperatures: in tropical systems, low temperatures are at best a secondary constraint.

Figure 4.3 The latitudinal distribution of tropical woodlands, savannas and grasslands derived from the Olson et al. (1983) data. The solid line represents Olson's class 43 ('savannas'), the long-dashed line is class 59 (thorn and succulent woodlands) and the short-dashed line is classes 32 (drought-deciduous tropical forests) and 48 (mainly Australian Eucalyptus and Acacia woodlands). Note the bimodal distribution on either side of the equator, extending to 35° S and 30° N. The equatorial region is mainly occupied by evergreen forest. At the higher latitudes, savannas are generally replaced by semi-desert shrublands and grasslands.

The high-altitude tropical grasslands share many characteristics with the temperate grasslands, including dominance by species with a C₃ photosynthetic pathway and a relatively high soil organic carbon content. They should be grouped ecologically with temperate rather than tropical grasslands, since low temperatures are an important constraint. Subtropical and high-altitude savannas also experience a degree of temperature constraint at the beginning and end of the growing season, but it is small relative to the water constraint. This is because the low temperatures typically occur during the dry season, where no growth is possible in any case. Frost can be an ecologically significant factor in tropical savannas, on the infrequent occasions when unseasonal rainfall coincides with low night temperatures.

The surface area, carbon store and net primary productivity of tropical woodlands, savannas and grasslands as estimated by previous studies and calculated according to the above criteria, are shown in Table 4.1.

4.2 FACTORS CONTROLLING CARBON FLUXES IN TROPICAL GRASSLANDS AND SAVANNAS

4.2.1 Primary production

Net primary production (NPP) is defined here as the sum of net increases in total plant standing crop over a given period, plus losses due to litterfall, root death and herbivory during the same period. In practice, the available data seldom satisfy this definition perfectly. It is often not possible to determine from the sources whether turnover losses and herbivory have been included in the estimate. The estimates given below are therefore generally conservative. We have excluded all estimates based simply on end-of-season standing crop. The sources quoted have measured NPP over a period of at least a year, and in most cases, several years.

Three main factors control primary production in tropical woodlands, savannas and grasslands:

1. water availability to the plants; 

2. nutrient availability to the plants; 

3. vegetation structure (particularly the tree-grass ratio, the plant density and cover, and the amount of standing dead material).

4.2.1 Water availability

Water availability is the relationship between water supply and water demand, seen from the plant point of view. Supply and demand have independent seasonal patterns, so although availability can be expressed as an annual mean, it must be calculated using a daily, weekly or monthly water balance model which takes into account precipitation, air temperature, radiation, humidity, wind, soil texture, rooting depth, vegetation attributes and landscape position. 

Precipitation in the lowland tropics is almost entirely in the form of rainfall, most of which occurs as short-duration, high-intensity convectional storms. The near-tropical zone occupied by savannas and woodlands is climatically characterized by:

1. stable, descending airmasses for protracted periods of the year, leading to highly seasonal rainfall, sometimes bimodally distributed; 

2. high continentality (much of the area is far from the ocean); 

3. very high solar radiation, since the cloud cover is relatively low. 

These conditions result in generally low humidity and high daytime temperatures, and thus very high rates of evaporation. Annual potential evaporation substantially exceeds annual rainfall. Therefore comparing production between tropical and temperate regions on the basis of equal annual rainfall alone is misleading, since rainfall is less effective in the tropics. A better approach (still imperfect, since it disregards soil and plant details) is to relate production to a water availability index calculated as the annual sum of the monthly ratio of rainfall to open-water evaporation. The water availability index (WAI) is calculated as the sum over 12 months

WAI =å _i=12d_iR_i/E_i  R_i/E_i < 1.0 

d_i is the number of days in month i, R_i is the monthly rainfall, and E_i is the monthly potential evaporation. Constraining R/E to less than 1.0 is equivalent to saying that evaporation cannot exceed rainfall and that stored soil water is not carried over between months. This is a conservative assumption, taken in the absence of detailed soil information. The climate data were from Leemans and Cramer (1990). Open-water potential evaporation was calculated using the Priestley-Taylor equation (Rosenberg et al. 1983, using a coefficient of 1.36, appropriate for semi-arid environments). Net radiation was calculated from the solar geometry and cloud cover (McMurtrie 1993). The sources of the NPP data are in Table 4.2. The sites for which both below ground NPP and aboveground NPP data were available were used to develop an equation for total NPP as a function of aboveground NPP (Figure 4.4 ), and this equation was used to predict total NPP at the sites which reported aboveground NPP only. This equation predicts that below ground production accounts for between 5 and 70% of the total NPP, the proportion decreasing with increasing water availability. The average proportion of below ground NPP is high, but in line with recent estimates by other techniques (Raich and Nadelhoffer 1989).

Figure 4.4 The relationship between total NPP and aboveground NPP for sites in Table 4.2 which report both aboveground and below ground NPP. The equation for this relationship is total NPP = 853 + 1.01 x aboveground NPP (n = 15, F = 141.3, p<0.001, r2 = 0.91). This relationship was used to calculate the total NPP for sites reporting only aboveground NPP, for inclusion in Figure 4.5. The meaning of the symbols is explained in the caption of Figure 4.5

The NPP predicted by the relationship shown in Figure 4.5 was calculated for a large number of randomly selected locations in each of the Olson et al. (1983) vegetation classes representing tropical woodlands, savannas and grasslands (Figure 4.6 and Table 4.3). The relations describing the upper and lower data envelopes in Figure 4.5 were used to estimate the likely error range.

Table 4.2 The primary production reported for tropical grasslands, savannas and woodlands

^along et al. (1989,1992) used a variety of methods for calculation of NPP. The sum-of-positive increments method is quoted here, which gives the third highest estimate on their data. The highest estimate, which includes consideration of decay between sampling periods, may be a better estimate, but was not used because that approach was not applied in any other study.

Slope, soil texture and depth control the fraction of water which leaves the ecosystem 'productively' (i.e. through transpiration) and 'unproductively' (runoff, drainage and evaporation from the soil). Where the soil is deep and rainfall is much less than evaporation, less water is lost 'unproductively' from sandy than from clayey soils. Where the rainfall is higher and the soils shallower, sandy soils lose water through drainage and runoff, while clay soils benefit from their higher water-holding capacity and nutrient availability (see below). In moist and mesic systems, therefore, clay soils support a higher NPP than sandy soils receiving the same rainfall, but the reverse is true in arid areas. The crossover occurs around 500 mm yr^-1 (Noy-Meir 1973).

Figure 4.5 The relationship between total NPP and an index of water availability for tropical grasslands and tree-grass mixtures, with 95% confidence limits. The water availability index (WAI) can be thought of as the number of days per year on which growth is not water limited. The derivation of WAI is discussed in the text. The least-squares best fit line has the equation NPP = 11 x WAI -539 (n = 37,F =108.6,p<0.001,r² = 0.75). The data are from Table 4.2, and a number of other studies which did not satisfy the criteria for inclusion in Table 4.2, for instance because they refer to regions rather than sites. The symbols denote different structural types: [ O] grasslands; [D] derived grasslands; [ Ñ] grassy savannas; [v] savannas; [w] woodlands; [�] plantations of exotic trees in savanna areas

The implicit assumption in much of the literature on water limitation of growth is that the mechanism of limitation is the reduction in stomatal conductance resulting from stomatal closure during periods of water stress. This reduces CO₂ assimilation and therefore dry matter production. The weakness in this argument is that there is little evidence to suggest that the rate of carbon assimilation constrains plant growth in the tropics. Photosynthetic carbon assimilation substantially exceeds the amount of carbon which appears as dry matter (Scholes and Walker 1993). The high carbon:nutrient ratios of tropical grasses, in particular, suggest that the availability of nutrients is the major production constraint, rather than the rate of carbon assimilation.

Figure 4.6 The probability distribution of total NPP for tropical woodlands, savannas and thorn scrub, calculated using the regression equation developed in Figure 4.5, the Olson et al. (1983) vegetation map and a global climate data set (Leemans and Cramer 1990). Total NPP includes trees and grass, above ground and below ground. The bars represent the fraction of all locations in the given type which have NPP in the range indicated on the x-axis

Table 4.3 Calculated net primary productivity in tropical woodlands, and savannas. The estimates are based on the relationship shown in Figure 4.5, along with its upper and lower data envelopes, applied to sites randomly drawn from the areas classified as having those vegetation characteristics, using a global climate data set. The 'best estimate' is based on the regression line defined in Figure 4.5. The 'lower limit' is based on the line which forms the lower envelope to the data in the figure, and the 'upper limit' applies the line defining the upper envelope

A more convincing case can be made that water availability controls plant growth via its influence on nutrient availability. The mineralization of nitrogen and phosphorus in litter and soil, the transport of solutes to the root, and the extension of roots into the soil are all under the direct control of soil water content. This line of argument is supported by fertilization experiments and by comparisons of NPP on adjacent soils with different fertility status. In both these cases productivity is increased without an increase in water a availability.

In reality, primary production in savannas (and other vegetation types) is probably controlled and constrained by a number of partially interdependent factors, either acting simultaneously or sequentially. The relative importance of individual factors shifts with time; the point is that it must not simply be assumed that the process of carbon assimilation is all-important, and that the stomate is the locus of its control by water availability.

4.2.1.2 Soil fertility and plant nutrition

Over half of the savannas and tropical grasslands worldwide occur on extremely old land surfaces, mostly formed from acid, crystalline igneous parent material. As a result, the soils are sandy and deficient in many nutrients. The small amount of clay is predominantly in minerals such as kaolinite and sesquioxides, which have a low surface area and low ion exchange capacity. Nitrogen and phosphorus playa central role in limiting primary production in tropical savannas (Bremen and de Wit 1983). Deficiencies of other elements in the natural vegetation are usually only obvious once nitrogen and phosphorus constraints have been alleviated. In the case of nitrogen, deficiency may be due to losses of nitrogen through repeated burning ('pyrodenitrification'); in the case of phosphorus it is usually associated with low supplies of weatherable phosphorus coupled with high phosphorus fixation. Secondary production is also constrained by nitrogen, phosphorus and trace elements. There is a striking difference in large mammal biomass between savannas on fertile and infertile soils (East 1984; Fritz and Duncan 1993).

Widespread nitrogen deficiency may seem odd, given that the ultimate nitrogen source is the atmosphere, which is the same over fertile and infertile soils. Although leguminous plants are a conspicuous (frequently dominant) component of savannas, there is very little nitrogen fixation by those on infertile soils, possibly because of deficiencies of cofactors such as molybdenum and phosphorus (Zietsman et al. 1988). The nitrogen cycle is dominated by mineralization of soil organic matter, which is in turn controlled by soil water content, soil temperature and the amount of readily mineralizable soil organic nitrogen present. The soil organic matter content of infertile savannas is low, for reasons discussed below. The principalpathway of nitrogen loss from savanna ecosystems is pyrodenitrification (Scholes and Walker 1993), which is high due to the combination of frequent fires and the accumulation of decomposition-resistant fuel.

The phosphorus supply in these old, weathered soils is controlled by two factors: the rate of microbial mineralization of organic phosphorus; and the sorption of inorganic phosphorus by the abundant aluminium and iron oxides in the soil. Large tracts of virgin tropical savanna and grassland remain on soils of this type, because of their unsuitability for low-input agriculture. Some ingenious traditional methods (such as chitemene ash-fertilization agriculture in Zambia; Lawton 1982; Stromgaard 1985) have evolved to use these soils. Advances in agricultural technology, notably the ability to supply high rates of lime, nitrogen and phosphorus fertilizer, make the infertile savannas highly attractive for agricultural expansion, such as occurred in the cerrado of Brazil during the 1980s.

In addition to the extensive areas of old, weathered, acidic soils, there are significant areas of basic intrusive and extrusive geology in the tropics. These not only have a more favourable nutrient content to begin with, but retain nutrients better as well. The basic igneous rocks weather to produce predominantly 'high activity' clays, with a large surface area and high surface charge. The soils formed from these materials store more carbon and nitrogen than soils formed from acid crystalline materials (Jones 1973). They also tend to occur on younger land surfaces, and therefore have a greater reserve of weatherable nutrients and have had less time to lose nutrients through leaching.

Due to the more favourable nutrient status, more phytomass is produced per unit of water transpired by ecosystems based on soils resulting from these recent, basic igneous parent materials, but they are more drought-prone in arid areas due to the high clay content of the soil. The areas of fertile soil are the focus of agriculture in the developing world. The only uncultivated patches are usually in extremely arid areas or in national parks. Where they have not been planted to crops they are heavily grazed (and therefore seldom burned) and have a sparse tree cover due to wood collection for domestic purposes.

The most dramatic difference between the infertile and fertile tropical ecosystems lies in the degree of herbivory. Whereas less than 10% of the NPP of tropical grasslands, savannas and woodlands on infertile soils is typically consumed by herbivores, up to 80% is consumed on fertile soils (Drent and Prins 1987; Scholes and Walker 1993). As a consequence, fire is less frequent and intense on grazed fertile soils than on infertile soils, since a fuel load does not accumulate. The high acceptability of the forage to herbivores is the main reason why most of the famous savanna game parks occur on these soils. The reasons for this difference are discussed below under herbivory.

Soils derived from sedimentary and calcareous rocks, and from alluvium and windblown sands, are also extensive in savanna regions. They have a fertility status and NPP intermediate to the above extremes.

4.2.1.3 Vegetation structure

The primary productivity of grasses in tropical mixed tree-grass systems is strongly reduced by small increases in the tree biomass (e.g. Walker et al. 1972). While a dense grass cover is able to suppress young trees (those still within the grass canopy and flame zone), it has little effect on mature trees (Knoop and Walker 1984). This competitive asymmetry is a recipe for structural instability. Where the herbivory is moderate (or intense but intermittent, such as in the migratory systems of East Africa), the combination of occasional fires and browsing keeps fertile savannas open or sparsely treed. Consistently heavy grazing by domestic stock combined with the exclusion of indigenous browsing mammals and fire leads to a rapid invasion by woody plants, culminating in a dense thicket with little grass cover (e.g. van Vegten 1984; Harrington and Johns 1990). This is known as 'bush encroachment' and is most common in cattle and sheep ranching areas on fertile soils, with a low human population. In densely populated areas, wood harvesting leads to decreased woody biomass, accompanied by high grazing pressure on the grass layer.

The inverse relationship between tree leaf biomass and grass production is typically non-linear: the steepest decline occurs while the woody plant biomass is low (Figure 4.7). The degree of curvature decreases as the productive potential of the site (i.e. water and nutrient availability) increases, suggesting that it is related to the degree of below ground competition (Scanlan and Burrows 1990). A few cases have been reported where the productivity of the grass layer increases with initial increases in tree cover, before declining at higher levels of cover. These cases may be explained in terms of the higher nutrient supply and lower evaporative demand immediately below the tree canopy (Belsky et al. 1989).

Figure 4.7 Some published relationships between the quantity of trees in a savanna and the annual grass aboveground NPP. Almost all the studies show concave inverse relations; the inverse convex relationship is unusual

Reviews of savanna NPP have generally given the impression that they are unproductive systems (Bourliere and Hadley1970; Murphy 1977), especially when compared to plantations of exotic trees in the same environment (Jackson and Ojo 1971). This is partly due to the small number of studies and the application of inadequate methods of estimating NPP: in particular the use of end-of-season standing crops, and the failure to account for below ground NPP (Long et al. 1989). More than half of all savannas and woodlands are on infertile soils, and all are by definition water limited to some degree. These factors do constrain savanna productivity (and cause a large fraction of NPP to shift below ground), but the inherent capacity for production by savanna plants is not markedly lower than other systems. If the upper range of data in Figure 4.5 (i.e. non-nutrient-limited systems) is projected to a water availability of 365 days (i.e. no water limitation) the predicted annual NPP is 4000 g m^-2, which is well within the average-to-high range for other natural ecosystems not subject to strong water or nutrient constraints (Hall et al. 1993). Moist, fertile tropical grasslands achieve productivities which rival the highest published crop or natural ecosystem values (Long et al. 1989).

Although plants with the C₄ photosynthetic system generally have higher instantaneous peak rates of photosynthesis, there is no statistically significant difference in the total NPP achievable by plants with C₃ (tree) or C₄ (grass) photosynthetic systems, when grown under identical environments (Snaydon 1991). In dry areas the total aboveground NPP in mixed tree-grass ecosystems is usually higher than in pure grasslands derived from them through the removal of woody plants, despite the large increase in grass production which accompanies this action. The explanation for higher total production by partly wooded ecosystems in dry areas appears to lie in the greater leaf area duration of trees, permitted by their larger internal storage of water, nutrients and carbohydrates, as well as their deeper roots (Scholes and Walker 1993).

Primary production by tropical grasses typically increases by about 20% in the season after a fire (San Jose and Medina 1975; Grossman et al. 1981 ), although the reverse has also been recorded (Smith 1960; Brockington 1961). Where there is little grazing, production increases are partly due to the removal of the self-shading effect of old, dead standing material. The return of nutrients in the ash and changes in the soil microclimate increase the availability of nutrients, and therefore increase both quality and quantity of grass produced. The long-term consequence of frequent fire may be to reduce NPP, due to nutrient depletion. Grazing on nutrient rich sites has the same effects as fire: moderately grazed grasslands have greater NPP than ungrazed or excessively grazed grasslands (McNaughton 1983).

Soil fertility, water availability, fire, herbivory and vegetation structure interact to determine the fraction of NPP which is allocated below ground. It is typically higher in grasses (40-80%) than in woody plants (20-60%), and increases with nutrient and water stress, fire and herbivory. In the extreme (but widespread) case of savannas on deep, sandy soils, upto 80% of the plant biomass, and about 60% of the NPP, may be below ground (Rushworth 1978).

4.2.2 Carbon loss to the atmosphere

4.2.2.1 Controls on decomposition

The overriding control on the decomposition rate of plant residues in tropical grasslands and savannas is water availability. The chemical composition of the litter, and to a lesser extent its physical structure, control its decay rate during the periods when it is moist. The chemical composition of woody plant leaf litter is significantly different from that of grass litter: tree litter typically has a higher content of nitrogen, lignin and secondary compounds (Scholes and Walker 1993). The chemistry of savanna tree litters also varies substantially between trees growing on fertile soils and those growing on infertile soils (Scholes and Walker 1993). Leaves from the latter group are heavily defended with secondary chemicals, and can take several years to decompose in the field. Grass litter decomposes faster than tree litter provided it is in contact with the soil surface; however grass litter frequently remains attached to the plant, where it decays gradually until burned, grazed or trampled.

The lignin:nitrogen indices which have been useful in predicting litter decay rates in temperate regions frequently fail when applied in the tropics (e.g. Palm and Sanchez 1990). The main reason appears to be the high content of secondary plant compounds, such as tannins and other phenolics, in the litter. It is suggested that warm temperatures, which favour microbial attack, and the high incidence of mainly insect herbivory promote high concentrations of secondary chemicals in tropical plants. A low litter decay rate is a consequence of these defensive compounds, rather than the principal reason for their synthesis.

Soil fauna, notably termites and earthworms, process a large fraction of the dead organic matter in some locations and at some times, but in general only a small fraction ( < 5% ) of the ecosystem respiration results directly from detritivores (Lavelle et al. 1992; Scholes and Walker 1993). Their major impact is to move surface litter into the soil and fragment it, thereby speeding up decomposition and preventing them from being burned.

4.2.2.2 Fire

Even in annually burned savannas, a surprisingly small fraction (<20%) of the NPP ends up being consumed by fire (Scholes and Walker1993). In triennially burned savannas, this fraction decreases to about 5%. This is because a large part of the NPP is protected from fire, being underground or in the tree biomass, or having decayed in the interval between fires. Furthermore, most savanna fires are very patchy and incomplete, due to discontinuous fuel, dissected topography and variation of meteorological conditions between day and night. Some estimates of gaseous fluxes to the atmosphere from biomass burning in savannas are substantially too large, since they assume too large a fraction of the aboveground biomass is combusted, or they overestimate the fuel load or fire frequency (Menaut et al. 1991). About 40% of the total savanna area is too dry to accumulate substantial fuel loads except in high-rainfall years, or are too heavily grazed to burn fiercely.

Hao et al. (1990) estimate that 3.69 Pg DM is consumed annually by fires in savannas, of which 2.43 Pg is in Africa. This is close to the 2.52 Pg estimated by Delmas et al. (1991) for African savannas, but both are probably overestimates. Hao et al. (1990) base their calculation on a mean fuel load of 6.6 Mg ha^-1, a burn efficiency of 83%, and an annual area burned in Africa of about 450 million ha ( out of 1000 million ha of African savannas). Delmas et al. (1991) estimate an area burned of 470 million ha, a burning efficiency of 80% and a mean fuel load of 6.1 Mg ha^-1. These estimates translate to an average of l12 g C m^-2 yr^-1, or 1.68 Pg C for all the savannas, using the Hao et al. (1990) estimate of global savanna area (15 million km²). Menaut et al. (1991) provide a more resolved calculation for West Africa, in which the different types of savannas, fuel loads and fire frequencies are taken into account. For an area of 2.27 million km², they estimate an annual emission of 123 Tg C. This works out to only 54 g C m^-2 yr^-1.

Seiler and Crutzen (1980) calculated that vegetation burning in savannas amounts to 1.19 Pg DM yr^-1, or 17% of the global total for all forms of biomass burning. Hao et al. ( 1990) estimated 3.69 Pg D M yr^-1, or 59% of all biomass burning calculated by their method. We estimate that the true emission from savanna burning is somewhere between 1.94 Pg DM yr^-1 (0.87 Pg C yr^-1), derived from an extrapolation of the Menaut et al. (1991) rate using our savanna extent estimate, and 3.69 Pg DM yr^-1 (1.66 Pg C yr^-1) taken from Hao et al. (1990). Even allowing for some overestimation, fire in tropical grasslands and savannas (particularly in Africa) remains an important global source of CO, CH₄, N₂O, NO_x and tropospheric O₃.

4.2.2.3 Herbivory

The fraction of NPP which is consumed by herbivores in tropical grasslands and, savannas, and the type of herbivores involved, are strongly dependent on the soil fertility status (East 1984). The nitrogen content of the grasses growing on the old, infertile land surfaces is below the threshold (about 1% nitrogen) for efficient ruminant digestion for most of the year. Consequently, these areas have a low mammalian herbivore biomass, dominated by large-bodied herbivores (> 200 kg). Grasslands on fertile soils remain above this threshold all year, and therefore support a large number and diversity of grazers, many quite small-bodied ( < 100 kg). In both fertile and infertile savannas, a large (perhaps even dominant) part of the herbivory is by insects. In the grass layer, these are mainly locusts and grasshoppers. The trees of infertile savannas are periodically defoliated by lepidopteran larvae (Scholes and Walker 1993).

Although the fraction of production consumed by mammals in infertile savannas is small, herbivory has an important consequence on global warming through the production of methane by enteric fermentation. This source of methane is of the same order of magnitude as that from burning of tropical grasslands and savannas (Schutz et al. 1990).

4.3 CARBON STOCKS IN TROPICAL WOODLANDS, SAVANNAS AND GRASSLANDS

4.3.1 Biomass

The underlying factors which control the biomass in unharvested tropical woodlands, savannas and grasslands are water and nutrient availability. The pattern of biomass distribution in the plane defined by these two factors is complex, since a large number of modifying factors alter the basic pattern (Figure 4.8). Especially at the moist and fertile end of the spectrum, the control is mainly exerted through fire. There are two broad trends: in general, the low-fertility soils support greater tree biomass than the high-fertility soils, especially in arid areas; and generally the woody biomass increases with increasing water availability, up to the point at which the soils are saturated for several months a year, above which the tree biomass drops abruptly. Firewood collection, tree harvesting and charcoal burning are important biomass-reducing factors in populous savannas.

Figure 4.8 The generalized relationships between woody biomass and the factors which influence it. If an interaction has a positive sign, it means that an increase in the causal factor results in an increase in the affected factor. Note that there are several loops containing both positive and negative interactions. This contributes to the fundamental instability of woody biomass in savannas and the existence of multiple quasi-stable states

Table 4.4 Biomass standing stocks (DM) reported for tropical savannas, woodlands and grasslands. The site descriptions are those used by the authors of the studies. They have been regrouped according to our definitions and judgement of woodland, savanna and grassland, which frequently differs from the definition applied by Olson et al. (1983).

At this stage there are insufficient data to extrapolate these trends on a global basis. The best that can be done is to associate mean biomass estimates with the structural categories defined in one of the global vegetation databases, and multiply by the extent of that category. Table 4.3 contains data of this type. organized according to the categories in Olson et al. (1983). From the sparse below ground biomass data that are available, it is estimated that below ground biomass is 20% of the total biomass (both on a DM basis) in drought deciduous forests, 25% in moist, broad-leaved woodlands and savannas, and 30% in the arid savannas which comprise most of the 'thorn woods' category of Olson et al. (1983). Note that the fraction of the total biomass which is belowground is much smaller than the fraction of the NPP which occurs below ground (see section on primary production). Applying these estimates to the mean aboveground biomasses in Table 4.4. and assuming that DM is 45% carbon, the vegetation carbon density in dry forests is about 7.47 , in woodlands is 3.74, and dry savannas is 1.13 kg C m^-2.

4.3.2 Soil carbon

It is accepted wisdom that savannas have a low soil organic carbon (SOC) content by comparison with tropical forests (Ahn 1970; Askew et al. 1970; Kalpage 1974; Montgomery and Askew 1983) or temperate grasslands. This is generally true, but not necessarily so (Sanchez et al. 1982). Savannas, like all vegetation types, exhibit a range of soil carbon contents. At the woody end, where they grade into seasonally dry forests, the organic matter content is as high as in moist tropical forests (see later). Tropical montane grasslands, hydromorphic grasslands, grasslands on vertisols or andosols and tropical forests have relatively high SOC, while desert grasslands, sandy savannas and old fields have a very low SOC. With careful management of grazing and soil fertility, the SOC under pastures derived from tropical forests can exceed that of the original forest soil, although the total ecosystem carbon store will remain lower due to the substantial reduction in aboveground biomass which occurs when trees are removed (Lugo and Brown 1993).

The SOC in tropical grasslands, savannas and woodlands increases with increasing soil clay content, rainfall, tree cover and decreasing temperature (Birch and Friend 1956; Jenny and Raychaudhuri 1960), and is higher on soils dominated by high-activity clays (such as smectites and allophanes) than low-activity clays (kaolinite, aluminium and iron oxides: Jones 1973).

The generally low SOC in savannas may be due to one or more of the following factors:

(a) The high maximum temperatures lead to large soil respiration losses, since respiration is exponentially related to temperature;
(b) Many savanna soils are low in clay content, and the clays are of the low-activity type which do not stabilize large amounts of carbon;
(c) Some carbon is lost to the ecosystem due to burning;
(d) The alternation of wetting and drying favours decomposition over production, since soil respiration can continue at water potentials below which primary production has ceased (this is despite the higher below ground allocation in arid areas);
(e) Primary production in many savannas is low, due to low rainfall and soil fertility .

High densities of earthworms occur in some moist savannas. Epigeic earth-worms increase the rate of turnover of SOC, and could therefore reduce the SOC content (Lavelle et al. 1992). Termites dominate the soil fauna of savannas receiving less than 800 mm yr^-1 rainfall. In exceptional years, they can consume up to half of the aboveground primary production, although 2-5% is more typical (Lavelle et al. 1992; Scholes and Walker 1993). There is a suggestion from one study that exclusion of termites results in slight increases in SOC and aboveground litter accumulation (Parker et al. 1982).

Figure 4.9 The carbon density of 78 soil profiles under tropical savanna and 22 under tropical dry woodland. Data from Zinke et al. ( 1986). The bars represent the fraction of the soil profiles recorded as being under that vegetation which have a carbon density in the range given on the x-axis

This review made use of a global soil carbon database (Zinke et al. 1986) to arrive at estimates of soil carbon in tropical grasslands, savannas and woodlands; The data were corrected for bulk density, and are expressed as kg C m^-2 to a depth of 1 m. Only in the Olson ecosystem categories of tropical savannas (i.e. tall-grass, moist, infertile savannas) and seasonally dry tropical woodlands were there sufficient recorded profiles in the database for a useful analysis (78 and 22 respectively). Woodlands have a mean and standard deviation soil carbon density of 11.8 + 5.34 kg C m^-2 (very close to the   10.57 kg C m^-2 average for tropical moist forests) and the savanna mean is 5.65 + 4.60 kg C m^-2 (Figure 4.9). Five records for tropical thorn woodland gave a mean carbon density of 5.4 + 1.9 kg C m^-2, and three warm semi-arid woodlands 10.2 +1.3.

Cultivation of tropical grassland and savanna soils rapidly leads to large reductions in the SOC (Nye and Greenland 1960). It appears the decomposition-resistant fractions which make up a large part of carbon in temperate grassland soils, contribute a much smaller part in tropical grasslands (Scholes et al. 1992). Typically, about half of the SOC in tropical woodlands, grassland or savanna soils converted to croplands is lost in 20 years, whereas the average decay rate in temperate grassland soils is slower.

4.3.2.1 Carbon stored in tropical savannas and woodlands

The information detailed in Tables 4.3 and 4.4 is summarized in Table 4.5, after weighting by the aerial extent of the soil and vegetation types, and the degree of disturbance. The degree of cultivation and harvesting is a rough estimate, based on the experience of the authors. Note that the vegetation descriptions in Table 4.4 are those used by the authors of the studies cited, and do not necessarily match the definitions used in Olson et al. (1983). For instance, the south-central part of Africa is described as drought-deciduous forest in Olson et al. (1983), but is described as a woodland by African ecologists. Some judgement has therefore been exercised in assigning vegetation carbon densities to the Olson et al. (1983) categories in Table 4.5.

4.3.2.2 The potential for carbon sequestration in savannas

More than three-quarters of the ecosystem organic carbon in woodlands and savannas is in the soil (Table 4.5). The spatial distribution of SOC in savannas is notably patchy: it is substantially higher beneath the tree canopies than between them (Campbell et al.1988; Scholes and Walker 1993). Therefore, removal of trees leads to an overall decline in SOC over a period of years. Permitting the trees to become larger and denser (for example, through excluding fires) increases both the biomass carbon store and the SOC (Trapnell1959; Moore 1960; Jones et al. 1990). The upper limit of biomass plus soil stored carbon should be equivalent to the current carbon density of the tropical woodlands (14.9 kg C m^-2), or on average about double the current savanna soil carbon store (6.7 kg C m^-2). An estimate of the upper limit of the carbon sequestration potential of tropical savannas is therefore the product of the area which is currently not woodland ( 11.5 x 10¹² m²) and the difference in carbon density (8.2 kg C m^-2), about 94.3 Pg C. This sequestration could be achieved over a period of about 50 years, by excluding fire from savannas, giving an annual rate of around 2 Gt, or about one-quarter of the annual anthropogenic carbon emissions to the atmosphere as CO₂. Although this is a tempting option it is a risky one. Given the seasonally dry climate, it is nearly impossible to exclude fire reliably from savannas. When fire inevitably occurs, much of the stored carbon will be returned to the atmosphere. In the meantime, the ecology of the savannas will have been significantly altered. One of the long-term consequences of regular savanna burning is a small but steady sequestration of carbon in the form of elemental carbon, which is virtually inert (Seiler and Crutzen 1980).

Table 4.5 Total ecosystem carbon stock and turnover in tropical grasslands, savannas and woodlands

4.4 CARBON FLUXES IN TROPICAL WOODLANDS, SAVANNAS AND GRASSLANDS

The NPP model (Figure 4.5) based on the data collated in Table 4.2 and reported in Tables 4.3 and 4.5, predicts that the gross annual flux of carbon from the atmosphere into tropical grasslands, woodlands and savannas via photosynthesis is between 3.16 and 10.84 Pg C, with a best estimate of about 7.61 Pg C. Between 0.54 and 1.7 Pg C is returned to the atmosphere through burning, and a small amount (less than 0.5 Pg C) through enteric digestion. Assuming steady-state conditions, which is clearly an oversimplification, the remainder returns through decomposition in the litter layer and soil.

The net annual carbon flux from tropical grasslands and savannas is assumed to have been zero before 1800, when significant European colonization of the tropical regions began. Since that time there has probably been a small flux of carbon to the atmosphere due to woodland clearing and cultivation, and possibly a reverse flux due to bush encroachment and higher productivity as a result of rising atmospheric CO₂. There are two opposing land-use changes occurring in tropical tree-grass systems. The first is towards the removal of trees, mostly due to cultivation of the soil and for use as domestic fuelwood. This trend occurs in areas of dense human settlement. The extent of conversion has been limited by the inherent low fertility of the moist savannas, and the unsuitability of the dry savannas for rainfed crops. Where infrastructural development can support high-input agriculture, and the rainfall is above 600 mm, conversion of savannas to cropland can be expected at a large scale in the next decade. An example of this trend has occurred in the cerrado of Brazil, where the use of lime and phosphate fertilizers has allowed millions of hectares of savanna to be converted to croplands during the1980s. There are large areas of similar climate and soils in Central and Southern Africa, where the same technology could be applied.

The second trend occurs in areas of low human population, usually low rainfall, and pastoral use. Particularly on the more fertile soils, sustained heavy grazing and the exclusion of fire and browsers have resulted in an increase in woody plant density. This trend, known as 'bush encroachment', has been particularly evident in South and East Africa, Australia, Argentina and the southern United States (Walker et al. 1981)

There are no reliable data to indicate the relative importance of these two processes, the first of which releases carbon, while the second sequesters carbon. The future trend of increasing human population and agricultural technology is likely to favour carbon release.

4.5 RESEARCH NEEDS

There is no direct information on the impact of rising CO₂ concentrations on primary production, decomposition and herbivory in tropical grasslands and savannas. The amount of data on biomass, soil carbon and productivity is still small compared to some less extensive vegetation formations, but is gradually improving. There is a need for several secure, well-documented long-term ecological research sites in tropical grassland and savanna regions. Information about the extent, frequency and behaviour of vegetation fires in this particularly fire-prone system is inadequate, but will improve substantially within the next few years as research currently in progress is published. The dynamics and mechanisms of the interaction between woody plants and grasses in mixed tree-grass systems are a research theme which has been adopted for a future SCOPE project. Perhaps the largest single uncertainty in understanding and predicting the carbon budgets for the tropical woodlands, savannas and grasslands is the extent, degree and nature of land-use change, along with the human processes which drive it.

4.6 ACKNOWLEDGEMENTS

The calculations in this chapter would not have been possible without access to the global soil carbon database, made available by the Oak Ridge National Laboratory, and the Global Ecosystems Database, a joint US Environmental Protection Agency and National Oceanic and Atmospheric Administration/National Geophysics Data Centre programme.

4.7 REFERENCES

Ahn, P. M. (1970) West African Soils. Oxford University Press, Oxford.

Ajtay, G. L., Ketner, P. and Duvigneaud, P. (1979) Terrestrial primary production and phytomass. In: Bolin, B., Degens, E. T., Kempe, S. and Ketner, P. (Eds) The Global Carbon Cycle, pp. 129-182. SCOPE Vol. 13. Wiley & Sons, New York .

Askew, G. P., Moffat, D. J., Montgomery, R. F. and Searle, P. L. (1970) Interrelationships of soil and vegetation in the savanna-forest boundary zone of north-eastern Mato Grosso. Geogr. J. 136, 370-376.

Aucamp, A. J. Dankwerts, J. E., Teague, W. R. and Venter, J. J. (1983) The role of Acacia karroo in the false thornveld of the Eastern Cape. J. Grassl. Soc. Sthn. Afr. 8, 151-154.

Beale, I. F. (1973) Tree density effects on yields of herbage and tree components in south-west Queensland mulga (Acacia aneura F. Meull.) scrub. Tropical Grasslands 7, 135-142.

Belsky, A. J., Amundson, R. G., Duxbury, J. M., Riha, S. J., Ali, A. R. and Mwonga, S. A. (1989) The effects of trees on their physical, chemical and biological environments in a semi-arid savanna in Kenya. J. Appl. Ecol. 26, 1005-1024.

Birch, H. F. and Friend, M. T. (1956) The organic matter and nitrogen status of East African soils. J. Soil Sci. 8, 156-167.

Bourliere, F. and Hadley, M. (1970) The ecology of tropical savannas. Annu. Rev. Ecol. Syst. 1, 125-151.

Bremen, H. and de Wit, C. T. (1983) Rangeland productivity and exploitation in the Sahel. Science 221, 1341-1347.

Brockington, N. R. (1961) Studies of the growth of Hyparrhenia-dominant grassland in Northern Rhodesia. II. The effect of fire. J. Br. Grassl. Soc. 16, 54-64.

Campbell, B. M., Swift, M. J., Hatton, J. and Frost, P. G. H. (1988) Small-scale vegetation pattern and nutrient cycling in miombo woodland. In: Verhoeven, J. T. A., Heil, G. W. and Werger, M. J. A. (Eds) Vegetation Structure in Relation to Carbon and Nutrient Economy, pp. 69-85. SPD Academic Publishing, Amsterdam.

Christie, E. K. (1978) Ecosystem processes in semi-arid grasslands. I. Primary production and water use of two communities possessing different photosynthetic pathways. Aust. J. Agric. Res. 29, 773- 787.

Collins, N. M. (1977) Vegetation and litter production in a southern Guinea savanna, Nigeria. Oecologia 28, 163-175. 

Dayton, B. R. (1978) Standing crops of dominant Combretum species at three browsing levels in the Kruger National Park. Koedoe 21, 67-76.

Delmas, R. A., Loudjani, P., Podaire, A. and Menaut, J.-C. (1991) Biomass burning in Africa: an assessment of annually burned biomass. In: Levine, J.S. (Ed.) Global Biomass Burning: Atmospheric, Climatic and Biospheric Implications, pp. 126-132. MIT Press, Cambridge, Mass.

De Ridder, N., Stroosnijder, L., Cisse, A. M. and van Keulen, H. (1982) Productivity of Sahelian Rangelands. Vol. 1. Dept of Soil Science and Plant Nutrition, University of Wageningen, The Netherlands.

Donaldson, C. H. and Kelk, D. M. (1970) An investigation of the veld problems of the Molopo area. 1. Early findings. Proc. Grassl. Soc. 5, 50-57.

Donaldson, C. H., Rethmann, G. and Grossman, D. (1984) Long-term nitrogen and phosphorus application to veld. J. Grassl. Soc. South. Afr. 1, 27-32.

Drent, R. H. and Prins, H. H. T. (1987) The herbivore as prisoner of its food supply. In: van Andel, J. et al. (Eds) Disturbance in Grassland, pp. 131-147. Dr. W. Junk Publishers, Dordrecht.

East, R. (1984) Rainfall, soil nutrient status and biomass of large African savanna mammals. Afr. J. Ecol. 22, 245-270.

Endean, F. (1967) The productivity of miombo woodland in Zambia. For. Res. Bull. 14, Government Printer, Lusaka, Zambia. 

Fatubarin, A. (1984) Biomass estimates for the woody plants in a savanna ecosystem in Nigeria. Trop. Ecol. 25, 208-213. 

Fritz, H. and Duncan, P. (1993) Large herbivores in rangelands. Nature 364, 292--293.

Greenland, D. J. and Nye, P. H. (1959) Increases in the carbon and nitrogen contents of tropical soils under natural fallows. J. Soil Sci. 10, 284-299.

Grossman, D., Grunow, J.O.. and Theron, G. K. (1981) The effect of fire with and without subsequent defoliation on the herbaceaous layer of Burkea africana savanna. Proc. Grassl. Soc. South. Afr. 16, 117-120.

Gupta, R. K., Sharm, S. and Sharm, S. K. (1972) Aboveground productivity of grasslands at Jodhpur, India. In: Golley, P. M. and Golley, F. B. (Eds) Tropical Ecology with Emphasis on Organic Production, pp. 75-93. Univ. of Georgia, Athens, Ga. 

Gurumurti, K., Bhandari, H. C. S. and Dhawan, M. (1986) Studies of yield, nutrients and energy conversion efficiency in energy plantations of Acacia nilotica. J. Tree Sci. 5, 36-42.

Guy, P. R. (1981) Changes in biomass and productivity of woodlands in the Sengwa wildlife research area, Zimbabwe. J. Appl. Ecol. 18, 507-519.

Hall, D. O., Scurlock, J. M. O., Bolhar-Nordenkampf, H. R., Leegood, R. C. and Long, S. P. (1993) Photosynthesis and Production in a Changing Environment. Chapman & Hall. London.

Hao, W. M., Lui, M. H. and Crutzen, P. J. (1990) Estimates of annual and regional release of CO₂ and other trace gases to the atmosphere from fires in the tropics, based on F AO statistics for the period 1975-1980. In: Goldammer, J. G. (Ed.) Fire in the Tropical Biota: Ecosystem Processes and Global Challenges, Ecol. Stud. 84, 440-462. Springer- Verlag, Berlin.

 Harrington, G. N. and Johns, G. G. (1990) Herbaceous biomass in a Eucalyptus savanna woodland after removing trees and/or shrubs. J. Appl. Ecol. 27, 775-787.

Jackson, J. K. and Ojo, G. O. A. (1971) Productivity of natural woodland and plantations in savanna zones of Nigeria. Nigerian J. For. 1, 53-90.

Jenny, H. and Raychaudhuri, S. P. (1960) Effect of Climate and Cultivation on Nitrogen and OM Reserves in Indian Soils. Indian Council of Agricultural Research, New Delhi.

Jones, C. L., Smithers, N. L., Scholes, M. C. and Scholes, R. J. (1990) The effect of fire frequency on the organic components of a basaltic soil in the Kruger National Park. South Afr. J. Plant Soil 7(4),236-238.

Jones, M. J. (1973) The organic matter content of the savanna soils of West Africa. J. Soil Sci. 24, 42-53.

Kadeba, O. and Aduayi, E. A. (1985) Litter production, nutrient recycling and litter accumulation in Pinus caribaea Morelet var hondurensis stands in the northern Guinea savanna of Nigeria. Plant Soil 26, 197-206.

Kadeba, O. and Aduayi, E. A. ( 1986) Dry matter production and nutrient distribution in a Pinus caribaea stand planted in a subhumid tropical savanna site. Oikos 46, 237-242.

Kalpage, F .S. C. F. (1974) Tropical Soils. St Martin's Press, Macmillan, New York.

Kelly R. D., Schwimm, W. F. and Barnes, D. L. (no date) Bush clearing trial at Nuanetsi. Annual Report of the Rhodesian Dept of Specialist Services, Harare, Zimbabwe. About . 1975.

Kelly, R. D. and Walker, B. H. (1976) The effects of different forms of land use on the ecology of a semi-arid region of South Eastern Rhodesia. J. Ecol. 64, 555-576.

Knoop, W. T. and Walker, B. H. (1984) Interactions of woody and herbaceous vegetation in two savanna communities at Nylsvley. J. Ecol. 73, 235-253.

Lavelle, P., Blanchart, E., Martin, A., Spain, A. V. and Martin, S. (1992) Impact of soil fauna on the properties of soils in the humid tropics. In: Myths and Science of Soils of the Tropics. SSS A Special Publication 29, pp. 157-185. Soil Sci. Soc. Am., Madison.

Lawton, R. M. (1982) Natural resources of miombo woodland and recent changes in agricultural and land-use practices. For. Ecol. Manage. 4, 287-297.

Leemans, R. and Cramer, W. P. (1990) The IIASA database for monthly values of temperature, precipitation and cloudiness on a global terrestrial grid. Working paper, August 1990. WP-90-41, Biosphere Dynamics Project, International Institute for Applied Systems Analysis, A-2361, Laxenburg, Austria.

Long, S. P ., Garcia Moya, E., Imbamba, S. K., Kamnalrut, A., Piedade, M. T. F ., Scurlock, J. M. O., Shen, Y. K. and Hall, D. O. (1989) Primary productivity of natural grass ecosystems of the tropics: a reappraisal. Plant Soi1 115, 155-166. 

Long, S. P., Jones, M. B. and Roberts, M. J. (Eds) (1992) Primary Productivity of Grass Ecosystems of the Tropics and Sub-tropics. Chapman & Hall, London.

Lugo, A. E. and Brown, S. (1993) Management of tropical soils as sinks or sources of atmospheric carbon. Plant Soi1 149, 27-41.

McMurtrie, R. E. (1993) Modelling of canopy carbon and water balance. In: Hall, D. O., Scurlock, J. M. O., Bolhar-Nordenkampf, H. R., Leegood, R. C. and Long, S. P. (Eds) Photosynthesis and Production in a Changing Environment, pp. 220-231. Chapman & Hall, London.

McNaughton, S. J. (1983) Compensatory plant growth as a response to herbivory. Oikos 40,329-336.

Malaisse, F., Freson, R., Goffinet, G. and Malaisse-Mousset, M. (1975) Litterfall and litter breakdown in miombo. In: Golley, F. B. and Medina, E. (Eds) Tropical Ecological Systems: Trends in Terrestrial and Aquatic Research, pp.137-152. Springer Verlag, New York.

Menaut, J.-C., Abbadie, L., Lavenu, F., Loudjani, P. and Podaire, A. (1991) Biomass burning in West African savannas. In: Levine, J. S. (Ed.) Global Biomass Burning: Atmospheric, Climatic and Biospheric Implications, pp. 133-142. MIT Press, Cambridge, Mass.

Menaut, J.-C. and Cesar, J. (1979) Structure and primary production of Lamto savannas Ivory Coast. Ecology 60, 1197-1210.

Montgomery, R. F. and Askew, G. P. (1983) Soils of tropical savannas. In: Bourliere, F. (Ed.) Ecosystems of the World 13: Tropical Savannas, pp. 63-78. Elsevier, Amsterdam.

Moore, A. W. (1960) The influence of annual burning on a soil in the derived savanna zone of Nigeria. Trans. 7th Int. Congr. Soil Sci. 4, 257-264.

Moore, A. W., Russel, J. S. and Coaldrake, J. E. (1967) Dry matter and nutrient content of a subtropical semi-arid forest of Acacia harpophylla F. Muell. (Brigalow). Austr. J. Bot. 15, 11-24.

Murphy, P. G. (1977) Rates of primary productivity in tropical grassland, savanna and forest. Geo-Eco-Trop 1, 95-102.

Noy-Meir, I. (1973) Desert ecosystems: environment and producers. Annu. Rev. Ecol. Syst. 4, 25-51.

Nye, P. H. and Greenland, D. (1960) The Soil Under Shifting Cultivation. Technical Communication 51, Commonwealth Agriculture Bureau, Wallingford, UK.

Olson, J. S., Watts, J. A. and Allison, L. J. (1983) Carbon in Live Vegetation of Major World Ecosystems. Report ONRL-5862, Oak Ridge National Laboratory, Oak Ridge, Tenn.

Palm, C. A. and Sanchez, P. A. (1990) Nitrogen release from the leaves of some tropical legumes as affected by their lignin and polyphenol contents. Biotropica 22, 330-338.

Pandey, C. B. and Singh,  J. S. (1992) Rainfall and grazing effects on net primary productivity in a tropical dry savanna, India. Ecology 73, 2007-2021.

Parker, L. W., Fowler, H. G., Ettershank, G. and Whitford, W. G. (1982) The effects of subterranean termite removal on desert soil nitrogen and ephemeral fauna. J. Arid Environ. 5, 53-59.

Raich, J. W. and Nadelhoffer, K. J. (1989) Below ground carbon allocation in forest ecosystems: global trends. Ecology 70, 1346-1354.

Rajvanshi, R. and Gupta, S. R. (1985) Biomass, productivity and litterfall in a tropical Dalbergia sissoo Roxb. forest. J. Tree Sci. 4, 73-78.

Rosenberg, N. J., Blad, B. L. and Verma, S. B. (1983) Microclimate: The Biological Environment 2nd Edn. John Wiley, New York.

Rushworth, J. E. (1978) Kalahari sand scrub: something of value. Rhod. Sci. News 12(8), 193-195.

Rutherford, M. C. (1982) Woody plant biomass distribution in Burkea africana savannas. In: Huntley, B. J. and Walker, B. H. (Eds) The Ecology of Tropical Savannas, pp. 120-141. Ecological Studies 42, Springer-Verlag, Berlin.

Sanchez, P. A., Gichuru, M. P. and Katz, L. B. (1982) Organic matter in major soils of the tropical and temperate regions. 12th Int. Congr. Soil Sci. 99-114.

San Jos, J. J. and Medina, E. (1975) Effects of fire on organic matter production in a tropical savanna. In: Golley, F .B. and Medina, E. (Eds) Tropical Ecological Systems, pp. 251-264. Springer-Verlag, Berlin.

Sarmiento, G. (1984) The Ecology of Neotropical Savanna. Harvard University Press., Cambridge, Mass.

Scanlan, J. C. and Burrows, W. H. (1990) Woody overstory impact on herbaceous understory in Eucalyptus spp. communities in central Queensland. Austr. J. Ecol. 15, 191-197.

Scholes, M. C., Scholes, R. J. and McKean, S. K. (1992) The origin and stability of soil organic matter in a semi-arid savanna. Proceedings of a Seminar on Stable Isotopes in Plant Nutrition, Soil Fertility and Environmental Studies, pp. 57-60. Institute of Soil, Climate and Water, Private Bag X79, Arcadia, South Africa.

Scholes, R. J. (1988) Response of three semi-arid savannas on contrasting soils to removal of the woody component. Ph.D. thesis, University of the Witwatersrand, Johannesburg, South Africa. 

Scholes, R. J. and Walker, B. H. (1993) An African Savanna: Synthesis of the Nylsvley Study. Cambridge University Press, Cambridge, UK.

Schütz, H., Seiler, W. and Rennenberg, H. (1990) Soil and land-use related sources and sinks of methane in the context of the global methane budget. In: Bouwman, A. F. (Ed.) Soils and the Greenhouse Effect, pp. 269-288. John Wiley & Sons, Chichester.

Seiler, W. and Crutzen, P. J. (1980) Estimates of gross and net fluxes of carbon between the biosphere and atmosphere from biomass burning. Clim. Change 2, 207-247.

Smith, E. L. (1960) Effects of burning and clipping at various times during the wet season on tropical grass range in northern Australia. J. Range Manage. 13, 197-203.

Snaydon, R. W. (1991) The productivity of C₃ and C₄ plants: a reassessment. Funct. Ecol. 5, 321-330.

Stromgaard, P. ( 1985) Biomass, growth and burning of woodland in a shifting cultivation area of south-central Africa. For. Ecol. Manage. 12, 163-178.

Swaminath, M. H. (1988) Studies on the response of fast-growing forestry species for biomass production under irrigation. Myforest 24, 117-123.

Tietema, T. (1993) Possibilities for the management of indigenous woodland in southern Africa: a case study from Botswana. Proceedings of the Symposium on Ecology and Management of indigenous Forests in Southern Africa, Victoria Falls, July , pp. 134-142. Zimbabwe Forestry Commission, Harare.

Trapnell, C. G. (1959) Ecological results of woodland burning in Northern Rhodesia. J. Ecol. 47, 161-172.

Van Vegten, J. A. (1984) Thornbush invasion in a savanna ecosystem in eastern Botswana. Vegetatio 56, 3-7.

Varshney,C. K. (1972) Productivity of Delhi grasslands. In: Golley, P. M. and Golley, F. B. (Eds) Tropical Ecology with Emphasis on Organic Production, pp. 27-42. Univ. of Georgia, Athens, Ga.

Vyas, L. N., Garg, R. K. and Agarwal, S. K. (1972) Net aboveground production in the monsoon vegetation at Udaipur. In: Golley, P. M. and Golley, F. B. (Eds) Tropical Ecology with Emphasis on Organic Production, pp. 95-99. Univ. of Georgia, Athens, Ga.

Walker, B. H., Ludwig, D., Holling, C. S. and Peterman, R. S. (1981) Stability of semi-arid savanna grazing systems. J. Ecol. 69, 473-498.

Walker, J., Moore, R. M. and Robertson, J. A. (1972) Herbage response to tree and shrub thinning in Eucalyptus populnea shrub woodlands. Austr. J. Agric. Res. 23, 405-410.

Ward, H. K. and Cleghorn, W. B. (1964) The effects of ringbarking trees in Brachystegia woodland on the yield of veld grasses. Rhod. Agric. J. 61, 98-107.

Whittaker, R. H. and Likens, G. E. (1973) Carbon in the biota. In: Woodwell, G. M. and Pecan, E. V. (Eds) Carbon and the Biosphere. AEC Symposium Series 30, 281-302. US Dept of Commerce, Springfield, Va.

Zietsman, P. C., Grobbelaar , N. and van Rooyen, N. ( 1988) Soil nitrogenase activity of the Nylsvley Nature Reserve. South Afr. J. Bot. 54, 21-27.

Zinke, P. J., Strangenberger, A. G., Post, W. M., Emanuel, W. R. and Olson, J. S. (1986) Worldwide organic soil carbon and nitrogen data. ORNL/CDIC-18 NDP-018.Carbon Dioxide Information Centre, Oak Ridge, Tenn.