CSIRO is calibrating drone LiDAR against grass cut and weighed by hand at Boorowa

LiDAR fires hundreds of laser pulses a second and times how long each one takes to come back ^[12]. The usual product is a point cloud, a record of where each pulse stopped. A ray cloud keeps the full path the pulse traveled as well, so the empty space it passed through is also data, and the file becomes a measure of how porous or solid a volume is ^[14]. Edwards' team built that approach with Wine Australia to measure canopy structure in vineyards, using a set of tools called RayCloudTools from CSIRO's Technology research unit ^[13].

Grapevines sit above a gap. The ground is a surface you can find, and everything above it is canopy ^[16]. Grass gives you nothing like that, because it grows from the soil up, and CSIRO calls separating plant from ground the hardest technical problem in the whole system ^[16]. A biomass estimate is built from what stands above the inferred ground surface, so an error in that surface passes into the estimate, and it costs most where the grass is shortest. "Ground detection is a huge challenge for this project," Kikuchi said ^[4].

"Graziers need to know the exact amount of living pasture biomass," Kikuchi said ^[3]. His field team flies the plots, then walks back into the same paddock with shears and scales and cuts square meter by square meter ^[2][21]. A cut quadrat is single-use, so each one yields one paired observation. To get a spread of values to fit a curve against, the team slashes and grazes the trial plots on purpose, producing high and low covers on demand ^[17].

The phys.org account, published in September, describes promising results and does not include an accuracy figure or the number of quadrats cut ^[15][20]. Eight months separate the project's start from that account ^[19].

Satellite NDVI, the free end of the range, is derived from how plants reflect and absorb different wavelengths of light ^[7]. Edwards, who leads CSIRO's Agrisensing Analytics team ^[5], said of NDVI: "It's considered a reliable indicator of how much live green vegetation is present." He also said that "accuracy drops away in dense canopies, where the index quickly saturates, an issue compounded by the lack of plant height information from satellites" ^[6]. "We are currently missing a method in between these two ends of the scale," he said ^[10]. The rising plate meter at the other end is itself a calibration curve built from many hand-cut estimates ^[8]. Both methods answer to the same reference, and both can be run over the same plots on the same day.