CQUniversity research explores how AI could help predict bushfire behaviour
As Australians prepare for an El Niño summer, new CQUniversity research is exploring how artificial intelligence (AI), satellite data and autonomous drones could support bushfire monitoring and emergency decision-making.
Recent CQUniversity graduate Dr Shouthiri Partheepan has developed a research framework that combines long-term bushfire analysis, drone-based monitoring and short-term fire spread forecasting to help turn complex data into practical insights for emergency management.
Bushfires can change direction and intensity rapidly, leaving emergency responders with limited time to assess evolving conditions and make critical decisions. Through her doctoral research, Dr Partheepan investigated how information gathered before, during and after a bushfire could be transformed into timely intelligence to support situational awareness and planning.
"Bushfire management is not only about detecting a fire. It is also about understanding where the risk is, how the fire is developing and what might happen next," Dr Partheepan said.
"My research explored how we could connect these different pieces of information to support more informed and timely decisions."
The research began with an analysis of 12 years of bushfire observations across Australia, examining historical fire patterns and the environmental factors influencing fire severity. By combining satellite imagery with vegetation, climate and topographical data, Dr Partheepan developed machine learning approaches to analyse bushfire severity and investigate patterns relevant to risk assessment.
The study identified areas where historical environmental conditions and fire patterns could help inform future planning, preparedness and bushfire management strategies.
The project also explored how drone technology could complement satellite observations by collecting detailed information about active fire events. While satellites provide valuable information across large geographical areas, drones can capture high-resolution observations of specific locations in near real time.
"Satellite data helps us understand the broader picture, while drones can provide a closer view of what is happening in a particular area," Dr Partheepan said.
"Connecting these perspectives could help bridge the gap between understanding bushfire risk and monitoring changing conditions on the ground."
A key part of the research investigated how AI could help drones navigate forest environments and monitor developing fires. Using advanced computer vision techniques, Dr Partheepan evaluated systems capable of identifying obstacles such as trees and branches while also developing methods to estimate the size of a fire using standard camera imagery.
The estimation methods were tested in both simulation environments and controlled outdoor trials, demonstrating the potential to rapidly assess fire size without relying solely on specialised thermal imaging equipment.
The research also explored how multiple autonomous drones could communicate and coordinate their movements to support cooperative bushfire monitoring, with the proposed approaches evaluated in simulated environments.
Beyond monitoring active fires, the research also explored whether AI could help predict how a fire might develop over short periods of time. Using a high-fidelity virtual fire environment, Dr Partheepan trained predictive models to estimate future changes in fire location, movement and area.
The findings demonstrated the potential of combining simulation data with machine learning to generate short-term forecasts that could eventually support planning, monitoring and emergency response activities.
"One of the important questions in bushfire management is not simply what is happening now, but what could happen next," Dr Partheepan said.
"Even short-term predictive information could eventually help support planning, monitoring and situational awareness when combined with reliable observations."
Alongside developing AI-based bushfire monitoring techniques, Dr Partheepan also published geospatial, forest obstacle-detection, and fire-and-smoke detection datasets to support further work in bushfire detection, environmental monitoring and autonomous systems research. The datasets have already attracted interest, with downloads and reported practical reuse demonstrating their potential value beyond the original PhD project.
While much of the research was evaluated in controlled simulation environments, the findings provide a foundation for future investigation into AI-assisted bushfire monitoring and decision support systems. Future work will focus on extending the technologies towards real-world drone operations and evaluating their performance under operational conditions.
"The long-term goal is to turn complex bushfire data into useful information that can help emergency management professionals make better-informed decisions," Dr Partheepan said.
"Although further real-world testing is needed, I hope this research can contribute to the development of smarter and more reliable tools for bushfire monitoring and management."
Dr Shouthiri Partheepan completed her PhD through CQUniversity's School of Engineering and Technology under the supervision of Dr Farzad Sanati and Associate Professor Jahan Hassan. Her research contributes to growing international efforts to apply AI, remote sensing and autonomous systems to environmental monitoring and natural hazard management.
