Irrigating Low-Flammability Vegetation: How Simulation Reduces Bushfire Risk in Australia
A study from the University of Sunshine Coast shows that irrigating low-flammability plants can decrease the spread and intensity of bushfires. Discover the method, AI models used, and implications for bushfire prevention in Australia.
A recent simulation suggests that irrigating low-flammability vegetation could provide Australian communities with crucial extra time to prepare for bushfires by reducing both the extent and intensity of fires.
What the study reveals: Reducing fire risk through irrigation of low-flammability vegetation
Researchers at the University of Sunshine Coast integrated irrigation scenarios into a predictive bushfire model. Results show a notable reduction in fire spread and attenuation of its intensity, even in areas where vegetation is naturally very dry. Simulations suggest that targeting species with low flammability could create safety corridors that slow down the fire front. Exact figures for reduction remain unconfirmed information at this stage, but the observed effect is significant enough to warrant field studies.
This approach contrasts with traditional strategies that mainly focus on direct fire suppression. By modifying the vegetation composition, it acts on the very source of fuel, which according to authors could reduce the prediction uncertainty linked to fuel variability. The model also indicates that communities near these irrigated zones would gain several hours, or even a day, of additional time for evacuation or implementing defense measures.
How: Research method, predictive model, and satellite data used
To generate these scenarios, the team combined several atmospheric and terrestrial data sources. Satellite data from Copernicus and the Sentinel program provided fine-scale vegetation cover mapping, while ECMWF reanalyses supplied high-resolution wind, humidity, and temperature fields. The core of the system relies on an auto-learning neural network, similar to architectures used in GraphCast and Pangu-Weather, to simulate fire dynamics faster than classical models.
The predictive model integrates variables such as fuel density, soil moisture content, and topography. Through supervised learning, the neural network could calibrate combustion parameters based on historical fire observations, reducing prediction uncertainty. Simulations were run on supercomputers, allowing exploration of tens of different irrigation scenarios, from simple spot watering to large-scale irrigation systems.