An empirical-based model for predicting the forward spread rate of wildfires in eucalypt forests

Abstract

Reliable and accurate models of the speed of a wildfire front as it moves across the landscape are essential for the timely prediction of its propagation, to devise suitable suppression strategies and enable effective public warnings. We used data from outdoor experimental fires and wildfires to derive an empirical model for the rate of fire spread in eucalypt forests applicable to a broad range of wildfire behaviour. The modelling analysis used logistic and non-linear regression analysis coupled with assumed functional forms for the effect of different environmental variables. The developed model incorporates the effect of wind speed, fine dead fuel moisture, understorey fuel structure, long-term landscape dryness and slope steepness. Model evaluation against the data used for its development yield mean absolute percentage errors between 35 and 46%. Evaluation against an independent wildfire dataset found mean percentage errors of 81 and 84% for two landscape dryness conditions. For these wildfires, the mean error was found to decrease with increasing rates of spread, with this error dropping below 30% when observed rates of spread were greater than 2 km h−1. The modular structure of the modelling analysis enables subsequent improvement of some of its components, such as the dead fuel moisture content or long-term dryness effects, without compromising its consistency or function.