UAV vs terrestrial lidar is an ongoing question in forest measurement: can a drone effectively replace a tripod-based scanner?
Researchers at Wageningen University and the Center for International Forestry Research explored this question by flying a RIEGL RiCOPTER equipped with a VUX-1UAV scanner over two hectares of Dutch forest and comparing the results against a VZ-400 terrestrial laser scanner.
Two days. Fifty-eight tripod setups, a theodolite, retro-reflective targets and a lot of walking. That is what it took to scan two hectares of Dutch forest from the ground. The drone did it in nine minutes. The real question is whether the drone data was any good.
The research team found canopy heights agreeing within centimeters and stem diameters correlating at 0.98, from 40 points per trunk against the tripod’s 5,522. However, where the drone struggled, and why, is the interesting part.
Why UAV lidar?
Crewed airborne lidar covers landscapes but typically yields only 1 to 10 points per square meter, forcing a statistical treatment of canopy structure. Terrestrial scanning effectively maps individual stems and branches, but co-registering scan positions makes projects slow, nearly three to six days per hectare on average. The research team set out to test whether a commercially available, off-the-shelf UAV system could deliver terrestrial-scale point density at airborne speed.
Two hectares, two sensors
Fieldwork took place at the Speulderbos Fiducial Reference site in the Dutch Veluwe, covering roughly 2 ha: a beech and oak stand dating to 1835 with an open understory, plus Norway spruce, giant fir, beech and Douglas fir plots chosen for contrasting density and canopy architecture. The overstory was mid bud-out, so most trees carried very little foliage.
The RiCOPTER flew at 90 meters above ground at 6 meters per second, with the scanner at its full 330 degree field of view, a 550 kHz pulse repetition rate and 58 scan lines per second. That produced about 8 cm point spacing and 140 points per square meter from a single nadir pass, with overlapping lines building plot averages near 3,000 points per square meter, though the terrestrial scanner was recording about a hundred times denser than that. Active scanning took nine minutes across 2,300 m of flight line. The terrestrial reference required 58 scan positions spread over two days.
Canopy Height
Canopy height models from the two systems agreed closely, with most cells falling within half a meter. The UAV models ran higher on average, by 6.1 cm over giant fir and 12.2 cm over old beech and oak, and 11.5 cm across all cells, consistent with the known difficulty of registering the uppermost canopy from beneath it. Importantly, the authors note that 11.5 cm sits well inside the roughly 50 cm precision of conventional field height measurement. Crowns read too tall on their eastern sides and too short on their western sides, a sign the two point clouds sat 0.5 to 1 m out of alignment at crown level, most likely because wind shifted the crowns between scans. This effect was found to be strongest over the taller, Douglas fir.
Stem Diameter
Of 58 stems extracted from both point clouds, 39 supported circle fitting between 120 and 140 cm height, with rejections concentrated in the dense conifer plots and the young beech. UAV stem rings averaged 40 points against 5,522 for the terrestrial scanner, yet agreement held, with a correlation coefficient of 0.98, an RMSE of 4.24 cm and UAV estimates running 1.71 cm larger on average. That still sits well above the roughly 0.3 cm precision of a diameter tape.
Takeaways
None of this drone success retires the tripod. The terrestrial scanner still recorded a hundred times more points, resolved stems the drone could barely see, and supplied the reference values the UAV was judged against, since no field-tape measurements entered the comparison. What the study establishes is that the two instruments fail in opposite directions: scanning upward from beneath the crown, the tripod misses treetops, while scanning downward from 90 m, the drone thins out on trunks. That points to a choice between priority outcomes. Rather than drones succeeding terrestrial methods, surveyors need to weigh the costs and benefits of each platform. The drone’s speed advantage counts for most in exactly the places ground campaigns are hardest… tropical and remote forest, which are also where its reliance on GNSS base station data and its 25 kg of hardware (near the legal limit of sUAS) become a problem. The authors’ next step is to push past these two metrics into quantitative structural modeling of whole trees, and from there to volume and above-ground biomass.
Read the full research paper here.
