Arie Staal

Forests and savannas span vast areas of the Earth, and changes in their distributions would significantly affect humans, biodiversity, and regional and global climates. It has been hypothesized that across a range of climates, tropical forest and savanna can be “alternative stable states”. This means that both forest, which is characterized by high tree cover, and savanna, characterized by low tree cover, are stable under the exact same climate. Such alternative stable states are the result of positive feedbacks and are accompanied by tipping points. Positive feedbacks reinforce change. Think, for example, about a speaker that is directed towards a microphone feeding that speaker: sound gets amplified, which in turn results in a louder input into the microphone. These occur either when a system state is not resilient against a disturbance or when the resilience breaks down altogether. In either case, positive feedbacks propel the system to an alternative stable state. Such a transition can be persistent: reverting the environmental conditions may not be sufficient to also revert the transition itself, a phenomenon called hysteresis (also see chapter 1).

Especially the Amazon rainforest is feared to be susceptible to tipping when climatic changes or deforestation would surpass a critical level. Climatic changes such as drying may reduce the resilience of forests. Disturbances to the forest like deforestation may have several knock-on effects on the forest through positive feedback loops. On a regional scale this involves moisture recycling: trees pump water from the soil and “transpire” it through their leaves to the atmosphere, from which it can rain out again. This, in turn, benefits the forest itself and increases the pump effect. A local-scale feedback involves fire: tree cover suppresses fire and fire kills trees. A loss of tree cover can thus result in further loss in tree cover through a positive feedback with fire.

In this dissertation I explore mechanisms behind, and possible effects of, the possibility that tropical forest and savanna can be alternative ecosystem states. Together with colleagues I studied tree-cover feedbacks with fire and with rainfall in the Amazon basin and other tropical regions.

We attempted to bridge gaps between theory and observation on the resilience and stability of tropical forest and savanna by linking modelling to the analysis of broad-scale empirical data from satellites. In chapter 2 we started by developing and analyzing a simple dynamical model with alternative tree-cover states. We applied this model to the tree-cover distributions in the “arc of deforestation” in the southeastern Amazon. This allowed us to show that interactions between local deforestation and regional drying could more easily induce tipping points of tree-cover loss than they could separately.

Theoretical results suggest that spatial interactions that take place between patches of forest and savanna can affect their hysteresis. Spatial interactions can for instance mean the dispersal of forest seeds or a savanna fire spreading into an adjacent forest. Such spatial interactions can narrow the range of climatic conditions where the two states can locally coexist: the least stable of the two is not resilient against the establishment of the most stable one. In the ultimate case of strong spatial interactions, a boundary between both states is only possible at conditions where forest and savanna are equally stable; this is despite the fact that in isolation, they can be alternative stable states. Such conditions of equal stability are called the ‘Maxwell point’. In chapter 3 we scanned tropical tree cover distributions for indications of this narrowing of hysteresis. We based our search on theoretical predictions of a mismatch between the conditions at which there are alternative stable states and those at which the stable local-scale coexistence of the alternative states is possible. Tree-cover data suggest that alternative stable states can occur between mean annual rainfall levels between 1100‒2000 mm for South America and between 1300‒2000 mm for Africa. However, our empirical analysis provided little evidence that spatial interactions between patches of forest and savanna are strong enough to eliminate their hysteresis.

In chapter 4 we explicitly considered the mechanism that may cause tipping points between forest and savanna. By analyzing fire occurrences across the tropics, we showed that the frequency of fires rises abruptly when tree cover drops below 40%. Using a simple empirical model based on these observations, we show that the steepness of this pattern can cause instability of intermediate tree cover under a broad range of assumptions on tree-cover growth. This is consistent with observed frequency distributions of tropical tree cover, implying that a feedback between fire and tree cover may be sufficient for generating alternative stable states of forest and savanna. We also show that percolation of fire through the open savannas may explain the rise of fire frequency around a critical tree cover.

In chapter 5 we build upon chapter 4 by analyzing time series of tree cover and fire observations to quantify the strength of the fire-tree cover feedback loop along climatic gradients. From these empirical results we developed a spatially explicit and stochastic fire-tree cover model. The model predicts that forest and savanna are alternative stable states across rainfall conditions, but the exact rainfall range depends strongly on rainfall variability. Both higher seasonal and inter-annual variability in rainfall increase fire frequency, but only seasonality widens the hysteresis of forest and savanna. The strength of the fire-tree cover feedback also depends on the spatial configuration of tree cover and disturbances to it: landscapes with clustered deforestation are more susceptible to cross a fire-driven tipping point than landscapes with scattered deforestation.

Chapter 6 presents regional-scale effects of tree cover on rainfall in the Amazon basin. We used output from a hydrological model to estimate the local forest transpiration in the 21st century. We then simulated the trajectories of that transpired water through the atmosphere to where it rains down. We estimate that one-third of all rainfall in the Amazon basin originates from the basin; two-thirds of that water has been transpired by trees at least once. Our calculations show that forests in the southern half of the basin contribute most to the resilience of other forests, whereas forests in the south-western Amazon are most dependent on tree transpiration from elsewhere in the basin. The relative contribution of tree transpiration to rainfall is higher in drier months and in drier years. This means that forests buffer against droughts through transpiration.

In chapter 7 I connect the presented results across scales and introduce some new analyses to discuss how the integration of the previous chapters may help identifying critical limits to land-use change in the Amazon.

Considering the results from across this dissertation, I conclude that tree cover will not change smoothly with climate change and that possible transitions between forest and savanna will likely be relatively abrupt and hard to revert. The main mechanism behind such tipping points is a feedback between tree cover and fire. Increasing rainfall seasonality will strengthen that feedback and in the Amazon, reduced tree transpiration resulting from tree-cover loss will enhance that seasonality.