In recent times there have been increasing efforts to integrate technology into wildfire management, especially in the fields of tactical monitoring and simulation. On the one hand, thermal infrared imaging (TIR) systems have been installed aboard surveillance aircraft including unmanned systems (UAS). On the other, there exists a variety of models and simulators able to forecast the fire spread. However, both fields currently present significant limitations. While relevant information is still extracted manually from aerial thermal imagery and is most times merely qualitative, simulators’ accuracy on fire spread prediction has proved insufficient. To solve these issues, this article presents a twofold methodology to couple meaningful automated wildfire monitoring with accurate fire spread forecasting. The main goals are to, firstly, automatically process aerial TIR imagery so that valuable information can be produced in real time during the event and, secondly, use this information to adjust a Rothermel-based simulator in order to improve its accuracy on-line. The fire perimeter location is tracked automatically through an unsupervised edge detector. Afterwards, an assimilation module uses the remotely sensed data to optimise the simulator's fuel and wind parameters, which are assumed to remain constant for a certain period of time. Subsequently, the optimum parameters’ values are used to issue a fire evolution forecast. All outputs are projected onto the corresponding Digital Terrain Model (DTM) and integrated into a Geographic Information System (GIS) for visualization. The global system was validated using two large-scale experiments. If these algorithms can be applied to a sufficiently rich and varied set of experimental data and further developed to cope with more complex scenarios, they could eventually be incorporated into a fire management decision support system.

Peer Reviewed