Randomizing activity in fire spreading simulations

UMR CNRS 6134, University of Corsica Campus Grimaldi, 20250 Corte, France

^a e-mail: ino@univ-corse.fr.

Abstract

In this paper, we suggest the use of activity paradigm to model the uncertain and imprecise nature of the local dynamics in fire spreading simulations. Formerly based on a very simple set of rules, Cellular Automata (CA) models are remarkable in their ability to easily simulate fire spreading behaviour. A new algorithm to track activity concept in a (CA) model is presented. It is called random-activity and it combines activity and stochastic precepts. This algorithm enhances the realism of the spatial simulations. It resolves the problem of the spurious symmetries of the square lattice usually encountered which affect negatively outputs. Experiment shows that the approach enhances fire spreading simulations which are based on local description of the fire dynamics with general descriptive laws. A comparison between classical and random-activity algorithms is performed. The random-activity algorithm proposed outperforms the classical algorithm in terms of graphical restitutions’ realism. It might be used to develop fire risk-management tools. A new way is presented to combine activity concept and stochastic modelling in a (CA) model.