Validation of the Australis Wildfire Simulator Using A Large-Scale Historical Fire
Presentation at Research Forum of the 2012 Bushfire CRC and AFAC Annual Conference.
The Australis wildfire simulator allows the progression of a wildfire to be rapidly forecast and projected as a series of isochrons onto a map of the landscape. However, the accuracy of the fire spread predictions depends upon (i) the quality of the input data, such as the location of the fire perimeter, the vegetation type, fuel load, and attendant weather conditions; and (ii) the fire spread models used, including rate-of-spread estimates, fire shape assumptions and length-to-breadth ratios. Four distinct phases of a large-scale fire in Western Australia, the Boorabbin fire, were independently simulated using the Australis simulator. Rate-of-spread estimates through the predominantly sandplain heath vegetation of the fire-site were calculated according to a Mallee-heath rate-of-spread meter using weather observations from the Southern Cross AWS as input variables. The simulated fire spread perimeters were compared to a highly detailed reconstruction of the fire obtained from a report to a WA coronial inquiry. The study found that (i) the rate-of-spread meter under-predicted during two distinct phases of the fire when wind speeds exceeded 30 km/h for significant periods; (ii) conventional length-to-breadth ratios for grassfires over-predicted flank fire spread; and (iii) an elliptical fire shape model over-predicted fire shape sizes resulting from wind direction changes. The Mallee-heath rate-of-spread meter was refined to produce spread rate estimates consistent with both the severe and the less severe phases of the Boorabbin fire. Results confirm that the rate-of-spread meter under-predicted for severe weather. The results give guidance as to the appropriate rate-of-spread and length-to-breadth ratio for fires burning in the heterogeneous sandplain heath fuel of the fire-site in severe weather. We emphasize that the refined Mallee-heath meter derived for this study provides an approximate guideline only, as it is based on data from a single fire only.