Data-driven wildfire behavior modeling: Focus on front shape similarity measure for data assimilation

A wildfire can be represented at regional scales as a propagating front separating the burning area and the vegetation. A front data assimilation system named FIREFLY has been developed at CERFACS in collaboration with the University of Maryland to better estimate the environmental conditions (biomass properties, near-surface wind) that are spatially-distributed inputs to the front-tracking simulator used for forecasting wildfire behavior. FIREFLY relies on an ensemble Kalman filter (EnKF) for inferring correction on the inputs from the distance between the simulated and observed fronts. Current efforts focus on the formulation of the "distance" between simulated fronts and observations with the objective of better representing the topology of the fire front under heterogeneous environmental conditions. The "distance" corresponds to a front shape similarity measure that is currently investigated for application in FIREFLY. The resulting front level-set data assimilation algorithm is promising to directly assimilate wildfire mid-infrared images.