Photo caption: Wildfire research programme experimental burn.
Developed by an international team led by Dr Alberto Ardid, a research engineer in Civil and Natural Resources Engineering at Te Whare Wānanga o Waitaha | University of Canterbury (UC), the new tool uses machine learning (a type of artificial intelligence) to analyse ever-changing weather data.
Dr Ardid says the frequency and intensity of wildfires, such as the destructive blazes seen in California recently, is increasing due to climate change and there's an urgent need for more effective fire management strategies to protect lives, property, and ecosystems.
"Accurate and timely wildfire danger forecasting is necessary for preparedness and response, enabling efficient resource allocation and mitigation efforts. Weather conditions can change dramatically within hours, potentially leading to sudden wildfire outbreaks," he says. "We are developing new tools that are sensitive to these hourly weather conditions which can help identify high fire hazard conditions."
His team has developed a unique AI-based system that uses readily available weather data to predict wildfire danger in real-time, providing a powerful tool to inform timely decision-making.
"Our AI model analyses weather data every 30 minutes, capturing dynamic weather patterns that can contribute to sudden wildfire outbreaks," Dr Ardid says. "This approach offers a cost-effective solution for communities and agencies to enhance their wildfire prediction and response capabilities, contributing to community safety and resilience in the face of increasing wildfire danger."
The model outputs a prediction of the likelihood of fire in the next few days, continuously updated with new meteorological data. The real-time monitoring system uses existing data and infrastructure making it cost-effective in regions with limited resources.
"We hope this research will complement current monitoring techniques and advance wildfire management, offering a valuable tool for mitigating the increasing threat of wildfires and protecting communities," Dr Ardid says.
The AI model was developed and tested using historical weather and fire data from Queensland, Australia, and achieved a 47% improvement in predicting critical pre-fire conditions compared to the existing Forest Fire Danger Index.
Dr Andres Valencia-Correa, a senior lecturer in UC's Civil and Natural Resources Engineering department, collaborated on the research which was published in the International Journal of Wildland Fire today. He says the system's early and accurate warnings would facilitate faster evacuations.
"It would also improve fire-fighting strategies and allow fire management agencies to allocate their resources more effectively," Dr Valencia-Correa says. "This enhanced predictive capability could potentially save lives and lessen the adverse impacts of wildfires on communities and ecosystems."
This project was recognised recently for its potential impact and innovation, being selected among the top 10 finalists for the Allianz Climate Risk Award for early-career scientists.
The research team also includes UC School of Earth and Environment Associate Professor Marwan Katurji, UC Civil and Natural Resources Engineering Associate Professor David Dempsey, and fire ecologist Shana Gross, from Scion.
The other collaborators are Anthony Power, an Australian bush fire consultant, and Professor Matthias Boer from Western Sydney University.
