Citation

Clarke HG, Gibson R, Cirulis B, Bradstock RA, Penman TD. J. Environ. Manage. 2019; 235: 34-41.

Affiliation

School of Ecosystem and Forest Sciences, The University of Melbourne, Victoria 3363, Australia.

Copyright

(Copyright © 2019, Elsevier Publishing)

DOI

10.1016/j.jenvman.2019.01.055

PMID

30669091

Abstract

Considerable investments are made in managing fire risk to human assets, including a growing use of fire behaviour simulation tools to allocate expenditure. Understanding fire risk requires estimation of the likelihood of ignition, spread of the fire and impact on assets. The ability to estimate and predict risk requires both the development of ignition likelihood models and the evaluation of these models in novel environments. We developed models for natural and anthropogenic ignitions in the south-eastern Australian state of Victoria incorporating variables relating to fire weather, terrain and the built environment. Fire weather conditions had a consistently positive effect on the likelihood of ignition, although they contributed much more to lightning (57%) and power transmission (55%) ignitions than the 7 other modelled causes (8-32%). The built environment played an important role in driving anthropogenic ignitions. Housing density was the most important variable in most models and proximity to roads had a consistently positive effect. In contrast, the best model for lightning ignitions included a positive relationship with primary productivity, as represented by annual rainfall. These patterns are broadly consistent with previous ignition modelling studies. The models developed for Victoria were tested in the neighbouring fire prone states of South Australia and Tasmania. The anthropogenic ignition model performed well in South Australia (AUC = 0.969) and Tasmania (AUC = 0.848), whereas the natural ignition model only performed well in South Australia (AUC = 0.972; Tasmania AUC = 0.612). Model performance may have been impaired by much lower lightning ignition rates in South Australia and Tasmania than in Victoria. This study shows that the spatial likelihood of ignition can be reliably predicted based on readily available meteorological and biophysical data. Furthermore, the strong performance of anthropogenic and natural ignition models in novel environments suggests there are some universal drivers of ignition likelihood across south-eastern Australia.

Copyright © 2019 Elsevier Ltd. All rights reserved.

Language: en

Keywords

Bushfire; Ignition; Modelling; Risk; Wildfire