A local ensemble transform Kalman filter data assimilation for a regional Eulerian level set wildfire model

Wildfire prediction problem is highly complex and faces a wide range of challenges in all elements of data assimilation. The purpose of this study is to build a foundation for the improvement of the forecast wildfire locations and burn probability by data assimilation based on existing wildfire simulation-forecast system and observations. An efficient data assimilation method of local ensemble transform Kalman filter (LETKF) is applied to a Eulerian Level set Model of Fire, called the ELMFire. The ELMFire model, starting from initial conditions with burning points or areas, numerically tracks the burning fronts, defined as the interface between active burning areas and those with covered with combustible fuels, with physics options including surface fire spread rate and directions, ember lofting, trajectory tracking, spot fire formations. The LETKF, known for its computational efficiency, is combined with the ELMFire model to optimally estimate the model state variables (burning fronts) or model parameters (such as fuel distributions) following the observations from spaceborne instruments such as the Visible Infrared Imaging Radiometer Suite (VIIRS). In the case of estimating the best possible initial model state variable of burning fronts, the analyzed initial burning fronts that can generate better fire forecasts can be obtained by assimilating the detected fire in remote sensing observations. Additional experiments will be conducted aiming to improve the background fuel distributions as a demonstration of the capability for parameter estimation with the proposed ELMFire LETKF framework.