Having written (as in, programmed, based on process knowledge from fire departments) a model for forest fires in Australia myself over the last few months, I feel that I can say that while a fine demo application of CA based modeling, this model is not an accurate description of forest fir behaviour (well unless it's just to illustrate how planned burns interact with their crude approximation of fuel load). E.g. it doesn't even consider slope and wind direction, or vegetation type.