A rate of spread index for fires in spinifex fuels

Fires in spinifex occur throughout arid and semi-arid parts of Australia and in some cases can affect large tracts of the landscape with associated environmental impacts. In response to this environmental challenge an empirical model for the prediction of fire spread rate in spinifex fuels has recently been developed, based on a number of experimental burns conducted in Western Australia. In other research related to fires in grasslands, a simple rate of spread index for quasi-equilibrium fire spread was developed and, despite its simplicity, was shown to provide practically identical output to current operational grassland fire spread prediction models. This simple rate of spread index for grasslands conceptualises the rate of fire spread as wind speed divided by fuel moisture content, where fuel moisture content is estimated using a fuel moisture index (FMI). Such a conceptualisation embodies the notion that fires will spread faster in windier conditions and in fuels that are drier. The rate of spread index, as it applied to grassfires, also incorporates a term that accounts for an intensity-dependent indraft that counters the prevailing winds at the fire line. As such, the rate of spread index can be viewed as a two parameter model for quasi-equilibrium fire spread. In this paper we investigated the performance of the rate spread index when applied to the discrete spinifex fuels of arid and semi-arid Australia. The performance of the rate spread index was evaluated through the use of empirical data relating to fires in spinifex and through comparison with the existing spinifex fire spread model. The results indicated that the rate of spread index, as it was applied to grassfires, was only able to account for 68% of the variation in the observed rates of spread. Multiplication of the rate of spread index by fuel cover improved it's predictive ability to 73%, but this was still not as good as the existing spinifex model, which could account for 83% of the variation in the data. The main reason for this relatively poor performance of the rate of spread index was found to be due to the fact that the FMI did a poor job of estimating the moisture contents of spinifex fuels. As such, we concluded that application of the FMI should be restricted to more temperate fuel types, for which it has been shown to work quite well. An alternate form of the rate of spread index, using actual fuel moisture content rather than the FMI, was considered and found to produce much more accurate predictions. Indeed, when multiplied by fuel cover, this alternate rate of spread index was able to account for 85% of the variation in the observed rates of spread, thereby slightly outperforming the existing models for spinifex. The final version of the rate of spread index can be expressed as a function of fuel cover c, 2m wind speed U and profile fuel moisture content m: S(U, m, c) = 37c^{max(1, U)}, m with corresponding rates of spread well predicted by the model R^*(U, m, c) = 1.5S(U, m, c) + 600. These results have implications for the parsimony of fire behaviour models and demonstrate how conceptual and pedagogical simplifications can be incorporated into fire spread models with no practical loss in model performance. The results are also relevant to the possible unification of fire spread models across different fuel types.