The simplest landscape fire model is a grid of cells, each unburned, burning or burned, updated on a fixed time step by a local rule: a burning cell ignites a neighbour with a probability, or after a delay, computed from the spread rate between them and the distance between their centres. It is fast, trivial to write, and easy to make probabilistic by drawing a random number per neighbour, which is why so many published spread models are of this class. Its characteristic defect is geometric. A cell has only four or eight neighbours, so fire can travel only along a handful of directions, and a fire that should be a smooth ellipse comes out as a diamond on a four-neighbour grid or an octagon on an eight-neighbour one. Spread along a diagonal is also over-counted unless the rule corrects for the longer distance. Remedies exist — larger neighbourhood templates, correction factors calibrated to the true elliptical shape — and every one of them is an admission that the grid, not the fire, chose the shape. After this Concept you can recognise grid-artefact fire shapes and say what caused them.
Cellular automata are a computational modeling formalism in which a system is represented as a grid of discrete cells, each assigned a state from a finite set, that evolves over discrete time steps a…