A probabilistic analysis of fire-induced tree-grass coexistence in savannas

Fires play an important role in determining the composition and structure of vegetation in semiarid ecosystems. The study of the interactions between fire and vegetation requires a stochastic approach because of the random and unpredictable nature of fire occurrences. To this end, this article develops a minimalist probabilistic framework to investigate the impact of intermittent fire occurrences on the temporal dynamics of vegetation. This framework is used to analyze the emergence of statistically stable conditions favorable to tree- grass coexistence in savannas. It is found that these conditions can be induced and stabilized by the stochastic fire regime. A decrease in fire frequency leads to bush encroachment, while more frequent and intense fires favor savanna- to- grassland conversions. The positive feedback between fires and vegetation can convert states of tree- grass coexistence in semiarid savannas into bistable conditions, with both woodland and grassland as possible, though mutually exclusive, stable states of the system.