Predicting daily human-caused bushfire ignitions

Fire Note 123: This research developed models that predict the number and probability of human- caused bushfires per day in south west Western Australia from bushfire incident records and weather data.

Significant predictor factors used in these models included fuel moisture, the number of recent bushfires, the day of the week (or type of day, such as public holidays) and rainfall.

Analysis shows that days with human-caused fires are more likely to occur on weekends and public holidays that coincide with days of low fuel moisture content, as well as days that follow periods of high fire activity.

The models performed well in regions around Perth, with reasonable accuracy of between 81 and 99% of the daily fire occurrences having daily fire counts within the predicted range. The models for these areas would be suitable for agencies to use to inform their daily operational resource planning. However, the models did not exhibit much day-to-day prediction variation in the regions along the southern coast, which experiences much fewer human-caused ignitions.

This study has relied on fire incident records, and demonstrates the importance of high quality data for allowing similar sorts of analyses and modelling.