Change detection study

Challenge

Airborne LiDAR of the same ground is flown by different programs, years apart, to different densities and often on an older datum. As delivered, the surveys of these sites sat up to 0.45 m apart in height on ground that had not changed: larger than much of the change being measured. A denser survey also sees more of the canopy, which reads as growth.

Approach

Each earlier survey was brought into GDA2020 and aligned to the later one on stable ground: slopes under 25°, at least 30 m from any fire mapped between the flights, and off water. For ground, the scatter that remains there sets the pair's level of detection at 95 % confidence, per square kilometre. For canopy, the offset between the two sensors was measured on unburnt vegetation, class by class from TASVEG, and taken off first; each 5 m cell's level of detection is then set at 95 % confidence from its laser returns in both surveys, never less than 10 points of cover. Only change beyond the level of detection, in patches of at least 100 m², is counted. Ground change was computed on 1 m cells and summed to cut and fill volumes.

Outcome

Rosebery: 1.56 million m³ placed and 0.59 million m³ removed between 2015 and 2026, nine-tenths of each in one area south of the town. Lake Mackenzie: canopy cover fell beyond the level of detection across 10,695 ha between 2013 and 2016, 99 % of it inside the fire perimeters mapped between the two surveys. Forcett–Dunalley: canopy cover fell beyond the level of detection across 1,915 ha between 2011 and 2013, 97 % of it inside the fire perimeters mapped between the surveys. At both fires the change follows the mapped perimeters: what crosses the level of detection is fire, not survey error.