Towards Data-Driven Simulations of Wildfire Spread using Ensemble-based Data Assimilation

Real-time predictions of a propagating wildfire remain a challenging task because the problem involves both multi-physics and multi-scales. The propagation speed of wildfires, also called the rate of spread (ROS), is indeed determined by complex interactions between pyrolysis, combustion and flow dynamics, atmospheric dynamics occurring at vegetation, topographical and meteorological scales. As a wildfire generally features a front-like geometry at regional scales, current operational models simulate it as a propagating front at a ROS based on a semi-empirical model due to Rothermel. In these models, the ROS is treated as a simplified function of vegetation, topographical and meteorological properties. For the fire spread simulation to be predictive and compatible with operational applications, the uncertainty on the ROS model should be reduced. As recent progress made in remote sensing technology provides new ways to monitor the fire front position, a promising appro ach to overcome the difficulties found in wildfire spread simulations is to integrate fire modeling and fire sensing technologies using data assimilation (DA). For this purpose we have developed a prototype data-driven wildfire spread simulator in order to provide optimal estimates of poorly known model parameters [*]. The wildfire spread simulation capability is adapted for more realistic wildfire spread: it considers a regional-scale fire spread model that is informed by an assumed set of real-time observations of the fire front location. An Ensemble Kalman Filter algorithm based on a parallel computing platform (OpenPALM) was implemented in order to correct parameters of the ROS model that relate to vegetation, wind and topography. The EnKF algorithm shows its good ability to track a small-scale controlled grassland fire and ensures a good accounting for the sensitivity of the simulation outcomes to the control parameters.
[*] Rochoux, M.C., Delmotte, B., Cuenot, B., Ricci, S., and Trouvé, A. (2012) "Regional-scale simulations of wildland fire spread informed by real-time flame front observations", Proc. Combust. Inst., 34, in press

up arrow