Back

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

Butler, K.; Tartaglia, E.; Rennie, J.; Deutsch, S.; McCarthy, N.

2025-08-02 ecology
10.1101/2025.07.31.668020 bioRxiv
Show abstract

Initial attack refers to the first set of firefighting actions undertaken at a newly reported fire, with the aim to control the fires spread. When multiple fires are reported and resources constrained, decisions need to be made about triaging fires for an escalated response. We developed empirical initial attack models for grass and forest fires for the specific purpose of triaging fires. We aimed to improve on previous models and evaluate the models relative to their intended application. We used a recently developed dataset of spreading fires and tested three statistical modelling techniques to model the probability of unsuccessful initial attack. Our results show that the models have limited skill overall and are unable to distinguish fires where bottlenecks affect the initial attack suppression. The best performance was seen in the case of lightning-ignited forest fires. Our improvements to the approach and validation did not give a reliable model for operational triage of fires, because there are many on-the-ground factors that influence how successful initial attack will be that are difficult to capture in data. This study shows the importance of taking an application specific approach to developing and validating initial attack models.

Matching journals

The top 8 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.