The FireDST project used the fire behaviour modelling from the PHOENIX RapidFire bushfire simulator. PHOENIX RapidFire had been developed as part of the original Bushfire CRC as a proof of concept tool. However, it had been applied operationally with limited formal validation and assessment.
The researchers' goals for FireDST involved gauging past work and future possibilities. Associate Professor Kevin Tolhurst led the project, working with researchers Dr Thomas Duff and Derek Chong, also from the University of Melbourne. They sought to validate and assess the simulations from PHOENIX RapidFire, pushing the boundaries by investigating how much improvement there might be in fire behaviour predictions if higher spatial and temporal resolution weather data were used.
PHOENIX was tested under different conditions using three selected case studies: The Warragamba bushfire in the New South Wales Blue Mountains on 25 December 2001; the Wangary bushfire on the Eyre Peninsula, South Australia in January 2005; and the Kilmore East bushfire on Black Saturday Victoria, 7 February 2009. These fires were selected as they covered a range of terrain, topography and fire severity.
The sensitivity of PHOENIX to a range of inputs was tested and quantified using the Area Difference Index developed as part of this project. This was a novel method as no suitable methods or statistics existed.
The researchers developed practical mathematical methods to objectively assess the accuracy of fire spread predictions against known outcomes. They used a range of spatial and temporal resolution weather data in PHOENIX and quantified the effect on fire-spread prediction. They also modified PHOENIX to incorporate upper level winds in the ember transport and spotting process.
The research has reinforced the importance of spotting in fire behaviour and its inclusion in the modelling process.
Through this study, it is now possible to objectively quantify the accuracy of a fire spread prediction and identify aspects of the prediction that are most in error, such as orientation or extent.
Although PHOENIX RapidFire is used operationally and for training purposes by many Australian fire agencies there is still a need to use this new research to improve the output information that shows the level of uncertainty in predictions.
Three postgraduate studies contributed to this work. The University of Melbourne’s Lisa Cheong focused her PhD research on representing and communicating uncertainty in visual information (maps) with specific reference to bushfires. She conducted a series of experiments on whether people stayed or left during a bushfire.
Kangmin Moon, also of the University of Melbourne, examined the impact of wind on fire behaviour for his PhD. He measured and characterised wind in different forest environments to improve the accuracy of fire behaviour models.
PhD student Mahfuz Sarwar, of Victoria University, applied a fluid dynamics methodology to assess the performance of eddy viscosity models and their potential as forecasting tools. He has found that the models were particularly grid sensitive, with variations in resolution changing the outcomes of the forecasts.