We generalize the forest-fire model of P. Bak et al., which contains a tree nearest growth probability p and fire spreading to the neighbors, by including a lightning probability f and an immunity g which is the probability that a tree catches no fire although one of its neighbors is burning. The model becomes self-organized critical in the limit f/p→0, provided the time scales of tree growth and burning down of forest clusters are separated. The size distribution of forest clusters obeys a power law. We calculate the critical exponents in one dimension. A continuous phase transition is observed in the general forest-fire model when g reaches its critical value. We determine the critical line gC(p) and show that the critical fire propagation represents a new type of percolation. Finally, we point out similarities between the forest-fire model and excitable media, which comprise such different systems as chemical reactions, spreading of diseases and populations, and propagation of electrical activity in neurons.
Modeling information diffusion in online social networks using a modified forest-fire model