Increase in the quality of the prediction of a computational wildfire behavior method through the improvement of the internal metaheuristic

• ESS-IM is an uncertainty reduction method applied to wildfire spread prediction.
• ESS-IM is based on Fire Simulation, HPC and Parallel Evolutionary Algorithms.
• ESS-IM has been validated using a set of real burns.
• ESS-IM uses multiple overlapping solutions which benefits the fire prediction.
• ESS-IM increases the quality of the prediction by incorporating the Island Model.

Wildfires cause great losses and harms every year, some of which are often irreparable. Among the different strategies and technologies available to mitigate the effects of fire, wildfire behavior prediction may be a promising strategy. This approach allows for the identification of areas at greatest risk of being burned, thereby permitting to make decisions which in turn will help to reduce losses and damages. In this work we present an Evolutionary-Statistical System with Island Model, a new approach of the uncertainty reduction method Evolutionary-Statistical System. The operation of ESS is based on statistical analysis, parallel computing and Parallel Evolutionary Algorithms (PEA). ESS-IM empowers and broadens the search process and space by incorporating the Island Model in the metaheuristic stage (PEA), which increases the level of parallelism and, in fact, it permits to improve the quality of predictions.