An investigation of the predictability of fire size and occurrence in central Australia on both daily and seasonal scales is described. Using a ten-year record of daily fire incidence, it is shown that using persistence and accurate forecasts of wind direction and pressure tendency, it is possible to skilfully forecast fire occurrence. Application of the same data to the problem of defining days when fire containment will be a problem is less successful. The McArthur Fire Danger Index (Luke and McArthur 1978), as determined using the Mark IV meter for grasslands, is shown to be a better indicator of the area likely to be burned out on a given day than any of the meteorological parameters which constitute it. A spatial average three-year cumulative antecedent rainfall, the blocking index (BI), as defined by the National Meteorological Centre (NMC), the Southern Oscillation Index (SOI) (Troup 1965), and sea surface temperature (SST) anomalies near Indonesia are all correlated with the fire data. The forecasting skill of these parameters is assessed and the SO1, rainfall and the BI are all found to have skill comparable with that of persistence.
Love et al. (Mon,) studied this question.