A shrubland fire behaviour dataset was assembled using data from experimental studies in Australia, New Zealand, Europe and South Africa. The dataset covers a wide range of heathlands and shrubland species associations and vegetation structures. Three models for rate of spread are developed using 2-m wind speed, a wind reduction factor, elevated dead fuel moisture content and either vegetation height (with or without live fuel moisture content) or bulk density. The models are tested against independent data from prescribed fires and wildfires and found to predict fire spread rate within acceptable limits (mean absolute errors varying between 3.5 and 9.1 m min À1 ). A simple model to predict dead fuel moisture content is evaluated, and an ignition line length correction is proposed. Although the model can be expected to provide robust predictions of rate of spread in a broad range of shrublands, the effects of slope steepness and variation in fuel quantity and composition are yet to be quantified. The model does not predict threshold conditions for continuous fire spread, and future work should focus on identifying fuel and weather factors that control transitions in fire behaviour.

In many landscapes, an important fire management objective is to reduce the negative impacts from unplanned fires on people, property and ecological values. In Australia, there exists an inherent assumption that high spatial variability in fire ages and hence fuel loads will have negative effects on both the incidence and spread of subsequent fires, and will enhance ecological values. A recent study using the process-based computer simulation model FIRESCAPE-SWTAS predicted several relationships between prescribed burn treatment levels and spatial patterning and management objectives in south-west Tasmania, Australia. The present study extended this investigation to additionally explore the effects of prescribed burning treatment unit size on unplanned fire incidence and area burned both in the general landscape and specifically in fire-intolerant vegetation. Simulation results suggest that treatment level had the greatest influence on modifying fire effects, whereas treatment unit size had the least effect. The model predicted that all three parameters interacted to determine the mean annual area burnt by unplanned fires. In fire-intolerant vegetation, treatment unit size did not influence the incidence of unplanned fires and the area burnt by unplanned fires in these communities. Where significant differences were evident, fire risk was reduced by higher treatment levels, deterministic spatial patterns of burning units, and smaller burning unit sizes.

We develop a method for estimating equilibrium moisture content (EMC) and fuel moisture response time, using data collected for Eucalyptus twig litter. The method is based on the governing differential equation for the diffusion of water vapour from the fuel, and on a semi-physical formulation for EMC (Nelson 1984), based on the change in Gibbs free energy, which estimates the EMC as a function of fuel temperature and humidity. We then test the model on data collected in Western Australian mallee shrubland and in Tasmanian buttongrass moorland. This method is more generally applicable than those described by Viney and Catchpole (1991) and Viney (1992). The estimates of EMC and response time are in broad agreement with laboratory-based estimates for similar fuels (Anderson 1990a ; Nelson 1984). The model can be used to predict fuel moisture content by a book-keeping method. The predictions agree wellwith the observations for all three of our data sets.

Summary1. Natural area managers use fire and grazing to achieve nature conservation ⁄ production goals and to prevent the loss of life and property. Yet, little is known of the effects of post-fire grazing on fuel load and the proportion of days on which fire can be sustained (fire potential). This knowledge could help managers in planning interventions to achieve their goals. 2. At seven sites in Tasmania, Australia, including sedgeland, heathy forest and grassland, fire potential and fuel load were measured before, and for 2 years after fire. Measurements were made in burning, fencing and burning plus fencing treatments, and in control quadrats. 3. Burning followed by grazing, largely by native vertebrates, resulted in lower fuel loads than either grazing by itself or burning by itself. A new steady state was established in two grasslands. Fire potential at the oligotrophic sites was largely a function of time elapsed since the last fire, while at grassy sites was increased by grazing without fire, but depressed or slightly increased by grazing after burning. 4. Synthesis and applications. Effects of grazing after burning on flammability are not predictable from the single or additive effects of grazing and burning, varying between vegetation type and environment. In highland grassy ecosystems fire potential can be reduced by excluding grazing animals after fire, while in scleromorphic ecosystems grazing after fire does not affect fuel or fire potential. Intense grazing after fire can cause an, often desirable, shift from tussock to lawn grassland. Burning and subsequent grazing of tussock grassland vegetation in the lowlands may reduce the chance of wildfire damaging property and conservation ⁄ production values, while in highland tussock grassland burning followed by grazing will be largely ineffective in reducing the already low chance of such damaging fire.