Estimation of new ignited fires using particle filters in wildfire spread simulation

Assimilating real time data into wildfire spread simulations has the potential to improve simulation results of wildfires, which are complex and dynamic in nature. Our previous work developed a data assimilation method based on particle filters (PF) to estimate the state of a wildfire. This method, however, does not work effectively when there are significant events, such as new ignited fires, that change the fire spread behavior. This paper proposes a new method to estimate new ignited fires based on PF. The developed data assimilation method uses prior knowledge to design the proposal distribution of PF. Experiment results of this method, compared to those from several other proposal distributions, indicate that the developed method can effectively improve simulation results.