Spark: A better way to predict the spread of bushfires
Bushfires are complex processes, making it difficult to accurately predict their progress across the landscape. So we have developed Spark, a software framework that combines our bushfire behaviour knowledge with state of the art simulation science. Spark will give fire-fighting agencies a more accurate view of fire behaviour, informing decisions that could minimise property damage and save lives.
Modelling of dynamic radiant heat flux within Spark
Testing the ability of homes to withstand a wildfire is crucial for building in wildfire-prone areas. One of the main factors for the survivability of a structure is the amount of radiant heat it is exposed to. This demonstration shows a wildfire approaching an urban edge and the calculation of radiant heat at two homes. The incoming radiation on these homes is shown on two spheres, coloured from blue (low heat) to red (very high heat). Simulations such as this can help us understand ways to mitigate the dangers to homes from wildfires.
Modelling of firebrand dynamics within Spark
A 3D visualisation of what Spark’s firebrand model can do. Spark allows the modelling of hundreds of thousands of individual firebrands from creation, interaction with the wind and local effects to landing and creating a new spot fire. The firebrand models are entirely user-defined, similar to the rate-of-spread models within Spark. This allows complete flexibility in creating, testing and deploying firebrand models. This example shows a fire spread simulation with firebrands dynamically created depending on the intensity of the fire, ballistically travelling downwind and having a chance to create a new spot fire.
Comparison of Spark simulation to simple Pyrotron experiment
A comparison of a small-scale experimental fire in the CSIRO Pyrotron to a Spark simulation (white line) using a new ‘near-field’ pyrogenic model. The new model accounts for air currents generated by the fire itself and shapes the predicted fire into a parabolic shape, in close agreement with the experiments. The near-field model allows rapid and accurate modelling of small-scale fire effects and is a step between previous basic two-dimensional models of fire propagation and highly-detailed, but very slow to run, three-dimensional models of fire.
Comparison of Spark simulation to Pyrotron fire attraction experiment
Nearby fires show a an effect in which they are attracted towards each other, shown here for two lines of fire in the CSIRO Pyrotron. The fires move towards the centre of the lines and eventually merge together. This effect has been replicated in Spark simulation (white line) using a new ‘near-field’ pyrogenic model. The new model accounts for air currents generated by the fire itself and shows the attractive effect, in close agreement with the experiments. The near-field model allows rapid and accurate modelling of small-scale fire effects and is a step between previous basic two-dimensional models of fire propagation and highly-detailed, but very slow to run, three-dimensional models of fire.
Large-scale fire predictions with Spark 1
Demonstrations of large-scale fire prediction in a Spark graphical application. The application allows input data sets, fire models, pre and post-processing steps to be defined and output data to be written in standard GIS formats. The first example shown is for a simulated fire of 24 hour duration in Australia near Hobart, Tasmania. Various visualisations include a shaded fire footprint outline, isochrones (shown here for every two hours) fire and a hue-shaded colour scheme representing arrival time of the fire. The second example is an ensemble of fires starting over a grid, the resulting footprint with isochrones and a hue-shaded map of impact probability from all fire simulations in the ensemble.
Sparking interest in CSIROs Pyrotron
This is an experiment being conducted in CSIROs Pyrotron. The results are then rectified and compared with a Spark simulation in order to improve our fire behaviour models.
UAV view of an experimental grass fire
UAV video feed of an experimental grass fire. The results are then rectified and compared with a Spark simulation in order to improve our knowledge of fire spread and our fire behaviour models.
Spark in virtual reality
CSIRO Data61’s work in virtual reality visualisation applied to a Spark bushfire simulation.