Comments on “The Origin of the Savanna Biome”

“The Origin of the Savanna Biome” by David Beerling & Colin Osborne, Global Change Biology (2006) 12: 2023–2031

In this paper, the authors posit that the savanna biome spread rapidly around 8 million years ago as the result of positive feedback cycles involving several factors, including not only a drop in atmospheric carbon dioxide concentration, but also the effects of fire and smoke and the simultaneous emergence of large herbivores in those particular regions of the globe.

The plant composition in these areas is determined by the particular environmental conditions, which give the advantage to either C₃ trees or C₄ grasses to prosper relative to one another. The standard view was that the concentration of CO₂ in the atmosphere fell to the point where grasses were able to outcompete trees, since they employ the C₄ photosynthetic pathway, which is more efficient at lower CO₂ levels than the C₃ pathway. Beerling and Osborne argue that this could not have been the only factor, as the appropriate CO₂ level had already been reached 20 million years earlier.

Wildfires are a common feature of savannas. Beerling and Osborne contend that this biome owes its existence to fire, and that grasses are adapted to facilitate the spread of fire, which limits the survival of trees in the area. The pyrocumulus clouds above wildfires contain many suspended particles, increasing the occurrence of lightning through friction, which may start more fires. These airborne particles also hold clouds together longer, and they absorb more heat from the sun, warming the troposphere but shading the ground, thus reducing evaporation. Both of these processes reduce the amount of rainfall and make fires more common. This multiple-positive-feedback cycle ensures that young trees are eliminated, exposing grasses to more sunlight, which allows them to grow fast and dominate the vegetation.

This cycle seems reasonable in principle, but, as the authors themselves admit, it is not clear to what extent the effects of these processes will be felt within the system. Wind, for example, might play a role in redistributing aerosols or keeping the temperature gradient between the surface and the troposphere around a certain level. The authors also do not mention the local increase in CO₂ concentration due to fires. I suppose the effect is negligible over an entire growing season, but some quantitative data on this negative feedback loop would be useful.

Low atmospheric CO₂ levels further drive the fire cycle by limiting the growth of tree seedlings, which keeps them short enough to be devastated by fire.

In the text, Beerling and Osborne say that “Global shrinking of forests […] slows biotic weathering,” (bottom of p. 2027) but that this effect may be offset by expansion of grasslands. Their flowchart (Figure 4a), however, shows only an increase in weathering with grass expansion; the negative feedback loop is omitted without explanation. There are also two opposing pathways relating CO₂ and weathering: h-k-g-c-n-m (warmer surface causes more tree growth, which decreases weathering) is a positive feedback loop, while h-l-m (warmer surface increases weathering) constitutes a negative one. Empirical data is required to decide the relative importance of the two pathways and their net effect.

The third major set of contributing factors to the creation and maintenance of savannas proposed by Beerling and Osborne is that of herbivores, which are adapted to their habitat but also have an effect on the local vegetation. The important distinction here is between browsers, which increase in numbers to exploit trees as food sources and thus limit tree survival (a negative feedback cycle tending towards a dynamic equilibrium in predator and prey populations), and grazers, which thrive in grasslands and destroy trees by trampling but also promote their survival by reducing the biofuel available for fire (grasses).

The effects of grazing and browsing presumably do not have the same magnitude or direction. This article does not specify which has the greater influence, and the authors call for quantitative modeling to settle the question.

The herbivory factor seems to act on a significantly different timescale to weathering and changing CO₂ levels, for example. Considering the susceptibility of large animals to extinction events and tectonic and climatic shifts, it seems prudent to distinguish between short-term and long-term effects. Changes in elephant populations over hundreds of years might affect the extent of grasslands during that period, but are unlikely to be as important as changes in climate over a million years. Similarly, weathering should affect CO₂ concentrations over long periods of time, but are unlikely to expand or restrain grasslands over a thousand years. This distinction is not clear from their diagrams, and it might be important in assigning significance to the various factors discussed.

The investigation of complex systems over large spaces and time periods is an onerous task, and scientists often prefer to limit themselves to simple, controllable experiments where results are easier to interpret. This study by David Beerling and Colin Osborne is thus both commendable in its aims and interesting in its conclusions. The concept of multiple feedback cycles is likely to be invaluable in understanding many other complex systems with numerous interacting factors.

Two factors that seem to have been left out of Beerling and Osborne’s analysis are groundwater and soil composition. If this was a deliberate choice, it is not explicitly motivated. I do not know to what extent these variables could affect vegetation patterns, or whether this question has been addressed elsewhere, but it seems a conspicuous omission.

A significant limitation of this paper is the fact that most of the empirical data seem to come from studies of single factors. Little evidence is given for the interactions between them that are proposed by the authors. Beerling and Osborne themselves call for complex simulations to provide quantitative data. However, mathematical models in general are prone to gross error if important factors are ignored or underestimated. Although practically difficult, a controlled large-scale common garden−type experiment, in which CO₂ concentration, temperature, and water, air, and soil quality were regulated, might provide useful data for confirming or refining the model proposed by Beerling and Osborne.