An application of regression models for analysis and forecasting of fire extent

Abstract

Every year wildfires cause destruction that results in the loss of human and animal lives and economical and ecological losses. In addition, unpredictable forest fires make it very difficult to plan suppression actions in light of saving monetary resources while effectively fighting all the active fires. By creating a model that uses meteorological measurements such as wind and relative humidity, it should possible to prepare better in order to fight wildfires more effectively and ideally decrease the amount of physical and economic damages. A dataset of forest fires identified in Montesinho Natural Park, located in the mountainous northeast of Portugal, was used for application of our model. While the analyzed data were not collected real time, our design can be applied to real-time fire management. The study identifies specific actions combating wildfires with low cost equipment, suggesting when more specialized equipment like satellite imaging and smoke scanners provide the required information to fight fires, especially the largest ones.