On Thursday 26 May 2016, members of the "Fire" Project and the Environment Sciences Laboratory (CNRS-University of Corsica) held a seminar on the topic "Assimilation of data in forest fire simulation" intended for a scientific audience with specialist subject knowledge. It was facilitated by Mélanie Rochoux, a researcher at the Basic and Applied Research Centre specialised in Modelling and Numerical Simulation (CERFACS).
Assimilation of forest fire simulation data
Predicting forest fires remains a challenge as the rate and direction of spread depends on multi-level interactions between the plant matter, soil topography and weather conditions. Regional-level models are therefore not very suitable in terms of taking into account the detail of the physical processes at play.
Modelling is negatively affected by a number of uncertainties (incomplete modelling, insufficient knowledge of the terrain, vegetation and flame/atmosphere interactions, etc.) that need to be measured and corrected. This thesis proposes regional fire modelling with better simulation and prediction capabilities, based on model evaluation and data assimilation.
An evaluation of the models entailed developing multi-physical simulations at flame level, including solving reactive Navier-Stokes equations, evaluating the radiative transfer occurring in the direction of the vegetation, constructing a model of pyrolysis of plant matter and modelling of the flame/vegetation interface to better understand fire dynamics. The second approach put forward comprises putting into place a prototype for the assimilation of data for monitoring spread of the firefront. The idea is to correct the simulated trajectory of the firefront as new observations become available, where the difference between the observed and the simulated positions of the firefront translates into correction of the rate of spread parameters, or the firefront position directly using the Kalman filter algorithm as a whole.
Considering the existing uncertainties both in terms of fire modelling and the available observations, these approaches allow for an improvement of fire dynamics predictions and atmospheric emissions, which constitutes a major challenge for civil and environmental protection.