A wildland fire model with data assimilation (Q960358): Difference between revisions

The authors here present tools accounting for the simulation of ground layer fire. They start from the coupled non-steady system \(dT/dt=\nabla (k\nabla T)-\upsilon \nabla T+A(Se^{-B(T-T_{a})}-C(T-T_{a})\), \( dS/dt=-C_{S}Se^{-B(T-T_{a})}\), which describes the evolutions of the temperature and of the fuel supply mass fraction. Initial known values are given. This system involves several physical parameters which have to be determined. The authors show how some of these coefficients can be deduced from observations on the wildland fire, that is fire in a ground layer with unspecified thickness \(h\). The main part of the paper is devoted to the description of the notion of data assimilation in order to build a precise numerical simulation of the forest fire. The authors describe the notion of ensemble Kalman filter as a collection of independent simulations. This tool has already been applied with success in different other contexts. Within the present fire context, the authors show how the ensemble Kalman filter can be applied within the current context in a 2D situation. The authors first calibrate the values of some parameters in a 1D situation. They also quote some other values from the literature. In the 2D situation, the authors use central finite differences with respect to the space variables and an explicit Euler method with respect to the time variable. The numerical simulations are obtained introducing additive random perturbations to a comparison solution. The authors present numerical simulations of the ensemble mean and of the ensemble variance. The paper ends with a long list of references.