A fire spread index for grassland fuels

Abstract

Fires in grasslands occur throughout the temperate parts of the globe and in some cases, such as was seen in the 2005 Eyre Peninsula (Wangary) fires, can have significant impacts. To address these issues a number of different models for predicting the rate of spread of fire in grasslands have been developed. In this paper we introduce and discuss a novel index, which incorporates fire weather information and relates to the rate of fire spread in grassland fuels. This so-called spread index has the following form: . The spread index model has two parameters, α and μ. The parameter α serves as a calibration constant that transforms values of the spread index into values that can be interpreted as rates of spread (in km h -1 ), whereas the parameter μ acts to temper the growth of the spread index as FMI → 0, and can be thought of as representing an intensity dependent indraft wind that counters the prevailing winds at the head of the fire. It is found that the values α = 2.4 and μ = 6 provide for reasonable overall performance. The performance of the index in predicting rate of spread was evaluated through the use of three datasets. The first of these comprises data recorded at Port Lincoln Airport (SA) surrounding the disastrous Wangary Fire (10-11 January 2005). The second is more representative of moderate summer conditions and is comprised of data recorded at Canberra Airport (ACT) in January 2007. The third dataset comprises information relating to twenty significant historical wildfires. These datasets were used as the basis for comparison of the spread index with an established model for grassfire rate of spread (the CSIRO Grassland Fire Spread model). The fire spread index was found to deliver rate of spread predictions that are practically equivalent to those derived from the established model. On average the differences in rate of spread predictions derived from the spread index and the CSIRO model were 0.2 - 0.3 km h -1 , with the biggest differences encountered during the most extreme fire weather associated with the Wangary fire. While neither model did a particularly good job of reproducing the estimated rates of spread of the historical wildfires, in the case of the most extreme fire behaviour during the Wangary fires, the spread index was found to produce the better rate of spread estimates. Overall, given the vagaries of estimating field values for rate of spread and the lack of experimental data for fires burning under extreme conditions, the simple, linearly-structured spread index was found to give reasonable estimates of rate of spread. Indeed, such estimates were not substantially different to those delivered by the nonlinear CSIRO rate of spread model. These results have implications for the parsimony of fire behaviour models; in particular they suggest a number of conceptual and pedagogical simplifications that could be made with only minimal changes in model performance. The spread index is also briefly considered in the context of a possible unification of fire spread models across different fuel types.