Ecosystems might show an abrupt, non-linear response to a small, gradual change in the environmental driver. Theory suggests that such abrupt changes occur in systems where multiple states can exist under similar climatic conditions. Although mathematical models confirm the existence of alternative stable states in various systems, empirical evidence for multistability in large scale ecosystems is limited. In the past decade, a few studies have used remotely-sensed data to test for multistability in terrestrial ecosystems. These studies showed distinct modes in frequency-distribution of vegetation cover in Africa and South America (Hirota et. al. 2011, Staver et. al. 2011). To analyse the frequency distribution of vegetation cover, these studies used MODIS (Moderate-Resolution Imaging and Spectroradiometer) based estimates of tree-cover. However, the reliability of these results have been questioned because they were based on data that had been pre-processed using a clustering algorithm that may have overrepresented the occurrence of certain ecosystem types (Hanan et. al 2012). Therefore, despite the importance of studying bi-stability for the resilience of the global forest system, reliable evidence of alternative stable states in terrestrial ecosystems remains lacking.
Results/Conclusions
To evaluate the possibility for bimodality in the global forest system, we use a unique forest cover dataset which is entirely independent of MODIS data and any associated data clustering. This dataset contains over 1,00,000 estimates of forest cover , each measured by photo-interpretation at a high spatial resolution (< 1m) . Histograms of tree cover reveal three distinct modes of forest cover when looked at an appropriate scale, supporting the idea of three stable states in terrestrial ecosystems as proposed by Hirota et. al 2011. In addition, our independent dataset (that has not been clustered into forest cover types) enables us to build on this observation by accurately quantifying each of the three ecosystem states. Specifically, these modes correspond to grassland or desert with zero tree cover, savanna with intermediate (30%) tree cover and forest with high (100%) tree cover. Massive-scale perturbations to the global forest system might have the potential to shift forests between these three stable states, with direct consequences for global biogeochemistry.