Analysis of Sequential Monte Carlo Methods in Dynamic Data Driven Simulation of Wildfire Spread

摘要

Dynamic data driven simulation (DDDS) represents a new simulation paradigm where a simulation system continuously assimilates real time data for better simulation results. In previous work, we developed DDDS that assimilates real time data into a wildfire spread simulation model. The developed DDDS uses Sequential Monte Carlo (SMC) methods, also called particle filters, to recursively adjust the estimations of a real wildfire's system states when new observation data becomes available. In this paper, we analyze the robustness of the SMC-based method in dynamic data driven simulation of wildfire. Our analysis is based on several metrics that measure the convergence, degeneracy, and sample impoverishment of the developed method for wildfire spread simulations. Results of this analysis show the effectiveness of DDDS based on SMC methods and provide a guideline for developing more advanced techniques to improve simulation results.