Fighting wildfires: predicting initial attack success across Victoria, Australia

Abstract

Background

The small portion of fires that escape initial attack (IA) have the greatest impacts on communities and incur most suppression costs. Early identification of fires with potential for escaping IA can prompt fire managers to order additional suppression resources, issue timely public warnings and plan longer-term containment strategies when they have the greatest potential for reducing a fire’s impact.

Aims

To develop IA models from a state-wide incident dataset containing novel variables that can be used to estimate the probability of IA when a new fire has been reported.

Methods

A large dataset was compiled from bushfire incident records, geographical data and weather observations across the state of Victoria (n = 35 154) and was used to develop logistic regression models predicting the probability of initial attack success in grassland-, forest- and shrubland-dominated vegetation types.

Key results

Models including input variables describing weather conditions, travel delay, slope and distance from roads were able to reasonably discriminate fires contained to 5 ha.

Conclusions and implications

The models can be used to estimate IA success – using information available when the location of a new fire can be estimated – and they can be used to prompt planning for larger fires.