Ajuste evolutivo de um modelo baseado em autômatos celulares para representar a dinâmica de propagação de incêndios florestais

Abstract

Due to climate change, the occurrence of fires in various biomes, whether fire-adapted or not, has significantly increased in recent decades, causing environmental damage, some of which is irreparable. In this context, computational models based on Cellular Automata (CA) can simulate and predict fire propagation behavior, assisting authorities in making control and prevention decisions. However, constructing and tuning the parameters of these models are complex tasks due to the number of variables involved. Environmental characteristics directly influence fire behavior. Thus, a tool capable of performing automatic parameter tuning based on Genetic Algorithms (GA) has proven to be promising. Therefore, the objective of this research is to develop a GA capable of properly adjusting a set of parameters of a forest fire propagation model based on CA to replicate the fire dynamics represented in a dataset of monitored region fires. Additionally, a new parameter, “sampling rate” (SR), is proposed to improve the adaptability of the CA model to fire propagation dynamics, regardless of the sampling interval in the reference data. The impact of different fitness functions on the algorithm’s convergence was also evaluated. During the experiments, artificial datasets were used as references, generated from simulations with our own CA model as well as another fire propagation model from the literature, to validate the robustness of the approach. The results showed that the GA successfully found appropriate parameters for the fire propagation model, including the number of CA steps required to represent fire evolution in accordance with the reference data generated by both models.