Inferring resilience of forest ecosystems from satellite data

Background/Question/Methods

Predicting ecosystem resilience to climate and disturbance events is one of the most urgent and challenging problems in ecology today. We used the MODIS remotely-sensed estimations of tree cover (%) at 1 km² to explore the patterns of tree cover distribution in tropical regions of South America, Africa, and Australia, and their relationship with climate. We used the tree-cover frequency distributions to infer ecological resilience and to map the potential vulnerability of tropical ecosystems in the three continents. 

Results/Conclusions

We found striking tri-modal distributions corresponding to treeless (< 5 %), savanna (20 %), and forest (> 60 %) states which are strongly associated to mean annual precipitation. The transitional conditions between treeless-savanna (5-10 % tree cover) and between savanna-forest (55-60 % tree cover) occur at very low frequencies, which we interpret as an indication of transient unstable conditions. We found that not only changes in mean annual precipitation, but also rainfall variability can affect  the probability of shifting between alternative ecosystem states.