An analysis of arid lands around the world shows how patterns in vegetation may serve as harbingers of things to come.
Society has an increasing awareness that there are finite limits to what we can expect the planet to absorb and still provide goods and services at current rates1. Both historical reconstructions and contemporary events continue to remind us that ecological regime changes are often abrupt rather than gradual. This reality motivates researchers who seek to discover leading indicators for impending ecosystem change. Berdugo et al.2 report an important advance in our ability to anticipate the conversion of arid lands from self-organized, self-maintaining and productive ecosystems, to a state characterized by disorganization and low functionality. Such conversions have important implications for our understanding of ‘desertification’ — which is a shift from arid to desert-like conditions.
Theoretical studies have suggested that patterns in the patchiness of vegetation might indicate how close a system is to making an abrupt change to desert-like conditions3,
Various approaches to identifying leading indicators of ecological change have been identified10. When trend data are available, characteristics of temporal dynamics, such as increasing fluctuations, are thought to provide useful predictive signals11. It is perhaps less obvious that spatial configurations may serve as indicators of potential change as well12. Contemporary progress in the study of spatial patterns began with the observation that vegetation tends to have a characteristic, non-random appearance in arid lands. The characteristic pattern, in general terms, is one where there are a few large and many small patches, with vegetation-free spaces between (Fig. 1).
Theoretical work on desertification indicators leads to the expectation that the sizes of patches in a sample follow a mathematical pattern called the power law13. In this type of equation, only a single exponent is needed to describe the relationship between patch sizes and the frequency of sizes in a sample. A presumed mechanism behind this pattern is that vegetation patches increase water percolation into the soil in the vicinity of patches, improving their favourability for plant growth. This produces a facilitative mechanism where vegetation promotes vegetation (and the absence of vegetation promotes the loss of water and absence of vegetation)12. However, it has been noted that in nature, vegetation does not always adhere to a strict power law4,6,8,9. The mix of power-law and non-power-law distributions that have been observed is of great interest as a potential signal that locations vary from those that are strongly self-organized to those that lack self-organization, respectively.
Berdugo et al.2 tackled the task of attempting to resolve alternative claims about the merits of using total cover versus patch-size distributions as indicators by starting with a pluralistic approach capable of identifying the roles of either vegetation indicator type. To develop a robust answer, they sampled 115 sites across 4 continents. In addition to vegetation, they sampled soils and obtained climate data. Three kinds of analysis were performed to provide the ingredients for their recipe for success. First, they attempted to fit vegetation patch-size distributions from each site to a power law. The results led them to distinguish between sites that did fit a power-law function and those best described as of non-power-law form. Second, they computed an index of overall ecosystem functionality for each site. If the soil conditions suggested a correlated suite of conditions that promote the supply of water and nutrients, the index was given a high score. If only a few or none of the soil conditions were favourable for productive conditions, the functionality index was given a low score. A major discovery from this part of the study was that ecosystem functionality followed a bimodal (double humped) distribution, indicating the presence of two alternative states with regard to functioning. This bimodality was not detected in total vegetation cover, but instead by patch size distribution metrics. Thus, it appears that an integrative measure of function was required to detect alternative system condition. Finally, the authors used causal network models to provide for a simultaneous test of the various ideas. The results were consistent. Organized systems that followed the power-law distribution were shown to be controlled by different mechanisms than disorganized systems that deviated from the power-law expectation. More specifically, results showed that biotic processes mediated through productivity and facilitation are important for organized systems while disorganized systems lacked internal controlling processes.
There are interesting possibilities for extension of this work. The authors demonstrated that many of the necessary measurements for understanding these questions are feasible. Therefore, more extensive sampling could permit tests of the robustness of their conclusions, as well as integration of our understanding of temporal and spatial indicators. It is also noted that only a subset of the sites in their study included measures of facilitation. Finally, the effect of aridity on facilitation as a mechanism for this process was untested in their analysis. There is great societal value in this scientific pursuit and we can expect additional major advances from this field in the future.