Monitoring the 2009 Victorian Bushfires with Alos Palsar Multi-temporal and Coherence Images

The Victorian bushfires in February 2009 were one of the most terrifying recent natural disasters in Australia. Synthetic Aperture Radar (SAR) is an all–weather remote observation system which can penetrate through cloud and smoke to image the Earth’s surface at anytime. Hence, it is an ideal tool for change detection, and in particular for fire monitoring. This paper describes the use of ALOS/PALSAR multi-temporal images and interferometric coherence data to study the 2009 Victorian Bushfires. Two areas where several ALOS PALSAR images had been acquired before, during and after the fire period were selected for this study. Landsat TM images acquired for a similar period over the same areas were used for validation purposes. Various techniques were applied to detect burnt areas, including backscatter intensity difference, coherence analysis and Principal Component Analysis. However, none of these methods could clearly locate the fire-affected regions on their own. A combined approach was significantly better. Burnt scars extraction was carried out using Maximum Likelihood, Artificial Neural Network, and Support Vector Machine classification techniques, and the results were compared to identify the best classifier – that produced the most accurate fire damage map.