Research findings

The HighFire Risk Project



Citation Sharples, J.J. & Matthews, S. (2011). Evaluation of some simplified models for predicting the moisture content of fine, dead fuels. 19th International Congress on Modelling and Simulation, Perth.

Evaluation of some simplified models for predicting the moisture content of fine, dead fuels.

Author(s) J.J. Sharples and S. Matthews.
Abstract Operational prediction of wildfire behaviour requires assessment of the moisture content of fine, dead fuels to within an acceptable degree of accuracy. Ideally the methods of assessment should be simple enough to implement in most operational settings, including those where computational power is a constraining factor. In this paper we compare fine fuel moisture observations relating to two different fuel types with predictions derived from a number of fuel moisture models. The models considered are: the empirical fuel moisture sub-model of the Western Australian Forest Fire Behaviour Tables (“Red Book”); a process-based model which accounts for heat and moisture fluxes within surface litter; two simplifications of the process-based model; and a very simple model based on a fuel moisture index defined in terms of the difference between air temperature and relative humidity.
The study utilises two sets of fuel moisture data collected, respectively, in Jarrah (Eucalyptus marginata) and Karri (Eucalyptus diversicolor) forests. These two species dominate the forested areas of southwest Western Australia. Specifically, the study considers surface fuel moisture content, which is the moisture content of the top 10mm of the litter bed, measured daily at 14:00 hours over the period 13 October 1982 – 15 March, 1983.
The predictive ability of the various models is evaluated through comparison of the model predictions with observed fuel moisture contents. While changes in the climate over southwest WA since 1975 mean that the data may not be entirely representative of the full range of fine fuel conditions under typical current climatic conditions, the data sets do permit exploration of the various model’s performance over the space of relevant meteorological variables.
In the present paper the predictive ability of the fuel moisture models is mainly considered in the context of fire management. Of particular interest is the predictive ability of the simpler models at the lower end of the fuel moisture continuum, especially at or below the flammability limit (approximately 25%). The specific interest in the performance of the simpler models is due to the fact that these models are the most easily implemented in a field setting and hence will be of the most direct use to fire-ground personnel. Model performance is evaluated through consideration of correlation and error statistics, a simple measure of prediction bias and the proportion of model predictions that match observed fuel moisture contents to within a specified tolerance.
It is found that the simple models perform quite well; in fact they outperform the more sophisticated models in a number of the evaluation measures. The results of the study have implications for model engineering and parsimony that are of particular relevance to the application of fine fuel moisture models during bushfire operations.
Keywords Fuel moisture content, fuel moisture modelling, bushfire, fire management