Diversity of Tropical Biome Structure and Function

Savanna can be considered to encompass a spectrum extending from almost pure grassland to open forest or closed woodland. Numerous factors can be invoked to account for these differences and within a given area of more or less uniform rainfall it is possible to see great variations in structure and to relate these variations to non climatic effects:

  • Heavy soils and seasonal water-logging tend to favour grasses over trees
  • More fertile soils tend to favour trees over grasses
  • More frequent fires tend to favour grasses over trees
  • Human or animal activity can either increase or decrease the relative density of trees.

Edaphic effects are more-or-less static but fire and other disturbances can change in their effects on much shorter time scales. All effects have their own potential feedbacks – by themselves and through interactions with each other. For example, an increased fire frequency will favour grasses over trees. Fire spreads more easily when grasses are present and this increases the probability of fire. Likewise, increased woodiness in response to more soil fertility should decrease fire frequency, reinforcing the positive effect of soil fertility in favouring woody vegetation in the first place. A secondary effect of this might then be a reduction in fire-associated nutrient losses which a commensurate long term increase in the overall ecosystem fertility resulting in a gradual aggradation of the system. Savannas can thus be considered to exist in a multi-dimensional environmental space with the consequences of each dimension potentially influencing the effects of the other three.

At a larger scale, variations in precipitation become important. Fig. 4 shows the distribution of different categories of African savanna with woodier types tending to occur in higher rainfall regions. There are models of tree-grass interactions in savanna systems, but these generally only consider grazing and/or fire effects; not edaphic effects or large scale patterns in precipitation. In the majority of Dynamic Global Vegetation Models (DGVMs) functional types of species are the focus of ecological modelling and dynamics. Biomes are then a mix of functional types that can occur in a particular climate. The functional types are, at a minimum, evergreen and deciduous broadleaved and needle leaved trees, shrubs, C3 and C4 grasslands. Savannas are then simulated as a mixture of (typically) broadleaved deciduous trees and C4 grassland. That evergreen broadleaved tree savannas are rarely predicted by the DGVMs may relate to the extended dry season and it may be that soil type, or even water logging, plays a major role here: a feature that is not considered in most current DGVMs. Most DGVMs generalise savannas to an unacceptable degree. They consider only rainfall (and maybe fire) as a determinant of structure and function. TROBIT is currently developing a new approach to address these limitations.

Many of the mechanisms causing differences within savanna are the same as those causing differences between savanna and rain forest (Section 1). The difference is that the savanna-forest transition is more of a “phase transition” than a continuous process (Section 2). We can modify Fig 3 (again for a small region of uniform rainfall) with the disjunction between savanna and rainforest emphasised. Combining this with the precipitation/feedback effect in Fig 2 gives rise to Fig 5. At high rainfall there is only a small area suitable for savannas (very low fertility soils and/or hyperseasonal savannas), but as rainfall decreases then the available space increases with soil fertility increasing its effects on the vegetation. Fire frequency will change as will the other effects (the soil texture effect may in fact become reduced for example, as water-logging becomes less severe). We term the above the n- Dimensional Vector Model (nDVM). It changes in shape as well and size as rainfall amount and seasonality changes.

School of Geography | Enquiries to : j.lloyd@leeds.ac.uk | Updated: 09/01/2009 | Privacy Statement | NERC |