Remote sensing of wildland bushfire fuels in the Royal National Park, NSW

Field based techniques assessing wildland bushfire fuel characteristics are limited by the spatial extent at which they can be implemented due to cost, labour intensity, and terrain accessibility. Within Australia, research using remote sensing instruments to assess fuel characteristics has utilised low resolution imagery, simultaneously assessing all fuel strata. Within this research, fuel load is used to quantify wildland fuel distribution. Fuel load has been suggested to not account for fuel particle arrangement. This thesis presents two manuscripts: the first investigates the fuel hazard classification of pan-sharpened imagery to improve bushfire fuel assessment resolution, and the second investigates machine learning algorithms and the fusion of LiDAR and high resolution imagery to classify a single fuel stratum, the understory. Results suggest that the use of pan-sharpening to improve bushfire fuel assessment resolution is plausible, and that understory fuels can be classified with moderate accuracy using Support Vector Machine classification of a fusion on imagery and LiDAR metrics.