Currently there are limited techniques to undertake space-based mapping of the development of fire fronts during bushfire episodes. Using the 2019-2020 Kangaroo Island fires as an example, this presentation will illustrate how sensors on-board five different GEO and LEO satellites provide sequences of fire hot-spot detections and demonstrate how spatio-temporal interpolation of these hot spots can be used to model the movement of the fire front.
Geostationary (GEO) and low Earth orbit (LEO) satellites are used to identify bushfire hot spots wherein the temporal resolution of observations is generally inversely proportional to the spatial resolution for single satellite constellations.
LEO satellites are frequently used to map bushfire burn scars during or post fire episodes and record recovery, especially in respect to vegetation. However, currently there are limited techniques to undertake space-based mapping of the development of a fire front during the fire episode. Using the 2019-2020 Kangaroo Island fires as an example, this presentation will illustrate how fire initiation points are associated with satellite lightning data, how sensors on-board five different GEO and LEO satellites (Himawari, EOS Terra, EOS Aqua, NOAA and Suomi-NPP) provide sequences of fire hot spot detections and demonstrate how spatio-temporal interpolation of these hot spots can be used to model the movement of the fire front. These models are compared with fire scar mapping from imagery captured by four other higher spatial resolution LEO satellite imaging sensors (Landsat 8 OLI, Sentinel 2 MSI, Planet Scope and Sentinel 1) operating in both the optical and microwave domains.
Use of the microwave domain largely eliminates issues associated with cloud and smoke, which significantly affect optical satellite observations. Despite its excellent 10 min repeat coverage, the 2 km spatial resolution of the 16 band multi-spectral AHI sensor on Himawari 8, particularly in the middle and far IR bands, results in a poor ability to model fire propagation. However, this situation improves slightly with the use of the 36 band multi-spectral MODIS sensors on EOS Terra and Aqua at 1km spatial resolution. At 375m spatial resolution the five L multi-spectral bands of the VIIRS sensors on Suomi-NPP and NOAA 20 provide a considerable improvement over MODIS and thus result in greater sensitivity in mapping hot spot distributions. The latter two satellites are LEO and only provide 12 hours repeat coverage, thus requiring the use of spatio-temporal interpolation of the vector point data that is output from the various hot-spot detection algorithms, especially for fires that progress rapidly as did the Ravine fire on 30 December, 2019.
Wind speed and direction from Global Forecast System of National Centers for Environmental Prediction (NCEP)-NOAA was utilised to assist this spatio-temporal interpolation. The efficacy of fire scar mapping using differential algorithms such the normalised burn ratio (NBR), normalised difference vegetation index, vegetation structure perpendicular Index, and the forest disturbance index is being explored in separate research, but here Differential NBR was applied to optical imagery and a modified version of the Differential Radar Vegetation Index was applied to the dual polarization SAR images. This produced maps of bunt – unburnt vegetation at different times during the fires periods from 16 Dec 2019 to 15 January 2020 and these were compared with the corresponding spatio-temporal interpolation of fire positions from hot-spot detections, to show that the latter process was most successful when using VIIRS data.
Assoc Prof David Bruce (Flinders University)
Dr Parwati Sofan
Time & room
1:45 pm–2:15 pm
City Room 2
Date & venue
Friday 17 September 2021 at the Adelaide Convention Centre