Evaluation of the probabilistic model of fire propagation using cellular automata applied to small areas

Abstract

The Municipal Park of Serra de São Domingos, located in Poços de Caldas, Minas Gerais state, Brazil, presents occurrences of annual fires, which promote environmental and socioeconomic losses. The use of a fire propagation model that could aid in the planning of combat, and even in the prevention of such fires, would be extremely beneficial for the municipality and for the Fire Department. The objectives of this work are the presentation of the fire studies in the area and the parameterization and evaluation of a probabilistic model of fire propagation, based on the cellular automata methodology, like the models used in the Cerrado biome region. The characterization is based on the use of remote sensing tools and meteorological data, and the parameterization of the model used data from a fire with well-defined affected area and well-known ignition point, and the model was adjusted by applying the metaheuristic algorithm of firefly swarm. The efficacy of the model was later tested with two distinct fire occurrences, both with well-struck area hit, but an unknown point of ignition. The model was able to represent up to 57% of the real area affected by the fire, and some parameters were considered satisfactory, such as the elementary probabilities of sustainability and fire ignition. However, the studies indicate that there is a need for greater refinement in the data used in the model, mainly data on the moisture and ignition point of fires.