A machine learning predictive model for bushfire ignition and severity: The Study of Australian black summer bushfires

Abstract

Australian bushfires are catastrophic, and their impacts span social, economic, and environmental factors. To reduce the damages experienced by bushfires, predicting Australian bushfire ignition allows for an early warning system to give first responders and disaster managers prompt and accurate information. Traditional methods of bushfire ignition prediction suffer from incorporating large-dimensional data and take extensive computational time. Applying machine learning (ML) models enhances accuracy and reduces the computational time required to predict bushfire ignition. This study proposes a predictive model that can take meteorological and topographical data and determine the probability of Australian bushfire ignition and severity using historical fire detection gathered from the Black Summer Bushfire Disaster. The Black Summer Bushfire Disaster occurred between December 2019 and February 2020. The fires affected numerous towns throughout Victoria, New South Wales, and the Australian Capital Territory; hence, the varying topographical and meteorological conditions allow fire ignition and severity influences to be explored. The proposed methodology incorporates Random Forest (RF), Support Vector Machine (SVM), Gradient Boosting Trees (GBT), Convolutional Neural Networks (CNN), and K-nearest Neighbour (kNN) Algorithms. The proposed method relies on five datasets. Meteorological data is sourced from the Bureau of Meteorology (BOM), Australia. Topographic data is sourced from Geoscience Australia and the National Aeronautics and Space Administration’s (NASA’s) Aqua and Terra satellites, which utilize a Moderate Resolution Imaging Spectroradiometer (MODIS). Active Fire Point Data is also sourced from NASA MODIS, which can detect fires. The proposed methodology aims to act as an early warning system by providing a fire occurrence and fire intensity warning map and the probability of fire occurrence and fire intensity depending on the current meteorological climate.