Evaluation of the performances of the fire simulators FOREFIRE and FARSITE in Mediterranean areas
Résumé
Abstract This work presents the comparison between two fire simulation codes that implement two distincts wildfire front rate of spread (RoS) models. Farsite simulator, developed by Missoula's firelab, is based on the Rothermel model, and has been used for years by firefighters around the world. ForeFire, developed at the university of Corsica, is a novel simulation code that aims to perform simulation at very high resolution while keeping very quick simulation time on regular computers. Farsite benefits form a much larger user base and has been very successful at predicting fire front shape and behaviour of fires occurring in the rather large forests that may be found in North America. ForeFire aims mostly to simulate fires occurring in very complex landscapes, trying to take into account meter large details (small roads, large rock bodies), often present in Mediterranean region and that may have significant impact on the fire front behaviour. The fire RoS model used in ForeFire differs from the Rothermel model and provides a more physical formulation of the phenomenon's involved in wildfire propagation. The objective here is not to compare performance of the two simulation codes, but to verify and validate the ability of the two codes in Mediterranean landscape. Both codes serves different purposes as it is easier to change and try novel formulation of RoS and fuel models in ForeFire, and Farsite benefits from years of development to be readily useable for foresters and fire fighters around the globe. To perform the comparison, both codes are presented and compared on 4 fires, two in Corsica and two in Sardinia that have been selected to be representative of typical fires that may occurs in the region. For all simulations, inputs were strictly identical, at the exception of the fuel models that differs (slightly) for the two RoS models. Comparison is made using typical error indices, and presented using intensity and arrival time fields. If simulation time for both models is in the same (sub-minute) range, the simulation resolution was much higher for ForeFire. It appears that simulation results are rather similar, and compare well with observations available from those case studies.