Evaluating wildfire simulators using historical fire data

Abstract

High-performance wildfire simulators allow the future location of a wildfire to be rapidly predicted. The accuracy of such simulators needs to be evaluated; this can be achieved by comparing simulated and observed spread for documented historical fires. A key issue relates to the accuracy of data obtained from historical fires, such as the time-varying fire location, fire-ground weather and accuracy of fuel type, load and structure data. A methodology used to evaluate the accuracy of wildfire simulators using historical fire data is presented and applied to the AUSTRALIS wildfire simulator using the four distinct phases of a large-scale wildfire occurring in Western Australian sand-plain heathlands and a fire reconstruction report on this fire produced by a wildfire expert. Challenges encountered in performing this validation exercise are highlighted.