An application-specific approach to modelling the probability of unsuccessful initial attack on wildfires in Victoria, Australia

Abstract

BackgroundInitial attack refers to the first firefighting actions undertaken at a newly reported fire, with the intention to control the fires spread. When multiple ignitions are reported and resources constrained, decisions need to be made about triaging fires for an escalated response. AimsWe developed empirical initial attack models for grass and forest fires for the specific purpose of triaging fires, and evaluated the models relative to their intended application. MethodsWe used a recently developed dataset of spreading fires and tested three statistical modelling techniques to model the probability of unsuccessful initial attack. Model performance was evaluated with case studies, including major fire episodes. Key resultsThe models showed limited skill overall. The best performance was observed for lightning-ignited forest fires in remote areas, but performance was not adequate to support implementation on days with extreme fire behaviour. ConclusionsOur approach did not yield suitable models for operational triage of fires, because there are many on-the-ground factors that influence the success of initial attack, and they are difficult to capture as data. ImplicationsThis study shows the importance of developing and testing empirical models relative to their intended application. It also demonstrates the significance of unmeasured bottlenecks in wildfire suppression on initial attack outcome. Summary text for Table of ContentsThis study develops models to aid operational staff in triaging fires for an escalated response. We determined that model performance was not suitable for this use case. We found that using both modelling visualisations and case studies were fundamental in assessing the models applicability for operations.