AI trained to spot fires
Print
AI might soon help to predict bushfires.
Researchers have developed a machine learning model that significantly improves the prediction of fire outbreaks based on real-time weather data.
Trained on 15 years of weather data from Queensland’s Sunshine Coast region, the model has been shown to outperform traditional methods, showing a 47 per cent improvement in accuracy over the standard fire danger index.
The model uses random forest algorithms to classify pre-fire conditions, analysing data such as temperature, wind speed, and humidity at 30-minute intervals.
It was able to predict 75 per cent of fires in simulated real-time conditions, identifying areas at risk from weather patterns occurring hours to days before fires ignited.
“This method provides objective, quantifiable information, enhancing the precision and effectiveness of fire warning systems,” the research team says.
The study highlights the system's potential to support decision-makers in implementing timely evacuations and effective fire suppression strategies.
By continuously integrating new weather data, the model adjusts to changing conditions, offering more adaptive fire danger forecasts compared to static traditional indices.
The researchers suggest that integrating this model with existing fire management systems could improve early warning capabilities, especially in rapidly changing environmental conditions influenced by climate change.
The full study is accessible here.
