Computational science for forest fire prediction

Abstract

Forest fires are a very serious hazard that, every year, causes significant damage around the world from the ecological, social, economical and human point of view. These hazards are particularly dangerous when meteorological conditions are extreme with dry and hot seasons or strong wind. The fire fighting should have at its disposal the most advanced resources and tools to help the use of available resources in the most efficient way to diminish fire effects as much as possible. The problem of forest fire spread prediction presents a high degree of complexity due in large part to the limitations for providing accurate input parameters in real time (e.g., wind speed, temperature, moisture of the soil, etc.). The inaccuracies present in the measurements, the models, and the computational implementation constitute different sources of uncertainty. This uncertainty has led to the development of computational methods that seek to obtain better predictions. In this article, we present a line of research for the development of uncertainty reduction methods for the prediction of propagation phenomena (so called DDM-MOS in the taxonomy). Each method in this family combines the strength of different elements: evolutionary computing, statistics, parallel computing, and novelty search, to make decisions according to the result and trend of a set of simulations. Due to the characteristics of the methods, this approach is also feasible to be applied for the prediction of other types of propagation phenomena such as floods, avalanches, etc.