without or with very few trees and shrubs (Figure 8.1). Moisture, geomor-phological factors, fire and herbivory are the principal determinants of this heterogeneity. So for example in the savannas of the Orinoco, the density of the woody vegetation varies with soil depth and with the age of the deposits (Silva and Sarmiento 1976a,b). This heterogeneity raises the question whether savannas should be considered as a single ecosystem, or whether they represent a diversity of ecosystems with poorly defined borders.
Most savanna syntheses have emphasized the similarities in structure and function rather than the differences in savanna ecosystems (Bouriiere 1983; Sarmiento 1984; Tothill and Mott 1985). One such approach was the RSSD (Responses of Savannas to Stress and Disturbance) program of the Decade of the Tropics, sponsored by IUBS, that developed a set of hypotheses that predict the function of tropical savannas (Frost et ai. 1986; Walker 1987; Sarmiento 1990; Werner 1991; Young and Solbrig 1993). The RSSD postulated four principal selective forces - which were called determinants - to explain some of the common features and differences in savanna structure and function. These are: (1) plant-available moisture (PAM); (2) plant-available nutrients (PAN); (3) fire; (4) herbivory. These determinants interact at all ecological scales from landscapes to local patches, but their relative importance differs with scale (Medina and Silva 1990; Solbrig 199ta).
According to the RSSD model, PAM and PAN are the principal detcrmi-
nants of savanna structure at the higher scales. They circumscribe what was called the PAM AN plane. Where PAM has high values mesic woody elements dominate, and as PAM increases the savanna eventually gives way to a moist forest. When PAM has very low values drought-adapted species become more numerous, and if the values of the PAM-AN plane get very low the savanna is replaced by a semi-desert. Between these two extremes the gamut of savanna types is encountered. To a limited extent PAM- and PAN compensate each other: low humidity regimes with relatively high nutrient levels, such as the Serengeti in Kenya, have a savanna-grassland and not a semi-desert vegetation; likewise areas with high rainfall but low nutrients, such as the American Llanos del Orinoco and the west African Guinea savannas in the Lamto area, have a savanna rather than a forest vegetation. Within savanna ecosystems, the local effects of the patchy distribution of soil types and topographic features modify PAM and PAN, and together with fire and herbivory determine the density of the tree layer, the productivity of the system, and the rates of nutrient and water flow through the system (Frost et al. 1986). Yet PAM and PAN are general determinants of vegetation, and their power in predicting some savanna properties cannot be considered sufficient evidence for the uniqueness of the savanna ecosystem.
In this chapter we address the following null hypothesis: "Removal and additions of species that produce changes in spatial configuration of landscape elements will have no significant cflect on ecosystem functional properties of savannas over a range of time and space scales" (Solbrig 1991b). Addition or removal of species from an ecosystem will change both the species richness and its evenness, (Pielou 1975),and it is important that both these aspects be considered. Furthermore, the effect of the addition or removal of species will depend on the morphological, physiological, demographic and trophic characteristics of the species. Clearly the effect will not be the same if the dominant tree is removed from a savanna, or if a rare leaf-mining insect is removed. While there are well-established quantitative procedures to measure the number and relative abundance of species, there is no universally accepted measure of the relative importance of different species in ecosystem function. This remains one of the principal outstanding problems in assessing the importance of biodiversity in ecosystem function.
The RSSD program primarily addressed questions regarding the function of savanna ecosystems and largely ignored the behavior of individual species. Yet the physical factors of climate and geology - such as rainfall, temperature, soil structure and soil nutrients - operating on individual organisms, as well as interactions between organisms, constitute the evolutionary forces that configure the characteristics of ecosystems. System properties such as productivity, structure and resilience are not under direct selection, but are modified as a result of changes in species populations and their properties. All ecosystem properties are the result of a particular mix of species in time and space possessing a given set of characteristics. Therefore, the subtraction or addition of species from a savanna ecosystem ought to modify its structure and function at some scale. The interesting question is then at what scale, and by how much, are the properties of a savanna ecosystem modified when its species composition changes. We would also like to know the mechanism responsible for the changes.
Tropical savannas in different continents when growing under similar values of PAM and PAN exhibit very similar ecosystem properties in spite of being composed of an entirely different set of Linnaean spccies (Medina and Huber 1992). In other words, two savanna ecosystems can be functionally very similar even when their species composition is not. It is valid to conclude that in such cases convergence in, the relevant species properties has taken place. The interesting question is therefore what those relevant properties are. arid by how much do sets of species from different savanna ecosystems have to resemble each other to produce similar ecosystem characteristics.
invasion of American and Australian savannas by African grasses, removal and/or reduction in the abundance of grass species through overgrazing, and the removal of shrubs and trees through intensification of fire regimes and mechanical means are examples of additions and removals of species that can be used to test the null hypothesis.
Was this article helpful?