Abstract
Bushfires in Australia are amongst the most frequent, severe and extensive natural disasters in the world. Frequent bushfires destroy many natural resources and damage the forest’s biochemical chain and cause soil damage. Therefore, assessment of bushfire scars is a very
important theme in observing Australia from space.
Remotely sensed data is used operationally to map burned area and fire severity after major
bushfires and to research the impact of prescribed burns. Several types of Remote Sensing
data were collected in order to assess fire effects from both optical sensors and active
microwave sensors. Optical data from more recent sensors are being used for mapping
burned scars and fire severity. Active microwave Remote Sensing data, like RADAR can
observe the Earth’s surface day and night under all cloud conditions. The usage of optical
and Synthetic Aperture RADAR (SAR) Remote Sensing technologies can provide valuable
information, analysis capability and more confidence for determining the extent of burned
areas and estimating fire severity. A fusion of optical and active Remote Sensing would
greatly improve post-fire scars mapping.
This research focuses on two study areas. In the first burned area, Dargo - Dippsland, the
operational application of both Advanced Land Observing Satellite L-band synthetic
aperture RADAR (ALOS PALSAR) and Landsat Thematic Mapper (TM) data for mapping
burned scars was evaluated. Both ALOS and Landsat data were acquired before and after
the bushfires. Several different methodologies were analysed and compared to assess the
bushfire scars, in terms of the image enhancement techniques, the band ratioing and texture
transformation techniques, the principal components analysis (PCA) and the fusion of
ALOS PALSAR and Landsat TM data. Results show that the technique based on PCA
extracted from the fusion of Landat TM and filtered ALOS PALSAR images indicated a
clear superiority on bushfire scars detection, which obtained an overall accuracy of 67.62%
and a Kappa coefficient of 0.2443, followed by the classification based on the PCA
extracted from the fusion of Landat TM and filtered ALOS PALSAR images (overall
accuracy is 59.88% and a Kappa coefficient is 0.2026). It is noted that the ALOS PALSAR
and Landsat TM fused images are much better than individual images to identify burned
scars.
Furthermore, the severity of a fire’s impact on vegetation was mapped in the second study
area, Sharps Unit - Otway, using Landsat TM and ground truth data. Methods in this study
highlight pre- and post-fire changes by computing the difference in the Normalised Burn
Ratio (NBR) and the Normalised Difference Vegetation Index (NDVI) between the period
of pre- and post-burn. Then the results show that the differenced NBR (dNBR) algorithm
represented better severity classes’ result than the differenced NDVI (dNDVI). What means
by this is NBR and dNBR were more sensitive to fire effects than NDVI and dNDVI.
However, some issues should be considered in this process, for instance the threshold of
severity classes should be considered carefully when placing the field plots thereby
preventing the loss of fire severity evidence. The analysis will be valuable for
understanding the different methodologies of mapping the bushfires scars and severity, and
forming ecologically-sensitive approaches to fire management.