Vegetation-fire feedbacks as savanna determinants

Background/Question/Methods

Savannas are widespread ecological communities characterized by a discontinuous overstory of trees in a graminoid-dominated groundcover.  Disturbance is one process put forward to explain the co-occurrence of grasses and trees in savanna communities.  Disturbance models of savannas propose that recurring external disturbances impose strong limits on overstory tree density, preventing the competitive displacement of the graminoid understory. The frequency and intensity of some disturbances such as fire, however, are largely determined internally to the savanna system by vegetation composition and characteristics rather than by external climatic conditions. Disturbance models that include feedbacks between vegetation and fire have demonstrated that these internal feedbacks can stabilize ecosystems in a savanna state.  Models that incorporate fire-vegetation feedbacks, however, do not usually consider space, despite fire spread being an inherently spatial process.  We investigate a spatial version of a disturbance model that incorporates feedback between vegetation and fire using a cellular automaton (CA).   Our savanna CA incorporates state transitions between grass, savanna trees, and forest trees, while allowing for fire and hurricane disturbance in the landscape. 

Results/Conclusions

We show that a feedback of grass and savanna trees on spatially-localized fire probability can maintain an ecosystem in a savanna state.  This result is consistent with non-spatial disturbance models that also incorporate vegetation-fire feedbacks.  Our CA predicts that distinct spatial patterns are associated with vegetation-fire feedbacks.  An over-dispersed spatial patterning of large trees, for example, results from positive vegetation-fire feedbacks, which is consistent with empirical observations from some savannas.  The absence of this internal feedback in our CA results in an abrupt transition from grassland to forest both spatially within the landscape and globally as fire becomes less frequent:  no intermediate savanna state is reached.  We suggest that internal feedbacks may be widespread in savanna systems, facilitating the coexistence of grasses and trees.