Savannas cover approximately 12% of the global land cover, and harbour both a high biodiversity as well as a large proportion of the world’s pastoral systems. Many savannas experience sudden increases in woody cover, known as bush encroachment. A possible mechanism is the positive feedback between grass biomass and fire intensity. Model studies suggest that when grazing reduces grass biomass below a certain threshold, fire intensities are not high enough to suppress tree biomass, thus allowing trees to dominate. This feedback can cause sudden shifts from savanna vegetation to woody dominance. Moreover, hysteresis can occur. This implies that to go back to the state with both trees and grasses, grazing has to be reduced to levels far below the point where bush encroachment occurred. Fire is important in this feedback, but the effect of fire on trees might not be the same everywhere. Savanna fires vary spatially in intensity, temperature, flame height and rate of spread, which is caused by weather conditions, fire regime, local climate, and the vegetation itself. This results in a mosaic in severity of fire damage on the woody component in the landscape and accordingly tree survival. What are the causes of this heterogeneity in tree-grass ratios, and what are the consequences for savanna stability?

Results/Conclusions

Using the long-term experimental burning plots in Kruger National Park, South Africa, we show that fire can create patterns in the woody cover. Trees are more clustered when fire frequency is high. This spatial heterogeneity of savanna vegetation as a result of fire has large consequences for their vulnerability to become bush encroached. We present a model study where we investigate the effect of fire as spatial process (the effect of fire is spatially explicit) on the stability of savannas. We show that sudden increases in woody cover as result of increasing grazing are replaced by a more gradual change from tree-grass co-occurrence to woody dominance when fire is included as a spatial process. However, without a sudden shift, the hysteresis effect still remains. This means that although the change observed may be gradual, it is still irreversible. We also analyzed the spatial scale of fires. We conclude that small-scale fire positively affects the stability of savannas, while large-scale fire has a negative effect. We argue that understanding the interaction between the scale of spatially-explicit processes and spatial heterogeneity is crucial to conserve the balance of trees and grasses in savannas.