Fire behaviour modelling in semi-arid mallee-heath shrublands of southern Australia

Classify & Cross-ref
Fire Behaviour
TitleFire behaviour modelling in semi-arid mallee-heath shrublands of southern Australia
Publication TypeJournal Article
Year of Publication2012
AuthorsCruz, MG, McCaw, LW, Anderson, WR, Gould, JS
JournalEnvironmental Modelling & Software
Date Published8/2012
AbstractKnowledge of fire behaviour potential is necessary for proactive management of fire prone shrublands. Data from two experimental burning programs in mallee-heath shrublands in semi-arid southern Australia were used to develop models for the sustainability of fire spread, fire type, i.e., surface or crown fire, forward spread rate and flame height. The dataset comprised 61 fires burned under a wide range of weather conditions. Rates of fire spread and fireline intensity varied between 4 and 55 m min−1 and 735 and 17,200 kW m−1 respectively. Likelihood of sustained fire spread and active crown fire propagation were modelled using logistic regression analysis. Fire spread sustainability was primarily a function of litter fuel moisture content with wind speed having a secondary but still significant effect. The continuity of fine fuels close to ground level was also significant. Onset of active crowning was mostly determined by wind speed. Rate of fire spread was modelled separately for surface and crown fires through nonlinear regression analysis with wind speed, litter fuel moisture content and overstorey canopy cover as significant variables. Flame height was modelled as a function of fireline intensity. A model system to predict the full range of fire behaviour in mallee-heath shrubland is proposed relying on a combined method that links the surface and crown fire rate of spread models. This model system was evaluated against independent data from large scale prescribed burns and wildfires with encouraging results. The best models for fire-spread sustainability and active crown fire propagation predicted correctly 75% and 79% respectively of the fires in the evaluation dataset. Rate of spread models produced mean absolute percent errors between 53% and 58% with only small bias. The models have applicability in planning and conducting prescribed fire operations but can also be extended to produce first order predictions of wildfire behaviour.
Short TitleEnvironmental Modelling & Software
Refereed DesignationRefereed